Discourse Cohesion in Ancient Greek

Amsterdam Studies in
Classical Philology

Edited by

Albert Rijksbaron
Irene J.F. de Jong
Caroline Kroon

VOLUME 16
Discourse Cohesion in
Ancient Greek

Edited by

Stéphanie Bakker and Gerry Wakker

LEIDEN • BOSTON
2009
This book is printed on acid-free paper.

Library of Congress Cataloging-in-Publication Data

Wakker, Gerry.
Discourse cohesion in ancient Greek / by Stephanie Bakker, Gerry Wakker.
p. cm. — (Amsterdam studies in classical philology ; v. 16)
Papers presented at the 6th International Colloquium on Ancient Greek Linguistics,
held June 27–29, 2007, Groningen, Netherlands.
Includes bibliographical references and index.
ISBN 978-90-04-17472-6 (hardback : alk. paper) 1. Greek language—Verb—
Congresses. 2. Greek language—Particles—Congresses. 3. Greek language—Tense—
Congresses. 4. Cohesion (Linguistics)—Congresses. 5. Discourse analysis—Congresses.
I. Bakker, Stéphanie J. II. Title.
PA337.W35 2009
485—dc22
2009023935

ISSN 1380-6068
ISBN 978 90 04 17472 6

Copyright 2009 by Koninklijke Brill NV, Leiden, The Netherlands.

Koninklijke Brill NV incorporates the imprints Brill, Hotei Publishing,
IDC Publishers, Martinus Nijhoff Publishers and VSP.

All rights reserved. No part of this publication may be reproduced, translated, stored in
a retrieval system, or transmitted in any form or by any means, electronic, mechanical,
photocopying, recording or otherwise, without prior written permission from the
publisher.

Authorization to photocopy items for internal or personal use is granted by

Koninklijke Brill NV provided that the appropriate fees are paid directly to
The Copyright Clearance Center, 222 Rosewood Drive, Suite 910,
Danvers, MA 01923, USA.
Fees are subject to change.

printed in the netherlands

 CONTENTS

List of contributors ............................................................................ vii

Preface .................................................................................................. ix
Introduction ........................................................................................ xi
Stéphanie J. Bakker and Gerry C. Wakker

Chapter One Discourse Cohesion Through Third Person

Pronouns. The Case of κεῖνος and αὐτός in Homer ................ 1
Anna Bonifazi

Chapter Two Pragmatic Presupposition and

Complementation in Classical Greek ......................................... 21
Luuk Huitink

Chapter Three On the Curious Combination of the Particles

γάρ and οὖν .................................................................................... 41
Stéphanie J. Bakker

Chapter Four ‘Well I Will Now Present My Arguments’.

Discourse Cohesion Marked by οὖν and τοίνυν in Lysias ...... 63
Gerry C. Wakker

Chapter Five The Particles αὖ and αὖτε in Ancient Greek as

Topicalizing Devices ...................................................................... 83
Antonio R. Revuelta Puigdollers

Chapter Six Καὶ μήν, καὶ δή and ἤδη in Τragedy and

Comedy ........................................................................................... 111
A. Maria van Erp Taalman Kip

Chapter Seven Dicourse Cohesion in Dialogue.

Turn-Initial ἀλλά in Greek Drama ............................................. 135
Annemieke Drummen

Chapter Eight Greek Particles: Just a Literary

Phenomenon? ................................................................................. 155
Coulter H. George
vi contents

Chapter Nine Towards a Typology of the Narrative Modes

in Ancient Greek. Text Types and Narrative Structure in
Euripidean Messenger Speeches .................................................. 171
Rutger J. Allan

Chapter Ten The Use of the Imperfect to Express Completed

States of Affairs. The Imperfect as a Marker of Narrative
Cohesion ......................................................................................... 205
Louis Basset

Chapter Eleven Involving the Past in the Present.

The Classical Greek Perfect as a Situating Cohesion Device 221
Sander Orriens

Chapter Twelve Discourse Cohesion in the Proem of Hesiod’s

Theogony .......................................................................................... 241
Albert Rijksbaron

Bibliography ........................................................................................ 267

Index Locorum ................................................................................... 279

General Index ..................................................................................... 283
 LIST OF CONTRIBUTORS

Rutger J. Allan is a Lecturer in the Department of Classics at the

VU University Amsterdam

Stéphanie J. Bakker is a postdoc fellow at Leiden University and

teaches Greek at the University of Groningen

Louis Basset is emeritus professor of Ancient Greek Linguistics,

Lumière University Lyon 2

Anna Bonifazi is holder of an Outgoing International Marie Curie

Fellowship (2005–2008), Harvard University (Classics) and University
of Turin (Center for Cognitive Science)

Annemieke Drummen is MA student of Classics at the VU University

Amsterdam

A. Maria van Erp Taalman Kip is emeritus professor of Ancient

Greek, University of Amsterdam

Coulter H. George is an assistant professor in the Department of

Classics at the University of Virginia

Luuk Huitink is PhD student of Classics and teaches Greek at Oxford

University

Sander Orriens is MA student of Classics at the University of Groningen

Antonio R. Revuelta Puigdollers is a Lecturer in the Department

of Classics at Madrid Autonomous University

Albert Rijksbaron is emeritus professor of Ancient Greek Linguistics,

University of Amsterdam

Gerry C. Wakker is professor of Ancient Greek Linguistics at the

University of Groningen
 PREFACE

In 1986 the Department of Classics at the University of Amsterdam ini-

tiated a series of International Colloquia on Ancient Greek Linguistics.
Within this framework every fourth or fifth year a new colloquium is
organised on a specific theme: grammatical questions (Amsterdam);
the language of Homer (Grenoble); particles (Amsterdam), comple-
ment clauses (Saint Étienne), word classes (Madrid). The 6th edition
of this International Colloquium on Ancient Greek Linguistics on the
theme of (phenomena of) discourse cohesion took place in Groningen
(the Netherlands) from June 27th to 29th 2007. This colloquium was
the starting point for the present volume Discourse Cohesion in Ancient
Greek in that most of the contributions are revised versions of papers
presented and discussed at the colloquium.
It is our pleasure to thank a number of organizations and persons
for their support. First of all, we would like thank the GUF (Stichting
Groninger Universiteitsfonds), the research institute CLCG (Center of
Language and Cognition Groningen), OIKOS (the National Research
School in Classical Studies) and the KNAW (Royal Netherlands Academy
of Arts and Sciences) for their financial support of the colloquium. Our
(former) students Maarten van Putten, Patricia Spuij and above all Inez
van Egeraat and Sander Orriens provided a lot of practical help. Sander
Orriens was also of great help in the final stage of the realization of the
volume, when the various contributions had to be converted into one
coherent typescript. Furthermore, we are grateful to the authors for their
patience and their willingness to revise their contribution. Finally, we
would like to thank the editors of this series Albert Rijksbaron, Caroline
Kroon and Irene de Jong, the anonymous reviewer, and Brill for their
help and valuable comments and suggestions.

Groningen, March 2009

Stéphanie Bakker
Gerry Wakker
 INTRODUCTION

Stéphanie J. Bakker and Gerry C. Wakker

The 6th International Colloquium on Ancient Greek Linguistics, which

took place in Groningen (the Netherlands) from June 27th to 29th 2007,
was the starting point for this volume. Because the colloquium did not
have a traditional set-up, this volume does not have the standard char-
acter of the acts of a colloquium. Instead of presenting a paper during
the colloquium, the participants were asked to submit a paper on the
theme of ‘discourse cohesion’ beforehand, so that all participants could
read them. At the colloquium, most time was spent on the discussion
of the papers. These discussions were opened by a reaction prepared
in advance by one of the other participants. This approach appeared to
be very fruitful and led to many stimulating discussions, both on the
topic of the paper at issue and on more general aspects of discourse
cohesion and the Ancient Greek language. This set-up inspired some of
the speakers to revise and submit their contribution and inspired other
scholars (some of them not even present at the colloquium) to write
and submit a paper on Discourse Cohesion in Ancient Greek.
Central in this book is the study of classical texts as discourse, i.e.
as texts functioning within a communicative situation.1 Most authors
in this volume explicitly or implicitly adopt a functional or cognitive
framework. These frameworks have in common that they consider
language an instrument for communicative interaction within a specific
communicative context. This principle enables the study of linguis-
tic phenomena that are often neglected in a strictly sentence-based
approach, like particles, word order, active-passive variation, speech acts,
anaphoric reference and tense usage. All these topics can be related to
the subject of this book: discourse cohesion in Ancient Greek.
What is understood by the term discourse cohesion? Although much
has been written about this topic, especially about the use of cohesion

1
By text we refer to a static object, the product of speech or writing, detached from
its situational context, whereas by discourse we refer to the dynamic process of speech
and writing in its situational context. Cf. Brown & Yule (1983: 23–5), Hoey (1991:
212–213) and Kroon (1995: 30 note 50).
xii stéphanie j. bakker and gerry c. wakker

devices in specific types of discourse (e.g., advertorials, conversations,

etc.),2 there is no consensus on what exactly discourse cohesion is (nor
on the question whether it really exists, see below p. xiv). Most scholars
working on discourse cohesion, however, will more or less accept the
ideas of Halliday and Hasan (1976), the fathers of the concept of cohe-
sion, which Tanskanen summarizes as follows:
(1) [They regard] cohesion as the grammatical and lexical elements on
the surface of a text which form relations that connect parts of the
text. (Tanskanen (2006: 19–20)).
Or as Halliday & Hasan (1976: 4) formulate it themselves, ‘cohesion
occurs where the interpretation of some element in the discourse is
dependent on that of another.’ Cohesion devices are thus linguistic ele-
ments that help the addressee arrive at the interpretation intended by
the speaker, since they make the unity of the discourse explicit.
Discourse cohesion is contrasted with discourse coherence, an even
more debated term. The term coherence is generally understood to refer
to the degree in which a text coheres with its (textual and non-textual)
context, i.e. the degree in which parts of a text make up a unified whole
instead of being a set of unrelated utterances.3 It is often stressed that
the presence of coherence is subjective in that it depends on the relation
between the reader/receiver and the text. Whether a text is perceived
as coherent is largely dependent on the background knowledge of its
receiver (see Enkvist (1985); Tanskanen (2006: 20–1)). The following
example, for instance, is only coherent for an addressee who happens
to know that garbage is collected every day except on Mondays:4

2
See e.g. Abelen et al. (1993); Knott et al. (2001); Stede (2004) (who all mention
further literature).
3
See e.g. Halliday & Hasan (1976: 23), Brown & Yule (1983: 223–71), Taboada (2004:
1–4) and Tanskanen (2006: 1, 7, 20).
4
Blakemore (1992: 35–6) provides the following examples to illustrate that coherence
is dependent on the (background knowledge of) the perceiver of the text:
(i) The river had been dry for a long time. Everyone attended the funeral.
(ii) If a river has been dry for a long time, then a river spirit has died. Whenever a
spirit dies, there is a funeral. The river had been dry for a long time. Everyone
attended the funeral.
Although the first example is incoherent and therefore incomprehensible for most
Western addressees, speakers of Sissal (an African language spoken in Burkino Fasso
and Ghana) will have no problems in understanding the coherence of (i). They would
probably judge the information provided by example (ii), coherent in the eyes of Western
addressees, as redundant.
 introduction xiii

(2) Speaker A: Would you mind putting the garbage out?

Speaker B: It is Monday.
Tanskanen (2006: 7) therefore argues that coherence ‘resides not in the
text, but is rather the outcome of a dialogue between the text and its
listener or reader.’
The previous example shows that discourse can display coherence
without the presence of cohesion devices. The same holds for the fol-
lowing, often-quoted example, which is originally from Widdowson
(1978: 29):
(3) Speaker A: That’s the telephone.
Speaker B: I’m in the bath.
Speaker A: O.K.
Although there are no linguistic means that specify the relation between
the question by speaker A and the answer by speaker B, speaker A can
still infer what the relation between his question and the answer is. On
the basis of his background knowledge, he infers that speaker B’s answer
is an excuse for not answering the request that is implicit in ‘that’s the
telephone’ (viz. ‘can you answer it, please?’) and therefore concludes
that he has to answer the telephone himself (‘O.K.’).5 Whereas Halliday
and Hasan (1976: 2) defended the view that a text is only coherent if
it contains cohesion devices,6 examples like (2) and (3) gave rise to the
idea that cohesion is less relevant for the creation of unity in discourse
than coherence (Enkvist (1978/1990), Widdowson (1978: 28–9), Brown
and Yule (1983: 195–9)). This idea is strengthened by the fact that the
presence of cohesion devices does not guarantee coherent discourse:
(4) Courses ended last week. Each week has seven days. Each day I feed
my cat. It has four legs, and it is in the garden. The latter has six letters.
(example from Enkvist (1978), slightly modified)

5
Cf. the cooperative principle of Grice (1975: 45–58), stating that discourse par-
ticipants should make their contribution as required, at the stage at which it occurs,
by the accepted purpose or direction of the discourse in question. The fact that the
addressee comes to this conclusion may be seen as a conversational implicature, i.e.
an inference the addressee can make from the presumption that speakers are seeking
to provide useful information.
6
‘If a passage of English containing more than one sentence is perceived as a text,
there will be certain linguistic features present in that passage which can be identified
as contributing to its total unity’ (Halliday and Hasan 1976: 2).
xiv stéphanie j. bakker and gerry c. wakker

Despite the presence of clear cohesive means (anaphoric reference,

repetition, coordination; all in italics), the coherence of the discourse
in question is difficult, if not impossible, to establish.
Although the fact that the presence of cohesion devices is neither
a necessary, nor a sufficient condition for discourse coherence made
scholars like Morgan (1978) and Carrell (1982) doubt about the
importance and explanatory value of cohesion devices altogether,7 the
general opinion is that cohesion devices do help the reader to perceive
the coherence of a text. A strong argument in favour of this view is
that examples like (2) and (3), in which there are no cohesion devices,
are relatively rare. The abundant usage of cohesion devices in most
discourse would be hard to explain if these devices were not helpful
for the addressee for his interpretation of the discourse. In this book,
we share the communis opinio that although cohesion devices cannot
create coherence, they do help the reader understand the unity of the
discourse by making relations explicit. In other words: the relation
between parts of the discourse would be the same without the cohesion
device, but the speaker can choose to specify the relation by the help
of the cohesion device.
Cohesion devices are usually divided into grammatical and lexical
devices, a division that goes back to Halliday and Hasan (1976). As the
term itself says, grammatical cohesion devices create cohesion by gram-
matical means. Standard examples are the use of pronouns to refer to
previously mentioned entities (example 5) and the use of conjunctions
to specify the semantic relation between successive utterances (example
6), but substitution (example 7) and ellipsis (example 8), for instance,
also belong to the grammatical cohesion devices.8
(5) Wash and core six cooking apples. Put them into a fireproof dish.
(6) She was never really happy here. So she’s leaving.
(7) My axe is too blunt. I must get a sharper one.
(8) Joan bought some carnations, and Catherine (. . .) some sweat peas.

7
Morgan (1978), for instance, argues that the fact that we perceive the link between
a pronoun and the preceding full noun phrase to which it refers is due to our assump-
tion that the text is coherent. Carrell (1982) also argues that cohesion is an ‘illusion’
created by the coherence of the text.
8
Examples were taken from Halliday and Hasan (1976: 2, 241, 89 and 143, respec-
tively).
 introduction xv

Lexical cohesion devices create cohesion by lexical means. Devices like

lexical repetition (example 9a), the use of synonymy (example 9b) and
hyponymy (9c) belong to this category:
(9) There’s a boy climbing that tree.
(9a) The boy’s going to fall if he doesn’t take care.
(9b) The lad’s going to fall if he doesn’t take care.
(9c) The child’s going to fall if he doesn’t take care.
In this book, attention is paid mainly to grammatical cohesion devices
in Ancient Greek. Besides the use of pronouns and conjunctions
(especially conjunctive particles), less standard cohesion devices, like
the use of tense and the grammatical form of complements, are taken
into consideration.
The first three chapters study the way in which reference is made to
information that is already familiar on the basis of preceding discourse,
or focus on cohesion devices used to mark that the information pro-
vided by the utterance is already familiar. Anna Bonifazi discusses the
use of third person pronouns in Homer. She claims that although the
choice between zero anaphora, demonstrative pronouns and pronouns
like μιν, νιν or ἑ is commonly assumed to be metrically constrained, a
cognitive-pragmatic approach that takes the accessibility and discourse
relevance of the referent into account is much more fruitful. Against the
background of the so-called ‘referent in the mind’ model (a cognitive
model that assumes that any anaphoric expression triggers the men-
tal representation of the referent) the paper focuses on the difference
between κεῖνος and αὐτός. In the case of κεῖνος, Bonifazi argues that it
is not the deictic or anaphoric distance that is traditionally said to be
inherent in κεῖνος, but the establishment of a visual or social relation-
ship between the speaker and the referent that is decisive for its use.
Αὐτός shares with κεῖνος that it re-activates the mental representation
of a referent to provide relevant information about this referent. It
differs from κεῖνος in that it is a so-called intensifier, which contrasts
the referent, being the centre of the visual field or frame at issue, with
alternatives in the periphery and is therefore very appropriate to create
an isolating or ‘singling-out’ effect.
Luuk Huitink demonstrates that the accessibility of the information
is not only relevant in the case of pronominal reference. He argues
that the syntactic form of the complement of cognitive verbs (like
‘know’, ‘understand’ and ‘believe’) depends on the question whether
the information provided by the complement clause already belongs
xvi stéphanie j. bakker and gerry c. wakker

to the knowledge shared by speaker and addressee. Traditionally, the

difference between participle phrases, ὅτι-clauses and complementary
infinitives is described in terms of semantic presupposition: participle
phrases and ὅτι-clauses would presuppose the truth of the proposition
expressed in the complement, whereas complementary infinitives would
not. The complement clauses in Huitink’s corpus give the impression,
however, that it is pragmatic instead of semantic presupposition that is
relevant. In participle clauses the information is pragmatically presup-
posed in that the information provided has already been asserted in the
preceding context. Ὅτι-clauses, on the other hand, provide information
that is new for the addressee.
Stéphanie Bakker shows that beside pronouns and the choice for
a specific kind of complement, particles can also be used to indicate
the familiarity of the information provided. Although it is commonly
agreed that οὖν is a presentational particle that marks the utterance as
more relevant or more to the point than what precedes, it is debatable
whether οὖν has the same function when it is combined with γάρ. Apart
from the fact that the two particles would contradict each other (γάρ
marking an utterance as less relevant than the preceding discourse),
the context often blocks such an interpretation of οὖν. On the basis of
an analysis of the use of γὰρ οὖν in a number of Platonic dialogues,
both in ongoing discourse and in reactions, Bakker proposes a different
function for οὖν when combined with γάρ. Whereas γάρ still has a text
organizing function, οὖν provides information about the accessibility
of the information, marking that the information provided is to be
expected in view of the preceding discourse or the general knowledge
of the conversation partners.
Particles appear to be used not only to mark the familiarity of the
information provided by the utterance, but are also often used to struc-
ture the discourse. The chapters four to eight deal with these text struc-
turing (or: presentational) devices. Gerry Wakker discusses the function
of the particles οὖν and τοίνυν when they connect sentences or discourse
units. Her corpus consists of the forensic speeches of Lysias. She also
briefly discusses the differences between these particles and connectors
in the traditional syntactic sense (i.e. coordinating conjunctions such
as δέ) and cases of asyndeton. Wakker shows that both οὖν and τοίνυν
have their own semantic value: οὖν indicates that the speaker proceeds
to a new important point (thereby having the effect of indirectly char-
acterizing the preceding unit as relevant but subsidiary to or prepara-
 introduction xvii

tory for what he is going to say now), whereas τοίνυν on the one hand
performs a similar function (by νυν), but on the other adds the nuance:
‘you (= the addressee) must take notice of it because possibly you do
not expect this’ (τοι). This means that both particles prefer their own
type of context, in accordance with their semantic value. Sometimes,
however, they seem to be used in similar contexts. This observation is
argued to be compatible with the so-called Prototype Theory.
Antonio Revuelta Puigdollers describes the contribution of the parti-
cles αὖ and αὖτε to topic management in Ancient Greek. He argues that
both particles (at least in one of their uses) are cohesion devices marking
the introduction of a different discourse topic, and hence having the
effect of opening a new thematic section. He provides the reader with
many examples from Homer and Classical Greek texts, showing that
both particles are used for the introduction of a subtopic, the resump-
tion of a given topic, or (very rarely) the introduction of a topic that is
altogether new. In all these examples the particles function as bound-
ary markers that highlight the transition between different discourse
units. As a consequence, αὖ and αὖτε are often found after topic closing
devices (such as closing formulas and the particle combination μὲν οὖν)
and together with other topic change markers, such as left dislocation,
prolepsis and τί δέ questions.
Marietje van Erp Taalman Kip studies the difference between καὶ
μήν and καὶ δή in tragedy and comedy. In the standard commentaries
these particle combinations are often said to be (nearly) equivalent,
but Van Erp Taalman Kip argues that the choice is determined by the
nature of the context, notably by the question whether there is a shift
in the focus of attention. Coherence, then, is marked in different ways
by these particle combinations. In the case of καὶ μήν there is a shift
in the focus of attention in that the speaker either marks an entry that
was not prepared for by the words that immediately precede or marks
the transition to a new subject. When there is no such shift, the speaker
corrects or contradicts his addressee. Καὶ μήν thus marks some kind
of incision, and is accordingly placed at the opening of a sentence. Καὶ
δή, on the other hand, is used to mark an entry that has been prepared
by the preceding words, or something said or done that is related to
the subject under discussion. When καὶ δή marks agreement, it does
so without contestation. Van Erp Taalman Kip argues that these char-
acteristics probably make that καὶ δή is sometimes removed from the
sentence opening.
xviii stéphanie j. bakker and gerry c. wakker

Annemieke Drummen focuses on another particle in tragedy and

comedy and studies the ways in which turn-initial ἀλλά enhances dis-
course cohesion. Turn-initial ἀλλά frequently occurs in both genres,
and in comedy it appears to be even the most frequent turn begin-
ning. Drummen tests the hypothesis set up by Basset (1997) on the
use of ἀλλά in Aristophanes’ Ranae on the basis of a larger corpus.
She shows that Basset’s hypothesis is basically correct and she presents
some refinements. In all cases investigated, she argues, the function
of turn-initial ἀλλά can be interpreted as marking a correction of the
preceding words or actions. The corrected (substituted) element is either
an explicitly stated or a presupposed element, an implication or the
discourse topic. Whereas ἀλλά is predominantly found in corrections
of explicit elements, at turn beginnings it most often corrects implica-
tions. All occurrences of turn-initial ἀλλά have in common that they
mark a relation that would also have been present without the particle;
but when ἀλλά is used the relation is explicit and consequently easier
to discern for the addressee.
Coulter George’s paper is the last chapter on particles. He does not
focus on one or two specific particles and their function as cohesive
discourse markers, but takes two opposite observations as his starting
point, the first by Denniston (1954: lxxii–iii), viz. that Greek conversation
is characterized by the use of many particles and the second by Duhoux
(1997a, 1997b, 1998, 2006), saying that particles are more typical of writ-
ten than of spoken Greek. To solve this discrepancy George makes use
of the notion of diaphony. By means of a detailed study of the context
of the particles μήν, οὖν, and γε μήν, George argues that particles are in
fact more typical of dialogical Greek than Duhoux argued, and secondly
that those particles that are more common in non-dialogical Greek
function primarily at the representational and presentational level of
the clause. George further argues that that there is a clear discrepancy
between the frequency of particles in dialogical and non-dialogical
Greek (as particles are more common in narrative than in speeches in
Homer) that is independent of the variable of ‘orality’.
The chapters nine to twelve show that Greek particles are not the
only devices used for text structuring and hence cohesion, but that
Greek tenses can also serve this purpose. In chapter nine, Rutger
Allan argues that narrative is a mixed genre with different text types
or narrative modes. He aims to make a typology of narrative modes in
Ancient Greek narrative. The narrative modes, he argues, depend on
the relation between the point of view of the narrator and the presenta-
 introduction xix

tion of the text. He thus distinguishes the displaced diegetic mode, the
immediate diegetic mode, the descriptive mode, and the discursive mode,
and argues that tense-aspect-marking is the most important distinctive
linguistic feature of the narrative modes: aorist and imperfect are typi-
cally used in the displaced diegetic mode, the historical present in the
immediate diegetic mode, the imperfect in the descriptive mode and
the present, perfect and future (all moods) in the discursive mode. His
analysis of Euripidean messenger speeches shows that other relevant
linguistic phenomena are, for instance, negations, particles, moods, the
narrator’s point of view and the type of statement. He next addresses
the relationship between the narrative modes and plot-structure. The
function of the narrative modes within the larger organisation of the
narrative is demonstrated by a refined analysis of the messenger speech
in Euripides’ Andromache.
Louis Basset focuses on the aspectual opposition between the present
and aorist stem when a Greek verb in a narrative is accompanied by
an adverbial expression of duration. The traditional description—the
present stem has an imperfective value, expressing ‘a not-completed
state of affairs’, whereas the aorist stem has a perfective or confective
value, expressing ‘a completed state of affairs’—appears to be not fully
correct. In his corpus, Herodotus’ Histories, there are examples where
the present stem is, unexpectedly, used for completed states of affairs
and does not have an imperfective meaning. In all such examples the
state of affairs in question is inserted into a natural narrative sequence:
instead of imperfective, it may be said to be continuative and opposed
to a discontinuative aorist. Basset argues that this use of the present
stem is related to the structure of the narrative and reinforces narra-
tive cohesion. What is in question here is the cohesion between states
of affairs that hang together, each of them leading to the next. It may
thus be said that the imperfect is a narrative past (i.e. a past not bound
to the narration time), just like the French passé simple, whereas the
aorist may be characterized as a speech past (a past bound to the speech
time), just like the French passé composé.
In the last paper on tense and aspect, Sander Orriens discusses the
role of the Greek perfect as a cohesion device. Orriens’ description is
based on a broad definition of cohesion that is not restricted to intra-
textual relations, but also to relations between the text and the extra-
textual or situational context. Within this framework, Orriens argues
that the Classical Greek perfect can be considered to play a role as a
situating cohesion device. This role is directly related to the properties
xx stéphanie j. bakker and gerry c. wakker

of its semantic value, which he describes as follows: the Greek perfect

is concerned with a reciprocal relationship between a completed past
state of affairs and the moment of speech. By this, Orriens means
that a speaker, when using the perfect in discourse, refers to a com-
pleted past state of affairs and simultaneously links this state of affairs
explicitly to the moment of speech. By doing this he directly involves the
state of affairs in the present communicative context. The perfect thus
functions as a kind of objective modality marker, in that it underlines
the actuality that the speaker ascribes to the past state of affairs within
the context of the present communicative situation. In this respect the
perfect contrasts with the neutral aorist, which lacks this actuality and
merely refers to the past state of affairs.
The final chapter is different in that it does not focus on a specific
cohesion device, but discusses various cohesion devices that occur in
one corpus, viz. the proem of Hesiod’s Theogony. Albert Rijksbaron
argues that although at first glance the proem does not seem a very
coherent passage, as its various parts seem rather independent, a closer
look reveals that what he calls ‘the continuity of progress’ of the proem
is used as a unifying principle. By discussing the use of the various
tense forms and adverbs used in the proem in the form of a running
commentary, Rijksbaron shows that a correct interpretation of these
tenses and adverbs leads to a much more coherent interpretation than
is usually assumed. Furthermore, Rijksbaron shows that Hesiod also
creates coherence by using lexical cohesion devices, mainly in the form
of the repetitive use of verb forms expressing a sound and recurring
reference to Hesiod as servant of the Muses.

We hope to have provided the readers with a volume that gives a good
impression of recent research in the field of Greek linguistics. As all
examples are translated into English this volume will not only be of
interest for classical scholars, but also for general linguists interested
in discourse coherence and cohesion within a functional-cognitive
framework.

Stéphanie Bakker
Gerry Wakker
 CHAPTER ONE

DISCOURSE COHESION THROUGH THIRD PERSON PRONOUNS.

THE CASE OF ΚΕIΝΟΣ AND ΑYΤΟΣ IN HOMER

Anna Bonifazi

1. Introduction

Within the Homeric diction various lexical choices are available to recall
a referent that has been previously introduced,1 that is, demonstrative
pronouns such as οὗτος, κεῖνος, weak demonstrative pronouns such as
ὁ, enclitic pronouns such as μιν, νιν, and no pronouns at all, namely
null anaphors.2 The oblique cases of αὐτός should be included as well,
as they are commonly said to equal plain third person pronouns. The
choice of the narrator in the individual instances is usually explained
by a general rule of co-reference, which concerns the distance to the
previous mention of the referent (or verbal antecedent): demonstrative
pronouns are expected to be used for farther antecedents, whereas weak
demonstratives and null anaphors are expected to be used for nearer
antecedents. However, in a number of cases this rule seems not to be
valid, as the antecedents of οὗτος and of κεῖνος may occur quite near
the pronouns themselves. Furthermore, grammars and lexica do not
make clear what makes the difference in the use of null anaphors vs.
weak demonstratives vs. oblique cases of αὐτός for referents that are
easy to retrieve. Metrical constraints could play a role in all of this, but
it is very unlikely that they are the only or the main factor involved in
the choice of the anaphor.
The topic of third persons pronouns is wide and complex; nonethe-
less, the current contribution is meant to offer some insights concerning
an alternative view of the Homeric uses of two third person pronouns
in particular, that is, κεῖνος and αὐτός (section 3 will explain the reason

1
The focus of interest of this contribution is backward reference. Cataphoric or
proleptic reference will be left out.
2
A relevant fact to bear in mind is that Greek is a pro-drop language; that is, gram-
matical subjects and objects may be omitted.
2 anna bonifazi

of this choice). This alternative interpretation springs from a pragmatic

and cognitive approach to the linguistic phenomenon of anaphora.
‘Anaphora processing’, which refers to how we interpret anaphoric
expressions (noun phrases and pronouns) in both spoken and written
texts and how we keep track of the people and objects involved, is
currently one of the most productive topics of research in disciplines
such as computational linguistics, philosophy of language and cognitive
psychology. Such activity significantly contributes to our evaluation of
the discourse cohesion characterizing those texts. Already in Apollo-
nius Dyscolus (Περὶ ’Αντωνυμίας) anaphora, from ἀνά + φέρω, ‘carry
back’, is defined with respect to the status of referents in memory, as
‘a reference to objects that have already previously figured in discourse
or are generally known’ (cf. Bosch (1983: 7)).
In the research I have done on third person pronouns in Homer so
far, I apply some contemporary theoretical frameworks—from cognitive
psychology and discourse studies—centered on the so-called ‘referent
in the mind’ model,3 which contrasts with the so-called ‘referent in
the text’ model. The latter assumes that any anaphoric expression has
an ‘antecedent’, commonly meant to be a more or less recent verbal
mention of the referent of the anaphoric expression. Conversely, the
former model—a cognitive one—assumes that any anaphoric expression
triggers the mental representation of the referent, the referent being an
extralinguistic entity, either fictional or non-fictional. The ‘referent in
the mind’ model presents outstanding advantages. First, it is possible
to account for anaphoric expressions missing any verbal antecedent,
as in the following example (Cornish (2002: 473)):
(1) A neighbor’s father has been in hospital for a week already
Anne to her neighbor, seeing her looking haggard:
How is he?
‘He’—that is, the neighbor’s father—does not refer to an individual that
has previously figured in discourse, but to an individual about whom
the discourse participants share some previous knowledge. Second, it
is possible to account for indirect anaphors, as the following instance
shows:

3
Landmark works for that are Emmott (1997) and Cornish (1999).
 the case of ΚΕIΝΟΣ and ΑyΤΟΣ in homer 3

(2) Kill an active, plump chicken. Prepare it for the oven, cut it into four
pieces and roast it with thyme for 1 hour.4
The referent of ‘it’ actually is not the entity verbally mentioned by
means of ‘an active, plump chicken’, since what is prepared to put in
the oven is a dead and no longer active chicken. Yet, readers of the
recipe successfully interpret the anaphor as they monitor the chicken
in the various phases of preparation, because ‘it’ gives access5 to the
entity ‘chicken.’ Third, the distinction between anaphora and deixis
collapses: in both cases the verbal sign for the referent is processed
as the retrieval of the mental representation of that entity, the only
difference being whether the referent is present to the sight or to the
memory of the recipient.

2. An alternative reading of Homeric third person pronouns

The application of the ‘referent in the mind’ model to anaphoric

expressions occurring in the Homeric poems may affect exegesis and
literary interpretation of passages much more than one would expect.
I posit that the extraordinary richness of third person pronouns in
Homeric language can be better grasped in the light of the ‘referent in
the mind’ model.
Any performer of Homer had a sufficient linguistic and poetic com-
petence to formulate their stories in such a way that characters could be
verbally recalled felicitously,6 for his own and his listeners’ convenience.
Being able to proceed with no difficulty—which I call ‘convenience’—is
something that inseparably involves the metrical, the syntactic, the
semantic, the pragmatic and the cognitive level, all of this at the same
time. The present line of argumentation deals with the pragmatic and
cognitive convenience linked to the utterance of some third person

4
Example from Brown and Yule, quoted by Emmott (1997: 201).
5
Accessibility has become a key concept of cognitive studies about anaphora
processing: the kind of anaphora chosen by the speaker may reflect the accessibility
degree of the referent. A crucial work on that is Gundel et al. (1993), concerning the
so-called ‘Givenness Hierarchy.’
6
‘Felicitous’ deliberately reminds the reader of Austin’s view of performative
utterances, which can be felicitous or infelicitous; that is, they can be successfully or
unsuccessfully conveyed. Cf. Austin (1962).
4 anna bonifazi

pronouns. In other words, discourse cohesion through anaphora is here

considered in terms of discourse memory and discourse relevance.
Two background assumptions are to be premised: first, the narrator
and the recipients of narrative texts constantly monitor the presence of
a certain character or of a certain object ‘on the fictional stage’; second,
any anaphoric retrieval of a certain character or object reflects both the
cognitive status of that character or object in memory and their conspi-
cuity degree at any particular moment of narration. To use Cornish’s
words, the referents available to the recipient ‘are ranked in terms of
their saliency, (. . .) and are located in different regions of working
memory according to their current levels of activation.’7 If a referent
is immediately accessible to the discourse memory of the recipient an
unaccented pronoun or null anaphor is probably used. Conversely, if
the referent is not immediately accessible but is familiar to the recipi-
ent and is going to be re-activated, an accented third person pronoun
or a demonstrative pronoun is probably used.8 Let us apply this to the
beginning of Odyssey 2, for example:
(3) ἦμος δ’ ἠριγένεια φάνη ῥοδοδάκτυλος Ἠώς,
ὤρνυτ’ ἄρ’ ἐξ εὐνῆφιν Ὀδυσσῆος φίλος υἱός,
εἵματα ἑσσάμενος, Ø περὶ δὲ ξίφος ὀξὺ θέτ’ ὤμῳ,
Ø ποσσὶ δ’ ὑπὸ λιπαροῖσιν ἐδήσατο καλὰ πέδιλα,
Ø βῆ δ’ ἴμεν ἐκ θαλάμοιο θεῷ ἐναλίγκιος ἄντην.
Ø αἶψα δὲ κηρύκεσσι λιγυφθόγγοισι κέλευσε
κηρύσσειν ἀγορήνδε κάρη κομόωντας Ἀχαιούς.
οἱ μὲν ἐκήρυσσον, τοὶ δ’ ἠγείροντο μάλ’ ὦκα.
αὐτὰρ ἐπεί Ø ῥ’ ἤγερθεν ὁμηγερέες τ’ ἐγένοντο,
Ø βῆ ῥ’ ἴμεν εἰς ἀγορήν, Ø παλάμῃ δ’ ἔχε χάλκεον ἔγχος

7
Cf. Cornish (1999: 255). An earlier account of monitoring activities is in Givón
(1992). Some seminal works introduce the notion of ‘focus framework’ for discourse
comprehension (cf. Sanford and Garrod (1981)) and the notion of attentional state of
discourse structure (cf. Grosz and Sidner (1986)). A pioneering cognitive-oriented study
on participant tracking in Latin literature is in Bolkestein and Van de Grift (1994).
The monitoring of Homeric characters in a cognitive-oriented perspective was first
introduced by Bakker (1997c). He analyzes the uses of ὁ δέ-clauses with respect to
the insertion of full names or name-epithet structures within what he calls ‘Homeric
framings’ (1997c: 86–122; see in particular 108–11). Bakker’s distinctions about the
accessibility of the referent are expressed in Chafe’s terms of the active, semiactive and
inactive status of the character in the consciousness of poet and listeners, on which
cf. Chafe (1994: 71–6).
8
Cf. Gundel et al. (1993) and Chafe (1994) among others. By ‘accented pronoun’
Cornish (1999: 119–24) means a pronoun whose utterance is characterized by a high
pitch level (iconically corresponding to capital letters, in writing). Accented pronouns
may share a deictic value with demonstrative pronouns (for example, in front of several
individuals one could say ‘I’d like to have HER in the group’).
 the case of ΚΕIΝΟΣ and ΑyΤΟΣ in homer 5

Now when the young Dawn showed again with her rosy fingers,
the dear son of Odysseus stirred from where he was sleeping,
and put on his clothes, and slung a sharp sword over his shoulder.
Underneath his shining feet he bound the fair sandals
and went on his way from the chamber, like a god in presence.
He gave the word now to his clear-voiced heralds to summon
by proclamation to assembly the flowing-haired Achaeans,
and the heralds made their cry, and the men were assembled swiftly.
Now when they were all assembled in one place together,
he went on his way to assembly, in his hands holding a bronze spear
(Od. 2.1–10; the narrator) (translation Lattimore)
The only actor ‘on the stage’ is Telemachus, already at the center of
the performer’s attention at the end of the preceding book (cf. Od.
1.437–44). He is recalled as Ὀδυσσῆος φίλος υἱός (‘the dear son of
Odysseus’, 2.2) and is the only agent (and grammatical subject) of the
following 5 clauses (cf. lines 3–7) without any anaphoric expression
occurring for him. In other words: lines 3–7 include 4 null anaphors
(see the conventional symbol Ø), which express that Telemachus is a
highly accessible referent. At lines 6–7 Odysseus’ son is said to call some
heralds to convene the assembly of the Achaeans. οἱ μὲν and τοὶ δ’ (line
8) activate the easily accessible entities corresponding to the heralds
and Achaeans respectively, who have just been mentioned. Finally, at
lines 9–10 both the latter (the heralds + the Achaeans) and the former
(Telemachus) are recalled via null anaphors: at that point indeed they
have all become highly accessible to the recipients.
However, as far as the unfolding of the Homeric narrative is con-
cerned, the accessibility degree of referents does not suffice to justify
the choice of anaphoric expressions. Several passages, for instance,
include different third person pronouns occurring relatively close to
each other and having the same referent (one of these passages will be
commented at the end of this contribution). In order to explain, for
example, what the cognitive difference is between the Homeric use
of μιν, νιν and that of ὁ/τό, other factors may play a role, such as the
switch between different discourse acts9 and the relevance of the referent
in what comes next. Here is where discourse analysis and pragmatics
can complementarily help.

9
By discourse act I mean the smallest unit of communicative behavior, after Hannay
and Kroon: each act denotes ‘each single step which language producers take in order
to achieve their communicative aims’ (2005: 95).
6 anna bonifazi

3. Spotlighting κεῖνος and αὐτός

The attention given to κεῖνος10 and αὐτός is, generally speaking, lim-
ited to their anaphoric function and to their main use as simple third
person pronouns.11 What seems to characterize κεῖνος and αὐτός in
their so-called ‘emphatic’ use—especially in the nominative case—is
basically remoteness for the former, and either sameness or selfness
for the latter. A major difference between κεῖνος and αὐτός is that the
latter can agree with all three persons as both adjective and pronoun,
which is not the case for κεῖνος: we can have ‘I αὐτός’, ‘you αὐτός’,
‘he/she/it’ αὐτός’ but only ‘he/she/it κεῖνος’.
What I would like to argue about the Homeric occurrences of κεῖνος
and αὐτός as pronouns is twofold: first, they never work as plain co-ref-
erences; second, their occurrence corresponds to specific cognitive and
pragmatic purposes. In line with the cognitive framework introduced,
both of them work as demonstrative pronouns: in the retrieval of the
referent they imply a greater cognitive effort than the one required by
unaccented third person pronouns. The reason may be either that the
referent is re-activated in a new discourse unit, or that a different or
new aspect concerning the referent is going to be conveyed, or both.12
In particular, it is suggested that their utterance serves cognitive pur-
poses, that is, to re-activate referents that figured in preceding sections
of discourse—often immediately preceding, in the case of αὐτός—, and
pragmatic purposes, that is, to make them a conspicuous piece of infor-
mation of what comes next. At the pragmatic level, different commu-
nicative intentions may be underlying the choice of each pronoun; this
concerns the orientation of the recipient’s attention and the speaker’s
involvement in recalling a referent. The result is a rather sophisticated
picture of the Homeric uses of κεῖνος and αὐτός, which can shed light
even on the poetic relevance of using such grammatical marks.

10
The Iliad and the Odyssey include 18 forms of ἐκεῖνος vs 184 forms of κεῖνος.
κεῖνος is the usual form in epic and lyric (the aeolic form being κῆνος). Herodotus
seems to prefer ἐκεῖνος; attic prose has ἐκεῖνος.
11
The attention of scholars to the pronouns ὅδε and οὗτος, by contrast, commonly
involves the borderline between the mere anaphoric (or cataphoric) and the deictic
functions. As for Homer, cf. Monro (1891: 216–8); Chantraine (1953: 169); on the
deictic and dialogic values of οὗτος cf. Bakker (1999). On deictic ὅδε and οὗτος in
Pindar, cf. Bonifazi (2001: 39–43 and 2004a: 398–401).
12
For a discourse-oriented interpretation of the anaphoric choices ille and hic in
Ovid, cf. Kroon (2007b).
 the case of ΚΕIΝΟΣ and ΑyΤΟΣ in homer 7

A Homeric feature that is notable on a large scale—here just briefly

mentioned—is, for example, the masterful interplay between Odysseus
as κεῖνος and Odysseus as αὐτός in the Odyssey.13 Discourse cohesion in
this case encompasses the whole sequence of 24 books. Indeed, Odysseus
is κεῖνος in 60% of the total occurrences of κεῖνος in the poem (possibly
covering each male character involved); he is αὐτός in 43% of the total
occurrences of αὐτός. None of the other characters is pronominalized in
such a way as Odysseus is. It is possible to observe not only a dynamic
poetic treatment of the two pronouns in line with the evolution of the
perceptions of Odysseus by other characters as well as by himself, but
also a subtle playfulness involving what the narrator puts in the mouths
of different characters in order to create polysemy. The significance of
Odysseus κεῖνος essentially covers the mourning about his supposed
death, his imagined appearance, and, finally, his actual appearance in
front of the Ithacan people. The significance of Odysseus αὐτός covers
his isolation from the companions, his prominence as the (visual) center
of attention, and the perception of his true identity.

4. Previous accounts

The influential study on (ἐ)κεῖνος by Havers (1906) identifies three

major meanings: ‘that’ (deictically referring to what is far in space or in
time), ‘the other’ (i.e., what is on the other side, such as enemies or dead
people) and ‘he/she/it’ (regular third person pronoun).14 The referents
of (ἐ)κεῖνος are typically excluded from the utterance situation (they
are neither the speaking ‘I’ nor the co-present ‘you’).15 Remoteness in
time, related to the past, suggests the idea of ‘well-known’ (cf. Latin
ille); in the positive sense it conveys venerability.16 Remoteness related
to the co-text, on the other hand, refers to the use of the pronoun when

13
The topic is extensively treated by the author of this article in a submitted mono-
graph.
14
Havers (1906: 3–5). Its fundamental characterization as demonstrative of the third
person agrees with the seminal tripartition by Brugmann (1904), namely Ich-Deixis (cf.
ὅδε), Du-Deixis (cf. οὗτος) and Jener-Deixis (cf. (ἐ)κεῖνος).
15
In a recent article about the use of ὅδε, οὗτος and (ἐ)κεῖνος, Ruijgh (2006: 159)
states: ‘In drama it [(ἐ)κεῖνος] refers to what is not present on the stage. (. . .) It is, in
fact, mainly used to refer to something outside the situation of utterance.’
16
Cf. Havers (1906: 4–5); Kühner-Gerth (1898/1904: 1.650); Humbert (1945: 33).
Cf. also LSJ s.v. 2.
8 anna bonifazi

the verbal trace of the referent is remote (or it occurs far above, in
the case of a written text). Interestingly, the latter idea is affirmed but
also contradicted in the literature. Thus, LSJ reports ‘when οὗτος and
ἐκεῖνος refer to two things already mentioned, ἐκεῖνος prop. belongs
to the more remote one, in time, place or thought (. . .); but ἐκεῖνος
sts. [sometimes] = the latter.’17 In an early dictionary of Xenophon’s
Anabasis, White and Morgan (1892: 68) write: ‘[ἐκεῖνος] used prop.
of a person or thing remote in thought or actual distance from the
speaker or subject in hand, but sometimes refers to one near at hand
or lately mentioned, and even to the subject itself (. . .); it is often used
as a strong form of the pers. pron.’ An analogous contradictory remark
is about the reference to what is presumably not present in the utter-
ance situation. Beside several anaphoric instances, LSJ records also
deictic uses of ἐκεῖνος, ‘with simple demonstrative force’, and quotes
Homer (Od. 18.239 Ἶρος ἐκεῖνος ἧσται, translated as ‘Iros sits there’)
and Thucydides (1.51 νῆες ἐκεῖναι ἐπιπλέουσιν, translated as ‘There are
ships sailing up’).18
Thus, the descriptions of the uses of (ἐ)κεῖνος in Greek literature
indicate a deictic or an anaphoric distance, but they also tell us that
this distance can be blurred. They link its basic meanings to characters
that are out of the utterance situation par excellence (third persons), but
they also say that from time to time it marks somebody or something
that does appear to the sight of the speaker. As for the Homeric uses
of κεῖνος, secondary literature does not substantially change the frame
outlined above.19
Classical Greek αὐτός is said to convey sameness in the attributive
position and selfness in the predicative one; it often occurs with other
personal or demonstrative pronouns; in the non-nominative form the
αὐτός pronoun is said to equal simple third person pronouns.20 Kühner-

17
Cf. also Smyth (1920: 309). The examples given are from Xenophon and Aristotle.
18
LSJ s.v. 4. Italics in the text. In Homer κεῖνος has a deictic value not only when
used as a predicative adjective but also when used as a pronoun. Cf. Il. 19.344 and
Od. 22.165.
19
Cf. Monro (1891: 217); Chantraine (1953: 169–70); LfgrE s.v.; Wace and Stubbings
(1967: 138–9); Ebeling (1885: 382); Autenrieth (2002: 102). Magnien (1922: 157–8)
states: ‘ἐκεῖνος, ou κεῖνος, ne sert jamais à désigner une personne ou une chose aperçues
au loin. Il s’applique le plus souvent à une personne absente, à une chose placée hors
de la vue.’ However, interesting remarks on the recurrence and focalizing function of
Homeric κεῖνος, appear in De Jong 2001: ad Od. 1.163 and De Jong 1987: 234–6 .
20
Cf. for example, Kühner-Gerth (1898/1904: 1.651) and Smyth (1920: 302).
 the case of ΚΕIΝΟΣ and ΑyΤΟΣ in homer 9

Gerth (1898/1904: 1.651) say its status is between that of a personal

pronoun and that of a demonstrative pronoun. Humbert (1945: 34 and
40) includes αὐτός in the demonstrative elements; its demonstrative
meanings, he says, relate to the notion of identity. Also Sadoulet (1984)
focuses on the relation of identity underlying third person pronominal
uses, and concludes that they reflect the meta-linguistic intention to give
prominence to the topic of the utterance.21 Taillardat (1987) unifies the
meanings relating to selfness (‘ipséité’) and those relating to sameness
(‘identité’) by pointing out a common semantic nucleus ‘one’ from
which both isolation (‘oneself ’, ‘alone’) and unity (‘the same’) stem.
Biraud (1990) sees in αὐτός a marker of individualization (‘déterminant
d’individualisation’) that connotes the boundaries and the fullness of
the designated referent at the same time.22
Striking remarks—cognitive ante litteram—come already from Her-
mann (1827). He states that αὐτός refers to something the thinking
of which is going to be continued, kept, and reiterated; its main trait
is to show the prominence and distinctiveness of that something.23
As for Homer, literature underscores the unsystematic use of αὐτός
as reflexive, the sense of aloneness, of isolation as well as the expres-
sion of identity and of opposition.24 Smyth (1920: 302) summarizes
the meaning of αὐτός as follows: ‘In Homer [it] denotes the principal
person or thing, in opposition to what is subordinate, and is intensive
by contrast.’ However, the most important source of insights into the
different meanings of αὐτός in Homer is a monograph by Wagnon
(1880): beside the marking of sameness and the reflexive uses, Wagnon
highlights the sense of contrast or prominence with respect to other
objects (usually a plurality of objects), the pointing to the physical body,

21
‘Comme c’est de ‘lui’ que l’on parle, ce ‘lui’ est en fort relief dans la macro-structure
que les interlocuteurs se construisent à partir de ce qui vient d’être dit (ou écrit). C’est
le référent qu’on rappellera le plus vite à la conscience.’ Cf. Sadoulet (1984: 62).
22
‘Aὐτός permet donc à la fois de restreindre le dénoté du substantif à ses limites
propres et de le considérer dans sa plénitude; c’est à dire qu’il assure l’individualisation
de ce dénoté.” ’ Cf. Biraud (1990: 98).
23
‘Sive Graece αὐτός, sive Latine ipse (. . .) id proprie nihil aliud significat, quam hic,
de quo loquor. Qui autem tali pronomine notionem eius rei, de qua loquitur, iterat,
is facit hoc eo consilio, ut se quam maxime de ista re loqui ostendat. Hinc primarius
istius pronominis usus in eo versatur, ut rem ab aliis rebus discernendam esse indicet.’
(original italics). Cf. Hermann (1827: 308–9).
24
Cf. respectively Chantraine (1953: 153 and 157); Adrados (1992: 319); Monro
(1891: 218); Chantraine (1953: 155–6).
10 anna bonifazi

the reference to the nucleus of an object as distinct from ornaments,

the mark of possession toward an object, the marking of identity of
somebody contrasting his/her current actions, and, finally, the mark
of full identity (‘identité complète’).

5. A cognitive and pragmatic interpretation of Homeric κεῖνος

In the Homeric poems the antecedent of κεῖνος can be found at some

distance from the pronoun, but it may also occur close or very close
to κεῖνος itself.25 Generally speaking, independently of the grammatical
case in which it occurs, the pronoun κεῖνος cognitively makes the recipi-
ents (both the internal addressees of speeches and the external listeners
or readers) re-activate the mental representation of the referent—which
may or may not be mentioned in the previous sentences—and it signals
that a conspicuous aspect of that referent is about to be communicated.
In the majority of the cases the referent is an individual. I submit that
κεῖνος referring to somebody who re-appears in the discourse memory
of the recipient shares some pragmatic properties with κεῖνος referring
to somebody present to the eyes or mind’s eyes of the recipient: in both
cases a visual or a social or both a visual and a social26 relationship is
established between the speaking ‘I’—who utters the pronoun—and the
referent κεῖνος refers to. The conspicuity of the referent that is picked up
coincides with the establishment of these relationships. The visual one
may include the physical appearance of somebody previously absent or
far, in front of the speaker (in ‘epiphanic’ situations), or the momentary
connection in the mind’s eyes with somebody who is supposedly far
away or dead. The social relationship may include a sense of distance
that is either displayed in a negative way, through wrath and contempt,
or in a positive way, through veneration and praise.27 A most noticeable
characteristic of κεῖνος is that it is hardly ever uttered by the narrator
in the Iliad and in the Odyssey; with a very few exceptions, it is only

25
Cf., for example, Il. 5.894; 14.48; 24.44; Od. 4,109 and 145; 14.163.
26
The term ‘social’ here captures the fact that the relationship established involves
an individual, but it actually mirrors a social appraisal by an entire community.
27
Pindar seems to use κεῖνος in quite similar ways, except, of course, for any overt
sign of wrath or contempt, as the latter tend to be silenced in songs of praise. Cf.
Bonifazi (2004b). On κεῖνος as a marker of social deixis, cf. Bonifazi (2001: 48).
 the case of ΚΕIΝΟΣ and ΑyΤΟΣ in homer 11

uttered by characters.28 The following examples exemplify the kinds of

conspicuity being suggested.
(4) ἔγνω δ’αἶψ’ ἐμὲ κεῖνος, ἐπεὶ ἴδεν ὀφθαλμοῖσι
At once that one knew me, when he saw [me] with his eyes.29
(Od. 11.390; Odysseus about Agamemnon’s soul)
This is an instance of epiphanic κεῖνος, that is, κεῖνος marking the
sudden appearance of a certain figure to someone’s eyes, along with
some wonder. Even though the soul of Agamemnon has been actually
introduced a few lines before (cf. 387–8), it is only at 390 that the con-
spicuity of the referent is conveyed, through the eye-contact established
between Odysseus (the speaking ‘I’) and Agamemnon’s soul (the referent
of κεῖνος). Furthermore, κεῖνος may convey the venerability of divine
figures or dead heroes, as in the following passage:
(5) (. . .) ἀτὰρ γνώσεσθε καὶ ὔμμες·
ῥηΐτεροι γὰρ μᾶλλον Ἀχαιοῖσιν δὴ ἔσεσθε
κείνου τεθνηῶτος ἐναιρέμεν. (. . .)
(. . .) You will know this as well:
you will be the easier for the Achaeans
to slaughter, now that that one is dead.
(Il. 24.242–44; Priam to the Trojans)
On his way to Achilles’ tent, Priam warns the Trojans about the lethal
dangers they have to face since Hector has died. κεῖνος does not express
only the love and respect of a father, but also the praise and attitude of
veneration by the community toward a hero who has died in war.
An opposite—but equally emphatic—feeling is expressed by means
of κεῖνος in the following case:

28
As for the exceptions: at Il. 2.37 and 482, 4.543 and 21.517 the narrator uses the
temporal phrase ἤματι κείνῳ to indicate the shared knowledge (between him and the
audience) of a well-known day; these are, incidentally, cases in which κεῖνος does
not refer to individuals. Αt Il. 16.648 the narrator is reporting Zeus’ thoughts about
the chance for Hector to kill Patroclus (Ζεὺς / (. . .) / (. . .) / (. . .) μερμηρίζων / (. . .)
καὶ κεῖνον (. . .) / (. . .) φαίδιμος Ἕκτωρ / χαλκῷ δῃώσῃ ‘Zeus / (. . .) / (. . .) pondering /
(. . .) [whether] that one as well [Patroclus] (. . .) / (. . .) glorious Hector / should kill’).
At Od. 24.19 κεῖνος refers to Ajax in a line that resembles an embedded collective
lament; at Od. 13.111 the narrator uses κείνῃ for the gods’ access to the cave of the
Nymphs.
29
Unless otherwise cited, all translations of the Iliad and the Odyssey are (adapta-
tions) from Lattimore.
12 anna bonifazi

(6) Ἀτρεΐδη κύδιστε ἄναξ ἀνδρῶν Ἀγάμεμνον

μὴ ὄφελες λίσσεσθαι ἀμύμονα Πηλεΐωνα
μυρία δῶρα διδούς· ὃ δ’ ἀγήνωρ ἐστὶ καὶ ἄλλως·
νῦν αὖ μιν πολὺ μᾶλλον ἀγηνορίῃσιν ἐνῆκας.
ἀλλ’ ἤτοι κεῖνον μὲν ἐάσομεν ἤ κεν ἴῃσιν
ἦ κε μένῃ· (. . .)
Son of Atreus, most lordly and king of men, Agamemnon,
I wish you had not supplicated the blameless son of Peleus
with innumerable gifts offered. He is a proud man without this,
and now you have driven him far deeper into his pride. Rather
we shall pay him no more attention, whether he comes in with us
or stays away. (Il. 9.697–702; Diomedes to Agamemnon) (translation
Lattimore)
‘We shall pay him no more attention’, that is, Lattimore’s translation,
assumes κεῖνος to be a plain third person pronoun. I suggest that the
choice of κεῖνος actually deals with the speaker’s communicative inten-
tion to convey anger against Achilles. Achilles is recalled as ‘that one’,
as the speaker, namely Diomedes, wants to keep distance from him, no
differently from what prosodically marked ‘that’ in English may imply.30
κεῖνος expresses social distance; this is what the conspicuity of Achilles
in the discourse act performed by Diomedes consists in.

6. A cognitive and pragmatic interpretation of Homeric αὐτός

What follows is an effort to combine several insights and inputs from

earlier literature in a consistent framework. The starting point is
the notion according to which ancient Greek αὐτός—just like Latin
ipse—belongs to the large category of intensifiers.31 A crucial property
of intensifiers is the following: they ‘structure the set of referents under
consideration (. . .) in terms of center and periphery’; they ‘evoke a
set of alternatives’, the focus of the intensifier being the center, and
the alternatives being the periphery (König (2001: 749)). This center-
periphery idea can be shown to underly a remarkable quantity of
Homeric instances of αὐτός. In Homer the mention of the entity
αὐτός refers to, that is, the center, is very frequently accompanied by

30
On emotional implications of some uses of ‘that’ and ‘this’ in English, see Lakoff
(1974).
31
Cf. König (1998, 2001) on intensifiers in many different languages; Puddu (2005)
on intensifiers in ancient IE languages; Bertocchi (2000) on Latin ipse as intensifier.
 the case of ΚΕIΝΟΣ and ΑyΤΟΣ in homer 13

the verbal mention of the periphery. In the majority of the cases the
center is represented by a single individual (it is seldom represented
by an object). Conversely, the periphery is often a plurality of entities
that could potentially constitute the alternatives. What is the center
and what is the periphery in the Iliad and in the Odyssey involves
the informational structure—the referent of αὐτός becomes the most
relevant topic of the discourse—and visualization. αὐτός indicates that
the referent is put at the center of the current visual field (or frame),
and a spotlight is on it at that moment.32 The cognitive convenience
and the pragmatic relevance of this communicative operation rest on
the overarching effect of singling out. Singling out may in turn imply
either isolating (keeping out, excluding) or centering (assigning a central
position and role within the surroundings).
Meanings such as ‘being separated from others’, ‘being the only one’,
‘being alone’, ‘being without anything else’ (naked or without arms or
dead) stem from the isolating effect. Let us consider, for example, the
following passage from the Catalogue of Ships:
(7) τῶν οἱ ἀδελφεὸς ἦρχε βοὴν ἀγαθὸς Μενέλαος
ἑξήκοντα νεῶν· ἀπάτερθε δὲ θωρήσσοντο·
ἐν δ’ αὐτὸς κίεν ᾗσι προθυμίῃσι πεποιθὼς
ὀτρύνων πόλεμον δέ· (. . .)
These were led by Agamemnon’s brother, even Menelaus, good at
the war-cry,
with sixty ships; and they were marshalled apart.
And himself he moved among them, confident in his zeal,
urging his men to battle.
(Il. 2.586–9; the narrator) (translation Murray, Loeb)
Menelaus is visually and thematically isolated. He is also a leader; quite
often αὐτός isolates an individual high or higher in rank (which is typi-
cal of intensifiers also in other languages).
The centering effect concerns meanings such as ‘being the pivot
character’ or ‘being at the center of the visual field’.
(8) Ἀλκίνοος δέ μιν οἶος ἐπεφράσατ’ ἠδ’ ἐνόησεν
ἥμενος ἄγχ’ αὐτοῦ, βαρὺ δὲ στενάχοντος ἄκουσεν.
Alcinous was the only one who understood and noticed this;
he was sitting next to him and heard him groaning heavily.
(Od. 8.94–6 (= 533–4); the narrator)

32
On the basic idea of Homeric epic as a movie running in the poet’s mind, cf.
Minchin (2001), Bakker (1997c) and Bonifazi (2008).
14 anna bonifazi

In the previous sentence Odysseus (grammatical subject) is said to weep

after Demodocus’ song (cf. 92–3). Here the hero remains the center of
the thematic and visual attention for Alcinous and for the recipients
of the performance as well. This example illustrates the basic cognitive
and pragmatic implications of αὐτός: by means of αὐτός the recipients
are invited to re-activate the mental representation of the referent even
though it is usually mentioned in the immediately preceding co-text,
because some conspicuous aspect of that referent is communicated.
The fact that the phrase in question syntactically requires a pronoun
does not conflict with the discourse relevance of the choice of αὐτοῦ.
In example (8) αὐτοῦ serves to mark topic continuity (the topic being
Odysseus) but it also gives the character visual conspicuity. In Homer,
putting a character in the middle of the visual field—that is, visual cen-
tering—is frequently conveyed by locative expressions including non-
nominative forms of αὐτός (cf. ἄγχ’ αὐτοῦ in the above example and ἐπ’
αὐτῷ below). The grammatical case in which αὐτός may occur does not
cancel or diminish its cognitive and pragmatic force. In other words,
non-nominative forms of αὐτός have the same cognitive and pragmatic
relevance as the nominative forms, the latter just reinforcing it.
A further effect of singling out concerns references to heroes’ corpses
or injured bodies, which play a significant role particularly in the Iliad.
(9) (. . .) ἀλλὰ φόβηθεν
πάντες, ἐπεὶ βασιλῆα ἴδον βεβλαμμένον ἦτορ
κείμενον ἐν νεκύων ἀγύρει· πολέες γὰρ ἐπ’ αὐτῷ
κάππεσον, εὖτ’ ἔριδα κρατερὴν ἐτάνυσσε Κρονίων.
(. . .) but [the Lycians] were scattered,
all of them, as they saw their king with a spear in his heart,
lying under the pile of dead men. Yes, because many others had
fallen
on his corpse, once Zeus has strained fast the powerful conflict
(Il. 16.659–62; the narrator)
Sarpedon’s corpse works as the center, while the pile of other dead
men work as the periphery. That corpse is put in the middle of the
visual field of the recipients of the text (as well as of the visual field of
the surrounding characters, that is, the Lycians themselves) and at the
same time it is singled out from the other corpses.33

33
Interestingly enough, intensifiers in different languages were originally expres-
sions for body parts (‘body’, ‘body parts’ or ‘soul’). Cf. König (1998: 10) and Puddu
(2005: 90–1).
 the case of ΚΕIΝΟΣ and ΑyΤΟΣ in homer 15

Centering may also imply the speaker’s empathy towards the refer-
ent of αὐτός. The latter sometimes turns out to be the pivotal figure
whose point of view is adopted. The narrator seems to have access to the
internal mental states of the individual αὐτός refers to. Many languages
encode these phenomena by means of the so-called indirect reflexives.
Linguists offer different explanations for reflexives violating syntactic
rules.34 Some scholars refer to them as ‘logophoric reflexives’. Logopho-
ricity—first explored during the seventies in African languages—deals
with a reflected or reported point of view.35 An instance of a logophoric
reflexive in English is the following:
(10) John was furious. The picture of himself in the museum had been
mutilated.36
I propose to consider αὐτός as a logophoric reflexive also in some
Homeric passages where the referent of αὐτός does not coincide with
the grammatical subject, but nevertheless it is the subject of conscious-
ness which the speaking ‘I’ has access to, empathically.
(11) ὣς δ’ αὔτως Μενέλαον ἔχε τρόμος· οὐδὲ γὰρ αὐτῷ
ὕπνος ἐπὶ βλεφάροισιν ἐφίζανε· μή τι πάθοιεν
Ἀργεῖοι, (. . .).
So, likewise, trembling seized Menelaus. Neither on himself, indeed,
slumber was descending on the eyelid. [He was afraid that] the
Argives might suffer some hurt (. . .) (Il. 10.25–7; the narrator)
After depicting Agamennnon’s sleepless night (cf. lines 3–4), the nar-
rator talks about Menelaus’ parallel non-sleep. By means of αὐτός he
not only takes Menelaus as the next ‘center’ (according to the ‘center/
periphery’ notion), but he also adopts Menelaus’ point of view and
starts conveying the hero’s feelings and thoughts (cf. μή τι πάθοιεν /
Ἀργεῖοι, 26–7).37

34
In Chomsky’s terms the antecedent has to lie inside the governing category of the
reflexive; in Reinhart and Reuland’s terms (1993), the reflexive mark and its antecedent
have to be arguments of the same predicate. These views account for direct reflexivity
involving basically grammatical subjects and direct objects, as in ‘The lawyer who was
young defended himself ’. Cf. Burkhardt (2002: 14). Chantraine (1953: 157) quotes a
few Homeric passages showing ‘ungrammatical’ reflexivity.
35
On the relationship between reflexives and logophoricity, cf. Hagège (1974), Sells
(1987), Culy (1997) and Burkhardt (2002).
36
Cf. Culy (1997: 846), after Pollard and Sag.
37
The narrator arguably does the same with Odysseus at Od. 5. 280 and at Od. 22.116,
with Achilles at Il. 17.407 and with the Achaeans at Il. 5.607. To quote another instance
outside of Homeric diction, see Xen. An. I 1.5 [Κῦρος] ἐπεμελεῖτο ὡς [οἱ βάρβαροι]
16 anna bonifazi

Indirect reflexives may have other functions as well. For example,

they may convey discourse prominence or else they may mark indirect
discourse. αὐτός may convey discourse prominence when its referent
becomes the main topic of the upcoming discourse;38 it may mark
indirect discourse when its referent represents the source of commu-
nication.39 The results of my analysis of Homeric αὐτός confirm that
these functions of indirect reflexives are all expressed.
A last cognitive and pragmatic aspect of Homeric αὐτός concerning
individuals is the possibility that the speaker expresses the recogni-
tion of someone’s full (or true) identity. As a result, αὐτός works as
a demonstrative pronoun that identifies who a certain character is.40
Odysseus and Diomedes, back from the expedition to the enemy camp
(Il. 10), are about to reach their companions. Nestor hears the sound
of galloping horses and hopes it is the two heroes coming back (‘May
these be really Odysseus and strong Diomedes / driving away from the
Trojans’, Il. 10.536–7). Immediately afterwards the narrator tells:
(12) οὔ πω πᾶν εἴρητο ἔπος ὅτ’ ἄρ’ ἤλυθον αὐτοί.
He had not yet spoken all his words, when they really did arrive,
[Odysseus and Diomedes] in person. (Il. 10.540; the narrator)

πολεμεῖν τε ἱκανοὶ εἴησαν καὶ εὐνοϊκῶς ἔχοιεν αὐτῷ ‘He took care that they should
be capable soldiers and should feel kindly toward himself’ [translation adapted from
Brownson, Loeb]. On anaphors expressing empathy in literary texts, cf. in particular
Daneš (1990) and Conte (1999: 75–81).
38
As an instance of αὐτός conveying discourse prominence I cite Od. 8.572–6
‘So come now tell me this and give me an accurate answer: / where you were driven
off your course, what countries peopled / by men you came to, the men themselves
(αὐτούς) and their strong-founded / cities, and which were savage and violent, and
without justice, / and which were hospitable and with a godly mind for strangers.’
(translation Lattimore). Not only is the speaker Alcinous keeping individuals distinct
from objects, but he also makes those individuals the prominent topic of discourse;
indeed, clauses at 575–6 focus on them as well. On ‘locally free reflexives’ marking
discourse prominence in English literature, cf. Baker (1995).
39
As an instance of αὐτός marking reflexivity in indirect discourse I cite Il. 3.87–91.
Hector is reporting Alexander’s plan to the Trojans and the Achaeans: ‘the word of
Alexandros (. . .) / he would have all the rest of the Trojans and all the Achaeans / lay
aside on the beautiful earth their splendid armour / while he himself (αὐτόν) and war-
like Menelaus / fight alone for the sake of Helen and all her possession’ [translation
Lattimore]. On indirect reflexives in indirect dicourse, cf. in particular Culy (1997).
40
This is what I argue about αὐτοῦ at Il. 9.193 (marking the fact that the ambassa-
dors face Achilles ipse), and, most of all, about several Odyssean occurrences of αὐτός
that more or less explicitly point to Odysseus’ true identity (cf. for example Od. 20.88;
21.107; 22.38; 23.55 and 24.321).
 the case of ΚΕIΝΟΣ and ΑyΤΟΣ in homer 17

7. Final remarks

I already mentioned the fact that in Homer it is not unusual to find

third person pronouns occurring relatively close to each other and
having the same referent. As a conclusive example, let us focus on a
striking instance of that.
(13) ἧσο παρ’ αὐτὸν ἰοῦσα, θεῶν δ’ ἀπόεικε κελεύθου,
μηδ’ ἔτι σοῖσι πόδεσσιν ὑποστρέψειας Ὄλυμπον,
ἀλλ’ αἰεὶ περὶ κεῖνον ὀΐζυε καί ἑ φύλασσε,
εἰς ὅ κέ Ø σ’ ἢ ἄλοχον ποιήσεται ἢ ὅ γε δούλην.
κεῖσε δ’ ἐγὼν οὐκ εἶμι· νεμεσσητὸν δέ κεν εἴη
κείνου πορσανέουσα λέχος· (. . .)
Go and sit beside him, abandon the way of the gods,
do not turn your feet back again to Olympus,
and, instead, express your suffering for that one forever, and look
after him,
until he will make you his wife, or he will make you his concubine.
Over there I am not going at all. It would arouse nemesis
to serve the bed of that one. (. . .) (Il. 3.406–11; Helen to Aphrodite)
Aphrodite has just saved Alexander from Menelaus’ hands; she is
addressing Helen in order to persuade her to join the charming man.
Helen’s reply to the goddess includes four different third person pro-
nouns—one of them occurring twice—within 5 lines, and all of them
have the same referent, that is Alexander. In this case the accessibility
degree of the referent is not relevant, as Alexander is the unquestion-
able center of attention of everybody at this moment of narration (of
Helen, of Aphrodite, of the narrator and of the audience). It is com-
monly acknowledged that the speech in question by Helen (cf. 399–412)
conveys a mixture of wrath, passion, bitterness and self-pity. In my view
the different third person pronouns have to be interpreted as belong-
ing to different discourse acts,41 and, consequently, as corresponding
to different communicative intentions. The communicative purpose of
αὐτόν at 406 is to express the centrality of the referent at the thematic,
visual and even psychological level. This is in line with Helen’s sar-
donic depiction of Aphrodite’s concern and passion toward the man:
the goddess is supposed to perceive Alexander as the central subject
around whom different actions may be accomplished (going to him,

41
For a definition of discourse act, see above, note 9.
18 anna bonifazi

suffering for him, taking care of him). (περὶ) κεῖνον at 408 belongs to
a discourse act expressing lament and veneration.42 Moreover, it echoes
κεῖνος uttered by Aphrodite at 391 (κεῖνος ὅ γ’ ἐν θαλάμῳ καὶ δινωτοῖσι
λέχεσσι), where it deictically points at Alexander’s appearance (‘Look at
him! There he is, in the room with the bed of circled patterns’) and to
κεῖνος at 411, with which Helen seems to be determined to keep herself
away from Alexander, both physically and emotionally (‘I am not going
to serve the bed of that one’). ἑ43 at 408 marks the immediate accessibil-
ity of Alexander within the same act of lament/veneration underlying
περὶ κεῖνον ὀΐζυε. Finally, if compared with the null anaphor of the first
half of line 409, ὁ in the second half involves a stronger activation of
the referent in the conciousness of the listener: the referent is always
Alexander, but a different discourse act is going to be performed.44 Far
from simply resuming the referent ‘Alexander’, ὁ introduces a contras-
tive idea: ‘he—the same one you respect and love—could even make
you his slave’. This quasi-paradox could be conveyed by emphasis in
the pronunciation of ὁ, as the particle γε may suggest.45
To sum up the points of this paper, the ‘referent in the mind’
model of anaphoric comprehension helps understanding better the
great variety of third person pronouns in Homeric language. Differ-
ent types of pronoun help monitoring the participants in the narrated
events according to their different accessibility or activation degree.
In this sense discourse cohesion matches cognitive convenience, that
is, the ways of recalling characters and objects are meant to facilitate
the retrieval of the referent in the discourse memory. However, the
choice of different pronouns—especially when they have the same

42
Several instances of κεῖνος in the first 4 books of the Odyssey—having Odysseus
as the referent—occur in semantic contexts of grief and sorrow about his presumed
ruin/death; such utterances are performed by Odysseus’ relatives or friends who love
and venerate him. Cf., for example, Od. 1.233; 2.351; 3.88; 4.109, 152, 182 and 832.
Further instances of κεῖνος expressing veneration, especially towards dead individu-
als, beyond Homer occur in Pind. Nem. 85–6; Soph. Aj. 437; Soph. OT 139; Eur. Alc.
866–7; Eur. Ion 1008.
43
ἑ is traditionally considered as a ‘regular’ anaphoric pronoun, whereas accented
ἕ is considered as a reflexive pronoun. Recently Puddu (2005: 150–61) has shown that
even the accented version in Homer does not convey per se any reflexivity.
44
As far as Latin is concerned, Bolkestein and Van de Grift (1994: 298) point out
a similar phenomenon in terms of discourse discontinuity: ‘As an overall generaliza-
tion we may state that pronouns are used rather than 0 at points where the discourse
develops in what might be called a “non-default” way: the introduction of Future Topic,
a switch in perspective, a jump towards a different level of the discourse’.
45
On γε as a demarcating scope particle, cf. Wakker (1994: 308).
 the case of ΚΕIΝΟΣ and ΑyΤΟΣ in homer 19

referent—may depend on specific communicative intentions as well.

Different discourse acts may involve different anaphoric expressions to
convey additional meanings. This is a pragmatic convenience (passages
like the one concerning Helen’s words on Alexander can hardly be
explained otherwise). Homeric third person pronouns arguably mark
discourse cohesion far beyond syntactic and metrical needs. This holds
at least for αὐτός and κεῖνος. Their use match cognitive convenience in
anaphora processing and mark the pragmatic relevance of the subjects
involved at a certain point of the narration.
 CHAPTER TWO

PRAGMATIC PRESUPPOSITION AND COMPLEMENTATION

IN CLASSICAL GREEK

Luuk Huitink

1. Introduction1

Many cognitive verbs in Classical Greek can be complemented by both

participle phrases and finite ὅτι-clauses,2 apparently without any differ-
ence in meaning. In this paper I will argue that there is in fact a differ-
ence between the two constructions, which concerns their information
structure. In analyzing the difference I will make use of the concept of
presupposition. In my opinion there is considerably more to be said
about this difficult notion than has been customary in Greek linguistics,
so that my explanation of it in section 2 will be fairly elaborate. In sec-
tion 3 I will analyse my data. As the number of instances I looked at is
large, this analysis will be of a provisional nature. The verbs selected for
the investigation are the cognitive verbs οἶδα, γιγνώσκω and ἐπίσταμαι;
my text sample consists of a number of Classical prose texts,3 but I will
occasionally refer to other texts as well.

2. Semantic and pragmatic presupposition

The grammars have long recognized that after cognitive verbs there is
a semantic opposition between complementary infinitives on the one
hand and participles and ὅτι(ὡς)-clauses on the other: whereas the

1
The research for this paper was made possible by a Postgraduate Award from the
AHRC and a Martin Senior Scholarship from Worcester College, Oxford. I would like
to thank Dr. S. Scullion, Prof. A. Willi (Oxford) and the editors of this volume for their
valuable remarks. Any shortcomings are due to my own stubbornness.
2
Ὡς-clauses after such verbs are much rarer than ὅτι-clauses and presumably not
synonymous with them. For data and some considerations, see section 3 below.
3
Andocides; Antiphon; Herodotus; Lysias (except 11 & 35); Plato, Euthyphro, Apol-
ogy, Euthydemus, Protagoras and Gorgias; Xenophon, Anabasis and Memorabilia.
22 luuk huitink

former express an allegation (or possible fact), the latter express facts.4
This semantic difference may be illustrated by the following examples
with ἐπίσταμαι, which may be complemented by all three types of
complement:
(1) τὴν δὲ Σάμον ἐπιστέατο δόξῃ καὶ Ἡρακλέας στήλας ἴσον ἀπέχειν.
[The Greeks] supposed too that Samos was as far away from them as
the Pillars of Heracles. (Hdt. 8.132.3)

(2) ὃν ὑμεῖς ἐπίστασθε ὑμᾶς προδόντα.

Of whom [Dexippus] you know that he betrayed you. (X. An.
6.6.17)

(3) ἐπίστασθε δέ, ὦ ἄνδρες δικασταί, ὅτι παράδειγμα τοῖς ἄλλοις ἔσται.
But you must understand, members of the jury, that it will be an
example for the others. (Lys. 30.24)
In (1), the narrator of the Histories does not share the opinion of the
Greeks; note the addition of δόξῃ ‘in their opinion’. We are dealing
with a mere opinion, and ἐπίσταμαι may be translated by ‘suppose’. By
contrast, the speakers in both (2) and (3) do vouch for the truth of the
propositions expressed in the complement (as is clear from the context),
and ἐπίσταμαι may be rendered by ‘know/understand’.5
The distinct translations ‘know/understand’ and ‘suppose’ correspond
to a distinction between factive and non-factive verbs in English, as
presented by Kiparsky & Kiparsky (1970). According to these schol-
ars, there exists in English a class of factive verbs, like ‘know’, ‘realize’,
‘understand’ and ‘regret’, which belong to those expressions in the
language which trigger presuppositions: it is said that they presup-
pose the truth of their complement.6 As such, they may be contrasted
with non-factive verbs like ‘believe’, ‘suppose’ and ‘maintain’, which

4
Kühner-Gerth (1898/1904: 2.357): ‘Nach den Verben des Glaubens, Denkens (. . .)
folgt in der Regel der Infinitiv (. . .); nach den Verben des Wissens, Erkennens (. . .) folgt
entweder ὅτι oder ὡς oder das Partizip.’ Cf. also Schwyzer & Debrunner (1950: 395): the
participle expresses ‘Tatsächliches’, the infinitive ‘Vermutetes, Gedachtes, Mögliches’.
5
The semantic oppositions between the several complements illustrated in examples
(1)–(3) are valid in most contexts. Exceptions concerning the participle and ὅτι-clauses
occur in a limited set of contexts and can, I believe, be handled by regarding them
as instances of presupposition-cancellation. I hope to deal with this phenomenon at
greater length elsewhere.
6
For a general introduction to presupposition, cf. Levinson (1983: 167–225). Among
many other expressions, the set of presupposition-triggers includes the definite article
(‘The king is ill’ presupposes that there is in fact an individual to whom ‘the king’
 pragmatic presupposition and complementation 23

trigger no such presupposition. The following utterances illustrate the

difference:
(4) John believed that Mary was at Trinity College, but in fact she was
at Worcester.

(5) John regretted that Mary was at Trinity College, * but in fact she was
at Worcester.
John’s belief in (4) can be contradicted by the speaker, but this is not
the case with John’s regret in (5). This is because ‘regret’ is a factive
verb, which gives rise to the presupposition that the complement is
true: a subsequent denial of that truth would result in a contradiction
on the part of the speaker. A similar contradiction could arise in (1), if
ἐπίσταμαι were rendered by ‘know’ or ‘understand’ instead of ‘suppose’.
Struck by the similarities between the semantic properties of English
factive/non-factive verbs and those of Greek complement clauses, many
scholars have argued that in Greek factive presupposition resides in
complement clauses, not in matrix verbs: after cognitive verbs, parti-
ciples and ὅτι(ὡς)-clauses trigger a presupposition, while infinitives do
not.7 In this way they recast in presuppositional terms a semantic dis-
tinction which, as we have seen, was long recognized by the grammars.
In Greek linguistics, then, factive presupposition is regarded as a
semantic notion, which pertains to a speaker’s commitment to the truth
of the proposition expressed in a complement clause. Although such an
account of the phenomenon along semantic lines is not lacking from
general linguistics, many linguists have felt that it does not adequately
explain the nature of presuppositions or how they originate. Therefore,
an alternative account has been developed, in which presupposition is
regarded as a pragmatic notion, which does not pertain to truth condi-
tions but to information structure.8 Central to the pragmatic account is
the idea that not every part of an utterance can be considered equally

refers) and a conjunction like ‘before’ (‘Before he had dinner, John washed his hands’
presupposes that John in fact had dinner).
7
Cf. Lightfoot (1975: 41–3), who introduced the concept to Greek linguistics; he
was followed by e.g. de Boel (1980); Rijksbaron (1986; 20073); Cristofaro (1996: 154);
Basset (1999); Ruijgh (1999: 216–7).
8
Cf. especially Lambrecht (1994: 51–73). Levinson (1983: 167–225) discusses both
semantic and pragmatic explanations of the phenomenon, but prefers the latter kind.
Pragmatic explanations have been put forward especially by Stalnaker (1974; 2002);
Karttunnen (1974); Lewis (1979). Kiparsky & Kiparsky (1970) also favour a pragmatic
explanation; the way their seminal paper has been used in Greek linguistics reveals a
misunderstanding of their original purposes. See also below, notes 11 and 16.
24 luuk huitink

informative. In fact, successful communication crucially depends on

a speaker’s ability to estimate the amount of relevant knowledge the
addressee already possesses and to increase the addressee’s knowledge
by gradually adding new information to the pool of already existing
knowledge. The process of adding new information to old knowledge is
to be taken quite literally: many utterances can be neatly divided into an
‘old’ and a ‘new’ part. Consider for example the following utterance:
(6) I finally met the woman who moved in downstairs.
(= (2.11) in Lambrecht (1994: 51))
This utterance is most naturally interpreted as an assertion that the
speaker ‘finally met someone’. Only this part of the utterance is new
and truly informative. By contrast, the material in the restrictive rela-
tive clause does not have a high information value: it is pragmatically
presupposed that the addressee already knows that a certain woman
moved in downstairs from the speaker. That fact already belongs to
the pool of shared knowledge between the speaker and addressee, the
so-called common ground. It is merely mentioned in order to enable
the addressee to identify the individual whom the speaker finally met.
In other words, (6) cannot normally be uttered in order to inform an
addressee that a woman has moved in downstairs.
The presupposition in (6) is evoked by the grammatical structure
of the sentence. Indeed, presuppositions are conventionally associated
with certain lexemes or grammatical constructions, like the restrictive
relative clause in (6) or, indeed, factive verbs and their that-comple-
ments. Apparently, factive verbs are conventionally used to describe
someone’s emotional response to, or cognitive perception of, a piece of
already established information. Thus, the first clause in (5) can only be
felicitously uttered if it is already known to speaker and addressee that
Mary went to Trinity College, most naturally because this information
is ‘old’ and has been introduced in an earlier part of the conversation.
If this is the case, the presupposition functions in much the same way
as an anaphora.9 The fact that the continuation ‘but in fact she was at
Worcester’ leads to incongruity is due to the fact that one cannot deny
what has not been asserted.

9
Van der Sandt (1992) offers an influential theory of presupposition based on this
similarity. See below for other possible ways in which presuppositions may arise.
 pragmatic presupposition and complementation 25

So, when speaking about presupposition, pragmaticists do not believe

that factuality or certainty on the part of the speaker is at issue; rather,
they emphasize the pre in presupposition: a presupposed proposition
is a proposition which the speaker assumes to be already part of the
common ground. The relevant opposition is that between pragmati-
cally presupposed and asserted information,10 not that between facts
and possible facts.
To further elaborate the contrast between the semantic and prag-
matic account of presupposition, we may briefly consider the semantic
explanation offered by Rijksbaron in his textbook on the Greek verb;
it involves the factive verb ‘see’:
(7) Someone who utters the sentence ‘John saw that the house was on
fire’ does not only tell us something about a certain perception on
John’s part, but also presents ‘the house was on fire’ as an independent
fact. The same information could be conveyed by two independent
statements: ‘The house was on fire and John saw that’. This property
of the verb ‘see’ is even more clearly illustrated when the main verb
is negated: ‘John did not see that the house was on fire’. Here, ‘the
house was on fire’ is still presented as an independent fact. (Rijksbaron
20073: 50; original italics)
Rijksbaron’s ‘negation-test’ is a standard feature of accounts of presup-
position. According to the pragmatic explanation, presuppositions owe
their survival under negation to the fact that they are not part of the
assertion of an utterance, so that the negation has no scope over them.
Rijksbaron’s ‘paraphrase-test’, by contrast, has no place in a pragmatic
account.11 For in such an account it would be emphasized that the two
simple sentences ‘The house was on fire and John saw that’ do not in
fact convey the same information as the complex sentence ‘John saw
that the house was on fire’. This is because the latter can only be felici-
tously uttered if it is already established that the house was on fire; it
cannot be used to inform (‘tell’) an addressee of that fact, as the simple
sentence ‘The house was on fire’ can.

10
Other terms which are often used to make more or less the same distinction are
‘old’ and ‘new’, ‘topic’ and ‘focus’, and ‘theme’ and ‘rheme’. See Lambrecht (1994) for
an elaborate discussion of the intricate relationships between these notions. For my
present purposes it is unnecessary to distinguish between them.
11
The test is Rijksbaron’s own. Introducing it, he explicitly stated that this was ‘a
property not mentioned by the Kiparskys’ (1986: 178). However, the Kiparskys had
good reasons for not doing so.
26 luuk huitink

Problems for the pragmatic account of presupposition arise when it

appears that in many contexts factive verbs are used by speakers without
there being a reasonable assumption that the proposition expressed in
the complement clause already belongs to the common ground at the
time of utterance. Consider the following examples:
(8) faq-page on the web for freshers at Newcastle University:
I know that Newcastle has a great reputation for its nightlife but how
cheap is it to go out?12

(9) Sign on the door of a library:

We regret that the library is closed.
(8) presents the first ‘utterance’ of a (virtual) dialogue between a
speaker and an addressee who presumably do not know each other at
all. This raises the question why the speaker may felicitously present
the proposition ‘Newcastle has a great reputation for its nightlife’ as
belonging to the common ground by subordinating it to the factive verb
‘know’. Presumably, what happens here is that the speaker establishes
(rather than merely reflects) a common ground between himself and
his addressee. He conveys that, given his familiarity with the reputation
of Newcastle’s nightlife, the addressee may direct the ‘conversation’ to
more detailed and informative matters; the reputation of Newcastle’s
nightlife itself does not need to be commented on: it is presented as
something both participants in the dialogue will probably agree on. So,
even though there was no prior presupposition in (8), when the speaker
utters the sentence, the presupposition simply springs into existence.
In technical terms, the presupposition is accommodated in the context
of the conversation.13
Example (9) shows how presuppositions may be rhetorically exploited.
Instead of a straightforward ‘The library is closed’, the addressee is
confronted with an ‘utterance’ which precludes a discussion with the
librarians about limited opening hours: the fact that the library is
closed is presented as presupposed and therefore as information which
is simply to be taken for granted. Examples (8) and (9) illustrate how
the pragmatic theory turns a weakness (not all presuppositions can be
easily explained as common ground) into a strength (presuppositions
may serve rhetorical ends). Although the that-complements in these

12
Source: http://www.talesofnewcastle.net/universitylife/faqs. The example was found
through a Google-search with the terms ‘I know that’.
13
Cf. Lewis (1979) for this concept.
 pragmatic presupposition and complementation 27

examples do not express old information, they do express information

which is not asserted because it is meant to be taken for granted.
This is not the case with the that-complement of the factive verb
‘discover’ in the following little dialogue, which poses considerably
more difficult problems:
(10) A: ‘Where was Harriet yesterday?’
B: ‘Henry discovered that she had a job interview at Princeton.’
(= (1) in Simons (2007))
Here, the that-clause seems to contain the main point of the utterance,
as it, and not the matrix clause, answers the addressee’s question. In
other words, the that-complement counts as asserted. According to
Simons (2007), in such cases the pragmatic presupposition does not
rise at all; the factive quality of ‘discover’ is simply suspended. Surpris-
ingly, it would seem that Rijksbaron’s ‘paraphrase-test’ works quite
well here, the sentence conveying the same information as the two
separate assertions ‘Harriet has a job interview at Princeton and Henry
discovered that’.
It is at present not quite clear under what conditions it is possible to
felicitously assert propositions dependent on a factive verb. What does
seem clear, however, is that some factive verbs can be more easily so
used than others. Evaluative factives like ‘regret’, for instance, are often
regarded as stronger triggers of pragmatic presupposition than cognitive
verbs like ‘know’, ‘understand’ and ‘realize’. Hooper (1975) has in fact
argued for a division of the category of factive verbs into ‘true factives’
and ‘semi-factives’. The latter subcategory is meant to include the cog-
nitive factives and is set apart from the ‘true factives’ by its members’
ability to be complemented by asserted propositions. Lambrecht (1994)
has emphasized the role played by intonation in determining whether a
proposition is pragmatically presupposed, arguing that even a sentence
like (9) and many other classic examples could receive a non-presup-
positional reading if intoned in a certain way. And in a recent article,
Spenader (2003) has shown that in a large corpus of spoken English,
almost 65% of complement clauses dependent on cognitive factive verbs
contain material with a high information value.14

14
As Spenader (2003: 351) herself points out, part of the explanation for this
surprising figure is that truly pragmatically presupposed propositions often appear
not as complement clauses but as anaphoric pronouns: ‘The house was on fire. John
regretted that’ is more common than ‘The house was on fire. John regretted that the
28 luuk huitink

Does all this mean that we are back at square one and that the prag-
matic explanation of presupposition is not adequate? I do not think we
have to go that far. I believe that both the semantic and the pragmatic
account capture something essential about complement clauses: both
semantic presupposition (or speaker’s commitment) and pragmatic
presupposition are relevant parameters.15 Going further, I believe that
we need both in order to be able to account for the distribution of
complement clauses in Greek: I claim that after cognitive verbs par-
ticiple phrases present pragmatically presupposed information, while
ὅτι-clauses present semantically presupposed (that is asserted) informa-
tion. Thus, the participle phrase in (2) is pragmatically presupposed, the
fact of Dexippus’ betrayal being one of the main underlying themes of
the speech from which (2) is taken. The ὅτι-clause in (3), by contrast,
contains new information which is not pragmatically presupposed but
only semantically, the speaker being committed to its truth and urging
his addressees to accept it as well. The infinitive phrase in (1), finally, is
neither pragmatically nor semantically presupposed. In the next section
I will substantiate this claim.

3. Participles and ὅτι-clauses: a pragmatic distinction

3.1. General arguments

Before discussing individual instances, I will first advance a number
of general arguments from typology and specific usages, in order to
place my claim on a secure theoretical footing. As our starting point
we may take the only earlier attempt to find a pragmatic distinction
between complementary participles and finite complement clauses. It
is here quoted in full:
(11) En effet, on y suggère que, plus la complétive est syntaxiquement
intégrée à la principale, cas de la construction participiale, plus elle

house was on fire’. See, however, examples (17) and (18) below for Greek examples
of the latter kind.
15
Cf. Lambrecht (1994: 61–4), whose terminological distinction between semantic
and pragmatic presupposition I follow. The question whether the two kinds of presup-
position should theoretically be reduced to one may be left unanswered here. I refer
to the literature cited in note 8 above.
 pragmatic presupposition and complementation 29

a tendance à porter le poids rhématique de l’énoncé. Cela peut aller

jusqu’à réduire le verbe principal au rôle d’un simple auxiliaire de
modalisation (. . .). En revanche, les complétives conjonctives, moins
intégrées, permettent de mieux détacher le prédicat principal et de
faire porter sur lui l’accent. (Basset 1999: 38)
Basset’s analysis can be readily understood in terms of pragmatic
presupposition.16 He seems to claim that participles express asserted
information, while ὅτι-clauses present presupposed information. The
following arguments substantiate my claim that the reverse is true.
To begin with, one may note that Basset’s main syntactic argument—
the less syntactically integrated a complement is, the more it permits
the matrix verb to ‘stand apart’ and carry focus—can be turned on its
head: the greater syntactic independency of finite complements may
as well permit the complement itself rather than the matrix verb to be
stressed. That this is in fact the better explanation has been shown for
a number of languages by Givón (1980), who establishes the following
language-independent rule for complements:
(12) The less a clause/proposition is independent in its expression of
asserted information, the less it is going to resemble an independent
clause and the more it is going to exhibit syntactic/structural inte-
gration into the main clause. (Givón 1980: 371; original italics).
This predicts that ὅτι-clauses, which have a finite verb and therefore
resemble independent clauses, are more likely to express asserted
information than complementary participles, which are much more
syntactically integrated into their matrix clause.17
Secondly, we may consider Basset’s claim that in the case of a comple-
mentary participle the matrix verb becomes something of a parenthetical

16
The fact that Basset assumes without any further explanation a pragmatic distinc-
tion after calling attention to the factivity of the complements involved, shows that
he is not fully aware of the pragmatic account of presupposition. Indeed, he explains
presupposition in purely semantic terms as a ‘contenu de vérité’ and ‘une modalité
de certitude’ (1999: 35); this differs in terminology but not in substance from his
earlier work in which he states that ‘le participe permet de définir un fait’ (1988: 127;
original italics).
17
Similar observations on the correlation between degrees of syntactic integration
and assertibility have been used by Noël (1997) and Jary (2008) to account in explicit
information-structural terms for the difference between that-clauses and to + infini-
tive phrases after the English verb ‘believe’. Buijs (2005) reaches similar conclusions
regarding the information structure of adverbial participles and finite subordinate
clauses in Classical Greek prose.
30 luuk huitink

addition to the sentence. Actual Greek usage suggests otherwise. I am

referring to the parenthetical use of οἶδ’ ὅτι:18
(13) τίς ἂν ἡμᾶς τοιούτους ὄντας ἐπαινέσειεν; ἡμεῖς μὲν γὰρ οἶδ’ ὅτι
πονηροὺς ἂν φαίημεν εἶναι τοὺς τὰ τοιαῦτα ποιοῦντας.
Who would praise us if we were like that? For we ourselves, I’m
sure, would say that people who perform such deeds are scoundrels.
(X. An. 5.7.33)
With the second sentence the speaker Xenophon explains (γάρ) why
nobody would praise the Greek army if they behaved brutally, thus mak-
ing explicit the implied answer to the rhetorical question. Importantly,
the explanation revolves around the low opinion the Greeks themselves
would hold about such men, not about the speaker’s knowledge of that
opinion; the main assertion is to be found in the ‘complement’ clause,
not in the main clause. Furthermore, if οἶδ’ ὅτι had the status of a full
main clause, we would perhaps expect to find the accusative ἡμᾶς as
the proleptic object of οἶδα, not the nominative ἡμεῖς. As the sentence
stands, οἶδ’ ὅτι seems to have been inserted into an independent sen-
tence as a parenthetical ‘tag’ which almost fulfils the role of a modal
particle and can hardly carry focus. It is hard to see how this parentheti-
cal usage of οἶδ’ ὅτι could have come about if οἶδα would be the main
assertion in utterances in which it is complemented by ὅτι.
A third argument against Basset’s claim is that it is incompatible with
the most common analysis of prolepsis, as developed in Panhuis (1984)
and Chanet (1988). They show that a proleptic constituent, which usu-
ally functions as the object of the main verb, while it is also the subject
of the finite complement clause, has the pragmatic status of a theme
(or topic), while the complement clause acts as the rheme (or focus)
of the sentence. An example from my own sample is:
(14) If you just start worshipping the gods,
γνώσῃ τὸ θεῖον ὅτι τοσοῦτον καὶ τοιοῦτόν ἐστιν, ὥσθ’ κτλ.
predicate [theme] [comp.] [rheme]19
Then you will find out that the greatness and nature of the divine
is such that (. . .) (X. Mem. 1.4.18)

18
Also (εὖ) ἴσθ’ ὅτι and δῆλον ὅτι (δηλονότι); cf. Kühner-Gerth (1898/1904: 2.368);
most of their examples come from Demosthenes. In my text sample, the construction
seems to be in its early stages, the only other convincing examples being Pl. Euthphr.
6c6–7 and Euthd. 272d2–3.
19
Analysis according to the general schema proposed by Chanet (1988: 76).
 pragmatic presupposition and complementation 31

Τὸ θεῖον has been the discourse topic of the entire conversation to

which (14) forms the conclusion. The ὅτι-clause contains Socrates’ final
assertion about the divine. As the prolepsis seems to originate precisely
from the difference in information value between the discourse topic
τὸ θεῖον and the ὅτι-clause, this is a strong argument in favour of my
claim that ὅτι-clauses are a locus of asserted information.
Fourthly, I would like to draw attention to a specific context in which
only participles occur. The context I mean is that of cognitive factives
which occur in restrictive relative clauses. As we have seen in the
discussion of example (6) above, such clauses cannot contain asserted
information which the speaker does not assume to be already known
or at least identifiable to the addressee. An example is:
(15) ἐὰν δέ τις ὃν ἂν γνῷ καλόν τε κἀγαθὸν ἐραστὴν ὄντα, τοῦτον φίλον
ἑαυτῷ ποιῆται, σώφρονα νομίζομεν.
But when someone makes a friend of someone whom he knows
to be a good and honourable lover, we regard him as prudent.
(X. Mem. 1.6.13)
The restrictive relative clause determines the referent of τοῦτον and
does not contain asserted information. The presupposition triggered by
γνῷ (. . .) ὄντα is that there exists a class of καλοί τε κἀγαθοὶ ἐρασταί
and that the addressee knows this and can identify it.20
Finally, something may be said about the data. Table 1 presents
the figures concerning the complementation of οἶδα, γιγνώσκω and
ἐπίσταμαι in my text sample alongside those for λέγω/εἶπον in the
same sample.21

Table 1: The complementation of οἶδα/γιγνώσκω/ἐπίσταμαι and of

λέγω/εἶπον
Participle ὡς ὅτι Totals
οἶδα/γιγνώσκω/ἐπίσταμαι 30.4% (138) 7.9% (36 [22]) 61.7% (280) 100% (454)
λέγω/εἶπον 0% (0) 35.7% (161[49]) 64.3% (290) 100% (451)

20
Cf. [And.] 4.29; Hdt. 1.5.3, 110.1; 9.94.2; Lys. 12.33; 14.37; X. An.1.9.20; Mem.
1.6.13 (bis); 4.6.9.
21
Infinitives were left out. Absolute numbers are given in brackets. In the case of ὡς,
the numbers for Herodotus are given in square brackets, as there is reason to assume
that in the Ionic dialect ὡς fulfills at least partly the role played by ὅτι in Attic and
therefore occurs much more often; cf. Monteil (1963: 399); Willi (2002b: 115–6).
32 luuk huitink

A first observation concerning these data is that the rate of occurrence

of ὅτι-complements and therefore of asserted information after cognitive
verbs does not differ much from the findings of Spenader (2003) for
spoken English (see above, section 2), which suggests that the cognitive
verbs are not particularly strongly associated with pragmatic presup-
position in Greek either. This interpretation of the data is corroborated
by the fact that a verb which may be expected to be more strongly
associated with pragmatic presupposition, namely ὁρῶ/εἶδον ‘see’,22 is
much more often complemented by a participle than by a ὅτι-clause.
A count of all instances in Xenophon’s Anabasis reveals that the ratio
participle to ὅτι-clause is close to 28:1 in the case of this verb.23
Secondly, while the rate of occurrence of ὅτι-clauses does not differ
much between the cognitive verbs and those of speaking, it seems that
there is a correlation between a frequent rate of occurrence of parti-
ciples and a limited rate of occurrence of ὡς-clauses. If ὡς-clauses and
participles by and large have the same pragmatic status, this correlation
would be easily explained, at least from a synchronic point of view.
We could then say that the possibility of complementing a verb with a
participle leads to a significant reduction of the number of ὡς-clauses
found after such verbs.24 Now, in her work on Greek complementation,
Cristofaro (1996; 1998) has offered the following explanation for the
difference between ὡς and ὅτι: ‘ὅτι introduces foreground information
with high communicative value, ὡς background information with low
communicative value.’ (Cristofaro 1998: 74). This assessment of the

22
Hooper (1975) classified English ‘see’ as a ‘semi-factive’ predicate (see section
2 above), mainly because it can occur in parenthetical clauses like other semi-factive
predicates (‘I see’, ‘I know’). However, Greek knows no such parenthetical usage of
ὁρῶ/εἶδον, and the verb can accordingly be regarded as more truly factive.
23
Participle.: 85x; ὅτι: 3x, ὡς: 2x. These figures present a slightly distorted picture,
because when ὁρῶ/εἶδον is used of direct perception, the participle is the only pos-
sible complement, so that there is no competition with ὅτι. Unfortunately, it is often
impossible to draw a clear distinction between direct and indirect perception (cf. Dik
& Hengeveld 1991), so that the figures are difficult to adjust. However, cases of direct
perception clearly form a minority, so even if we take this complication into account,
the participle occurs in an overwhelming majority of cases.
24
Chantraine (1953: 2.326–7) suggests that the complementary participle originated
from a reinterpretation of circumstantial participles after verbs denoting direct percep-
tion; cf.: ὁρῶ/πυνθάνομαι [αὐτὸν] [ποιοῦντα ταῦτα]: ‘I see/perceive him, while he is
doing this’> ὁρῶ/πυνθάνομαι [αὐτὸν ποιοῦντα ταῦτα] ‘I see/perceive that he is doing
this’. It seems reasonable to suppose that when verbs like πυνθάνομαι took on a more
abstract cognitive meaning, the participle could also refer to indirectly or cognitively
perceived propositions, and hence spread analogously to other cognitive verbs, but not
to verba dicendi, where ὡς already fulfilled a similar role.
 pragmatic presupposition and complementation 33

pragmatic value of ὡς-complements closely resembles my claim for

the participle (about whose pragmatic status Cristofaro has nothing to
say), while ὅτι is described in terms similar to mine.25 The pragmatic
status of ὡς-complements can be seen most clearly in instances like
the following:
(16) σχεδὸν οἶμαι ὑμᾶς ἐπίστασθαι ὡς πολλὰ καὶ δεινὰ μετὰ ταῦτα τῇ
πόλει ἐγένετο.
I think you know well enough how many miserable things befell
the city afterwards. (Lys. 13.43)
In this and most other instances with ὡς from my sample, the comple-
mentizer can best be interpreted as ‘how’, which is itself a presupposi-
tion-trigger, in that it focuses on the extent to which the city was beset
by dangers, while the fact that the city was in danger is itself taken for
granted (cf. Cristofaro (1998: 72)).

3.2. Presupposed, accommodated and exploited propositions

Having advanced a number of general arguments in favour of my claim,
I will now discuss a number of concrete examples. I will start with some
clear-cut examples and then proceed to an analysis of some passages in
which presupposition accommodation and exploitation play a role.

25
In fact, Cristofaro (1996: 70–5; 1998) elaborates the general claim cited above
by drawing up an intricate table in which the occurrence of ὅτι or ὡς is linked to
the three parameters [+/-focus], [+/-topic] and [+/-known information]. She argues
that only ὅτι is found in the configuration [+focus] [-topic] [-known], while only ὡς
is found in the configuration [-focus] [+topic] [+known]. In all other configurations
both complementizers are found. I admit that I have difficulty understanding the
table, all the more so as the parameters receive only the briefest of definitions (1996:
36–7) and because only two examples are offered; it is unclear to me, for instance,
what concrete example would satisfy the configuration [+focus] [+topic] [-known],
as topic and focus are most naturally seen as mutually exclusive, while there is also a
strong tendency for topical information to be known (or at least identifiable, which
would perhaps have been a better parameter). It is therefore unclear to what extent
her divisions correspond to my notion of pragmatic presupposition. I should like to
add that in Cristofaro (1996) the category of ‘factive verbs’ (which do not include the
cognitive factives, which rather arbitrarily constitute their own category) are shown
to be complemented by ὅτι on most occasions, not by ὡς (or the participle). Given
her assessment of the differences between ὅτι and ὡς this surely needs an explanation,
but it does not receive one, presumably because Cristofaro thinks of factivity in purely
semantic terms. So, while I agree with Cristofaro’s basic characterization of the differ-
ence, I believe more research is necessary.
34 luuk huitink

The following two passages provide very clear examples of pragmati-

cally presupposed participle phrases:
(17) The Greeks are impressed by the large Persian naval forces at
Artemisium and want to withdraw, leaving the Euboeans to them-
selves:
καταρρωδήσαντες δρησμὸν ἐβούλευον ἀπὸ τοῦ Ἀρτεμισίου ἔσω τὴν
Ἑλλάδα· γνόντες δέ σφεας οἱ Εὐβοέες ταῦτα βουλευομένους ἐδέοντο
Εὐρυβιάδεω προσμεῖναι χρόνον ὀλίγον.
They became afraid and planned to flee from Artemisium to Greece.
When the Euboeans realized they were planning this, they asked
Eurybiades to stay for a little longer. (Hdt. 8.4.1–2)

(18) καὶ ὃς ἐθαύμασεν· οὕτως ἔτι νέος τε καὶ εὐήθης ἐστί.—Κἀγὼ γνοὺς
αὐτὸν θαυμάζοντα (. . .).
And he was amazed. He is still so young and naive. And when I
realized that he was amazed (. . .). (Pl. Euthd. 279d7–8)
The complementary participle phrases are licensed by the fact that the
information they contain has already been asserted in the (immediately)
preceding context. The γνόντες/γνούς-clauses increase the addressees’
knowledge only by telling them that certain individuals cognitively
perceived the presupposed information. Observe that in both cases the
participle phrases are rather short, repeat the verb used in the preceding
context and contain anaphoric material which is bound by the preceding
context (σφεας and ταῦτα in (17), and αὐτόν in (18)). In other words,
the participle phrases have very little descriptive content, which is a
general property of pragmatically presupposed propositions.26 Most
other equally clear examples also come from Herodotus and Plato,
whose oral style favours such forms of repetition which clearly mark
the boundaries of successive stages of the narrative.27
To be sure, it is not a necessary condition that participle phrases
repeat material verbatim from the assertion they refer back to. Often,
too, there is rather more space between the first time information is

26
Cf. van der Sandt (1992). Both Scott Scullion and the editors suggest to me that
the fact that the participle in (17) is in the middle, while the main verb it refers back
to is active, may imply that the Euboeans think the other Greeks are acting in their
own interest.
27
Cf. Hdt. 1.140.2; 9.33.5, 9.108.1; Pl. Euthd. 275d5–7, 276d2–4; X. Mem. 4.2.40. Den-
niston (1952: 92–6) finds such ‘repetition for the sake of clearness’ typical of Herodotus
and Plato and finds only very few examples in Xenophon or the orators.
 pragmatic presupposition and complementation 35

asserted and the moment it is referred back to by a presupposed par-

ticiple phrase.28 In the following example, five OCT-pages separate the
assertion of the proposition that ‘Cyrus died’ from the first time it is
referred to by a pragmatically presupposed participle. Another three
pages later, it is referred to again in the same way. This example shows
that it is sometimes possible to ‘track’ a presupposed proposition over
longer stretches of text:
(19) Κῦρος δὲ αὐτός τε ἀπέθανε (. . .) ὡς δ’ ᾔσθετο Κῦρον πεπτωκότα
(. . .) οὐ γὰρ ᾔδεσαν αὐτὸν τεθνηκότα.
Cyrus himself died (. . .). When he perceived that Cyrus had fallen
(. . .). For they did not know that he was dead. (X. An. 1.8.27, 9.31,
10.16)
We first hear about Cyrus’ death from the perspective of an omniscient
narrator before most of the characters within the story find out. It is
precisely such a narrator which enables pragmatically presupposed
participles like those in (19) to be used. However, the narrator of the
Anabasis does not present himself as omniscient throughout his work.
Recently, Rood (2007: 148) has identified a peculiar narrative strategy
in the Anabasis, by which the narrator sometimes holds back vital
information until it becomes available to the Greek army or even the
character Xenophon. This gives the impression of immediacy and eye-
witness report: the reader experiences with Xenophon the vicissitudes of
the army trying to make its way back to Greece. A good example of this
strategy is the narrator’s (professed) ignorance of the whereabouts of
Artaxerxes after the battle of Cunaxa, which is in marked contrast with
his treatment of Cyrus’ death in the same part of the story. Artaxerxes
is last seen leaving a stronghold (1.10.13), but then vanishes from the
scene. The intelligence the Greeks receive does not reveal his where-
abouts. The resulting anxiety is as much that of the Greek soldiers as
that of the reader, because the narrator does not tell us more than the
Greek troops could have known at the time. Towards the end of the
Greek army’s first day of marching, we read the following:

28
Cf. e.g. And. 1.14 (cf. 12); [And.] 4.6 (cf. 3); Hdt.1.971.1 (cf. 96.3); Lys. 6.49 (cf.
46, where the definite noun phrase τὴν συμφοράν shows that the disaster of the Pelo-
ponnesian War is regarded as common knowledge); 17.8 (cf. 5); 23.12 (cf. ibidem); Pl.
Grg. 517a1 (note οἱ ἔμπροσθεν λόγοι, which makes the reference to 503b-c explicit);
X. An. 3.4.36 (cf. 34).
36 luuk huitink

(20) ἔτι δὲ ἀμφὶ δείλην ἔδοξαν πολεμίους ὁρᾶν ἱππέας (. . .) ἐν ᾧ δὲ

ὡπλίζοντο ἧκον λέγοντες οἱ προπεμφθέντες σκοποὶ ὅτι οὐχ ἱππεῖς
εἰσιν, ἀλλ’ ὑποζύγια νέμοιντο. καὶ εὐθὺς ἔγνωσαν πάντες ὅτι ἐγγύς
που ἐστρατοπεδεύετο βασιλεύς· καὶ γὰρ καπνὸς ἐφαίνετο ἐν κώμαις
οὐ πρόσω.
But while it was still afternoon, they thought they saw horsemen of
the enemy (. . .) While they were arming themselves the scouts that
had been sent ahead returned, saying that they weren’t horsemen
but pack animals grazing. And immediately everybody realized that
the king was camping somewhere nearby; in fact, smoke was seen
in villages not far away. (X. An. 2.2.14–6)
It is with the ὅτι-clause after ἔγνωσαν that both the characters within
the story and the readers learn that the king is much closer than they
had assumed. The shocking information that he is in the vicinity is not
pragmatically presupposed, but new and salient.
In the following passage a complementary participle and a ὅτι-clause
occur in close proximity:
(21) Polus’ first question concerns king Archelaus of Macedon.
ΠΩΛ. εὐδαίμων οὖν σοι δοκεῖ εἶναι ἢ ἄθλιος; ΣΩ. οὐκ οἶδα, ὦ Πῶλε·
οὐ γάρ πω συγγέγονα τῷ ἀνδρί. ΠΩΛ. Τί δέ; συγγενόμενος ἂν
γνοίης, ἄλλως δὲ αὐτόθεν οὐ γιγνώσκεις ὅτι εὐδαιμονεῖ; ΣΩ. μὰ Δί’
οὐ δῆτα. ΠΩΛ. δῆλον δή, ὦ Σώκρατες, ὅτι οὐδὲ τὸν μέγαν βασιλέα
γιγνώσκειν φήσεις εὐδαίμονα ὄντα.
P.: Then do you think he is happy or miserable? S.: I don’t know,
Polus. For I have never met the man. P.: What? You could tell if
you had met him, but otherwise you do not recognize that he is
happy? S.: No, of course not by Zeus. P.: Then it’s clear, Socrates,
that you will say you cannot even recognize that the Persian king
is happy. (Pl. Grg. 470d8–e7)
Although Polus’ initial question is an open one, it is clear which answer
he solicits from Socrates. However, when Socrates declines to acknowl-
edge that ‘Archelaus is happy’, this proposition is not admitted into the
common ground of the conversation. The ὅτι-clause in the next question
shows this: while Polus is committed to the truth of the proposition,
he cannot present it as shared knowledge by putting it in the form of a
participle phrase. Determined not to be caught out again and to show
the absurdity of Socrates’ position, Polus next comes up with an even
more obvious example of a happy king. Polus’ remark derives its ironic
force from the fact that it is simply inconceivable that the Persian king
might be regarded as unhappy. Public opinion made the Persian king
 pragmatic presupposition and complementation 37

the paragon of happiness, which facilitates the accommodation of the

proposition in the conversation.29 The complementary participle also has
the effect of trying to forestall any objection Socrates might have—as
it will turn out in vain.
Another example of a pragmatically accommodated complemen-
tary participle—this time for text-structural reasons—can be found in
Plato’s Phaedo, in the passage in which Socrates’ guard bids his farewell
to the philosopher. This one, too, is found together with an asserted
complement:
(22) σὲ δὲ ἐγὼ καὶ ἄλλως ἔγνωκα ἐν τούτῳ τῷ χρόνῳ γενναιότατον καὶ
πρᾳότατον καὶ ἄριστον ἄνδρα ὄντα τῶν πώποτε ἀφικομένων, καὶ
δὴ καὶ νῦν εὖ οἵδ’ ὅτι οὐκ ἐμοὶ χαλεπαίνεις, γιγνώσκεις γὰρ τοὺς
αἰτίους, ἀλλ’ ἐκείνοις.
During the time you have been here I have come to know you as the
noblest, the gentlest and best man of all those who have ever come
here. And especially now I know that you do not blame me—for
you know who are responsible—but them. (Pl. Phd. 116c4–8)
The guard’s utterance may be analyzed as a complex communicative
move in Kroon’s (1995: 64–7) terms, whose ultimate goal is to elicit
an assurance from Socrates that the guard is not to blame. The move
can be subdivided into two acts, a main and a subsidiary one. The sub-
sidiary act comes first: while the participle phrase has a large amount
of descriptive content and while it is possible that this is in fact the
first time that the guard praises Socrates in this way, this is not the
main point of his uttering ἔγνωκα plus a participle. The connection
ἄλλως (. . .) καὶ δὴ καὶ νῦν between the two sentences shows that the
second one contains the main point:30 ‘surely (εὖ οἶδ’ ὅτι), Socrates will
not blame him, but others’. It is this assertion to which Socrates is to
respond (which he does by not refuting it, thus accepting its addition
to the common ground).31 In other words, the sentence with ἔγνωκα
plus participle provides the preconditions under which the guard
feels confident to make his main and possibly controversial point. As
such, it is presented as presupposed information, which is not to be

29
Cf. Dodds (1959: ad loc.): ‘the King of Persia embodies the popular ideal of supreme
εὐδαιμονία.’ The presupposition may thus be said to be culturally determined.
30
‘The idea conveyed is one of climax’ (1954: 256 on καὶ δὴ καί).
31
Cf. Lewis (1979: 339) for the strategy of lending consent by silence.
38 luuk huitink

commented on further but easily accommodated; one may compare

example (8) above.32
Finally, I will give an example of the rhetorical exploitation of a
pragmatically presupposed proposition. For this I will not look at
my own text sample but instead focus on a long-standing puzzle, the
complementation by a participle of verba dicendi, which occasionally
occurs in tragedy.33 First, however, let us look at one of the very few
examples in prose, a passage from Plato’s Laws:
(23) λέγω γὰρ οὖν ταῦτα οὕτως ἔχοντα, ὡς σὺ λέγεις.
I say that these things are as you say. (Pl. Lg. 893e5)
This instance does not differ from examples (17) and (18) above: all
constituents of the participle phrase are anaphorically bound by the
preceding context (and note also the particle combination γὰρ οὖν,
which can be regarded as a presupposition trigger; cf. Bakker, this
volume). It seems clear that a ὅτι-clause would be less felicitous here:
that construction would turn the sentence into a strong performative
utterance, emphatically asserting the proposition expressed in the ὅτι-
clause. An example of such an utterance is:
(24) ΣΩ. λέγε δή, τί φῂς εἶναι τὸ ὅσιον καὶ τί τὸ ἀνόσιον; ΕΥΘ. λέγω
τοίνυν ὅτι τὸ μὲν ὅσιόν ἐστιν ὅπερ ἐγὼ νῦν ποιῶ (. . .).
S.: Then tell me, how do you define piety and impiety? E.: ‘Very well:
I say that piety is just what I am doing now [namely prosecuting
my own father]. (Pl. Euthpr. 5d7–9).
Here, Euthyphro emphatically and controversially answers Socrates’
question (note τοίνυν in this respect), introducing his own ‘definition’
of piety into the conversation, which will be the subject of the subse-
quent discussion.34

32
Participle phrases dependent on first person cognitive verbs which provide pre-
liminary information are quite common; cf. e.g. Hdt. 7.9α.1; Pl. Grg. 487c1; X. An.
2.5.3; Mem.1.2.53. To this category also belong the ‘knowledge claims’ with οἶδα which
the narrator of the Memorabilia regularly makes at the beginning of a new episode.
These are invariably complemented by participles, as they provide the subject matter or
immediate cause of the ensuing dialogue: cf. X. Mem. 2.9.1; 2.10.1; 3.3.1; 4.4.5; 4.5.2.
33
Cf. Kühner-Gerth (1898/1904: 2.72); Fournier (1946: 184–5) for lists of instances.
34
In fact, Euthyphro singularly fails to answer Socrates’ question. Had he under-
stood Socrates’ dialectic methods, he would have echoed Socrates’ question by using a
φημὶ εἶναι-type answer, postponing any definite assertion until some mutually agreed
conclusions had emerged from the discussion.
 pragmatic presupposition and complementation 39

In the instances of λέγω plus participle in tragedy it happens quite

often that the complementary participle phrase contains what looks
like very salient and crucial information. An example is:
(25) ἈΓΓΕΛΟΣ· ἄνδρες πολῖται, ξυντομωτάτως μὲν ἂν
τύχοιμι λέξας Οἰδίπουν ὀλωλότα·
Messenger: Citizens, to briefly sum up the news: Oedipus is gone.
(S. OC 1579–80)
The messenger’s difficult task is to bring the tidings of Oedipus’ death,
of which his addressees are as yet unaware. We may wonder why we
do not find a ὅτι-clause here, and whether the distinction between
the two complements I have proposed is not invalidated by this and
similar instances. However, had the messenger selected a ὅτι-clause
here, the news could have been interpreted as a strongly performative
assertion along the lines of (24), which could be the starting point of
further discussion. As it stands, the messenger waives any personal
responsibility by presenting the news as a fact which does not admit
of further comments or contradiction but should be taken for granted.
Explaining the passage as a case of presupposition exploitation along
the lines of example (9) above seems to me a fruitful interpretation.
However, it remains to be seen if all comparable instances can be
explained in this way.35

4. Conclusion

In this paper the concept of presupposition was defined in pragmatic

and information-structural terms. This constitutes a departure from
the way in which the notion has been used in Greek linguistics. The
advantages of adding a pragmatic approach to the existing semantic one
have been demonstrated by a contrastive analysis of complementary
participles and ὅτι-clauses dependent on three cognitive verbs. In sec-
tion 3.1 several arguments of a general order were adduced in support
of the hypothesis that participles express presupposed and ὅτι-clauses
asserted information. My concern here was to place my analysis on a

35
In my view, the participle in the following instances can be understood in terms of
presupposition exploitation or at least as a way of avoiding a ὅτι-clause, which would
result in an unwarranted assertion: A. A. 269, 583, Ch. 682; S. Ant. 995, El. 676; E. Alc.
1158, IA 803, Hel. 537, 823, 1076, Or. 1581.
40 luuk huitink

secure theoretical footing. This was necessary as the data analysis in

section 3.2 was of a provisional nature. It focused on some clear-cut
examples and on passages in which accommodation and exploitation
of presuppositions play a role. It has been shown that the distinction
drawn in this paper may be put to use in interpreting the communi-
cative intentions and narrative strategies of speakers and narrators in
relevant passages of Greek literature.
Further research is still required and may focus on two points in
particular. First, information is often not entirely ‘old’ or entirely ‘new’;
rather these two options are to be seen as extremes on a scale, with
more or less contextually prepared information in between (cf. Buijs
(2005: 17–21)). It would therefore be worth asking whether there are
certain contexts in which a speaker has a genuine choice between a
participle and a ὅτι-clause. Secondly, cases of presupposition exploita-
tion as the one in example (25) above raise a methodological point as
to the validity of the claim put forward here. Further research into the
communicative motives and effects of this use of presupposition may
help to make for a less ad hoc-approach to such instances.
 CHAPTER THREE

ON THE CURIOUS COMBINATION OF THE PARTICLES

ΓAΡ AND ΟYΝ

Stéphanie J. Bakker

1. Introduction1

In Plato the combination of the particles γάρ and οὖν occurs no fewer
than 181 times. This is perhaps surprising, given the seemingly con-
tradictory nature of these two particles. According to recent studies,
both particles have a text-organizing function and therefore contribute
to discourse cohesion. However, there is a certain contradiction here:
γάρ marks the utterance in question as an explanation, examplifica-
tion or digression with regret to the main story line, and thus as less
relevant than the preceding utterance(s).2 Oὖν, by contrast, indicates
a new relevant step, one that marks the preceding utterances as sub-
sidiary.3 As Slings (1997a) notes, γάρ is a ‘push’ particle that pushes
the utterance to a lower level than the preceding discourse, while οὖν
is a ‘pop’ particle, ‘popping’ the utterance to a higher level than what
precedes it.4

1
The research for this article was funded by the Leiden ‘Stichting voor de Weten-
schapsbeoefening’. I would like to thank the members of the Cratylus team of the
‘Limits of Language Project’ (financed by NWO), the participants in the Conference
on Greek and Latin Linguistics in Katwijk (2007) and Gerry Wakker for their valuable
suggestions and comments.
2
See Rijksbaron et al. (2000: 160) and Sicking and Van Ophuijsen (1993: 20). The
latter state: ‘anyone who marks the beginning of a section of his text with γάρ is sub-
ordinating the stretch which comes within the scope of the particle to a (preceding or
following) item in his narrative or argument.’
3
See Sicking and Van Ophuijsen (1993: 27): ‘the speaker marks that what precedes
as relevant, and for the present purpose subsidiary, to what follows’ and Sicking and
Van Ophuijsen (1993: 91) ‘οὖν indicates, then, that what follows comes nearer to the
point that what precedes; that what precedes owes its relevance to what follows.’ A
similar view on the function of οὖν can be found in De Jong (1997), Rijksbaron et al.
(2000) and Slings (1997a).
4
The terms push and pop are borrowed from Polanyi and Scha (1983).
42 stéphanie j. bakker

Although γάρ and οὖν themselves have received considerable atten-

tion in recent particle studies, we have to rely on Denniston and Des
Places for a description of the function of these particles in combination.
Denniston (1954: 445–8) states that in Homer οὖν in the combination
γὰρ οὖν ‘always has a backward reference’. In the examples he quotes,
Denniston paraphrases οὖν as ‘as described above’ or ‘in fact’. In post-
Homeric Greek, by contrast, οὖν adds to γάρ ‘the idea of importance
or essentiality’ (Denniston 1954: 446). ‘Really’, ‘in truth’, ‘certainly’ are
some of the translations with which Denniston attempts to do justice
to this intensifying aspect of οὖν. Des Places (1929: 122–131), who
confines himself to the use of γὰρ οὖν in Plato, agrees with Denniston
that by adding οὖν, the speaker is insisting on the truth of the utter-
ance introduced by γάρ, but only where γὰρ οὖν is used in an ‘exposé
continu’.5 According to Des Places, οὖν does not have an intensifying
function in dialogical contexts (i.e., after a turn-taking), but is added
to γάρ in order to underline the relation between the utterance at issue
and the preceding lines. Although Des Places’ interpretation of οὖν
is different in these cases, his translation remains the same, viz. ‘en
effet’ (a translation which, in my view, is also inadequate to express
an intensifying function).
Regardless whether the combination γὰρ οὖν functions differently
after a turn-taking and in ongoing discourse, it is questionable whether
in this combination οὖν ever has an ‘intensifying’ function. In several
examples given by Denniston, the intensfying function attributed to
οὖν should be ascribed to other factors. For example, in Ar. Th. 164
(καὶ Φρύνιχος,—τοῦτον γὰρ οὖν ἀκήκοας ‘and Phrynichos, for you
have heard of him’) the emphasis on τοῦτον should not, in my view,
be attributed to an intensifying function of οὖν, but rather to the fact
that τοῦτον is in focus position.6 In Hdt. 1.49 (οὐ γὰρ ὦν οὐδὲ τοῦτο
λέγεται, ‘for that is not been said’), it is not the particle οὖν, but the
combination οὐ . . . οὐδέ that serves to emphasize τοῦτο.
In view of the fact that Denniston’s examples are not always in line
with his conclusions and that he and Des Places differ when it comes to
the function of γὰρ οὖν after a turn-taking and seeing that the combina-
tion of the particles γάρ and οὖν has not been studied since Denniston

5
Des Places argues that οὖν may also be added to γάρ to resume the main story
line after a digression (Des Places 1929: 126). Des Places illustrates this use of οὖν with
my example (13), among others.
6
For the position of focus in the Ancient Greek clause, see H. Dik (1997).
 the curious combination of the particles ΓAΡ and ΟYΝ 43

and Des Places, there is more than enough reason to examine afresh
the curious combination of the particles γάρ and οὖν. In this paper, I
will examine what the particles γάρ and οὖν ‘do’ when combined in
one utterance, and how this function relates to the usual functions of
γάρ and οὖν. Does the combination of these particles indeed contribute
to discourse cohesion, as γάρ and οὖν do individually, and if so, how?
In trying to answer these questions I have relied mainly on examples
from Plato.7

2. The most natural interpretation

Despite the conflicting nature of the particles γάρ and οὖν as described
in the introduction, one might expect the function of γὰρ οὖν to repre-
sent the sum of the functions of γάρ and οὖν respectively. This would
imply that although in a formal sense γάρ places the utterance on a
lower level than the preceding discourse, οὖν marks its content as more
appropriate for the present purpose. Examples (1) to (3) suggest that
this is indeed the function of the combination γὰρ οὖν:
(1) κομιδῇ δέ γε, ὦ ξένε, ἔοικεν ἀληθὲς εἶναι τὸ περὶ τὸν σοφιστὴν κατ’
ἀρχὰς λεχθέν, ὅτι δυσθήρευτον εἴη τὸ γένος. φαίνεται γὰρ οὖν προ-
βλημάτων γέμειν, ὧν ἐπειδάν τι προβάλῃ, τοῦτο πρότερον ἀναγκαῖον
διαμάχεσθαι πρὶν ἐπ’ αὐτὸν ἐκεῖνον ἀφικέσθαι.
It certainly seems, Stranger, that what you said at first about the
sophist—that he was a hard kind of creature to catch—is true; for he
seems to have no end of defences, and when he throws one of them
up, his opponent has first to fight through it before he can reach the
man himself. (Pl. Sph. 261a6)8

(2) σπουδάζοντα δ’ εἴ με τιθέναι βούλεσθε, τίθετε· πάνυ γὰρ οὖν

προσδοκῶ νῦν ὑμᾶς εὑρήσειν τῷ λόγῳ ἑπομένους ὃν ὀλίγον ἔμπροσθε
προυθέμεθα, τῆς τῶν βασιλέων τε φθορᾶς καὶ ὅλου τοῦ διανοήματος
οὐ δειλίαν οὖσαν τὴν αἰτίαν, (. . .).

7
For this paper, I examined all instances of γὰρ οὖν in the Apology, Charmides,
Cratylus, Symposium, Republic, Sophist, Politicus, Timaeus and Leges. By opting for these
dialogues, I have tried to cover the various periods of Plato’s writings. Another factor
in the selection was the frequency of the combination of the particles γάρ and οὖν in
ongoing discourse (i.e. without a turn-taking). Examples of γὰρ οὖν in Herodotus and
Xenophon were also studied, but less systematically.
8
The translations of the examples in this paper are (slight adaptations) of the
translations in the Loeb series.
44 stéphanie j. bakker

If you want to take my words seriously, do so. For I certainly expect

that, as you follow the argument recently propounded, you will now
discover that the cause of the ruin of those kingdoms, and of their
whole design, was not cowardice (. . .). (Pl. Lg. 688c1–5)

(3) πάντες γὰρ κεκοινωνήκατε τῆς φιλοσόφου μανίας τε καὶ βακχείας—

διὸ πάντες ἀκούσεσθε· συγγνώσεσθε γὰρ τοῖς τε τότε πραχθεῖσι καὶ
τοῖς νῦν λεγομένοις. οἱ δὲ οἰκέται, καὶ εἴ τις ἄλλος ἐστὶν βέβηλός τε
καὶ ἄγροικος, πύλας πάνυ μεγάλας τοῖς ὠσὶν ἐπίθεσθε. ἐπειδὴ γὰρ
οὖν, ὦ ἄνδρες, ὅ τε λύχνος ἀπεσβήκει καὶ οἱ παῖδες ἔξω ἦσαν, ἔδοξέ
μοι (. . .).
Every one of you has had his share of philosophic frenzy and trans-
port, so all of you shall hear. You shall stand up alike for what then
was done and for what now is spoken. But the domestics, and all else
profane and clownish, must clap the heaviest of doors upon their ears.
‘Well, gentlemen, when the lamp had been put out and the servants
had withdrawn, I determined (. . .).’ (Pl. Smp. 218b3–c1)
In example (1), it can be maintained that although the utterance intro-
duced by γὰρ οὖν is formally a justification of the statement that the
stranger’s observations about the sophist were correct, Theaetetus is in
fact drawing attention to the problem-raising strategy of the sophist.
As regards example (2), it can likewise be maintained that whereas γάρ
presents the utterance as an explanation to the preceding order, οὖν
indicates that the speaker is coming to the point after the intermezzo.
Something similar happens in example (3). Here, it is not the argu-
mentation, but rather a story that is interrupted in order to address the
conversation partners. While γάρ indicates that Alcibiades has finally
redeemed the promise evoked by ἀκούσεσθε,9 it can be argued that
οὖν marks the fact that the speaker has finally come to the point and
is about to relate what happened that night.
These three examples suggest that although the combination γὰρ οὖν
seems problematic at first sight, it is quite possible that an utterance is
both an explanation of—and thus subordinate to—what precedes, and
at the same time more to the point. However, the problem is that in a
large number of examples the utterance introduced by γὰρ οὖν cannot
be seen as more relevant or more to the point than what precedes. The
most problematic are those instances in which γὰρ οὖν introduces a
parenthesis:

9
For this use of γάρ to introduce an embedded narrative, see De Jong (1997).
 the curious combination of the particles ΓAΡ and ΟYΝ 45

(4) ὁ δὴ γέρων—σφόδρα γὰρ οὖν μέμνημαι—μάλα τε ἥσθη καὶ

διαμειδιάσας εἶπεν·
Whereat the old man (for I remember the scene well) was highly
pleased and said with a smile (. . .). (Pl. Ti. 21c2–4)

(5) καὶ Φρύνιχος,—τοῦτον γὰρ οὖν ἀκήκοας,—

αὐτός τε καλὸς ἦν καὶ καλῶς ἠμπίσχετο.
And Phrynichus (for you have heard of him), what a dandy was he
and how careful in his dress! (Ar. Th. 164–5)

(6) τῶν δὲ Ἑλλήνων τῶν Μαρδονίου συμμάχων οἶδε μὲν οὐδεὶς ἀριθμόν
(οὐ γὰρ ὦν ἠριθμήθησαν), ὡς δὲ ἐπεικάσαι, ἐς πέντε μυριάδας
συλλεγῆναι εἰκάζω.
As for the Greek allies of Mardonius, no one knows the number
of them (for they were not counted), I suppose them to have been
mustered to the number of fifty thousand. (Hdt. 9.32.2)
In examples like these, which are quite common—in Herodotus more
than half of the instances of γὰρ οὖν is marked as a parenthesis by the
editor of the text—it is difficult to maintain that the utterance intro-
duced by γὰρ οὖν is more relevant or more to the point than what
precedes.
The same holds true for many of the instances—especially in
Plato—in which γὰρ οὖν is used in a brief response indicating approval
of the argumentation of the conversation partner. These reactions can
not always be analyzed as a subsequent and more relevant step in the
discussion, or as a reaction to a more relevant step in the argumenta-
tion. In example (7), for instance, Socrates tries to convince Charmides
of the fact that it is awkward to assume that knowledge has itself as
its object. He argues that desire, wishes and love, too, never focus on
themselves, but rather on another object:
(7) ἀλλ’ ἐπιθυμία δοκεῖ τίς σοι εἶναι, ἥτις ἡδονῆς μὲν οὐδεμιᾶς ἐστὶν
ἐπιθυμία, αὑτῆς δὲ καὶ τῶν ἄλλων ἐπιθυμιῶν;
—οὐ δῆτα.
οὐδὲ μὴν βούλησις, ὡς ἐγᾦμαι, ἣ ἀγαθὸν μὲν οὐδὲν βούλεται, αὑτὴν
δὲ καὶ τὰς ἄλλας βουλήσεις βούλεται.
—οὐ γὰρ οὖν.
ἔρωτα δὲ φαίης ἄν τινα εἶναι τοιοῦτον, ὃς τυγχάνει ὢν ἔρως καλοῦ
μὲν οὐδενός, αὑτοῦ δὲ καὶ τῶν ἄλλων ἐρώτων;
—οὔκ, ἔφη, ἔγωγε.
Socr. Now, do you think there is any desire which is the desire, not
of any pleasure, but of itself and of the other desires?’
46 stéphanie j. bakker

Charm. Surely not.

Socr. Nor, again, is there a wish, I imagine, that wishes no good, but
wishes itself and the other wishes.
Charm. Quite so; there is not.
Socr. And would you say there is any love of such a sort that it is
actually a love of no beauty, but of itself and of the other loves?
Not I, he replied. (Pl. Chrm. 167e1–9)
In the enumeration of the various analogies (ἐπιθυμία, βούλησις, ἔρως,
δόξα and φόβος), only the reaction to the second analogy is marked with
γὰρ οὖν. I fail to see why this reaction or this analogy should be more
to the point than the others. In example (8), the utterance marked by
γὰρ οὖν is also a reaction to an intermediate step in the argumentation.
In this episode, the stranger tries to persuade Theaetetus that mental
purification differs from other kinds of purification:
(8) ΞΕ. πονηρίαν ἕτερον ἀρετῆς ἐν ψυχῇ λέγομέν τι;
ΘΕΑΙ. πῶς γὰρ οὔ;
ΞΕ. καὶ μὴν καθαρμός γ’ ἦν τὸ λείπειν μὲν θάτερον, ἐκβάλλειν δὲ
ὅσον ἂν ᾖ πού τι φλαῦρον.
ΘΕΑΙ. ἦν γὰρ οὖν.
ΞΕ. καὶ ψυχῆς ἄρα, καθ’ ὅσον ἂν εὑρίσκωμεν κακίας ἀφαίρεσίν τινα,
καθαρμὸν αὐτὸν λέγοντες ἐν μέλει φθεγξόμεθα.
Str. Do we say that wickedness is distinct from virtue in the soul?
Theae. Of course
Str. And purification was retaining the one and throwing out whatever
is bad anywhere?’
Theae. Yes, it was.
Str. Hence whenever we find any removal of evil from the soul, we
shall be speaking properly if we call that a purification. (Pl. Sph.
227d4–11)
Here γὰρ οὖν is a reaction to the second step in the argumentation,
which in my view is no more relevant than the first step, where the
reaction is not marked by γὰρ οὖν.
Apart from parentheses and reactions of approval, which together
form the greater part of the instances of γὰρ οὖν, there are others
in which it is doubtful whether οὖν really marks a step in the argu-
mentation or story that is more relevant or to the point than what
precedes:
(9) (. . .) μήτε ἄλλον, γέροντα ἢ νέον, ἐᾶν πάσχειν ταὐτὸν τοῦθ’ ἡμῖν,
ἄρρενα ἢ θῆλυν, ἁπάντων δὲ ἥκιστα εἰς δύναμιν τὸν ἀρτίως νεογενῆ·
κυριώτατον γὰρ οὖν ἐμφύεται πᾶσι τότε τὸ πᾶν ἦθος διὰ ἔθος. ἔτι δ’
ἔγωγ’, εἰ μὴ μέλλοιμι δόξειν παίζειν, φαίην ἂν δεῖν καὶ τὰς φερούσας ἐν
 the curious combination of the particles ΓAΡ and ΟYΝ 47

γαστρὶ πασῶν τῶν γυναικῶν μάλιστα θεραπεύειν ἐκεῖνον τὸν ἐνιαυτόν

ὅπως μήτε ἡδοναῖς τισι πολλαῖς ἅμα καὶ μάργοις προσχρήσεται ἡ
κύουσα μήτε αὖ λύπαις, (. . .).
(we should not become prone to pleasure, nor be devoid of pain)
and not allow any other person—old or young, man or woman—to
be in this condition and least of all, so far as possible, the new-born
baby. For because of the force of habit, it is in infancy that the whole
character is most effectually determined. I should assert further—were
it not that it would be taken as a jest—that women with child, above
all others, should be cared for during their year of pregnancy, lest any
of them should indulge in repeated and intense pleasures or pains,
(. . .). (Pl. Lg. 792d7–e6)
In example (9) the Athenian defends the view that we should strive to
free ourselves from pleasure and pain. Moreover, we should protect
others from these evil influences, especially young babies and even—
although the Athenian admits that this may sound a bit ridiculous—the
unborn baby. Following his statement that babies must be protected
against the evil influence of pain and pleasure, he explains that our
character is determined mainly in our early years. This explanation
is introduced by the combination of the particles γάρ and οὖν. The
addition of οὖν can hardly be taken as an indication that the fact that
our character is determined at an early age is the main point in the
Athenian’s argumentation.
Because of the many problematic cases, I cannot agree that the most
natural interpretation of utterances introduced by γὰρ οὖν, i.e. that the
utterance is formally subordinate to what precedes, but more to the point
with respect to its content, is correct. Although γάρ undeniably offers a
measure of explanation or additional information, it is debatable whether
οὖν actually marks the information as more relevant or to the point than
the preceding utterances. In the next section, I will therefore examine
in more detail the function of οὖν in the γὰρ οὖν combination.

3. An alternative interpretation

If one compares the utterances with γὰρ οὖν in Herodotus with the
utterances introduced by γάρ alone, it is striking that γὰρ οὖν utterances
generally provide very little if any new information. The explanation
provided in these utterances appears to be either largely familiar or
easily inferable from the preceding discourse:
48 stéphanie j. bakker

(10) νόμος δὲ τοῖσι Λακεδαιμονίοισι κατὰ τῶν βασιλέων τοὺς θανάτους

ἐστὶ ὡυτὸς καὶ τοῖσι βαρβάροισι τοῖσι ἐν τῇ Ἀσίῃ· τῶν γὰρ ὦν
βαρβάρων οἱ πλέονες τῷ αὐτῷ νόμῳ χρέωνται κατὰ τοὺς θανάτους
τῶν βασιλέων. ἐπεὰν γὰρ ἀποθάνῃ βασιλεὺς Λακεδαιμονίων, (. . .).
The Lacedaemonians have the same custom at the deaths of their
kings as the foreigners in Asia. For most foreigners use the same
custom at their kings’ deaths. When a king of the Lacedaemonians
dies, (. . .). (Hdt. 6.58.2)

(11) The god [Belos, a Babylonian god] visits the shrine and sleeps with
a woman who is appointed by the god, as in Thebes of Egypt (. . .)
καὶ κατά περ ἐν Πατάροισι τῆς Λυκίης ἡ πρόμαντις τοῦ θεοῦ, ἐπεὰν
γένηται· οὐ γὰρ ὦν αἰεί ἐστι χρηστήριον αὐτόθι· ἐπεὰν δὲ γένηται,
τότε ὦν συγκατακληίεται τὰς νύκτας ἔσω ἐν τῷ νηῷ.
And as does the prophetess of the god at Patara in Lycia, whenever
she is appointed; for there is not always a place of divination there;
but when she is appointed she is shut up in the temple during the
night. (Hdt. 1.182.1–2)
The explanation in the second clause of example (10) offers very little
new information in comparison to the first clause, as the expression
‘the barbarians in Asia’ makes it clear that the barbarians, at any rate
the majority of them, have the same customs with respect to royal
deaths. Otherwise, Herodotus would not have used the expression ‘the
barbarians’, but rather a phrase like ‘some of the barbarians in Asia’.
Similarly, the information in the utterance introduced by γὰρ ὦν in
example (11) is already more or less familiar. Here, we are told that at
night the god Belos sleeps with a woman in this temple, as in Thebes of
Egypt and as does the prophetess of the god at Patara in Lycia (whenever
she is there, for she is not always there). After the subordinate clause
ἐπεὰν γένηται, the utterance introduced by γὰρ ὦν is no longer very
informative. It notes more explicitly that the prophetess is not always
in Patara, but the reader would already have been aware of this fact
from the preceding subordinate clause.
Whereas Herodotus uses utterances introduced by γὰρ ὦν for
information which is (largely) given or easily inferable, he usually uses
utterances introduced by γάρ alone for information that is completely
new to the reader:
(12) ἔχοντες Ὀνομάκριτον, ἄνδρα Ἀθηναῖον χρησμολόγον τε καὶ
διαθέτην χρησμῶν τῶν Μουσαίου, ἀνεβεβήκεσαν, τὴν ἔχθρην
προκαταλυσάμενοι. ἐξηλάσθη γὰρ ὑπὸ Ἱππάρχου τοῦ Πεισιστράτου
ὁ Ὀνομάκριτος ἐξ Ἀθηνέων, (. . .).
 the curious combination of the particles ΓAΡ and ΟYΝ 49

They [the Pisistratidae] had come up to Sardis with Onomacritos, an

Athenian diviner who had set in order the oracles of Musaios, after
having reconciled their previous hostility with him. For Onomacritos
had been banished from Athens by Pisitratus’ son Hipparchus, (. . .).
(Hdt. 7.6.3)

(13) τῶν δὲ οὔ φασι θεῶν γινώσκειν τὰ οὐνόματα, οὗτοι δέ μοι δοκέουσι

ὑπὸ Πελασγῶν ὀνομασθῆναι, πλὴν Ποσειδέωνος. τοῦτον δὲ τὸν
θεὸν παρὰ Λιβύων ἐπύθοντο· οὐδαμοὶ γὰρ ἀπ’ ἀρχῆς Ποσειδέωνος
οὔνομα ἔκτηνται εἰ μὴ Λίβυες, καὶ τιμῶσι τὸν θεὸν τοῦτον αἰεί.
The gods whose names they say they do not know were, as I think,
named by the Pelasgians, except Poseidon, the knowledge of whom
they learned from the Libyans; for alone of all nations the Libyans
have had among them the name of Poseidon from the beginning,
and they have always honored this god. (Hdt. 2.50.2–3)
Example (12) is similar to example (11) in that the second sentence
clarifies the last part of the first sentence. But while the explanation in
example (11) offers no new information, but is only a more explicit
formulation of information that is already familiar, the explanation
about of the hostile relationship with Onomacritos in example (12) is
completely new to the reader. Similarly, the explanation of the assump-
tion that the Egyptians have learned the name of Poseidon from the
Lybians in example (13) would have been completely new information
to the average reader.
According to Van Ophuijsen, an utterance introduced by οὖν provides
information already known or highly expectable for the addressees. He
maintains that: ‘the most we can say that applies to all three [examples,
SB] is that we have been prepared for the information contained in the
sentence with οὖν by that preceding it’ (Van Ophuijsen 1993: 90). On
the next page, however, he argues that there is a more important aspect
of the function of οὖν: ‘οὖν indicates, then, that what follows comes
nearer to the point than what precedes; that what precedes owes its
relevance to what follows’ (Van Ophuijsen 1993: 91). As I attempted to
show in section 2, however, it is highly questionable whether ‘coming
nearer to the point’ is really part of the function of οὖν in utterances
introduced by γὰρ οὖν. Before dealing with the question of how the
‘common’ use of οὖν, as described by Van Ophuijsen and others, is
related to the use of οὖν in utterances introduced by γὰρ οὖν (and
whether we should distinguish two separate functions for οὖν), I will
return to Plato and, on the basis of additional examples, demonstrate
that utterances marked by γὰρ οὖν provide already familiar or at least
highly expectable information.
50 stéphanie j. bakker

4. γὰρ οὖν in ongoing discourse in Plato

Although the examples from Plato are less obvious than those from
Herodotus, it can still be argued that the addressee has been prepared
for the explication or exemplification introduced by γὰρ οὖν, in any
case in those instances where the utterance introduced by γὰρ οὖν is
expressed by the same speaker as the preceding utterances. After turn-
takings, i.e., in reactions introduced by γὰρ οὖν, it is more difficult to
establish this function of οὖν, as I will demonstrate in the next section.10
In those instances where there is no change of speaker, however, the
utterance at issue arguably provides little or no new information, but
is inferable or highly expectable due to either the preceding discourse
or the general knowledge shared by the conversation partners.11
A clear example of an utterance introduced by γὰρ οὖν providing
information already familiar on the basis of the preceding discourse, is
example (3). Before the brief interruption of his story, Alcibiades has
told his audience that Socrates stayed overnight with him (because it
was very late), and that there was no one else in the room. After the
interruption, Alcibiades resumes his story with a temporal subordinate
clause in which he briefly summarizes the last part of his story before
the interruption. Thus the temporal subordinate clause introduced by
γὰρ οὖν provides already familiar information.
Another clear instance of an utterance introduced by γὰρ οὖν which
provides information already familiar on the basis of the preceding
discourse, is example (14):
(14) μὴ θορυβεῖτε, ὦ ἄνδρες Ἀθηναῖοι, ἀλλ’ ἐμμείνατέ μοι οἷς ἐδεήθην
ὑμῶν, μὴ θορυβεῖν ἐφ’ οἷς ἂν λέγω ἀλλ’ ἀκούειν· καὶ γάρ, ὡς ἐγὼ
οἶμαι, ὀνήσεσθε ἀκούοντες. μέλλω γὰρ οὖν ἄττα ὑμῖν ἐρεῖν καὶ ἄλλα
ἐφ’ οἷς ἴσως βοήσεσθε· ἀλλὰ μηδαμῶς ποιεῖτε τοῦτο.
Do not make a disturbance, men of Athens; continue to do what
I asked of you, not to interrupt my speech by disturbances, but to
hear me; for I believe you will profit by hearing. For I am going to

10
In the Platonic dialogues mentioned in note 7 γὰρ οὖν occurs 33 times in ongoing
discourse, as opposed to 70 times after a turn-taking.
11
In some cases the information provided by the utterance introduced by γὰρ οὖν
is not inferable from what was said in the preceding discourse, but from the mere
fact that something was said at all. This is for instance the case in example (4). In this
example, the function of οὖν can be paraphrased as follows: ‘as is clear from the fact
that I mention this’. In such cases, the preceding discourse from which the information
in the γὰρ οὖν-clause can be inferred is limited to the preceding words.
 the curious combination of the particles ΓAΡ and ΟYΝ 51

say some things to you at which you will perhaps cry out; but do
not do so by any means. (Pl. Ap. 30c2–5)
Although Socrates has not made this explicit before the utterance with
γὰρ οὖν, on the basis of the fact that he again asks his audience to remain
silent, it is clear that he is going to say something that will meet with
considerable resistance, and thus shouting and crying. Therefore, the
explanation offered in the utterance marked by γὰρ οὖν may be said
to be expected on the basis of the preceding discourse.
As indicated above, the content of the utterance introduced by γὰρ
οὖν may be familiar or inferable not only from the preceding discourse,
but also on the basis of general knowledge shared by the conversation
partners, as in example (9).12 There, the Athenian defends the view
that wherever possible we should refrain from pleasure and pain, and
that this is especially important for young babies, who are most liable
to acquire bad habits. By adding οὖν to this argument, the Athenian
indicates that he regards the sensitivity of young babies as a generally
known and accepted phenomenon.
In example (15), too, οὖν may indicate that the explanation offered
by the speaker for his unusual statement that fear leads to victory is
generally known. As his conversation partner knows, victory is ensured
not only by confidence in the face of the enemy, but also by fear of
being seen as a coward by friends:
(15) ΑΘ. ἆρ’ οὖν οὐ καὶ νομοθέτης, καὶ πᾶς οὗ καὶ σμικρὸν ὄφελος,
τοῦτον τὸν φόβον ἐν τιμῇ μεγίστῃ σέβει, καὶ καλῶν αἰδῶ, τὸ τούτῳ
θάρρος ἐναντίον ἀναίδειάν τε προσαγορεύει καὶ μέγιστον κακὸν ἰδίᾳ
τε καὶ δημοσίᾳ πᾶσι νενόμικεν;
ΚΛ. ὀρθῶς λέγεις.
ΑΘ. οὐκοῦν τά τ’ ἄλλα πολλὰ καὶ μεγάλα ὁ φόβος ἡμᾶς οὗτος
σῴζει, καὶ τὴν ἐν τῷ πολέμῳ νίκην καὶ σωτηρίαν ἓν πρὸς ἓν οὐδὲν
οὕτως σφόδρα ἡμῖν ἀπεργάζεται; δύο γὰρ οὖν ἐστὸν τὰ τὴν νίκην
ἀπεργαζόμενα, θάρρος μὲν πολεμίων, φίλων δὲ φόβος αἰσχύνης πέρι
κακῆς.
ΚΛ. ἔστι ταῦτα.

12
Other languages do not necessarily express the two kinds of familiarity in the
same manner. In Dutch, for instance, information familiar on the basis of the preced-
ing discourse is generally marked with dus, e.g. hij zei dus dat hij niet wilde komen,
‘so he said (as I told you before) that he didn’t want to come’ (although immers is also
possible). By contrast, information that is familiar on the basis of general knowledge is
always marked with immers, e.g. de winkels zijn immers na 20u gesloten, ‘for the shops
are closed after 8 p.m.’ (dus is not possible in these cases).
52 stéphanie j. bakker

ΑΘ. ἄφοβον ἡμῶν ἄρα δεῖ γίγνεσθαι καὶ φοβερὸν ἕκαστον· ὧν δ’

ἑκάτερον ἕνεκα, διῃρήμεθα.
Ath. Does not, then, the lawgiver, and every man who is worth
anything, hold this kind of fear in the highest honour, and name
it ‘modesty’; and to the confidence which is opposed to it does he
not give the name ‘immodesty,’ and pronounce it to be for all, both
publicly and privately, a very great evil?
Klin. Quite right.
Ath. And does not this fear, besides saving us in many other impor-
tant respects, prove more effective than anything else in ensuring
for us victory in war and security? For victory is, in fact, ensured
by two things, of which the one is confidence towards enemies, the
other, fear of the shame of cowardice in the eyes of friends.
Klin. That is so.
Ath. Thus each one of us ought to become both fearless and fear-
ful; and that for the several reasons we have now explained. (Pl. Lg.
647a8–c1)
The fact that the speaker may add οὖν to indicate that the explanation
or justification is known, inferable or expected does not mean that
that information is always known, inferable or expected, nor that an
explanation or justification introduced by γάρ alone cannot provide
known or inferable information:
(16) τὰ μὲν οὖν παρεληλυθότα τῶν εἰρημένων πλὴν βραχέων ἐπιδέδεικται
τὰ διὰ νοῦ δεδημιουργημένα· δεῖ δὲ καὶ τὰ δι’ ἀνάγκης γιγνόμενα τῷ
λόγῳ παραθέσθαι. μεμειγμένη γὰρ οὖν ἡ τοῦδε τοῦ κόσμου γένεσις
ἐξ ἀνάγκης τε καὶ νοῦ συστάσεως ἐγεννήθη·
The foregoing part of our discourse, save for a small portion, has
been an exposition of the operations of Reason; but we must also
furnish an account of what comes into existence through Necessity.
For this Cosmos in its origin was generated as a compound, from
the combination of Necessity and Reason. (Pl. Ti. 48a)

(17) τοῦτο μὲν τοίνυν ἀπὸ τῆς ἐν τοῖς ὅπλοις ὀρχήσεως ἡγούμενοι
τεθῆναι ὀρθῶς ἄν, ὡς ἐγᾦμαι, ἡγοίμεθα· τὸ γάρ που ἢ αὑτὸν ἤ τι
ἄλλο μετεωρίζειν ἢ ἀπὸ τῆς γῆς ἢ ἐν ταῖς χερσὶν ‘πάλλειν’ τε καὶ
‘πάλλεσθαι’ καὶ ὀρχεῖν καὶ ὀρχεῖσθαι καλοῦμεν.
Those of us are right, I fancy, who think this name is derived from
armed dances, for lifting oneself or anything else from the ground
or in the hands is called shaking and being shaken, or dancing and
being danced. (Pl. Cra. 406e2)
In example (16), the content of the utterance marked with γὰρ οὖν is
not familiar from the preceding discourse, nor, presumably, is it part
 the curious combination of the particles ΓAΡ and ΟYΝ 53

of the general knowledge of the conversation partners. However, it is

convenient for the speaker to present the information as such, because
he wants to legitimate the transition to the subject of the influence
of Necessity. Example (17) shows that general knowledge used as an
explanation is not necessarily marked by the particle combination γὰρ
οὖν. In this example, the speaker chooses not to mark the content of the
utterance as generally known beforehand, but to leave his conversation
partner some room by using the particle που.13 The last two examples
show that the difference between utterances introduced by γὰρ οὖν and
utterances introduced by γάρ alone does not hinge on whether or not
the information is (generally) known or inferable, but rather on whether
or not the speaker wants to present the information as such.

5. γὰρ οὖν in reactions

As we have seen, many of the examples of γὰρ οὖν in Plato consist of

a sign of approval on the part of the conversation partner. Whether in
these cases οὖν also indicates that the reaction is to be expected in view
of the preceding discourse or the general knowledge of the conversation
partners, is much more difficult to determine, since in general reactions
introduced by γὰρ οὖν do not contain any more information than the
repetition of a word from the preceding lines and the combination of
the particles γάρ and οὖν:
(18) ΞΕ. καὶ τοῦ πτηνοῦ μὴν γένους πᾶσα ἡμῖν ἡ θήρα λέγεταί πού τις
ὀρνιθευτική.
ΘΕΑΙ. λέγεται γὰρ οὖν.
Str. And the hunting of water creatures goes by the general name
of fishing?
Theae. That is indeed the general name. (Pl. Sph. 220b4–6)

(19) δεῖ δέ γε οὔχ, ὡς ἄρτι ἡμῖν ὁ λόγος ἐσήμαινεν· ᾧ πειστέον, ἕως ἄν

τις ἡμᾶς ἄλλῳ καλλίονι πείσῃ.—οὐ γὰρ οὖν δεῖ.
But that must not be, as our reasoning but now showed us, in
which we must put our trust until someone convinces with a better
reason.—No, it must not be. (Pl. R. 388e2–4)

13
In his reaction (πάνυ μὲν οὖν) Hermogenes appears to maintain that the content
of the utterance is inferable from general knowledge. For the use of οὖν in reactions
without γάρ, see note 15.
54 stéphanie j. bakker

The function of γάρ in these reactions is clear: it serves to clarify the

reaction of approval (‘you are right/that is correct, for . . .’). The function
of the addition of οὖν, on the other hand, is much more difficult to deter-
mine, mainly because of the nature of the somewhat contentless reaction.
But although there may be no strict differences between reactions with
and without οὖν, there are some clear tendencies. In the first place,
reactions introduced by γάρ alone may contain new information:
(20) ΣΩ. ἦ καὶ ὁπόσα ἂν φῇ τις ἑκάστῳ ὀνόματα εἶναι, τοσαῦτα ἔσται
καὶ τότε ὁπόταν φῇ;
ΕΡΜ. οὐ γὰρ ἔχω ἔγωγε, ὦ Σώκρατες, ὀνόματος ἄλλην ὀρθότητα ἢ
ταύτην, ἐμοὶ μὲν ἕτερον εἶναι καλεῖν ἑκάστῳ ὄνομα, ὃ ἐγὼ ἐθέμην,
σοὶ δὲ ἕτερον, ὃ αὖ σύ.
Socr. And whatever the number of names anyone says a thing has,
it will really have that number at the time when he says it?
Herm. Yes, Socrates, for I cannot conceive of any other kind of cor-
rectness in names than this; I may call a thing by one name, which
I gave, and you by another, which you gave. (Pl. Cra. 385d5–9)
Whereas the γάρ-reaction in this example clearly provides new infor-
mation, this is never the case in reactions where γάρ is combined with
οὖν since, as noted above, they contain only a repetition of a word from
the preceding lines, together with the combination of the particles γάρ
and οὖν (see examples 18 and 19).
Of course, the fact that reactions introduced by γάρ alone may
contain new information does not necessarily mean that they always
do so. There are various examples like (21) and (22) which, like reac-
tions introduced by the combination of γάρ and οὖν, contain only the
repetition of a word from the preceding lines:
(21) ἀλλὰ μὴν Ἔρωτά γε ὡμολόγηκας δι’ ἔνδειαν τῶν ἀγαθῶν καὶ καλῶν
ἐπιθυμεῖν αὐτῶν τούτων ὧν ἐνδεής ἐστιν.—ὡμολόγηκα γάρ.
But you have admitted that Love, from lack of good and beautiful
things, desires these very things that he lacks.—Yes, I have. (Pl. Smp.
202d1–4)

(22) καὶ γυμναστικὴ μέν που περὶ γιγνόμενον καὶ ἀπολλύμενον

τετεύτακεν· σώματος γὰρ αὔξης καὶ φθίσεως ἐπιστατεῖ.—φαίνεται.—
τοῦτο μὲν δὴ οὐκ ἂν εἴη ὃ ζητοῦμεν μάθημα.—οὐ γάρ.
And gymnastics, I take it, is devoted to that which grows and per-
ishes; for it presides over the growth and decay of the body.—Appar-
ently.—Then this cannot be the study that we seek.—No. (Pl. R.
521e3–522a1)
 the curious combination of the particles ΓAΡ and ΟYΝ 55

In addition to the fact that reactions with γάρ alone may contain new
information, whereas reactions with γὰρ οὖν never do so, there is a
second tendency. A reaction with οὖν is especially frequent in situations
in which the conversation partner can expect a reaction of approval,
because the person who utters the reaction has already been very posi-
tive about the preceding steps in the argumentation and often has even
explicitly agreed with those steps. In example (23), for instance, we
find a reaction of approval which is definitely expected in view of the
preceding discussion.14 After all his positive reactions to the preceding
steps in the argumentation, the stranger may expect Theaetetus to agree
with the last statement:
(23) ΞΕ. πρῶτον μὲν κίνησιν, ὡς ἔστι παντάπασιν ἕτερον στάσεως. ἢ
πῶς λέγομεν;
ΘΕΑΙ. οὕτως.
ΞΕ. οὐ στάσις ἄρ’ ἐστίν.
ΘΕΑΙ. οὐδαμῶς.
ΞΕ. ἔστι δέ γε διὰ τὸ μετέχειν τοῦ ὄντος.
ΘΕΑΙ. ἔστιν.
ΞΕ. αὖθις δὴ πάλιν ἡ κίνησις ἕτερον ταὐτοῦ ἐστιν.
ΘΕΑΙ. σχεδόν.
ΞΕ. οὐ ταὐτὸν ἄρα ἐστίν.
ΘΕΑΙ. οὐ γὰρ οὖν.
Str. Take motion first; we say that it is entirely other than rest, do
we not?
Theae. We do.
Str. Then it is not rest?
Theae. Not at all.
Str. But it exists, by reason of its participation in being.
Theae. Yes, it exists.
Str. Now motion again is other than the same.
Theae. You’re about right.
Str. Therefore it is not the same.
Theae. Indeed, it is not. (Pl. Sph. 255e11–256a6)
Reactions with γάρ alone, by contrast, are more frequent in situations
where a reaction of approval is unexpected. For instance, reactions
with γάρ occur in cases where the person responding has been very
reserved in his preceding reactions (see examples 22 and 24), where
a reaction of approval would place him in an uncomfortable position

14
Example (18) is another clear example of a reaction of approval that is to be
expected on the basis of the preceding discussion.
56 stéphanie j. bakker

(see examples 21 and 25), or where the question is formulated very

cautiously (see example 26):
(24) ἔοικεν ἄρα, ἦν δ’ ἐγώ, ὁ μὲν δίκαιος τῷ σοφῷ καὶ ἀγαθῷ, ὁ δὲ ἄδικος
τῷ κακῷ καὶ ἀμαθεῖ.—κινδυνεύει.—ἀλλὰ μὴν ὡμολογοῦμεν, ᾧ γε
ὅμοιος ἑκάτερος εἴη, τοιοῦτον καὶ ἑκάτερον εἶναι.—ὡμολογοῦμεν
γάρ.
Then the just man is like the wise and good, and the unjust is like
the bad and the ignoramus.—It seems likely.—But furthermore we
agreed that such is each as that to which he is like.—Yes, we did.
(Pl. R. 350c4–9)

(25) ἐὰν ἡ ἐμή, ἔφη, νικᾷ, τὸν τοῦ ἐπιεικοῦς μιμητὴν ἄκρατον.—ἀλλὰ μήν,
ὦ Ἀδείμαντε, ἡδύς γε καὶ ὁ κεκραμένος, πολὺ δὲ ἥδιστος παισί τε καὶ
παιδαγωγοῖς ὁ ἐναντίος οὗ σὺ αἱρῇ καὶ τῷ πλείστῳ ὄχλῳ.—ἥδιστος
γάρ.
If my vote prevails, he said, the unmixed imitator of the good.—Nay,
but the mixed type also is pleasing, Adeimantus, and far most pleas-
ing to boys and their tutors and the great mob is the opposite of
your choice.—Most pleasing it is. (Pl. R. 397d4–9)

(26) οἶμαι γάρ σε οὑτωσί πως εἰπεῖν, ὅτι τοῖς θεοῖς κατεσκευάσθη τὰ
πράγματα δι’ ἔρωτα καλῶν· αἰσχρῶν γὰρ οὐκ εἴη ἔρως. οὐχ οὑτωσί
πως ἔλεγες;—εἶπον γάρ, φάναι τὸν Ἀγάθωνα.
What you said, I believe, was to the effect that the gods contrived the
world from a love of beautiful things, for of ugly there was no love.
Did you not say something of the sort?—Yes, I did, said Agathon.
(Pl. Smp. 201a3–7)
In examples (22) and (24), a reaction of approval is not likely after
the preceding—very reserved—reaction with φαίνεται and κινδυνεύει
respectively. In this respect, the reaction in these examples clearly
differs from those in examples like (19) and (23), in which a reaction
of approval was highly probable, in view of the preceding discourse.
Example (25) is almost the direct opposite of example (23), in that
in view of the preceding discussion a reaction of approval is virtually
impossible, for approval would conflict with what the speaker previ-
ously expressed as his opinion. In example (26), it is clear from the
cautious formulation of the question with οἶμαι and πως that a reaction
of approval is desirable, but not necessarily expected.
The fact that a reaction introduced by γὰρ οὖν, unlike those intro-
duced by γάρ alone, is a) to be expected and b) never contains new
information makes it likely that the function of οὖν in reactions intro-
 the curious combination of the particles ΓAΡ and ΟYΝ 57

duced by γὰρ οὖν is comparable to that of οὖν in utterances in ongoing

discourse. Just as in ongoing discourse, οὖν in reactions may mark
that the approving content of the reaction is inferable, or at least to be
expected, from the preceding discussion or general knowledge.15 How-
ever, the problem is that the differences between reactions introduced
by γὰρ οὖν and reactions introduced by γάρ alone are merely tenden-
cies. Although reactions with γὰρ οὖν never contain new information
and are always to be expected, reactions with γάρ alone do not always
contain new information and may also be expected. It is therefore ques-
tionable whether there are any conclusions to be drawn with respect
to the many examples in which the difference between reactions with
and without οὖν is unclear, on the basis of the examples in which we
do find a clear difference. In example (27), for instance, the reaction of
approval introduced by γάρ seems just as inferable from the preceding
discussion16 as the reaction in example (28) introduced by γὰρ οὖν is
inferable from general knowledge:
(27) οὐκοῦν τὴν πρὸς τοὺς Ἕλληνας διαφοράν, ὡς οἰκείους, στάσιν
ἡγήσονται καὶ οὐδὲ ὀνομάσουσιν πόλεμον;—οὐ γάρ.
Will they not then regard any difference with Greeks who are their
own people as a form of faction and refuse even to speak of it as
war?—Indeed. (Pl. R. 471a1–3)

(28) ἆρ’ οὖν ὧδε πέφυκεν ὄψις πρὸς τοῦτον τὸν θεόν;—πῶς;—οὐκ
ἔστιν ἥλιος ἡ ὄψις οὔτε αὐτὴ οὔτ’ ἐν ᾧ ἐγγίγνεται, ὃ δὴ καλοῦμεν
ὄμμα.—οὐ γὰρ οὖν.
Is not this, then, the relation of vision to that divinity?—What?—Nei-
ther vision itself nor its vehicle, which we call the eye, is identical
with the sun.—No, it isn’t. (Pl. R. 508a9–b2)

15
Οὖν is not only used in reactions introduced by γάρ, but also appears in other
kinds of reactions. In the Cratylus, for instance, there are 16 examples of πάνυ μὲν οὖν
(‘certainly’) and 9 of τί οὖν; (‘yes, so?’). In these reactions, too, the use of οὖν seems
to indicate that the reaction can be inferred from, or is to be expected on the basis of,
the preceding discussion or the general knowledge of the conversation partners. For
clear examples of this function of οὖν in πάνυ μὲν οὖν and τί οὖν reactions, see Cra.
400b4, 407a3 and 398a3 respectively.
16
At any rate, the question to which οὐ γάρ is a reaction is marked as inferable from
the preceding discussion by the use of the inferential future. An inferential future is a
future that indicates that the content of the utterance at issue can be inferred from the
preceding discourse, from general knowledge or experience of the conversation partners
(e.g. ‘if a is b and b is c, then a will be c’). For details, see S. Bakker (2002).
58 stéphanie j. bakker

Of course, the fact that the reaction of approval in example (27) is

expected in view of the preceding discussion does not mean that it is
marked as such. Thus it is still possible that οὖν marks the reaction in
example (28) as expected on the basis of general knowledge, whereas
the reaction without οὖν in example (27) does not explicitly indicate
that approval is to be expected on the basis of the preceding discus-
sion. It is, however, impossible to conclude with certainty whether this
is indeed the case.17

6. Concluding remarks

In the preceding sections, I have argued that the function of γὰρ οὖν
in utterances introduced by these particles cannot be simply the sum of
the respective functions of γάρ and οὖν. Only in a very limited number
of examples did the utterance at issue prove to be formally marked as
subordinate to what went before and at the same time more to the
point with respect to its content (see examples 1 to 3). In many more
examples, especially parentheses and reactions to less important steps
in the argumentation, an interpretation of οὖν as a ‘pop’ particle has
proved somewhat problematic. Therefore, it seems very unlikely that
οὖν in the combination γὰρ οὖν marks the utterance as more to the
point than the preceding discourse.
However, this is not to say that οὖν in the combination γὰρ οὖν
does not contribute to discourse cohesion, for what the utterances
introduced by γὰρ οὖν discussed above have in common is that the
information provided is not new to the addressee. Especially in utter-
ances with γὰρ οὖν in ongoing discourse (without a turn-taking) it is
clear that the content of the utterance marked by γάρ as an explanation
or background information is to be expected, i.e., it is either already
familiar or inferable from the preceding discussion or general knowl-
edge shared by the conversation partners. Although the overall picture
provided by reactions introduced by γὰρ οὖν is less clear, it can be
argued that οὖν marks the fact that the reaction of approval is to be

17
Although it cannot be proved on the basis of my data that the addition of οὖν
to a reaction with γάρ indicates that the approval is to be expected on the basis of the
preceding discourse or general knowledge, it is clear that οὖν does not indicate that
the reaction should be interpreted as the next relevant step in the argumentation, as
maintained in section 2.
 the curious combination of the particles ΓAΡ and ΟYΝ 59

expected in these instances as well. Therefore, οὖν does contribute to

discourse cohesion, but not in the same way that γάρ does. Whereas
γάρ has a presentational or text-organizing function, οὖν plays a more
interactional role, as it provides some indication of the accessibility
of the information.18
The problem in assuming that οὖν in utterances with γὰρ οὖν marks
the information as to be expected, so that οὖν’s function is interactional
rather than presentational, is that the function of οὖν in these utterances
would not be in line with its ‘common’ function. At any rate, not in line
with what is generally assumed to be the common function of οὖν, for
even if οὖν occurs alone, in many cases the audience has somewhat been
prepared for the information provided by the utterance.19 For example,
a conclusion, one of the typical environments for οὖν,20 tends not to
contain new information, but rather to combine information from the
preceding discourse and/or general knowledge, as in:
(29) ‘μηχανὴ’ γάρ μοι δοκεῖ τοῦ ‘ἄνειν ἐπὶ πολὺ’ σημεῖον εἶναι· τὸ γὰρ
‘μῆκός’ πως τὸ πολὺ σημαίνει· ἐξ ἀμφοῖν οὖν τούτοιν σύγκειται,
‘μήκους’ τε καὶ τοῦ ‘ἄνειν’, τὸ ὄνομα ἡ ‘μηχανή’.
For I think μηχανή signifies ἄνειν ἐπὶ πολύ (much accomplishment);
for μῆκος (length) has about the same meaning as τὸ πολύ (much).
The name μηχανή is thus composed of these two, μῆκος and ἄνειν.
(Pl. Cra. 415a4–7)
Similarly, where οὖν is used to return to the main line of the argumen-
tation/narrative after a digression, the information in the utterance at
issue is generally not new either, since speakers tend to return to the

18
The fact that the particle οὖν in utterances introduced by γάρ has an interactional
function does not mean that it provides information on the attitude of the speaker with
regard to the truth of the utterance. In contrast to particles like δή and ἄρα, which
express a high and a low degree of commitment on the part of the speaker, respectively
(see Sicking and Van Ophuijsen 1993: 83), οὖν is neutral in this respect. It is clear that
οὖν is neutral from—among other things—the fact that οὖν in utterances introduced by
γάρ is frequently accompanied by the particle δή, indicating that the speaker commits
himself to the content of the utterance (and expects his addressee to do the same), see
e.g. Pl. Plt. 270b5, R. 615c4 and Ti. 84e10.
19
This has been noted by Van Ophuijsen (1993), see section 3. Cf. also Wakker’s
(elsewhere in this volume) observations on the difference between οὖν and τοίνυν.
Whereas οὖν marks a relevant step that follows on the previous one(s), τοίνυν marks the
next important step as lying outside or beyond the expectations of the addressee(s).
20
For the various uses of οὖν, see Van Ophuijsen (1993) and Rijksbaron et al. (2000),
Wakker (elsewhere in this volume).
60 stéphanie j. bakker

main line by repeating familiar information. In the following example

this is clear from the addition of ὅπερ εἶπον:
(30) ὥστε οὐδ’ εἴ με νῦν ὑμεῖς ἀφίετε Ἀνύτῳ ἀπιστήσαντες, ὃς ἔφη ἢ
τὴν ἀρχὴν οὐ δεῖν ἐμὲ δεῦρο εἰσελθεῖν ἤ, ἐπειδὴ εἰσῆλθον, οὐχ οἷόν
τ’ εἶναι τὸ μὴ ἀποκτεῖναί με, λέγων πρὸς ὑμᾶς ὡς εἰ διαφευξοίμην
ἤδη [ἂν] ὑμῶν οἱ ὑεῖς ἐπιτηδεύοντες ἃ Σωκράτης διδάσκει πάντες
παντάπασι διαφθαρήσονται,—εἴ μοι πρὸς ταῦτα εἴποιτε· (. . .)—εἰ
οὖν με, ὅπερ εἶπον, ἐπὶ τούτοις ἀφίοιτε, εἴποιμ’ ἂν ὑμῖν ὅτι (. . .).
Therefore, even if you acquit me now and are not convinced by
Anytus, who said that either I ought not to have been brought to
trial at all, or since I was brought to trial, I must certainly be put to
death, adding that if I were acquitted your sons would all be utterly
ruined by practicing what I teach—if you should say to me in reply
to this: (. . .); if you should let me go on this condition which I have
mentioned, I should say to you, (. . .). (Pl. Ap. 29b9–d2)
However, cases in which οὖν marks a new step in the argumentation
or narrative are less straightforward. Although there are instances in
which this new step is prepared for by what has gone before (see example
31), there are also many examples in which the information provided
is definitely not predictable or expected (see example 32):
(31) (Polemarchos, who meets Socrates and Glaucon in the city,
announces there will be a night festival which will be worth seeing
and invites Glaukon and Sokrates to stay).
καὶ ὁ Γλαύκων, ἔοικεν, ἔφη, μενετέον εἶναι. ἀλλ’ εἰ δοκεῖ, ἦν δ’ ἐγώ,
οὕτω χρὴ ποιεῖν. ἦιμεν οὖν οἴκαδε εἰς τοῦ Πολεμάρχου, καὶ Λυσίαν
τε αὐτόθι κατελάβομεν καὶ Εὐθύδημον, (. . .).
‘It looks as if we should have to stay,’ said Glaucon. ‘Well,’ said I,
‘if it so be, so be it.’ So we went with them to Polemarchus’ house,
and there we found Lysias and Euthydemus, the brothers of Pole-
marchus, (. . .). (Pl. R. 328b2–5)

(32) (. . .) προσευξάμενοι δὲ καὶ θεωρήσαντες ἀπῇμεν πρὸς τὸ ἄστυ.

κατιδὼν οὖν πόρρωθεν ἡμᾶς οἴκαδε ὡρμημένους Πολέμαρχος ὁ
Κεφάλου ἐκέλευσε δραμόντα τὸν παῖδα περιμεῖναί ἑ κελεῦσαι.
(. . .) after we had said our prayers and seen the spectacle we were
starting for town. Polemarchus, the son of Cephalus, caught sight
of us from a distance as we were hastening homeward and ordered
his boy run and bid us to wait for him. (Pl. R. 327b1–4)21

21
Of course, the information ἀπῇμεν πρὸς τὸ ἄστυ raises the question ‘what hap-
pened next?’, so that the addressee may expect that something will happen next.
 the curious combination of the particles ΓAΡ and ΟYΝ 61

It is therefore unclear whether the general function of οὖν can and

should be described in terms of a high degree of predictability of the
utterance.
But perhaps we should not strive to determine a single general func-
tion that can account for the use of οὖν in all examples. For the Dutch
dus (‘so’), for instance, there is no such general function either. On the
basis of a diachronic analysis, Evers-Vermeul (2005) concludes that dus
today may be used both as a connector and as a discourse marker. In
its role as connector dus marks a conclusion, in its role as discourse
marker, by contrast, it provides information on the accessibility of the
information: it marks the information as already familiar. These two
functions may overlap when dus signals that the conclusion is already
accessible to the addressee, since the information was provided earlier
in the discourse.
Whether οὖν functions similarly in Ancient Greek, meaning that
we must distinguish two functions of οὖν (one more interactional
function, where οὖν marks the information as to be expected, and one
more presentational function, where οὖν marks the information as
more to the point), which can occur separately, but can also partially
overlap, cannot be answered on the basis of my analysis of utterances
introduced by γὰρ οὖν. However, my data do make it clear that the
function attributed to οὖν in recent particle studies, i.e., the marking
of a new step in the discourse, one which is more to the point than
what precedes, does not hold true for a substantial proportion of the
utterances marked by οὖν.

However, he has no clue yet as to the exact nature of this step. In this sense, the infor-
mation provided by the utterance marked by οὖν is not predictable or expected.
 CHAPTER FOUR

‘WELL I WILL NOW PRESENT MY ARGUMENTS’.

DISCOURSE COHESION MARKED BY ΟYN
AND ΤΟINYN IN LYSIAS

Gerry C. Wakker

1. Introduction1

Nowadays it is generally agreed that (connective) particles play an

important role in marking the transition from one discourse unit to
another and in signalling hierarchical relations between the discourse
units in question. Thus, in the speeches of Lysias we often find at the
end of the proem an announcement of the speaker that he is going to
tell what has happened (in the narrative section of the speech). In the
sentence announcing the narrative, various connective particles are
found, e.g. τοίνυν (example 1), οὖν (example 2), and δέ (example 3).
The first sentences of the narratives in Lysias appear to be nearly
always marked by γάρ2 (examples 1 and 2), but sometimes we find an
asyndeton (example 3).
(1) ἐγὼ τοίνυν ἐξ ἀρχῆς ὑμῖν ἅπαντα ἐπιδείξω τὰ ἐμαυτοῦ πράγματα,
οὐδὲν παραλείπων, ἀλλὰ λέγων τἀληθῆ· ταύτην γὰρ ἐμαυτῷ μόνην
ἡγοῦμαι σωτηρίαν, ἐὰν ὑμῖν εἰπεῖν ἅπαντα δυνηθῶ τὰ πεπραγμένα.
ἐγὼ γάρ, ὦ Ἀθηναῖοι, ἐπειδὴ ἔδοξέ μοι γῆμαι καὶ γυναῖκα ἠγαγόμην
εἰς τὴν οἰκίαν, τὸν μὲν ἄλλον χρόνον οὕτω διεκείμην ὥστε (. . .).

1
My thanks are due, first, to the participants of the 6th International Colloquium on
Ancient Greek Linguistics (Groningen, June 2007), and especially to Stéphanie Bakker,
for their remarks on my concept paper and, second, to Monique Swennenhuis, who
corrected my English.
2
The occurrence of γάρ in these contexts may be explained as follows: γάρ being
explanatory, i.e. providing answers to possible questions raised by the speaker’s utter-
ances, may be said to explicitly answer the question how the speaker will make good
the promises or announcements he has made in the proem of his speech. In example
(1) γάρ introduces an embedded narrative consisting of a rather large text unit (Lys.
1.6–27). See e.g. Sicking & van Ophuijsen (1993: 23); De Jong (1997: 182).
64 gerry c. wakker

Well, I shall therefore set forth to you the whole of my story from the
beginning; I shall omit nothing, but will tell the truth. For I consider
that my own sole deliverance rests on my telling you, if I am able, the
whole of what has occurred. When I, Athenians, decided to marry,
and brought a wife into my house, for some time I was in such a
way that (. . .). (Lys. 1.5–6)3

(2) Many people are surprised because of my accusing the corn-deal-

ers, and they tell me that they consider those who accuse them as
slanderers.
ὅθεν οὖν ἠνάγκασμαι κατηγορεῖν αὐτῶν, περὶ τούτων πρώτον εἰπεῖν
βούλομαι. ἐπειδὴ γὰρ οἱ πρυτάνεις ἀπέδοσαν εἰς τὴν βουλὴν περὶ
αὐτῶν, (. . .).
I therefore propose to speak first of the grounds on which I have
found it necessary to accuse them. When the Committee of the time
brought up their case before the Council, (. . .). (Lys. 22.1–2)

(3) ὅμως δὲ πειράσομαι ὑμᾶς ἐξ ἀρχῆς ὡς ἂν δύνωμαι δι’ ἐλαχίστων

διδάξαι.
οὑμὸς πατὴρ Κέφαλος ἐπείσθη (. . .).
Nevertheless I will try to inform you of the matter from the begin-
ning, as briefly as I can. My father Cephalus was induced (. . .). (Lys.
12.3–4)
In this article I want to study in more detail the function of the particles
οὖν and τοίνυν, which are often used in seemingly similar contexts. I
will do so in one homogeneous corpus, the forensic speeches of Lysias,
hoping that a better understanding of the ways in which one author
deals with these particles and, consequently, with discourse cohesion,
may further our knowledge about discourse cohesion in general. The
main questions I want to investigate are the following:

• What exactly is conveyed by the particles in question? What is the

exact contribution of these particles to the discourse cohesion?
• Given the fact that both connectors in the traditional syntactic sense
(i.e. coordinating conjunctions such as δέ) are found in between
discourse units, as well as connective particles (οὖν and τοίνυν), one

3
Unless stated otherwise my translations are borrowed from the translation by
Lamb in the Loeb series, sometimes slightly adapted. The Greek text used is the text
of the OCT-series edited by Hude (1912).
 discourse cohesion marked by OYN and ΤΟINYN 65

may wonder what this implies as to the syntactic status of the latter
group4 and as to the definition of asyndeton.

After a brief survey of the recent relevant literature (section 2) I will

study in more detail the particles οὖν and τοίνυν in Lysias (section
3) to see in which respect the two particles differ from each other. At
the same time I hope to make a first start in answering the question
in which respect examples with these particles differ from examples of
asyndeton and δέ in comparable contexts. This last part of the research
is only based on a limited number of examples and certainly needs
further research. In section 4 I will present my conclusions.

2. State of research

2.1. The framework

Particles traditionally belong to a category of linguistic elements that
is very difficult to describe: particles usually fall outside the syntactic
structure of the clause in which they occur, and their meaning is elusive.
In a strictly semantic approach it is difficult to account for the diversity
of—seemingly unconnected—uses of most particles. With the relatively
recent development of functionally oriented linguistic theories a new
(pragmatic) approach presented itself and has proven to be extremely
rewarding in the field of particle research.5 Within this framework,
roughly speaking, two types of meaning are distinguished: referential
meaning (i.e. the meaning that contributes to the representation of an
event, a situation, an action etc.), and functional meaning, which has to
do with the placing of the state of affairs described in the communicative
(textual and non-textual) context.6 Particles clearly have the second type
of meaning: they primarily have a pragmatic (rather than a grammatical
or strictly semantic, i.e. referential) function. From the point of view
of the addressee, particles may be considered a kind of road signs in

4
Cf. the introduction in Rijksbaron (ed.) (1997: 2, 13).
5
For the necessity of a pragmatic approach of particles see e.g. Levinson (1983:
100), Abraham (1986: 87–100), Kroon (1992: 53–8; 1994; 1995: 34–57). As to Ancient
Greek, pragmatic descriptions of particles may be found in Bakker (1986, 1988, 1993),
Sicking (1986), Sicking & van Ophuijsen (1993), Wakker (1994: 301–64, 1995) and
Rijksbaron (ed.) (1997).
6
See, for instance, Kroon (1992: 55–6; 1995: 41, 61–2).
66 gerry c. wakker

the text which help him keep track of the structure of the text or find
out the communicative purpose or expectations of the speaker. From
the perspective of the speaker, particles may be described as a means
of placing the unit they have in their scope into a wider perspective,
which may be the surrounding context (and its implications) or the
interactional situation the text forms part of.
Central to a pragmatic approach are two further distinctions: first,
every discourse can be analysed at at least three levels: the represen-
tational, the presentational or text structuring and the interactional
level.7 At each level particles may occur. Second, discourse is usually
structured hierarchically. One may discern embedding sequences and
embedded sequences. The starting point of an embedded sequence (also
called PUSH) and the point where a speaker returns to the embedding
sequence (POP) are very often marked by particles and other relators.
In Greek, as Slings8 argues, the most typical PUSH particle is γάρ (see
example 1 above), the most typical POP particle is οὖν.

2.2. Literature on οὖν and τοίνυν

Scholars working within the pragmatic framework sketched above usu-
ally give the following rough descriptions of the particles studied here:9
οὖν is a text structuring and connective particle marking a new, relevant
step (that is in some way related to the previous one) in a narrative or
argumentative context. The relationship may vary dependent on the
context. It may simply indicate a next event in the (continuing) narra-
tive, a conclusion, or it may resume the main line of thought after some
digression. τοίνυν is described both as a modal (τοι) and a connective
(νυν) particle: ‘I come to the point now (νυν) and I claim something
you must take notice of/you must note’ (τοι). Both οὖν and τοίνυν
are often used where δέ and asyndeton are also found. As to δέ, this
coordinating particle is said to mark discontinuity and to introduce a
new discourse unit that, together with the previous one(s), forms part
of a larger argumentative or narrative whole (see Bakker (1993) and
Sicking & van Ophuijsen (1993: 10–3)). Asyndeton (the lack of an overt

7
These distinctions are based on Kroon (1995). For the three levels and Greek
particles see Wakker (1997: 211–3), who mentions further literature.
8
Slings (1997a), who mentions further literature.
9
Sicking & van Ophuijsen (1993), Rijksbaron (ed.) (1997), Rijksbaron et al. (2000:
ch. 17) and Crespo (2007).
 discourse cohesion marked by OYN and ΤΟINYN 67

expression of connection in between sentences) is as yet not studied in

detail in Ancient Greek. It is said that asyndeton is sometimes found
when the connection is obvious (after an announcement has been made
and when certain backwards pointing pronouns or demonstratives are
present);10 in a series of coordinated items asyndeton is also found,
where it is often explained as being used for stylistic reasons (cf. Den-
niston (1954: xliii–xlv)).
I will try to show with detailed observations on οὖν and τοίνυν that
the above descriptions apply to most cases in Lysias (section 3). Even
then, however, it is not at once clear which factors determine the actual
choice of the particles.

3. οὖν and τοίνυν in Lysias

3.1. οὖν in Lysias

Most examples of οὖν neatly illustrate the above description. In Lysias
there are 222 examples of οὖν. When οὖν is found alone (without other
particles), it very often marks the next important step in the ongoing
narrative or argumentation. The nature of the relationship with the
preceeding context may vary. Thus, after an introductory statement
containing a subsidiary explanation introduced by γάρ, οὖν may mark
the return to the main line of the story and introduce the next relevant
step, e.g.:
(4) They said that they needed money; their enemies were rich and against
the government. They easily persuaded their hearers, for (γάρ) killing
meant nothing to them, but money everything.
ἔδοξεν οὖν αὐτοῖς δέκα συλλαβεῖν, τούτων δὲ δύο πένητας, (. . .).
So they resolved to seize ten of them, of whom two should be poor
men (Lys.12.7)11
Here the οὖν-unit contains the consequence and the next step in the
story after the statement (and its explanation) that the hearers were

10
E.g. ταῦτα εἰποῦσα (Lys. 1.17); οὕτως (Lys. 1.2, 27, 32), ὥστε (Lys. 3.48), a new
vocative address after an interruption (Lys. 1.29, 30, 32). See Sicking & van Ophuijsen
(1993: 40, 44–5).
11
Such οὖν-units may have either the form of a statement (e.g. Lys. 7.23; 9.15; 12.64;
16.9, etc.) or of a rhetorical question (e.g. Lys. 1.45; 8.7; 10.13; 12.88; 24.33, etc.).
68 gerry c. wakker

easily persuaded.12 That they will resolve to kill people is a—at this point
not unexpected—consequence; the new, important information is the
fact that they resolve to seize ten men (two of whom are to be poor).
The subsidiary digression is not always introduced by γάρ, cf. (5),
where the discourse unit with Ἁγνόδωρος δ’ ἦν gives background
information (cf. the use of the imperfect) about the newly introduced
person Hagnodorus. With οὗτος οὖν (referring to Hagnodorus) the
main line of the narrative is resumed by telling the next important step
in the narrative, cf. the use of the historic present παράγει.
(5) I am told that Agoratus attributes these depositions in part to Menes-
tratus.
ὁ Μενέστρατος οὗτος ἀπεγράφη ὑπὸ τοῦ Ἀγοράτου καὶ συλληφθεὶς
ἐδέδετο· Ἁγνόδωρος δ’ ἦν Ἀμφιτροπαιεύς, δημότης τοῦ Μενεστράτου,
Κριτίου κηδεστὴς τοῦ τῶν τριάκοντα. οὗτος οὖν, ὅτε ἡ ἐκκλησία
[ἐν] Μουνιχίασιν ἐν τῷ θεάτρῳ ἐγίγνετο, ἅμα μὲν βουλόμενος τὸν
Μενέστρατον σωθῆναι, ἅμα δὲ ὡς πλείστους ἀπογραφέντας ἀπολέσθαι,
παράγει αὐτὸν εἰς τὸν δῆμον, καὶ εὑρίσκονται αὐτῷ (. . .) ἄδειαν.
That Menestratos was informed against by Agoratus, and was arrested
and put in prison. Hagnodorus of Amphitrope, a fellow-townsman
of Menestratus, was a kinsman of Critias, one of the Thirty. Well
(οὗτος οὖν (. . .)), when the Assembly was being held in the theatre
at Munichia, this man, with the double aim of saving the life of
Menestratus and of causing, by means of depositions, the destruction
of as many people as possible, brought him (παράγει αὐτὸν) before
the people, that gave him impunity. (Lys. 13.55)13
The next step may also consist of an urgent appeal to the jury. This
may occur anywhere within the speech (6) or as the final appeal in the
epilogue of the speech (7):
(6) You know how the walls were demolished, you lost your private pos-
sessions, and were all expelled. And you, Agoratus, accused them of
intriguing against our democracy.
νῦν οὖν μνησθέντες καὶ τῶν ἰδίων ἕκαστος δυστυχημάτων καὶ τῶν
κοινῶν τῆς πόλεως τιμωρεῖσθε τὸν αἴτιον τούτων.

12
Comparable are the examples where after a presequence the core of the message
is brought home by a discourse unit marked by οὖν (see, for instance, 1.16; 12.14) and
examples where οὖν marks alternating subjects and marks the most relevant perspective
in the eyes of the speaker, cf. 12.9, where the speaker marks the introduction of his
own words (and not that of the interlocutor) with οὖν, cf. Sicking & van Ophuijsen
(1993: 26).
13
For the use of the historic present see Sicking & Stork (1997).
 discourse cohesion marked by OYN and ΤΟINYN 69

So now let each of you remember the misfortunes caused both to

individuals and to the common weal of the city, and take vengeance
on their author. (Lys. 13.48, end of the narratio)14

(7) ὑμεῖς οὖν καὶ τὰ εἰρημένα καὶ τὰ παραλελειμμένα ἀναλογισάμενοι

πολὺ μᾶλλον αὐτοῦ καταψηφίσασθε, ἐνθυμηθέντες ὅτι (. . .).
You have therefore to consider what has been told the tale of and
what has been omitted, and to be all the more for condemning him,
reflecting that (. . .). (Lys. 14.47)
We also find οὖν on its own at the transition to a new part of speech, e.g.
to prepare the testimony of witnesses or the reading aloud of a law, cf.
(8) ὡς οὖν καὶ ταῦτ’ ἀληθῆ λέγω, τούτων ὑμῖν τοὺς παραγενομένους
μάρτυρας παρέξομαι.
Now, to show that here too I am speaking the truth, I will produce
to you as witnesses the persons who were present on the occasion.
(Lys. 3.20)
Only in the rare combination of δ’ οὖν we find a real coordinating
conjunction combined with οὖν. All 3 cases15 occur after a digression
or speculation of which the truth is as yet not certain and they resume
the main line of thought. Leaving the truth of the previous words out of
consideration, δέ here marks the beginning of a new section, whereas
οὖν marks the return to the main line of thought, and the transition to
the next important point. The combination may very well be translated
by: ‘anyhow/in any case the next important point to tell is’, cf. (9):
(9) Andocides committed himself to prison, having assessed the penalty
at imprisonment if he failed to hand over his attendant. He knew,
however, that this man was dead. Must it not have been some god
that destroyed his reason?
ἐκ δ’ οὖν τούτου τοῦ τιμήματος ἐδέδετο ἐγγὺς ἐνιαύτον.
Anyhow, as the result of this proposal he lay for nearly a year in
prison. (Lys. 6.23)
All the above examples illustrate that the term ‘POP particle’ seems
justified when οὖν resumes the main line after a presequence or a
digression or comes up with some logically following conclusion and/or
the next relevant step that follows out of the previous one(s). In some

14
Other examples are, for instance, 11.7; 15.11; 16.9; 17.1.
15
The examples are Lys. 6.23; 9.11; 20.26.
70 gerry c. wakker

way or another this description applies to all cases mentioned above.

One could say that, as a side effect, in many cases οὖν marks as it were
a difference in what may be called ‘status’ in terms of ‘relevance’: the
speaker marks what precedes as relevant, and for the present purpose
subsidiary to or preparatory for what follows and what is introduced
by οὖν (cf. Sicking & van Ophuijsen 1993: 27).
Οὖν is further found in the combination μὲν οὖν (. . .) δέ. In this
combination the μὲν οὖν-clause often rounds off the preceding discourse
unit by summarizing it, whereas the δέ-clause marks a new section
(opening it, by a balancing or contrasting new point, new argument
or even a new part of speech), cf. (10):
(10) περὶ μὲν οὖν τούτων τοσαῦτά μοι εἰρήσθω· ὑπὲρ ὧν δέ μοι προσήκει
λέγειν, ὡς ἂν οἷόν τε διὰ βραχυτάτων ἐρῶ.
Well, in regard to those matters, let these few words of mine suffice;
I will now speak as briefly as I can on the points with which I am
here concerned. (Lys. 24.4, transition from proem to narrative)16
Instead of offering a summary of the previous text the μὲν οὖν-clause
itself may also contain new information, balanced by the information
of the δέ-clause:
(11) I have to defend my father and myself as best as I can.
τὴν μὲν οὖν παρασκευὴν καὶ <τὴν> προθυμίαν τῶν ἐχθρῶν ὁρᾶτε
(. . .)· τὴν δ’ ἐμὴν ἀπειρίαν πάντες ἴσασιν, ὅσοι ἐμὲ γιγνώσκουσιν.
αἰτήσομαι οὖν ὑμᾶς (. . .).
You see, of course, the artifice and the alacrity of my enemies;
whereas everone who knows me is aware of my inexperience. I shall
therefore beg of you . . . (the just and easy favour of hearing us with
the same absence of anger as when you listened to our accusers).
(Lys. 19.2)17
Sometimes, however, the expected δέ-clause is omitted as in 1.28
(πρῶτον μὲν οὖν ἀνάγνωθι τὸν νόμον ‘first read the law now’).18 As

16
Some other examples 10.5; 14.3 (transition from proem to narrative); 17.2; 19.11,
23.10, 11; 32.18 (transition from narrative to proofs), 1.47 (transition from proofs to
epilogue).
17
Cf. Lys. 1.9 (see also the comments by Sicking & van Ophuijsen 1993: 27–8); 3.2.
18
Cf. Lys. 8.19; 25.8; 26.21; In 13.51 the same holds: after the reading aloud of the
decrees, saying that Agoratus has denounced others, this point is at once resumed (ὡς
μὲν οὖν οὐκ ἀπέγραψεν) and brought further to the next relevant statement (he can find
no means of showing it). Instead of the expected balancing δέ the text continues with
‘he must therefore (δεῖ τοίνυν) prove that he was justified in giving that information,
 discourse cohesion marked by OYN and ΤΟINYN 71

Sicking (Sicking & van Ophuijsen 1993: 28) convincingly argues, the
effect of the use of μὲν οὖν without a balancing δέ-clause is that the
speaker appears at once to conclude what precedes and to proceed to
what he wants to say next: ‘this said I now come to what is important
now (. . .)’. It thus seems to perform at once the same function as the
two steps in Lys. 12.47: ‘so much then, I would say in regard to them
(πρὸς μὲν οὖν τούτους (. . .)). Now call my witnesses (τοὺς δὲ μάρτυράς
μοι κάλει).’19
One may ask whether οὖν in the combination μὲν οὖν (. . .) δέ is still
functioning as a POP particle. I would say it is. I would argue that οὖν
has its usual function, resuming the main line of thought and bringing it
to a new important point. While in an οὖν-clause this happens straight
away, in a μὲν οὖν (. . .) δέ-clause the preceding is first summarized (in
the μὲν-clause) and then the next point is introduced (δέ-clause), or
both clauses contain new information, the δέ-clause containing new
balancing information. The scope of οὖν, then, is the combination of
both clauses: οὖν (μέν—δέ). From the perspective of the organization
of the text, μέν—δέ thus functions at one level below that of οὖν (and
as I will show below, of τοίνυν).
To prepare my section 3.4, I may already point out that, apart from
the combination δ’ οὖν, formally we would have to classify all examples
of οὖν as asyndeton. As indicated in section 2.2 (notably note 10),
often other backwards pointing linguistic elements are present and
help to clarify the discourse coherence (e.g. demonstrative pronouns
or demonstratives, cf. examples (5), (7), (8) and (10)).
Up till now the observations on οὖν sufficiently illustrate the rough
sketch given above (section 2.2). Even the seemingly difficult examples
mentioned in notes 18 and 19, may be explained along the same
lines.20

but I think (οἴομαι δέ) that he will not attempt to show this either.’ Perhaps τοίνυν is
chosen because it is this point that is as controversial as it is crucial for the defendant,
whereas δέ would simply mark a balancing statement.
19
The same holds good for Lys. 9.15; 22.3; 29.13–4, where after a μὲν οὖν-clause
(after an intervening γάρ-explanation or an absolute genitive-construction) a fully
developed combination of μὲν οὖν . . . δέ or οὖν alone picks up the main line of thought
and introduces the next relevant point.
20
There is one example of οὖν that falls outside the framework of my study; it does
not introduce a new discourse unit, but occurs within a generalizing relative clause
(but note that it is omitted in manuscript P): ‘The law orders when someone detects an
adulterer, to deal with him in whatever way he pleases (ὅ τι ἂν οὖν βούληται χρῆσθαι)’
(Lys. 1.49). In my opinion ‘in whatever way he pleases and thinks relevant’ might be a
72 gerry c. wakker

3.2. τοίνυν in Lysias

As to τοίνυν, this particle is usually described (see also section 2.2)
both as a modal (τοι) and a connective (νυν) particle: ‘I come to the
point now (νυν) and I claim something you must take notice of / you
must note (τοι)’. As compared to οὖν the speaker emphasizes that what
he will say is important for the addressee. It is especially this idea in
which τοίνυν differs from οὖν.21 The reasons why the speaker claims his
utterance to be important for the addressee may vary, as the examples
below will show.
It is time to look again at example (1), quite elaborately discussed in
Sicking & van Ophuijsen (1993: 28–31), where τοίνυν announces the
transition to the defendant’s narrative. Sicking argues that in speeches
of the accuser usually δέ is used when the transition to the narrative is
prepared, since an account of the events is expected by the audience, cf.
(3). In (1), however, it is the defendant’s speech. An elaborate account
of the events is not necessary, the accuser already having presented
his speech. In (1), however, the defendant wants the jury to hear a
significantly different version. In the proem he first sketches the bur-
den of his proof and the importance of his evidence for his case. The
defendant’s use of τοίνυν here seems motivated by the fact that the
narrative is presented as possibly not expected in this elaborate form,
but as crucial for his case. τοίνυν thus seems to mark a new section as
being highly important, but possibly outside or beyond the expecta-
tions of the audience.22 τοίνυν may, therefore, be characterized as a
modal POP particle.
The above description seems—in some way or another—to apply to all
132 examples of τοίνυν in Lysias. Just like οὖν we find τοίνυν alone or
in the combination μὲν τοίνυν (. . .) δέ (formally all cases of asyndeton).

good paraphrase, οὖν marking the relevance—in the perspective of the victim—of the
punishment chosen. Further research is however necessary. In this non-connective use
it seems difficult to speak of a POP particle. Perhaps in prototype theory one would
have to classify this use as a peripheral use and less prototypical, cf. Bakker (1988:
14–8). See also my section 4 (concluding remarks). See also S. Bakker (this volume)
for the discussion of a puzzling use of οὖν in the combination γὰρ οὖν.
21
Νυν alone seems often more or less equivalent to οὖν ‘I come to the point now’.
22
Compare also the description of τοίνυν in Plato (Sicking & van Ophuijsen 1993:
164): τοίνυν is employed ‘to correct or forestall a discrepancy between the pragmatical
information (. . .) of the speaker and that of the hearer.’. So here too, one can say that
τοίνυν marks that the speaker presents something that is outside the expectations of
the addressee.
 discourse cohesion marked by OYN and ΤΟINYN 73

Many examples resume the main line of the story or argumentation

after the recitation of evidence, laws, other documents or testimonies
of the witnesses. Compare, for instance, (12) with (13):
(12) Witnesses come forward who declare that the defendant tried to run
away, and that no body was hurt right there; the defendant continues
his own speech by saying:
ὅτι μὲν τοίνυν οὗτος ἦν ὁ ἀδικήσας, ὦ βουλή, καὶ ἐπιβουλεύσας ἡμῖν,
καὶ οὐκ ἐγὼ τούτῳ, ὑπὸ τῶν παραγενομένων μεμαρτύρηται ὑμῖν.
μετὰ δὲ ταῦτα τὸ μὲν μειράκιον εἰς γναφεῖον κατέφυγεν, οὗτοι δέ
(. . .) ἦγον αὐτὸν βίᾳ.
That this man, then, was the wrongdoer, gentlemen, and that he had
designs on us, and not I on him, has been testified to you by those
who were then present. After this the boy took refuge in a fuller’s
shop, but these men laid violent hands on him. (Lys. 3.15)

(13) Call the witnesses.—witnesses

τῶν μὲν οὖν μαρτύρων ἀκηκόατε: ἐνθυμεῖσθε δὲ ὅτι (. . .).
Well then, you have heard the witnesses; and now reflect that (. . .).
(Lys. 19.60, cf. 3.21)
Example (12) mentions a very important point, for the speaker has
just told the audience that he is accused of having injured Simon
(who is in love with the same boy as he is) on purpose; the defendant
now, unexpectedly, more or less reverses the accusation, as if he is the
accuser and not the defendant. This is underlined by the use of τοίνυν.
One might even say that, in this way, he manipulates his audience and
causes them to believe that in fact he is the victim, not the aggressor.
In the at first sight rather comparable example (13) the situation is less
dramatic, μὲν οὖν summarizing in a neutral way and δέ introducing
the next important point to consider.
The particle τοίνυν may have the same nuance without μέν:
(14) ἐπειδὴ τοίνυν ὁ χρόνος οὗτος ἐξήκει, αὐτὸς γεωργῶ.
Since the termination of that time, then, I have cultivated it myself.
(Lys. 7.11)
The speaker is accused of having cleared away an olive tree. Just before
the text of (14) the former renters and cultivators of the land have testified
that there was no olive or stump. This is precisely the key question,
for directly afterwards the speaker became the cultivator himself, as he
indicates in (14). If there was no olive, the accusation is groundless, for
74 gerry c. wakker

it is not possible that the speaker, as the one who became the cultivator
from that moment on, cleared away what was not there before. It is
thus the crucial point in his defence.
With a similar nuance we find examples of τοίνυν after the narrative
of an incident, resuming the main line of the story/argumentation, cf.
(15):
(15) ἐγώ τοίνυν, ὦ βουλή, ἡγούμενος μὲν δεινὰ πάσχειν, αἰσχυνόμενος δέ,
ὅπερ ἤδη καὶ πρότερον εἶπον, τῇ συμφορᾷ, ἠνειχόμην, καὶ (. . .).
So I, gentlemen, feeling myself grossly ill-used, but ashamed—as I
have already told you before—at my misfortune, put up with it (. . .).
(Lys. 3.9)
In the main narrative, the speaker/defendant states that Simon found out
where he and his guests were dining; Simon called them out of doors
and tried to hit the speaker, and to pelt him with stones. He missed
the defendant, but Aristocritus (one of Simon’s friends) was struck by
a stone which broke his forehead. The latter incident about Aristocri-
tus is of course a digression. The main line is now resumed with (15).
Far from fighting or aggressing the others, so the speaker says, he put
up with what had happened. Of course in his defence it is crucial to
bring home the point that he was not the aggressor, unexpected as this
point may be in view of the accusation uttered and in view of the fact
that the jury may be surprised that he did not act when he met such a
misfortune. His shame, however, prevented him from action.
The same holds good for the transition to (the introduction of) new
arguments, cf.
(16) ἔτι τοίνυν, ὦ βουλή, καὶ ἐκ τῆς μάχης τῆς γενομένης ῥᾴδιον γνῶναι
ὅτι ψεύδεται.
τὸ γὰρ μειράκιον (. . .) φεῦγον ᾤχετο, (. . .) ἐγὼ δὲ ἑτέραν ἀπελθὼν
ὁδὸν ᾠχόμην.
And besides, gentlemen, from the very fight that took place you
can easily perceive that he lies. For the boy (saw what was on hand,
flung off his cloak and) ran away. (these men pursuing him,) I took
myself off by another street. (Lys. 3.35)
The speaker states how he and the boy left the city for 4 years in order
to avoid problems and came back to Athens when he expected the
problems to be forgotten, but they unfortunately (not on purpose)
encountered Simon and a fight started (as the accuser claimed). He
now continues with the crucial and perhaps unexpected outcome of
 discourse cohesion marked by OYN and ΤΟINYN 75

the fight he has been accused of: there was no real fight, as he and
the boy ran away, each in another direction. This extra argument is
underlined by the use of ἔτι τοίνυν. It is crucial in his defence. There-
fore the jury must take notice of it, it is another proof of his not being
guilty.23
All 132 examples of τοίνυν may be explained along the same lines,
τοίνυν (as a modal POP particle) marking the transition to the next
important, for the speaker even crucial point, which is presented as
lying outside or beyond the expectations of the addressee(s).

3.3. Supportive arguments

There are several linguistic elements in the τοίνυν-clauses that reinforce
the idea of cruciality, importance or urgency, however unexpected this
may be for the addressee. Of course, these linguistic elements are not
exclusively found in τοίνυν-clauses, but they occur in a significantly
higher frequency in τοίνυν than in, for instance, οὖν-clauses. Moreover
the contexts often seem to be of a slightly different nature.

Table 1: Comparison between τοίνυν- and οὖν-clauses24

Linguistic element τοίνυν-clause οὖν-clauses
(ex. studied: 132 = 100%) (ex. studied 222 = 100%)
a. ἐγὼ (μέν) PART24 19 ex. = 14.4% 13 = 5.9%
consisting of: consisting of: consisting of:
—ἐγὼ μὲν PART (. . .) δέ 2 ex. 2 ex.
—ἐγὼ μὲν PART – 9 ex.
—ἐγὼ PART 17 ex. 2 ex.
b. οὗτος PART 12 = 9% 3 = 1.9 %
c. direct address 36 = 27 % 20 = 9 %

23
Comparable examples are, for instance, Lys. 25.15; 32.14.
24
PART is the abbreviation of particle; the particle is either τοίνυν or οὖν.
76 gerry c. wakker

a. ἐγὼ (μὲν) οὖν/τοίνυν (. . .) etc.

Often when the narrative/argumentation is resumed at a crucial point
for the speaker to prove that he is in the right, the story/argumenta-
tion is presented from the viewpoint of the speaker and ἐγὼ τοίνυν is
found, often also implying a topic switch, as is clearly shown by example
(15) above,25 where after the digression about Aristocritus the story
is resumed and there is a topic switch to the speaker himself. Often
such switches go along with a switch from general remarks to specific
ones (e.g. Lys. 7.27; 12.37; 25.12) or from speculative utterances to the
presentation of the relevant facts (e.g. Lys. 7.18; 7.37; 27.15).
Also ἐγὼ μὲν οὖν and ἐγὼ οὖν occur, but in rather different frequen-
cies as the above table shows: 11 of these 13 examples are cases of ἐγὼ
μὲν οὖν, two with balancing δέ,26 and 9 without a balancing δέ. The
verb is always a verb of opinion or of request; without balancing δέ
the speaker at once concludes what precedes and proceeds to his next
point.27 In three of these cases28 ἐγὼ μὲν οὖν occurs in the epilogue,
indicating that it is the next important step that follows from everything
said above. This seems more natural than indicating by τοίνυν that
what the speaker is going to say is crucial, but outside the expectations
of the audience. For this would amount to saying that the speaker has
not effectively delivered his speech!29
There are only two instances of ἐγὼ οὖν30 (also with a verb of opin-
ion), whereas ἐγὼ/ἐμοῦ etc. τοίνυν is found 17 times. Compare, for
instance, (17)—(18) as an illustration of the fact that the choice of the

25
Other examples are Lys. 1.5, 34 (ἐμοῦ τοίνυν; for these two examples see Sicking
& van Ophυijsen (1993: 28–9); 7.12,18 (ἐμοί), 27, 30, 37; 8.18; 10.15; 12.37; 14.22; 16.7
(ἐμὲ τοίνυν); 18.27 (ἡμεῖς); 25.11, 12 (ἐμοὶ τοίνυν); 27.15 (ἡμεῖς μὲν τοίνυν); 30.35
(ἡμεῖς μὲν τοίνυν); 31.7.
26
Lys.15.12; 28.16.
27
οὖν here marks the speaker’s request or his own opinion—while others may have
other opinions, but this is left implicit: Lys. 1.47; 2.81; 12.3; 14.46; 18.27; 19.45; 21.22;
29.14; 34.3. Cf. also note 21.
28
Lys. 1.47, 14.46, 18.27.
29
Crespo (2007) argues that with μὲν τοίνυν . . . δέ in 12.79 the transition to the
epilogue is made. I do not think this plausible, for as argued above, normally the jury
will expect the epilogue to come and the proofs to be sufficient to make this transi-
tion. With Sicking (Sicking & van Ophuijsen 1993: 30) I assume the epilogue to start
in 81 (κατήγορηται δὴ—such is the accusation against Eratosthenes) or, alternatively,
with Edwards (1999 ad loc.) in 92, where the appeal to the members of the jury really
starts.
30
Lys. 13.3; 32.9.
 discourse cohesion marked by OYN and ΤΟINYN 77

particle seems very well understandable, given the above descriptions

of their semantic value.
(17) ΝΟΜΟΣ (stating that throwing away one’s shield is an act of slander)
ἐγὼ τοίνυν, ὦ ἄνδρες δικασταί, ὑμᾶς μὲν πάντας εἰδέναι ἡγοῦμαι ὅτι
ἐγὼ μὲν ὀρθῶς λέγω, τοῦτον δὲ οὕτω σκαιὸν εἶναι ὥστε οὐ δύνασθαι
μαθεῖν τὰ λεγόμενα. βούλομαι οὖν αὐτὸν καὶ ἐξ ἑτέρων νόμων περὶ
τούτων διδάξαι.
Well, gentlemen, I think you have all perceived that my statement
is correct, whereas this man is so stupid that he cannot understand
a word that is said. So I would like to avail myself of some other
laws. (Lys. 10.15)31

(18) After a general statement (‘it is our duty (προσήκει) to avenge the
good and democratic men who were put to death by the thirty’) the
accuser explains that he therefore wants to avenge Dionysodorus,
his brother-in-law and a good democrat, who has been put to death
by the defendant Agoratus, who has committed an act of informing
against him. (a digression follows on the losses the speaker and the
city thus suffered).
ἐγὼ οὖν, ὦ ἄνδρες δικασταί, δίκαιον καὶ ὅσιον ἡγοῦμαι εἶναι καὶ
ἐμοὶ καὶ ὑμῖν ἅπασι τιμωρεῖσθαι καθ’ ὅσον ἕκαστος δύναται.
I therefore, gentlemen, consider it an act of justice and piety in all
of you as well as myself to take vengeance as far as each of us is
able. (Lys. 13.1–3).
In (17) the speaker accuses Theomnestus of the fact that he interprets
the law correctly when he wants to punish his enemies for slander, but
when he himself is slandering he claims to righteously escape punish-
ment. After the reading of the law in question, the accuser’s main point
is resumed, and a switch from the general law to his own opinion is
made: Theomnestus is too stupid to understand his words. This is a
crucial statement, presented as possibly unexpected for the addressee.
For this reason the speaker likes to continue (next important point, cf.
οὖν) by reading other laws. Whether or not the crucial statement is
really unexpected is not relevant, it is a question of presentation, giv-
ing the speaker the opportunity to support his statement by reading
other laws.

31
Cf. Lys. 12.37; 14.22; 25.11.
78 gerry c. wakker

In (18) the οὖν-statement implies a resumption and conclusion after

the digression on the losses they suffered, and the content is presented
as following from the above said. Note that προσήκει is explained,
amplified and resumed by δίκαιον καὶ ὅσιον in the οὖν-clause. This is,
of course, comprehensible and very apt in the speech of the accuser.

b. οὗτος οὖν/τοίνυν, ‘he’, the bad guy

Just as the perspective of the ‘I’ can be underlined by τοίνυν in crucial
and possibly unexpected utterances (or at least utterances that are pre-
sented as such), the perspective of the enemy and his unjustified actions
or accusations may be underlined in order to blacken his character.
Compare example (5) with
(19) Simon heard that the boy was at my house. He came here at night
and broke into the women’s rooms: within were my sister and my
nieces (absolute genitive), whose lifes have been so well ordered
that they are ashamed to be seen even by their kinsmen.
οὗτος τοίνυν εἰς τοῦτο ἦλθεν ὕβρεως ὥστ’ οὐ πρότερον ἠθέλησεν
ἀπελθεῖν πρὶν (. . .).
This man, then, carried insolence to such a pitch that he refused to
go away until [his own and other people drove him out by force].
(Lys. 3.7, cf. 31.12)
Here τοίνυν underlines the possible unbelief of the jury. Once again the
difference between τοίνυν and οὖν depends on the question whether
the speaker supposes (or presents it as such) that the statement falls
outside or beyond the expectation of his public (such as in ex. 19) or
not (see ex. 5). Of course in (19) the information is crucial (if so the
speaker is not guilty as charged), but possibly difficult to believe for
the members of the jury. Therefore, τοίνυν is used.

c. Direct addresses
Addressing the jury directly at crucial points in the story/argumentation
means getting the jury involved as much as possible, because what is
said is important. Of course this linguistic means is compatible with
a lot of situations, but surely those marked by τοίνυν, cf. (12), (15),
(16), (17). In 27,3 % (= 36 examples) a direct address to the jury is
added, whereas in the case of οὖν this is only in 9% of the examples,
cf. (18). Once again this may be seen as a signal that the use of τοίνυν
is preferred in crucial passages in which it is very important to gain
commitment/belief of the jury. For the same reason τοίνυν-clauses
 discourse cohesion marked by OYN and ΤΟINYN 79

often contain exhortations, expressions of the speaker’s will, or of what

is appropriate to do. It always concerns states of affairs that are crucial
for the speaker, and (presented as) possibly outside or beyond the
expectations of the addressee(s). A combination of all these factors of
course reinforces the emotional and urgent tone of the discourse unit
in question, such as in (17) above.

3.4. Connective particles and asyndeton

As stated above, asyndeton is traditionally defined as the lack of a
coordinator (i.e. coordinating conjunction) in between sentences. This
means that, apart from the rare examples of δ’ οὖν, all other examples
would have to be classified as cases of asyndeton. The above description
shows, however, that both particles οὖν and τοίνυν establish a transition
from the previous discourse unit to the next one.32 So in my opinion
they both function on the presentational level, just as the traditional
coordinators δέ, καί and ἀλλά.33 As for τοίνυν, I would claim that it
functions on two discourse levels at the same time: the presentational
level and the attitudinal/interactional level, combining its two original
parts: presentational/text structuring νυν and attitudinal τοι (‘perhaps
you won’t believe it, but take it from me that (. . .)’). I would argue,
then, in favour of a redefinition of the term ‘asyndeton’ so as to exclude
those connective particles that function on the presentational level. As
to the suggestion of Denniston (1954: xliii n.2) that these connective
particles may have a mitigating effect on the harshness of an asyndeton,
one may ask whether this is really the case, because also in δέ-clauses
(notably in clauses with topic switch) the other so called mitigating
means he mentions (backwards pointing pronouns and demonstra-
tives etc.) occur, cf. μετὰ δὲ ταῦτα; ἐγὼ δέ, οὕτως δέ etc.). So one may
wonder what this ‘mitigation’ amounts to, and what is the effect of
‘unmitigated asyndeton’.34
As an illustration of the problem as to the status and definition of
asyndeton, compare (20)–(21).

32
At least in Lysias, apart from one example where οὖν is used in a relative clause.
Since it is only one example in Lysias I have not taken it into consideration. See also
note 20.
33
For the difference between coordinators and connectors see Kroon (1995: 40) and
Pinkster (1972: 153), with further literature.
34
Cf. the introduction in Rijksbaron (ed.) (1997: 2, 13).
80 gerry c. wakker

(20) οὕτω δὲ σφόδρα ἠπορούμην ὅ τι χρησαίμην, ὦ βουλή, τῇ τούτου

παρανομίᾳ, ὥστε ἔδοξέ μοι κράτιστον εἶναι ἀποδημῆσαι ἐκ τῆς
πόλεως.
I was so perplexed, gentlemen, in face of this man’s lawless behav-
iour, that I decided that it would be best for me to reside abroad.
(Lys. 3.10)

(21) οὕτω σφόδρ’ αὐτοὺς ἐφυλλαττόμην, καὶ πάντα τὰ ὑπὸ τούτων

γιγνόμενα μεγάλην ἐμαυτῷ συμφορὰν ἐνομιζον.
(With this conclusion I took myself off by another street.) So much
I was on my guard against them, and I regarded all the proceedings
of these men as a grievous misfortune to myself. (Lys. 3.13, cf. in
this speech also ὥστε in 3.32; 3.45)
(20) is part of the narrative (3.5–26) and within this narrative it starts
a new discourse unit (the story about his reasons to leave the city), cf.
the presence of the coordinator δέ, which marks discontinuity (new
discourse unit) within a larger whole (the narrative); οὕτω here points
forwards to the following ὥστε. In (21), on the other hand, οὕτω is
pointing backwards, resuming and explaining the information given
and thus rounding off this discourse unit.
Without claiming to have found a definite answer to the question
what asyndeton is used for, I hope to have shown that at least it is
better to make a redefinition of the term ‘asyndeton’, to the effect that
if a clause is introduced by οὖν or τοίνυν (in between sentences), they
are not considered ayndetic any longer.

4. Concluding remarks

In my opinion, both connective particles studied above have their

own semantic value: οὖν indicates that the speaker proceeds to a new
important point (thereby having the effect of indirectly characterizing
the preceding unit as relevant but subsidiary to or preparatory for what
he is going to say now), whereas τοίνυν on the one hand performs
a similar function (by νυν), but on the other hand adds the nuance:
you (= the addressee) must take notice of it because possibly you do
not expect this (τοι). All examples and interpretations may be taken as
illustrations of these basic semantic values. Accordingly, I would charac-
terize οὖν as a text structuring POP particle, τοίνυν as a modal and text
structuring POP particle. This means that both particles prefer their own
 discourse cohesion marked by OYN and ΤΟINYN 81

type of context, in accordance with their semantic value. Sometimes,

however, they seem to be used in similar contexts. This observation is
compatible with the so called Prototype theory:35 linguistic elements all
have their own semantic value. In actual usage some are more central,
others more peripheral. Sometimes the semantic differences between
the particles might seem to be neutralized or in any case minimalized.
In these contexts the choice is a question of presentation and of what
the speaker wants to convey and emphasize in his message. A speaker
does not use them at random as argued above.
By way of summary, I have tried to capture all the above said about
the semantic value and the pragmatic implications of the particles
concerned in a table (which holds for at least Lysias):

Table 2: Function of sentence connections36

Sentence Discourse level Basic Explicitly Relevance Commitment Commitment
connection on which semantic related to discourse avowed by S presupposed
the particle value previous text unit in A
primarily
functions

τοίνυν presentational next highly yes POP- high: low

+ attitudinal important particle: ‘you take it
point; you highly from me that’
must note important
point
οὖν presentational next yes POP- neutral neutral
important particle:
point important
point
δέ presentational next new yes and no; neutral neutral neutral
item discontinuity
within larger
textunit
‘asyndeton’ – – no neutral neutral neutral
(traditional
sense of the
word)

35
For a brief and clear description of Prototype theory see, for instance, Bakker
(1988: 14–8).
36
Note that καί, ἀλλά, δέ are also used interclausally, whereas οὖν and τοίνυν are
only found intersententially. I note that by ‘neutral’ I mean that this type of sentence
connection does not give information on this point (as part of its semantic value or
as a conventional pragmatic side effect of its basic meaning).
 CHAPTER FIVE

THE PARTICLES ΑY AND ΑYΤΕ IN ANCIENT GREEK AS

TOPICALIZING DEVICES

Antonio R. Revuelta Puigdollers

1. Introduction

The purpose of this paper is to describe some of the meanings of the

particles αὖ and αὖτε: in particular I am going to concentrate on those
that contribute to topic management in Ancient Greek.1
Any attempt to explain the function of the connectors αὖ and αὖτε
poses many different problems. First of all, they are very often ill-
explained or ill-translated in dictionaries, commentaries and transla-
tions, as LSJ shows (s.v. αὖ / αὖτε).2 The dictionary’s first entry for these
particles ascribes a meaning completely absent from their actual use:
the expression of repetition (again).3 Although both particles co-occur
with repetition adverbs (αὖθις, πάλιν)—as well as with many other
adverbs, not only temporal—, in isolation, they do not exhibit this
meaning. Second, both connectors are almost absent from Denniston’s

1
As LSJ s.v. shows, the difference between αὖ and αὖτε is a matter of literary
genre and style. Much of the analysis applied to both particles is based on Kroon’s
(1995) account of autem in Latin. For some differences between the Latin and Greek
particles see Revuelta (1996). The Greek particles can be involved in more types of
focusing constructions than the Latin ones: additive, selective, parallel and absolute.
For the focus types and devices see, among others, Dik et al. (1981), Hannay (1983)
and König (1991).
2
See also Kühner-Gerth (1898/1904: 2.278–79). More information about these
particles is available in Hartung (1832–3), Hoogeveen (1782), Klein (1988), Revuelta
(1996) and Shorey (1928).
3
The concept of Repetition is understood in this paper as appears in the example
‘Peter went to the library again’, where the adverb again works as a ‘presuppositional
trigger’ introducing the presupposition that Peter had gone to the library at least once
before. In these cases the adverbs operate as ordinals quantifying over occurrences of
States of Affairs. See Dik (1997 I: 236) and Dik et al. (1990) for this kind of σ1 and σ2
aspectual terms. As for the use of πάλιν and αὖθις in this way in Ancient Greek see
Revuelta (1996). The particles αὖ and αὖτε do not express repetition in this narrow
sense of the word.
84 antonio r. revuelta puigdollers

book4 and as a negative consequence of the well-deserved success of

this work both of them have vanished from the catalogue of Ancient
Greek particles.5
This arbitrary exclusion of αὖ and αὖτε from the catalogue of particles
and the confusion about their actual meaning would wholly justify the
choice of this topic for the present paper. But the main reason is that—as
I shall try to prove—these particles contribute to discourse coherence
by marking the introduction—at least in one of their uses—of a differ-
ent discourse Topic. That is to say, they work as cohesion devices that
reflect and create text coherence.6

2. Topicality

One of the most frequent ways to organize discourse is to order the

material according to thematic similarity.7 When writing a book, for
example, content is divided into chapters: frequently the material of
every chapter is thematically8 coherent and differs from the other
chapters.9 Within every chapter, on the other hand, the general theme
is subdivided into more coherent and specific strands, and so on into
subunits within them. Therefore there is a connection between the
hierarchical structure of a text and its thematic content.
One of the ways to establish thematic unity is to discuss the same
entities or Topics. In the following paragraphs I will explore the con-
nection between topicality and the use of the particles αὖ and αὖτε

4
They are mentioned once or twice, but they are not discussed.
5
See Thrall (1962), Blomqvist (1969) and more recent works.
6
On the difference between cohesion and coherence see Kroon (1995: 30, 60 passim).
7
Apart from thematicity there are other ways to organize the discourse, as for
example the argumentative relationships, as happens, in a high degree, in forensic and
scientific discourses, where the schemes Theory-Argument or Premises-Conclusion are
formally marked by particles in Ancient Greek (‘Theory: γάρ Arguments/Evidences’
and ‘Premises: ἄρα/οὖν Conclusions’). For other organizing principles see the Geneva
School (Moeschler 1985, Roulet 1985) and the Rhetorical Structure Theory (Mann &
Thompson 1988).
8
In works organized from a thematical point of view, of course. For other organiz-
ing principles see the previous note.
9
The degree of coherence depends on many factors apart from the speaker’s linguistic
ability. Coherence is more apparent in well-structured scientific texts, whereas other
genres may be less coherent. In any case coherence contributes to make the linguistic
message more easily understandable and it should be the aim of any (in Gricean terms)
cooperative speaker.
 the particles ay and ayte as topicalizing devices 85

in Ancient Greek. I will take as my starting point the classification of

Topics established in Dik (1997), although I will introduce some minor
changes in order to account for the Greek data.10 This classification can
be roughly summarized as follows:11

(i) New Topic: an entity is introduced for the first time into the dis-
course.
(ii) Given/Accessible Topic: the entity is already available within the
discourse because it has previously been introduced (Given) or can
be inferred (Accessible) from another available topic. In subsequent
units, this topic can be kept or eliminated from the discourse.
(iii) Resumed Topic: an entity already available within the speaker’s
and addressee’s common knowledge, but which has been given up
for a while and revived later as a discourse topic.

First of all, some previous remarks must be taken into account before
applying this typology. An entity’s status as Topic is dynamic: an entity
can be a New Topic at a certain point of the discourse, turn into a
Given Topic in the next, be kept as such for a while, be given up and
disappear, be resumed later, and so on.
Second, I will establish no difference between Topic (intra-clausal
pragmatic function) and Theme (extra-clausal pragmatic function) and
I will make use of both labels as if they were equivalent, although they
are not.12 The reason for paying no attention to this difference is that
it is not relevant for a description of the particles at issue .
Finally, when I speak about Topics I am referring to discourse Topics
and not necessarily to sentence Topics. Sentence and discourse Topics
may or may not overlap. For example, in the next sentence the Subject
a happy prince is not the sentence topic, but it is the new discourse
topic of the discourse stretch opened by the sentence:13

10
For example, the labels Closing Topic (see § 3.4.) and Excursus (see first note at
§ 2.2.) will be introduced in order to account for the Greek data. As we shall see, some
(combinations of ) particles are specialized in highlighting the closing of the current
topic (duplicated μέν, μὲν δή, μὲν οὖν) or the introduction of a temporary short-lived
topic or excursus (γάρ). The notion of Promoted Topic is not explicitly stated in Dik,
but it has been developed out of a remark in his work (1997 I: 323, example 28).
11
See also Givón (1983), Hannay (1985) and Prince (1981).
12
For this difference see Dik (1997 I: 310–1, II: 389–95).
13
On the sentence level a happy prince is rather (part of) the sentence Focus (the
whole sentence is the focus), whereas the whole sentence introduces this entity (the
happy prince) as Topic on the Discourse level (the whole narration). For the distinction
86 antonio r. revuelta puigdollers

(1) Once upon a time there was a happy prince (. . .)

In the next sections I will examine the role of the particles at issue
within a Topic typology of Ancient Greek.

2.1. Second or Further Given and Accessible Topic

Once a topic has been established (a Given or Accessible Topic) the
speaker can deal with it in many different ways. The topic can be
abandoned, it can be kept, or it can be divided into different subunits
in order to be processed in a more detailed way. The particles αὖ and
αὖτε can be used in these cases in order to highlight the introduction
of the second or further subdivision of a Given Topic. In the following
three sections I will describe three different possibilities for dealing with
a Given or Accessible Topic. In the remainder of this section, I will
describe Announced Topics, Promotion of Topic and Subtopics.
Sometimes the Given Topic is composed of various entities introduced
or announced jointly in the previous units (see below A+ B [+ . . . n]),
but in subsequent sections these entities are discussed separately. The
connectors αὖ and αὖτε can be used in these cases for marking the
introduction of the second or further Given Topic within a topic chain
(Announced Topics). As the next figure shows, this use has a cohesive
effect since it clarifies the text structure:

Discourse

A + B [+ . . . n] A B [ . . . n] [A+ B [+ . . . n]]14

αὖ [αὖ]

In the following passage Critias and Alcibiades are introduced jointly

in the first unit (A+B). This Given Topic is decomposed in the next
sections and the writer comments on them separately, first on Critias

between Sentential Topic and Discourse Topic, see Brown & Yule (1983: 70–82) and
Goutsos (1997: 2–17), among others. This differentiation has arisen as a consequence
of Michel Buijs’ observations on a former version of this paper.
14
Brackets indicate that the presence of the element is possible, but not necessary.
 the particles ay and ayte as topicalizing devices 87

(A) and then on Alcibiades (B). The particle αὖ establishes the boundary
between them. In the fourth section they are again treated jointly:15

(2) X. Mem. 1.2.24–6:16

A+ B καὶ ΚριτίαςA δὴ καὶ ἈλκιβιάδηςB, ἕως μὲν Σωκράτει συνήστην,

ἐδυνάσθην ἐκείνῳ χρωμένω συμμάχῳ τῶν μὴ καλῶν ἐπιθυμιῶν
κρατεῖν· ἐκείνου δ᾿ ἀπαλλαγέντε,
And indeed it was thus with Critias and Alcibiades. So long as they
were with Socrates, they found in him an ally who gave them strength
to conquer their evil passions. But when they parted from him, (24)

A ΚριτίαςA μὲν φυγὼν εἰς Θετταλίαν ἐκεῖ συνῆν ἀνθρώποις ἀνομίᾳ

μᾶλλον ἢ δικαιοσύνῃ χρωμένοις,
Critias fled to Thessaly, and got among men who put lawlessness
before justice; (24)

B ἈλκιβιάδηςB δ’ αὖ16 διὰ μὲν κάλλος ὑπὸ πολλῶν καὶ σεμνῶν γυναικῶν
θηρώμενος, διὰ δύναμιν δὲ τὴν ἐν τῇ πόλει καὶ τοῖς συμμάχοις ὑπὸ
πολλῶν καὶ δυνατῶν [κολακεύειν] ἀνθρώπων διαθρυπτόμενος,
ὑπὸ δὲ τοῦ δήμου τιμώμενος καὶ ῥᾳδίως πρωτεύων, ὥσπερ οἱ τῶν
γυμνικῶν ἀγώνων ἀθληταὶ ῥᾳδίως πρωτεύοντες ἀμελοῦσι τῆς
ἀσκήσεως, οὕτω κἀκεῖνος ἠμέλησεν αὑτοῦ.
While Alcibiades, on account of his beauty, was hunted by many
great ladies, and because of his influence at Athens and among her
allies he was spoilt by many powerful men: and as athletes who gain
an easy victory in the games are apt to neglect their training, so the
honour in which he was held, the cheap triumph he won with the
people, led him to neglect himself. (24–5)

A+ B τοιούτων δὲ συμβάντων αὐτοῖνA+B, (. . .) τί θαυμαστὸν εἰ ὑπερηφάνω

ἐγενέσθηνA+B;
Such was their fortune: (. . .) what wonder if they grew overbearing.
(25–6)

15
As in the rest of the paper the format applied to the text tries to capture its hierar-
chical structure while keeping its conventional disposition. The letters on the left refer
to the topic of the discourse unit on the right. As for the same use of αὖτε see Hom.
Il. 15.332–8, especially 337. In almost all examples translations are taken with minor
changes from the Loeb Classical Library.
16
The particles at issue frequently appear with coordinating conjunctions (δέ, τε,
and so on). This is just a case of compatibility rather than redundancy: coordinating
conjunctions introduce a linguistic element belonging to the same syntactic hierarchi-
cal level as the previous one, whereas αὖ/αὖτε introduce an entity (topic or focus) in
contrast with some other entity (mentioned or not in the previous context). For more
details see Revuelta (1996).
88 antonio r. revuelta puigdollers

Topicalization is not restricted to core elements of the predication

(arguments), but can also apply to more peripheral constituents
(satellites) that establish the settings of the described events (Time,
Circumstances and the like) and the status as topic of these entities
can therefore be underlined by αὖ/αὖτε, as happens in the following
passage taken from Herodotus. Darius gives his opinion about three
political systems. The introduction of the third system is marked by αὖ
and its form is that of a circumstance (‘when the mass has the power’,
absolute genitive):

(3) Hdt. 3.82.4–19:

A+B+C τριῶν γὰρ προκειμένων καὶ πάντων τῷ λόγῳ ἀρίστων ἐόντων,

δήμουA τε ἀρίστου καὶ ὀλιγαρχίηςB καὶ μουνάρχουC, πολλῷ τοῦτο
προέχειν λέγω.
For the choice lying between these three, and each of them,
democracy, oligarchy and monarchy being supposed to be the
best of its kind, I hold that monarchy is by far the most excellent,
(4–6)

C ἀνδρὸςC γὰρ ἑνὸς τοῦ ἀρίστου οὐδὲν ἄμεινον ἂν φανείη· (. . .)

Nothing can be found better than the rule of the one best man;
(. . .) (7–10)

B ἐν δὲ ὀλιγαρχίῃB πολλοῖσι ἀρετὴν ἐπασκέουσι ἐς τὸ κοινὸν ἔχθεα

ἴδια ἰσχυρὰ φιλέει ἐγγίνεσθαι· (. . .)
But in oligarchy the desire of many to do the state good service
sometimes engenders bitter enmity among them; (. . .) (10–6)

A δήμουA τε αὖ ἄρχοντος ἀδύνατα μὴ οὐ κακότητα ἐγγίνεσθαι·

(. . .)
Again, the rule of the commonality must of necessity engender
evil-mindedness (. . .). (16–9)
 the particles ay and ayte as topicalizing devices 89

In other cases an entity with low topicality in the previous discourse

unit is promoted to the topic in the next unit. The particles at issue
contribute to mark that promotion (Topic Promotion):17

Discourse

A+ b B (+ c) ... [N+ o]

αὖ [αὖ] [αὖ]

In the following passage of the Iliad the poet describes the inheritance
of the royal sceptre. Each donor appears as a Given Topic, whereas his
heir appears at the end of the same unit as a New Topic and turns into
the main Topic in the next unit. The particle αὖτε is placed in two of
the boundaries marking off one donation from the other:18

17
The difference between this type of topic change is justified from a typological
point of view, since the Spanish discourse connector a su vez is specialized in marking
it: ‘Pedro le vendió su coche a Luis; éste se lo regaló a su hijoi y éstei, a su vez, me lo
vendió a mí’ (Pedro sold his car to Luis; this one gave it to his soni and this onei in
turn sold it to me). Capital letters mark high topicality (main Given Topic), whereas
small letters indicate low topicality (New Topic). The entity that appears as New Topic
in one unit (e.g. ‘b’) turns into the main topic in the next (‘B’). As for the same use of
αὖ see Hdt.7.23.2–8 (specially 7.23.7).
18
The particle does not appear in all similar cases of the example and this fact should
be explained, since it happens in many other passages. The reason is that particles
contribute to the cohesion of the text underlining the topical status of the entities, but
they do not make the entities into topics. For this distinction between coherence (the
topical structuring of the text in this case) and the cohesion structures (in this case
the particles), see once more Kroon (1995: 30): ‘From such a viewpoint text structure
involves on the one hand semantic and interactional patterns (what might be called
coherence structures) and, on the other hand, the ways in which these patterns are
formally expressed on the other (cohesion structures).’ Put in other words, particles
just help to clarify the relationships already existing within the text.
90 antonio r. revuelta puigdollers

(4) Hom. Il. 2.102–7:

A+b ἭφαιστοςA μὲν δῶκε Διὶb Κρονίωνι ἄνακτι,

Hephaestus gave it to lord Zeus, son of Cronos,

B+c αὐτὰρ ἄρα ΖεὺςB δῶκε διακτόρῳ ἀργεϊφόντῃc·

and Zeus gave it to the messenger Argeïphontes;

C+d ἙρμείαςC δὲ ἄναξ δῶκεν Πέλοπιd πληξίππῳ,

and Hermes, the lord, gave it to Pelops, driver of horses,

D+e αὐτὰρ ὃ αὖτε ΠέλοψD δῶκ᾿ Ἀτρέϊe ποιμένι λαῶν,

and Pelops, in turn, gave it to Atreus, (105)

E+f ἈτρεὺςE δὲ θνῄσκων ἔλιπεν πολύαρνι Θυέστῃf,

and Atreus at his death left it to Thyestes, rich in flocks

F+g αὐτὰρ ὃ αὖτε Θυέστ᾿F Ἀγαμέμνονιg λεῖπε φορῆναι, / πολλῇσιν

νήσοισι καὶ Ἄργεϊ παντὶ ἀνάσσειν.
and Thyestes again left it to Agamemnon to carry, to be lord of
many isles and of all Argos.

In other cases the Given Topic has not been explicitly mentioned, but
can be inferred (as a subtopic) from some entity in the previous context
(supertopic). As in the previous cases, the particles at issue highlight
the introduction of the second or further inferrable subtopic, as the
following picture describes:

Discourse

A A1 A2 [An] [A]

αὖ [αὖ]
 the particles ay and ayte as topicalizing devices 91

The relationship between the supertopic and its subtopics may vary
in nature: it can be inclusion in a class, or a part and a whole, among
others. In the following passage from Homer, once the Achaeans are
mentioned any further mention of a particular Achaean (class) is not
unexpected:19

(5) Hom. Il. 17.596–602:

A νίκην δὲ Τρώεσσι δίδου, ἐφόβησε δ᾿ ἈχαιούςA.

He [Zeus] gave victory to the Trojans, but the Achaeans he drove in
rout.

A1 πρῶτος Πηνέλεως ΒοιώτιοςA1 ἦρχε φόβοιο. / βλῆτο γὰρ ὦμον δουρὶ

πρόσω τετραμμένος αἰεὶ / ἄκρον ἐπιλίγδην· γράψεν δέ οἱ ὀστέον
ἄχρις / αἰχμὴ Πουλυδάμαντος· ὃ γάρ ῥ᾿ ἔβαλε σχεδὸν ἐλθών.
First to begin the rout was Peneleos the Boeotian. For as he stood ever
facing the foe he was struck on the top of the shoulder with a spear, a
glancing blow, but the spear point of Polydamas cut clean to the bone,
for he it was who cast at him from near at hand.

A2 ΛήϊτονA2 αὖθ᾿ Ἕκτωρ σχεδὸν οὔτασε χεῖρ᾿ ἐπὶ καρπῷ / υἱὸν

Ἀλεκτρυόνος μεγαθύμου, παῦσε δὲ χάρμης·
And Leïtus again, the son of great-hearted Alectryon, did Hector
wound in close fight, on the hand at wrist, and made him cease from
the fight.

19
As for the whole-part relation see Homer’s Iliad 3.328–33, where, once the armours
(A) are mentioned, the appearance of the greaves and corselets (parts of the armours,
A1 and A2) is easily evoked: αὐτὰρ ὅ γ᾿ ἀμφ᾿ ὤμοισιν ἐδύσετο τεύχεα καλὰA / δῖος
Ἀλέξανδρος Ἑλένης πόσις ἠϋκόμοιο. / κνημῖδαςA1 μὲν πρῶτα περὶ κνήμῃσιν ἔθηκε /
καλάς, ἀργυρέοισιν ἐπισφυρίοις ἀραρυίας·, / δεύτερον αὖ θώρηκαA2 περὶ στήθεσσιν
ἔδυνεν / οἷο κασιγνήτοιο Λυκάονος· (And he put on about his shoulders his beautiful
armor, noble Alexander, the husband of fairhaired Helen. The greaves first he set about
his legs; beautiful they were, and fitted with silver ankle pieces; next he put on about
his chest the corselet of his brother Lycaon and fitted it to himself, 328–33).
92 antonio r. revuelta puigdollers

2.2. Resuming a Given Topic

Sometimes the Given Topic is abandoned and a new one occupies
its place.20 If the speaker wants to revive or resume the previous topic
he has to make use of certain strategies. Strategies that are possible
are the use of either a strong form of anaphoric reference or particles
(Dik 1997 I: 325–326, 457). From the point of view of discourse
structuring, this kind of topic could be represented as in the following
figure:

Discourse

A B A

Given Topic Excursus Resumed Topic

αὖ/αὖτε/οὖν

The particle αὖ is used in the next example to mark the reintroduc-

tion of a previous topic. In the passage taken from Homer, the main
topic is Nestor, which is given up for a short reference to Neleus; the
main topic, Nestor, is reintroduced after the short excursus through
the formula Νέστωρ αὖ:21

20
This new topic can be a main topic or just a secondary topic or excursus, as hap-
pens in the examples provided in this section.
21
For an example with αὖτε see Hom. Il. 4.127–33, where the main Topic is Athena
and there is a short excursus with a comparison with mothers.
 the particles ay and ayte as topicalizing devices 93

(6) Hom. Od. 3.404–12:

A ἦμος δ᾿ ἠριγένεια φάνη ῥοδοδάκτυλος Ἠώς, / ὤρνυτ᾿ ἄρ᾿ ἐξ

εὐνῆφι Γερήνιος ἱππότα ΝέστωρA, / ἐκ δ᾿ ἐλθὼν κατ᾿ ἄρ᾿ ἕζετ᾿
ἐπὶ ξεστοῖσι λίθοισιν, / οἵ οἱ ἔσαν προπάροιθε θυράων ὑψηλάων /
λευκοί, ἀποστίλβοντες ἀλείφατος·
So as early Dawn appeared, the rosy-fingered, up from his bed
rose the horseman, Nestor of Gerenia, and went forth and sat
down on the polished stones which were before his lofty doors,
white and glistered as with oil.

BEXCURSUS οἷσ᾿ ἔπι μὲν πρὶν / ΝηλεὺςB ἵζεσκεν, θεόφιν μήστωρ ἀτάλαντος·
/ ἀλλ᾿ ὁ μὲν ἤδη κηρὶ δαμεὶς ᾌuδόσδε βεβήκει,
On these of old was wont to sit Neleus, the peer of the gods in
counsel; but ere he had been stricken by fate and had gone to
the house of Hades,

A ΝέστωρA αὖ τότ᾿ ἐφῖζε Γερήνιος, οὖρος Ἀχαιῶν, / σκῆπτρον

ἔχων.
And now there sat upon them in his turn Nestor of Gerenia, the
warder of the Achaeans, holding a sceptre in his hands.

The previous scheme may be used to confirm the reading of the manu-
scripts in the following passage taken from Thucydides: the Athenians
receive news about Hegesandridas; there is a short excursus about him
and afterwards the narration goes back to the Athenians. Although
the textual transmission is almost unanimous, Bekker proposed to
change αὖ into οὖν and, although it is not accepted in most editions,
his emendatio appears in almost every apparatus criticus. As the previ-
ous examples taken from Homer and many others prove, there is no
reason to doubt the soundness of this passage, and it can be analyzed
in the same way:
94 antonio r. revuelta puigdollers

(7) Th. 8.94.1–3:

ἐπειδὴ δὲ ἐπῆλθεν ἡ [ἐν Διονύσου] ἐκκλησίαA καὶ ὅσον οὐ

A+b ξυνειλεγμένοι ἦσαν, ἀγγέλλονται αἱ δύο καὶ τεσσαράκοντα νῆες καὶ
ὁ Ἀγησανδρίδαςb ἀπὸ τῶν Μεγάρων τὴν Σαλαμῖνα παραπλεῖν· καὶ
πᾶς τις [τῶν πολλῶν ὁπλιτῶν] αὐτὸ τοῦτο ἐνόμιζεν εἶναι τὸ πάλαι
λεγόμενον ὑπὸ Θηραμένους καὶ τῶν μετ᾿ αὐτοῦ, ὡς ἐς τὸ τείχισμα
ἔπλεον αἱ νῆες, καὶ χρησίμως ἐδόκει καταπεπτωκέναι.
When the time had come for the meeting of the assembly in the
precinct of Dionysus and they had all but gathered there, it was
announced that Hegesandridas with his forty two ships had left
Megara and was sailing to the coast of Salamis; and the hoplites to
a man believed that this was precisely the move that had long since
been predicted by Theramenes and his party and that the ships were
coming in to occupy the fort; and they felt that its demolition had
served a useful purpose. (1–2)

BEXCURSUS ὁ δὲ ἈγησανδρίδαςB τάχα μέν τι καὶ ἀπὸ ξυγκειμένου λόγου περί

τε τὴν Ἐπίδαυρον καὶ ταύτῃ ἀνεστρέφετο, εἰκὸς δ᾿ αὐτὸν καὶ
πρὸς τὸν παρόντα στασιασμὸν τῶν Ἀθηναίων, δι᾿ ἐλπίδος ὡς κἂν
ἐς δέον παραγένοιτο, ταύτῃ ἀνέχειν.
Now it may well be that Hegesandridas was acting in accordance
with some prearranged understanding when he hovered about
Epidaurus and that neighbourhood, but it is probable that in
tarrying there he also had regard to the dissension prevailing
among the Athenians, and was hopeful that possibly he might
arrive in the very nick of time. (2–3)

A οἱ δ᾿ αὖ ἈθηναῖοιA, ὡς ἠγγέλθη αὐτοῖς, εὐθὺς δρόμῳ ἐς τὸν Πειραιᾶ

πανδημεὶ ἐχώρουν, ὡς τοῦ ἰδίου πολέμου μείζονος [ἢ] ἀπὸ τῶν
πολεμίων οὐχ ἑκάς, ἀλλὰ πρὸς τῷ λιμένι ὄντος.
However this may be, when his movements were reported to the
Athenians they immediately advanced at a run with all their forces
to the Peiraeum, thinking that a new war, launched by the enemy
and more serious than their own domestic feud, was not far away,
nay, was actually at their port. (3)

This use of αὖ/αὖτε is a side effect of its more general use as topic
change marker, as explained in § 2.4.22

22
The particle οὖν would explicitly state that a previous topic had been recovered,
whereas αὖ simply indicates that there is a change of topic; in this case the introduced
topic happens to be a previous topic.
 the particles ay and ayte as topicalizing devices 95

This having been said, it is true that in Ancient Greek the particle
οὖν is more frequently used than αὖ/αὖτε in connection with a resumed
topic, as the next passage, taken from Lysias, shows.23 An old woman
(Topic A) approaches Euphiletus to inform him about Eratosthenes’ love
affair with his (Euphiletus’) wife; there is a short excursus about another
woman (Topic B)—a former lover of Eratosthenes—who is sending the
messenger of this bad news; and afterwards the topic of the old woman
(A) is reintroduced through (i) the use of a strong anaphoric form
(ἡ ἄνθρωπος), (ii) the repetition of the previous words (see προσέρχεταί
μοί and προσελθοῦσα μοι ἐγγὺς) and (iii) the presence of οὖν:

(8) Lys. 1.15–16

A+b μετὰ δὲ ταῦτα, ὦ ἄνδρες, χρόνου μεταξὺ διαγενομένου καὶ ἐμοῦ

πολὺ ἀπολελειμμένου τῶν ἐμαυτοῦ κακῶν, προσέρχεταί μοί τις
πρεσβῦτις ἄνθρωποςA, ὑπὸ γυναικὸςb ὑποπεμφθεῖσα ἣν ἐκεῖνος
ἐμοίχευεν, ὡς ἐγὼ ὕστερον ἤκουον·
After this, sirs, an interval occurred in which I was left quite unaware
of my own injuries; I was then accosted by a certain old female, who
was secretly sent by a woman with whom that man was having an
intrigue, as I heard later.

B αὕτηB δὲ ὀργιζομένη καὶ ἀδικεῖσθαι νομίζουσα, ὅτι οὐκέτι ὁμοίως

EXCURSUS ἐφοίτα παρ’ αὐτήν, ἐφύλαττεν ἕως ἐξηῦρεν ὅ τι εἴη τὸ αἴτιον.
This woman was angry with him because he no longer visited
her so regularly, and she kept a close watch on him until she
discovered what was the cause.

A προσελθοῦσα οὖν μοι ἐγγὺς ἡ ἄνθρωποςA τῆς οἰκίας τῆς

ἐμῆς ἐπιτηροῦσα, ‘Εὐφίλητε’ ἔφη ‘μηδεμιᾷ πολυπραγμοσύνῃ
προσεληλυθέναι με νόμιζε πρὸς σέ·
So the old creature accosted me where she was on the look-out,
near my house, and said: ‘Euphiletus, do not think it is from any
meddlesomeness that I have approached you.’

The particle οὖν is rather a POP marker, that is to say, a connector

that takes the discourse strand out of an embedded unit (Polanyi &
Scha 1983), and its use as a topic resuming marker is a side effect of

23
See Denniston’s (1954: 428–9) short, but very informative explanation on this
resumptive use of οὖν.
96 antonio r. revuelta puigdollers

this more general use. A particle whose main function is to mark a

transition from an embedded unit to a main unit is very suitable for
many purposes: it is useful for drawing conclusions from previous
premises (the embedded material), as happens in οὖν’s inferential
meaning (see Denniston (1954: 426)), or to go back to the main topic
after an excursus (the embedded narrative material), as in the example
here discussed.

2.3. New Topic

Although still fewer in number,24 there are some cases where the par-
ticles at issue seem to contribute to the introduction of a completely
new discourse topic that can’t be inferred from the actual Given Topic
nor is it related to any previous Topic, as in this passage from Eu-
ripides’ Orestes:

(9) E. Or. 132–9

A Helen and Electra’s conversation (at the end Helen leaves the stage
and Electra remains on it)

B αἵδ᾿B αὖ πάρεισι τοῖς ἐμοῖς θρηνήμασι / φίλαι ξυνῳδοίA· τάχα

μεταστήσουσ᾿ ὕπνου / τόνδ᾿ ἡσυχάζοντ᾿, ὄμμα δ᾿ ἐκτήξουσ᾿B ἐμὸν
/ δακρύοις, ἀδελφὸν ὅταν ὁρῶ μεμηνότα. / [ὦ φίλταται γυναῖκεςB,
ἡσύχῳ ποδὶ / χωρεῖτε, μὴ ψοφεῖτε, μηδ᾿ ἔστω κτύπος. / φιλία
γὰρ ἡ σὴ πρευμενὴς μέν, ἀλλ᾿ ἐμοὶ / τόνδ᾿ ἐξεγεῖραι συμφορὰ
γενήσεται.]’
(Enter by eisodos B Argive women as chorus) ELECTRA. Here they
are again [αὖ], my friends who sing in harmony with my laments!
They will awaken my brother at once from his peaceful sleep, making
my eyes melt with tears when I see his insanity! [Dear friends, walk
with quiet step, make no noise, let there be no clattering! To be
sure, your friendship is kindly meant, yet still! It would be a great
misfortune to wake this man up.]

24
It is not possible to give clear-cut statistics of the particles’ uses, since in many
cases the borders between them are fuzzy (see § 2.4). Other examples: E. Tr. 706–8
(τίν᾿ αὖ . . .), Hom. Il. 5.418–9 (αἳ δ᾿ αὖτ᾿ . . .).
 the particles ay and ayte as topicalizing devices 97

As the translation shows,25 above all in these difficult passages the

particle is frequently interpreted as a repetition adverb (‘once more’,
‘again’ and the like), although no other mention has been made of the
chorus previously.26 I think that the temporal adverb interpretation is
just an easy way out: a temporal relationship is always easy to establish
and compatible with almost any context.27 There are more reasons to
prefer its interpretation as a New Topic Marker: (i) this interpretation
gives a better account of this and other examples, where no repetition
takes place, (ii) it is coherent with the other topicalizing uses of these
particles seen in the previous sections28 and (iii) in this particular case
the presentative structure (πάρεισι ‘here they are’) indicates that the
speaker is introducing a new entity into the discourse domain (New
Topic).
Nevertheless, it is important to distinguish between New Topic and
Absolute First Topic of Discourse introduced by formulas like ‘once
upon a time’ and the like.29 Very frequently the latter introduce an entity
completely unknown to the reader and fix it from the beginning of the
narration as its general topic, as in the next example:

25
The translation has not been corrected in order to show how the particles are
understood in these passages.
26
See West (1987) and Willink’s (1986) comments. ‘again: if this is the right trans-
lation, it is a unique suggestion of previous songs which the audience has not heard,
and this may anyway be implied by “my friends who sing with me”. Alternatively
αὖ may mean “here’s another interruption.” ’ (West); ‘αὖ: i.e. as a further upsetting
disturbance’ (Willink). In this passage αὖ is very frequently translated as if it were
αὖθις (‘once more’).
27
There are many difficult words in Ancient Greek rendered through temporal
expressions: translators and lexicographers feel compelled to offer a description and a
temporal relationship is innocuous and easy to find. In this case the etymology con-
tributes to wrongly identify αὖ/αὖτε’s and αὖθις’ meanings (see Boisacq (1916: 99),
Chantraine (1968–80: 137), Kühner-Gerth (1898/1904: 2.279).
28
It is coherent as well with the particles’ use as focusing and discourse listing
devices. See Revuelta (1996: chapters 9–11).
29
When describing the concept of New Topic, linguists (see Dik 1997 I: 315) usu-
ally exemplify with Absolute First Topic of Discourse, probably because this is a very
clear example of a New Topic. Nevertheless, speakers introduce new entities in their
discourse and make them into topics at other stages of their speech and not only at
the beginning.
98 antonio r. revuelta puigdollers

(10) ‘Once upon a time,’ said the Linnet, ‘there was an honest little
fellow named Hansi.’ ‘Was hei very distinguished?’ asked the
Water-rat. ‘No,’ answered the Linnet, ‘I don’t think hei was dis-
tinguished at all (. . .). (Oscar Wilde, The devoted friend, Gutenberg
Project)

The particles αὖ/αὖτε are never used in this second sense. Under New
Topic I understand an entity not previously mentioned nor able to
be inferred from the previous linguistic context, although it could be
inferred from extra linguistic data, as probably happens in the Greek
example, where the chorus physically appears. The chorus is a new
topic in the sense that it has not been previously mentioned (linguistic
context) nor noticed (physical context) nor can be inferred from the
preceding information. The particle αὖ is most probably used here in
order to mark the transition between Helen, who was previously talking
to Electra and has just left the scene, and the first appearance of the
chorus on the stage. This general meaning is coherent with the mean-
ings described in previous sections: the speaker introduces a further
(second, third, and so on) topic into his/her discourse, but this time it
is new in the sense that it is unrelated to the previous one.

2.4. Topic change markers

From the previous sections (§ 2.1–3) we can conclude that the particles
αὖ/αὖτε are used in order to mark or highlight the introduction of a
further (second, third and so on) topic that is more or less related to the
previous one(s). For that reason they are usually found at the beginning
of a new discourse section. This broad general meaning is the reason
why the particle is compatible with different topics (Given, Accessible,
Resumed, New Topic) and can be interpreted in different ways. For
example, in the next passage taken from Homer the introduction of
Apollo in the last unit could be interpreted as either (i) the introduction
of a second/further component of the previous Given Topic announced
in the first unit, or (ii) as the resumption of the topic (Apollo) that
appears in the second unit and is abandoned in the third:
 the particles ay and ayte as topicalizing devices 99

(11) Hom. Il. 7.17–43 (conversation between Athena and Apollo):

A+B τοὺς δ᾿ ὡς οὖν ἐνόησε θεὰ γλαυκῶπις ἈθήνηA / Ἀργείους

ὀλέκοντας ἐνὶ κρατερῇ ὑσμίνῃ, / βῆ ῥα κατ᾿ Οὐλύμποιο καρήνων
ἀΐξασα / Ἴλιον εἰς ἱερήν· τῇ δ᾿ ἀντίος ὄρνυτ᾿ ἈπόλλωνB / Περγάμου
ἐκκατιδών, Τρώεσσι δὲ βούλετο νίκην· / ἀλλήλοισι δὲ τώ γε
συναντέσθην παρὰ φηγῷ.
But when the goddess, flashing-eyed Athene, noticed them as they
were slaying the Argives in the mighty combat, she darted down
from the peaks of Olympus to sacred Ilios. And Apollo rushed to
meet her, for he had looked down from Pergamus and seen her,
and was eager to have victory for the Trojans. So the two met one
another by the oak tree.

B τὴν πρότερος προσέειπεν ἄναξ Διὸς υἱὸς ἈπόλλωνB· (. . .)

Then to her spoke first the king Apollo, son of Zeus:

A τὸν δ᾿ αὖτε προσέειπε θεὰ γλαυκῶπις ἈθήνηA· (. . .)

And in answer to him spoke the goddess, flashing-eyed Athene:

B τὴν δ᾿ αὖτε προσέειπεν ἄναξ Διὸς υἱὸς ἈπόλλωνB· (. . .)

In answer to her spoke king Apollo, son of Zeus:

No matter how we classify this example, it is clear that the particles

αὖ/αὖτε can be described as discourse boundary-markers that contribute
to increasing text cohesion.30

3. Heuristics

There are many formal clues that support the previous account of αὖ/
αὖτε as topic change markers. In the next paragraphs I shall comment
on some of them: fronting and left dislocation, topicalizing questions,
metadiscursive expressions and topic closing devices.
First of all, in a high percentage of cases, the entity introduced by the
particles αὖ/αὖτε appears at the beginning of the sentence (Ziv 1994), as
the many examples already mentioned prove, except when this entity is

30
See Bakker (1993) for the function as boundary-marker of δέ.
100 antonio r. revuelta puigdollers

a New Topic.31 These constituents work as ‘headings’ that announce the

thematic content of the new section. In some cases the nominal occu-
pies an extra-clausal position32 that highlights its thematic relevance,
as when it appears out of the clause (e.g. left dislocations in questions
in example 12) or in a proleptic position out of the subordinate clause
to which the nominal belongs (e.g. in example 13).33 In (12) Socrates
proposes to classify some types of deception. Every type of deception
is presented through a conditional clause34 and recaptured in the main
clause with an expression of the kind τὴν ἀπάτην ταύτην/ταύτην τὴν
ἀπάτην or τοῦτο (with αὖ in two of the instances). In two cases (A2
and A3) these expressions precede the question word (ποῖ, ποτέρωσε,
ποτέρωθι) and are therefore placed in an extra-clausal position:

(12) X. Mem. 4.2.16–17:

A βούλει οὖν, ἔφη, ταῦταA οὕτω θέντες διορισώμεθα πάλιν πρὸς μὲν
τοὺς πολεμίους δίκαιον εἶναι τὰ τοιαῦτα ποιεῖν, πρὸς δὲ τοὺς φίλους
ἄδικον, ἀλλὰ δεῖν πρός γε τούτους ὡς ἁπλούστατον εἶναι; πάνυ μὲν
οὖν, ἔφη ὁ Εὐθύδημος.
‘Then I propose to revise our classification, and to say: it is just to
do such things to enemies, but it is unjust to do them to friends,
towards whom one’s conduct should be scrupulously honest.’ ‘By all
means.’
A1 τί οὖν; ἔφη ὁ Σωκράτης, ἐάνA1 τις στρατηγὸς ὁρῶν ἀθύμως ἔχον τὸ
στράτευμα ψευσάμενος φήσῃ συμμάχους προσιέναι καὶ τῷ ψεύδει
τούτῳ παύσῃ τῆς ἀθυμίας τοὺς στρατιώτας, ποτέρωθι τὴν ἀπάτην
ταύτηνA1 θήσομεν; δοκεῖ μοι, ἔφη, πρὸς τὴν δικαιοσύνην. (. . .)
‘Now suppose that a general, seeing that his army is downhearted,
tells a lie and says that reinforcements are approaching, and by
means of this lie checks discouragement among the men, under
which heading shall we put this deception?’ ‘Under justice, I think’
(. . .).

31
New topics prefer to appear later in the sentence, as in the introductory formula of
fairy tales ‘once upon a time there was a happy prince’. Other positions of the particles
are possible since they are used also as focusing devices (see Revuelta 1996).
32
For extra-clausal positions, see Dik (1997 II: 379–407).
33
See Hoffmann (1989: 192).
34
See Haiman (1978) for this use of conditional clauses for introducing topics.
 the particles ay and ayte as topicalizing devices 101

A2 ἐὰνA2 δέ τις υἱὸν ἑαυτοῦ δεόμενον φαρμακείας καὶ μὴ προσιέμενον

φάρμακον ἐξαπατήσας ὡς σιτίον τὸ φάρμακον δῷ καὶ τῷ ψεύδει
χρησάμενος οὕτως ὑγιᾶ ποιήσῃ, ταύτην αὖ τὴν ἀπάτηνA2 ποῖ
θετέον; Δοκεῖ μοι, ἔφη, καὶ ταύτην εἰς τὸ αὐτό. (. . .)
‘Suppose, again, that a man’s son refuses to take a dose of medicine
when he needs it, and the father induces him to take it by pretending
that it is food, and cures him by means of this lie, where shall we
put this deception?’ ‘That too goes on the same side, I think (. . .).’

A3 τί δ᾿; ἐάνA3 τις, ἐν ἀθυμίᾳ ὄντος φίλου, δείσας μὴ διαχρήσηται

ἑαυτόν, κλέψῃ ἢ ἁρπάσῃ ἢ ξίφος ἢ ἄλλο τι τοιοῦτον, τοῦτοA3 αὖ
ποτέρωσε θετέον; καὶ τοῦτο νὴ Δί᾿, ἔφη, πρὸς τὴν δικαιοσύνην. (. . .)
‘And again, suppose one has a friend suffering from depression,
and, for fear that he may make away with himself, one takes away
his sword or something of the sort, under which heading shall we
put that now’ ‘That too goes under justice, of course.’

The second possibility is exemplified in example (13), where the

δημιουργούς are the topical entity. For this reason the term is promoted
to the beginning of the sentence and appears as an argument of the
main verb (prolepsis):

(13) τοὺς ἄλλους αὖ δημιουργοὺς σκόπει εἰ τάδε διαφθείρει, ὥστε καὶ

κακοὺς γίγνεσθαι.
Consider whether these are the causes that corrupt other crafts-
men too so as positively to spoil them. (Pl. R. 421d1–2)

A second indication of the topicalizing value of the particles at issue

is the fact that they appear with questions used as topic markers,
as for example τί δέ ‘what about’ (Kühner-Gerth 1898/1904: 2.518,
Anmerkung 4). In the following example, after the introduction of the
main topic (which sort of men can be taught and which cannot) and
the enumeration of some subtopics (drinkers, sluggards and men in love
are excluded), the next subtopic (men who have a passion for lucre) is
introduced by a τί δέ question and highlighted by the particle αὖ:35

35
The particle αὖ appears after the relative pronoun, but it introduces as topic the
entity referred to by the whole relative clause: τί δέ A αὖ (A = relative clause). The
position of the particle misleads the modern reader into thinking that it operates
within the subordinate clause, when in fact it operates upon the whole relative. The
102 antonio r. revuelta puigdollers

(14) τί δέ, ἔφην ἐγώ, οἵτινεςA αὖ ἐρωτικῶς ἔχουσι τοῦ κερδαίνειν, ἦ

καὶ οὗτοι ἀδύνατοί εἰσιν εἰς ἐπιμέλειαν τῶν κατ᾿ ἀγρὸν ἔργων
παιδεύεσθαι;
And what about the men who have a passion for lucre? Are they
also incapable of being trained to take charge of the work of a
farm (. . .)? (X. Oec. 15)

Besides, in many cases the particles are accompanied by metadiscursive

expressions that allow the speaker to explicitly state the transition to a
different topic (‘now I’m going to speak about X αὖ/αὖτε’), as shown
in the next passage:

(15) X. Mem. 4.25–5.1

A1 τοιαῦτα λέγων τε καὶ πράττων δικαιοτέρους ἐποίει τοὺς

πλησιάζοντας.
By such words and actions he encouraged Justice in those who
resorted to his company.

A2 ὡς δὲ καὶ πρακτικωτέρους ἐποίει τοὺς συνόντας ἑαυτῷ, νῦν αὖ

τοῦτοA2 λέξω.
The fact that he did also try to make his companions efficient in
affairs, that is what I will now show.

Finally, since the particles αὖ/αὖτε mark a different topic and a new
discourse section, a different topic must precede them in the previous
unit. This previous topic can be just abandoned, or it can be explicitly
closed. In the latter case its closing can be stated in a formal way with
a formula (‘so far we have been speaking about topic X’) or with its
summary. This closing formula is especially frequent when the previous
topic is very long or complex or the speaker wants to round it up as
clearly as possible before proceeding to the next topic. The fact that in
many cases a topic is formally closed just before the sections headed
by the particles αὖ/αὖτε appear is itself a proof that both contribute to

same phenomenon happens when the particle appears after the conjunction heading a
conditional clause (e.g. εἰ δ’ αὖ . . .), a very frequent collocation of this particle.
 the particles ay and ayte as topicalizing devices 103

mark a new thematic unit. In these cases the discourse structure has
the following scheme:

Discourse

A B

Opening Closing αὖ

In case the speaker considers it necessary to explicitly close his/her

previous topic, this closing can be marked by particles, stated through
a metadiscursive expression or introduced by some subordinating
construction, as I will show in the next paragraphs.
When using particles to explicitly close a preceding topic, Ancient
Greek usually uses μέν36 and some of its combinations37 with other
particles, μὲν δή38 and μὲν οὖν.39 In the following passage Socrates
closes the topic of his previous conversation with Ischomachos about
his wife’s tasks—a topic which has extended for four whole chapters
(7–10)—making use of both a metadiscursive comment (‘I have heard
enough’) and the particle combination μὲν δή. Once this topic has been
closed, the particle αὖ highlights the introduction of the next related
subtopic (Ischomachos’ tasks in their household) and a new conversa-
tion begins (discourse structuring effect):

36
An example of closing μέν followed by αὖ can be seen in Hom. Il. 9.135–6. This
meaning is almost explicitly stated in Denniston’s (1954: 384–6) description of the
‘duplicated μέν’.
37
Further research is needed as far as the differences between μέν, μὲν δή and μὲν
οὖν are concerned. Nevertheless, combinations of particles cannot always be analyzed
as the addition of the single particles’ meanings.
38
This combination appears at the end of almost every conversation of the book
4 of Xenophon’s Banquet: see 4.6.1 (Calias), 4.10.1 (Niceratus), 4.28.6 (Critoboulos),
4.50.1 (Hermogenes), 4.64.10 (Socrates). Denniston’s (1954: 258) description of this
combination is clear and incredibly modern in its wording: ‘μὲν δή is frequently used
by the historians as a formula of transition, the μέν clause often summing up the
preceding section of the narrative.’
39
Denniston’s (1954: 472) comment on μὲν οὖν is clear and very modern in its
formulation; see too Sicking & Van Ophuijsen (1993: 27–8) for this and other related
senses of the combination, and Wakker (in this volume). For this particle combination
see Pl. R. 362e3–364a1. The particle αὖ appears at e5 short after μὲν οὖν at e3.
104 antonio r. revuelta puigdollers

(16) X. Oec. 7–11:

A Ischómachos details his wife’s tasksA in their householding (7–

10)

ACLOSING ὦ Ἰσχόμαχε, τὰ μὲν δὴ περὶ τῶν τῆς γυναικὸς ἔργωνA ἱκανῶς

μοι δοκῶ ἀκηκοέναι τὴν πρώτην, καὶ ἄξιά γε πάνυ ἐπαίνου
ἀμφοτέρων ὑμῶν.
At this point I said, ‘Ischomachos, I think your account of
your wife’s occupations is sufficient for the present -and very
creditable it is to both of you.’ (11.1)

B τὰ δ᾿ αὖ σὰ ἔργαB, ἔφην ἐγώ, ἤδη μοι λέγε, (. . .).

But now tell me of your own (. . .). (11.1)

In other passages the speaker makes a metadiscursive comment in

order to state the end of the previous topic (‘so far we have been
speaking about topic X’ and the like). In the following example taken
from Euripides Apollo addresses Menelaos and speaks to him about
Helen. When he finishes this point (Helen), he states it explicitly (see
the metalinguistic expression τὰ μὲν καθ᾿ Ἑλένην ὧδ᾿ ἔχει ‘that is how
things stand with Helen’); simultaneously he redundantly40 makes use
of the particle μέν (see § 3.4.1). Once the previous topic is closed he
proceeds to the next one, viz. Orestes, marked off by αὖ, whom he
addresses and tells what to do:

40
The metadiscursive comment may appear without the particle and viceversa.
 the particles ay and ayte as topicalizing devices 105

(17) E. Or. 1629–45:

A ἙλένηνA μὲν ἣν σὺ διολέσαι πρόθυμος ὢν (. . .).

As for Helen, whom you were eager to kill in your anger (. . .).
(1629–43)

ACLOSING τὰ μὲν καθ᾿ ἙλένηνA ὧδ᾿ ἔχει·

That is how things stand with Helen. (1643)

B σὲB δ᾿ αὖ χρεών, Ὀρέστα, γαίας τῆσδ᾿ ὑπερβαλόνθ᾿ ὅρους

Παρράσιον οἰκεῖν δάπεδον ἐνιαυτοῦ κύκλον.
But you, Orestes, must cross this land’s frontier and for the circuit
of a year live in the plain of Parrhasia. (1644–5)

Another frequent way to close a topic is to use a subordinate clause, a

participle or an absolute genitive that summarizes the previous section
(repeating its content), before the new one is opened:

(18) Pl. R. 362d1–3:

A Glaucon’s speechA (361d8–362c8)

A ταῦτ᾿ εἰπόντος τοῦ ΓλαύκωνοςA,

CLOSING
When Glaucon had thus spoken (362d1)

B ἐγὼB μὲν αὖ ἐν νῷ εἶχόν τι λέγειν πρὸς ταῦτα, ὁ δὲ ἀδελφὸς αὐτοῦ

ἈδείμαντοςB, οὔ τί που οἴει, ἔφη, ὦ Σώκρατες, ἱκανῶς εἰρῆσθαι
περὶ τοῦ λόγου;
I had in mind to make some reply thereto, but his brother
Adeimantus said, ‘You surely don’t suppose, Socrates, that the
statement of the case is complete?’ (. . .). (362d1–3)

As can be deduced from the examples, the three devices described

can—but do not have to—be used simultaneously to increase text
cohesion.
106 antonio r. revuelta puigdollers

4. Summary and further research

The purpose of this paper has been to show one of the uses of the
particles αὖ and αὖτε. In one of their uses both particles mark the
introduction of a thematic discontinuity (different topic), whose effect
on the discourse is the opening of a new thematic section. As a result,
these particles work as boundary-markers that highlight the transition
between different discourse units and therefore both contribute in a
very clear way to the cohesion of discourse in Ancient Greek.
Apart from the thorough discussion of the focusing and argumenta-
tive uses of both particles, many other open questions remain. One of
these questions is the way topic is managed in Ancient Greek and the
role that other discourse connectors apart from αὖ/αὖτε play in this
task. We have seen that οὖν is used more frequently than αὖ/αὖτε for
resuming a topic that has temporarily been given up by the speaker.
Other connectors compete with αὖ/αὖτε in highlighting the introduc-
tion of topics into the discourse, as for example γε μήν:

(19) φράσω δὲ καὶ τὸ πλῆθος ἑκατέρωνA+B. συνελέγησαν γὰρ ὁπλῖται

ΛακεδαιμονίωνA μὲν εἰς ἑξακισχιλίους, (. . .) αὕτη μὲν δὴ <ἡ> μετὰ
Λακεδαιμονίων δύναμιςA ἦν. ἥ γε μὴν τῶν πολεμίωνB ἡθροίσθη
Ἀθηναίων μὲν εἰς ἑξακισχιλίους ὁπλίτας, (. . .) αὕτη μὲν δὴ
ἑκατέρων ἡ δύναμιςA+B ἐγένετο.
And now I will state the numbers on either side. As for the
hoplites, there had gathered together about six thousand of the
Lacedaemonians (. . .) This, then, was the force on the side of the
Lacedaemonians. But the force of the enemy which was gathered
together included, about six thousand hoplites of the Athenians
(. . .) This, then, was the force on either side. (X. HG. 4.2.16–8)

The hierarchical structure of this passage could be represented as in the

figure below. The particles appear distributed along the text marking
the boundaries between the topics and the units that discuss them.
As the figure clearly shows, the combination γε μήν behaves as the
particles αὖ/αὖτε in many of the examples discussed in this article
introducing the second or further element of an Announced Topic
(e.g. example 2).
 the particles ay and ayte as topicalizing devices 107

Complex movement

Central movement Subsidiary movement 1 Subsidiary

(elaboration of the movement 2
topics)

γάρ

A+B A B A+B

Announced Topic Topic 1 Topic 2 Topic Closing

The troops of both The Lacedaemonian The Athenian The troops of

side troops troops both sides

Opening Closing

μὲν δὴ γε μὴν μὲν δὴ

The particle γάρ, on the other hand, plays a different role in topic man-
agement. It seems very suitable for introducing a topic elaboration (see
example 19)41 or an excursus about an entity with (low) topicality in the
previous discourse unit, as happens in the following example, where ὦν
(= οὖν) is used by the speaker to resume the main topic (the Colchians)
after the parenthesis about the Caucasus introduced by γάρ:42

41
The unit introduced by γάρ redeems the promise evoked by φράσω (metadiscursive
expression) and at the same time it elaborates on the description and enumeration of
the military forces of both sides (τὸ πλῆθος ἑκατέρων) that have been mentioned in
the previous chapters.
42
The use of γάρ for marking the beginning of an excursus or an elaboration on
any previous topic is a side effect of its more general use as a PUSH particle (see
Polanyi & Scha 1983): that is, a particle that appears where a discourse embedded
section begins.
108 antonio r. revuelta puigdollers

(20) ΚόλχοιA δὲ ταξάμενοι ἐς τὴν δωρεὴν καὶ οἱ προσεχέες μέχρι

Καυκάσιος ὄρεοςb (ἐς τοῦτο γὰρ τὸ ὄροςB ὑπὸ Πέρσῃσι ἄρχεται,
τὰ δὲ πρὸς βορέην ἄνεμον τοῦ Καυκάσιος Περσέων οὐδὲν
ἔτι φροντίζει), οὗτοιA ὦν δῶρα τὰ ἐτάξαντο ἔτι καὶ ἐς ἐμὲ διὰ
πεντετηρίδος ἀγίνεον, ἑκατὸν παῖδας καὶ ἑκατὸν παρθένους.
The Colchians also had set themselves among those who brought
gifts, and with them those who border upon them extending as far
as the range of the Caucasus (for the Persian rule extends as far as
these mountains, but those who dwell in the parts beyond Caucasus
toward the North Wind regard the Persians no longer),—these, I
say, continued to bring the gifts which they had fixed for them-
selves every four years even down to my own time, that is to say,
a hundred boys and a hundred maidens. (Hdt. 3.97.14–9)

Discourse

A B A

Κόλχοι τὸ ὄρος Καυκάσιος οὗτοι

Given Topic Excursus Resumed Topic

γάρ ὦν

Which are the exact differences between αὖ/αὖτε, γε μήν, γάρ and
οὖν? Do other particles and connectors contribute to topic manage-
ment in Ancient Greek? Although these and other questions are still
to be answered, a preliminary version of topic management in Ancient
Greek and the role that particles play in it could be described in the
following table, which summarizes part of the observations scattered
in this article:
 the particles ay and ayte as topicalizing devices 109

Table 1: The role of particles in topic management

Particles Topic typology
New Topic Given/Inferrable Topic Resumed Excursus Topic
Topic Closing Topic elaboration
αὖ/αὖτε not absolute 2nd/nth topic of side effect
first topic of a topic chain: of the main
discourse ≠ (a) announced value (topic
‘once upon a topic change)
time’ (b) promoted
topic
(c) subtopic
(inferred topic)
γε μήν ? 2nd/nth topic of
a topic chain
μέν +
μὲν δή +
μὲν οὖν +
γάρ + +
οὖν/ὦν +
...
 CHAPTER SIX

ΚΑI ΜHΝ, ΚΑI ΔH AND HΔΗ IN TRAGEDY AND COMEDY

A. Maria van Erp Taalman Kip

1. Introduction1

My starting point is Sophocles Ajax 540–4. Aias wants to see his son.
Tekmessa has assured him that the child is nearby, guarded by servants,
whereupon Aias impatiently asks:
(1) Αι. τί δῆτα μέλλει μὴ οὐ παρουσίαν ἔχειν;
Τεκ. ὦ παῖ, πατὴρ καλεῖ σε. δεῦρο προσπόλων
ἄγ’ αὐτὸν ὅσπερ χερσὶν εὐθύνων κυρεῖς.
Αι. ἕρποντι φωνεῖς, ἢ λελειμμένῳ λόγου;
Τεκ. καὶ δὴ κομίζει προσπόλων ὅδ’ ἐγγύθεν.
Ajax Why then the delay in his appearance?
Tec. My son, your father calls you. Bring him here, servant, you
who lead him by the hand.
Ajax Is he coming, the man to whom you call, or does he fail to
hear your word?
Tec. Here comes the servant bringing him near. (S. Aj. 540–4)2
It is interesting to note what has been said about καὶ δή (544) by various
scholars. In Jebb’s commentary we find: ‘καὶ δή “already”: OC 31 n.’
Kamerbeek comments: ‘καὶ δή: non-connective; “signifies, vividly and
dramatically, that something is actually taking place at the moment”
(Denniston, G.P., 250). It is here not very different from καὶ μήν: cf. Eur.
Medea 1118: καὶ δὴ δέδορκα τόνδε τῶν ᾿Ιάσονος / στείχοντ’ ὀπαδῶν.’
Stanford has: ‘καὶ δή: as often = “Well, here you are (. . .)”, the French
et voilà; cf. Denniston 251).’ And Garvie: ‘both καὶ δή (more often καὶ
μήν; see Denniston 251) and the deictic ὅδε regularly mark the arrival
of a new character on stage.’

1
I am grateful to the editors of this volume for their advice and their support.
2
Translation Garvie (1998). The other translations from Ajax are likewise his.
112 a. maria van erp taalman kip

The influence of Denniston is immediately clear. It is only Jebb,

writing in the pre-Denniston era, who renders καὶ δή as ‘already’. But
Denniston could hardly have objected to this, since he himself (252)
argues that καὶ δή ‘frequently approximates in sense (. . .) to ἤδη, though
it is always more vivid and dramatic in tone.’ As an example he men-
tions e.g. OC 31, the same line referred to by Jebb. But if it is true that
καὶ δή may have approximately the same value as ἤδη, Denniston’s
decision to explain Ajax 544 differently, seems a bit arbitrary, since
‘already’ would suit the context very well.
However, I shall first focus on the relation between καὶ δή and καὶ
μήν. Stanford and Garvie refer to Denniston (1954: 251), where he says
of καὶ δή: ‘sometimes used (as, far more often, καὶ μήν) to mark the
entrance of a character on the stage.’ Kamerbeek, though not referring
to the same page, also parallels καὶ δή with καὶ μήν. Nevertheless, it
is improbable that the dramatists chose the one or the other particle
combination at random. If καὶ μήν is the rule, why then an exception
in Ajax 544? There must be a difference, which is somehow related to
the context, and thus to the discourse coherence and the way in which
that coherence is marked by linguistic means.

2. Entry markings in tragedy

2.1. καὶ μήν

In her admirable article on the particle μήν Wakker argues that it has
a strongly affirmative (and corrective) value; it is ‘especially at home in
adversative contexts: it corrects or eliminates the previous statement or
its implications’ (1997: 229).3 The use of entry-marking καὶ μήν fits in
with this interpretation. In her view the difference between entry-mark-
ing καὶ μήν and entry-marking καὶ δή ‘seems to be that the former is
connected with unexpected events, while the latter is connected with
events that are expected.’ She points out that καὶ μήν is never used
when the entrance has been announced in the preceding lines: ‘Καὶ μήν
thus expresses, at the interactional level, “really, there we have” (the

3
Cf. Sicking (1993: 54): ‘The particle μήν seems to be at home in expressing the
contrary of what the person addressed might either (1) suppose or (2) wish.’ Although
Wakker is not satisfied by this description, it certainly points in the same direction.
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 113

speaker expresses his surprise that at that very moment . . . is coming

and wants his adressee(s) to share that feeling). On the other hand, at
the presentational level, μήν highlights a—surprising—new turn in the
course of the events/discussion’ (1997: 228). I think this is essentially
correct, but I would like to explore the matter in somewhat greater
detail, not only in tragedy, but also in comedy.
Let us compare Ajax 544 with 1168. Urged by the Chorus, Teukros is
about to leave the stage to look for a place where Aias may be decently
buried. But there is a short delay, caused by the arrival of Tekmessa
and the child. Teukros marks their arrival by saying:
(2) καὶ μὴν ἐς αὐτὸν καιρὸν οἵδε πλήσιοι
πάρεισιν ἀνδρὸς τοῦδε παῖς τε καὶ γυνή, (. . .).
Look in the very nick of time here come his child and wife, (. . .).
(S. Aj. 1168–9)
After giving instructions to the child and the Chorus, he leaves as
planned, and the Chorus sings the third stasimon. This entry occurs
at the end of an epeisodion, shortly before a choral song. It does not
result in a new turn in the course of events or the discussion, nor is
it wholly unexpected, since Teukros has sent Tekmessa away to fetch
the child (985–9). If there is surprise, it is at most, as Wakker (1997:
228) puts it, surprise that they have come at this very moment. We find
a comparable case in OC 549, where the Chorus, after an emotional
amoibaion with Oedipus, announce the entry of Theseus as follows:
(3) καὶ μὴν ἄναξ ὅδ’ ἡμῖν Αἰγέως γόνος
Θήσευς κατ’ ὀμφὴν σὴν ἀποσταλεὶς πάρα.
See, here is our king, the son of Aegeus, who was summoned accord-
ing to your words. (S. OC 549–50)4
Here, too, the entry is not really unexpected, since the Chorus has
assured Oedipus that Theseus will come as soon as he hears of Oedipus’
presence in Kolonos. And although they could not know that he would
come at exactly this moment, there is no reason for great surprise. This
is even less the case in 1249, where Antigone announces the arrival of
Polyneikes:

4
Translation Lloyd-Jones (1991–4). The other translations from tragedies by
Sophocles, Ajax excluded, are also his.
114 a. maria van erp taalman kip

(4) καὶ μὴν ὅδ’ ἡμῖν, ὡς ἔοικεν, ὁ ξένος (. . .).

Why, here, it seems, is the stranger! (S. OC 1249)
Oedipus has reluctantly agreed to a meeting with his son (1204–7). Both
he and Antigone know that he is on his way, so his arrival is surely not
unexpected, and cannot be a surprise. However, we are not prepared
for it in the preceding lines, since 1240 follows on a choral ode.
The difference between the above examples and Ajax 544 seems to be
that the arrival of Eurysakes is the topic of the discussion in the lines
that immediately precede. Aias has asked for his son in 530. Tekmessa
has explained why he is absent, she has summoned a servant to bring
him to his father, and then, in 544, he is actually there. With καὶ μήν
there is a shift in the focus of attention; with καὶ δή there is no such
shift.

2.2. καὶ δή
Entry-marking καὶ δή is rarely found in tragedy. Euripides Medea
1118, referred to by Kamerbeek, is a case in point. After an anapestic
intermezzo Medea says to the Chorus:
(5) φίλαι, πάλαι τοι προσμένουσα τὴν τύχην
καραδοκῶ τἀκεῖθεν οἷ προβήσεται.
καὶ δὴ δέδορκα τόνδε τὸν Ἰάσονος
στείχοντ’ ὀπαδῶν (. . .).
My friends, for a long time now I have been expecting the event,
waiting to see how matters in that quarter will turn out. And look,
here I see one of Jason’s servants coming (. . .). (E. Med. 1116–9)5
Medea has heard from the paedagogus (1002–4) that Jason’s bride has
gratefully accepted her gifts. She is certain now that her plan has suc-
ceeded and that the princess must already be dead. She is only waiting
for confirmation, as is clear from 1116–7, and then she sees the mes-
senger who will no doubt bring this confirmation.
Ajax 544 and Medea 1118 are quoted by Denniston as examples of
entry-marking καὶ δή, but, as we have seen, he considers καὶ δή in OC
31 the equivalent of ἤδη. However, this line parallels with the lines
from Ajax and Medea. Oedipus is not sure that the region he and his

5
Translation Kovacs (1994–2002). The other translations from plays by Euripides
are also his.
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 115

daughter have entered is inhabited (line 27). Antigone assures him it

is; there is no need for any investigations, since she sees a man nearby.
Oedipus asks whether he is coming in their direction, whereupon she
answers: καὶ δὴ μὲν οὖν πάροντα (. . .). Here, too, καὶ δή marks the
actual arrival of a man whose appearance has been prepared for in the
foregoing lines.
I have searched in vain for other examples from tragedy,6 but there
is one in Euripides’ satyr-play Cyclops. The Chorus has endorsed
Odysseus’ plan to blind the Cyclops with a burning olive branch and
has promised to help him carry it out. In the anapests that follow they
discuss the order in which they will line up when launching the attack,
but in 488 they interrupt these speculations:
(6) σίγα σίγα. καὶ δὴ μεθύων
ἄχαριν κέλαδον μουσιζόμενος
σκαιὸς ἀπωιδὸς καὶ κλαυσόμενος
χωρεῖ πετρίνων ἔξω μελάθρων.
φέρε νιν κώμοις παιδεύσωμεν
τὸν ἀπαίδευτον.
πάντως μέλλει τυφλὸς εἶναι.
Hush! Hush! For now the Cyclops, drunk and making graceless
melody, comes forth from the rocky cave, a singer who is inept and
who shall pay dearly. Come, let us with our revelling songs impart
some culture to this trout. In any case he shall be blind. (E. Cyc.
488–94)
Seaford does not comment on καὶ δή in 488, while Ussher offers the
usual explanation: ‘look, here he comes, a rarer use than καὶ μήν (. . .)
to mark a new arrival on the stage. See (for example) Med. 1118, Ar.
Ec. 500, Denniston 251.’ But why καὶ δή? In this case, the arrival of
the Cyclops has not been referred to in the preceding lines, but it has
been prepared for immediately before. Manuscript L notes, just before
488, ᾠδὴ ἔνδοθεν. Seaford and even Taplin (1977: 15, note 1) deem it
possible that this stage direction goes back to Euripides himself, but it
does not really matter, since—as Seaford remarks—it is ‘inferable from
the text.’ Even without a stage direction it would be obvious that the

6
Denniston mentions Euripides Supp. 1114, but I cannot believe that the text of that
line (τάδε δὴ παίδων καὶ δὴ φθιμένων) is sound. Diggle adopts Musgrave’s ἤδη instead
of καὶ δή. In Aeschylus we find no examples of entry-marking καὶ δή and only one of
entry-marking καὶ μήν (Th. 327). In Sophocles we find entry-marking καὶ μήν eight
times and two times καὶ δή, in Euripides two times καὶ δή and 24 times καὶ μήν.
116 a. maria van erp taalman kip

Chorus (and probably the audience as well) hears the drunken song the
Cyclops is singing in his cave. And it is this song which prepares them
for the Cyclops’ arrival. At 488 there is no shift in the focus of attention.
We may suppose that Polyphemus comes into sight during 488–94 but
lingers at the opening of the cave, too drunk and as yet too far off to
understand what the Chorus is saying. Their words are obviously not
destined for his ears, but the song that follows upon the anapests is
meant to be heard by him. By then he must be closer to them.7

2.3. Some more examples of καὶ μήν

After discussing the use of entry-marking καὶ δή in Sophocles and
Euripides, I will now continue with a number of examples of καὶ μήν
in Euripides. Sometimes καὶ μήν marks an entry that causes genuine
surprise and brings about a turn in the events. This is, for instance, the
case in Andromache 545 (arrival of Peleus) and Alcestis 1006 (arrival of
Heracles). But other examples are different. In Hippolytus 1342 Hip-
polytus’ entry is marked by καὶ μήν, although the Chorus knows, as of
1265, that he will come and must now be on his way. In the meantime,
however, Artemis has informed Theseus of Hippolytus’ innocence and
between her last words and his entry there is a clear shift in the focus
of attention. More noteworthy is Andromache 1166. A messenger has
told the Chorus of Neoptolemus’ death and the way he died. In 1158–60
he has informed them that he and his companions are bringing the
body with them and at 1166 the Chorus marks the arrival of the bier:
(7) καὶ μὴν ὄδ’ ἄναξ ἤδη φοράδην
Δελφίδος ὲκ γῆς δῶμα πελάζει.
See, here is our lord, his body carried home from the land of Delphi.
(E. Andr. 1166–7)
Here the interval between announcement and actual arrival is quite
short; nevertheless there is a shift in the focus of attention, since the
messenger’s final lines (1161–5) do not deal with the body, but with
Apollo, the god who has taken revenge on Neoptolemus, as if he were
a spiteful human being.

7
Taplin (1977: 174, note 3) assumes that the Cyclops’ entry occurs at 503 or even
511. In view of the Chorus’ καὶ δὴ (. . .) χωρεῖ I do not consider this possible.
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 117

A more difficult case is Ion 1257. A servant has told the Chorus how
Kreousa’s attempt to poison Ion has been discovered. The Chorus pan-
icks and is joined at 1250 by an equally panicky Kreousa, exclaiming:
(8) Κρ. πρόσπολοι, διωκόμεσθα θανασίμους ἐπὶ σφαγάς,
Πυθίᾳ ψήφῳ κρατηθεῖσ’, ἔκδοτος δὲ γίγνομαι.
Χο. ἴσμεν, ὦ τάλαινα, τὰς σὰς συμφοράς, ἵν’ εἶ τύχης.
Κρ. ποῖ φύγω δῆτ’; ἐκ γὰρ οἴκων προύλαβον μόγις πόδα
μὴ θανεῖν, κλοπῇ δ’ ἀφῖγμαι διαφυγοῦσα πολεμίους.
Χο. ποῖ δ’ἂν ἂλλοσ’ ἢ ’πὶ βωμόν;
Κρ. καὶ τί μοι πλέον τόδε;
Χο. ἱκέτιν οὐ θέμις φονεύειν.
Κρ. τῷ νόμῳ δέ γ’ ὄλλυμαι.
Χο. χειρία γ’ ἁλοῦσα.
Κρ. καὶ μὴν οἵδ’ ἀγωνισταὶ πικροὶ
δεῦρ’ ἐπείγονται ξιφήρεις.
Χο. ἵζε νυν πυρᾶς ἔπι.
Kr. Serving women, I have been condemned by the Delphians’
verdict! They are looking for me to put me to death! My life
is forfeit!
Cho. We know, poor lady, where you stand in misfortune.
Kr. Where shall I take refuge? I barely ran out of the house in time
to escape death. I gave my enemies the slip and came here by
stealth.
Cho. Where else but at the altar?
Kr. What good will that do to me?
Cho. It is unlawful to slay a suppliant.
Kr. But the law is putting me to death!
Cho. Yes, if it can capture you.
Kr. But see, here come on hurrying feet, sword in hand, my hated
adversaries. (E. Ion 1250–8)
Is there actually a shift in the focus of attention here? Kreousa says
that she is being pursued by her enemies, and if their actual arrival had
followed immediately after 1251 or 1252, she might well have marked
their entry by καὶ δή. But as it is, their arrival has been postponed,
interrupting the brief discussion of the safety which the altar may offer.
The shift in the focus of attention is not very marked, but apparently
just marked enough to justify καὶ μήν.

3. Entry-markings in comedy

I now turn to comedy, where καὶ δή and καὶ μήν are more evenly dis-
tributed: entry-marking καὶ δή occurs seven times, while entry-marking
118 a. maria van erp taalman kip

καὶ μήν occurs eight times.8 It is noteworthy that καὶ δή, when used in
this way, is not found at the beginning of a sentence, but is preceded
by other words or even other particles. Presumably this mid-verse posi-
tion was not imperative, since in comparable cases we do find καὶ δή
at the beginning of a sentence, as in Pax 178. Trygaios feels that he is
nearing the gods, and then his feelings are confirmed:
(9) καὶ δὴ καθορῶ τὴν οἰκίαν τὴν τοῦ Διός.
In fact, I can see the house of Zeus. (Ar. Pax 178)9
Since obviously the house does not enter the stage I have not included
this line in my total of seven, but an example like this one is, of course,
quite similar. Nonetheless, there is in this respect a difference with
tragedy, which must be related to the value of καί. I shall return to this
question in section 5.
Apart from this difference in position, the use of entry-marking καὶ δή
in comedy is largely the same as in tragedy. We find clear instances in
the prologue of Lysistrata. The heroine remarks indignantly (lines 61–2)
that even the women of Acharnai have not responded to her summons,
whereupon Kalonike, after a joke about one of them, exclaims:
(10) ἀτὰρ αἵδε καὶ δή σοι προσέρχονταί τινες.
But look, here you are, some of them are arriving now. (Ar. Lys. 65)
Henderson comments: ‘καὶ δή often marks the entrance of a new char-
acter (. . .); here an additional nuance is the fulfilment of something
anticipated or required by the circumstances (Denn. 251), (. . .).’ This is
a mix of Denniston’s paragraphs 2 (ii), on entry-marking καὶ δή, and
2 (iii): ‘marking the provision or completion of something required
by the circumstances.’ Henderson’s own addition (‘the fulfilment of
something anticipated’) is accurate, but it is not only here that this
additional nuance appears. As I am trying to show, it is found in all
cases of entry-marking καὶ δή.
Next there is the arrival of Myrrhine, who wants to know what
Lysistrata has to say. However, the latter prefers to wait until the women

8
I do not include Ranae 285 and 288, where Xanthias frightens Dionysos by
announcing the entry of imaginary monsters.
9
Translation Sommerstein (1981–98). The other translations from Aristophanes
are likewise his.
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 119

from Sparta and Boiotia have arrived (lines 74–6a), whereupon Myr-
rhine answers:
(11) πολὺ σὺ κάλλιον λέγεις.
ἡδὶ δὲ καὶ δὴ Λαμπιτὼ προσέρχεται.
Yours is a much better idea. And look, here comes Lampito now.
(Ar. Lys. 76b–7)
In 74–6a Lysistrata apparently does not doubt that the Spartan women
will also come, now that the other women are marching in. Hence
there is no need for Myrrhine to use a corrective ἀτάρ, like Kalonike
in Lysistrata 65.10
In Vespae 1299–1323 we find a very extensive preparation. After
accompanying Philokleon to a symposion, the slave Xanthias returns
somewhat earlier than his master’s father and describes to the Chorus
how scandalously the man misbehaved: he offended all the other guests
and now he is returning home, in a drunken state beating everyone he
meets. In this case we are not only prepared for an arrival, we also learn
how the newcomer will, in all probability, behave. Small wonder, then,
that the actual arrival does not result in a conversation between the
slave, who announces it, and Philokleon, who is announced. Xanthias
hastily disappears:
(12) ὁδὶ δὲ καὶ δὴ σφαλλόμενος προσέρχεται.
ἀλλ’ ἐκποδὼν ἄπειμι πρὶν πληγὰς λαβεῖν.
Look, here he comes, reeling along. I’m going to get out of the way
before I get hit. (Ar. V. 1324–5)
Out of the seven occurrences I have mentioned, two are borderline cases.
The first one is Vespae 1482. After emerging from the house, Xanthias
once again describes the misbehaviour of Philokleon, who this time
clamorously attemps to revive the old tragedy of Thespis. At this point
Philokleon himself appears and calls out from the doorway:
(13a) τίς ἐπ’ αὐλοίσι θύραις θάσσει;
Who sitteth at the outer door? (Ar. V. 1482)
Responding to this line, probably a quotation from a tragedy, Xanthias
says:

10
Comparable to Lys. 65 is Aves 268.
120 a. maria van erp taalman kip

(13b) τουτὶ καὶ δὴ χωρεῖ τὸ κακόν.

This really gets worse and worse. (Ar. V. 1483)
We find these same words in Nubes 906–7 and, in a slightly different
order, in Ranae 1018. However, in these cases the κακόν consists in
the annoying words of the speaker’s interlocutor; it is purely abstract.
But in Vespae Philokleon’s annoying words conincide with his appear-
ance. It is possible, therefore, to see this as an entry-marking καὶ δή,
as MacDowell does, referring to Vespae 1324 (example 13). But there
is a difference, since in 1482 Xanthias speaks his καὶ δή after the new-
comer has spoken. This seems to be ‘against the rules’, so I think that
in this line καὶ δή is a reference to Philokleon’s words rather than to
his appearance.
A second borderline case is Ranae 604. Dionysus’ slave Xanthias has
once again taken his master’s Heraclean gear, but he is worried about
the consequences. However, he resolves to bear them courageously
and then continues:
(14) δεῖν δ’ ἔοικεν, ὡς ἀκούω
τῆς θύρας καὶ δὴ ψόφον.
And it seems I’ll need to, because listen, I hear the door creaking.
(Ar. Ra. 604)
Immediately after these words the janitor appears and orders his ser-
vants to arrest him. Although Xanthias does not know exactly who is
about to appear at the door, he does realize that the newcomer means
danger. Καὶ δή serves to mark the sound of the creaking door, a sound
that confirms his forebodings, since it clearly announces an entry.
Strictly speaking, the particles mark the sound rather than the appear-
ance of the janitor, but in any case the contrast with, say, Sophocles
Electra 78 is clear. There the paedagogus reacts to the sound of Electra’s
off stage lament with the words:
(15) καὶ μὴν θυρῶν ἔδοξα προσπόλων τινός
ὑποστενούσης ἔνδον αἰσθέσθαι, τέκνον.
Why, I thought I heard one of the slaves behind the door groaning,
my son! (S. El. 78–9)
Electra’s lament comes fully unannounced. In this case καὶ δή could
not possibly be used.
My study of all the instances in Aristophanes has convinced me that
the criterion I proposed holds true: καὶ δή marks an entry that has been
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 121

prepared for in the preceding lines, καὶ μήν does not; καὶ μήν implies a
shift in the focus of attention, καὶ δή does not. But before concluding
this part of my argument, I feel obliged to discuss, as I did in the case of
tragedy, the most troublesome example of καὶ μήν, viz. Vespae 899.
After Bdelykleon’s announcement that the trial of the dog is about
to start, Philokleon asks:
(16a) τίς ἆρ’ ὁ φεύγων οὗτος; ὅσον ἁλώσεται.
So who is this defendant? How thoroughly he’ll be convicted!
(Ar. V. 893)11
Then Bdelykleon reads the complaint that Kyon has lodged against
Labes: he has eaten all the Sicilian cheese himself; the punishment must
be a figwood collar. But Philokleon corrects him:
(16b) θάνατος μὲν οὖν κύνειος, ἢν ἅπαξ ἁλῷ.
No, a dog’s death, if once he is found guilty. (Ar. V. 898)
In the next line Bdelykleon marks the entry of Labes:
(16c) καὶ μὴν ὁ φεύγων οὑτοσὶ Λάβης πάρα.
And here is the defendant, Labes, present. (Ar. V. 899)
Labes has been the subject of conversation in the preceding lines and it
is only to be expected that the defendant will be present when his case
is heard. Nevertheless Bdelykleon, who is running this show, suggests
a shift in the focus of attention, marking his entry with καὶ μήν. And
perhaps there is such a shift, since Philokleon is not really interested
in Labes’ arrival. He has not asked where he is or when he will arrive,
since he does not need to hear him. He has never in his life acquitted
a defendant and this one, too, has already been found guilty. I suggest,
therefore, that Bdelykleon’s καὶ μήν implies: I am not interested in your
offhand opinions, we are going to focus on a regular trial. The effect
may well be humorous.

11
Dobree has suggested the following reading: Phil. τίς ἆρ’ ὁ φεύγων; Bd. οὗτος. This
reading was adopted in the Budé-edition, but has been rightly rejected by MacDowell
and Sommerstein. According to Sommerstein, Labes already makes his appearance
between 893 and 899, during the reading of the complaint. In that case καὶ μήν would
not mark his entry but rather his arrival at the place reserved for the defendant.
122 a. maria van erp taalman kip

4. Other instances of καὶ δή and καὶ μήν

It is not only in the case of entry-marking καὶ δή that Denniston,

by means of a cross reference, suggests that καὶ δή and καὶ μήν are
interchangeable. On pp. 250–1 he describes another use of καὶ δή as
follows: ‘In general marking vivid perception by mind, ear or eye: “lo”,
“hark”, “see there!”. (Cf. καὶ μήν, (7).’ And on καὶ μήν (7) we read:
‘(. . .) calling attention to something just seen or heard. “See!”. “Hark!”.
(Cf. καὶ δή, 2.i).’ However, a comparison of his examples reveals the
difference. Compare, for instance, Euripides Hercules 867 with Bac-
chae 918, In Hercules Lyssa is describing how she will drive Herakles
mad and what he will do in his madness. This madness is at the core
of her speech, so it does not come as a surprise when she mentions
its initial signs:
(17) ἢν ἰδού· καὶ δὴ τινάσσει κρᾶτα βαλβίδων ἄπο (. . .).
See. He has left the starting gate. He shakes his head about (. . .).
(E. HF 867)
The Chorus cannot actually see this, since Herakles is inside the palace,
but Lyssa can.
In Bacchae Dionysus has assured Pentheus that he looks like one of
Pentheus’ daughters, whereupon Pentheus says:
(18) καὶ μὴν ὁρᾶν μοι δύο μὲν ἡλίους δοκῶ (. . .)
Look, I seem to see two suns in the sky! (E. Ba. 918)
This statement has nothing to do with Dionysus’ words about his
appearance. There is a clear-cut shift in the focus of attention, as in
Sophokles Elektra 78, already discussed above.
A comparison can also be made between Aristophanes Thesmo-
phoriazusae 769 and Aves 1462. In example (19) Euripides’ relative
is wondering how he can send a message to Euripides, and then
continues:
(19) οἶδ’ ἐγὼ καὶ δὴ πόρον
ἐκ τοῦ Παλαμηδοῦς (. . .).
Ah yes, I know a trick from his Palamedes (. . .). (Ar. Th. 669–70)
He is searching for a solution and he finds one.12

12
The words of the relative are not meant to be heard by the characters on stage.
He is addressing himself, or perhaps the audience.
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 123

In Aves 1462, on the other hand, the sycophant is longing to be given

wings by Peisetairos, and he explains to him the ignoble use he intends
to make of them. He believes that his interlocutor applauds his scheme
and, pleased with himself, he concludes:
(20a) βέμβικα οὐδὲν διαφέρειν δεῖ.
One has to be no different from a whipping-top. (Ar. Av. 1461)
Hereupon Peisetairos retorts:
(20b) μανθάνω
βέμβικα. καὶ μὴν ἐστί μοι, νὴ τὸν Δία,
κάλλιστα Κορκυραῖα τοιαυτὶ πτερά.
I understand your ‘whipping-top’; and in fact, by Zeus, I have
some splendid wings of just that kind—Corcyraean ones! (Ar.
Av. 1461–3)
With καὶ μήν he steers the conversation in quite an unexpected direc-
tion. Instead of providing the sycophant with wings, he is going to
chase him with a whip.
There is one more use of καὶ δή and καὶ μήν I would like to discuss.
On p. 251 (iii) Denniston mentions the use of καὶ δή ‘in response to a
definite command, often with a word of the command echoed.’ This is a
well-known use, also discussed by Wakker (1997: 216–7), who points out
that ‘expressing the assent means at the same time performing the state
of affairs assented to.’ We find this use in both tragedy and comedy. Here
I will concentrate on a command (or permission) to speak. In Sophocles
Antigone 245 the guard, after being pressed by Kreon, finally delivers his
message with the opening words καὶ δὴ λέγω σοι. In Sophocles Electra
558, Electra opens her speech with the same words, when Clytaemnestra
has permitted her to speak, and later on in the play the opening words of
Chrysothemis, in response to Electra’s half-hearted permission, are: καὶ
δὴ λέγω σοι πᾶν ὅσον κατειδόμην (line 892). In Nubes 778 Strepsiades
reacts on Socrates’ order εἰπὲ δή with καὶ δὴ λέγω, and in Equites 22
Slave A obeys the order of Slave B (λέγε δὴ· μόλωμεν) by echoing him:
καὶ δὴ λέγω· μόλωμεν. In all these cases καὶ δή expresses agreement
with the interlocutor’s order, while the clause that is introduced by the
particle combination does not as yet contain information pertaining to
the content of the story that is going to be told.13

13
The only exception is Ar. Aves 555. The Chorus has asked Peisetairos what they
must do to regain their empire, and his answer starts with the following sentence: καὶ
δὴ τοίνυν πρῶτα διδάσκω μίαν ὀρνίθων πόλιν εἶναι, . . . However, this juxtaposition
124 a. maria van erp taalman kip

So far, so good. But what about the use of καὶ μήν described by
Denniston (1954: 355)? There we read: ‘A person who has been invited
to speak expresses by the particles his acceptance of the invitation:
“Well”, “Very well”, “All right”.’ He does not refer to καὶ δή here, but
again we are justified in searching for the difference between the two
particle combinations. Obviously it has nothing to do with the difference
between a command and and an invitation, especially since, more often
than not, the invitations are expressed by means of an imperative.
It is noteworthy that this use of καὶ μήν is largely confined to comedy
and Plato.14 According to Denniston, the only examples from tragedy
are Aeschylus Agamemnon 1178 and Sophocles OT 345, but even these
examples are highly questionable. In Agamemnon the old men of the
Chorus, although convinced that Kassandra’s prophecies are inspired
by some malign power, fail to understand their meaning: τέρμα δ’
ἀμηχανῶ. At this point Kassandra starts a rhesis that opens with καὶ
μήν (. . .). But I think Fraenkel (1950: ad loc.) is right when he says of
the Chorus’ words: ‘all I am able to perceive is the ring of complete
resignation.’ It is rather arbitrary to consider them, as Denniston does,
an ‘implied appeal for plainer speaking.’
OT 345, too, is a doubtful example. The passage runs as follows.
Teiresias angrily says:
(21a) οὐκ ἂν πέρα φράσαιμι. πρὸς τάδ’, εἰ θέλεις,
θυμοῦ δι’ ὀργῆς ἥτις ἀγριωτάτη.
I will explain no further; in the face of that, pray rage with the
most ferocious anger! (S. OT 343–4)
Thereupon Oedipus retorts:
(21b) καὶ μὴν παρήσω γ’ οὐδὲν, ὡς ὀργῆς ἔχω,
ἅπερ ξυνίημ’ (. . .).
Well, I am so angry that I will leave untold nothing of what I
understand. (S. OT 345–6)
Kamerbeek refers here to Denniston (1954: 355 (5)) and comments:
‘Oedipus readily accepts the challenge of Teiresias’ words.’ It is indeed
a challenge; it can hardly be called an invitation, let alone an invitation

of καὶ δή and τοίνυν is highly unusual; according to Denniston (1954: 578, note 1)
it only occurs two times in Gorgias’ Palamedes. The sentence seems to be a kind of
contamination of καὶ δὴ διδάσκω (. . .) and πρῶτα τοίνυν διδάσκω (. . .).
14
For this reason it is not discussed by Wakker.
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 125

to speak. Teiresias puts an end to the altercation, no matter how furious

Oedipus will be. And Oedipus makes it clear that if Teiresias thinks
he is intimidated, he is mistaken. He will say everything he has on his
mind. This cannot be called an ‘acceptance of an invitation to speak’,
but it fits in with Wakker’s assertion that καὶ μήν ‘corrects or eliminates
the previous statement or its implications’ (1997: 229).
Let us now turn to comedy. Denniston has twelve examples. I
have found no others, but since I do not include Aves 639, the total
is eleven.15 In most cases the ‘invitation’ originates from the Chorus.
The only exception is Ranae 1248, where it is Dionysus who, now fed
up with Aeschylus’ ληκύθια, orders Euripides to turn to the songs of
his opponent.
The invitations, or rather exhortations, of the Chorus are found
mainly in the (anti)katakeleusmos of an agoon, but two times in a song
(Equites 620–3 and Ranae 1099–1118). In nine cases the Chorus uses an
imperative or χρή/χρῆν, in one case (Nubes 1034–5) there is no formal
exhortation at all, but rather a warning. Here the Chorus points out to
the ἥττων λόγος that his task will be very difficult, but the ἥττων λόγος
is not impressed and makes it clear that the Chorus is mistaken:
(22) καὶ μὴν πάλαι ᾿γὼ ’πνιγόμην τὰ σπλάγχνα κἀπεθυμοῦν
ἅπαντα ταῦτ’ ἐναντίαις γνωμαῖσι συνταράξαι.
Well, actually I’ve been choking in my insides for some time with
eagerness to make mincemeat of all this by counterarguments. (Ar.
Nu. 1036–7)
His task is not difficult at all; on the contrary, he cannot wait to vent
the numerous arguments he has in store.
This combative tone is characteristic of the καὶ μήν clauses discussed
here. In most cases the characters are exhorted to start a verbal contest
with a third party,16 and by καὶ μήν they express their eagerness to attack,
trying to eliminate beforehand any possible doubt about their ability to
win the contest. So for instance Strepsiades in Nubes. The Chorus asks
him to tell them how his quarrel with Pheidippides started:
(23a) ἀλλ’ ἐξ ὅτου τὸ πρῶτον ἠρξαθ’ ἡ μάχη γενέσθαι
ἤδη λέγειν χρὴ πρὸς χόρον· πάντως δὲ τοῦτο δράσεις.

15
Equites 335, 624, Nubes 1036, 1353, Vespae 548, Aves 462, Lys. 486, Ranae 907,
1249, 1119. Eccl. 584. As for Aves 639 Dunbar, ad 462, rightly notes that we find here
‘a related but distinct use’.
16
Exceptions are Equites 624 and Aves 462; see below.
126 a. maria van erp taalman kip

Now you must tell the Chorus what the quarrel originally arose
from. But you’ll do that anyway. (Ar. Nu. 1351–2)
Whereupon Strepsiades answers:
(23b) καὶ μὴν ὅθεν γε πρῶτον ἠρξάμεσθα λοιδορεῖσθαι
ἐγὼ φράσω.
I certainly will tell you what first caused us to to begin using hard
words. (Ar. Nu. 1353–4)
At first glance this is a simple acceptance of an invitation to speak.
He is asked to tell a story and he will do so. He even picks up on the
Chorus’ λέγειν χρή by φράσω, and their ἐξ ὅτου τὸ πρῶτον ἤρξαθ’ ή
μάχη γενέσθαι by ὅθεν γε πρῶτον ἠρξάμεσθα λοιδορεῖσθαι. Neverthe-
less, this story is the opening of his contest with Pheidippides, and in
its introductory song (1345–50) the Chorus has cast doubt on Strepsia-
des’ chances. But Strepsiades is convinced that the outrageousness of
his story will be a strong argument in his favour. With καὶ μήν and a
combative ἐγώ he counters the Chorus’ doubts.
In the clauses that open with καὶ μήν we never find a simple accep-
tance of an invitation to speak. In Ecclesiazusae 583–4 Praxinoa is fully
convinced that she has strong arguments, but she is afraid that the audi-
ence will be refractory to new ideas—a danger not mentioned by the
Chorus. In Aves 462 Peisetairos is more than willing to tell the Chorus
why he has come to them, but he emphasizes the fact that it is not, as
the Chorus has suggested, his fault that he has not told them before. My
last example is Equites 624. The saugage-seller has returned victorious
from the Boulè, and the Chorus is longing to hear his story:
(24a) ὡς ἐγώ μοι δοκῶ
κἂν μακρὰν ὁδὸν διελθεῖν
ὥστ’ ἀκοῦσαι. πρὸς τάδ’, ὦ βέλ-
τιστε, θαρρήσας λέγ, ὡς ἅ-
παντες ἡδόμεσθά σοι.
For I fancy I’d be willing to make a long jouney to hear it. So,
admirable friend, take courage and speak; we’re all delighted with
you. (Ar. Eq. 620–3)
And the sausage-seller answers:
(24b) καὶ μὴν ἀκοῦσαί γ’ ἄξιον τῶν πραγμάτων.
Well, it’s certainly worth hearing the story. (Ar. Eq. 624)
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 127

After this introductory line he embarks on his tale.

The sausage-seller is asked to relate what has happened and he is
eager to do so; there is no adversary who must be attacked and he
is in agreement with the Chorus. Here the difference with the καὶ δή
examples seems to be purely formal: the sausage-seller does not agree
to do anything. He does not respond to the Chorus’ λέγε by saying
something like λέγω; he links up with ὥστ’ ἀκοῦσαι. But he may also
be responding to the Chorus’ somewhat paternalizing θαρρήσας. He
himself is fully aware that his story is worthwhile.

5. The basic values of καὶ μήν and καὶ δή

The uses of καὶ μήν discussed above fit in with the basic value of μήν
as outlined by Wakker (see 2.1). Where the basic value of δή is con-
cerned, things are less clear. According to Wakker, the speaker uses
δή to demand ‘the addressee’s special attention for the (interesting and
important) proposition’ he presents (1997: 216). She does not agree with
the view of Sicking, who speaks of ‘the aptness of δή to convey to the
hearer the speaker’s suggestion that the two of them share information
in a wider sense including opinions’ (1993: 52). This view is shared
by Van Ophuijsen, but Wakker argues that ‘obviously’ etc. is not the
general or basic meaning of δή: ‘this semantic nuance is rather a later
development of δή, which occurs only in specific contexts’ (1997: 216,
note 17). Nonetheless this nuance is perhaps not incompatible with the
uses of καὶ δή discussed above, since the speaker suggests that he has
common ground with his addressee; he assumes that the other person
has been prepared for what he is going to hear by what has been said
before.
Another problem is the value of καί. According to Denniston, in all
the cases of καὶ δή discussed above, it is non-connective, while in καὶ
μήν it is always connective. As for καὶ μήν Wakker agrees, but not in
the case of καὶ δή, as is clear from her discussion of καὶ δὴ λέγω and
the like: ‘καί expresses (. . .) the close link with the previous utterance’
(1997: 216–7). I think she is right and that originally καί, in combination
with δή, was connective. But I assume that later on the combination
was felt as a unity, so much so that it could be removed from its posi-
tion at the opening of a sentence. We have seen that this is not unusual
in comedy, where it may even be preceded by other particles such as
128 a. maria van erp taalman kip

ἀτάρ, δέ or ἀλλά. We do not find this position in tragedy, with one

exception: Sophocles OC 173. When Oedipus asks Antigone to touch
him, she answers: ψαύω καὶ δή.17
In conclusion: Wakker is correct in maintaining that δή and μήν
are essentially different from one another, and so are καὶ δή and καὶ
μήν. In the case of καὶ μήν we find a shift in the focus of attention:
the speaker marks an entry that is not prepared for by the words that
immediately precede it or he embarks on a new subject. When there
is no such shift, the speaker corrects or contradicts his addressee.
And since καὶ μήν marks some kind of incision, it is at home at the
beginning of a sentence. These characteristics are absent in καὶ δή. It
is used to mark an entry that has been prepared for by the words that
immediately precede it, or something said or done which is related to
the subject under discussion. And when the speaker marks his agree-
ment, he does so without contestation. It is probably by reason of these
differences that καὶ δή, unlike καὶ μήν, can forsake its original position
at the beginning of the sentence.

Appendix

Leaving aside καὶ μήν, there is one more problem that I would like to
discuss: may καὶ δή be used as a near equivalent of ἤδη?
We have seen that Jebb considered it self-evident that καὶ δή could
be used instead of ἤδη and, since the publication of Denniston’s book,
commentators have often referred to his view on this matter, as it
appears on p. 252 (iv): ‘The line between “actually happening” and
“happening now” is often difficult to draw. Hence καὶ δή frequently
approximates in sense (particularly in the historians) to ἤδη, though
it is always more vivid and dramatic in tone.’18 However, although
‘already’ or ‘now’ may seem to be apt translations in many cases, I
doubt whether καὶ δή may actually be used this way. It is noteworthy
that Denniston himself is not very consistent in the choice of examples
destined to illustrate the various categories he distinguishes. In this

17
Herwerden has changed the word order (καὶ δή ψαύω) and Dawe has adopted
this change. However, Kamerbeek and Lloyd-Jones/Wilson retain the word order; Jebb
does not even comment on it.
18
I shall confine myself to a discussion of examples from tragedy and comedy.
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 129

connection I have already discussed Sophocles Ajax 544 (example 1)

and OC 31 (cf. section 2.1. following example 5). But there are other
examples as well. In Euripides Hercules 867 (example 17) ‘already’
would suit the context very well,19 but Denniston ranges this line in
section (i): ‘marking vivid perception by mind, ear or eye’. Then there
is Aristophanes Lysistrata 909. Myrrhine does not want to make love
to Kinesias in the presence of their child, whereupon Kinesias orders
his slave to take the child away and says:
(25) ἰδού· τὸ μέν σοι παιδίον καὶ δὴ ’κποδών.
There you are, look, you’ve got the baby out of the way. (Ar. Lys.
909)
Here again ‘already’ would be apt, but this time we find καὶ δή in sec-
tion (iii): ‘marking the provision or completion of something required
by the circumstances’.
The most telling example of καὶ δή in the sense of ἤδη is, accord-
ing to Denniston, Ranae 647. The janitor has announced that he will
strike Xanthias and Dionysus in turn in order to discover the truth
about their professed divinity. After striking Xanthias, he turns to his
second victim with the words: ἀλλ’ εἶμ’ ἐπὶ τονδὶ καὶ πατάξω. Dionysus
asks πηνίκα; and the janitor retorts: καὶ δὴ ’πάταξα. Neither Dover nor
Sommerstein comment on this καὶ δή, but Sommerstein’s translation
reads: ‘I already have’. Stanford has: ‘Actually I have already’; he seems
to render καὶ δή twice.
Most editors agree that the janitor hits Dionysus before he asks
πηνίκα,20 and in that case ‘already’ seems the only possible explana-
tion of καὶ δή. Likewise Denniston (1954: 252): ‘καὶ δή cannot mean
“there!”, since the blow precedes the question πηνίκα.’ However, I fail
to see any reason why this should be the case. I rather believe that καὶ
δή marks the blow Dionysus was prepared for. ‘There! I have struck
you’ is perfectly possible.
In drama καὶ δή is usually followed by a present tense or a perfect.
Ranae 647 is an exception; another one is Sophocles Ajax 49. Athena
tells Odysseus what Aias was trying to do during the night:

19
Kovacs does not render it this way, but Barlow (1996) does: ‘Look at him! He is
already shaking his head (. . .)’
20
Radermacher (1954) seems to disagree, since he comments on πηνίκα: ‘Dionysos
stellt sich alsob er seinen Hieb gar nicht erwarten könnte.’
130 a. maria van erp taalman kip

(26a) νυκτὼρ ἐφ’ ὑμᾶς δόλιος ὁρμᾶται μόνος.

He set out alone against you, by night and stealthily. (S. Aj. 47)
Odysseus, shocked and surprised, asks:
(26b) ἦ καὶ παρέστη κἀπὶ τέρμ’ ἀφίκετο;
Did he really get there and arrive at his goal? (S. Aj. 48)
whereupon Athena retorts:
(26c) καὶ δὴ ’πὶ δισσαῖς ἦν στρατηγίσιν πύλαις.
Yes, he was already at the gates of the two generals. (S. Aj. 49)
Kamerbeek comments: ‘καὶ δή: approaches ἤδη in meaning (“though
it is always more vivid and dramatic in tone” (Denniston, G.P., 252;
cf. e.g. O.C. 31)).’ Jebb has: ‘καὶ δή, “already”; O.C. 31 n.’ and Garvie,
too, has ‘already’ in his translation. Stanford, however, explains: ‘καὶ δή
with vivid and dramatic force (Denniston 252): “there he was, actually,
at the twofold command gates.”’ Note that he does refer to Denniston
(iv), but only to the ‘vivid and dramatic force’; he does not mention
the supposed equivalence with ἤδη. And indeed, although ‘already’
suits the context perfectly, it does not, in my view, completely capture
the value of καὶ δή. It is basically the same as an entry-marking καὶ δή,
but this time embedded in a story. The presence of Aias at the gates
is prepared for by the preceding lines, but Odysseus cannot actually
see him. By using καὶ δή Athena makes him an eye-witness après
la lettre, thus lending her story an extra thrill. The ‘vivid and dramatic
force’ seems to result from the combination of καὶ δή and the past
tense.
A complicating factor is the value of καί. We have seen that Denniston
considers καί non-connective in all the cases of καὶ δή discussed above.
And if we assume that καὶ δή may be used as ἤδη, a non-connective καί
would indeed appear to be a necessary condition, since otherwise δή in
itself would carry temporal value. However, as I argued above, there is
reason to suppose that καί, originally at least, is connective, and when we
find καὶ δή at the beginning of a sentence—as always in tragedy, except
for OC 173—it is not always easy to determine whether this original
function of καί is still present. The question has been brought up by
Barrett, who, while agreeing that καί is in most cases non-connective,
claims that it is connective in Euripides Hippolytus 1447, as well as in
Aristophanes Pax 178 (example 9) and Euripides Orestes 1108, one of
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 131

Denniston’s examples of καὶ δή in the sense of ἤδη.21 In Orestes 1108

Pylades has suggested killing Helen and Orestes has asked how this
can be done. Pylades anwers:
(27a) σφάξαντες. ἐν δόμῷ δὲ κρύπτεται σέθεν.
Cut her throat. She is hiding in your house. (E. Or. 1107)
Orestes confirms that she is in his house with the words:
(27b) μάλιστα· καὶ δὴ πάντ’ ἀποσφραγίζεται.
Yes, in fact she is putting the whole house under her seal. (E. Or.
1108)
I agree with Barrett that καί must be connective here,22 so καὶ δή can-
not be used as ἤδη. By opting for καὶ δή Orestes calls Pylades’ atten-
tion to something that, in his view, ensues from the subject of their
discussion: Helen’s presence in the house. However, he cannot see or
know what Helen is doing. He is simply convinced that this is precisely
what she is doing, and he suggests that Pylades, too, can ‘see’ what he
himself sees. His καὶ δή conveys a certain recklessness, that would not
be there if he had imparted the same information accompanied by,
say, the particle που.
Another example in Denniston is Orestes 1214. Electra has suggested
taking Hermione hostage, on her return from Clytaemnestra’s grave.
By threatening to kill her, they can keep Menelaos from taking revenge.
Orestes asks when she will be back:
(28a) ἥξει δ’ἐς οἴκους Ἑρμιόνη τίνος χρόνου;
ὡς τἄλλα γ’ εἶπας, εἴπερ εὐτυχήσομεν,
κάλλισθ’, ἑλόντες σκύμνον άνοσίου πατρός.
But when will Hermione return? All else you have said is excellent
provided we can succeed in catching this whelp of a godless sire.
(E. Or. 1211–3)

(28b) καὶ δὴ πέλας νιν δωμάτων εἶναι δοκῶ·

τοῦ γὰρ χρόνου τὸ μῆκος αὐτὸ συντρέχει.

21
Barrett (1964: ad loc.) In Hipp. 1447 (ὄλωλα καὶ δὴ νερτέρων ὁρῶ πέλας) Den-
niston (1954: 249) assumes asyndeton between ὄλωλα and καὶ δή. According to Barrett
this is ‘impossibly artificial’.
22
Willink (1986: ad loc.) is of the same opinion, but as is clear from his comment
on Orestes 1214, he nonetheless considers καὶ δή a more vivid equivalent of ἤδη.
132 a. maria van erp taalman kip

Well, I suppose she is quite near the house. The length of her
absence in itself supports this. (E. Or. 1214–5)
West, in contrast to Kovacs, follows Denniston and renders καὶ δή by
‘already’. And here again, this suits the context quite well. But I rather
think that what we have is a kind of entry-marking καὶ δή. The par-
ticle combination suggests an entry, but the listener realizes at the end
of the line that this entry is not yet a fact, and only reflects Electra’s
assessment.
Willink compares Orestes 1214 with Medea 1065. Medea knows
that, after receiving the poisoned robe she has sent her, Jason’s bride
will surely die:
(29) πάντως πέπρακται ταῦτα κοὐκ ἐκφεύξεται·
καὶ δὴ ’πὶ κρατὶ στέφανος, ἐν πέπλοισι δὲ
νύμφη τύραννος ὄλλυται.
These things are settled and cannot be undone. Already the crown
is on her head and the royal bride is perishing in the robe. (E.
Med. 1064–6)
This time it is Kovacs who opts for ‘already’,23 while Mastronarde refers
to Denniston’s ‘imaginary realization’. He compares 1065 with line 386
of this same tragedy, where Medea, after meditating on the possibility
of killing her enemies, speaks the words:
(30) καὶ δὴ τεθνᾶσι· τίς με δέξεται πόλις;
Now let us suppose they have been killed. What city will receive
me? (E. Med. 386)
Mastronarde does not note, however, that Denniston ranges only
example (30) in this category (2,v); example (29) is found in 2.iv (καὶ δή
= ἤδη). There is, of course, a difference. In (29) Medea has every reason
to imagine the death of Jason’s bride, whereas in (30) the death of her
enemies exists, for the time being, exclusively in her imagination. How-
ever, this difference is not essential and Mastronarde rightly explains
καὶ δή in these lines in the same way. In both cases Medea sees with
her inner eye how the death of her enemies becomes a reality.
Finally two examples from Aristophanes Ecclesiazusae, to start with
line 786. Chremes is ready to hand over his belongings to the State.

23
The Berlin papyrus has the varia lectio ἤδη, but Kovacs does not adopt this
reading.
 KΑΙ ΜHΝ, ΚΑΙ ΔΗ and ΗΔΗ in tragedy and comedy 133

Another citizen thinks he is crazy and asks whether he is really going

to do this. Chremes retorts:
(31) ναὶ μὰ Δία, καὶ δὴ μὲν οὖν
τωδὶ ξυνάπτω τὼ τρίποδε.
Yes, indeed—or rather, I am tying together those two tables at this
moment! (Ar. Ec. 786–7)
He is tying two tripods together, and by means of καὶ δή he calls his
interlocutor’s attention to what he is doing; ‘vois plutôt’ is the rendering
van Van Daele. The man can see it for himself. According to Ussher,
however, καὶ δή is used in the sense of ‘here and now’. He follows
Denniston in considering it the equivalent of ἤδη, although this choice
is far from compelling.
Line 581 is a different case. In its introductory song the Chorus has
told Praxinoa what she must do and in the katakeleusmos they once
again sum it up:
(32) ἀλλ’ οὐ μέλλειν, ὰλλ’ ἅπτεσθαι καὶ δὴ χρῆν ταῖς διανοίαις,
ὡς τὸ ταχύνειν χαρίτων μετέχει πλεῖστον παρὰ τοῖσι θεαταῖς.
You shouldn’t waste time, but get started right now on your idea,
because pace is what wins the most favour with audiences. (Ar. Ec.
581–2)
In Sommerstein’s translation we read ‘right now’. Ussher comments: ‘καὶ
δή: “here and now” (ἤδη), 786. See GP 252’, and this time Van Daele,
too, has recourse to a temporal adverb and renders it as ‘à l’instant’.
And yet I do not believe that this temporal value is implied by καὶ δή.
The Chorus call Praxinoa’s attention to something that ensues from
the preceding song. In this case, that ‘something’ is a course of action
she must follow, and καὶ δή implies, in my view, that Praxinoa herself
is bound to understand this. I am aware, of course, that a translation
like ‘as you can see for yourself’ would be overkill, but an explanation
is not always the same as a translation.
All in all I am far from convinced that, in both tragedy and comedy,
καὶ δή may be the near equivalent of ἤδη. I have already noted that
Denniston’s classification seems arbitrary; he fails to explain why he
ranges some examples in 2.iv and others in 2.i or 2.iii. Moreover, there
is the complicating factor of the value of καί, especially in the case of
tragedy, and finally it is, in my view, no easy task to relate the basic
characteristics that καὶ δή seems to possess with a temporal value. Those
characteristics risk disappearing behind the supposed temporal value.
 CHAPTER SEVEN

DISCOURSE COHESION IN DIALOGUE.

TURN-INITIAL AΛΛA IN GREEK DRAMA

Annemieke Drummen

1. Introduction

In Greek comedy and tragedy, speaking turns often start with ἀλλά.
How can the function of this particle be described, in general and spe-
cifically at turn beginnings? What is the contribution of ἀλλά to the
discourse cohesion of a dialogue? The present paper deals with these
questions. To understand them better, we will first consider discourse
cohesion in dialogues in general.

1.1. Cohesion in dialogue

The contribution of linguistic elements to the coherence of a discourse
is called cohesion. It is generally agreed that cohesion consists of gram-
matical and lexical elements forming connections between parts of a
discourse.1 In many cases discourse relations are made explicit, either
by intonation or by the use of a discourse marker, often a particle.2
When a discourse marker is present, the relation between the utter-
ances is immediately clear to the hearer, who does not need to interpret
them first. In this way, a discourse marker can ease the cognitive effort
of a hearer, as Jucker (1993) and Mosegaard Hansen (1996) have pointed
out for two partial equivalents of ἀλλά: English well and French eh
bien, respectively. As Mosegaard Hansen (1996: 337) writes, ‘[g]iven
the ephemeral, linear structure of (at least relatively) unplanned spoken

1
Tanskanen (2006: 7). Vuchinich (1977: 233 and passim) uses the term cohesion
differently, stating that ‘cohesion is not linguistically marked.’
2
Blakemore (1989: 23; 1992: 136); Mosegaard Hansen (1996: 315); Tanskanen
(2006: 16–7). Kroon (1995: 36; 1997: 17) explains that particles belong to the larger
category of discourse markers, which indicate how a text unit is integrated into the
discourse context. Rijksbaron (1997: 3) agrees that ‘particles are an important means
to signal coherence.’
136 annemieke drummen

discourse, (. . .) any element that works to ensure coherence takes on a

great deal of importance’ for smooth and successful communication.
A specific location for cohesive devices is at the beginning of a turn of
speaking. Lenk (1998: 3) writes about these turns: ‘connections [between
turns] are not always clearly recognizable unless they are properly
indicated. Lack of such indication prevents analyzers—and sometimes
even participants in the conversation—from understanding different
parts of a conversation as coherent’ (cf. also Blakemore (1989: 23)).
Thus, there is an important role for cohesion to play at the beginning
(and at the end)3 of a speaking turn. In the words of Schegloff (1987: 72),
‘turn beginnings (. . .) are important structural places in conversation.’
In English, the language studied by Schegloff, turn beginnings are either
sentence starts or discourse markers like well, but and so.4 When a turn
beginning is a sentence start, the hearer can infer information about the
turn from this turn beginning, as a new speaker usually projects features
of the type and shape of his turn through the sentence start.5 When,
for instance, a turn has begun with a question word, the addressee will
expect the turn to be a question and himself to be the next speaker
(again) (Jefferson, Sacks & Schegloff (1974: 719, note 32)).
When a turn starts with a discourse marker, on the other hand, the
speaker does not necessarily reveal or even have a plan for the turn’s
construction (Jefferson, Sacks & Schegloff (1974: 719); Schegloff (1987:
74)). A speaker uses the discourse marker to mark a specific relation
between the preceding discourse and the following turn6 and thus
enhances the cohesion. The relation between the turns is still under-
standable without the discourse marker,7 which is shown convincingly
by Schegloff ’s (1987: 80–1) observation that turn-initial discourse
markers (‘pre-placed appositionals’), as opposed to sentence starts,

3
Cf. Duncan (1972: 286–7; 1974: 165) on turn-yielding signals in English, and
Hopper (1992: 103–6, 114–5) on turn endings in English telephone conversations.
4
Jefferson, Sacks & Schegloff (1974: 719, with note 32); Schegloff (1987: 76). Dis-
course markers are called ‘appositionals’ by Schegloff. See Jucker (1993) for a detailed
analysis of well.
5
Hopper (1992: 104–5); Jefferson, Sacks & Schegloff (1974: 719, note 32); Schegloff
(1987: 71–4).
6
Discourse markers can react not only to a preceding turn, but also to events or
actions. For example, a sad face often speaks volumes, and may elicit a reaction like:
‘but look at the bright side’. Cf. Basset (1997: 83); Blakemore (1992: 139).
7
The same holds true for discourse relations inside a turn; cf. e.g. Slings (1980:
118).
 turn-initial ΑΛΛΑ in greek drama 137

are not repeated (‘recycled’) after an overlap of two speakers talking at

the same time. An overlap of the discourse marker at the turn begin-
ning does not influence the analysability and the understanding of the
sentence started (Hopper (1992: 123–5); Jefferson, Sacks & Schegloff
(1974: 719); Schegloff (1987: 74)).
So when a turn-initial discourse marker is not heard because of an
overlap, the addressee will still be able to understand the implied relation
between the two turns, although in that case alternative relations might
be possible. As Mosegaard Hansen (1996: 321) has described for French
eh bien, a certain ‘relation between the utterances would be the same
even without the marker, but once such a relation exists, the speaker
may choose to emphasize it through the use of eh bien.’ The discourse
marker ‘helps to weed out unwanted alternative interpretations.’8 In
other words, discourse markers seem only to make relations explicit,
not to create them. To anticipate slightly on the use of turn-initial ἀλλά,
consider a turn beginning from Aristophanes with a clearly correcting
value, but without ἀλλά.
(1) Χο. ἔστιν ὀπὴ δῆθ’ ἥντιν’ ἂν ἔνδοθεν οἷός τ’ εἴης διορύξαι,
εἶτ’ ἐκδῦναι ῥάκεσιν κρυφθεὶς ὥσπερ πολύμητις Ὀδυσσεύς;
Φι. πάντα πέφρακται κοὐκ ἔστιν ὀπῆς οὐδ’ εἰ σέρφῳ διαδῦναι.
ἀλλ’ ἄλλο τι δεῖ ζητεῖν ὑμᾶς.
Then is there a chink that you could excavate from inside
and then slip out disguised in rags, like wily Odysseus?—
Everything’s sealed up; there isn’t enough of a chink for even
a gnat to slip through. You’ve got to think of something else.9
(Ar. V. 350–3)
The chorus suggest to the character Philocleon that he might escape
the house through some hole, but he denies that this possibility exists.
To make the relation between the two utterances explicit, Philocleon
could have used ἀλλά, but for some reasons he chooses not to. Instead,
he first explains the unfortunate situation, and then substitutes it with
a new suggestion, this time using ἀλλά.

8
Mosegaard Hansen (1996: 337). Cf. Blakemore (1989), who argues the same in
other words: discourse markers, she writes, ‘constrain the hearer’s interpretation.’
9
The translations of the Vespae cited are those by Henderson (1998).
138 annemieke drummen

We are of course unable to know anything about overlap in Ancient

Greek conversations. However, if we assume, like Slings (2002a), with
Willi (2002a: 15–6), ‘that the logical operations of the human mind
are not fundamentally different in different places and societies, we
may make inferences about the functioning of an ancient language by
comparing how modern languages structure information.’ Likewise, I
will try to show in this paper that the Greek discourse marker ἀλλά,
when placed at the beginning of a turn, behaves in a similar way as
described for English and French turn-initial discourse markers.

1.2. The language of comedy and tragedy

As research corpus I have chosen two comedies of Aristophanes (Ranae
and Vespae), two tragedies of Aeschylus (Agamemnon and Persae), two
of Sophocles (Antigone and Trachiniae), and two of Euripides (Ion and
Medea).10 However stylized these texts are as part of a literary tradition,
they were also meant to simulate real conversations.11 A realistic rep-
resentation should include turn-taking devices, and therefore a written
version should show the ones that are expressed in a linguistic way.
Although both Greek comedy and Greek tragedy are composed of
an artificial language, a fact that is most obvious in the use of metre
(cf. Willi (2002b: 114–6)), the language of comedy is generally
considered to simulate ‘everyday language’ (Willi (2002a: 13–4)).
Willi (2002a: 18) formulates a balanced position: ‘a comparison with
non-comic data proves the obvious point that this [Aristophanes’]
literary language is much closer to spoken Attic than, for instance, the
language of tragedy.’ Comedy displays a ‘greater linguistic realism’

10
The texts of the Oxford editions were used: Davies (1991); Dover (1993); Dunbar
(1995); Lloyd-Jones & Wilson (1990); MacDowell (1971); Olson (2002); Page (1972);
Ussher (1973). I do not distinguish between the start of a choral song and that of a
speaking turn in a ‘true’ dialogue. The chorus start talking (or singing) with ἀλλά at
e.g. Aristophanes’ Vespae 365, 1009, Aeschylus’ Agamemnon 276, 1302, 1304, 1652,
and Choephori 306, 340, 400, 775, 1044, 1063, Sophocles’ Electra 137, and Euripides’
Medea 759.
11
Cf. Dik (2007: 7). The prose texts of Plato might be an even closer approxima-
tion to spoken language. A quick glance at his Cratylus, Crito, Euthyphro, Phaedo and
Phaedrus creates the impression that turn-initial ἀλλά occurs frequently in Platonic
dialogue as well; it would be interesting to know whether it is used in Plato in the same
way and frequency as in drama. I leave Plato out of account in this paper, because
comedies and tragedies are more alike in structure and function. Therefore, they can
be compared to each other more easily.
 turn-initial ΑΛΛΑ in greek drama 139

than tragedy.12 Furthermore, comedy contains dialogues of a more

aggressive nature than tragedy: characters are often interrupted and
therefore not granted the possibility to talk as long as they might have
wanted (Slings (2002a: 101)).
Sommerstein (2002b: 46, 57) points out some linguistic differences
within the area of tragedy. Sophocles would have been ‘fond of unusual,
sometimes innovative syntactic constructions’, while ‘Euripides’
language is closer than that of Sophocles to the vocabulary and syntax
of ordinary speech.’ Goward (2005: 103) describes Aeschylus’ diction
as ‘bombastic, heavy and obscure.’ An important difference with the
other two tragedians, furthermore, is the larger role of the chorus in
Aeschylean drama,13 which leads to a much smaller number of turns.
In section 2, I will discuss the differences between comedy and tragedy
and among the various tragedians with regard to the use of a specific
turn-taking device: the discourse marker ἀλλά.

1.3. The discourse marker ἀλλά

In this paper I will focus on turn-initial ἀλλά because it is the most
frequent turn beginning in Aristophanes’ comedies and because it
occurs in tragedy as well. Before turning to more detailed observa-
tions on its usage in the corpus, I will give a brief overview of earlier
descriptions of ἀλλά.
Its most typical function is said to be substitution, also called
‘eliminative’, ‘replacing’, or ‘the sondern-type’ after the German discourse
marker sondern.14 In this meaning, ἀλλά is found in a construction of
the type οὐ Α ἀλλά Β, in which A and B are incompatible elements.
Although this use of the particle is in general most frequent, the study
of Basset (1997: 93; 96, note 12) suggests that this use is rare in or
maybe even absent from turn beginnings in drama.
What function does ἀλλά have, then, at the beginning of a turn of
speaking? LSJ mention ‘to introduce a general objection’ and ‘to break
off a subject abruptly’, but note that ἀλλά is used ‘also in affirmative
answers.’ Denniston (1954: 7; 15–20) describes several possibilities
for the meaning of turn-initial ἀλλά. It can ‘[object] to the previous

12
Willi (2002a: 18; 2002b: 118, 121, 124). Cf. Dik (2007: 6).
13
Conacher (1980: 146); Smyth (1963: x, xxvi). See note 22 below.
14
Basset (1997: 75, note 3, 97); Denniston (1954: 1); Ruijgh (1971: 135–6); Sicking
(1993: 36–7, 49–50); Slings (1980: 107, 109; 1997a: 105); Wakker (1995: 265–6).
140 annemieke drummen

speaker’s words or behaviour’, in which case it ‘simply expresses

opposition, and it is left undetermined whether the opposite ideas are,
or are not, incompatible.’15 When turn-initial ἀλλά occurs in wishes
and prayers, Denniston writes, ‘[t]here is no strong break-off (. . .):
ἀλλά merely marks a gentle transition from the known present to the
unknown and desired future.’ According to Denniston, turn-initial
ἀλλά can even express assent, in two different ways. Either ‘agreement
is presented as self-evident and inevitable. The speaker not only agrees,
but repudiates the very idea that dissent is possible.’ Secondly, it is
possible that ‘agreement is presented, not as self-evident, but as wrung
from the speaker malgré lui.’
Now it is hard to believe that the very same particle could express
disagreement as well as agreement, and a strong break-off as well as a
gentle transition, unless these opposite interpretations are not inherent
in the meaning of ἀλλά, but rather arise from the different contexts with
which ἀλλά is compatible.16 Still, if ἀλλά has some meaning of its own
at all, I believe this compatibility has to exist because of some similarity
between these contexts (cf. Jucker (1993: 437–8)). In other words, I
believe the particle has only one basic function, viz. the marking of
corrections. This function may lead to several possible interpretations,
depending on the context.
Basset (1997) has shown in his study of ἀλλά in Aristophanes’ Ranae
that it is a more fruitful approach to distinguish between different ele-
ments that are corrected by ἀλλά than to divide its use into several kinds
of adversativity or even into totally different meanings. The meaning
of the particle is more coherent if one follows Basset in attributing a
major role to the context. In each occurrence, ἀλλά corrects an element
of the preceding words or actions, which can be either an explicitly
stated element, a presupposed element, an implicitly understood ele-
ment (henceforth ‘implication’), or the discourse topic.17

15
Cf. Blakemore (1989: 29) for a similar comment on ἀλλά’s English counterpart
but.
16
Cf. Sicking (1993: 45). Wakker (1995: 263, with note 29; 1997: 227, with n. 41)
similarly criticises Denniston’s treatment of μήν.
17
Basset (1997: 82). Cf. Sicking (1993: 49–50) on ἀλλά substituting an implication,
and Sicking (1993: 38) on ἀλλά’s replacement of a discourse topic. ‘Implication’ is
Rijksbaron’s equivalent of Basset’s ‘sous-entendu’, called ‘expectation’ by Slings (1980;
1997a); cf. Rijksbaron (1997: 6, note 6). I will use ‘implication’, for I consider it the
most convenient and neutral term.
 turn-initial ΑΛΛΑ in greek drama 141

An explicit(ly stated) element is situated in the text itself, and directly

heard and understood by the addressee without the need of interpre-
tation. Now when explicit elements are corrected, explicit negations
are to be expected, yielding the true eliminative construction οὐ Α
ἀλλά Β. As explained by Wakker (1995: 266), the affirmed element
B becomes stronger when element A is explicitly dismissed. Whereas
explicit elements are easy to discern, both presupposed elements and
implications are left implicit. The difference between the latter two lies
in the fact that a presupposed element is part of the utterance itself,
and the speaker can therefore be held responsible for its content. The
presupposed element still holds true when the utterance is negated or
transformed into a question. Therefore, such negation and interrogation
tests are important means to recognize (the correction of) presupposed
elements.18 Implications, on the other hand, cannot be found in the
utterance itself, but are dependent on interpretation. The negation or
interrogation test does not hold for implications, and the speaker cannot
be held responsible for their content.19 The discourse topic, finally, is
the theme about which the participants in the conversation are talking.
When this is corrected by a second speaker, there is no disagreement
between the two turns with regard to their content: only the topic of
conversation is changed.
To illustrate these four elements and their corrections with a clear
example, consider (2) and (2a–d), showing the corrections of an explicit
element, a presupposed element, an implication, and the discourse topic,
respectively. The corrections are marked explicitly with the discourse
marker but.

(2) He quit smoking.

(2a) But he still smokes.
(2b) But he has never smoked at all.
(2c) But you will never succeed.
(2d) But it is time to go now.

18
Ducrot (1984: 18, 20–1). For example, the utterance ‘you know my brother’ as
well as ‘you don’t know my brother’ and ‘do you know my brother?’ presuppose that
the speaker has a brother.
19
Ducrot (1984: 19, 21). For example, the possible implication ‘I will not go’ of the
utterance ‘it is dangerous to go’ is not shared by ‘it is not dangerous to go’ and ‘is it
dangerous to go?’ (example based on Aeschylus’ Agamemnon 1311–13).
142 annemieke drummen

When a second speaker reacts to (2) with (2a), he corrects the explicit
element of the quitting. With (2b), on the other hand, it is a presupposed
element that is corrected, viz. that the person spoken about has been
a smoker some day. The first speaker can be held responsible for this
element, which can be made clear by a negation or interrogation test.
Both ‘he did not quit smoking’ and ‘did he quit smoking?’ presuppose
that ‘he’ has been a smoker. Reaction (2c), however, corrects an element
that is part of the hearer’s interpretation, viz. that the first speaker
meant that he would be able to quit smoking as well. The first speaker
does not bear responsibility for this implication and may always deny
that it was intended. (2d), finally, switches (corrects) the discourse
topic: from the (former) smoker to the departure of the two speakers.
Note that these four corrections would also be present if the discourse
marker but were omitted.
In section 3, I will arrange all occurrences of turn-initial ἀλλά in
my corpus according to these four different corrected elements. First,
however, I will give a quantitative analysis.

2. A quantitative analysis of turn-initial ἀλλά

In the six comedies of Aristophanes that I investigated, the most

frequently used expression at turn beginnings is ἀλλά, followed by μὰ
(τὸν) Δία or νὴ (τὸν) Δία. Another word often attested at the start of
turns is the negation οὐ (and variants or similar words like οὐχ and
οὐδέν). I counted the total number of speaker’s changes in each play
as well, in order to calculate the relative frequencies of the turn-initial
expressions. In the six Aristophanic plays, about 590 speaker’s changes
were attested on average.
As can be seen in table 1, in these comedies, ἀλλά is on average used
as the first word of a speaker almost thirty times per play, or in 5% of
all turn takings. The frequency in the Lysistrata stands out: almost 8%
of the approximately 560 turns begin with ἀλλά. This play is exceptional
in another aspect, too: it starts with ἀλλά. So, strictly speaking, this
occurrence is not after a change of speaker, since there is no previous
speaker, but it is precisely the start with ἀλλά that creates the impression
of a beginning in mediis rebus,20 since ἀλλά always indicates a reaction
to some preceding element, whether words or actions.

20
Henderson (2002a: 66): ‘Only in the subsequent dialogue do we discover the
thought with which ἀλλά is contrasted.’
 turn-initial ΑΛΛΑ in greek drama 143

Table 1: Number of turn-initial ἀλλά in comedies of Aristophanes21

Comedy Turn-initial ἀλλά
Acharnenses 27
Aves 30
Ecclesiazusae 28
Lysistrata 42
Ranae 25
Vespae 26
Average number per play 29.7
Average number per change of speaker 0.05

In contrast to the expressions μὰ (τὸν) Δία and νὴ (τὸν) Δία, turn-

initial ἀλλά occurs in tragedy as well. Tables 2 to 4 show the number
of occurrences in six plays of each of the three major tragedians, as
well as the average relative frequency, calculated by dividing the average
number of ἀλλά by the average number of speaker’s changes.21

Table 2: Number of turn-initial ἀλλά in tragedies of Aeschylus

Tragedy Turn-initial ἀλλά
Agamemnon 12
Choephori 12
Eumenides 7
Persae 7
Septem contra Thebas 6
Supplices 8
Average number per play 8.7
Average number per change of speaker 0.06

21
I do not include the occurrences of turn-initial μἀλλά, because in these cases
the turns in fact start with μή, not ἀλλά. Turn-initial μἀλλά is found at Aristophanes’
Acharnenses 458, Aves 109, Ranae 103, 611, 745, 751, and the uncontracted form μὴ
ἀλλά at Aeschylus’ Choephori 918. Similarly, the more frequent turn-initial expression
οὔκ, ἀλλά is not included either (e.g. Ar. Ach. 108, 425, 1114; Av. 71, 105, 912; Ec. 331;
S. El. 1453, OT 1040, Ph. 993, 997; E. El. 964; Ion 1407).
144 annemieke drummen

Table 3: Number of turn-initial ἀλλά in tragedies of Sophocles

Tragedy Turn-initial ἀλλά
Ajax 10
Antigone 17
Electra 26
Oedipus Tyrannus 25
Philoctetes 25
Trachiniae 23
Average number per play 21.0
Average number per change of speaker 0.06

Table 4: Number of turn-initial ἀλλά in tragedies of Euripides

Tragedy Turn-initial ἀλλά
Andromache 2
Electra 6
Hecuba 4
Hippolytus 1
Ion 6
Medea 7
Average number per play 4.3
Average number per change of speaker 0.01

These tables show strikingly that, even with regard to such a small
linguistic phenomenon as turn-initial ἀλλά, the three writers have
a characteristic diction, different from that of the others. Aeschylus
shows a relatively low absolute number of turn-initial ἀλλά (about 9 per
play), but because his plays contain only about 150 speaker’s changes
on average, the particle does cover 6% of all these changes: even more
than the relative frequency in the comedies investigated. The same high
percentage is found in the six Sophoclean tragedies, but the absolute
number of turn-initial ἀλλά is much higher (21 per play), since these
plays contain on average about 350 changes of speaker each. Euripides,
with an average number of about 320 speaker’s changes, comes out
lowest both in absolute and in relative numbers: he uses ἀλλά at turn
beginnings about 4 times each play, which means in only 1% of all
speaker’s changes.
The different numbers of speaker’s changes show that comedy char-
acters switch turns more often than tragedy characters, yielding a rather
 turn-initial ΑΛΛΑ in greek drama 145

different, more natural (or, as stated above, more aggressive) kind of

dialogue. As a consequence, turns in tragedies are on average longer
than those in comedies. The length of choral songs, largest in Aeschylus,
plays a major role in this difference.22

3. A qualitative analysis of turn-initial ἀλλά

In this section, I will investigate the use of turn-initial ἀλλά in eight of

the plays mentioned above: Aristophanes’ Ranae and Vespae, Aeschy-
lus’ Agamemnon and Persae, Sophocles’ Antigone and Trachiniae, and
Euripides’ Ion and Medea. Each occurrence will be arranged according
to the element that is corrected by ἀλλά: an explicitly stated element, a
presupposed element, an implication, or the discourse topic.

3.1. Preliminary remarks

Basset (1997) has already arranged all instances of ἀλλά found in the
Ranae by means of the above classification of four possible corrected
elements, but I do not agree completely with his assessments of the
nature of the corrected elements. The instances of ἀλλά in lines 134
and 258, which he classifies as corrections of presupposed elements, in
my view correct an implication because the negation or interrogation
test23 does not hold.
(3) Βα. βρεκεκεκὲξ κοὰξ κοάξ.
Δι. οἰμώζετ’· οὐ γάρ μοι μέλει.
Βα. ἀλλὰ μὴν κεκραξόμεσθά γ’ ὁπόσον ἡ φάρυξ ἂν ἡμῶν
χανδάνῃ δι’ ἡμέρας.
Brekekekex koax koax!—Wail away; what do I care?—In fact we’ll
bellow as loud as our gullets will stretch, all the livelong day!24 (Ar.
Ra. 256–9)

22
The Prometheus Vinctus deviates from the general impression of the Aeschylean
plays: in this play ἀλλά is used at merely 4 turn beginnings, covering 2% of the total
number of turn beginnings (about 190). I leave the Prometheus Vinctus out of account,
because the authorship of Aeschylus is disputed. Cf. Conacher (1980: 141–74, esp. 146,
158): the smaller role of the chorus and the particle usage in this play are among the
arguments against its authenticity.
23
See above, notes 18 and 19.
24
The translations of the Ranae cited are those by Henderson (2002b).
146 annemieke drummen

Basset (1997: 86) argues that the order of Dionysos preceding the frogs’
utterance is considered superfluous by them. According to Basset, ἀλλά
corrects the presupposed element that would be present in every order:
that the addressee has not yet started the requested action. However,
Dionysus’ οἰμώζετε is a present imperative, so it does not presuppose
that the order is not yet carried out (viz. that the frogs are not yet
bellowing). The chorus rather correct an implication of the imperative,
viz. ‘go on wailing like you do know’, by stating that they will croak
extra loud and long. With μήν, they mark the truth of this statement
as personally guaranteed, even though the addressee (Dionysus) may
find it incredible or unexpected (cf. Wakker (1995: 257–8, 262; 1997:
213, 229)). In the two other instances of turn-initial ἀλλὰ μήν in my
corpus (Aeschylus’ Persae 226 and 233) I think the combination of the
particles marks the substitution of an implication for an alternative that
was not expected by the speaker (the Persian queen Atossa).
(4) Where is Athens?—Far away
Βα. ἀλλὰ μὴν ἵμειρ’ ἐμὸς παῖς τήνδε θηρᾶσαι πόλιν;
Can it then really be that my son had the keen desire to make booty
of this city?25 (A. Pers. 233)
Contrary to the view of Wakker (1997: 214, note 13), I do not think the
queen is particularly interested in a true answer to her question, because
she already knows the answer; she wants to express her surprise about
it. I do agree that it is the queen herself who finds this implied answer
hard to believe. Xerxes’ desire to capture far-away Athens runs counter
to her own expectation that he would not pursue the impossible, or
the like. She therefore corrects this (implied) expectation with ἀλλά
and reinforces her question with μήν, inviting her addressee to assure
her of the truth of this unexpected conclusion and to give, perhaps,
an explanation for it.
Moreover, I think Basset (1997: 78) is wrong in excluding the ἀλλ’
in Ranae 928 from the corpus, because it would be used as an adverb
rather than as a coordinator.
(5) Εὐ. σαφὲς δ’ ἂν εἶπεν οὐδὲ ἕν—
Δι. μὴ πρῖε τοὺς ὀδόντας.
Εὐ. ἀλλ’ ἢ Σκαμάνδρους ἢ τάφρους ἤ . . .

25
The translation of the Persae cited is that by Smyth (1963).
 turn-initial ΑΛΛΑ in greek drama 147

And he wouldn’t say a single intelligible word—Stop gnashing your

teeth!—But only Scamanders, or moats, or. . . . (Ar. Ra. 927–8)
The combination ἀλλ’ ἤ can be considered an idiomatic expression,
usually meaning ‘except (that)’, which may have originated either
from (οὐδὲν) ἄλλο ἤ or from ἀλλά ἤ (see Denniston (1954: 24–7)).
However, the repetition of ἤ makes such an idiomatic interpretation
improbable here. The translation ‘except’ would be odd in this case,
since the character Euripides considers the Σκαμάνδροι and τάφροι
typical examples of Aeschylus’ obscurity, not exceptions to it. Moreover,
even if we are dealing with the idiomatic expression here, I believe the
correcting function of the particle can certainly be felt, whether ἀλλά
was part of the true origin of the expression or not.26 In this instance,
the speaker corrects the explicit element σαφὲς ἕν (‘one clear thing’)
of his own previous turn (note the negation οὐδὲ in line 927). Thus,
despite a possible contamination with ἄλλο, there is no need to exclude
this instance as completely deviant from ἀλλά’s normal function, as
suggested by Basset.
A special case in another play, furthermore, is the occurrence of
ἀλλά in Vespae 1152:
(6) Βδ. οὐκ ἀναβαλεῖ;
Φι. μὰ Δί’ οὐκ ἔγωγ’.
Βδ. ἀλλ’, ὦ ’γαθέ—
Φι. εἴπερ γ’ ἀνάγκη, κρίβανόν μ’ ἀμπίσχετε.
Please put it on.—I absolutely refuse.—But good sir—If this is
compulsory, dress me in an oven instead. (Ar. V. 1152–3)
We cannot know which of the four elements Bdelycleon wanted to
correct (although the discourse topic is an unlikely candidate in the
situation), because he is interrupted by Philocleon before he has had
the chance to make this clear. Incidentally, this interruption is a very
strong example of the aggressiveness of comic dialogue.

26
Perhaps the absence of an accent on the first syllable may also be taken as an
indication for the association of ἀλλ’ ἤ with ἀλλά, even if originally a ‘loss of accent
[of ἄλλο was] caused by fusion with the following word’, as suggested by Denniston
(1954: 24). Kühner-Gerth (1898/1904: 2.285) and LSJ also treat ἀλλ’ ἤ as a combination
containing ἀλλά, while at the same time accepting a possible origin from ἄλλο.
148 annemieke drummen

3.2. Corrections of explicit elements27

When an explicitly stated element is contradicted, the preceding turn
usually contains a negation. However, the corrected element is not
necessarily one of the preceding turn; a speaker can also correct his
own previous words, like in example (7).
(7) Χο. πότερα δ’ ὀνείρων φάσματ’ εὐπιθῆ σέβεις;
Κλ. οὐ δόξαν ἂν λάβοιμι βριζούσης φρενός.
Χο. ἀλλ’ ἦ σ’ ἐπίανέν τις ἄπτερος φάτις;
Dost thou pay regard to the persuasive visions of dreams?—I would
not heed the fancies of a slumbering brain.—But can it be some
pleasing rumour that hath fed thy hopes?28 (A. A. 274–6)
The chorus correct their own previous suggestion, after it is negated by
Clytaemnestra in the intervening turn.
In Medea 1366, on the other hand, the utterances of two different
speakers together form an eliminative construction:
(8) Ἰα. οὔτοι νιν ἡμὴ δεξιά γ’ ἀπώλεσεν.
Μη. ἀλλ’ ὕβρις οἵ τε σοὶ νεοδμῆτες γάμοι.
It was not my hand, you know, that killed them.—No: it was the
outrage of your new marriage.29 (E. Med. 1365–6)
The tenor of the two turns is that ‘not Jason’s hand, but his marriage’
killed the children: a normal eliminative construction in which ἀλλά
is frequently found. The remarkable aspect is that it is expanded over
two turns. It makes no difference for the interpretation of ἀλλά that
Medea does not want to stress her agreement with Jason’s words, but
her disagreement with the implication that the children’s death was
not his fault.

3.3. Corrections of presupposed elements30

A presupposition is left implicit, but is enclosed in an utterance itself.
Therefore, the speaker can be held responsible for a presupposed

27
Ar. Ra. 56, 57, 488, 928, 1058, 1298; A. A. 276; S. Tr. 397, 1211, 1216; E. Med.
1366.
28
The translation of the Agamemnon cited is that by Smyth (1963).
29
The translation of the Medea cited is that by Kovacs (1994).
30
Ar. Ra. 45, 287, 292, 481, 1130; V. 165, 945; S. Ant. 567; Tr. 67, 1151; E. Ion
1287; Med. 816.
 turn-initial ΑΛΛΑ in greek drama 149

element of his utterance. In example (9), the correction of a presupposed

element is used in a joke.
(9) Φι. διατρώξομαι τοίνυν ὀδὰξ τὸ δίκτυον.
Ξα. ἀλλ’ οὐκ ἔχεις ὀδόντας.
Then I’ll gnaw through this netting with my teeth!—You haven’t
any teeth! (Ar. V. 164–5)
Biting through a net with one’s teeth (ὀδάξ) naturally presupposes that
this person has teeth—this is, however, contradicted by Xanthias in his
turn starting with ἀλλά. Note that the relation between the two utter-
ances would still be one of correcting a presupposed element if ἀλλά
were absent; in that case, however, the relation would be implicit and
more difficult for the hearers to recognise.
Example (10) also contains the correction of a presupposed element.
(10) Ἰω. κἄπειτ’ ἔκαινες φαρμάκοις τὸν τοῦ θεοῦ;
Κρ. ἀλλ’ οὐκέτ’ ἦσθα Λοξίου, πατρὸς δὲ σοῦ.
You tried to kill me [with poison]. Was I not the god’s boy?—You
were no longer Loxias’ boy but your father’s.31 (E. Ion 1286–7)
Although it is explicitly stated in Ion’s utterance that Creusa’s sup-
posed victim was ‘the god’s boy’, it is presupposed that this boy was
Ion himself. The further context of the dialogue demonstrates that Ion
and Creusa are talking about the attack on Ion, described before in the
same tragedy (lines 1181–1216). Through correction of this presupposed
element, Creusa denies to have attacked the god’s boy, but it is the pre-
supposition that Ion belonged to Apollo that is corrected with ἀλλά.

3.4. Corrections of implications32

In most cases, turn-initial ἀλλά corrects an implication, as in example
(11). Antigone has been caught burying her brother and Creon has
just asked her sister Ismene whether she has taken part in the crime
as well.

31
The translation of the Ion cited is that by Kovacs (1999).
32
Ar. Ra. 134, 136, 226, 240, 258, 568, 1025, 1039, 1136, 1215, 1389; V. 8, 28, 75,
175, 181, 214, 223, 365, 453, 457 (first ἀλλά), 461, 832, 920, 1129, 1190, 1434, 1502,
1504; A. A. 944, 1050, 1206, 1248, 1302, 1304, 1313, 1649, 1652, 1662, 1666; Pers. 226,
233, 246, 697, 703, 795, 796; S. Ant. 48, 84, 89, 98, 217, 473, 520, 538, 540, 556, 568,
834, 1253, 1336; Tr. 72, 86, 229, 389, 472, 490, 588, 592, 594, 600, 620, 727, 981, 1128,
1179, 1245, 1257; E. Ion 755, 769, 1288, 1325, 1443; Med. 326, 619, 1389.
150 annemieke drummen

(11) Ἰσ. δέδρακα τοὔργον, εἴπερ ἥδ’ ὁμορροθεῖ,

καὶ ξυμμετίσχω καὶ φέρω τῆς αἰτίας.
Ἀν. ἀλλ’ οὐκ ἐάσει τοῦτό γ’ ἡ δίκη σ’, ἐπεὶ
οὔτ’ ἠθέλησας οὔτ ἐγὼ ’κοινωσάμην.
Ἰσ. ἀλλ’ ἐν κακοῖς τοῖς σοῖσιν οὐκ αἰσχύνομαι
ξύμπλουν ἐμαυτὴν τοῦ πάθους ποιουμένη.
I did the deed, if she agrees, and I take and bear my share of the
blame.—Why, justice will not allow you this, since you refused
and I was not your associate!—But in your time of trouble I am
not ashamed to make myself a fellow voyager in your suffering.33
(S. Ant. 536–41)
The first ἀλλά corrects the implication—inferred by the interpretation
of the character Antigone—that Ismene will be allowed to be punished
as well. It is not a presupposition, since it cannot be deduced from the
negated ‘I do not take my share of the blame.’ The second ἀλλά also
corrects an implication, viz. that Ismene would be ashamed to share
in the punishment of Antigone. As a negation or interrogation test
may show, the speakers of the utterances preceding those containing
turn-initial ἀλλά cannot be held responsible for the content of these
implications.
In the next example, the element corrected by ἀλλά might not be
clear at once.
(12) Δη. χαίρειν δὲ τὸν κήρυκα προὐννέπω, χρόνῳ
πολλῷ φανέντα, χαρτὸν εἴ τι καὶ φέρεις.
Λι. ἀλλ’ εὖ μὲν ἵγμεθ’, εὖ δὲ προσφωνούμεθα.
I welcome you the herald, who have now at last appeared, if indeed
your news is welcome.—I am happy in my coming and happy in
your salutation.34 (S. Tr. 227–9)
Davies (1991: 106), following Denniston (1954: 19), interprets ἀλλά as
‘a sympathetic reaction to the previous speaker’s words or actions’ and
thereby contradicts Jebb’s (1892: 38) opinion that ἀλλά ‘replies to the
doubt implied in χαρτὸν εἴ τι καὶ φέρεις.’ However, the sympathetic
reaction arises out of the other words of Lichas, not out of ἀλλά. What
does ἀλλά correct, then, in this utterance? Clearly it is not an explicit
element of Deianeira’s words. It is no presupposition either, because
the subordinate εἰ-clause, containing a present indicative (φέρεις),

33
The translations of the Antigone cited are those by Lloyd-Jones (1994).
34
The translations of the Trachiniae cited are those by Lloyd-Jones (1994).
 turn-initial ΑΛΛΑ in greek drama 151

expresses a neutral condition. This means that the condition enclosed

in ‘if you bring good news’ may be realized or not; the clause does
not presuppose that you do not bring good news. Since Lichas’ reply
is a greeting succeeding a greeting, furthermore, there is no change
of discourse topic. The only possible interpretation left is close to the
one of Jebb: ἀλλά corrects an implication, namely that Lichas would
be bringing bad news.
In reactions to requests, a similar weak implication of the preceding
speaker’s utterance can be corrected by ἀλλά: viz. that the current
speaker would not be willing to fulfil the request.35 Denniston (1954:
17) even calls ἀλλά in these cases an ‘expression of willingness to act in
a required way’, but there is no need to abandon the particle’s general
correcting function.

3.5. Corrections of discourse topics36

The particle ἀλλά can also mark the substitution of one discourse topic
for another: for example, after Creon and Teiresias have talked for
some time about the latter’s thrustworthiness, the dialogue now turns
to Teiresias’ prediction proper.
(13) Κρ. μόνον δὲ μὴ ’πὶ κέρδεσιν λέγων.
Τε. οὕτω γὰρ ἤδη καὶ δοκῶ τὸ σὸν μέρος;
Κρ. ὡς μὴ ’μπολήσων ἴσθι τὴν ἐμὴν φρένα.
Τε. ἀλλ’ εὖ γέ τοι κάτισθι μὴ πολλοὺς ἔτι
τρόχους ἁμιλλητῆρας ἡλίου τελῶν.
Only do not speak for the sake of profit!—That is what you already
think I do.—Know that you will never be able to trade on my

35
Ar. V. 457 (first ἀλλά); S. Tr. 86, 389, 490, 620, 1257. An example is:
Λι. κείνου τε καὶ σὴν ἐξ ἴσου κοινὴν χάριν
καὶ στέργε τὴν γυναῖκα καὶ βούλου λόγους
οὓς εἶπας ἐς τήνδ’ ἐμπέδως εἰρηκέναι.
(. . .)
Δη. ἀλλ’ ὧδε καὶ φρονοῦμεν ὥστε ταῦτα δρᾶν
Both for his sake and your own show kindness to the woman, and wish the things
you said regarding her not to have been said in vain. (. . .)—Why, I am indeed
minded to do this. (S. Tr. 485–90)
The character Deianeira does not use ἀλλά to express her obedience to Lichas’ request.
Rather, she marks the correction of an implication of his words, viz. that she would
be planning to harm the girl her husband is in love with, or something of similar
purport.
36
Ar. Ra. 108, 123, 646; V. 173, 428, 457 (second ἀλλά), 715, 1009; S. Ant. 327,
1064; Tr. 627; E. Med. 688, 759.
152 annemieke drummen

judgement!—Then know well that you shall not accomplish many

racing courses of the sun. (S. Ant. 1061–5)
With ἀλλά, Teiresias marks his change of discourse topic explicitly. The
preceding discussion is broken off, and whether Creon will believe him
or not, Teiresias now tells him the prediction.
Likewise, in example (14), the utterance starting with ἀλλά is used
to put an end to the dispute of Philocleon and Bdelycleon about who
is the best donkey seller.
(14) Φι. ἀποδόσθαι βούλομαι
τὸν ὄνον ἄγων αὐτοῖσι τοῖς κανθηλίοις·
νουμηνία γάρ ἐστιν.
Βδ. οὔκουν κἂν ἐγὼ
αὐτὸν ἀποδοίμην δῆτ’ ἄν;
Φι. οὐχ ὥσπερ γ’ ἐγώ.
Βδ. μὰ Δί’, ἀλλ’ ἄμεινον.
Φι. ἀλλὰ τὸν ὄνον ἔξαγε.
I just want to take the donkey and its panniers out and sell them.
It’s market day.—Surely I could do that, couldn’t I?—Not the way
I would.—That’s right, I’d do it better.—All right, let the donkey
out. (Ar. V. 169–73)
In fact, Philocleon does not really care about the discussion: he only
wants to escape from the building by means of the donkey. He there-
fore changes the discourse topic and summons his son Bdelycleon to
let the donkey out.

4. Conclusions

The general function of the discourse marker ἀλλά is said to be

that it substitutes one element for another. This turned out to be its
fundamental value at turn beginnings as well. In all cases investigated,
the function of turn-initial ἀλλά can be interpreted as marking a
correction of the preceding words or actions. The corrected (substituted)
element is either an explicitly stated element, a presupposed element,
an implication, or the discourse topic. Thus, Basset’s classification (see
section 1.3) is very useful to understand ἀλλά’s function in different
contexts. My investigation has shown that while in general the particle
is predominantly found in corrections of explicit elements, at turn
beginnings it most often corrects implications. This prevalence of
 turn-initial ΑΛΛΑ in greek drama 153

Table 5: Element corrected by turn-initial ἀλλά37

Corrected element Aristophanes’ Aeschylus’ Sophocles’ Euripides’ Total
Ranae and Agamemnon Antigone and Ion and
Vespae38 and Persae Trachiniae Medea
Explicit element 6 1 3 1 11
Presupposed element 7 0 3 2 12
Implication 29 18 31 8 86
Discourse topic 8 0 3 2 13
Total 50 19 40 13 122

implication corrections is present in all four authors of my corpus:

Aristophanes, Aeschylus, Sophocles, and Euripides. The numbers of
occurrences are given in Table 5. 37 38
The high frequency of implication corrections (about 70% of all
instances) is not surprising in view of the fact that speakers usually cor-
rect an element of someone else’s words at the beginning of their turn. It
is to be expected that a complete contradiction of a preceding utterance
(yielding a pointless discussion without any progression) occurs less
frequently than the denial of an implication. After all, implications may
often be more relevant to the reacting character than the explicit utter-
ance of the preceding speaker. In other words, the prevalence of cor-
rections of implications by means of turn-initial ἀλλά can be explained
because of the change of speaker, whereas inside turns a speaker can
more easily create true eliminative constructions, by explicitly mention-
ing and directly negating the element to be substituted.39 Incidentally,
most turn-initial corrections of explicit elements attested in my corpus
indeed correct elements of the speaker’s own words.40
Furthermore, the quantitative analysis of turns and turn-initial
expressions in several Ancient Greek comedies and tragedies has shown
that the plays of Aristophanes contain on average many more turns

37
For all individual instances, see notes 27 (explicit elements), 30 (presupposed
elements), 32 (implications), and 36 (discourse topics).
38
The instance in Vespae 1152 cannot be classified: see section 3.1.
39
Further research on ἀλλά both at turn-initial position and inside turns might
clarify this difference more. Beside that, it would be interesting to find out whether
other discourse markers, in both Ancient Greek and other languages, also have a more
or less specific use at turn beginnings.
40
6 of all 11 occurrences: Ar. Ra. 56, 57, 928; A. A. 276; S. Tr. 1211, 1216.
154 annemieke drummen

than those of Aeschylus, Sophocles and Euripides. This means that tragic
characters usually talk longer than comic characters before another
speaker takes the floor, a difference increased by the longer songs of the
tragic chorus, especially in Aeschylus. As a result, comic dialogues are
of a more realistic and more aggressive nature than tragic dialogues.
With regard to expressions at turn beginnings, the discourse marker
ἀλλά is the most frequent one in comedy, followed by μὰ (τὸν) Δία or
νὴ (τὸν) Δία, which is absent from tragedy. In absolute numbers, of
the four writers investigated Aristophanes uses turn-initial ἀλλά most:
almost 30 times on average per play, as against about 9, 21 and 4 times
in the tragedies of Aeschylus, Sophocles and Euripides, respectively.
The Lysistrata contains by far the most occurrences: 42 turns begin
with ἀλλά. In relative numbers, however, Aeschylus and Sophocles use
turn-initial ἀλλά most: in 6% of all turn beginnings, as compared with
5% of Aristophanic turn beginnings and 1% in Euripides’ work.
All occurrences of turn-initial ἀλλά have in common that they mark a
relation that would also have been possibly present without the particle.
But when ἀλλά is used, alternative interpretations are made impossible.
The relation is explicit, and consequently easier to discern. Thus, turn-
initial ἀλλά reduces the cognitive effort of the addressees and increases
the cohesion of a dialogue. In this way, the use of ἀλλά is similar to
that of (turn-initial) discourse markers in English and French. Even
the study of particles in ancient written sources can thus prove that
the structure of spoken conversation contains features that transcend
the borders of any single language.41

41
I wish to thank Rutger Allan, Mia Drummen, Coulter George and the editors of
this volume for their helpful comments on earlier versions of this paper.
 CHAPTER EIGHT

GREEK PARTICLES: JUST A LITERARY PHENOMENON?

Coulter H. George

1. Introduction

When most modern readers approach Ancient Greek, one of the first
features of the language to attract their attention is the large number
of particles that evade easy translation.1 Examples such as the follow-
ing give the impression that they are particularly common in Platonic
dialogues and drama:
(1) ἀλλ’ οὐδ’ ἐγὼ μέντοι πεσών γε κείσομαι, (. . .).
(Strepsiades) Well, I won’t lie down either after this fall! (Ar. Nu. 126)

(2) ἀλλὰ μὴν οὐδέν γε τῶν ἀγαθῶν βλαβερόν· ἦ γάρ;

But nothing that’s good is harmful, right? (Pl. R. 379b)
This state of affairs led Denniston, in his comprehensive study of Greek
particles, to claim: ‘It cannot be doubted that Greek conversation was
full of particles: at moments of excitement the dialogue of tragedy
and comedy fairly bristles with them’ (1954: lxxii–iii). This assertion,
however, was not supported by any rigorous numerical analysis until
Duhoux applied statistics to the question and published the results in
several articles dealing with different aspects of the problem (1997a,
1997b, 1998, 2006). His findings in fact contradicted Denniston’s intu-
ition: particles are actually less common in dialogical texts and in drama,
leading him to conclude that particles are more typical of written than

1
I would like to extend heartfelt thanks to Gerry Wakker and Stéphanie Bakker, not
only for their keen editorial advice, but also, as organizers of the Sixth International
Colloquium on Ancient Greek Linguistics, held in Groningen in June 2007, for pro-
viding the opportunity to present this material in a setting so conducive to lively and
constructive debate. In particular, the perspicacious comments of Elizabeth Koier were
invaluable in helping to tighten the argumentation of the following paper.
156 coulter h. george

of spoken Greek. But this result, in turn, prompts one to ask how a
scholar with Denniston’s undisputed expertise in the subject could be
mistaken on so fundamental a question. Accordingly, in the following
paper, I aim to build on Duhoux’s work in two ways, effecting in the
end a rapprochement with Denniston’s views. First, I will argue that
Duhoux is at times too quick to label particular instances of particles as
non-dialogical and therefore as characteristic of the written language.
Second, a reassignment of several particles from a non-dialogical group
to a dialogical group disrupts the distinction proposed by Duhoux
between connective particles being non-dialogical and the other (modal
and focus) particles being dialogical. As connective particles are found
in both dialogical and non-dialogical Greek, a more accurate statement
of the difference will turn out to be that representational particles are
more frequent in non-dialogical texts, interactional particles in dialogi-
cal texts; presentational particles, occupying an intermediate position,
can align with either category.

2. When is a particle dialogical?

It is best to begin with a summary of the results presented in Duhoux

1997a, the article in which the issues discussed here are most thor-
oughly examined. Essentially, Duhoux selected five texts as samples
and recorded the number of times various particles occurred in sections
that are ‘dialogué’ or ‘non dialogué’. He then compared these figures
with what one would expect if the particles were evenly distributed in
both types of text. The following table provides a simplified account
of Duhoux’s results:2

2
Particles which occur too evenly in both discourse types for the difference to be
statistically significant are omitted. Included in brackets, however, are particles whose
distribution becomes significant if one requires only 90% accuracy rather than the figure
of 95% accuracy used by Duhoux. In other words, of the forty instances where Duhoux
assigns a particle to one category or the other, two are likely to be statistical flukes and
not, in the end, significant; of the seven bracketed instances that I have added, one is
likely to be such a fluke. It seemed tolerable to admit this additional likelihood of an
error in the interests of fleshing out more fully the individual particles’ inclination to
one side or other of the scale.
 greek particles: just a literary phenomenon? 157

Table 1: Particles that are significantly more common in dialogical or

non-dialogical sections of the stated texts (data from Duhoux 1997a: 22–9)
Text Dialogical Non-dialogical
Xen. Symposium αὖ, οὐκοῦν, [δή, οὖν, τοίνυν] δέ, καί, μέν, μήν, τε
Plato Protagoras ἄρα, ἆρα, γε, δή, ἤ, μέντοι, καί, μέν
οὐκοῦν, οὖν
Plato Meno + ἄρα, ἆρα, γε, μήν, οὐκοῦν, γάρ, καί, οὔτε, [μέντοι]
Apology τοίνυν
Sophocles OC γε, δῆτα, ἦ, [γάρ] μέν, οὐδέ, οὔτε, τε
Ar. Thesm. γε, δῆτα, [οὖν] δέ, ἤ, καί, οὐδέ, τε, [μέν]

We can go one step further in clarifying the proclivity of the individual

particles towards dialogical or non-dialogical discourse types by assign-
ing them one point for each occurrence in the dialogical column and
a negative point for each occurrence in the non-dialogical column to
arrive at a simple dialogical ranking:

Table 2: Particles arranged from most to least dialogical (cf. Duhoux 1997a: 31)
4 3 2 1 0 –2 –3 –4
γε οὐκοῦν δή αὖ ἤ δέ τε καί
οὖν τοίνυν ἦ μέντοι οὔτε μέν
ἄρα μήν οὐδέ
ἆρα γάρ
δῆτα

This chart reveals two facts worth pointing out. First is that there are
ten different particles which prefer dialogical discourse, but only six
that prefer non-dialogical discourse. Now Duhoux is still right to say
that particles are equally likely in either discourse type insofar as the
particles that line up on the non-dialogical side include some of the
most common particles; indeed, the three most common particles in
Classical Attic are, in descending order of frequency, καί, δέ, and μέν.
But as far as sheer diversity of particles is concerned, we find greater
variety among those that are more at home in dialogical sections of
texts. The second point to consider is the nature of the particles towards
either end of the scale. At first glance, it is tempting to characterize the
non-dialogical particles as primarily connective and the dialogical par-
ticles as more mixed in nature. This is, however, a somewhat misleading
158 coulter h. george

oversimplification, as will be discussed in the second half of this paper.

But before we can turn to this second question, we must first examine
more closely the division between dialogical and non-dialogical.
On the one hand, a straightforward dichotomy of this kind is a
natural starting point for conducting the sort of statistical tests that
should help us to see the big picture; on the other, it also runs the risk
of failing to capture nuances only detectable by examining individual
usages of the particles in their native habitat. To illustrate the benefits of
supplementing the results of Duhoux’s study with this latter approach,
let us consider two particles whose location in table 2 merits further
attention: μήν and μέντοι.

2.1. How dialogical is μήν?

Now μήν is a particle that has generally been seen as operating at the
interactional level of discourse, to use Kroon’s terminology: in other
words, it is a particle that helps place what the speaker is saying into
the context of what the addressee is saying or thinking, as opposed to
one that is chiefly directed towards structuring the speaker’s utterance
on its own terms.3 It is thus surprising that, in Xenophon’s Symposium,
Duhoux found μήν to be more common in non-dialogical than in dia-
logical passages: twenty-five of the former, compared to only eight of
the latter.4 But a closer look at these thirty-three examples of μήν shows
that it is rather more dialogical a particle than Duhoux’s figures sug-
gest. The easiest way to show this is to place the examples in question
along a scale of ‘dialogicality’, moving from those examples which are
unquestionably characteristic of dialogue to those which can equally
safely be assigned to a non-dialogical heading.
Most dialogical are the seven examples that occur in sentences that
are the first in a speaker’s turn. In addition to this structural reason

3
For Kroon’s levels of discourse, see Kroon (1995: 59–62, 69–95). In addition to the
interactional level, she also speaks of the representational level, at which the speaker
is ‘[portraying] some real or imaginary world outside the language itself, and often,
but not necessarily, also outside the immediate discourse situation’ (Kroon 1995: 69),
and the presentational level, at which the speaker is ‘[imposing] an organizing and
rhetorical perspective on the ideas conveyed’ (Kroon 1995: 61). For the interactional
use of μήν, see especially Wakker (1997), as well as Sicking (1993: 54–5) and Oréal
(1997: 234, 242).
4
All such figures in this paper are based on all occurrences of μήν, including those
in combinations like καὶ μήν.
 greek particles: just a literary phenomenon? 159

for considering these the most dialogical, they also have incidental
features, such as a first- or second-person pronoun, a vocative, or an
interrogative, that also suggest a context highly conducive to the use
of interactional particles. Typical is example (3):5
(3) καὶ ὁ Φίλιππος, Νὴ Δί’, ἔφη, καὶ μὴν ἔγωγε ἡδέως ἂν θεῴμην Πείσανδρον
τὸν δημηγόρον μανθάνοντα κυβιστᾶν εἰς τὰς μαχαίρας, (. . .).
And Philippus said, ‘By Zeus, I know that I would gladly watch
that populist Pisander learning to do acrobatics with swords (. . .).’
(X. Smp. 2.14)
The next category consists of three examples rather close to the first
seven: μήν occurs in the second or third sentence of a turn, and in all
three sentences there is a first-person pronoun. See example (4):6
(4) οὔτε γὰρ ἔγωγε σπουδάσαι ἂν δυναίμην μᾶλλον ἤπερ ἀθάνατος
γενέσθαι, οὔτε μὴν ὡς ἀντικληθησόμενος καλεῖ μέ τις, ἐπεὶ πάντες ἴσασιν
ὅτι ἀρχὴν οὐδὲ νομίζεται εἰς τὴν ἐμὴν οἰκίαν δεῖπνον προσφέρεσθαι.
For I could no more be serious than become immortal; nor, I say,
does anyone invite me to dinner thinking they’ll be invited in return,
since everyone knows that I don’t make a habit at all of having dinner
at my place. (X. Smp. 1.15)
Although Duhoux does not specify which examples he includes under
his eight dialogué examples, one suspects that they must include all of
the first set (that is, the set including example 3), but only one of these
three: but surely all three examples of this class ought to be considered
dialogical.
The third, and central, category consists of twelve examples that
occur in the middle of a continuous speech, so at first glance it might
seem reasonable to count them as non dialogué. At this point, how-
ever, it is apposite to bring into play a term introduced by Kroon in
her discussion of interactional particles, namely diaphony. She notes
that, even in the monological discourse type—that is, even in a passage
all in the voice of a single speaker—there are times when such a text
‘contains explicit references to the communicative frame in which it is
integrated’ (1995: 113). Passages of this sort are marked by, inter alia,
first- and second-person pronouns, metadiscursive expressions (such as
as I was saying), and rhetorical questions. It is this discourse type that

5
See also X. Smp. 3.11, 3.12, 3.13, 4.23, 4.55, 5.7.
6
See also X. Smp. 4.47, 8.7.
160 coulter h. george

Kroon calls diaphonic, and her work shows that particles, like Latin at
and enim, that are characteristic of fully interactional passages are also
found frequently in such diaphonic passages. And so it is with μήν in
Xenophon’s Symposium as well: twelve of the times that it occurs in
the middle of a speech, there is a first-person pronoun or verb in the
sentence, a clear sign that μήν gravitates towards sentences where the
speaker makes explicit reference to himself. See example (5):7
(5) ἐπειδάν γε μὴν ἐν τῇ οἰκίᾳ γένωμαι, πάνυ μὲν ἀλεεινοὶ χιτῶνες οἱ
τοῖχοί μοι δοκοῦσιν εἶναι, (. . .).
Then, when I’m at home, I think the walls are like a perfectly warm
tunic (. . .). (X. Smp. 4.38)
A fourth category consists of eight passages that also have good claim to
be considered diaphonic, although there are no criteria quite as objective
as a first-person pronoun to mark them as such. Still, two examples
occur in rhetorical questions;8 two examples occur in a speaker’s
response to a rhetorical question;9 in three, a first-person point-of-view
has carried over from a preceding sentence (if without explicit restate-
ment of the pronoun);10 and in the final example the collocation ἦ μήν
occurs in indirect speech right after the verb συνομόσαι:
(6) <ὥστε> (. . .) ἅπαντας συνομόσαι ἂν ἦ μὴν τὸν παῖδα καὶ τὴν παῖδα
ὑπ’ ἀλλήλων φιλεῖσθαι.
(. . .) such that all would have agreed in swearing that, assuredly, the
boy and the girl were loved by each other. (X. Smp. 9.6)
In the end, there remain only three examples in Xenophon’s Sympo-
sium where μήν occurs in a context that does not have any dialogical
or diaphonic characteristics. See example (7):11
(7) καὶ μὴν ἐν μὲν τῇ τῆς μορφῆς χρήσει ἔνεστί τις καὶ κόρος, ὥστε ἅπερ
καὶ πρὸς τὰ σιτία διὰ πλησμονήν, ταῦτα ἀνάγκη καὶ πρὸς τὰ παιδικὰ
πάσχειν·
And there is also a certain satiety that comes with repeated exposure
to physical beauty, such that quite the same thing one feels towards

7
See also X. Smp. 4.13, 4.15, 4.32, 4.38 (there is a second example, in addition to
the one given in (5), 4.44, 7.3, 8.2, 8.3 (2×), 8.31, 8.37.
8
See X. Smp. 8.18, 8.21.
9
See X. Smp. 8.20, 8.21.
10
See X. Smp. 4.42, 7.3, 8.13.
11
See also X. Smp. 8.26, 9.4.
 greek particles: just a literary phenomenon? 161

food on account of becoming full, one also necessarily feels towards

a lover (. . .). (X. Smp. 8.15)

2.2. How dialogical is μέντοι?

Nor is it only μήν that is more dialogical than Duhoux’s figures sug-
gest. Consider μέντοι as well. Duhoux had classified thirteen of its
uses in the Symposium as dialogical, and seven as non-dialogical: not
a sufficient difference to be statistically significant considering the low
figures. But once again, if examples of μέντοι are classified into the same
five categories of dialogicality, thirteen (presumably Duhoux’s thirteen
dialogical examples) fall into the first two categories, as they occur in
the first, second, or last sentence of a turn;12 another three occur in
the middle of a turn, but with a first- or second-person pronoun or
vocative explicitly signaling the diaphonic discourse type;13 in another
two examples, such a signal does not occur in the unit with μέντοι, but
does occur in the preceding unit;14 and there are only two remaining
examples where the discourse type is purely monological.15

2.3. Plato and Thucydides

This state of affairs extends to other authors as well. Another pair of
texts examined by Duhoux is Plato’s Meno and Apology, in which he
counts six examples of μέντοι as dialogué and thirteen as non dialogué
(1997a: 22). Again, it is true that we only find seven examples of
μέντοι in the first or second sentence of a turn.16 But every one of the
other twelve examples occurs in a clause that also has some feature
of diaphony, most often a first-person pronoun, but also oaths and
directives.17 Or, to return to μήν, we find that it is used only nineteen

12
See X. Smp. 1.12, 3.13, 4.5, 4.33, 4.61, 6.1, 6.8, 6.10, 8.5 (first sentence of the turn);
2.25, 4.10, 4.49 (second sentence); 4.4 (last sentence).
13
See X. Smp. 4.63, 8.9, 8.15.
14
See X. Smp. 4.17, 4.24.
15
See X. Smp. 2.3, 4.59. Possibly only 2.3 should count as purely monological, as at
4.59 the μέντοι comes in the first unit of a turn, and it is only the fact that it is a turn
of the narrator that prevents this from falling into the first category.
16
See Ap. 26e (2×), Men. 72d, 92e, 94b (μέντοι in first sentence), Ap. 26c, Men. 86c
(μέντοι in second sentence).
17
See Ap. 17b (with oath), 17c, 20d (with directive), 29c (in the middle of a turn,
but at the start of a direct quotation), 30d, 32c, 35d (with oath and directive), 37c, 38d,
41e, Men. 80d (2×). The only one of these examples that could be seen as approaching
162 coulter h. george

times in all of Thucydides. Compared to the thirty-three examples in

Xenophon’s Symposium, this low figure in what is primarily a narrative
work of history suggests right from the start that μήν is indeed more
at home in dialogical contexts. Now this relative infrequency could
of course be due solely to some idiosyncratic distaste of Thucydides’
for the particle, but the contexts in which it occurs show otherwise.
Thucydides most often uses μήν in the collocation ἦ μήν (8×), which
is always found introducing indirect discourse after verbs of swearing,
promising, or threatening.18 The second most common environment for
μήν is in speeches (6×), where it occurs five times in diaphonic contexts
and only once in a monological context.19 Finally, the remaining five
examples, while they all come from what might be considered narra-
tive in a broad sense, also all occur in contexts colored by diaphony.
In particular, they are predominantly found in well-known passages
where Thucydides’ persona as a historical observer comes to the fore to
comment on the situation at hand.20 This authorial intrusion is signaled
explicitly by first-person pronouns at 1.3.3 (Homer does not speak of
βάρβαροι because there are as yet no Ἕλληνες) and 6.55.3 (Hippias,
not Hipparchus, was the elder son of Pisistratus, pace popular belief ).
Another signal of diaphony, a switch to the present tense, occurs at
2.97.6 (the Scythians, if united, would conquer all, despite their not
equalling other nations in other respects) and 3.82.6 (the discussion of
στάσις in connection with Corcyra).21 The final example comes from
the narrative of the Sicilian Expedition; although there are no similarly
explicit grammatical features of diaphony here, the running use of the
imperfect and the evaluative content of the passage still suggest that
Thucydides’ persona as an engaged observer is prominent:22

the monological type is Ap. 30d, where the clause with μεντἄν does not itself contain a
first-person form, but even here οἴομαι is found in neighboring clauses on both sides: οὐ
γὰρ οἴομαι θεμιτὸν εἶναι ἀμείνονι ἀνδρὶ ὑπὸ χείρονος βλάπτεσθαι. ἀποκτείνειε μεντἂν
ἴσως ἢ ἐξελάσειεν ἢ ἀτιμώσειεν· ἀλλὰ ταῦτα οὗτος μὲν ἴσως οἴεται καὶ ἄλλος τίς που
μεγάλα κακά, ἐγὼ δ’ οὐκ οἴομαι.
18
See 4.88.1, 4.118.1, 5.38.1, 5.50.1, 6.72.5, 8.33.1, 8.75.2, 8.81.3.
19
There is a first-person pronoun in the same sentence as μήν at 1.82.1, 2.38.1,
and 4.86.1; there is a second-person pronoun in the same sentence at 1.70.4; there is
a first-person pronoun in the preceding sentence at 6.17.5; the monological example
is 1.142.2.
20
For a discussion of the different personas adopted by Thucydides and the rami-
fications they have for the expression of tense, see Bakker (1997b).
21
For the present tense as a signal of diaphony, see Kroon (1995: 114).
22
Indeed, this is one passage that Bakker mentions as illustrating what he calls
Thucydides’ mimetic mode, in which the discourse is ‘presented from the internal
 greek particles: just a literary phenomenon? 163

(8) ἔφερον δὲ οὐδὲ ταῦτα ἱκανά· σῖτος γὰρ οὐκέτι ἦν ἐν τῷ στρατοπέδῳ.

καὶ μὴν ἡ ἄλλη αἰκία καὶ ἡ ἰσομοιρία τῶν κακῶν, ἔχουσά τινα ὅμως
τὸ μετὰ πολλῶν κούφισιν, οὐδ’ ὣς ῥᾳδία ἐν τῷ παρόντι ἐδοξάζετο,
(. . .).
And not even what they carried was enough, for there was no longer
any food in the camp. And also the rest of their disgrace and their
equal share in the misery, although it still provided some lightening
of the burden, the fact of being one among many, even so did not
seem easy in the present circumstances (. . .). (Th. 7.75.5–6)
In short, μήν is a particle that Thucydides uses only under a very limited
number of circumstances: in introducing the content of an oath with
ἦ μήν, in speeches, or in his more explicitly personal observations about
the history he is writing.
More generally, we have seen that further consideration of what
should be regarded as dialogical or non-dialogical can significantly
affect our conclusions about whether an individual particle tends more
towards one context or the other. In particular, an awareness of the
middle ground of diaphony, with its underlying inclination towards
the dialogical despite a superficially monological appearance, shifts the
balance much more in favor of the view that particles are relatively par-
tial to dialogical contexts. Now, one could of course question whether
diaphony is a valid concept to bring to this discussion: pseudo-dialogical
isn’t necessarily the same as dialogical. But Kroon’s work on Latin sug-
gests that diaphonic and dialogical text-types do go closely together,
and, in addition to the findings from Xenophon and Plato, the use of
μήν in Thucydides provides especially clear evidence that this affinity
exists in Greek as well. As far as I can tell, not one of the Thucydidean
examples would have counted as dialogical by Duhoux’s standards, yet
oaths, speeches, and personal observation all stand out as being more
closely allied with spoken, dialogical language than ordinary narrative
is. It is reasonable to expect an analysis of the dialogicality of particles
to take this into account. In passing, we have also come across one of
the pitfalls of applying statistics to solve linguistic problems: statistical
tests are only as good as the data one feeds into them, and Xenophon’s
Symposium is too short for one to gather meaningful data on μήν or
μέντοι unless one reduces the categories to a stark dichotomy between

standpoint of the observer, and largely conducted with imperfect verbs’ (1997b: 39,
42). For further discussion of narrative modes in Ancient Greek, see Allan (2007, this
volume).
164 coulter h. george

dialogical and non-dialogical, ignoring the crucial fact that there are
shades of grey in between.23

3. How connective are dialogical particles?

But even if Duhoux has counted as non-dialogical contexts that have

good reason to be considered dialogical, his figures still provide a valu-
able source of information about the dialogicality of particles relative
to each other—for his decision about where to draw the line between
the dialogical and the non-dialogical will have skewed the results for all
particles in the same direction.24 And Duhoux comes to the conclusion
that what the most consistently non-dialogical particles have in common
is that they act as ‘conjonctions de coordination’; furthermore, as these
connective particles are thus less frequent in dialogical discourse, he
suggests that spoken Greek will have had more asyndeton than written
Greek (1997a: 42–3).25 If anything, this distribution would in fact sup-

23
In theory, one could refine Duhoux’s study by selecting a larger corpus, subdividing
it into perhaps four or five text types graded along a finer-grained scale of dialogicality,
then applying the same statistical tests to see whether the particles under investiga-
tion have any proclivity to one or more of these more narrowly defined text types. In
practice, I am doubtful that this would be an efficient way to study Greek particle use.
First, one would have to devise objective tests by which one could assign every single
sentence in the corpus to one category or another. (Such strictness is necessary in order
to determine the base word-count for each text type.) But the more text types one sets
up, the more difficult it is to achieve such objectivity, as the linguistic features one
might choose as distinguishing criteria do not always line up in parallel. For instance,
the presence of a first-person pronoun and the fact that a sentence is first in a turn
would both appear to be clear objective features of some degree of dialogicality. But
in setting up a more precise scale, how would one rank the relative dialogicality of a
first-in-turn sentence in a stilted third-person-only philosophical dialogue as against
that of a rhetorical question enlivened with both first- and second-person pronouns,
but occurring in the middle of a monologue? Second, even if one expands the number
of text types to four or five, this would still be insufficient to capture everything. Con-
sider the Thucydidean usage of μήν, the accurate assessment of which would require
the setting up of additional categories of dialogicality to account for oaths in indirect
speech, direct speech in the context of a narrative history, and personal observation.
It is not hard to imagine such categories multiplying still further.
24
In other words, we can be reasonably sure that particles that Duhoux labels dia-
logical are indeed dialogical: αὖ, for instance, which Duhoux says is 9× dialogical and
2× non-dialogical in Xenophon’s Symposium (1997a: 25), occurs in the first, second,
or last sentence of a turn all but once (8.30).
25
To be precise, the argumentation at this point of Duhoux’s article is no longer
based on whether the particles in question are found more often in dialogical or non-
dialogical contexts, but rather on whether they are more typical of drama or prose
 greek particles: just a literary phenomenon? 165

port Denniston’s comment, which Duhoux claims to refute, about the

liveliness of the particles in Greek dialogue. To place the quotation in
context, Denniston earlier in the paragraph had specifically excluded
the connective particles: ‘Particles, apart from the necessary connec-
tives, are like ignition sparks’ (1954: lxxii). Be that as it may, this idea
that non-dialogical written Greek is particularly characterized by con-
nective particles and dialogical spoken Greek by asyndeton is also one
that merits further attention.
In particular, it is not clear exactly which particles are to be included
in the category of ‘conjonctions de coordination’.26 Duhoux takes the
label to mean particles whose function it is to specify ‘une liaison
entre deux éléments de même niveau hiérarchique’ (1997a: 15–6). This
definition should probably be understood in fairly broad terms. First,
Duhoux counts μέν and δή as connective (1997a: 42).27 Yet both of these

(‘textes scéniques ~ non scéniques’). In the present paper, I have collapsed my response
to these two arguments together because they both tend in the same direction anyway,
with Duhoux considering both dialogical and dramatic texts to be closer to the spoken
language than non-dialogical and prose texts (1997a: 18).
26
Much has been written on how to categorize the various types of particles,
starting with the ancient grammarian Dionysius Thrax (see Sluiter (1997)), whose
classification of σύνδεσμοι (conjunctions) as παραπληρωματικοί or συλλογιστικοί
developed into Denniston’s division of particles into those that express a mode of
thought or mood of emotion in isolation (‘adverbial’) (1954: xxxvii–ix) and those that
‘[establish] a relationship between ideas’ (‘connective’) (1954: xxxix–l). Recent work
on such classifications in Greek and Latin has primarily been based on the work of
Dik (1968: 34–41) and Pinkster (1972); Van Ophuijsen offers a useful summary of this
work, noting in particular the potential distinction between coordinators in a strict
sense and connectors in a slightly broader sense (1993: 77–9). Both Van Ophuijsen
(1993: 79) and, in the same volume, Sicking (1993: 45) decide that matters like the
classification of particles and the concept of connectivity are better discussed as scalar
phenomena than as discrete categories, a conclusion with which I agree wholeheart-
edly. As Denniston drily noted, ‘[t]he line between connectives and non-connectives
cannot be rigidly drawn’ (1954: xliii note 2). Kroon combines both a top-down and a
bottom-up approach to connective particles: that is, she looks both at the relations they
mark (1995: 7–33) and the particles themselves that mark them (1995: 34–57). She too
arrives at the conclusion that there is often no clear distinction between connective
particles and situating particles (which situate their host units against the background
of the extra-textual reality) (1995: 63–4). More recently, Duhoux has argued for a
three-way split between ‘des conjonctions de coordination, des modalisateurs, et des
intonateurs’ (2006: 522). Moving beyond Greek and Latin, we find that Fraser (1999)
has argued that the term ‘discourse marker’ should be restricted to essentially con-
nective particles that signal a relationship between the host clause and the preceding
clause. But cf. Schiffrin’s broader definition of discourse particles, as discussed in her
convenient overview of the issue (2001).
27
Although δή comes across as a relatively dialogical particle in table 2, it is
more common in prose than in drama, and Duhoux thus includes it in the list of
166 coulter h. george

particles can combine with καί or ἀλλά, and so fail the juxtaposition
test, whereby connective particles, in the strictest sense, ought not to
occur in combination with each other.28 Second, in later work, Duhoux
explains more fully that the ‘liaison’ created by such particles ‘peut
être de mot à mot, de syntagme à syntagme, de proposition à proposi-
tion ou de phrase à phrase’ (2006: 521 note 7). This list of functions
is extensive enough that it would seem to include any particle that
articulates the logical or semantic relationship between its host clause
and the preceding clause.29
But a broad definition along these lines means that one must abandon
Duhoux’s further conclusion, that spoken Greek had more asyndeton
than written Greek. For, with so wide a range of connective particles,
the fact that dialogical texts have fewer examples of, say, καί or μέν
does not mean that they had more asyndeton: it could just mean that
they use other connective particles instead. Indeed, the list of particles
that Duhoux sees as typical of drama (and therefore of spoken Greek)
includes not only γάρ, which one might just be able to exclude from
his connective category on the grounds that its host clause is not at the
same ‘niveau hiérarchique’ as the preceding clause, but also μηδέ, μήτε,
οὐδέ, and τε (1997a: 39). But surely these particles could be seen as
decreasing the amount of asyndeton in the texts that Duhoux himself
uses as a proxy for spoken Greek. Furthermore, other, less obvious
candidates for connective particles may also have blocked asyndeton in
such texts. The combination γε μήν, not typically grouped together with
the connectives, consisting as it does of two tone-of-voice particles, has
good reason to be considered connective in Xenophon on distributional
grounds: in ten of its thirteen occurrences in the Symposium, it occurs
after a full stop or question mark, and, if it were omitted, asyndeton
would result.30

connective particles he believes to have been more typical of the written language. Cf.
my note 25.
28
For the juxtaposition test and related phenomena, see Dik (1968: 34–5), Pinkster
(1972: 158–62), and Van Ophuijsen (1993: 78).
29
For this view of connective particles, see Fraser (1999).
30
See X. Smp. 3.11, 3.12, 4.13, 5.7, 8.3, 8.7, 8.13, 8.18, 8.37, 9.4; twice it is used
similarly after a raised stop (4.38, 8.2); only once is it used after a comma (4.38, second
occurrence), and even here it is best considered connective, as its host clause contains
a new main verb: ἐπειδάν γε μὴν ἐν τῇ οἰκίᾳ γένωμαι, πάνυ μὲν ἀλεεινοὶ χιτῶνες οἱ
τοῖχοί μοι δοκοῦσιν εἶναι, πάνυ δὲ παχεῖαι ἐφεστρίδες οἱ ὄροφοι, στρωμνήν γε μὴν
οὕτως ἀρκοῦσαν ἔχω ὥστ’ ἔργον μέγ’ ἐστὶ καὶ ἀνεγεῖραι (‘Then, when I’m at home, I
think the walls are like a perfectly warm tunic, and the roof like a very thick cloak, and
 greek particles: just a literary phenomenon? 167

In the end, a more fruitful approach to understanding why some

particles prefer dialogic and diaphonic contexts is to examine the level
of discourse at which they function.31 The more non-dialogical particles
(καί, τε, οὔτε) tend to operate on the representational level of discourse,
whereas the more dialogical particles are generally interactional (οὐκοῦν,
δή, τοίνυν, ἆρα, δῆτα; many occurrences of γε probably belong here as
well); the presentational particles, as befits the intermediate position of
this level, are split between the more non-dialogical (μέν, δέ, οὐδέ) and
the more dialogical (οὖν, ἄρα).32 This distribution is not surprising: (i)
interactional particles are going to cluster in dialogical texts almost by
definition; (ii) presentational particles, as they organize the discourse
from a rhetorical standpoint, presuppose a second-person audience to a
greater extent than representational particles, and so might be expected
to occur more in dialogical texts—but they also play an important role
in managing the flow of information in speakers’ longer turns, and so
have good reason to be found in non-dialogical passages as well;33 (iii)
representational particles are left as the ones most suited to monologi-
cal texts, mostly for the negative reason that they serve no function
peculiar to dialogical discourse that would elicit them there to the
same extent as particles operating on the other two levels.34 The only
potential problem is γε, which, though very characteristic of dialogi-
cal texts, is generally considered representational because of its role as

I have blankets that are so satisfactory that even waking me up is a big deal.’). Insofar
as Greek nearly always avoids asyndeton except (i) when a connective demonstrative
is present, (ii) after an interruption, or (iii) for expressive effects (Sicking 1993: 40–4),
a combination like γε μήν can be considered connective in at least a broad sense of
the term. Denniston, noting Xenophon’s partiality towards γε μήν, remarks that it is
often used as a variant of ἀλλὰ μήν and καὶ μήν (1954: 347).
31
For the levels of discourse, see note 3.
32
That δέ in particular should be considered a presentational particle is evident
(i) from Bakker’s work, which establishes that, unlike καί, which extends an existing
discourse unit, δέ opens a new unit (1993, 1997c: 54–84); (ii) from Kroon’s work,
which describes autem—which she considers a presentational particle—in terms very
similar to those in which Bakker describes δέ (Kroon 1995: 226–80, especially 247–69
and the explicit comparisons at 230, 235). But Kroon also signals the closeness of the
representational and presentational levels by pointing out the occasional difficulty of
assessing the boundary between the two (1995: 71).
33
I am grateful to the editors of the volume for calling my attention to this latter
point.
34
Here one must bear in mind the cautionary note sounded at the beginning of
section 3: considering the extent to which Duhoux’s figures are skewed in favor of
labeling particles as non-dialogical, these representational particles are probably not
actively favoring monological texts so much as they are evenly distributed in both
non-dialogical and dialogical contexts.
168 coulter h. george

a scope particle. But in fact, many of its occurrences should instead

be assigned to the interactional level: the limitations it imposes on its
host-unit are often rhetorical and challenging rather than objective in
nature,35 it frequently collocates with interactional particles,36 and, as
Duhoux showed, it is the particle most likely to occur in dialogue. It
need not bother us to have a particle operating on two different levels
in this way. Indeed, Kroon argues similarly that vero can be either
representational or interactional.37 Still, even though the trait shared
by Duhoux’s non-dialogical particles is not so much that they are
connective as that they operate at the representational and, to a lesser
extent, presentational levels, it remains the case that there is consider-
able overlap between the non-dialogical particles and the connective
particles. This is not surprising, however, since also in English, Schiffrin
observes, connective particles operate primarily at the representational
level (1987: 315–7).38

4. Conclusion

In conclusion, we have seen (i) that particles are in fact more typical
of dialogical Greek than has recently been argued, and (ii) that those
particles that are more common in non-dialogical Greek have as a com-

35
A full exploration of this statement is beyond the scope of this article, but note
the considerable space Denniston allots to ‘exclamatory γε’ and ‘γε in answers’ (1954:
126–30 and 130–8 respectively).
36
Note in particular the collocation γε μήν, which, as Buijs observes, is characteristic
of those sections of Xenophon’s Agesilaus that contain the author’s evaluative com-
mentary on Agesilaus’ life (as opposed to the sections with simple diegetic narrative)
(2007: 126–7). For the use of such collocations as a general guide to the level at which
a particle is functioning, see Kroon (1995: 117); for a particular instance similar to the
one at hand, cf. her observations of the collocations of vero (as against autem) with
interactional particles as illustrative of its interactional status (1995: 330).
37
See Kroon (1995: 281–4). Another similarity between vero and γε is their shared
affinity for the first-person pronoun: Kroon cites the frequent occurrence of ego vero
(1995: 302), and the combination of ἐγώ and γε is common enough to have caused
the univerbation of ἔγωγε and ἔμοιγε.
38
Schiffrin, following Halliday and Hasan (1976: 26), calls this level ‘ideational’.
While it is true that we would not necessarily expect particles in written Greek and
English to work in the same way—formal written English clearly allows asyndeton to
a greater extent than Greek does—Schiffrin’s findings apply to the spoken language,
a register where we might expect cross-linguistic parallels to be more abundant: lan-
guage-specific training in school is more likely to affect writing than speech, e.g. by
discouraging the ‘unnecessary’ use of sentence-initial And and But.
 greek particles: just a literary phenomenon? 169

mon denominator not that they are connective (otherwise we would

expect a particle like οὖν, or even the combination γε μήν, to number
among them) but that they function primarily at the representational
and presentational levels. To what extent, then, can we extrapolate from
the fact that a particle is relatively non-dialogical to the conclusion that
it was more typical of the written than of the spoken language?39 At
first glance, this seems reasonable enough, insofar as conversation is
more prototypical of the spoken language than monologue is (Levinson
(1983: 284), Mey (2001: 136)). But further consideration suggests that
the dialogicality of a word is potentially a very misleading proxy for
its orality. This is perhaps most easily demonstrated with Homer. The
Homeric corpus is richer in particles than most classical Greek texts are
(Duhoux (1998: 25–7, 37–8; 2006: 525)); this suggests right from the
start that they were not merely a literary device to aid the comprehen-
sion of the written word. Obviously, the diachronic discrepancy means
that we cannot directly compare Homer with fifth-century Attic and
conclude that spoken Greek had more particles than written Greek.
But it is also the case that particles are more common in narrative
than in speeches in Homer (Duhoux (1998: 20–3, 35–7; 2006: 526)).
There is thus a clear discrepancy between the frequency of particles in
dialogical and non-dialogical Greek that is independent of the variable
of ‘orality’, thus rendering questionable the use of the one as a proxy
for the other.

39
For a general discussion of the difficulties with setting up a spoken/written
dichotomy for Ancient Greek, see Dickey (1996: 30–42, especially 30–1).
 CHAPTER NINE

TOWARDS A TYPOLOGY OF THE NARRATIVE MODES

IN ANCIENT GREEK. TEXT TYPES AND NARRATIVE
STRUCTURE IN EURIPIDEAN MESSENGER SPEECHES

Rutger J. Allan

1. Introduction1

Narrative is a mixed genre. Since Plato made the well-known distinction

between διήγησις and μίμησις (Rep. 392c–394c), it has been recognized
that narratives do not consist purely of narrative sentences. Typically,
narrative discourse also includes a variety of non-narrative ingredients
such as descriptions, character discourse and metanarrative elements.
The various text types or narrative modes (the term I will use here)2 of
which narratives are composed have been the object of both linguistic
and narratological studies, such as Genette (1972), Chatman (1978,
1990), Bonheim (1982), Fleischman (1990), Chafe (1994), Longacre
(19962), Fludernik (2000), Roulet, Fillietaz & Grobet (2001), Smith
(2003) and Adam (2005). Within the field of Ancient Greek linguistics,
however, the issue of narrative modes has not yet been addressed in a
comprehensive manner.3 This paper aims to make a first step towards
a typology of narrative modes in Ancient Greek narrative.
The narrative modes, in my conception, hinge on a central conceptual
aspect of narration—the relation between the point of view of the nar-
rator and the presentation of the text. Texts typically involve a range

1
I wish to thank Gerard Boter, Inez van Egeraat and Irene de Jong for their valuable
comments on an earlier version of this paper.
2
I owe this term to Genette (1972), Bonheim (1982), and Chafe (1994). Smith (2003)
uses the more general term discourse mode.
3
Examples of studies touching on the issue of text types or related notions in Ancient
Greek are Basset (1989b), who builds on Benveniste’s distinction between discours and
récit to explain the occurrence of augmented verbs in Homer, and Bakker (1997b) who
distinguishes two narrative modes, the mimetic and the diegetic mode, to account for
certain aspect-alternations in Thucydides. With regard to Latin, Suzanne Adema is
currently working on a Ph.D.-thesis on the discourse modes in Vergil’s Aeneid. See
also Adema (2007) and Kroon (2002, 2007a).
172 rutger j. allan

of different relationships between the narrator’s point of view and the

world which is referred to verbally (e.g. Genette 1972, Bonheim 1982,
Fleischman 1990 and Chafe 1994). The narrator may refer to a narrated
world from an external retrospective point of view, or from a point of
view internal to the narrated world. The narrator can also refer to the
world outside the narrated world, shared by the narrator and the nar-
ratee. Furthermore, the relation between the narrator’s perspective and
the narrated world is also relevant to the particular manner in which
the text advances. For example, texts can display progress according to
temporal change, or according to spatial change (Smith 2003).
In this paper, narrative modes are primarily approached as linguistic
phenomena. The conceptual features of the narrative modes as they
were described above are reflected in formal linguistic properties, such
as tense and aspect, particles, and modality. Therefore text-segments
which are marked by one particular narrative mode constitute discrete
(textual) linguistic units. Although the narrative modes are primarily
thought of as linguistic notions, they are also associated with a number
of special narratological features.
In section (2) of my paper, I will introduce a typology of four distinct
narrative modes, and I will discuss their specific linguistic and narrato-
logical properties. The four narrative modes which I distinguish are the
displaced diegetic mode, the immediate diegetic mode, the descriptive
mode, and the discursive mode.4 It will be shown that tense-aspect-
marking is the most important distinctive linguistic feature of the
narrative modes.5 The textual corpus on which my analysis is based
consists of the Euripidean messenger speeches.6 My typology is however
intended to account for all Ancient Greek narrative texts.
The third section of my paper addresses the relationship between
the narrative modes and plot-structure. The linguistic units which are
marked by the narrative modes have a certain rhetorical function in

4
My typology of narrative modes resembles that of Roulet, Fillietaz & Grobet (2001),
who propose a trichotomy of discourse types: discours narratif (cf. my diegetic mode),
descriptif and déliberatif (my discursive mode). A difference between their typology
and mine is my distinction within the diegetic mode between an immediate and a
displaced subtype.
5
The crucial importance of tense and aspect as linguistic markers of the narrative
modes or discourse modes has also been recognized by Weinrich (2001), Fleischman
(1990), Chafe (1994), Smith (2003), and Adema (2007).
6
The corpus of Euripidean messenger speeches is based on De Jong (1990: 179).
However, I exclude Hel. 605–621 from my corpus since it lacks a substantial narrative
section.
 towards a typology of the narrative modes 173

the global structure of the narrative. In this way, the narrative modes
can be said to function as a bridge between the sentences (micro-level)
on the one hand, and the more abstract structure that the text evokes
(macro-level) on the other (Smith 2001: 9). The function of the nar-
rative modes within the larger organisation of the narrative will be
demonstrated by an analysis of the messenger speech in Euripides’
Andromache (section 4).

2. Four narrative modes

2.1. The two diegetic modes: displaced and immediate

It might not come as a surprise that the narrative mode in which
stories are typically told is the diegetic mode. This mode is the default
narrative text type. Without it, there is no story. In the diegetic mode,
the text progresses as narrative time advances (cf. Smith 2003: 14).7
It presents a sequence of events and states which are connected by a
causal or other consequential relation. The typical tenses used in the
diegetic mode are aorists, imperfects, and historical presents. Typical
discourse particles are γάρ, δέ, καί, οὖν, that is, particles that function
on the presentational (i.e. text-structural) discourse level.
The diegetic mode has two subtypes: the displaced and the immediate
diegetic mode.8 The most typical type of diegetic mode is the displaced
diegetic mode.9 In this mode of narration, the narrator is displaced
with respect to the narrated events, that is to say, there is a distance
between the Discourse-Now and the Story-Now. A characteristic feature
of the displaced diegetic mode is its high degree of narratorial control
(Kroon 2002: 191). The narrator recounts the events from a point of
view external to the story world, and, because of his/her retrospective
knowledge, (s)he has a complete overview of the complex of events.

7
I do not adopt the term Narrative Mode used by Smith in order to avoid a possible
confusion with the general term ‘narrative mode’ as it is used by Bonheim (1982) and
Chafe (1994). Note, generally, that the terminology used in the literature about narrative
modes/discourse modes/text types is rather confusing. Terms like diegetic or mimetic
are employed with sometimes slightly, sometimes very different senses. I fear that this
paper will not help to deconstruct the terminological tower of Babel.
8
In Allan (2007) I have analysed the use of these two modes in the narrative of
Thucydides. My displaced diegetic and immediate diegetic modes are quite similar to
Adema’s narrative mode and directing mode, respectively (Adema 2007).
9
The term displaced mode I owe to Chafe (1994).
174 rutger j. allan

This overview enables him/her to manipulate the presentation of events

in all manner of ways. Thus, (s)he may make a distinction between
foreground—typically marked by the aorist tense—and background
(or framework)10 of the story. Another consequence of the narrator’s
retrospective knowledge is the ability to indicate the exact temporal or
causal relation between two events. In this way, the narrator reveals
himself as an intermediary in the narrative.11
The immediate diegetic mode is marked with respect to the displaced
diegetic mode. In the immediate diegetic mode the narrator acts as
if there is no spatial and temporal distance between the experience
and the reporting of the events. In other words, there is no distance
between Discourse-Now and Story-Now. This gives the suggestion of
an eyewitness report. The consequence of this mode of narration is
that the narrator will pretend to have little control over the way in
which the story is told. The events are, therefore, necessarily narrated
in their chronological order, without variations in speed. All narrated
events are treated as equally significant and foregrounded. Finally, the
narrator, (creating the illusion of ) being ‘overcome’ by the impact of
the scene, will be less inclined to express his/her personal view on the
events. In other words, the narrator is less visibly present as an inter-
mediary voice.
The essential distinction between the two diegetic subtypes is their
differing use of tense. While the displaced diegetic mode is marked
by the past tense—typically an alternation of aorists and imperfects—
denoting conceptual distance, the immediate mode is signalled by
the use of the historical present.12 In my view, the historical present is the
tense used to create the impression of immediacy,13 of presence at the
scene.14 This effect has also been noted by Rijksbaron:

10
A framework is an atelic state of affairs within which other, telic, states of affairs
may occur (Rijksbaron 20023: 11).
11
For the notion of mediated narration, see Chatman (1978: 32–3, 146–7) and
Bonheim (1982: 39–41).
12
The deployment of the historical present as a marker of immediacy has also
been noted by Chafe (1994: 207–11). Similar observations with regard to the mimetic
mode can be found in Fleischman (1990) and Kroon (2002). The idea of immediacy
in Homeric narrative is discussed by Bakker (2005: 127–35, and elsewhere).
13
The abstract meaning of the Ancient Greek present tense (comprising both its
actual and its historical uses) can be viewed as signalling epistemic immediacy, whereas
the past tense (including its modal uses) can be seen as a marker of epistemic distance
(see also Langacker 1991: 240–9).
14
De Jong describes the function of the historical present in the Euripidean mes-
senger speeches along the same lines (De Jong 1991: 39). Compare also the historical
present in English, which is characterised by Langacker as ‘[T]he speaker describes a
 towards a typology of the narrative modes 175

(1) [In a number of nuances of the historical present] the notion of

‘present’ may play a part to the extent that a ‘pseudo-present’ or
‘pseudo-moment of utterance’ is created: the narrator plays the role
of an eyewitness. (Rijksbaron 20023a: 22)
Apart from tense usage, another linguistic difference between the
displaced and the immediate diegetic mode is the use of particles. As
stated above, in the immediate mode the narrator pretends to have less
control over the narration and to be less prominently present in his/
her intermediary role as an organizer and commentator of the story.
Therefore, narration in the immediate diegetic mode will involve fewer
text-structural and interactional particles. These particles specify the
relationship between different discourse units, or explicate the narrator’s
attitude with respect to the narrated events. The difference in particle
usage between the two subtypes of the diegetic mode can be observed
in the following table:

Table 1: Use of particles in Euripides’ messenger speeches15

Displaced Immediate Descriptive Discursive Total
Diegetic Diegetic
ἀλλά 16 1 13 18 48
ἄρα 1 0 0 1 2
ἆρα 1 0 0 1 2
αὖ 5 0 0 0 5
γε 3 1 0 4 8

previous sequence of events as if they were unfolding right now, before his eyes; he
takes a hearer through them step by step, achieving a sort of ‘vividness’ by portraying
them as immediate’ (Langacker 1991: 267). We should be careful, however, to resort
to the rather impressionistic epitheton ‘vivid’ too readily. For critical discussions of
the notion of vividness in connection with the historical present, see Sicking & Stork
(1997: 131–4), Rijksbaron (2002b: 257, 261–2). The visualizing character of the historical
present in narrative is also discussed by Fleischman (1990: 35–7).
15
The connective particles δέ and καί were not included since they function as
unmarked sentence connectives in messenger speeches (see also Rijksbaron 2006: 5). To
calculate the average numbers per line, the absolute numbers of occurrences have been
divided by the total number of lines of each of the four modes in Euripides’ messenger
speeches. The total numbers are: displaced diegetic: 752 lines (43 %); immediate diegetic:
172 (10 %); descriptive: 360 (21 %); discursive: 472 (27 %). The total number of lines of
all messenger speeches is 1756. In order to avoid possible circular argumentation and to
minimize subjective considerations in my method of assigning text-segments to one of
the narrative modes, I use tense and aspect as primary distinctive criteria. Text-segments
in which aorist and imperfects alternate will be ascribed to the displaced diegetic mode.
Text-segments marked by the historical present are labelled as immediate diegetic
mode. Text-segments displaying a sequence of imperfects are viewed as instances of
the descriptive mode. Imperfects in the displaced diegetic mode can be distinguished
 176 rutger j. allan

Table 1 (cont.)
Displaced Immediate Descriptive Discursive Total
Diegetic Diegetic
γάρ 20 216 12 26 60
δή 7 2 0 5 14
δῆτα 0 0 0 1 1
ἤτοι ‘either’ 1 0 0 1 2
καίτοι 0 0 0 0 0
μέν 42 7 28 18 95
μήν 1 0 0 0 1
μέντοι 3 0 0 1 4
νυν 1 0 0 1 2
οὖν 3 0 1 7 11
περ 4 1 0 9 14
που 1 0 0 1 2
τοι 0 0 0 1 1
τοίνυν 0 0 0 0 0
109 (0.14 p.l.) 14 (0.08 p.l.) 54 (0.15 p.l.) 95 (0.20 p.l.) 272

The table also shows the numbers relating to the two other modes
which I will discuss at a later point. Here it can be seen that the imme-
diate diegetic mode uses particles more sparingly than the displaced
diegetic mode (0.08 particles per line, against 0.14). Interestingly, the
frequency of particles in the four narrative modes is neatly paralleled
by the occurrence of negations.16

Table 2: Frequency of negations in the narrative modes

Displaced Immediate Descriptive Discursive Total
Diegetic Diegetic
οὐ, οὐδέ, οὔτε 63 8 33 68 172
μή, μηδέ, μήτε 6 1 2 13 22
TOTAL 69 (= 0.09 p.l.) 9 (= 0.05 p.l.) 35 (= 0.10 p.l.) 81 (= 0.17 p.l.) 194

from imperfects in the descriptive mode primarily by the fact that the latter shows a
hierarchical Topic structure (see section 2.3). Finally, text-segments characterized by
an alternation of actual presents, present perfects, futures and constative aorists are
identified as belonging to the discursive mode (see section 2.4). In section (4), I will
apply these criteria to the messenger speech of Euripides’ Andromache.
16
It should be noted that γάρ in combination with an historical present is always
of the special narrative-embedding type, never of the regular explanatory type (Rijks-
baron 2006: 6).
 towards a typology of the narrative modes 177

In table (2), we can see that the immediate diegetic mode has the
lowest number of negations (0.05 per line), followed by the displaced
diegetic mode (0.09 per line). The higher number of negations in the
displaced diegetic mode shows that the narrator is more in control of
the narration. The narrator does not simply register what is happening,
but because (s)he is narrating from a retrospective stance (s)he is also
able to contradict possible unjustified expectations of the narratee.17
Another linguistic property of the immediate diegetic mode is its
strong connection with the presentation of direct speech. This can be
observed in table (3).
Verbs of speech in the historical present virtually always introduce
direct speech.18 This link between immediate mode and direct speech
can, in my view, be explained by their conceptual relatedness. Both
linguistic phenomena hinge on the aspect of immediacy (cf. Rijksbaron’s
‘pseudo-eyewitness’) and the absence of a narrator as a perceivable
intermediary in the representation of the narrated world.19
Other linguistic properties of the historical present can also be
accounted for by means of the notion of immediacy. Rijksbaron (2006)
mentions three strong linguistic tendencies regarding the historical
present: (1) the rarity of passive historical presents; (2) the rarity of
negated historical presents, and (3) the rarity of historical presents in
subordinated clauses. Passives, negations and subordinate clauses are
linguistically marked expressions, i.e. deviations of the unmarked active,

Table 3: Tenses of speech-introducing verbs in Euripides’ messenger speeches

Direct Speech Indirect Speech
Historical Present 16 1
Aorist 48 15
Imperfect 9 11
Present Participle 7 2
Aorist Participle 2 1

17
Negations in narratives tend to be used by the narrator to contradict an expecta-
tion of the narratee (see De Jong 1987: 61–8).
18
The sole exception is Ion 1191–3.
19
For the conceptual relationship between the historical present and direct speech,
see also Chafe (1994: 218). Bonheim (1982: 52) describes this relation as follows: ‘For
direct speech suggests the closest possible nexus between character and reader, as the
term direct suggests. Indirect and reported speech, by contrast, blur the impression and
distance us from the character.’
178 rutger j. allan

positive main clause type. These marked ways of expression are, in my

view, not well compatible with the use of a historical present because
the use of linguistically marked expressions would reveal the narrator’s
role as a mediating instance in the narration, whereas the historical
present conceals the narrator’s role.
To illustrate the difference between the displaced diegetic mode
and the immediate diegetic mode, I would like to give two examples
from Euripides’ messenger speeches. A run-of-the-mill example of the
displaced diegetic mode is the following:
(2) Displaced diegetic mode:
ἔνθεν τις ἠχὼ χθόνιος ὡς βροντὴ Διὸς
βαρὺν βρόμον μεθῆκε, φρικώδη κλύειν·
ὀρθὸν δὲ κρᾶτ’ ἔστησαν οὖς τ’ ἐς οὐρανὸν
ἵπποι· παρ’ ἡμῖν δ’ ᾖν φόβος νεανικὸς
πόθεν ποτ’ εἴη φθόγγος. ἐς δ’ ἁλιρρόθους (1205)
ἀκτὰς ἀποβλέψαντες ἱερὸν εἴδομεν
κῦμ’ οὐρανῷ στηρίζον, ὥστ’ ἀφῃρέθη
Σκίρωνος ἀκτὰς ὄμμα τοὐμὸν εἰσορᾶν·
ἔκρυπτε δ’ Ἰσθμὸν καὶ πέτραν Ἀσκληπιοῦ.
There a great noise in the earth, like Zeus’ thunder, roared heavily-it made
one shudder to hear it. The horses pricked up their heads and ears to
heaven, while we servants were taken with a violent fear at the thought
where this voice came from. When we turned our eyes to the sea-beaten
beach, we saw a wave, immense and uncanny, set fast in the sky, so great
that my eye was robbed of the sight of Sciron’s coast, and the Isthmus
and Asclepius’ cliff were hid from view.20 (Hipp. 1201–1208)
In this text-segment, we find an alternation of the past aorist and
the imperfect—the former marking the main story line; the latter
(backgrounded) information forming a framework within which the
subsequent events occur.21
An example of the immediate diegetic mode is the following:
(3) Immediate diegetic mode:
δεύτερον δὲ παῖδ’ ἑλὼν
χωρεῖ τρίτον θῦμ’ ὡς ἐπισφάξων δυοῖν.
ἀλλὰ φθάνει νιν ἡ τάλαιν’ ἔσω δόμων
μήτηρ ὑπεκλαβοῦσα, καὶ κλῄει πύλας.

20
The translations in this paper are taken from Kovacs’ Loeb-edition.
21
In a number of examples in this paper, imperfect verbs are indicated in bold type.
Aorists are indicated in italics, and historical presents in bold italic type.
 towards a typology of the narrative modes 179

ὁ δ’ ὡς ἐπ’ αὐτοῖς δὴ Κυκλωπίοισιν ὢν

σκάπτει μοχλεύει θύρετρα (. . .).
Having killed his second son, he went off to sacrifice a third victim on
top of the other two. But before he could do so the boy’s mother snatched
him up, took him inside the chamber, and barred the door. Heracles, just
as if he were besieging Mycenae, dug under the door, pried it up, pulled
out the doorpost (. . .). (HF 995–9)
At this climactic moment in the story, the historical present serves to
enhance the feeling of immediate involvement with the horrible events.
The events referred to by the historical present all belong equally to the
foreground of the narrative. Heracles rushes off to kill his third child
and a rapid series of actions ensues (a scene in the sense of Bal 19972).
Even a door cannot stop Heracles’ mad fury. Note that the speed of
the actions is iconically expressed by the asyndetic juxtaposition of
σκάπτει μοχλεύει.

2.2. Descriptive mode

In the descriptive mode time is static. The text progresses spatially
through a scene, person or object (Smith 2003: 28–9).22 Descriptions
often begin with an explicit denomination of the theme that will be
described (Discourse Topic) and a phrase indicating the over-all location
of the described scene or object. The theme and the location function as
a framework, an ‘Erwartungshorizont’, so to speak, for the subsequent
description. The description itself consists of a series of parts of the
described entity (Sub-Topics).23 Of these Sub-Topics either a state (for
example, a property) is described, or an ongoing activity. The typical
tense used in descriptions is the imperfect, which either designates
states or ongoing activities (often iterative events).24 The subsequent

22
Note that descriptions are frequently internally focalized (Bal 1997: 36). A typi-
cal example in Homer is Od. 7.81–135 (description of Alcinous’ palace and garden,
where Odysseus is the focalizer). An example from our corpus is the description of the
shields of the Seven in Ph. 1104–38, which is internally focalized in an explicit manner
(εἰσορῶμεν, line 1099; παρῆν (. . .) μοι θέαματα, line 1139).
23
For the terms Discourse Topic, Sub-Topic, and New Topic, see Dik (1997b:
313–25). Descriptions typically display a tree-like, hierarchical organisation in that the
description of a Sub-Topic, in turn, may be divided in Sub-Sub-Topics, and so on.
24
The imperfect implies that the temporal point of reference is located within the
boundaries of the event. It represents the state of affairs from the inside, while it is
taking place. This internal point of view often gives the imperfect a certain visual quality
(‘a scene is painted’, Rijksbaron 2002: 12). This aspect of the imperfect also lies at the
180 rutger j. allan

changes in space during the progression of the description are mostly

inferred from lexical information. The frequency of particles (0.15 per
line) and negations (0.10) in the descriptive mode is approximately the
same as in the displaced diegetic mode (see table 1).
An example of an object description is the following ekphrasis from
the Ion (1146–58).
(4) ἐνῆν δ’ ὑφανταὶ γράμμασιν τοιαίδ’ ὑφαί,
Οὐρανὸς ἀθροίζων ἄστρ’ ἐν αἰθέρος κύκλῳ.
ἵππους μὲν ἤλαυν’ ἐς τελευταίαν φλόγα
Ἥλιος, ἐφέλκων λαμπρὸν Ἑσπέρου φάος·
μελάμπεπλος δὲ Νὺξ ἀσείρωτον ζυγοῖς (1150)
ὄχημ’ ἔπαλλεν, ἄστρα δ’ ὡμάρτει θεᾷ·
Πλειὰς μὲν ᾔει μεσοπόρου δι’ αἰθέρος
ὅ τε ξιφήρης Ὠρίων, ὕπερθε δὲ
ἌΡΚΤΟΣ στρέφουσ’ οὐραῖα χρυσήρη πόλῳ·
κύκλος δὲ πανσέληνος ἠκόντιζ’ ἄνω (1155)
μηνὸς διχήρης, Ὑάδες τε, ναυτίλοις
σαφέστατον σημεῖον, ἥ τε φωσφόρος
Ἕως διώκουσ’ ἄστρα.
On them was woven the following. Heaven was mustering the stars in the
circle of the sky. Helios was driving his horses toward his final gleaning,
bringing on the brightness of the Eveningstar. Night, robed in black, was
making her chariot, drawn by a pair with no trace horses, swing forward,
and the stars were accompanying the goddess. The Pleiades were pass-
ing through mid heaven and so was Orion with his sword, while above
them the Bear turned its golden tail about the Pole. The circle of the full
moon, as at mid month, darted her beams, and there were the Hyades,
clearest sign for sailors, and Dawn the Daybringer putting the stars to
flight. (Ion 1146–58)
The description is introduced by a verb-initial presentational clause,
containing an indication of the location of the described object (ἐνῆν)
on the covering of garments of the tent, and the new Discourse Topic
of the description (ὑφαί). Next, a series of Sub-Topics is introduced in
subject position (Οὐρανός, Ἥλιος, etc.). Of these Sub-Topics, an activity
is described, mostly by means of an imperfect verb and sometimes by
a participle. The spatial movement through the scene depicted on the
garment is mainly implied by the nature of the different subjects (such

heart of what Bakker (1997b) calls the mimetic mode. Note further that if the descrip-
tion refers to generic or habitual states or activities which (also) hold at the time of
narrating, the present tense can also be used.
 towards a typology of the narrative modes 181

as their usual position in the sky) and their specific activities. Explicit
indications of (relative) locations are ὡμάρτει (1151), μεσοπόρου δι’
αἰθέρος (1152), and ὕπερθε (1153). As an ekphrasis, this description
is, of course, somewhat special. Although it describes an unmoving
physical object—the garment—, the represented items on the garment
are portrayed as if they are in motion.
Most descriptions in messenger speeches relate to scenes. In the
beginning of the messenger speech in the Heracles, for example, we
find a description of the initial situation which sets the scene for the
subsequent fatal events.25
(5) ἱερὰ μὲν ἦν πάροιθεν ἐσχάρας Διὸς
καθάρσι’ οἴκων, γῆς ἄνακτ’ ἐπεὶ κτανὼν
ἐξέβαλε τῶνδε δωμάτων Ἡρακλέης·
χορὸς δὲ καλλίμορφος εἱστήκει τέκνων (925)
πατήρ τε Μεγάρα τ’· ἐν κύκλῳ δ’ ἤδη κανοῦν
εἵλικτο βωμοῦ, φθέγμα δ’ ὅστον εἴχομεν.
Sacrificial victims were standing before Zeus’ altar to purify the house,
since Heracles had killed the land’s ruler and flung him out of doors. His
children stood by as a lovely chorus, and his father and Megara too, and
the sacred basket had made its circular course about the altar: we were
all keeping reverent silence. (HF 922–7)
Unlike people and objects, scenes obviously do not have straightforward
names. As a result, this description—like most scene descriptions—lacks
an explicit indication of the descriptions’ theme, that is, the Discourse
Topic. The text leads us through the space of the scene by describing
every person and object (Sub-Topics) present at the purificatory sacri-
fice, starting—highly significantly—with the victims. The location of the
scene is specified by the phrase πάροιθεν ἐσχάρας Διός. All main clauses
describe simultaneous states (ἦν, εἱστήκει, εἵλικτο, εἴχομεν). There is no
temporal, only spatial progression.

2.3. Discursive mode

In the discursive mode, the states of affairs referred to by the speaker
are directly related to the communicative situation (Discourse-Now).26

25
Other examples of descriptions are: Ba. 677–88, 722–9, 739–47, 751–64, 1043–57,
1095–1102, 1129–36; El. 798–902; Med. 1179–89, 1197–1203, 1211–7.
26
My notion of discursive mode is related to Benveniste’s notion of discours. Benveniste
contrasts discours (of which the basic tenses are présent, futur, parfait and imparfait
182 rutger j. allan

The speaker and hearer are directly linked to, and concerned with,
what is described. Often, the communicative function of the discursive
mode is to influence the hearer in a certain way.27 Ways of influenc-
ing are to try to convince the addressee that a certain state of affairs
is the case, or to persuade the addressee to act in a certain manner. A
typical way of influencing the addressee is, of course, by appealing to
his or her emotions.28 However, the aim to influence the beliefs and
behaviour of the hearer does not always have to be predominant. The
speaker’s aim may also be primarily informative about the present state
of affairs (Discourse-Now), including information about the existence
of intentions.
The discursive mode appears in two narrative situations. On the
one hand, the narrator can directly address the narratee through nar-
ratorial intrusion, for example, by commenting on the narrated event
or characters, the development of the action, or the circumstances of
the narration (Bonheim 1982: 30–2). On the other hand, the discur-
sive mode appears when characters address one another by means of
direct speech, relating to their own Discourse-Now. From a narrato-
logical point of view, narratorial discourse and character discourse are,
of course, distinct phenomena because they relate to different levels of
narration.29 In the former case, we are dealing with the discourse of
an external narrator; in the latter, with character-bound discourse (see
Bal 19972). From a purely linguistic point of view, however, there is
no reason for making a distinction between the two types. Consider,
for example, the two following utterances: (1) τοιόνδε τοι στρατηγὸν

in French) with histoire, which is marked by the aoriste (passé simple), imparfait, plus-
que-parfait. According to Benveniste, histoire characterizes the narration (récit) of past
events (Benveniste 1966: 238–9). Benveniste’s histoire may be roughly compared to my
diegetic mode. Weinrich (2001) makes a similar distinction between Besprochene Welt
(Präsens, Perfekt, Futur), and Erzählte Welt (Präteritum, Plusquamperfekt).
27
Benveniste describes his notion of discours as follows: ‘Il faut entendre discours
dans sa plus large extension: toute énonciation supposant un locuteur et un auditeur,
et chez le premier l’intention d’influencer l’autre en quelque manière’ (Benveniste
1966: 241–2).
28
Some scholars distinguish an argumentative discourse type (Chatman 1978, 1990,
Smith 2003, Adam 2005). In my view, however, this term suggests too much that persua-
sion is brought about only (or typically) by means of rational (deductive or inductive)
argument. Argumentation is only one aspect of the discursive mode.
29
It should be noted that the narrator is not to be confused with the author. The
narrator is a fictitious text-internal (speaking or narrating) agent and, therefore, in this
respect not different from a character. See for the crucial distinction between narrator
and author, Genette (1972: 203–6), De Jong (1987: 29–40), (Bal 1997: 16).
 towards a typology of the narrative modes 183

αἱρεῖσθαι χρεών (E. Supp. 726) ‘Such a general one should choose’; (2)
Ἀφαρπάζειν χρεὼν/ οἰνηρὰ τεύχη σμικρά (E. Ion 1178–9) ‘We should
take away the small wine vessels’. Both are examples of the discursive
mode. They have the illocutionary force of a directive, uttered by the
speaker to recommend a certain action to the addressee (and thereby
to influence the addressee). The fact that the former is spoken by the
narrator to the narratee (a messenger to Adrastus and the chorus), while
the latter is spoken by a character to a character (an old man to the
servants), is of no relevance to the linguistic text type of the utterances.
The difference in narrative situation is analogous to the distinction
between the story told by the primary narrator and a story told by a
character. Narratologically, these narratives belong to different levels of
narration since the latter is embedded in the former. From a linguistic
point of view, however, it is unnecessary to make a distinction since,
after all, both are instances of the narrative text type.30
Although the discursive mode is a common ingredient of narratives,
it is clearly non-narrative in character. Likewise, the linguistic features
of the discursive mode diverge strongly from the modes discussed so
far. With regard to tense, (actual) present tense forms are typically
used.31 However, constative aorists and future forms can also occur. The
constative aorist is used in direct speech, and usually indicates that the
event is completed relative to the Discourse-Now, rather than relative
to another event in the discourse—as is the case in narrative.32 In other
words, this means that in the discursive mode time does not progress.
Future tenses are often used with special illocutionary forces, such as

30
Another issue which may come up is the relation between character speech and
the immediate mode. I have indicated earlier that there is a relation between the use of
the historical present and the insertion of direct speech. Indeed, from the point of view
of the primary narrator both involve the pretence of immediate presence at the scene.
With regard to character speech, however, the point of view of the primary narrator
is obviously of no relevance to the internal structure of the text (such as regarding its
text type) produced by a character. Instead, it is the relation between the actual speaker
(the character) and his/her discourse world that counts.
31
Examples from the Andromache (see Appendix) are: ἔστι (1073), ἔχει (1174),
ὁρᾶτε, διαστείχει (1092), βουλόμεσθ’ (1107), κτείνετ’ (1125), ὄλλυμαι (1126), κομίζομεν
(1159).
32
For the constative aorist, see Rijksbaron (2002: 28–9). An example is: αἷμα δ’
αἵματος/ πικρὸς δανεισμὸς ἦλθε τῷ θανόντι νῦν (El. 858) (‘For the one who has died
bloodshed has come as the bitter return for bloodshed’). Note the combination with
νῦν, which stresses the present relevance of the event. See also Ph. 1999. Examples of
the constative aorist in the Andromache (see Appendix) are: ἦλθε (1095), ᾔτησα (1107),
ἔβαλ’ (1157), ἔδρασ’ (1159), ἐμνημόνευσε (1164).
184 rutger j. allan

expressing an announcement, a promise, a command, or a threat (see

also Rijksbaron 20023: 33).33
As for particle use in the discursive mode, let us return to the figures
in table 1. As we have seen above, the immediate diegetic mode uses
particles more sparingly than the displaced diegetic and the descriptive
mode. The discursive mode, ultimately, displays the most abundant
use of particles (0.20 per line). This high frequency is due to the fact
that the discursive mode is characterized by the strongest interaction
between narrator and narratee, or, to put it in other terms, the discursive
mode is the most diaphonic of all modes.34 Typical particles found in
the discursive mode are: ἀλλά, ἄν, γάρ, γε, δή, μέντοι, οὖν, περ.
In the same vein, in table 2 it was shown that the descriptive mode
shows approximately the same frequency of negations as the displaced
diegetic mode. In turn, the discursive mode features the highest number
of negations (0.17 per line). This relatively high number of negations
in the discursive mode is connected to the fact that negations function
on the level of narrator-narratee interaction.35 Negations are typically
used by speakers to contradict an expectation of the hearer. Now, if we
compare the frequency numbers of particles and negations in all four
modes, we can locate them on a ‘scale of narratorial mediation’.
(6) Scale of narratorial mediation
Discursive mode

Descriptive mode, Displaced diegetic mode

Immediate diegetic mode

The ‘scale of narratorial mediation’ indicates the degree to which the
narrator is overtly present as a mediating voice in the narration (Bon-

33
In the discursive mode, the imperfect does not occur very often. A special use of
the imperfect in the discursive mode concerns situations in which the speaker draws
the conclusion with hindsight that some state of affairs had been different from what
one thought at the moment (often combined with ἄρα), e.g. Andr. 1088 (καὶ τοῦθ’
ὕποπτον ἦν ἄρ’). Examples of the same use of the imperfect in dialogue (i.e. discursive
mode) can be found in Hipp. 1169 and IT 1316.
34
The term diaphony has been introduced by Roulet and refers to the idea that a
speaker/narrator of a monological discourse may also represent—be it embedded—the
‘voice’ of the addressee/narratee (see Roulet, Fillietaz & Grobet 2001: 286). See also
Kroon (1995: 111–6).
35
Or, in terms of polyphonic organisation, a negation can be seen as introducing a
polyphonic structure in the sense that the speaker anticipates the reaction of the hearer
(see e.g. Roulet, Fillietaz & Grobet 2001: 277).
 towards a typology of the narrative modes 185

heim 1982: 40). Thus, we can say that the more particles and negations
present in a particular mode, the more prominent the mediating voice
of the narrator in the narration. The direct communication between
speaker and hearer (narrator and narratee) in the discursive mode is
manifested concretely by an increased use of explicit references to the
second person by means of vocatives, second person pronouns and
verbs.36 For example, in the messenger speech of the Andromache there
are 10 cases of references to the second person, all of which occur in
the discursive mode.37
In order to influence the hearer, finally, a speaker may exploit various
types of speech acts such as assertions, directives, exclamations, (rhetori-
cal) questions, and wishes. In the discursive mode, therefore, all moods
are employed: indicative, optative, imperative and subjunctive.38
I conclude section (2) with a synoptic table which shows a number
of typical linguistic and narratological features of the four narrative
modes. It is important to note, however, that the four modes as I have
described them above should be thought of as prototype-categories
(also see Adam 2005: 18), that is to say, narrative modes are gradual
notions.39 Actual instances of the narrative modes can be more or less
similar to the category prototype, depending on the number of prop-
erties which they have in common with the prototype. This implies
that a given sentence can be ambiguous as to which narrative mode it
displays since it is not excluded that a sentence possesses properties of
different narrative modes.40

36
The presence of 2nd person pronouns is also mentioned by Kroon (1995: 114)
as a feature of diaphony.
37
These are: σοι (1073), γέρον Πελευ (1073–4), ὁρᾶτε (1092), νεανία (1104), σοι (1104),
ἥκεις (1105). κτείνετ’ (1125), εἶδες (1135), σοί (1159), πρέσβυ (1160). De Jong (1990:
195–7) mentions 86 instances of second person pronouns and verbs and vocatives. In
this number, however, the cases occurring in character speeches are not included. Of
these 86 cases, 41 cases occur in the discursive mode. These are Andr. 1135, 1159, 1160;
Ba. 686, 712–3, 737, 740, 760, 769, 770, 1085; El. 855, 857; Hec. 518, 519, 580–1, 58;
Hel. 1616; Heracl. 832, 853, 856; Hipp. 1249, 1251; IT 336, 337, 338, 1417; Med. 1144,
1222, 1223; Or. 951, 953, 954, 955; Ph. 1150, 1219, 1259, 1260 (bis), 1262, 1357.
38
Examples from the Andromache (see Appendix) are 1092–5, 1104–5, 1105, 1125–6,
1126, 1165 (questions), 1105 (subjunctive), 1135 (counterfactual past indicative).
39
In Allan (2003), I have argued that also a grammatical category such as the Greek
middle voice can be described more adequately by means of the linguistic prototype
model.
40
A clear example of a mixture of two modes is indirect discourse, which can be
seen as an embedding of the discursive mode in the diegetic mode. For more examples
of borderline cases, see Bonheim (1982: 33–4).
 186 rutger j. allan

Table 4: Overview of typical linguistic and narratological features

of the narrative modes
Displaced Immediate Descriptive Discursive
Diegetic Diegetic
Tense and aorist, imperfect historical present imperfect present, perfect,
Aspect future

Particles presentational: presentational: presentational: all types:

ἀλλά, αὖ, γάρ, μέν μέν (. . .) δέ ἀλλά, γάρ, μέν ἀλλά, ἄν, γάρ, γε,
(. . .) δέ (. . .) δέ δή, οὖν, περ

Mood indicative indicative indicative all moods

Narrator’s from discourse- from story-now from discourse- from discourse-

Point of View now now now

Progression advancement in advancement in spatial advancement

narrative time narrative time advancement anchored to
through the discourse-now
scene or object

Rhythm summary, ellipse scene pause pause

Speech Acts assertions assertions assertions assertions,

directives,
exclamations,
questions, wishes

3. Narrative structure

In this section, I would like to demonstrate how the narrative modes

are linked to narrative structure. The various narrative modes tend to
appear in specific parts of a narrative. This relationship between narra-
tive mode and narrative structure will be demonstrated by an analysis
of the messenger speech in the Andromache. Before starting off with an
examination of the messenger speech, I will set out a structural model to
describe the organization of narratives, building on the work of Labov
(1972: 362–70) and Fleischman (1990: 135–54).41

41
I have applied this model before in an analysis of the narrative modes in Thucy-
dides (Allan 2007).
 towards a typology of the narrative modes 187

(7) The Global Structure of Narrative

a. Abstract: Point of story or summary of significant events
b. Orientation: Identification of the time, place, circumstances
and participants
c. Complication: Build-up of tension
d. Peak: Climax, decisive moment
e. Resolution: Outcome/result
f. Coda: Closure, bridge to time of narrating
g. Evaluation: Narrator’s comment
Narratives frequently start off with an Abstract which conveys the point
of the story (why is the story relevant to the hearer?), or by expound-
ing the story’s most significant events. This plot summary is usually
marked by a present tense which refers to the time of narrating, i.e. the
Discourse-Now (Fleischman 1990: 138). The use of the present tense
makes clear that the Abstract does not belong to the actual narrative,
but constitutes the ‘outer frame’ of the narration; grounding the narra-
tive in the social-communicative situation; and building a bridge from
the world of the narrator and narratee to the story-world.42 After the
Abstract, at the outset of the narrative proper, an Orientation section
may follow, in which the time, place, circumstances and participants
are presented. Although the natural position of the Orientation is at the
head of the narrative, Orientation sections may also occur embedded
throughout the text, in the Complication, the Peak, the Evaluation, and
also (less frequently) in the Resolution.
In the Complication, the stability of the initial situation is somehow
disturbed43 and a possibility of improvement is opened (Bremond 1973).
This triggers a ‘rising’ sequence of actions as a result of which tension
gradually builds up, eventually leading to the Peak, that is, the climax
of the story. Typically, the Peak coincides with the decisive moment
in the course of events. Using Bremond’s terminology, we may see
the Peak as an actualisation (or, non-actualisation) of the possibility
opened in the Complication. According to Longacre (1996: 38), the
Peak is a ‘zone of turbulence’ when compared to the other parts of the
narrative. Special narratological and linguistic properties of Peaks are

42
For the hierarchical integration of narrative in dialogical discourse, I refer to
Roulet (1989).
43
Compare Tomachevski’s definition of a ‘noeud’: ‘Pour mettre en route la fable,
on introduit des motifs dynamique qui détruisent l’equilibre de la situation initiale.
L’ensemble des motifs qui violent l’immobilité de la situation initiale et qui entame
l’action s’appelle le noeud’ (Tomachevski 2001: 278).
188 rutger j. allan

a concentration of participants (‘crowded stage’); a shift to dialogue or

direct speech; a change of pace (‘a slowing down of the camera’); the
absence of particles marking the storyline elsewhere; and a switch to
the historical present tense (Longacre 1996: 38–50).
After the Peak, the story attains its Resolution in which the outcome
of the story is told. The Resolution eventually leads to a restoration of
the equilibrium.44 Additionally, the story-telling may end with a Coda.
The narrated events usually do not extend to the narrator’s present
(Discourse-Now), but, by means of a Coda, the narrator can bridge the
gap between the end of the story and Discourse-Now. In other words,
it is a device for returning the verbal perspective from the time-frame
of the story (Story-Now) to the communicative situation of the nar-
rator and narratee (Discourse-Now) (see also Fleischman 1990: 138).
In Evaluation sections, the narrator comments on the content of the
story and its significance. Evaluative elements are usually interspersed
throughout the story.45 Evaluations which contain a comment on the
significance of the story as a whole (the ‘morale’) will obviously tend
to appear at the end of the story (or, rather, after the end).46
Stories typically consist of multiple episodes,47 that is, they tend to
show a recursive structure of Complications, Peaks and Resolutions,
thereby providing a profiled pattern of build-ups and relaxations of
tension. This pattern is represented schematically in (8):

44
The basic ‘triad’ Complication—Peak—Resolution can be seen as the ‘nucleus’ (plot)
of a narrative. However, even each of these three structural elements is dispensable (Bal
1997: 189). For example, in the case of an anti-climax, the Peak is absent.
45
It is important to note that subjective-evaluative elements may also be part of nar-
rative clauses. Because of their embedded character, however, these cannot be viewed as
separate Evaluation sections. For example, the clause in Andr. 1140–1 (see Appendix)
primarily refers to a narrative event (the flight of the Delphians), but it also contains
an evaluative element (the comparison of the Delphians with doves).
46
It is also possible to assign a hierarchical structure to the various elements of a
narrative (cf. Roulet 1989, Roulet, Fillietaz & Grobet 2001). For example, the elements
(a.) to (f.) in (7) can be seen as constituting together a subsidiary argumentative move
supporting the Evaluation, which can be interpreted as the principle move (the ‘point’
of the story). The Orientation can be thought of as subsidiary to the move which con-
sists of the triad Complication-Peak-Resolution. For the concept move (intervention
in French), see Roulet (1989), Kroon (1995) and Roulet, Fillietaz & Grobet (2001). A
detailed analysis of these hierarchical relations between narrative units is beyond the
scope of this paper.
47
Van Dijk defines episodes as ‘coherent sequences of sentences of a discourse,
linguistically marked for beginning and/or end, and further defined in terms of some
kind of “thematic unity”—for instance, in terms of identical participants, time, location
or global event or action’ (Van Dijk 1982: 41).
 towards a typology of the narrative modes 189

(8) Episodic Structure of Narrative:48

• Abstract
• Orientation
• [Complication—Peak—Resolution]Episode 1
• [Complication—Peak—Resolution]Episode 2
• (. . .)
• [Complication—Peak—Resolution]Episode n
• Coda
• Evaluation
On this schema, however, considerable variations are possible. Com-
mon deviations of the schema are the following: firstly, each of these
structural elements can be dispensed with. Especially the Abstract, Ori-
entation, Coda and Evaluation, (i.e. the elements outside the ‘nucleus’
of the story) may be absent. Secondly, separate episodes may have their
own local Abstract, Orientation, Coda or Evaluation. Thirdly, in some
cases episodes are enchained ‘bout à bout’ (Bremond 1973). Then, the
Resolution of Episode A is, at the same time, the Complication of the
subsequent Episode B.

4. Analysis of the messenger speech in the Andromache

In this section, an analysis will be made of the messenger speech in

the Andromache. The main objective will be to show the relationship
between the occurrence of the various narrative modes and the struc-
ture of the narrative as it has been set forth above. It will become clear
that the use of tenses and aspects will be of crucial importance to this
type of analysis.
In virtually all messenger speeches, the point of the narrative is sum-
marized by means of an Abstract. In the Andromache, the Abstract
is spoken by the messenger, a servant of Neoptolemus, to Peleus, the
intended narratee in lines 1073–5: οὐκ ἔστι σοι παῖς παιδός, ὡς μάθῃς,
γέρον/ Πηλεῦ· τοιάσδε φασγάνων πληγὰς ἔχει/ Δελφῶν ὑπ’ ἀνδρῶν καὶ
Μυκηναίου ξένου (see also the Appendix). The relatively detached
relation of Abstracts to the actual narrative is clear from its embed-
ded position in the dialogue preceding the story. The function of the
Abstract is to make clear why the story is here-and-now relevant to the

48
This episodic narrative schema is similar to the schema presented by Fludernik
(1996: 65).
190 rutger j. allan

narratee. It refers, therefore, to the social and communicative situation

of the narrator and narratee (Discourse-Now). This means that Abstracts
are verbalized in the discursive mode. Typical tenses of Abstracts are
presents,49 perfects,50 and constative aorists,51 that is, tenses that bear a
direct relationship with the moment of speaking.
The story of the messenger starts in line 1085. According to our
schema in (8), we would expect the narrative to start with an Orienta-
tion. However, like many other messenger speeches, the Andromache
does not contain an Orientation section. As Rijksbaron (1976) has
demonstrated, this type of messenger speech opens with an ἐπεί-clause
which clarifies (in a minimal way) the moment at which the narrated
events begin, by referring back to earlier-given information. In these
messenger speeches, there is no need for an Orientation section because
the addressee is already acquainted with the place, time and general
circumstances of the events. In the Andromache, Peleus already knows
of Neoptolemus’ journey to Delphi, as appears in his remark to Andro-
mache in l. 558: ὕπαρνος γάρ τις οἶς ἀπόλλυσαι,/ ἡμῶν ἀπόντων τοῦ τε
κυρίου σέθεν (‘For you are being put to death like some ewe with her
lamb while I and your master are away’).
An example of a messenger which does include an Orientation is
the Heracles printed in example (5). In lines 922–7, the Orientation
is described, which sets the scene to the upcoming events. The main
characters of the story are introduced—Heracles’ children, his father
and Megara—, and the specific location within the house is indicated:
πάροιθεν ἐσχάρας Διὸς. The presence of the victims makes clear that they
are about to make a sacrifice. There is no need for the messenger to
indicate the specific moment of time at which the event took place. After

49
Presents: Andr. 1073–5 (ἔστι, ἔχει), Ba. 667 (δρῶσι), Hipp. 1162 (ἔστιν), Heracl.
786 (νικῶμεν), IT 1315 (οἴχεται), Ph. 1339 (εἰσί). In several cases the Abstract is given in
indirect discourse, dependent on a verb of telling (or hearing), e.g., Alc. 157 (θαυμάσῃ
κλύων), Ba. 666–7 (ἥκω φράσαι), El. 762–4 (ἀγγέλλω), IT 239 (ἄκουε), Supp. 634–8 (ἥκω
(. . .) ἀγγελῶν). In the Hec., the actual Abstract—if there is one—is given indirectly by
σὴν παῖδα κατθανοῦσαν within a final clause: σὴν παῖδα κατθανοῦσαν ὡς θάψῃς, γύναι,/
ἥκω μεταστείχων σε (Hec. 508–9).
50
Perfects: Med. 1125 (ὄλωλεν), Ba. 1030 (ὄλωλεν), El. 770 (τέθνηκε), Hel. 1515,
1517, 1522 (βέβηκ’, ἐκπεπόρθμευται, βέβηκεν), HF 913 (τεθνᾶσι), Hipp.1163 (δέδορκε),
Ph. 1079 (ἑστᾶσ’, ἀνήρπασται), Ph. 1349 (τέθνηκ’).
51
Constative aorists: Ion 1117 (ἐξηῦρεν), Or. 858 (ἔδοξε). The messenger often tells
the story in reply to a question (of the type: ‘How did he/she/they die?’). Tenses used
in this question are also presents, perfects and constative aorists (i.e. discursive mode),
e.g. Andr. 1083 (οἴχεται), Ba. 1041 (θνῄσκει), Hipp. 1171 (διώλετ’), Med. 1134 (ὤλοντο),
Ph. 1353 (πέπρακται).
 towards a typology of the narrative modes 191

all, the Chorus were standing outside the house during the dreadful
events, hearing the cries within. As we have seen above, this Orienta-
tion section is in the descriptive mode (which is, of course, typical of
Orientations).52 I will now return to the Andromache.
The messenger speech of the Andromache consists of three episodes.
In the first episode, the failed ambush in the temple is narrated. The
second episode relates how Neoptolemus strikes back. The third, and
last, episode tells us how Neoptolemus is slain after Apollo’s interven-
tion. In line 1085, the Complication of the story’s first episode begins.
In the Complication section, the tension gradually builds up. Having
arrived in Delphi, Neoptolemus and his companions play the tourist
and in doing so arouse suspicion. There are spontaneous gatherings,
and Orestes sets the Delphians against Neoptolemus by means of a
slanderous accusation. The Delphians are alarmed and take precautions.
Neoptolemus and his companions, unaware of the scheme, approach
the altar. Neoptolemus states that he has come to make amends for his
earlier sin. The story line of the Complication (1085–1111) is given by
aorists, imperfects, and a pluperfect in the main clauses.53 The aorists
refer to sequential events, the imperfects and pluperfect refer to events
that create a framework for the following events. In other words, the
events are narrated in the displaced diegetic mode. However, we also
find parts of the narrative in the discursive mode: character speech is
represented three times (1092–5, 1104–5, 1106–8).54

52
Examples of messenger speeches with Orientations are: Ba. 677–88, 1043–57, El.
774–8, Hec. 521–2, Hipp. 1173–7, IT 260–4, Or. 860–4, Supp. 650–1. Orientations are
often introduced by μέν indicating that the Orientation is preparatory to the actual
narrative. An Orientation may also contain diegetic elements referring to events anterior
to the story (external analepsis). E.g., in the HF an analepsis is found in a subordinate
clause: γῆς ἄνακτ’ ἐπεὶ κτανὼν/ ἐξέβαλε τῶνδε δωμάτων Ἡρακλέης. The killing of Lycus
is referred to in line 754. Also the pluperfect εἵλικτο implies a (in this case, recent)
event in the past. Another example of an orientation with an analepsis is Hipp. 1175–7:
ἦλθε γάρ τις ἄγγελος (. . .) (note the ‘past-in-the-past’-aorist in combination with the
PUSH-particle γάρ).
53
Aorists: ἐτάξαντ’ (1099), εἶπεν (1104), εἶπε (1106); imperfects: ἐξεπίμπλαμεν (1087),
ἦν (1088), ἐχώρει (1099) (iterative?), ηὔδα (1091) (iterative?), ἐχώρει (1095), ᾖμεν (1102),
ἐφαίνεθ’ (1110); pluperfect: ἐφέσταμεν (1102).
54
Line 1088 is an interesting case. The narrator draws, with hindsight, the unex-
pected conclusion (note the presence of ἄρα) that their sight-seeing had been a cause
of suspicion (‘And that, apparently, was suspicious’). This line may well belong to the
discursive mode. A less far-reaching interpretation has been given by De Jong, who
takes the line as an instance of narrating focalization. The messenger is here—as in a
large part of the story in the Andromache—reconstructing events which, at that time,
192 rutger j. allan

The central scene of the first episode, the Peak, begins with Neo-
ptolemus entering (ἔρχεται) the temple-building, the location of the
ambush. Subsequently, a rapid sequence of actions is marked by a
remarkable cluster of 7 historical presents, indicating the immediate
mode. Neoptolemus finds Phoebus’ statue among burnt offerings,55 and
prays to the god. Orestes and his men try to stab him, but he retreats,
draws his sword, jumps on the altar and shouts ‘Why do you want to
kill me?’. The sequence of historical presents is interrupted three times
by a switch to other tenses. Two times the narrative gives background
information (1114–6: pluperfect, imperfect, 1120–1: imperfect + γάρ).
These are embedded Orientations, during which narrated time does
not progress. Once, the tense shifts to an aorist (ἔστη, 1123), which
is less easy to account for. Possibly, the fact that Neoptolemus takes
his stand on the altar is presented as an event which is somewhat less
important to the course of action than that the fact that he draws his
sword (ἐξέλκει, 1121)—which makes clear that he will try to strike back.
The switch back to a historical present βοᾷ is less remarkable. Prob-
ably because of its inherent dramatic character, the verb βοάω shows a
strong predilection for the historical present. In Euripides’ messenger
speeches we find 7 instances of βοᾷ but only 3 aorists (3 x ἀνεβόησε).
In line 1124, the historical present βοᾷ underscores the dramatic quality
of Neoptolemus’ question—which remains unanswered.
At this point, I would like to make a digression on the narrative
function of the historical present in Euripidean messenger speeches. I
distinguish three specific narrative functions of the historical present
in Greek texts. These are (in order of frequency),

were not known or visible to him and Neoptolemus (De Jong 1991: 53). Interpreted
thus, the line is part of the displaced diegetic mode.
55
I owe this interpretation to Albert Rijksbaron (p.c.) who proposes (though hesi-
tantly) to take Neoptolemus as the subject of τυγχάνει, and Apollo as the unexpressed
genitive object. In my view, this interpretation is supported by the fact that there is
no indication of a subject-switch (such as ὁ δέ), and by the fact that, on this reading,
τυγχάνει has a telic meaning, which is compatible with the use of the historical present.
The traditional stative interpretation of the verb ‘He [Apollo] happened to be engaged
in burnt offerings’ (sc. ὤν: see Stevens a.l.) is untenable since the historical present
never occurs with stative verbs. Incidentally, this means that the historical present
σκυθράζει in El. 830 shows that the verb σκυθράζω (of which σκυθράζει in El. 830 is
the only attestation) does not designate a state ‘be angry’ (as LSJ claim), but rather a
telic event ‘become angry’.
 towards a typology of the narrative modes 193

(i) marking a Peak section in an episode,56

(ii) marking a visually or verbally dramatic ‘shot’ in the Complication,
(iii) punctuating the narrative57 by indicating the start (incipit) of a
Complication.58
I would like to emphasize that the use of the historical present never
becomes an automatism on the part of the narrator. The historical
present is, rather, a rhetorical device which can be exploited by the
narrator to present the story subjectively.59 An illustrative passage is
the following from Medea (1160–6):

56
This use can, of course, be compared to previous characterizations of the historical
present as a marker of decisive events (e.g. Rijksbaron 20023: 22), or of the narrator’s
main concern (Sicking & Stork 1997: 156).
57
For the punctuating function of the historical present, I refer to Rijksbaron (2002:
24). The punctuating use can be explained as a secondary extension on the basis of
the prototypical Peak-marking use of the historical present. The notion of ‘important
turn of events’ is transferred from the Peak to the beginning of a new episode in the
narrative (see also Fludernik 1991: 375).
58
All instances of function (i) are (63 x): Alc. 176; Andr. 1111, 1113, 1117, 1119,
1120, 1121, 1124, 1140, 1149, 1152, 1153; Ba. 705, 1112, 1115, 1117; El. 854; Hec. 565;
HF 964, 969 (bis), 975, 976, 979, 986, 988, 995, 996, 997, 999 (bis); Hipp. 1212, 1218,
1221, 1224, 1237, 1246; Ion 1207, 1210, 1213, 1217, 1219; IT 284, 298, 301, 307, 330,
1345; Med. 1169 (bis), 1173, 1190, 1195; Or. 944; Ph. 1154, 1165, 1169, 1181, 1186, 1192,
1415, 1458; Supp. 706. The instances of (ii) are (32 x): Ba. 680 (ὁρῶ), 1063 (ὁρῶ); El. 779
(ἀυτεῖ), 783 (ἐννέπει), 790 (ἐννέπει), 814 (λέγει), 822 (ἀπωθεῖ), 830 (σκυθράζει, ἀνιστορεῖ),
838 (κόπτει); Hec. 528 (αἴρει), 529 (σημαίνει), 574 (πληροῦσιν); Hel. 1596 (βοᾷ); Hipp.
1188 (μάρπτει); HF 956? (σκευάζεται is dubious. See below); Ion 1193 (δίδωσι, λέγει),
1196 (ἐσπίπτει); Ion. 1143 (περιβάλλει); Med. 1161 (σχηματίζεται), 1163 (διέρχεται); Or.
871 (ὁρῶ), 879 (ὁρῶ), 1444 (ἄγει, ἄγει), 1461 (ἐννέπουσι); Ph. 1099 (εἰσορῶμεν), 1410
(ἀμφέρει), 1452 (τίθησι), 1475 (βοηδρημοῦμεν); Supp. 653 (ὁρῶ). The instances of (iii) are
(9 x): Alc. 186 (στείχει); Ba. 748 (χωροῦσι), HF 1001 (ἱππεύει); Med. 1205 (προσπίτνει); Or.
887 (ἀνίσταται), 902 (ἀνίσταται); Ph. 1401 (χωρεῖ), 1429 (προσπίτνει); Supp. 696 (χωρεῖ).
Note that they all involve verbs of motion. I have used the list of historical presents
compiled by De Jong (1991: 185–6) as a starting-point in this inventory. However,
with Rijksbaron (2006) I view the following forms as (unaugmented) imperfects: δεύετο
(instead of -ται) (Alc. 184), κομίζομεν (IT 334), κύνει (Alc. 183, Med. 1141, 1207), κύρει
(Ba. 728, El. 777), χειροῦμεθα (IT 330), ὤθει (IT 1395). Further, Dobree’s conjecture ἔχει
in HF 956 must be rejected given the fact that the historical present never occurs with
stative verbs. Rijksbaron (p.c.) points out to me that the ms.-reading ἐκεῖ is perfectly
understandable if one construes ἔφασκε (. . .) ὡς ἐκεῖ σκευάζεται. For the construction
of φημί with ὡς, see Kühner-Gerth (1898/1904: 2.356 Anm. 1 and 357, Anm. 3 transi-
tion infinitive to ὡς-clause). This means, of course, that σκευάζεται is not a historical
present either. With Kovacs, I take IT 329 (note the negation) as an imperfect. Κεῖται
(Ph. 1459) is an actual present.
59
In this connection, it is interesting to compare the use of the historical present in
messenger speeches with its use in historical texts. For example, in messenger speeches
it is much more common to find sequences of historical presents in which a complete
(visually) dramatic scene is narrated. In historiography, the historical present tends to
194 rutger j. allan

(9) χρυσοῦν τε θεῖσα στέφανον ἀμφὶ βοστρύχοις

λαμπρῷ κατόπτρῳ σχηματίζεται κόμην,
ἄψυχον εἰκὼ προσγελῶσα σώματος.
κἄπειτ’ ἀναστᾶσ’ ἐκ θρόνων διέρχεται
στέγας, ἁβρὸν βαίνουσα παλλεύκῳ ποδί,
δώροις ὑπερχαίρουσα, πολλὰ πολλάκις
τένοντ’ ἐς ὀρθὸν ὄμμασι σκοπουμένη.
(. . .) and setting the gold crown about her locks, she arranged her
hair in a bright mirror, smiling at the lifeless image of her body. And
then getting up from her seat she paraded about the room, her white
feet making dainty steps, entranced with the gifts, glancing back again
and again at the straight tendon of her leg. (Med. 1160–6)
The fact that Glauce arranges her hair and parades through the house
can in no way be construed as being crucial or decisive to the course of
events. Instead, in my view, by highlighting these actions, the historical
presents stress the dramatic contrast between the trivial vanity of these
actions and the horrific mutilation of Glauce’s body that will follow
shortly. The moment that the poison takes effect (1168–70) is marked
once again by two historical presents (χωρεῖ, φθάνει), here indicating
a narrative Peak (preceded by a comment on the part of the narrator:
τοὐνθένδε μέντοι δεινὸν ἦν θέαμ’ ἰδεῖν).
The subjective nature of the historical present especially becomes
clear in those cases in which there is a discrepancy between, on the one
hand, the crucial event in the plot structure and, on the other hand,
the linguistic Peak (signalled by the historical present). For example, in
Electra 838, the blow to the calf ’s head (κόπτει) is marked by a histori-
cal present, whereas the fatal blow to Aegisthus’ head is expressed by
an aorist form (ἔπαισε). The first action is rhetorically highlighted by
means of the present tense, perhaps as a dramatic foreshadowing of the
killing of Aegisthus. However, from an ‘objective’ point of view, it is,
of course, Aegisthus’ death which is most pertinent to the development
of the plot. Two comparable passages can be found in the Heracles. In
986–9, the poignant scene of Heracles’ second son begging his father not
to kill him is marked by the historical present (φθάνει, αὐδᾷ), whereas
the killing itself is related using two aorists (καθῆκε in 993, ἔρρηξε in
994). Likewise, the remarkable scene in which Heracles takes the door

single out just one single event. The visualizing aspect of the historical present therefore
appears to be somewhat less prominent. Accordingly, the second narrative function
(‘dramatic shot’) is less frequent in historiography.
 towards a typology of the narrative modes 195

from the door case to get to his third son and his wife (example 3) is
told in a continuous series of historical presents, while the actual murder
is narrated by an aorist verb (κατέστρωσεν in 1000).60
In my view, the rhetorical character of the historical present is con-
nected with an essential concept in narrative analysis—tellability (or,
reportability).61 Historical presents typically mark the most tellable
events. The idea of tellability relates to the idea that the person telling a
story claims a considerable amount of ‘social space’: (s)he holds the floor
for a considerable period of time, demanding the undivided attention
of the audience. To justify this action, the narrative must carry enough
interest for the audience. If not, the narrator violates a social norm. A
tellable event is an event that is of special interest to the audience, and
therefore worthy of being told. Tellable events are generally (presented
as) extraordinary, remarkable, unexpected or wonderful. To capture the
audience’s attention from the start, the most tellable event(s) are often
already related in the Abstract.62 Now it is important to note that the
standards of tellability are subjective. That is, which events are viewed
as tellable depends on the purposes of the story-teller and the audience,
and also on the social context of the story-telling.63 In the Heracles, for
example, the killing of the first son can be seen as a tellable (cf. histori-
cal present βάλλει (979)). However, the killings of the second and third
sons are presented by the narrator as less tellable, probably because after
the first killing they are less unexpected to the audience. The narrator,
instead, chooses to highlight the extraordinary events that preceded the
killings, thereby presenting them as more tellable.
After this digression on the historical present I now return to the
Andromache. Episode I of the story ends without a Resolution: there is

60
In some cases, the use of the historical present seems to involve a action-reaction
type of pattern. For example, in the Electra Aegisthus direct speech is systematically
marked by a historical present (El. 779, 783, 813 and 830), whereas Orestes’ verbal
responses are marked by past tenses (lines 781, 793 and 834). In the Orestes, when
a speaker gets up to speak against Orestes, it is marked by the historical present two
times (Or. 887 and 902). The two responding speakers (in favour of Orestes), however,
are introduced with past tenses (lines 898 and 917).
61
Cf. Labov (1972), Fludernik (1991).
62
For example, in the messenger speech of the Andromache the Abstract already
refers to the death and the mutilation of Neoptolemus.
63
A similar conception of the effect brought about by the use of the historical
present is advocated by Weinrich 2001: 52–3. He stresses that the historical present
effects a ‘Haltung der Gespanntheit’ within the hearer or reader, because the narrator
‘erzählt, als ob er bespräche’.
196 rutger j. allan

no relaxation of the tension, but the Complication of the second episode

starts immediately with a Topic-shift (τῶν δ’) from Neoptolemus to the
attackers. After a single displaced diegetic event (οὐδείς (. . .) ἐφθέγξατ’),
a descriptive passage begins in which the states of affairs referred to in
1128–1134 do not take place in sequential order, but simultaneously.
The scene is ‘painted’ (cf. Rijksbaron 20023: 12) by means of a series of
iterative imperfects. The textual progression resides in a form of spatial
movement: the ‘camera’, so to speak, switches from the attackers (ἀλλ’
ἔβαλλον) to Neoptolemus (προὔτεινε τεύχη κἀφυλάσσετ’), and back to
the attackers (ἀλλ’ οὐδὲν ἦνον), and their missiles (ἀλλὰ πόλλ’ ὁμοῦ
βέλη (. . .) ἐχώρουν).
In lines 1135–6, the messenger addresses Peleus directly by means
of a second person counterfactual (ἂν εἶδες),64 which turns Peleus in a
virtual eye-witness. This is a typical device which is used to draw the
narratee into the story (De Jong 1987: 53–60, De Jong 1991: 98, 105,
Lloyd 20052: 169). The effect of this appeal is enhanced, according to
Stevens, by the pathetic reminder of Peleus’ relation (παιδός ‘grandson’)
to the victim. In terms of narrative structure, lines 1135–6 are an Evalu-
ation in the discursive mode.
After this evaluative interruption, the narrative is resumed by means
of a preposed ὡς-clause, a common marker of textual discontinuity.65
Being pressed hard, Neoptolemus leaps from the altar and charges:
three actions following rapidly in succession. Neoptolemus’ charge is
the decisive turning point in this scene and is marked by a historical
present. After this Peak, the topic switches once again to the attackers
(οἱ δ’). The result of Neoptolemus’ strike (Resolution) is that many are
put to flight or are killed. Herewith, the equilibrium is restored.
The next episode starts after a slight temporal lapse indicated by the
adverbial phrase ἐν εὐδίᾳ δέ πως (translated by Kovacs as ‘in the calm
that somehow ensued (. . .)’). The equilibrium is disturbed by a divine
voice coming from the adyton. Urged by the god, Orestes’ men charge
again, and Neoptolemus is immediately killed (πίτνει). The Peak con-
tinues with a repeated rhetorical question alluding to another highly

64
With Wakker (2006), I consider the so-called ‘past potentials’ as identical to
counterfactuals.
65
Other examples of this type of preposed ὡς-clause (indicating a temporal, spatial,
referential or actional discontinuity) in Euripidean messenger speeches are: Ba. 1088,
1095; Hec. 546, 555; Hel. 1530; Ion 1168, 1170, 1198; IT 308, 322, 1354; Med. 1156; Ph.
1143, 1187, 1416, 1472; Supp. 695.
 towards a typology of the narrative modes 197

dramatic moment in literature, the killing of Hector (Il. 22.371). Lin-

guistically, this sentence displays a number of anomalies with respect
to the historical present. We find an historical present in a subordinate
clause (ὡς (. . .) πίτνει), and a negated historical present in a (rhetori-
cal) question (τίς οὐ (. . .) προσφέρει).66 The typical syntagmatic patterns
regarding the use of the historical present are apparently ‘overruled’ in
favour of the strongly dramatic effect of the epic allusion. The asyndeton
βάλλων ἀράσσων iconically conveys the speed of the actions.67
The appearance of a present perfect ἀνήλωται (1154) makes clear
that we have left the Story-Now, and have returned to the Discourse-
Now and the discursive mode. Remarkably, this transition has not
been noticed by Kovacs and Lloyd (who translate ‘was destroyed’ and
‘was rent’).68 With ἀνήλωται we enter the Coda of the story, the bridge
between the narrated world and the social-communicative situation
of narrator (messenger) and narratee (Peleus). Of course, the fact that
Neoptolemus is mutilated (again an epic reminiscence) is a present
source of grief to Peleus. The expressive contrast between καλλίμορφον69
and τραυμάτων ὑπ’ ἀγρίων stress the pathos of the situation. The aor-
ist ἐξέβαλον (1157) might not belong to the story proper but it is an
explanation of how the messenger (who was outside the temple during
the murder) knows the appearance of the corpse, and how they were
able to get hold of it (ἀναρπάσαντες ὡς τάχος χεροῖν) in order to take
it to Peleus (κομίζομέν νιν σοί). In that case, the aorist is not narrative,
but constative. Like ἀνήλωται, the actual present κομίζομεν refers the
moment of speaking. The vocative πρέσβυ makes clear that the utter-
ance is directly addressed to Peleus.
The messenger speech concludes, as usual, with an Evaluation. In
this case the Evaluation condemns the vindictiveness of the god.70 The

66
The repetition of πίτνει in the ὡς-clause immediately after the first πίτνει (a tail-
head construction typical of oral language) seems somewhat redundant because it does
not provide new information. (The adverbial phrase πρὸς γαῖαν, of course, does not
add much to πίτνει) Possibly, this case of ‘chunking’ of the information flow is to be
seen as a rhetorical device used to underscore the dramatic climax of the story. Slings
(2002b: 62–3) discusses similar cases of ‘literary’ chunking at dramatic climaxes in
Herodotus. Slings puts it as: ‘(. . .) the simpler the information supplied in the clause,
the higher the chances are that the distribution phenomena [i.e. chunking, RJA] were
experienced as being “literary”.’ (Slings 1997b: 165, 2002b: 63).
67
For asyndeton in a Peak section, see also example (4).
68
Stevens and Lloyd do not mention it in their commentaries.
69
For the pathetic undertones of epitheta, see De Jong (1991).
70
For a complete inventory of concluding Evaluations, I refer to De Jong (1991: 191).
198 rutger j. allan

aorists ἔδρασ’ and ἐμνημόνευσε are obviously not narrative (for a start,
they are not chronologically ordered). They are, rather, constative ao-
rists, referring to past events which are directly relevant to the pres-
ent. The events of the story are alluded to by the direct object τοιαῦθ’.71
The Evaluation concludes (οὖν) with a rhetorical question containing
a potential optative. The (rhetorical) question, the presence of the
particle οὖν (see table 1), and the optative are features typical of the
discursive mode.

5. Conclusion

Narratives are composed of linguistic units which can be characterized

as narrative modes. In Ancient Greek narrative four types of narra-
tive modes can be distinguished, each of which displays a number of
characteristic linguistic and narratological properties. Furthermore, the
narrative modes play an important part in the composition of the plot
since they show a tendency to appear at certain structural points in
the organization of the narrative. The relationship between the occur-
rence of the narrative modes and narrative structure is synoptically
represented in table 5.
Although there is no one-to-one-relationship between the occurrence
of the narrative modes and particular sections of narrative structure, a
number of strong tendencies can be discerned. Observing these struc-
tural patterns may offer a clearer view on the way in which linguistic
phenomena such as tense/aspect, particles and modality interact in the
organisation of narrative discourse.

71
Note that Evaluations of messenger speeches are often introduced by a form of
τοιοῦτος/τοιόσδε.
 towards a typology of the narrative modes 199

Table 5: Relation between narrative modes and narrative structure

Narrative structure Typical narrative mode Other possible narrative
modes
Abstract Discursive
Orientation Descriptive Diegetic (displaced)
Complication Diegetic (displaced) Diegetic (immediate),
Descriptive, Discursive
(character speech)
Peak Diegetic (immediate) Diegetic (displaced),
Discursive (character
speech)
Resolution Diegetic (displaced) Descriptive, Discursive
(character speech)
Coda Discursive
Evaluation Discursive
 APPENDIX

ANALYSIS OF THE MESSENGER SPEECH IN

THE ANDROMACHE1

Abstract:
οὐκ ἔστι σοι παῖς παιδός, ὡς μάθῃς, γέρον Discursive
Πηλεῦ· τοιάσδε φασγάνων πληγὰς ἔχει
Δελφῶν ὑπ’ ἀνδρῶν καὶ Μυκηναίου ξένου. (1075)
(. . .)

[Episode 1: Ambush, first attempt to murder]

Complication
ἐπεὶ τὸ κλεινὸν ἤλθομεν Φοίβου πέδον, (1085) Diegetic (displaced)
τρεῖς μὲν φαεννὰς ἡλίου διεξόδους
θέᾳ διδόντες ὄμματ’ ἐξεπίμπλαμεν.
καὶ τοῦθ’ ὕποπτον ἦν ἄρ’· ἐς δὲ συστάσεις (Discursive?)
κύκλους τ’ ἐχώρει λαὸς οἰκήτωρ θεοῦ.
Ἀγαμέμνονος δὲ παῖς διαστείχων πόλιν (1090)
ἐς οὖς ἑκάστῳ δυσμενεῖς ηὔδα λόγους·
Ὁρᾶτε τοῦτον, ὃς διαστείχει θεοῦ Discursive
χρυσοῦ γέμοντα γύαλα, θησαυροὺς βροτῶν,
τὸ δεύτερον παρόντ’ ἐφ’ οἷσι καὶ πάρος
δεῦρ’ ἦλθε, Φοίβου ναὸν ἐκπέρσαι θέλων; (1095)
κἀκ τοῦδ’ ἐχώρει ῥόθιον ἐν πόλει κακόν· Diegetic (displaced)

ἀρχαί δ’ ἐπληροῦντ’ ἐς τὰ βουλευτήρια,

ἰδίᾳ θ’ ὅσοι θεοῦ χρημάτων ἐφέστασαν,
φρουρὰν ἐτάξαντ’ ἐν περιστύλοις δόμοις.
ἡμεῖς δὲ μῆλα, φυλλάδος Παρνασίας (1100)
παιδεύματ’, οὐδὲν τῶνδέ πω πεπυσμένοι,
λαβόντες ᾖμεν ἐσχάραις τ’ ἐφέσταμεν
σὺν προξένοισι μάντεσίν τε Πυθικοῖς. Discursive
καί τις τόδ’ εἶπεν· Ὦ νεανία, τί σοι
θεῷ κατευξώμεσθα; τίνος ἥκεις χάριν; (1105)
ὃ δ’ εἶπε· Φοίβῳ τῆς πάροιθ’ ἁμαρτίας
δίκας παρασχεῖν βουλόμεσθ’· ᾔτησα γὰρ
πατρός ποτ’ αὐτὸν αἵματος δοῦναι δίκην.
κἀνταῦθ’ Ὀρέστου μῦθος ἰσχύων μέγα
ἐφαίνεθ’, ὡς ψεύδοιτο δεσπότης ἐμός, (1110)
ἥκων ἐπ’ αἰσχροῖς.

1
The text is taken from Kovacs’ Loeb edition.
202 appendix

Peak Diegetic (immediate)

ἔρχεται δ’ ἀνακτόρων
κρηπῖδος ἐντός, ὡς πάρος χρηστηρίων
εὔξαιτο Φοίβῳ· τυγχάνει δ’ ἐν ἐμπύροις·

Orientation (embedded) Descriptive

τῷ δὲ ξιφήρης ἆρ’ ὑφειστήκει λόχος
δάφνῃ σκιασθείς· ὧν Κλυταιμήστρας τόκος (1115)
εἷς ἦν ἁπάντων τῶνδε μηχανορράφος.

Peak Diegetic (immediate)

χὣ μὲν κατ’ ὄμμα στὰς προσεύχεται θεῷ·
οἳ δ’ ὀξυθήκτοις φασγάνοις ὡπλισμένοι
κεντοῦσ’ ἀτευχῆ παῖδ’ Ἀχιλλέως λάθρᾳ.
χωρεῖ δὲ πρύμναν· οὐ γὰρ εἰς καιρὸν τυπεὶς (1120)
ἐτύγχαν’· ἐξέλκει δὲ καὶ παραστάδος
κρεμαστὰ τεύχη πασσάλων καθαρπάσας
ἔστη ’πὶ βωμοῦ γοργὸς ὁπλίτης ἰδεῖν,
βοᾷ δὲ Δελφῶν παῖδας ἱστορῶν τάδε· Discursive
Τίνος μ’ ἕκατι κτείνετ’ εὐσεβεῖς ὁδοὺς (1125)
ἥκοντα; ποίας ὄλλυμαι πρὸς αἰτίας; —

[Episode 2: Neoptolemus strikes back]

Complication Diegetic (displaced)
τῶν δ’ οὐδὲν οὐδεὶς μυρίων ὄντων πέλας Descriptive
ἐφθέγξατ’, ἀλλ’ ἔβαλλον ἐκ χερῶν πέτροις.
πυκνῇ δὲ νιφάδι πάντοθεν σποδούμενος
προὔτεινε τεύχη κἀφυλάσσετ’ ἐμβολὰς (1130)
ἐκεῖσε κἀκεῖσ’ ἀσπίδ’ ἐκτείνων χερί.
ἀλλ’ οὐδὲν ἦνον· ἀλλὰ πόλλ’ ὁμοῦ βέλη,
οἰστοί, μεσάγκυλ’ ἔκλυτοί τ’ ἀμφώβολοι,
σφαγῆς ἐχώρουν βουπόροι ποδῶν πάρος.

Evaluation Discursive
δεινὰς δ’ ἂν εἶδες πυρρίχας φρουρουμένου (1135)
βέλεμνα παιδός.

Peak Diegetic (immediate)

ὡς δέ νιν περισταδὸν
κύκλῳ κατεῖχον οὐ διδόντες ἀμπνοάς,
βωμοῦ κενώσας δεξίμηλον ἐσχάραν,
τὸ Τρωικὸν πήδημα πηδήσας ποδοῖν
χωρεῖ πρὸς αὐτούς·
 analysis of the messenger speech in the andromache 203

Resolution Diegetic (displaced)

οἳ δ’ ὅπως πελειάδες (1140)
ἱέρακ’ ἰδοῦσαι πρὸς φυγὴν ἐνώτισαν.
πολλοὶ δ’ ἔπιπτον μιγάδες ἔκ τε τραυμάτων
αὐτοί θ’ ὑπ’ αὐτῶν στενοπόρους κατ’ ἐξόδους,
κραυγὴ δ’ ἐν εὐφήμοισι δύσφημος δόμοις
πέτραισιν ἀντέκλαγξ’·

[Episode 3: Divine intervention, death of Neoptolemus]

Complication
ἐν εὐδίᾳ δέ πως (1145)
ἔστη φαεννοῖς δεσπότης στίλβων ὅπλοις·
πρὶν δή τις ἀδύτων ἐκ μέσων ἐφθέγξατο
δεινόν τι καὶ φρικῶδες, ὦρσε δὲ στρατὸν
στρέψας πρὸς ἀλκήν.

Peak Diegetic (immediate)

ἔνθ’ Ἀχιλλέως πίτνει
παῖς ὀξυθήκτῳ πλευρὰ φασγάνῳ τυπεὶς (1150)
[Δελφοῦ πρὸς ἀνδρός, ὅσπερ αὐτὸν ὤλεσε,]
πολλῶν μετ’ ἄλλων· ὡς δὲ πρὸς γαῖαν πίτνει,
τίς οὐ σίδηρον προσφέρει, τίς οὐ πέτρον,
βάλλων ἀράσσων;

Coda Discursive
πᾶν δ’ ἀνήλωται δέμας
τὸ καλλίμορφον τραυμάτων ὑπ’ ἀγρίων. (1155)
νεκρὸν δὲ δή νιν κείμενον βωμοῦ πέλας
ἐξέβαλον ἐκτὸς θυοδόκων ἀνακτόρων.
ἡμεῖς δ’ ἀναρπάσαντες ὡς τάχος χεροῖν
κομίζ̣ομέν νιν σοὶ κατοιμῶξαι γόοις
κλαῦσαί τε, πρέσβυ, γῆς τε κοσμῆσαι τάφῳ. (1160)

Evaluation
τοιαῦθ’ ὁ τοῖς ἄλλοισι θεσπίζων ἄναξ,
ὁ τῶν δικαίων πᾶσιν ἀνθρώποις κριτής,
δίκας διδόντα παῖδ’ ἔδρασ’ Ἀχιλλέως.
ἐμνημόνευσε δ’, ὥσπερ ἄνθρωπος κακός,
παλαιὰ νείκη· πῶς ἂν οὖν εἴη σοφός; (1165)
 CHAPTER TEN

THE USE OF THE IMPERFECT TO EXPRESS COMPLETED

STATES OF AFFAIRS. THE IMPERFECT AS A MARKER
OF NARRATIVE COHESION

Louis Basset

1. Introduction

Often the aspectual opposition between present and aorist stems of the
verb in Ancient Greek is described as follows: the present stem has an
imperfective value, expressing ‘a not-completed state of affairs’,1 whereas
the aorist stem has a perfective or confective value, expressing ‘a com-
pleted state of affairs’ (e.g. Rijksbaron (1994: 1)). The correctness of this
description may be tested when in a narrative a verb is accompanied
by an adverbial expression of duration. If this expression of duration
indicates the total length of a state of affairs, we should have an aorist
stem, whereas, when it does not indicate a total duration, we should
have a present stem. However, whereas this is indeed often the case, we
also find, unexpectedly, the present stem for completed states of affairs.
We will argue that this use of the present stem reinforces the narrative
cohesion and is somehow related to the structure of the narrative. We
have chosen as corpus Herodotus’ Histories, which contain a lot of
adverbial expressions of duration in a narrative context.

2. Expressions of duration in Herodotus

In Herodotus, expressions of duration indicate generally a number of

years (very rarely a number of days or months). They correspond:

1
We will make use of the term state of affairs in its now well known meaning in the
Functional Grammar literature: all sorts of events (states, activities, accomplishments,
achievements), that are expressed by verbs with their subjects and complements.
206 louis basset

– either to the duration of the total accomplishment of a transformative

state of affairs.2 Then the utterance answers the question ‘in how many
years’ (e.g. ‘did he complete his journey?’). In French, the question
is: ‘en combien d’années?’.
– or to a duration which does not cover a total accomplishment. Then
we have two possibilities. Either the state of affairs, whether an
accomplishment, activity or state, has not been completed and was
still continuing at the end of the given duration, this end being taken
as reference point. The statement then answers the question ‘for how
many years at that moment’ (e.g. ‘had he been building his house?’,
or ‘had he been ill?’). In French, the question is: ‘depuis combien
de temps?’. Or the state of affairs has been completed, but is not a
transformative state of affairs (not an accomplishment, but an activity
or a state). Then the statement answers the question ‘for how many
years’ (e.g. ‘did he reign?’ or ‘was he ill?’). In French, the question
is: ‘pendant combien de temps?’.

Therefore three syntactic types must be distinguished:

– Type A answering the question ‘in how many years (did he complete
his journey)?’ is expressed in Ancient Greek by the preposition ἐν
with the dative. The verb is usually in the aorist stem, which has
its confective value, expressing a total accomplishment, unless this
accomplishment belongs to a repetition the end of which is not taken
into account.
– Type B1 answering the question ‘for how many years at that moment
(had he been building his house)?’ is expressed in Ancient Greek by
an accusative case with or without the preposition ἐπί. The verb is
always in the imperfective present stem, which means that the state
of affairs was not completed at the reference point.
– Type B2 answering the question ‘for how many years (was he ill /
did he reign)?’ is also worded in Ancient Greek by an accusative
case with or without the preposition ἐπί. But the verb is then, unlike
in type B1, sometimes in the present stem, sometimes in the aorist
stem. So there is a formal opposition between type B1 and type B2

2
For the classification of states of affairs into state, activity, accomplishment, achieve-
ment, see Vendler (1957). For the basic opposition transformative/non-transformative,
see in particular Ruipérez (1954). This ‘Aktionsart’ typology has also been applied to
Ancient Greek in Stork (1982: 33–8), Sicking (1991: 39–42), etc.
 the imperfect as a marker of narrative cohesion 207

only if the verb has an aorist stem, which excludes type B1. If the
verb is in the present (or perfect) stem, the distinction may be made
only by inspecting the context.

3. Type A: ‘in how many years (did he complete his journey)?’

There are only few instances of this type in Herodotus’ Histories, and
only one in a narrative context. Moreover, since it is found in a subor-
dinate purpose clause, it is not itself narrative. It nevertheless shows the
normal usage of the aorist stem in the context of such an expression.
(1) (. . .) ἵνα δή σφι οἱ παῖδες ἀνδρωθέωσι ἐν τούτοισι τοῖσι ἔτεσι
[They wanted a thirty years’ treaty] in order that their children grow to
be men in these years.3 (7.149.1)
In this example, the prepositional phrase ἐν τούτοισι τοῖσι ἔτεσι goes
as expected with the subjunctive aorist ἀνδρωθέωσι the meaning of
which is confective.
This type of adverbial expression of duration appears also in sec-
tions which are not narrative, for example when we have a general
statement:
(2) (. . .) τὴν περιήλυσιν δὲ αὐτῇ γίνεσθαι ἐν τρισχιλίοισι ἔτεσι.
[They say] that this circular tour is completed by it [the soul] in three
thousand years. (2.123.2)
In this example, the prepositional phrase ἐν τρισχιλίοισι ἔτεσι goes
exceptionally with a present infinitive. This infinitive indicates the
timeless repetition of a total accomplisment. It is this timeless and
therefore unfinished repetition which gives an imperfective meaning
to the verb.

4. The two types B1 and B2 ‘for how many years (had he reigned
at that time?/did he reign?)’

In Herodotus’ Histories, the distribution of stems and moods of verbs

accompanied by accusative expressions of duration is as follows. The

3
I have used Legrand’s text (Collection des Universités de France, Belles Lettres).
The translations are mine.
208 louis basset

presence or absence of the preposition ἐπί is also indicated (AO = aorist

stem, PR = present stem, PFT = perfect stem, FUT = future stem):

– Thirty-three AO:
– seven indicatives (one with ἐπί,4 six without it),5
– twenty-three participles (five with ἐπί,6 eighteen without it),7
– two infinitives of a reported speech (without ἐπί),8
– one dynamic9 infinitive (without ἐπί).10
– Twenty-one PR:
– twelve indicative imperfects (nine with ἐπί,11 three without it),12
– one indicative present in a general statement (without ἐπί),13
– four infinitives of a reported speech (three with ἐπί,14 one without
it),15
– two dynamic infinitives (with ἐπί),16
– two participles (one with ἐπί,17 one without it).18
– One PFT: pluperfect indicative (with ἐπί).19
– One FUT: infinitive in a reported speech (without ἐπί).20
– Three examples have a dubious indicative stem PR or AO as lessons
differ in the manuscripts (without ἐπί).21

The following observations may be made. First, we see that the presence
or absence of the preposition does not entail the choice of an aspect

4
3.59.2.
5
1.29.1; 1.163.2; 1.163.2; 2.157.1; 4.1.2; 7.154.1.
6
1.7.4; 1.130.1; 1.130.1; 2.161.2; 4.159.1.
7
1.14.4; 1.16.1; 1.25.1; 1.86.1; 1.102.1; 1.102.2; 1.106.3; 1.214.3; 2.30.3; 2.133.1;
2.161.1; 3.10.2; 3.66.2; 4.1.3; 4.159.1; 5.89.2; 7.4.1; 7.155.1.
8
2.127.1; 2.128.1.
9
For the term dynamic applied to infinitives which do not report a speech, see
Stork (1982).
10
5.89.3 (although it comes earlier in the sentence, the expression of duration seems
to bear on ἐπισχεῖν, cf. 5.89.2, where it bears on ἐπισχόντας).
11
1.46.1; 1.106.1; 1.166.1; 2.140.2; 2.175.3; 4.95.4; 5.55.1; 5.68.2; 7.20.1.
12
1.18.1; 1.18.2; 1.18.2.
13
3.22.4.
14
1.94.4; 2.137.2; 7.170.1.
15
2.111.2.
16
2.133.3; 3.67.3.
17
1.95.2.
18
4.157.1.
19
1.74.1.
20
1.29.2.
21
1.16.1; 2.157.1; 3.67.2.
 the imperfect as a marker of narrative cohesion 209

stem. There is only a slight correlation, the preposition being more often
used with non-AO stems (sixteen times with a non-AO stem, while
only six times with an AO stem). Conversely, it is more often absent
with AO stems (twenty-seven AO stems without it, and only seven
non-AO stems, not taking into account the three dubious examples).
The weakness of this correlation shows that the presence or absence of
the preposition does not characterize types B1 and B2.22
Second, we observe a stronger correlation between AO stem and
participial mood (twenty-three AO for only two PR). On the contrary,
the indicative mood is more often used with non-AO stem (twelve PR
in the imperfect and one PFT in the pluperfect, against only seven AO).
Such correlations are likely to be connected to subordination (parti-
ciple) or independence (indicative), and consequently to the structure
of the discourse.
Of the examples mentioned above we have as a rule retained only
those which belong to narrative sections, excluding the few cases where
the expression of duration goes with a dynamic infinitive, an infinitive
in the future, or a present indicative asserting a general statement. On
the other hand, we have retained the examples where an infinitive of
reported speech is the substitute of a past indicative, since the reported
speech is then a reported narrative. Finally, all participles have been
retained, although they do not all belong to a fully narrative section:
there is at least a temporal relationship with the main verb, which gives
a point of reference to the participle, as in narrative contexts.

5. Type B1 in narratives (‘for how many years had he reigned

at that time?’)

Investigating the contexts of the examples retained, we find only two

of them where the state of affairs in question is not completed at the
moment which is pointed to in the narrative (the reference point). It is
noteworthy that they are the only two with a participle in the PR stem,
which is in that case obviously imperfective. One of these two examples
is with the preposition ἐπί, the other is without it, which corroborates

22
It is very difficult to ascribe a meaning to the use of ἐπί in this context. It could
stress the length of a state of affairs. But this is a very subjective matter: two years of
illness may seem to last longer than twenty years of reign.
210 louis basset

our above observation that the presence or absence of this preposition

has nothing to do with the distinction of types B1 and B2. Here are
these two examples.
(3) Ἀσσυρίων ἀρχόντων τῆς ἄνω Ἀσίης ἐπ᾿ ἔτεα εἴκοσι καὶ πεντακόσια,
πρῶτοι ἀπ᾿ αὐτῶν Μῆδοι ἤρξαντο ἀπίστασθαι.
At the time when the Assyrians had ruled Upper Asia for five hun-
dred and twenty years, the Medians were the first to begin to shake
themselves free of them. (1.95.3)
The beginning of the Medians’ defection did not at once put an end to
the Assyrians’ rule upon Upper Asia.
(4) ταύτην (sc. Πλατέαν) οἰκέοντες δύο ἔτεα, οὐδὲν γάρ σφι χρηστὸν
συνεφέρετο, ἕνα αὐτῶν καταλιπόντες οἱ λοιποὶ πάντες ἀπέπλεον ἐς
Δελφούς· ἀπικόμενοι δὲ ἐπὶ τὸ χρηστήριον ἐχρέωντο, φάμενοι οἰκέειν
τε τὴν Λιβύην (. . .).
As they had dwelt there for two years (since nothing good had hap-
pened to them), leaving one of them there, all the others sailed away
to Delphi. When arrived at the seat of the oracle, they consulted it,
claiming to have their dwelling-place in Libya (. . .). (4.157.1)
The Thereans’ sailing away to the Delphi oracle had not broken off
their dwelling in Platea, where they had left one of their number, and
where they still claimed to dwell, believing it was in Libya, and where
they would return.
In these two examples, the present participles point to lengths of
time which precede the moment of the main action. Therefore, such
conjunct present participles do not imply as usual simultaneity but
correspond to anteriority. They only indicate that the state of affairs
expressed by the participle was not completed at the end of this pre-
ceeding length of time.

6. Type B2 ‘for how many years did he reign?’ in a narrative

with an AO stem

When the completed duration of a non-transformative state of affairs

is expressed in a narrative, the verb is in the AO stem for twenty-three
participles, seven past indicatives, and two infinitives of a reported
narrative.
 the imperfect as a marker of narrative cohesion 211

6.1. The AO participles

The twenty-three AO participles are conjuncted to a noun of the main
clause, or are in a genitive participial clause. A state of affairs is so set in
the background of the state of affairs expressed by the main verb. This
secondary state of affairs is then viewed from the moment of the main
state of affairs (the reference point). While, as we have seen above, the
PR stem means that this secondary state of affairs is not completed at
the reference point, the AO stem on the contrary means that the whole
secondary state of affairs is viewed from the reference point. Therefore,
when this reference point is given by the main state of affairs, it is
subsequent to the one expressed by the participle.
In that context, the preposition ἐπί is present only five times before
the accusative expression of duration. Here are two examples, the first
with ἐπί, the second without it.
(5) Ἀστυάγης μέν νυν βασιλεύσας ἐπ᾿ ἔτεα πέντε καὶ τριήκοντα οὕτω
τῆς βασιληίης κατεπαύσθη.
Astyages then, who had reigned for thirty-five years, lost his sover-
eignty in this manner. (1.130.1)

(6) Ἄρδυος δὲ βασιλεύσαντος ἑνὸς δέοντα πεντήκοντα ἔτεα ἐξεδέξατο

Σαδυάττης ὁ Ἄρδυος.
As Ardys had reigned during forty-nine years, Sadyattes the son of
Ardys succeeded to the throne. (1.16.1)
Sometimes the main verb does not belong to a narrative in the strict
sense. But the temporal succession between the first state of affairs and
the point of reference given by the main verb is still the same, as can
be seen in the following example.
(7) ἐλθεῖν οἱ μαντήιον ἐκ Βουτοῦς πόλιος ὡς μέλλοι ἓξ ἔτεα μοῦνον βιοὺς
τῷ ἑβδόμῳ τελευτήσειν.
[They said] that an oracle came to him from the town of Bouto,
foretelling that he would, after having lived for only six years, die in
the seventh. (2.133.1)
It happens also that the reference point is not given by the first main
verb. In the following example, the first main verb expresses a state of
affairs which began at the end of the one expressed by the participle
and lasted afterwards against expectation. It is into this stretch of time
212 louis basset

out of the expected one that the reference point is to be located. The
interrupting state of affairs which gives it is stated only in the second
main verb.
(8) τοὺς ὦν δὴ Αἰγυπτίους τρία ἔτεα φρουρήσαντας ἀπέλυε οὐδεὶς τῆς
φρουρῆς· οἱ δὲ βουλευσάμενοι καὶ κοινῷ λόγῳ χρησάμενοι πάντες
ἀπὸ τοῦ Ψαμμητίχου ἀποστάντες ἤισαν ἐς Αἰθιοπίην.
When the Egyptians had been on guard for three years, nobody came
to relieve them. So, taking counsel and coming to a common deci-
sion, they all revolted against Psammetichus, and went to Ethiopia.
(2.30.3)
Such an example may be opposed to examples (3) and (4) above which
are somehow similar, but have PR participles. Here the AO participle
implies that three years was the average length for the Egyptians’ guard.
This expected length of three years was completed and even overstayed
when they revolted against Pasammetichus.

6.2. The AO indicative

Contrary to the constant use of the AO stem with the participles in
type B2, only seven indicatives of this type are AO, opposed to twelve
imperfects (PR stem) and one pluperfect (PFT stem) having the same B2
value (‘for how many years did he reign?’). These seven AO indicatives
therefore contitute a minority. This is why we have tried to find which
contextual features may have caused the rarer choice of an AO stem.

6.2.1. AO indicative due to a flash-back in the narrative

In two of these examples, the AO stem is likely to be due to a flash-back
in the narrative. In such a flash-back to an earlier period, the duration
is described in its totality (confective aorist) from the viewpoint of a
posterior moment, i.e. the moment at which the narrative has arrived
in the preceding context.
(9) (. . .) ἐπεθύμησε ὁ Δαρεῖος τείσασθαι Σκύθας (. . .). τῆς γὰρ ἄνω
Ἀσίης ἦρξαν, ὡς καὶ πρότερόν μοι εἴρηται, Σκύθαι ἔτεα δυῶν δέοντα
τριήκοντα.
Darius became desirous of punishing the Scythians (. . .). For the
Scythians, as I have shown before, had ruled the Upper Asia during
twenty-eight years. (4.1.2)
 the imperfect as a marker of narrative cohesion 213

The reference point arrived at in the previous narrative, the moment

when Darius decided to punish the Scythians, was widely posterior to
the period of their supremacy.23
(10) Κλεάνδρου (. . .) τελευτήσαντος (. . .) ὃς ἐτυράννευσε μὲν Γέλης ἑπτὰ
ἔτεα, (. . .), ἐνθαῦτα ἀναλαμβάνει τὴν μουναρχίην Ἱπποκράτης.
After the decease of Cleandrus (. . .) who had been despot of Gela
for seven years (. . .), then the sovereignty passed to Hippocrates.
(7.154.1)
The participle τελευτήσαντος in the immediately preceding context
points to a moment subsequent to the period of Cleandrus’ sovereignty.
Moreover, the expression of duration is here in a relative subordinate
clause. So this example is analogous to the examples where the verb is
participial and subordinate.

6.2.2. AO indicative when the reference point is the moment of

narration
In the context of the five other AO indicatives (two in the same
example), different hints to the moment of narration suggest that the
past narrative sequence is somehow altered. It is no longer a mere suc-
cession of reported events, each of them constituting its own reference
point. But for one or two of these events the narrator’s viewpoint is
introduced.
In one of these examples, the narration starts again after a digression
which has interrupted it. This digression has brought into the text a
reference to the narrator’s view point. The narrative starts anew with an
AO indicative. It is followed as expected by a PR indicative (imperfect),
but there is an AO variant.
(11) ταῦτα μὲν οὕτω λέγουσι. Ψαμμήτιχος δὲ ἐβασίλευσε Αἰγύπτου
τέσσερα καὶ πεντήκοντα ἔτεα, τῶν τὰ ἑνος δέοντα τριήκοντα Ἄζωτον
(. . .) ἐπολιόρκεε (ἐπολιόρκησεν), ἐς ὃ ἐξεῖλε.

23
We could also say that the inserted clause, (‘as I have shown before’), introduces
in the text a reference to the present narration time. So this example is somehow
analogous to examples (11), (15), (17), where the narration time is taken as reference
point, and so we are no longer in a purely narrative sequence.
214 louis basset

Such is their [Egyptians’] account. Then Psammetichus ruled Egypt

(so they say) for fifty-four years. For twenty-nine of these he besieged
Azotus, untill he took it. (2.157.1)
The digression was about the Bouto’s oracular place. It began at 2.155.1
and interrupted the narrative sequence at the time when Psammetichus
became Egypt’s ruler (2.154.13). It seems that it is the foregoing λέγουσι
which induces an AO stem in the first verb of the resuming narrative:
in a way, although not directly embedding the following, it, as it were,
triggers the narrator’s viewpoint (see also 6.3 below).
In one example, where we have two coordinated AO indicatives,
a present outcome of the past events is mentioned in the sentence
itself.
(12) ἔμειναν δ᾿ ἐν ταύτῃ καὶ εὐδαιμόνησαν ἐπ᾿ ἔτεα πέντε, ὥστε τὰ ἱρὰ
τὰ ἐν Κυδωνίῃ ἐόντα νῦν οὗτοί εἰσι οἱ ποιήσαντες (. . .). ἔκτῳ δὲ ἔτεϊ
Αἰγινῆται αὐτοὺς (. . .) ἠνδραποδίσαντο.
They stayed there and prospered for five years; and it is they who
built the temples which are now at Cydonia (. . .) But in the sixth
year the Aegineans (. . .) took them as slaves. (3.59.2)
The subordinate consecutive clause mentions a present outcome and evi-
dence of the Samians’ long stay in Cydonia. So the whole sentence may
be viewed somehow as an interruption in the narrative sequence.
In two examples, the expression of duration states a completed dura-
tion although it was not completed at the time to which the narrative
context refers. Such a discrepancy implies that this completed duration
can be viewed only from the narration time.
(13) καὶ δὴ καὶ Σόλων ἀνὴρ Ἀθηναῖος, ὃς Ἀθηναίοισι νόμους κελε-ύσασι
ποιήσας ἀπεδήμησε ἔτεα δέκα, κατὰ θεωρίης πρόφασιν ἐκπλώσας.
[Some wisemen came to Sardis] (. . .) among them Solon of Athens,
who after having made laws for the Athenians at their request (. . .)
became a traveller for ten years, sailing the seas to see the world,
as he said. (1.29.1)
The following context relates Solon’s stay in Sardis. At that time, Solon’s
travel far from Athens had not yet ended. So the ten years during which
this exile lasted cannot belong to the narrative sequence. They can be
viewed only from the time of the narration. The relative clause inserts
therefore an interruption in the narrative sequence.
In the second of these examples we have two such AO indicatives.
 the imperfect as a marker of narrative cohesion 215

(14) ἀπικόμενοι δὲ ἐς τὸν Ταρτησσὸν προσφιλέες ἐγένοντο τῷ Βασιλέϊ

τῶν Ταρτησσίων, τῷ οὔνομα μὲν ἦν Ἀργανθώνιος, ἐτυράννευσε δὲ
Ταρτησσοῦ ὀγδώκοντα ἔτεα, ἐβίωσε δὲ τἀ πάντα εἴκοσι καὶ ἑκατόν.
τούτῳ δὴ τῷ ἀνδρὶ προσφιλέες οἱ Φωκαιέες οὕτω δή τι ἐγένοντο
(. . .).
When they came to Tartessos, they became friends with the king of
the Tartessians, whose name was Argantonios, who ruled Tartessos
for eighty years and lived in the whole one hundred and twenty
years. The Phoceans became so friends with that man that (. . .).
(1.163.2–3)
At the time when the Phoceans became friends with the king of the
Tartessians, obviously neither his reign nor his life had ended. Therefore
their whole length can be viewed only from the moment of narration.
Here also the relative clause inserts an interruption in the narrative
sequence.

6.3. The AO infinitives in a reported narrative

When a narrative is reported with infinitives substituting indicatives,
the verb stem is twice AO (without ἐπί), and four times PR. Both
AO stems follow immediately the verb of saying which introduces
the reported narrative (while, as we shall see later, the four PR stems
intervene further in the stream of the reported narrative). These AO
stems at the beginning of the reported narrative still depend directly
on the verb of saying which give them a reference point.
(15) βασιλεῦσαι δὲ τὸν Χέοπα τοῦτον Αἰγύπτιοι ἔλεγον πεντήκοντα ἔτεα.
As the Egyptians said, Cheops reigned for fifty years. (2.127.1)

(16) βασιλεῦσαι δὲ ἔλεγον Χεφρῆνα ἓξ καὶ πεντήκοντα ἔτεα.

As they said, Chephren reigned for fifty-six years. (2.128.1)

7. Type B2 ‘for how many years?’ in a narrative with a PR

(or PFT) stem

With a PR (or PFT) stem, there are twelve imperfect indicatives (of
which nine with ἐπί) and one pluperfect (with ἐπί), besides four infini-
tives in a reported narrative (three with ἐπί). The whole length of a state
216 louis basset

of affairs is still taken into consideration, but not from a later reference
point (neither the time of a later main event of the narrative nor the
narration time). Thus the expressed state of affairs, being a main event
in the narrative sequence, constitutes for itself its own reference point.
It creates then a framework for what follows. It is generally followed
by a subsequent state of affairs which is expressed on the same level in
the narrative sequence. However, events taking place during its length
of time may be uttered first.

7.1. The imperfect indicatives

Here are four of the twelve examples with an imperfect indicative. In
three of them there is a simple succession of events.
(17) Κροῖσος δὲ ἐπὶ δύο ἔτεα ἐν πένθεϊ μεγάλῳ κατῆστο τοῦ παιδὸς
ἐστερημένος· μετὰ δε (. . .) ἀπέπαυσε.
And Croesus, after the loss of his son, sat in deep sorrow for two
years. After this time (. . .) he ceased [his mourning]. (1.46.1)

(18) τοῦτον οἴκεον τὸν χῶρον ἓξ ἔτεα· ἑβδόμῳ δέ (. . .) ἔτεϊ (. . .) ἀνέγνωσαν

ἐκλιπεῖν.
They lived in that country for six years. The seventh year (. . .) They
decided to leave it. (4.158.1)

(19) ἐπὶ μέν νυν ὀκτὼ καὶ εἴκοσι ἔτεα ἦρχον τῆς Ἀσίης οἱ Σκύθαι.
(. . .) καὶ τούτων μὲν τοὺς πλέονας Κυαξάρης τε καὶ Μῆδοι (. . .)
κατεφόνευσαν.
Then the Scythians ruled Asia for twenty-eight years (. . .) The greater
number of them were slain by Cyaxare and the Medes. (1.106.1)
This last example is inserted into a narrative sequence. It may be com-
pared to example (9) above, where the same length of the Scythians’
rule over Asia is expressed with the AO ἦρξαν, because it is not as here
inserted into a narrative sequence, but corresponds to a flash-back into
the foregoing past, and so interrupts the narrative sequence.
In the following example, we have three imperfects. The first cor-
responds to a length of time which is then divided into two successive
parts.
(20) ταῦτα ποιέων ἐπολέμεε ἔτεα ἕνδεκα (. . .) τὰ μέν νυν ἓξ ἔτεα τῶν
ἕνδεκα Σαδυάττης ὁ Ἄρδυος ἔτι Λυδῶν ἦρχε (. . .)· τὰ δὲ πέντε
τῶν ἐτέων τὰ ἑπόμενα (. . .) Ἀλυάττης (. . .) ἐπολέμεε (. . .). τῷ δὲ
δυωδεκάτῳ (. . .) συνηνείχθη (. . .).
 the imperfect as a marker of narrative cohesion 217

He waged war in this way for eleven years (. . .). For six of these
eleven years, Ardys’ son Sadyattes was still ruler of Lydia (. . .). For
the following five, the war was waged by Salyattes (. . .) In the twelth
year, it happened that (. . .). (1.18.1–3)

7.2. PR infinitives in a reported narrative

All infinitives of a reported narrative which have a PR stem (substitut-
ing an imperfect of direct speech) are not at the beginning but in the
middle of the narrative, while expressing a foregrounded state of affairs
(four examples, of which three with ἐπί). Here are two examples.

(21) λέγουσι βαλεῖν (. . .) τυφλωθῆναι; δέκα μὲν δὴ ἔτεα εἶναί μιν τυφλόν·
ἑνδεκάτῳ δὲ ἔτεϊ ἀπικέσθαι οἱ μαντήιον.
They say that he threw (a spear into the river . . . Straightway after
this) he became blind. He then remained blind for ten years. In the
eleventh year, a soothsayer came to him (. . .). (2.111.2)

(22) τὸν μὲν δὴ τυφλὸν τοῦτον οἴχεσθαι φεύγοντα ἐς τὰ ἕλεα, τὸν δὲ

Αἰθίοπα βασιλεύειν Αἰγύπτου ἐπ᾿ ἔτεα πεντήκοντα, ἐν τοῖσι αὐτὸν
τάδε ἀποδέξασθαι.
[They say so and so . . .] Then the blind man fled in the marshes. And
the Ethiopian ruled Egypt for fifty years, during which he behaved
in that manner (. . .). (2.137.2)
In this last example, the following narrative does not lead to the next
events, but the whole specified length of time is taken as a reference
point for what follows. We have seen above the same situation in
example (11), which however has an AO stem. Therefore it is not this
peculiarity which induces the use of the PR stem.

8. Conclusions

When, in Herodotus’ Histories, a verb is accompanied by an expression

referring to the whole length of a non-transformative state of affairs and
answering the question ‘for how many years?’, the choice between AO
or PR stems depends on how the various components of the narrative
are tied to each other:

– The AO stem means either that the state of affairs in question is back-
grounded (mostly participia), and has his reference point specified
218 louis basset

by a main state of affairs which takes place afterwards, or that there

is a interruption into the narrative sequence. Such an interruption
may be caused either by returning to a preceding state of affairs in
the narrative (flash-back), or by introducing the narrator’s point of
view. In the first case, the preceding state of affairs is viewed from
the perspective of the subsequent states of affairs in the narrative. In
the second case, a past state of affairs is viewed from the narrator’s
viewpoint.
– The imperfect indicative (or PR infinitive) on the contrary sets the
state of affairs in question at the main level of a narrative sequence
(foregrounded): his whole length is inserted in the sequence of fore-
grounded events and is not considered from the perspective of any
external moment, neither from another moment of the narrative nor
from the moment of narration.

We can thus distinguish three differents modes of structuring a narra-

tion in Herodotus’ Histories.

– First an even and continuous moving forward of the narrative, where

states of affairs follow one after the other on the same level.
– Secondly, an uneven and hierarchical setting of the narrative, where
some states of affairs are put into the background and viewed from
foregrounded states of affairs.
– Thirdly, an interruption into the narrative sequence, either by return-
ing to a preceding state of affairs, or by introducing the moment of
narration as reference point.

In this context, the observed uses of the present and aorist stems lead
to the same conclusions as those made about various narratives texts.24
The present stem may be used even without an imperfective meaning,
when the state of affairs in question is inserted into a natural narrative
sequence. This use seems to correspond to the concept ‘frayage’ as it
has been formulated by A. Culioli (2000: 20–3). Instead of imperfective,
it may be said continuative and opposed to a discontinuative aorist.
Such a use, especially in imperfect indicatives, is the mark of narrative
cohesion, but not of cohesion with a remote context, as in Rijksbaron

24
See the works of the french GDR 1038 of the CNRS, in particular Basset (2000a:
315; 2003a: 1–7).
 the imperfect as a marker of narrative cohesion 219

(19942: 13). What is in question here is the cohesion between states

of affairs which hang together, each of them leading to the next. It
resembles somehow the use of present stems for the verbs ‘to convince’
or ‘to say’, as described by Rijksbaron (19942: 18–9).
It is thus suggested that, beyond the well known aspectual opposi-
tions, the opposition between the imperfect and the aorist indicative
in Herodotus’ Histories may correspond to the opposition between
narrative past (imperfect) and speech past (aorist), which had been
formulated by Benveniste (1959), and discussed by Weinrich (1964)
and Le Guern (1986). A narrative past is indeed defined as a past not
bound to the narration time, while a speech past is a past bound to the
speech time. The french passé simple ‘il chanta’ is a narrative past, while
the french passé composé ‘il a chanté’ is a speech past. The imperfects of
our examples correspond to the first, while most of the aorist indica-
tives could be characterized as the second.
As was explained by Le Guern (1986: 24–5), nothing prevents a
speech past to be introduced into a narrative, if only a link is established
between the reported state of affairs and the narration time. So it would
be better to oppose a subjective past linked to the present situation and
an objective past not linked to the present situation. In Ancient Greek, as
suggested in Basset (1989a: 246–50), the meaning of verbal tenses seems
to be more subjective than in modern languages, which would suggest
that past tenses involved generally a reference to the speaker’s sphere.
But some less marked verbal forms may have been used to lessen and
remove any link of a past state of affairs to the speaker’s sphere. It may
be that unaugmented verbal forms in the homeric poems had such a
function, opposed to more subjective augmented forms (Basset (1989b)).
But in classical times, the augment was generalised and so lost this first
subjective meaning.25 Perhaps in narrative sequences the imperfect,
which had a less marked aspectual meaning,26 sometimes performed
the same function as the epic unaugmented verbal forms. Having no
longer his polar imperfective value, the imperfect is then reduced to a
continuative value: successive states of affairs of a narrative are thus
linked to one another without any link to the speaker’s sphere.

25
In the same way, the french passé composé, in a language state where the passé
simple becomes obsolete, may become a narrative past (cf. the novel L’Étranger of
Albert Camus).
26
For this conception of the present stem as being a not marked term in the aspectual
opposition, see in particular Sicking (1991) and Basset (2000ab).
 CHAPTER ELEVEN

INVOLVING THE PAST IN THE PRESENT.

THE CLASSICAL GREEK PERFECT AS A
SITUATING COHESION DEVICE

Sander Orriens

1. Introduction1

The term discourse coherence, which refers to the existence of relation-

ships between different parts of discourse, is usually applied in a nar-
row sense, and refers to relationships that exist within the text itself,
i.e. ‘(. . .) between verbally expressed units of discourse’ (Kroon (1995:
63)). The linguistic means by which these relationships can be made
explicit are captured under the term discourse cohesion. Examples of
intratextual cohesion devices are the structuring and organizational
use of (connective) particles, anaphoric pronouns and tense-aspect
distinctions. In this respect attention with regard to discourse coher-
ence and cohesion has mostly been paid to narrative discourse, which,
due to its inherent structural characteristics, displays a large variety of
coherence strategies.2
However, in my opinion the study of coherence and cohesion should
not be restricted to intratextual relationships alone. After all, a text does
not stand on its own but is always part of an external communicative
situation. This connection, in turn, may yield relationships between the
text and the extratextual context and these relationships can equally
be analysed as a kind of coherence. They may, for instance, be made
explicit by means of situating cohesion devices. Kroon (1995: 63), from
whom I adopt this broader idea of discourse coherence and cohesion,

1
I would like to thank my respondent Toon van Wolferen and the participants
of the 6th International Colloquium on Ancient Greek Linguistics (Groningen, 27–29
June 2007) for their comments on the first version of this paper. Special thanks go to
Caroline Kroon, Gerry Wakker, Stéphanie Bakker and Inez van Egeraat for the many
interesting, stimulating and illuminating discussions we have had about this topic.
2
See e.g. for Greek: Bakker (1997b), Allan (2007). For Latin e.g.: Adema (2007),
Kroon (1998, 2007a). These publications especially pay attention to the clustering of
(structuring) linguistic devices in several so-called discourse modes.
222 sander orriens

says the following about particles that fulfil a role as a situating cohe-
sion device:
(1) Particles that somehow fit their host unit into the extratextual reality
(such as modal and focus particles (. . .)) I call, for lack of a better
term, situating particles: they ‘situate’ (or ‘evaluate’), so to speak, their
host unit against the background of (some element of) the extratex-
tual reality. Although they are not connective in a strict sense, these
particles can still be seen as relational devices, since they function as
a trait d’union between the textual and the extratextual/situational.
(my emphasis)
In this contribution I will argue that the Classical Greek perfect3 can be
considered to play a role as a situating cohesion device as well, which
is suggested by a number of its uses in non-narrative discourse. This
role is directly related to the properties of its semantic value, which in
my opinion has to be described differently than commonly assumed.
I will try to show that a speaker may use a perfect in non-narrative
discourse to mark a situating (extratextual) coherence relationship
between a past State of Affairs (SoA)4 and the present communica-
tive situation, which holds primarily on the representational level of
discourse. In this respect it bears a resemblance to situating particles
at the same discourse level. The perfect underlines the actuality that
the speaker ascribes to the past SoA within the context of the present
communicative situation (cf. Kroon (1995: 282–3)). At this point the
perfect contrasts with the aorist, which lacks this actuality and merely
refers to the past SoA.5

2. The semantic value of the Classical Greek perfect 6

If one takes a look at the literature that has been written about the
semantics and pragmatics of the (Classical) Greek perfect, the first thing

3
I use the term ‘perfect’ for the primary perfect indicative in Classical Greek, unless
stated otherwise.
4
I use the term State of Affairs (henceforth: SoA) to refer to ‘ “(. . .) that which is
expressed by a predication” (= roughly a verb form and its arguments, e.g. Agent and
Patient).’ (Rijksbaron (2002: 3)).
5
In this article I will focus on the opposition between the perfect and the aorist. An
other possible opposition, between the perfect and the present, exists in my opinion only
when dealing with inherently stative verbs. This falls outside the scope of this article.
6
This paragraph summarizes the main issues from my Groningen MA thesis
(Orriens (2007a)).
 greek perfect as a situating cohesion device 223

that is striking is the fact that this subject in general has been treated
quite stepmotherly. The reason for this lack of interest is probably the
fact that the perfect occurs significantly less often in the Greek texts
available to us than, for instance, an aorist, which apparently led many
scholars to conclude that no further attention is needed. The effect of
this thought is that treatments of the perfect in standard grammars,
syntaxes and monographs written in the late nineteenth and twentieth
centuries are virtually the same each and every time (apart from some
minor differences) and have hardly changed under influence of more
recent insights in linguistics, viz. the difference between semantics
and pragmatics.7 The communis opinio about the semantic value of
the perfect, which is present (in different, but comparable wordings)
in almost all the well known publications about the subject, can be
expressed as follows:
(2) – The perfect stem signifies both that a State of Affairs is completed
and that as a result a state exists (stative-confective value).
– The primary perfect indicative (commonly: perfect) locates the
state at the moment of utterance (the ‘present’).8
This semantic value, however, only holds well with clearly telic verbs,
i.e. verbs whose semantics contain a clearly distinguishable natural end
point. ‘Less’ telic or atelic verbs leave one in doubt about how to perceive
the result state exactly, because after the end of the atelic SoA there is
no clearly conceivable result. Consider the following two examples:
(3) ΕΥ. αἰαῖ, τέθνηκε. ποῦ δ’ ἐτυμβεύθη τάφῳ;
MN. τόδ’ ἐστὶν αὐτοῦ σῆμ’, ἐφ’ ᾧ καθήμεθα.
Eu. Oh no, is Proteus dead? And where was he buried?
Mn. This is his tomb whereon I am sitting now. (Ar. Th. 885–6)9

7
I have consulted the following publications (in chronological order): Goodwin
(1889), Gildersleeve (1900), Kühner-Gerth (1898/1904), Wackernagel (1904), Stahl
(1907), Smyth (1920), Chantraine (1926), Humbert (1945), Schwyzer & Debrunner
(1950), McKay (1965), Ruijgh (1971), Moorhouse (1982), Rijksbaron (1984), Ruijgh
(1985), Ruijgh (1991), Martinéz Vázquez (1993), Slings (1994), Sicking & Stork (1996),
Duhoux (2000), Rijksbaron et al. (2000), Sauge (2000), Rijksbaron (2002) and Ruijgh
(2004).
8
Rijksbaron (2002: 1 and 4 respectively). Cf. furthermore for instance the passages
concerned with the perfect in: Kühner-Gerth (1898/1904), Stahl (1907), Schwyzer &
Debrunner (1950), Humbert (1960), Ruijgh (1971) and Duhoux (2000).
9
All translations are taken from the Loeb editions. In the case of Aristophanes and
Sophocles I have used the recent editions by Henderson (1998–2002) and Lloyd-Jones
(1991–1994). The Lysias translations are taken from Lamb (1930). The translations are
224 sander orriens

(4) ΣΩ. οὐ τοίνυν τῆς ἐπιούσης ἡμέρας οἶμαι αὐτὸ ἥξειν ἀλλὰ τῆς ἑτέρας.
τεκμαίρομαι δὲ ἔκ τινος ἐνυπνίου ὃ ἑώρακα ὀλίγον πρότερον
ταύτης τῆς νυκτός· καὶ κινδυνεύεις ἐν καιρῷ τινι οὐκ ἐγεῖραί με.
So. Well, mark you, I think it will not come in today, but tomorrow.
And my reason for this thought is a dream which I saw a little
while ago in the course of this night. And perhaps you let me
sleep just at the right time. (Pl. Cri. 44a)

(5) Στ. ἀκήκοας μυριάκις ἁγὼ βούλομαι

περὶ τῶν τόκων, ὅπως ἂν ἀποδῶ μηδενί.
St. You have heard a million times what I want with regard to my
interest payments, a way to avoid paying them to anyone. (Ar.
Nu. 738–9)
In example (3) we see the telic verb ‘to die’ (θνῄσκω) which has a
natural end point at the moment the dying process is over and a per-
son, as a result, is dead. The stative-resultative value is unproblematic
with the perfect of this verb. However, in (4) we find the atelic verb ‘to
see’ (ὁράω), which does not entail a straightforward result state after
the ending of the SoA. Furthermore, the perfect is combined with a
clear marker of past time (ὀλίγον πρότερον ταύτης τῆς νυκτός): both
these elements are problematic for the traditional view of the perfect
expressing a present state resulting from a completed SoA.10 The same
problem is visible in example (5): ‘to hear’ likewise does not entail a
clear result state and the frequentative adverb μυριάκις points in the
direction of a number of past SoAs of ‘hearing’ instead of something
that (frequently) resulted from it. Examples like these have led me
to the conclusion that the semantic value of the perfect has to be
described differently.
Given the observation that it is the perception of a result state that
causes the problem in these examples, it seems plausible that the
semantics of the perfect should be reconsidered exactly at this point.
This means that the aspect of the perfect has to be analysed differently.
In the traditional descriptions of the semantics of this tense we are
dealing with the concept of phasal aspect, meaning that aspect as a
verbal quality relates to different phases of a SoA (e.g. its beginning,
continuation, end, result) which can be represented individually by

adapted where considered necessary. The Greek texts are taken from the online TLG
(http://www.tlg.uci.edu/).
10
See Rijksbaron’s definition in (2).
 greek perfect as a situating cohesion device 225

different aspect stems. In this view the perfect aspect concerns the
phase that exists after the completion of a SoA: the result state. I,
however, think that we have to conceptualize the perfect aspect dif-
ferently and analyse it in terms of perspectival aspect: according to
this idea aspect concerns different ways of perceiving a SoA in terms
of its relationship to an external point of view, usually the moment
of speech.11
Let us now turn towards the primary perfect indicative in Classi-
cal Greek (henceforth again referred to as ‘perfect’): I think that its
aspect, when combined with primary tense, is concerned with marking
a reciprocal (or: bilateral) relationship between a completed past SoA
and the moment of speech. By this I mean that a speaker, when using
a perfect in discourse, simultaneously both refers to a completed past
SoA and explicitly links this SoA to the moment of speech. By doing
this he directly involves the SoA in the present communicative context.
At this point the perfect differs from the closely related aorist, whose
aspect is concerned with a mere unidirectional (or: unilateral) relation-
ship between the moment of speech and the completed past SoA: it
only refers to this SoA without explicitly linking it to the moment of
speech.12 The difference in point of view is also apparent in the mor-
phology of both tenses, viz. in the primary endings of the perfect and
in the secondary endings and the augment of the aorist. The semantic
value of both tenses can be visualized by the following figures, in which

11
See Dik (1997a: 221) for a useful and concise overview of different conceptualiza-
tions of aspect, and Johanson (2000) for an elaborate discussion of viewpoint operators
in European languages. See also the well-known works of Comrie (1976; 1985).
12
This approach is partly based on the spear heads of Current Relevance Theory
(CRT) known from English linguistics, where it is applied to the present perfect (the
Current Relevance Perfect). This elusive and quite controversial theory is based on
the assumption that the present perfect expresses some kind of relevance of the past
SoA at the moment of speech, while it is absent in the simple past (cf. the aorist in
Greek). The exact idea of this relevance remains abstract because its exact interpretation
varies, depending on the verbal semantics and the context. This vagueness has been
the subject of heavy criticism. See for some illustrating discussions of CRT: McCoard
(1978), Fenn (1987) and Elsness (1997). The link between CRT and the Greek perfect
is laid in Slings (1994) and Rijksbaron et al. (2000: 71). Echoes of the theory can be
found in e.g. Duhoux (2000: 421). Using Reichenbach’s symbols these relationships
may roughly be formulized as follows: aorist: [E,R—S]; perfect: [E—R,S]. E is the time
of the event (i.e. the SoA); R is the reference time (the aspectual perspective); S is the
moment of speech. So in the case of the aorist R coincides with E, while in the case of
the perfect R coincides with S. See: Reichenbach (1947: 287–98).
226 sander orriens

the white dot represents the present moment of speech and the black
dot the past SoA:13

Aorist:

Perfect:

This quite abstract basic semantic value of the perfect gives rise to a
number of different interpretations, which depend on the combination
of the inherent semantics of the specific verb involved and its appli-
cation in context. The reciprocal link can for instance be interpreted
as a result state in the case of clearly telic verbs (like τέθνηκα: I am
dead), but has to be interpreted differently in the case of other verbs.14
I think that the perfect in the latter case is a very subtle alternative
for the aorist, which enables the speaker to grammaticalize a differ-
ent, explicitly involving perspective of a past SoA.15 It is here that

13
This reciprocal relationship of the perfect also explains why it is not used as a
narrative tense: the perfect cannot temporally link a past SoA to another past SoA, a
quality which is needed for narrative steps. Furthermore, being a primary tense, the
perfect usually does not allow an exact (i.e. temporally fixed) specification of the past
SoA (example (4) is one of the few exceptions I know in Classical Greek). In both cases
Greek uses the aorist (or the historical present), the imperfect or the pluperfect.
14
Cf. Johanson (2000: 103). My more recent research (Orriens (2007b)) suggests that
even so-called intensive perfects may be analysed eventually in terms of a reciprocal
relationship.
15
In my opinion the reciprocal relationship is also present, mutatis mutandis, in the
pluperfect and future perfect (i.e. with the reference point (which is at the moment of
speech in the case of the perfect) shifting to the past and future respectively). After all,
the problem concerning the result state equally holds for these tenses, which becomes
clear when one takes a look at the pluperfect, especially in narrative discourse: in the
 greek perfect as a situating cohesion device 227

coherence and cohesion enter the game: in the following I will argue
that the reciprocal relationship expressed by the perfect may be used
by a speaker to explicitly mark a situating coherence relationship on
the representational level of discourse, marking the actuality that the
speaker ascribes to the past SoA within the present communicative
situation. In this way, then, the perfect is used as a situating cohesion
device.

3. The Classical Greek perfect as a situating cohesion device

As already described in the introduction, I follow Caroline Kroon in

using the term discourse coherence in a broader sense than what is
usually done. In addition to the common idea that coherence rela-
tionships exist intratextually, i.e. between units within the text, I also
discern extratextual relationships, which hold between the text and
the external communicative situation in which the text is integrated.
In all cases these relationships help in making the act of communica-
tion successful: if a speaker wants an addressee to understand what
he tries to communicate, he has to make sure that his expression fits
into the communicative situation and that it relates coherently both
to the direct linguistic (intratextual) and non-linguistic (extratextual)
context. Such coherence relationships may occur on different discourse
‘levels’, which are concerned with different aspects of the act of com-
munication.
A very fruitful way of analyzing coherence with regard to levels of
discourse can be found in Kroon’s model for the analysis of Latin dis-
course particles (Kroon (1995: 58–96). She distinguishes between three
discourse levels: the representational level, the presentational level and
the interactional level. The first level concerns the representation of
SoAs in the real or imaginary world and their semantic (e.g. temporal,

case of atelic verbs this tense seems to be used as an alternative for the aorist. More
research on this topic is certainly needed. The (synthetic) future perfect is virtually
limited to telic verbs and therefore almost always yields plausible result state interpre-
tations. Furthermore, this revised approach of the core semantic value of the perfect
opens up new ways of approaching the elusive diachronic development of this tense/
aspect category, which should be done by means of a typologically and methodologi-
cally sound analysis based on modern linguistic insights. I have made a first start with
developing such an approach in my Cambridge MPhil thesis (Orriens (2008)), which
has yielded very stimulating results.
228 sander orriens

contrastive) relationships. The second level is concerned with the pre-

sentation and structuring of discourse units by means of organizational
(hierarchical) and rhetorical relationships. The third level refers to the
relationship between discourse units in the light of the conversational
exchange and the (extra-linguistic) communicative situation.16
These different coherence relationships are often (but not necessarily)
explicitly marked by linguistic elements which may be called cohesion
devices. Discourse particles are a clear example of such coherence
markers, because they are ‘(. . .) involved in signalling or emphasizing
the coherence of a stretch of discourse’ by signalling different kinds
of relationships (Kroon (1995: 58)). But tense and aspect distinctions
play an important role as well. A familiar example in Greek is the
organizational alternation between aorists and imperfects in narrative
discourse, which reflects coherence relationships on the presentational
level of discourse (see e.g. Rijksbaron (2002: 11–4)). For the present
purposes I will focus on extratextual coherence relationships.
Within her framework for analyzing discourse particles Kroon mainly
discusses connective particles, i.e. those particles that mark intratextual
coherence relationships. However, in addition to this group she also
distinguishes (what she calls) situating particles, ‘which are only con-
nective in a broad sense.’17 In order to understand what is meant by this
it will prove useful to first quote Kroon’s (1995: 35) broad definition
of the term particle:
(6) [P]articles are those invariable words which have in common that
they fit their host unit into a wider perspective, which may be the
surrounding verbal context and its implications [i.e. intratextual
relations, SO], or the communicative situation in which the text is
integrated [i.e. extratextual relations, SO]. (my underscoring)
Now, I think that we may apply this functional definition to tense
and aspect categories as well, given the fact that they also somehow
fit their host unit (i.e. the SoA) into the wider perspective of both the
text and the communicative context. Usually, this ‘hosting’ is merely
intratextual: verbal forms are used to represent SoAs from the real or
some imaginary world within the text and may also be deployed in
the organization (the presentation) of the text, especially in narratives,
where the order of events and the foreground or background status of

16
See Kroon (1995: 58–96). See also e.g.: Wakker (1997: 211–3). The order in which
these levels are treated is arbitrary and is not meant to reflect actual mental processes.
17
Kroon (1995: 282). Cf. also (1).
 greek perfect as a situating cohesion device 229

SoAs can be manipulated by tense and aspect variations; in this case

the function of tenses is comparable to that of connective particles in
the narrow sense, i.e. linking textual units to each other in some coher-
ent way. But what about extratextual ‘connections’? Let us look again
at Kroon’s (1995: 63) definition of situating particles, which I already
discussed briefly in the introduction of this article (see (1)):
(7) Particles that somehow fit their host unit into the extratextual reality
(such as modal and focus particles (. . .)) I call, for lack of a better
term, situating particles: they ‘situate’ (or ‘evaluate’), so to speak, their
host unit against the background of (some element of) the extratex-
tual reality. Although they are not connective in a strict sense, these
particles can still be seen as relational devices, since they function as
a trait d’union between the textual and the extratextual/situational.
Situating particles thus explicitly integrate their host unit into some
aspect of the present (extratextual) communicative situation, and may,
according to Kroon, function on either the representational level or the
interactional level, or on both at the same time.18 For now we will focus
on representational situating particles, and on the specific category of
objective modality markers (Kroon 1995: 283):
(8) Objective modality markers (e.g. indeed, really, actually, maybe) (. . .),
evaluate the actuality-status of a state of affairs in the represented
world as compared to a hypothetical or possible alternative. (my
underscoring)
Different modality markers (i.e. different representational situating par-
ticles) give rise to different ways of evaluation, but this is not relevant
for the present discussion. What I am interested in is the idea of the
marking of actuality status in general: I think that the Classical Greek
perfect bears some kind of resemblance to (but is certainly not identical
to)19 objective modality markers in that it makes explicit the speaker’s
ascription of actuality to a completed past SoA within the present com-
municative situation.

18
Kroon (1995: 282). It is to be expected that such particles do not act on the
presentational level as defined by Kroon, because this level is strictly concerned with
the organization of the text itself and is therefore by definition limited to intratextual
coherence relationships.
19
An important difference between objective modality particles (like Latin vero and
Greek δή) and the Classical Greek perfect is that the former are concerned with actuality
status in terms of ‘real’ (as opposed to ‘merely hypothetical’), while the latter is solely
concerned with actuality status in terms of ‘actual within the present communicative
situation’ (as opposed to ‘it (merely) happened there and then’).
230 sander orriens

When communicating, especially (but certainly not exclusively) in

dialogues, a speaker stands in direct confrontation with his addressee
and with the present communicative situation in which he finds him-
self. When referring to SoAs which took place in the past the speaker
may use the aorist. He may, however, also use the perfect, which adds
a situating dimension to the reference: by using the perfect the speaker
does not only (unidirectionally, like the aorist) refer to the SoA, but
also connects this SoA to the moment of speech and the extratextual
communicative situation. By doing this he represents the past SoA in a
different way than with the aorist, attributing actuality status to it. By
this I mean that the speaker considers the SoA to somehow be part of
the communicative situation and he wants his addressee to interpret
this direct relationship. The reasons the speaker may have for doing
this depend on the specific communicative context, as we will see later
on when I discuss individual cases. Let us first look again at the figures
of the semantic value of the aorist and the perfect in the light of the
issues just discussed:

Referring

Aorist:

Referring, and...
...attributing actuality
status (representational
Perfect: situating)

In this way (and by using the term discourse coherence in the broad
sense) the perfect can be seen as a cohesion device: it linguistically
makes explicit the extratextual relationship between a past SoA and
the present communicative situation. This relationship can be termed
coherence, because it is an example of two elements considered by the
speaker to be connected to each other. In the same way as a speaker
may intratextually link two units to each other by means of a cohe-
sion device, he may also link an extratextual idea (here: a SoA) to the
communicative situation, by means of language. Just like situating
 greek perfect as a situating cohesion device 231

particles, then, the perfect may serve as a situating cohesion device on

the representational level of discourse.20
The approach sketched above is the result of work in progress, and
obviously needs to be further refined. It is based on the corpus of (pre-
dominantly) non-narrative texts I used for my Groningen MA thesis. Of
course, more work on a larger corpus is needed.21 In the remainder of this
contribution, I will try to show how the principles discussed are operat-
ing in context by discussing a number of examples from my corpus. I
will also discuss related aorist forms, which enable a clearer revelation
of the perfect’s role in establishing coherence relationships.

4. The Classical Greek perfect in context

As I already indicated earlier the perfect is not a narrative tense. It is

mainly used in monological and dialogical discourse, where the speaker’s
base (or: point of view) is located in his own present.22 This observation
is not surprising when we take the semantic value of the perfect into
account: the past SoA is overtly connected to the present moment of
speech. The monological and dialogical discourse types are also typical
for communicative environments where a speaker is in direct contact
with his adressee(s), which prompts the presence of situating coher-
ence relationships and the explicit marking of these relationships by
means of situating cohesion devices, including the perfect. I will now
discuss a number of perfects in their contexts, comparing them to
related (non-narrative) aorist forms.23 We will be dealing with atelic
verbs, because in these cases the difference between the perfect and the
aorist becomes the most clear.

20
Slings (1994: 244) relates the use of the (non-resultative) Classical Greek perfect
to interactional motivations (as opposed to the (in his eyes) representational aorist
and presentational imperfect). Unfortunately he does not elaborate on how he sees
this exactly.
21
My corpus consists of the tragedies of Sophocles, the comedies of Aristophanes
and the speeches of Lysias. Fragments have not been taken into account. Other texts
that need to be taken into account are e.g. the tragedies of Aeschylus and Euripides,
the dialogues of Plato and narrator’s comment and speeches in e.g. Herodotus and
Thucydides.
22
See Adema (2007: 53–9) for the term base.
23
The main (or rather: best known) function of the aorist is its use as a narrative
tense. It may however also be used in other discourse environments where the speaker
wants to refer to the past.
232 sander orriens

The first example is a well known passage from Sophocles’ Antigone

that contains a striking minimal pair of a perfect infinitive and an aorist
infinitive, linked to a present tense form:
(9) Κρ. σὲ δή, σὲ τὴν νεύουσαν ἐς πέδον κάρα,
φής, ἢ καταρνῇ μὴ δεδρακέναι τάδε;
Αν. καὶ φημὶ δρᾶσαι κοὐκ ἀπαρνοῦμαι τὸ μή.
Cr. You, then, you that are bowing down your head towards the
ground, do you admit, or do you deny, that you have done
this?
Ant. I admit that I did it and do not deny it. (S. Ant. 441–3)
Creon’s guard has just told that he has seen Antigone burying the body
of her brother Polyneices (verses 422–31). For Creon it is now clear
who is guilty and he turns towards Antigone who is standing before
him. Two cohesion observations can be made: first, he explicitly marks
the obviousness with which his question follows from the guard’s story
by means of the modal (evidential) particle δή, thereby underlining a
situating interactional relationship between his utterance and the com-
municative context.24 Furthermore, while the guard has constantly used
aorists or imperfects to refer to Antigone’s act,25 Creon uses a perfect
(δεδρακέναι): for him personally the SoA is not a mere fait accompli
that belongs to the past, but a matter that has to be resolved here and
now. This attitude towards the SoA is made explicit by the choice for the
perfect instead of the aorist.26 Antigone herself, in turn, subtly parries
this attitude and simply refers to the fact that she committed the act
(aorist: δρᾶσαι), nothing more and nothing less: after all, in her eyes
she has done nothing wrong (cf. Rijksbaron (2002: 11–4)).
By means of the reciprocal relationship of the perfect, Creon explicitly
involves the SoA in the context of the conversation and ascribes actual-
ity to it, while the utterance as a whole is interactionally integrated by

24
See Kroon (1995: 94–5; 282–4). Interactional situating particles ‘(. . .) are some-
how oriented to indicating the involvement of the discourse partners in the speech
event (and their management of the conversation) rather than at conveying content’
(Kroon 1995: 283).
25
In non-narrative discourse: ἐξειργασμένη (line 384), ἣ καθῃρέθη τάφον κοσμοῦσα
(lines 395–396), αὐτὴ τὸν ἄνδρ’ ἔθαπτε (line 402), ταύτην γ’ ἰδὼν θάπτουσαν (line
404).
26
Cf. Johanson (2000: 103) who states that the semantic value of the (English)
perfect among other things ‘(. . .) allows a conclusive judgment at O [i.e. the moment
of speech, SO].’
 greek perfect as a situating cohesion device 233

means of the particle δή. The actuality is totally absent in Antigone’s

neutral response.
The sense of guilt that is often ascribed to a perfect in a context
like this can easily be linked to its semantic value and its situating
coherence marking: for it is exactly this characteristic that a speaker
enforces on his addressee’s interpretation when he connects the latter’s
considered negative (criminal) act to the moment of speech. The same
holds for positive acts and the awarding of credit.27 First consider the
next examples:
(10) παύσομαι κατηγορῶν. ἀκηκόατε, ἑωράκατε, πεπόνθατε, ἔχετε·
δικάζετε.
I will conclude my accusation. You have heard, you have seen, you
have suffered, you have him: judge him. (Lys. 12.100)

(11) φανερῶς ἔχετε αὐτὸν ἀσεβοῦντα· εἴδετε, ἠκούσατε τὰ τούτου

ἁμαρτήματα. ἀντιβουλήσει καὶ ἱκετεύσει ὑμᾶς· μὴ ἐλεεῖτε. οὐ γὰρ
οἱ δικαίως ἀποθνῄσκοντες ἀλλ’ οἱ ἀδίκως ἄξιοί εἰσιν ἐλεεῖσθαι.
You have him, caught in the the open commission of impiety; you
saw, you heard his offences. He will beseech and supplicate you:
have no pity. For it is not those who justly, but those who unjustly,
suffer death that deserve to be pitied. (Lys. 6.55)
Both these passages are parts of speech conclusions, a section where
one would expect a great deal of interaction between the orator and his
public; the presence of a perfect is therefore quite unsurprising within
the light of the discussion above. The motivation for the speaker’s aspect
choices in examples (10) and (11) can be explained when we take a
look at both the internal structure of the text and the communicative
situation. The perfects in the first example are located at the very end
of the speech and are thus among the last words the audience will hear.
By using the perfects, the speaker ascribes actuality to the past SoAs
(the hearing, seeing and suffering of his addressees), by connecting them
directly to the moment of speech. He thereby explicitly situates them in

27
Cf. e.g. the remarks of Stahl (1907: 113–4): ‘Ferner kann das Subjekt als Urheber
eines vergangenen Ereignisses bezeichnet werden (. . .). Insbesondere kann als einen
gegenwärtigen Zustand darstellend alles Vergangene aufgefasst werden, das zur Cha-
rakteristik eines Gegenwärtigen dient oder eine vorhandene Schuld oder ein vorhan-
denes Verdienst bekundet, ein Gebrauch, der besonders bei den Rednern vorkommt,
da hierin zugleich ein für die Verhandlung massgebendes Moment enthalten ist.’ Cf.
also Chantraine (1926: 164–6), McKay (1965: 13), Ruijgh (1971: 250), Sicking & Stork
(1996: 146), Duhoux (2000: 143), Rijksbaron (2002: 36–7).
234 sander orriens

the direct communicative context of the courtroom for reconsideration,

evaluation and, ultimately, judgement (δικάζετε).
The related aorists in example (11) have a different place and func-
tion: firstly, they do not stand in the ultimate line of the speech and,
secondly, they do not have the same hierarchical status as their perfect
counterparts in the previous example. The main issue of the sentence is
φανερῶς ἔχετε αὐτὸν ἀσεβοῦντα, which is subsequently illustrated by
the two aorists: they form a subsidiary unit towards the more central
unit and therefore are less demanding for situational marking. They
simply refer to past SoAs without the explicit connection to the moment
of speech and the communicative context.
The following example contains another interesting minimal pair of
a perfect and aorist:
(12) (Με) ἤδη ποτ’ εἶδον ἄνδρ’ ἐγὼ γλώσσῃ θρασὺν
ναύτας ἐφορμήσαντα χειμῶνος τὸ πλεῖν,
ᾧ φθέγμ’ ἂν οὐκ ἐνηῦρες, ἡνίκ’ ἐν κακῷ
χειμῶνος εἴχετ’, ἀλλ’ ὑφ’ εἵματος κρυφεὶς
πατεῖν παρεῖχε τῷ θέλοντι ναυτίλων.
οὕτω δὲ καὶ σὲ καὶ τὸ σὸν λάβρον στόμα
σμικροῦ νέφους τάχ’ ἄν τις ἐκπνεύσας μέγας
χειμὼν κατασβέσειε τὴν πολλὴν βοήν.
(Τευ) ἐγὼ δέ γ’ ἄνδρ’ ὄπωπα μωρίας πλέων,
ὃς ἐν κακοῖς ὕβριζε τοῖσι τῶν πέλας.
κᾆτ’ αὐτὸν εἰσιδών τις ἐμφερὴς ἐμοὶ
ὀργήν θ’ ὁμοῖος εἶπε τοιοῦτον λόγον,
‘ὤνθρωπε, μὴ δρᾶ τοὺς τεθνηκότας κακῶς·
εἰ γὰρ ποήσεις, ἴσθι πημανούμενος.’
τοιαῦτ’ ἄνολβον ἄνδρ’ ἐνουθέτει παρών.
ὁρῶ δέ τοί νιν, κἄστιν, ὡς ἐμοὶ δοκεῖ,
οὐδείς ποτ’ ἄλλος ἢ σύ. μῶν ᾐνιξάμην;
(Men) Once I saw a man of reckless speech, urging sailors to sail
during a storm. But one heard no word from him when he
was in the grip of the storm’s attack; he huddled up under
his cloak and allowed any sailor who wished to trample on
him. It is just so with you and your loud mouth: a mighty
tempest, blown from a small cloud, shall quickly put a stop
to all your shouting.
(Teu) And I have seen a man full of stupidity, who harried neigh-
bours in their time of troubles. And then a man like me
and of the same temper saw him, and spoke such words
as these: ‘Fellow, do not persecute the dead; for if you
do so, know that you shall suffer pain!’ That was how he
rebuked the miserable man directly; and, mark my words,
 greek perfect as a situating cohesion device 235

I see that man, and he is, I think, none other than you. Or
did I speak in riddles?
(S. Aj. 1142–58)
Menelaos and Teucer, arguing about the burial of Ajax, compare each
other with some disagreeable character they have encountered earlier in
their lives. Menelaos uses an aorist (εἶδον) to refer unidirectionally to
this moment of seeing, considering the experience to be there and then,
and subsequently compares this man with the person he is talking to
right now (οὕτω δὲ καὶ σὲ κτλ.). Teucer, in turn, immediately responds
to the issue (overtly marked by ἐγὼ δέ γ’) and uses the perfect (ὄπωπα)
not only to refer to the past SoA, but also to make this experience part
of the present communicative situation. This has a very subtle effect
here: Teucer’s perfect provides a signal of him actualizing his experi-
ence within the present situation, exactly because he wants us (and
Menelaos of course) to suspect that something is going on. But due to
the fact that the perfect itself remains vague about what the implication
exactly is, we are forced to read on further until the climax: ὁρῶ δέ τοί
νιν, κἄστιν, ὡς ἐμοὶ δοκεῖ, οὐδείς ποτ’ ἄλλος ἢ σύ. Note in this respect
the interactional (modal) particle τοι ‘mark my words’, which indicates
that the speaker wants the addressee to pay particular attention to what
he is saying. The hint coded in the situating coherence relationship
between the past experience and the present communicative context
marked by the perfect, is now totally made clear: the experience as well
as the disagreeable character are exactly the same, both in the past and
in the present communicative situation.28
This involving or actualizing use of the perfect and its functioning as
a situating cohesion device are mainly seen in cases of atelic verbs that
are concerned with the evaluation of some kind of past experience (verba
sentiendi, like in the previous two examples), or atelic verbs dealing
with evaluation in terms of positive (credit) or negative (guilt) liability
(as in example (9)). In all these cases the speaker attributes actuality to
the past SoA, and considers it to be part of the present communica-
tive situation. This observation can be explained by the fact that both
experiences and questions of liability often (inherently) entail a direct

28
Cf. also the perfect ἑώρακα in example (4): the past experience of the dream is
directly involved in the communicative context. The perfect ἀκήκοας in (5) actualizes
the experience of countless (μυριάκις) SoAs of ‘hearing’, from an undefined point in
the past right up to the moment of speech.
236 sander orriens

(reciprocal) relationship between the ‘now’ and ‘then’ and the speaker’s
choice for the perfect is therefore both semantically and pragmatically
quite logical, due to the latter’s reciprocal / bidirectional semantic value.
I will now provide some further examples from my corpus:
(13) ἐπίστασθε δέ, ὦ ἄνδρες δικασταί, ὅτι παράδειγμα τοῖς ἄλλοις ἔσται
μὴ τολμᾶν εἰς ὑμᾶς ἐξαμαρτάνειν οὐχ ὅταν τοὺς ἀδυνάτους εἰπεῖν
κολάζητε, ἀλλ’ ὅταν παρὰ τῶν δυναμένων λέγειν δίκην λαμβάνητε.
τίς οὖν τῶν ἐν τῇ πόλει ἐπιτηδειότερος Νικομάχου δοῦναι δίκην; τίς
ἐλάττω τὴν πόλιν ἀγαθὰ πεποίηκεν ἢ πλείω ἠδίκηκεν; ὃς καὶ τῶν ὁσίων
καὶ τῶν ἱερῶν ἀναγραφεὺς γενόμενος εἰς ἀμφότερα ταῦτα ἡμάρτηκεν.
Understand, gentlemen of the jury, that it will be an example to
the rest, and will deter them from committing offences against you,
if instead of punishing unskillful speakers you exact requital from
the skillful. And from whom amongst our citizens could it be more
suitably exacted than from Nicomachus? Who has rendered less
service or done more wrong to the city? Appointed to transcribe
our code of duties, secular and sacred, he has offended against both.
(Lys. 30.24–5)

(14) ῾Ερ. ἐλθοῦσά φησιν αὐτομάτη μετὰ τἀν Πύλῳ

σπονδῶν φέρουσα τῇ πόλει κίστην πλέαν
ἀποχειροτονηθῆναι τρὶς ἐν τἠκκλησίᾳ.
Τρ. ἡμάρτομεν ταῦτ’· ἀλλὰ συγγνώμην ἔχε·
ὁ νοῦς γὰρ ἡμῶν ἦν τότ’ ἐν τοῖς σκύτεσιν.
Her. She says that after the events at Pylos she came here of her
own accord, offering the city a basketful of treaties, and was
voted down three times in the Assembly.
Tr. We made the mistake, but do pardon us: at that time our
brains were in our shoe leather. (Ar. Pax 665–9)
In example (13) the orator is directly confronting his addressees dur-
ing his argumentation. Note in this respect the imperative ἐπίστασθε
and the vocative ὦ ἄνδρες δικασταί, which both are explicit markers
of interaction. The perfects also contribute to this interaction, again by
underlining the speaker’s ascription of actuality to the acts committed
by the accused in the wider context of evaluating (i.e. denouncing) them
within the communicative situation of the courtroom.
It is exactly the actuality that is (intentionally; cf. Antigone’s words
in (9)) absent from the aorist ἡμάρτομεν in example (14), which con-
cerns a quick apology for a rather big mistake: Lady Peace is angry
for not being properly treated by Trygaeus and the latter tries to find
a way out. By using the aorist (and asking forgiveness) he detaches
 greek perfect as a situating cohesion device 237

the mistake from the moment of speech and tries to counter possible
further allegations by giving an explanation why things went wrong
there and then (τότε). If he would have used a perfect, he would have
underlined a relationship between his act and the present moment of
speech, considering it to be part of the communicative situation and
thereby actually acknowledging his guilt.
The next example shows another interesting minimal pair of a perfect
and an aorist:
(15) Πει. ἀτάρ, ὦ ποητά, κατὰ τί δεῦρ’ ἀνεφθάρης;
Πο. μέλη πεποίηκ’ εἰς τὰς Νεφελοκοκκυγίας
τὰς ὑμετέρας κύκλιά τε πολλὰ καὶ καλὰ
καὶ παρθένεια καὶ κατὰ τὰ Σιμωνίδου.
Πει. ταυτὶ σὺ πότ’ ἐπόησας; Ἀπὸ πόσου χρόνου;
Πο. πάλαι, πάλαι δὴ τήνδ’ ἐγὼ κλῄζω πόλιν.
Πει. οὐκ ἄρτι θύω τὴν δεκάτην ταύτης ἐγώ,
καὶ τοὔνομ’ ὥσπερ παιδίῳ νυνδὴ ’θέμην;
Peis. Now why the hell did you come up here, poet?
Poet I have composed songs for your Cloudcuckooland, lots of
fine dithyrambs, maiden songs and songs à la Simonides.
Peis. When did you compose these songs? Starting when?
Poet I have been celebrating this city for a long, long time.
Peis. But I have just begun its tenth-day sacrifice, and named it,
like a baby, just moments ago! (Ar. Av. 915–23)
The situating relationship between the past SoA of composing and
the present communicative situation lies in the observation that the
poet uses the perfect to relate the SoA (and here also his utterance as
a whole) to the preceding question of Peisistratus: ‘I have composed
songs’ (with the conversational implicature: and that is why I am here).
The perfect thus helps to explicitly place the utterance in the line of the
actual conversation. The aorist in the reactive move is anchored to the
past (and not to the present) and the moment of composition, which
is made explicit by the interrogative πότε (‘when?’) and the subsequent
specifying question (Ἀπὸ πόσου χρόνου;).
The last examples we will discuss contain a perfect and an aorist of
the same verb in very similar contexts.
(16) Now, if it were to the advantage of your people that, while some
kept their own, others had to suffer the unjust confiscation of their
property, you would have some reason to neglect our arguments.
νυνὶ δὲ πάντες ἂν ὁμολογήσαιτε ὁμόνοιαν μέγιστον ἀγαθὸν εἶναι
πόλει, στάσιν δὲ πάντων κακῶν αἰτίαν, διαφέρεσθαι δὲ πρὸς
238 sander orriens

ἀλλήλους ἐκ τῶν τοιούτων μάλιστ’, ἐὰν οἱ μὲν τῶν ἀλλοτρίων

ἐπιθυμῶσιν, οἱ δ’ ἐκ τῶν ὄντων ἐκπίπτωσι. καὶ ταῦθ’ ὑμεῖς ἔγνωτε
νεωστὶ κατελθόντες, ὀρθῶς βουλευόμενοι· ἔτι γὰρ ἐμέμνησθε τῶν
γεγενημένων συμφορῶν, καὶ (. . .).
(. . .) but now in fact you must all acknowledge that unanimity is the
greatest boon to a city, while faction is the cause of all evils; and that
mutual dissensions chiefly arise from the desire of some for what is
not theirs, and the ejection of others from what they have. This you
judged shortly after your return, and your reasoning was sound; for
you still remembered the disasters that had occurred, and (. . .). (Lys.
18.17–8)

(17) νυνὶ δὲ λαχόντος ἐν τῷ Γαμηλιῶνι μηνὶ οἱ ναυτοδίκαι οὐκ ἐξεδίκασαν.

ἐπειδὴ δ’ ὑμῖν τὰ Ἐρασιφῶντος δημεύειν ἔδοξεν, ἀφεὶς τῇ πόλει τὼ
δύο μέρει τὰ Ἐρασιστράτου ἀξιῶ μοι ψηφισθῆναι, διότι ταῦτά γε
ἤδη καὶ πρότερον ἐγνώκατε ἡμέτερα εἶναι.
But now, although I was permitted to bring proceedings in the month
of Gamelion, the nautical court has not decided the case. Now that
you saw fit to confiscate the property of Erasiphon, I relinquish two
thirds to the State, and claim that the property of Erasistratus be
adjudged to me, because you already judged earlier that this property
is ours. (Lys. 17.5–6)
In passage (16) the speaker discusses the common values of the city in
connection with the confiscation of private property and actively asks
for the jury’s consent to his opinion (πάντες ἂν ὁμολογήσαιτε κτλ.).
He reinforces and substantiates this point by referring to their own
judgement they made in the past (ἔγνωτε), of which the details of are
subsequently related.
In (17) the speaker does a similar thing: he again refers to the judge-
ment the jury made earlier, here with regard to the awarding of property,
but uses a perfect (ἐγνώκατε) instead of an aorist. Given the importance
of this past judgment for his present purposes (getting back the prop-
erty of Erasistratus, of which he claims to be the rightful owner), he
chooses to explicitly connect the SoA to the moment of speech. He thus
directly involves the judgement in the present communicative situation
and thereby underlines its actuality, expecting his audience (the jury)
to subscribe to it. The need for such actuality is less prominent, if not
absent, in (16): the audience’s subscription is already prompted by the
consent-triggering (interactional) ἂν ὁμολογήσαιτε: the neutral aorist
is therefore enough for making the point.
 greek perfect as a situating cohesion device 239

5. Conclusion

In this contribution I have tried to show how the perfect may be used
as a situating cohesion device for the explicit marking of an extratextual
coherence relationship between a past SoA and the present communica-
tive situation. This usage is directly related to the core semantic value of
the perfect, which in my opinion should be described as the establish-
ment of a reciprocal relationship between a completed past SoA and
the moment of speech. Its subsequent interpretation mainly depends
on verbal semantics (especially in terms of the telicity or atelicity of a
SoA) and the context in which the verb is integrated. Only with telic
verbs, having an inherent natural end point, can the perfect be satis-
factorily interpreted as a result state. Especially (but not exclusively)
when used with atelic verbs, the perfect, in non-narrative discourse, is
an alternative for the aorist when it comes to referring to past SoAs. In
the latter case, for instance with regard to verba sentiendi, the perfect
may be used as a situating cohesion device.
A speaker may choose for this tense to underline his personal
perception of a relationship between a past SoA and the present com-
municative situation. In this respect the perfect bears a resemblance to
situating (representational) particles in that the SoA is explicitly situated
in the communicative context or considered to be part of it, instead
of being merely referred to (as in the case of the aorist). A speaker
may have several reasons for marking this relationship, which can all
be subsumed under the header actuality: the SoA is connected to the
communicative context for reasons of e.g. evaluation, judgment, and
addressee-oriented appeal.
In a wider perspective this use of the perfect versus the aorist may
perhaps be regarded as another example of the well-known choice
between immediacy and displacement, which, in a word, refers to the
speaker’s ability of assuming several (spatio-temporal) positions towards
the content of his utterance:29 in non-narrative discourse a speaker has
the choice to either involve the completed past SoA in the present and
make it part of the communicative situation by means of the recipro-
cal semantics of the perfect, or to maintain the temporal distance by
using the unidirectional referring of the aorist. Needless to say, more
research on this interesting topic is needed.

29
Following the research tradition after Chafe (1994).
 CHAPTER TWELVE

DISCOURSE COHESION IN THE PROEM OF

HESIOD’S THEOGONY

Albert Rijksbaron

Few pieces of Greek literature have sparked so much controversy as the

proem of Hesiod’s Theogony (lines 1–115). Among the controversial
issues are: the identity of the Muses, the strangely selective catalogue
of gods at lines 11–21, and the temporal and spatial coordinates, so
to speak, of Hesiod’s encounter with the Muses and especially those
of the movements of the Muses during the proem. It is the temporal
and spatial side of the proem that will be the subject of this paper. In
1972, W.J. Verdenius published a series of notes on the proem of the
Theogony, which he concluded with the following verdict: ‘The unity
of the proem does not lie in the interdependency of its parts, but in
the continuity of its progress’ (Verdenius 1972: 260). In what follows I
will argue that Verdenius rightly viewed the continuity of the progress
of the proem as a unifying principle, but underestimated the unifying
force of the interdependency of its parts.1

The Theogony sets off with what is often regarded as a quaestio maior,
whose status is formulated by Thalmann as follows:
Explanations of the different verb tenses in Theog. 1–10 have been sug-
gested that rival in their complexity those given for the similar mixture
in h. Apol. 1–13. This elaborateness by itself makes them suspect. Any
such explanation, to be convincing, would have to account for similar
inconsistencies in other passages cited by West (1966), 155, on Theog. 6
(i.e. h. 19.10–15, 27–29; h. Aphr. 260–61). His solution to the problem,
that the verbs in all these passages are timeless, is more economical and
more plausible than its convoluted rivals. Verdenius takes issue with West
on this point; but his own explanation, (1972, 227 on Theog. 5)—that the

1
As is also shown by Thalmann in his, on the whole convincing, analysis of the
proem (1984: 134–50). For the role of devices like ring composition I may refer to these
pages. At the end of this paper the full text of the proem (lines 1–115) is presented.
242 albert rijksbaron

narrative of a particular event begins at line 5 and continues through line

35—ignores ποτε (l. 22) and drives him to an untenable view (p. 249) of
τότε in line 68. (Thalmann 1986: 227, fn. 9)
I will now review the relevant tense forms and adverbs in the form of
a running commentary, following Verdenius in this respect, too. In the
process I will also discuss some other points that perhaps deserve further
or new comments. Present indicatives are printed in italics, imperfects
in bold type, and aorist indicatives in bold italic type.
Μουσάων ῾Ελικωνιάδων ἀρχώμεθ’ ἀείδειν,
αἵ θ’ ῾Ελικῶνος ἔχουσιν ὄρος μέγα τε ζάθεόν τε,
καί τε περὶ κρήνην ἰοειδέα πόσσ’ ἁπαλοῖσιν
ὀρχεῦνται καὶ βωμὸν ἐρισθενέος Κρονίωνος
5 καί τε λοεσσάμεναι τέρενα χρόα Περμησσοῖο
ἠ’ ῞Ιππου κρήνης ἠ’ ᾿Ολμειοῦ ζαθέοιο
ἀκροτάτῳ ῾Ελικῶνι χοροὺς ἐνεποιήσαντο,
καλοὺς ἱμερόεντας, ἐπερρώσαντο δὲ ποσσίν.
ἔνθεν ἀπορνύμεναι κεκαλυμμέναι ἠέρι πολλῷ
10 ἐννύχιαι στεῖχον περικαλλέα ὄσσαν ἱεῖσαι2

1 Μουσάων (. . .) ἀρχώμεθ’ ἀείδειν

The opening line of the Theogony makes us expect that the poet is about
to start singing,3 making the Muses the starting-point of his song. This
expectation, however, is only partially fulfilled, for after seven lines on
some of the characteristic activities of the Muses, the poet first explains
how he is able to undertake such an enterprise at all, and why the Muses
are his starting-point (lines 9–34). It will turn out that the self-hortatory
ἀρχώμεθα is the consequence of an order given to Hesiod by the Muses
themselves (ἐκέλοντο, line 33).4 After he has given the explanation he
makes a new start at line 36, where his Muses-song really begins.

2
‘Let us begin to sing from the Heliconian Muses, who possess the great and holy
mountain of Helicon and dance on thier soft feet around the violet-dark fountain
and the altar of Cronos’ mighty son. And after they have washed their tender skin in
Permessus or Hippocrene or holy Olmeius, they perform choral dances on highest
Helicon, beautiful, lovely ones, and move nimbly with their feet. Starting out from there,
shrouded in thich invisibility, by night they walked, sending forth their very beautiful
voice (. . .)’ (translation Most, except for στεῖχον in line 10, see below).
3
Temporally, the hortative subjunctive is uttered at a fictional moment of speaking,
which creates an implicit ‘now’. For a similar use of the first person future indicative
see Pfeijffer (1999: 19–20). See also Furley & Bremer (2001: 51–2) on self-exhortations
in hymns.
4
Verdenius wrongly says that ἀρχώμεθα ‘is a traditional formula used at the
beginning of a hymn.’ On the contrary, it is anything but traditional. He refers to h.
Dem. 1, but there we find ἄρχομ’ ἀείδειν, which is, indeed, a formula, cp. h. 9.8, 11.1,
 discourse cohesion in the proem of hesiod’s theogony 243

ἀρχώμεθα
The poet may have preferred the plural to the singular to involve the
audience right from the start in his enterprise.5 For this use of the plural
cp. Chantraine (1963: 34) on Od. 1.10 (to the Muse) (. . .) εἰπὲ καὶ ἡμῖν:
‘(. . .) le poète parle de lui-même en s’associant à ses auditeurs.’ Compare
a speaker starting a lecture with: ‘Ladies and gentlemen, let’s talk about
the future of the classics,’ where the talking of course will be done just
by the speaker. But see also below, on ἀρχώμεθα at line 36.

2 ἔχουσιν, 4 ὀρχεῦνται
‘Omnitemporal’ (Faulkner 2005: 66), rather than ‘timeless’ (e.g. Thal-
mann following West, see above), ‘eternal’ (Stoddard 2004: 130) states
of affairs. ‘Timeless’ states of affairs are ‘outside time altogether’ (Lyons
1977: 680), and are confined to mathematical truths like ‘two times five
equals ten’, or, in Greek, τὰ δὶς πέντε δέκα ἐστίν. As for ‘omnitem-
poral’, Lyons uses this term for situations which are ‘(. . .) time-bound
but temporally unrestricted’ (Lyons 1977: 680).6 This applies in our

13.1, 16.1, 22.1, 26.1, 28.1. The hortative subjunctive makes Th. 1 crucially different.
(Μουσάων ἄρχωμαι, at h. 25.1, is, just like the rest of this short hymn, a clear imitation of
Th. 1). According to Pucci (2007: 33), ἀρχώμεθα expresses the idea that ‘il cantore . . .
si mette in prima linea con una insolita enfasi sul proprio “io” precedendo le Muse.’
This ignores the reappearance of ἀρχώμεθα at line 36; see there.
5
For the role of the audience, and for other characteristics of hymns see Daniełewicz
(1976: 119), as quoted by Furley & Bremer (2001: 59): ‘The specific character of the
hymn [as a type—AR] is to be seen in the simultaneous existence of two communica-
tive settings, the first of which (the author/performer → the formal addressee, viz. the
god) is supplemented by a new one: the author performer → the real recipient, viz.
the listeners. The latter setting enables the poet to deal with the theme of the utter-
ance from a distance: to begin and conclude with a meta-textual formula, to treat the
addressed god, until the moment of salutation, as an object of description or narration,
to emphasize the presence of the performer, and to imply the virtual audience.’ All
these elements are present, at various places, in the proem of Th., and yet the proem is
entirely sui generis, since not only the god(s) but also the poet himself is ‘an object of
narration’, viz. in the narrative about his encounter with the Muses. The elements are:
line 1 meta-textual formula, emphasizing the presence of a performer and implying an
audience; lines 2–8 description passing into narrative (see below); lines 9–34 personal
narrative: Hesiod’s encounter with the Muses, perception of their songs, Dichterweihe;
lines 35–36 meta-textual formulas; lines 37–67 description of the Muses’ activities on
Olympus, interrupted by a narrative digression at lines 53–60: Mousogony; lines 68–71
and lines 75–9 personal narrative, continued: Hesiod and the Muses; lines 71–4 descrip-
tion of Zeus; lines 80–103 description of the Muses’ activities, through their servants,
the singers, among mortals; lines 104–115 salutation and prayer.
6
Omnitemporal states of affairs always involve iteration or habituality. There
are also temporally restricted habitual states of affairs, like ‘The Royal Ballet dances
at the Royal Opera House’. If the habitual dancing of the Muses lasts longer, so to
speak, than that of mortals, this is due to the characteristics of the subject and of the
244 albert rijksbaron

case: as long as there are Muses—this may change, of course—they are

the masters of Mount Helicon and dance there.7 Note that these two
descriptive omnitemporal statements occur in a relative clause, which
means that we are dealing with a so-called ‘attributive section’, i.e. a
section in which one or more fixed attributes of a god are mentioned.8
They are due to an omniscient narrator and must represent impersonal,
conventional knowledge.

ἔχουσιν
= ‘possess’ (Lombardo, Most), not ‘bewohnen’ (Marg). Compare lines
108 and 113 below: εἴπατε δ’ ὡς (. . .) τὰ πρῶτα πολύπτυχον ἔσχον
῎Ολυμπον (‘took possession of ’, Most). ‘Live’ is rather expressed by
οἰκία or δώματ’ ἔχειν and similar expressions, like ἔνθά σφιν λιπαροί
τε χοροὶ καὶ δώματα καλά (line 63). The Muses did indeed not live on
Mount Helicon, but on Olympus (lines 62–63): Ολύμπου· / ἔνθά σφιν
λιπαροί τε χοροὶ καὶ δώματα καλά.

5–9 καί τε (. . .) χοροὺς ἐνεποιήσαντο, ἐπερρώσαντο δὲ ποσσίν

Perhaps I may quote Ruijgh (1971: 900) here: ‘L’emploi de τε adver-
bial invite à considérer le fait exprimé comme permanent (aoriste
gnomique). Comme à priori, l’indicatif II de l’aoriste peut également
exprimer un fait du passé, il y a une transition graduelle vers la phrase
suivante (Th. 9 sqq.), qui exprime nettement un fait du passé (στεῖχον).
Ainsi, le poète prépare le récit de sa propre expérience, c’est-à-dire de
sa rencontre avec les Muses (Th. 22 sqq.).’ The aorists start, then, as
elements of an attributive section but must eventually be reinterpreted
as elements of the narrative that is initiated by στεῖχον (see further
below). For the sequence (omnitemporal) present indicative: (gnomic)
aorist indicative cp. e.g. Hdt. 7.10ε: φιλέει γὰρ ὁ θεὸς τὰ ὑπερέχοντα
πάντα κολούειν. οὕτω δὲ καὶ στρατὸς πολλὸς ὑπὸ ὀλίγου διαφθείρεται
κατὰ τοιόνδε· ἐπεάν σφι ὁ θεὸς φθονήσας φόβον ἐμβάλῃ ἢ βροντήν,
δι’ ὦν ἐφθάρησαν ἀναξίως ἑωυτῶν.

locative phrase; it is not formally marked on the verb (although in the epic language
the omnitemporal interpretation of a present indicative may be indicated by the pres-
ence of epic τε, see Ruijgh (1971: chapter 1)).
7
Clay (1988: 324) writes: ‘Hesiod does not tell us what the Muses sing on the peak
of Helicon.’ But they just dance (ὀρχεῦνται) there, there is no verb of singing in these
lines. The singing occurs during their descents from Helicon, see below.
8
For attributive sections see Bakker (2002: 66), Depew (2004: 129), and Faulkner
(2005: 60–1).
 discourse cohesion in the proem of hesiod’s theogony 245

9–10 ἔνθεν ἀπορνύμεναι κεκαλυμμέναι ἠέρι πολλῷ

ἐννύχιαι στεῖχον περικαλλέα ὄσσαν ἱεῖσαι

9 ἔνθεν
Not ‘puis’ (Mazon) but ‘from there’ (Evelyn-White, Most, Marg and
others), i.e. from the top, hence their being covered by ἀήρ.

10 στεῖχον
After the omnitemporal present indicatives of lines 2–4, and the
‘gnomic-cum-past’ aorist indicatives of lines 7–8, στεῖχον expresses
‘nettement un fait du passé’ (see Ruijgh as quoted above at lines 5–9).
More specifically, στεῖχον should be taken as a ‘focalising’ imperfect,
that is, an imperfect which presents a certain state of affairs from the
point of view of a character rather than that of the narrator.9 Who is
this character? At lines 22–3 the narrator mentions Hesiod (αἵ νύ ποθ’
῾Ησίοδον καλὴν ἐδίδαξαν ἀοιδήν, ἄρνας ποιμαίνονθ’ ῾Ελικῶνος ὕπο
ζαθέοιο), from which we must infer that he was the character perceiving
the στείχειν of the Muses. Observe that the Muses moved περικαλλέα
ὄσσαν ἱεῖσαι: they could be heard while they were moving. Next, this
perceiving character turns out to be none other than the narrator
himself: με (. . .) θεαὶ πρὸς μῦθον ἔειπον, line 24: here, the third person
narrative of lines 10–23 becomes a first person narrative and thus the
report of a personal experience.
The focalising use of the imperfect after a general description is not
always recognized by scholars. An illuminating example from Homer
occurs at the end of the passage Od. 7.104–32: (δμῳαὶ) (. . .) ἀλετρεύουσι
(. . .) ὑφόωσι καὶ (. . .) στρωφῶσιν (. . .) ἀπολείβεται (. . .) (line 113)
ἐλήλαται (. . .) πεφύκασι (. . .)· οὔ ποτ’ ἀπόλλυται οὐδ’ ἀπολείπει (. . .)
τὰ μὲν φύει, ἄλλα δὲ πέσσει. (line 120) (. . .) γηράσκει, (. . .) ἐρρίζωται,
(. . .) τέρσεται (. . .) τρυγόωσιν, (line 125) (. . .) τραπέουσι· (. . .) εἰσιν (. . .)
ὑποπερκάζουσιν. (. . .) πεφύασιν, which is followed and concluded by

ἐν δὲ δύω κρῆναι ἡ μέν τ’ ἀνὰ κῆπον ἅπαντα

130 σκίδναται, ἡ δ’ ἑτέρωθεν ὑπ’ αὐλῆς οὐδὸν ἵησι
πρὸς δόμον ὑψηλόν, ὅθεν ὑδρεύοντο πολῖται.
τοῖ’ ἄρ’ ἐν ᾿Αλκινόοιο θεῶν ἔσαν ἀγλαὰ δῶρα10

9
Smith (2002: 71) uses the term ‘inferred perception’ for this phenomenon, and
calls the sentences involved ‘perspectival’. See also Bakker (1997: 40) on the remark-
able combination, at Thuc. 7.70.7, of the imperfect ἐγίγνετο with νῦν, expressing ‘the
perception and experience of the crews on the ships.’
10
Cp. De Jong (2001: 176–7) on Od. 7.81–135.
246 albert rijksbaron

Here ὑδρεύοντο and τοῖ(α) ἔσαν are presented from the standpoint of
Odysseus there and then on the treshold of the palace.11 Some prose
examples are: Xen. An. 6.4.1–6: ηὐλίζοντο ἐπὶ τοῦ αἰγιαλοῦ πρὸς τῷ
λιμένι. τὸ δὲ χωρίον τοῦτο (. . .) ἔστι (. . .) καλεῖται (. . .) ἔστι (. . .) ἐστὶν
(. . .) ἐστιν (. . .) λέγονται (. . .) κεῖται (. . .) ἔστι δ’ (. . .) τὸ δὲ ὄρος (. . .)
ἀνήκει (. . .) φέρει γὰρ ἡ γῆ (. . .) ἡ μὲν χώρα ἦν τοιαύτη. ἐσκήνουν δ’ (. . .).
Here, a geographical description presenting conventional knowledge
in omnitemporal present indicatives is closed by an imperfect plus an
evaluative adjective (cp. τοῖα at Od. 7.132), whereby the whole preceding
description is turned into something that was perceived by the soldiers
camping there; Xen. An. 1.5.6 (. . .) ὁ δὲ σίγλος δύναται ἑπτὰ ὀβολοὺς
καὶ ἡμιωβέλιον ᾿Αττικούς· ἡ δὲ καπίθη δύο χοίνικας ᾿Αττικὰς ἐχώρει
(the soldiers found out that this was the contents of the kapithê); Xen.
An. 1.5.3 τὰς δὲ ὠτίδας ἄν τις ταχὺ ἀνιστῇ ἔστι λαμβάνειν· πέτονται
γὰρ βραχὺ ὥσπερ πέρδικες καὶ ταχὺ ἀπαγορεύουσι. This omnitemporal,
habitual, description of the behaviour of the bustards whenever someone
hunts them is followed by τὰ δὲ κρέα αὐτῶν ἥδιστα ἦν—for the soldiers
there and then.12 Also in a narrative, after an aorist indicative: X. HG
2.1.21 ἔπλευσαν εἰς Αἰγὸς ποταμοὺς ἀντίον τῆς Λαμψάκου· διεῖχε δ’ ὁ
῾Ελλήσποντος ταύτῃ σταδίους ὡς πεντεκαίδεκα. I have discussed this
use of the imperfect and further examples in Rijksbaron (1995), with
literature. For a rather spectacular instance of such an imperfect (not

11
In the brilliantly structured passage 81–132—called ‘inorganic’ by Heubeck
et al.—the focalising imperfect is already present before. There is a gradual movement
of the camera (which is Odysseus) from the outside to the inside of the palace of Alci-
nous. Odysseus was approaching (ἴε, line 82) the palace, and he pondered (ὅρμαινε)
many things while he was busy finding a place to stand (ἱσταμένῳ, line 82), until he
arrived (ἱκέσθαι, line 83) at the treshold. Next, lines 84–102 describe, in imperfects and
pluperfects, the things Odysseus pondered, things, we may infer, seen by him during
the time of his ἵστασθαι: first the walls ἐς μυχὸν ἐξ οὐδοῦ and elements belonging to
the walls (lines 84–94), and then the space enclosed by the walls as far as he could
see it, i.e. the courtyard (lines 95–102; notice line 95 ἐν δὲ θρόνοι, and cp. ὑπ’ αὐλῆς
οὐδόν at line 130). Then what is going on inside the palace itself, and what is situated
outside the αὐλή (line 112 ἔκτοσθεν αὐλῆς), viz. the orchard, neither of which can be
seen by Odysseus, are told by the omniscient narrator in omnitemporal present (and
perfect) indicatives (lines 103–31 ὑψηλόν). Finally, at line 133 the narrative is resumed;
ἔνθα (στάς) refers back to the οὐδὸν introduced in l. 83, while θηεῖτο expresses the
idea that once Odysseus stands on the threshold he is no longer pondering but full
of admiration. (De Jong (2001: 176) wrongly thinks that Odysseus steps over the
threshold in line 83.)
12
The imperfect may also appear before the omnitemporal present, as in Od. 3.291–3
ἔνθα διατμήξας τὰς μὲν Κρήτῃ ἐπέλασσεν, / ἧχι Κύδωνες ἔναιον ᾿Ιαρδάνου ἀμφὶ ῥέεθρα.
/ ἔστι δέ τις λισσὴ αἰπεῖά τε εἰς ἅλα πέτρη -ἔναιον, because perceived by Odysseus.
 discourse cohesion in the proem of hesiod’s theogony 247

discussed in the article just mentioned), which illustrates its function

very well, see the Appendix.
Στεῖχον at Theogony line 10 should be taken, then, in the same way,
the human perceiver being presupposed by the presence of περικαλλέα
ὄσσαν ἱεῖσαι, / ὑμνεῦσαι Δία etc. The words sung by the Muses in lines
11–21 are therefore heard by the poet.13 Note that the non-verbal sound
implied by the movement of the feet of the Muses (ἐπερρώσαντο (. . .)
ποσίν in line 8) may also have been heard by the poet, and may there-
fore prepare the ground for the meaningful sounds mentioned in lines
10–21. I think, moreover, that στεῖχον has iterative meaning: after the
impersonal, conventional information on the omnitemporal activities
of the Muses in lines 2–8 we are now dealing with a habitual activity
of the Muses in the past, in Hesiod’s past, that is, as witnessed (aurally)
by Hesiod.14 In other words, with στεῖχον we enter a narrative.15 See
the next note, on ποτε.

13
For the idea cp. A.R. 4.1381 Μουσάων ὅδε μῦθος, ἐγὼ δ’ ὑπακουὸς ἀείδω /
Πιερίδων.
14
West, in his commentary ad locum, considered στεῖχον a ‘typifying’, that is to
say a timeless, imperfect, which temporally would not differ from the present and aor-
ist indicatives of lines 2–8 (likewise already e.g. Evelyn-White, Loeb (1914): ‘Thence
they arise and go abroad by night . . .’). I will refrain from discussing this bizarre view,
and I can do this the more easily because West himself, in a much later publication
(West 1989), without discussing his earlier view, abandoned it, only to exchange it for
another untenable view. In West (1989) he considers στεῖχον an instance of a non-
past, so-called ‘injunctive’, use of augmentless forms with secondary endings. But this
injunctive use does not exist. Cp. Beekes (1995: 245): ‘In Greek the injunctive forms
are completely equal to the indicative forms (. . .); it thus has no sense to speak here
of an injunctive (because there is no meaning component). Only in Sanskrit can one
speak of an injunctive.’ Remarkably, West never tells us what is wrong with taking the
imperfect as an imperfect.—It will be recalled that according to Thalmann the use of
the tenses in Th. 1–10 should be explained in connection with ‘similar inconsistencies’
in other passages, notably h. 3.1–13, 19.10–15, 27–29; h. Aphr. 260–1. Why? The phrase
‘similar inconsistencies’ begs the question, of course; above I argued (with others) that
the tenses in Th. 1–10 are not used inconsistently at all but in accordance with the
normal rules of Greek. Which means that any inconsistencies that might be found in
the hymns—one may think notably of the imperfect μίμνε at h. Ap. 5, which is definitely
unlike στεῖχον at Th. 10—may, as a matter of principle, be sui generis, and demand a
different explanation from that of the tenses in Th. 1–10. For a recent, ingenious but
not quite convincing, analysis of the beginning of h. Ap. see Bakker (2002).
15
Cp. Clay (2003: 54): ‘In effect, the imperfect στεῖχον conveys precisely and viv-
idly the moment of transition from the eternal time of the gods to the temporality of
mankind.’ We are not dealing with ‘a moment’, however. That στεῖχον rather expresses
a habitual activity, as claimed above, is an effect of ποτε in l. 22: ‘once, on one occa-
sion’ presupposes the existence of a series of activities during which another activity
occurred. For this effect of ποτε cp. the sequence ἐφύλασσε (. . .) κοτε (. . .) διέφθειραν
at Hdt. 9.93.2: ἔνθα δὴ τότε ὁ Εὐήνιος οὗτος ἀραιρημένος ἐφύλασσε· καί κοτε αὐτοῦ
248 albert rijksbaron

22 αἵ νύ ποθ’ ῾Ησίοδον καλὴν ἐδίδαξαν ἀοιδήν

νυ
Νυ stresses the idea that the διδάξαι of the Muses did actually happen
as it is described in this sentence. Cp. Ruijgh (1971: 842): ‘(. . .) elle
[the particle νυ] signale que le fait exprimé par la phrase a la même
importance qu’un fait actuel.’

ποθ’
Cp. Thalmann (1984: 136): ‘The Muses’ gift of song to Hesiod was a
single instance, which occurred at a particular moment (ποτε, ‘once,’
line 22), of their often-repeated activities on the heights of Helikon
(lines 1–21).’ This is correct, but for the reference to lines 1–21 as a
whole; Thalmann wrongly followed West in assuming that all verbs
in lines 1–21 are timeless. Incidentally, how could the Muses possibly
have met Hesiod in the human world if their often-repeated activities
are timeless, i.e. outside time altogether? For this encounter to happen
it is necessary that they performed an activity in (and not outside) the
human world, and this activity is provided for by στεῖχον. The encounter
of the Muses with Hesiod occurred, indeed, at a particular moment,
namely during one of their descents from Mount Helicon.16

ἐδίδαξαν
A complexive aorist, which denotes a state of affairs (‘ “bestow an abil-
ity” rather than “teach” ’—Verdenius) that occurs within the frame-
work created by στεῖχον and comprises ἔειπον (line 24), ἔδον (line
30) and ἐνέπνευσαν (line 31).17 For the idea cp. Od. 8.480–1 πᾶσι γὰρ

κατακοιμίσαντος τὴν φυλακὴν παρελθόντες λύκοι ἐς τὸ ἄντρον διέφθειραν τῶν

προβάτων ὡς ἑξήκοντα. The first verb may also be explicitly marked as habitual, as
at Hes. Op. 633–5 ὥς περ ἐμός τε πατὴρ καὶ σός, μέγα νήπιε Πέρση, / πλωίζεσκ’ ἐν
νηυσί, βίου κεχρημένος ἐσθλοῦ· / ὅς ποτε καὶ τεῖδ’ ἦλθε πολὺν διὰ πόντον ἀνύσσας.
Pucci (2007: 43–4) confines himself to enumerating the various interpretations of
lines 9–10.
16
In spring or early summer, cp. ἄρνας ποιμαίνοντ(α), line 23 (lambs are typically
born in the spring; not ‘sheep’ (West and others), for this is ποιμαίνοντ’ ἐπ’ ὄεσσι, Il.
11.106), and probably at dawn, for Hesiod must have seen the ‘marvellous branch’
given to him—as he tells us—at lines 30–1. We are not told that he actually saw the
Muses, and must assume that they remained covered by mist, as indicated by the perfect
participle κεκαλυμμέναι (line 9).
17
For the complexive aorist see Kühner-Gerth (1898/1904: 1.155) and Rijksbaron
(2007: 11–2).
 discourse cohesion in the proem of hesiod’s theogony 249

ἀνθρώποισιν ἐπιχθονίοισιν ἀοιδοὶ / (480) τιμῆς ἔμμοροί εἰσι καὶ αἰδοῦς,

οὕνεκ’ ἄρα σφέας / οἴμας Μοῦσ’ ἐδίδαξε, φίλησε δὲ φῦλον ἀοιδῶν,
also 8.488.

24 τόνδε δέ με πρώτιστα θεαὶ πρὸς μῦθον ἔειπον

τόνδε should be connected with μῦθον, not with με (Waltz 1914: 232,
Clay 2003: 52, Bakker 1999: 10, Pucci 2007: 58–9 and others): deictic
pronouns are never combined with unemphatic clitic pronouns. This,
it is true, could be remedied by printing δ’ ἐμέ, but this, too, should
be rejected: the deictic use of ὅδε with first and, rarely, second person
pronouns is not found in narratives but only in direct speech: they
presuppose the presence of an addressee in a face to face situation.
Cp. Il. 19.40 ἐγὼν ὅδε, Od. 1.76 ἡμεῖς οἵδε, 16. 205 ὅδ’ ἐγώ, 21.207 ὅδ’
αὐτὸς ἐγώ, 2.45–46 ἐν ὑμῖν / τοίσδεσσιν, Th. 1.53.2 ἡμᾶς τούσδε. For
further examples, also of other combinations with ὅδε, see Kühner-
Gerth (1898/1904: 1.641–3). For τόνδε (. . .) μῦθον cp. Il. 15.202 τόνδε
φέρω Διὶ μῦθον (. . .) / ἦ τι μεταστρέψεις (. . .).

25 Μοῦσαι ᾿Ολυμπιάδες, κοῦραι Διὸς αἰγιόχοιο·

Probably the Muses are now called ‘Olympian’ because the hymn that
is to follow (lines 37–67) deals with their activities on the Olympus.18

33-4 καί μ’ ἐκέλονθ’ ὑμνεῖν μακάρων γένος αἰὲν ἐόντων,

σφᾶς δ’ αὐτὰς πρῶτόν τε καὶ ὕστατον αἰὲν ἀείδειν.

ἐκέλοντο
After the Muses have provided him with a purpose for his ‘divine voice’
(lines 31–32), they urge Hesiod to apply his newly acquired art to sing
a hymn to the gods collectively (γένος) and especially to the Muses
themselves. By ordering him to ὑμνεῖν μακάρων γένος αἰὲν ἐόντων they
ask him to do among mortals what they themselves do on Olympus:
line 37 ὑμνεῦσαι τέρπουσι (. . .) ἐντὸς ᾿Ολύμπου, further specified at
44 as θεῶν γένος αἰδοῖον πρῶτον κλείουσιν ἀοιδῇ.
As for the imperfect, this creates a certain expectation on the part of
the reader/listener as to the reaction of the addressee, a use of verbs of
saying that was first recognized and discussed by Blass (1889). Thus, at

18
‘There is (. . .) a shift from one mountain to the other as the focus of the Muses’
activities’ (Thalmann 1984: 135).
250 albert rijksbaron

Il. 1.386 αὐτίκ’ ἐγὼ πρῶτος κελόμην θεὸν ἱλάσκεσθαι· / ᾿Ατρεΐωνα δ’

ἔπειτα χόλος λάβεν (. . .), χόλος λάβεν describes Agamemnon’s reaction
to Achilles’ κελόμην. The aorist lacks this feature, as at Il. 6.110, 116:
῎Εκτωρ δὲ Τρώεσσιν ἐκέκλετο μακρὸν ἀΰσας· ‘(. . .)’. / ὡς ἄρα φωνήσας
ἀπέβη κορυθαίολος ῎Εκτωρ. There is no mention of the reaction of the
Trojans. See further Rijksbaron (2007: 18–9).

35 ἀλλὰ τίη μοι ταῦτα περὶ δρῦν ἢ περὶ πέτρην

With this line Hesiod breaks off the narrative about his encounter with
the Muses: the time has come to turn to the matter at hand.19 It will
turn out, however, that the narrative is not yet finished, for the poet
resumes it at line 68, quo vide.

36 τύνη (. . .) ἀρχώμεθα
This remarkable combination has not got the attention it deserves.
West has only a note on the form τύνη. Verdenius observes: ‘Hesiod
addresses himself, not the audience’, rightly, but this is not the only
point to be made.20 In fact, the use of a second person pronoun with a
first person plural hortative subjunctive is, as far as I could see, unique,
and understandably so. For what we have here is a speaker who is giv-
ing an order to himself as if he is somebody else: ‘You <and me>, let us
start singing (. . .).’21 This very daring construction, which is the more
remarkable because in the previous line μοι is present, must suggest
that Hesiod is addressing himself both as the Hesiod of lines 5–34, so
to speak, the Hesiod who was favoured by, and got orders from, the
Muses on Helicon, and as the accomplished singer he is now: the Hesiod
of the earlier lines should stop talking about how the Muses inspired
him, and finally start singing his hymn to them.

19
Whatever the exact meaning of ταῦτα περὶ δρῦν ἢ περὶ πέτρην, it can hardly stand
for speaking about ‘irrelevant matters’ (Most, in a note on p. 5), since it must refer
to the actions of the Muses described in lines 30–4, and may even refer to Hesiod’s
encounter with the Muses as a whole. Perhaps περὶ δρῦν ἢ περὶ πέτρην = ‘things
belonging to the past,’ cp. one of the scholia ad locum (τί μοι χρεία ἐστὶν ἀρχαιολογεῖν)
and West (1966: 168).
20
Pucci (2007: 76) mainly discusses its use in the Iliad. Most translates ‘come then’,
but τύνη is not a hortative particle like ἄγε.
21
To be sure, the ‘we’ of the hortative subjunctive may be subdivided into ‘you’ and
‘me’ but then two different persons are involved, as at Pl. Phdr. 237c6 ἐγὼ οὖν καὶ σὺ
μὴ πάθωμεν ὃ ἄλλοις ἐπιτιμῶμεν.
 discourse cohesion in the proem of hesiod’s theogony 251

τύνη, Μουσάων ἀρχώμεθα, ταί (. . .)

Having explained how he is able to sing, the poet now restarts, so to
speak, the hymn to the Muses and the gods, which was announced
by ἀρχώμεθα in line 1:22 ‘au vers 36, ce “commencement” constitue
clairement l’application directe des recommandations formulées par
les Muses au vers 34’ (Leclerc 1993: 170–1).23 This does not mean that
we have to imagine the poet starting to sing while the Muses are still
on Mount Helicon: the hymn to the Muses is executed within, not
outside, the framework of the proem. To have the latter interpretation
the poet should have continued with ‘and so I began to sing from the
Muses (. . .).’ Temporally, we are here still at the implicit ‘now’ of line
1; the poet really resumes his activity, be it in different circumstances,
for the audience has by now been fully informed about the legitimity
of his undertaking.
With ταὶ Διὶ πατρὶ / ὑμνεῦσαι τέρπουσι (lines 36–7) we leave the
narrative and enter the omnitemporal world of the gods on Olympus, in
another attributive section, which describes the hymnic activities of the
Muses on Olympus and their effect: ὑμνεῦσαι τέρπουσι (line 37), ῥέει
αὐδή (line 39), γελᾷ (. . .) δώματα (line 40), ἠχεῖ (. . .) κάρη (line 42), θεῶν
γένος (. . .) κλείουσιν (line 44), ὑμνεῦσαι τέρπουσι (line 55), μέλπονται
(line 66) and κλείουσιν (line 67). We have already been informed that
knowledge of the world of the gods has been granted to Hesiod by the
Muses themselves, viz. when he overheard them singing of the gods
of lines 11–21. The hymn first runs up to and including θεῶν γένος
κλείουσιν (line 44) plus the genealogical relative clauses (οὓς (. . .) ἔτικτεν,
line 45, οἵ τ’ (. . .) ἐγένοντο, line 46), thereafter up to and including line
52. This is followed by a ‘Mousogony’, a narrative digression within the
attributive section, where both ἐν Πιερίῃ and Κρονίδῃ (. . .) πατρί are
to be connected with μιγεῖσα:24 Mnemosyne was visited by Zeus in her
own abode, where she mingled in love with him as a father. Cp. lines
56–7: ἐννέα γάρ οἱ νύκτας ἐμίσγετο μητίετα Ζεὺς / νόσφιν ἀπ’ ἀθανάτων

22
By the presence of τύνη, ἀρχώμεθα here does not simply pick up the form of line
1, where I argued that it is used to involve the audience in the activity of the poet.
23
Strangely enough Minton seems to take Μουσάων as the addressees of these
words, since he writes (1970: 358) that line 36 is an ‘exhortation to the Muse to cel-
ebrate the god.’
24
Often, ἐν Πιερίῃ is taken with both μιγεῖσα and τέκε, or with τέκε only, but this
is less likely in view of line 62. See below, note 26.
252 albert rijksbaron

ἱερὸν λέχος εἰσαναβαίνων.25 Here, ἐννέα (. . .) νύκτας ἐμίσγετο specifies

the nature of the encounter of Zeus and Mnemosyne, which was first
presented by a complexive aorist (μιγεῖσα), and explains the number of
the Muses who will be born. Lines 60 and 62 conclude the Mousogony;
although the time of the birth of the Muses is not mentioned, the place
is specified: at line 60 ἔτεκ’ picks up the τέκε of line 53, while line 62
adds the place: ἡ δ’ ἔτεκ’ ἐννέα κούρας, ὁμόφρονας (. . .) /, (. . .) / τυτθὸν
ἀπ’ ἀκροτάτης κορυφῆς νιφόεντος ᾿Ολύμπου.26 In between there is a
brief attributive section at lines 60–1 (ᾗσιν ἀοιδὴ / μέμβλεται (. . .)),
which is continued at lines 63–7 (ἔνθα σφιν (. . .)).27
68 αἳ τότ’ ἴσαν πρὸς ῎Ολυμπον, ἀγαλλόμεναι ὀπὶ καλῇ,
ἀμβροσίῃ μολπῇ· περὶ δ’ ἴαχε γαῖα μέλαινα
70 ὑμνεύσαις, ἐρατὸς δὲ ποδῶν ὕπο δοῦπος ὀρώρει28
What does τότε refer to? Most scholars take it to refer to the birth of the
Muses mentioned in line 60. Thus e.g. West on line 68: ‘The first thing
a newborn god does—even if he is born practically on the summit of
Olympus!—is to go and join the other gods’, and likewise Minton (1970:
363): ‘Their dwelling place is just below the peak of snowy Olympus,
and there they join with the Graces and Desire in choral song prais-
ing the gods. Then (emphasis Minton) they proceed up to the halls of

25
Zeus visited Maia in the same way, cp. h. 4.1–7 (Hermes) ὃν τέκε Μαῖα / νύμφη
ἐϋπλόκαμος Διὸς ἐν φιλότητι μιγεῖσα / αἰδοίη· μακάρων δὲ θεῶν ἠλεύαθ’ ὅμιλον /
ἄντρον ἔσω ναίουσα παλίσκιον, ἔνθα Κρονίων / νύμφῃ ἐϋπλοκάμῳ μισγέσκετο νυκτὸς
ἀμολγῷ, / ὄφρα κατὰ γλυκὺς ὕπνος ἔχοι λευκώλενον ῞Ηρην.
26
Can this be reconciled with the information of line 53 if this is taken as ‘Mne-
mosyne bore her children in Pieria’? Or, to put it differently, can Pieria be said to be
located ‘at a small distance from the highest top of the Olympus’? I think not. ‘Pieria
is the region immediately to the north of Olympus, between it and the Haliacmon’
(West). In fact, the nearest it gets to Olympus seems to be as a slope of that mountain,
cp. E. Ba. 409–11 ἁ καλλιστευομένα Πιερία / μούσειος ἕδρα, / σεμνὰ κλιτὺς ᾿Ολύμπου.
Cp. also Eustath. Ad Iliadem 3.623.23, where Pieria is called an ἀκρώρεια (‘ridge’)
᾿Ολύμπου. Nowhere is Pieria mentioned as a region near the summit. I take it, then,
that the Muses were born not in Pieria but near the summit of Olympus, and that they
were begot in Pieria (see on line 36). Another god born on Olympus is Hermes, at
least according to a tradition that is found in Philostr. Im. 1.26.1: (Hermes) τίκτεται
μὲν ἐν κορυφαῖς τοῦ ᾿Ολύμπου, κατ’ αὐτὸ ἄνω τὸ ἕδος τῶν θεῶν, and that probably
goes back to Alc. fr. 2 D. χαῖρε, Κυλλάνας ὀ μέδεις, σὲ γάρ μοι / θῦμος ὔμνην, τὸν
κορύφαισιν † αὐγαῖς † / Μαῖα γέννατο Κρονίδαι μίγεισα / παμβασίληϊ. In the hymn
to Hermes, the god is born in Maia’s cave, cp. line 23, where Hermes, whose birth has
been mentioned in line 13, steps over οὐδὸν (. . .) ἄντροιο.
27
Line 62 can hardly belong to the attributive section, for in that case τυτθὸν (. . .)
᾿Ολύμπου would modify μέμβλεται. But why would the Muses’ care for song be locally
restricted?
28
‘They then went towards Olympus, exulting in their beautiful voice, and around
them the black earth resounded as they sang’ (translation Most, adapted).
 discourse cohesion in the proem of hesiod’s theogony 253

Zeus on the peak of Olympus (. . .)’,29 Thalmann (1984: 137): ‘68–79:

Muses’ procession to Olympos at the time of their birth,’ Clay (2003:
68): ‘Immediately after their birth, the goddesses, singing and dancing,
all the while, made their ascent to their father’ and Pucci (2007: 97): ‘le
Muse andarono a presentarsi al padre loro, Zeus’. Cp. also the transla-
tion of e.g. Bonnafé: ‘Sitôt nées, les voilà qui partaient vers l’Olympe.’
This must be rejected, however, since it yields an impossible situation.
First of all, the Muses, whether their birth is located in Pieria as part
of Olympus (cp. note 26), or, as I prefer, near the summit of Olympus
(cp. again note 26), are already on Olympus, so they cannot go there.
Note that in West’s explanatory note πρὸς ῎Ολυμπον is omitted, while
his ‘to join the gods’ is not in the Greek. In fact, the basic flaw of the
above interpretations is that they ignore line 62. Second, if a newborn
god does go to Olympus, he/she, being an infant, is brought there, as in
two of the parallels adduced by West for his claim that ‘[t]he first thing
a newborn god does (. . .) is to go and join the other gods’, viz.
h. 6.14–6 (subject: the Horai, object: Aphrodite)
αὐτὰρ ἐπεὶ δὴ πάντα περὶ χροῒ κόσμον ἔθηκαν
15 ἦγον ἐς ἀθανάτους· οἱ δ’ ἠσπάζοντο ἰδόντες
χερσί τ’ ἐδεξιόωντο καὶ (. . .).30
After Zephyros has brought Aphrodite to the shores of Cyprus (lines
3–4) and the Horai have taken her over (δέξαντ’, line 6), the latter start
clothing and adorning her (lines 6–13), which is followed by the lines
quoted above. There is no activity of Aphrodite herself: it is the Horai
who bring her to the gods.
h. 19.42 (τόν = Pan)
40 τὸν δ’ αἶψ’ ῾Ερμείας ἐριούνιος εἰς χέρα θῆκε
δεξάμενος, χαῖρεν δὲ νόῳ περιώσια δαίμων.

ῥίμφα δ’ ἐς ἀθανάτων ἕδρας κίε παῖδα καλύψας

δέρμασιν ἐν πυκινοῖσιν ὀρεσκῴοιο λαγωοῦ·
πὰρ δὲ Ζηνὶ καθῖζε καὶ ἄλλοις ἀθανάτοισιν,
45 δεῖξε δὲ κοῦρον ἑόν·31

29
Minton’s idea that line 68 belongs to the same time sphere as lines 60–7 (see his
‘Then they proceed’), as if there is no imperfect (or did he take this as timeless?), is
not supported by the text.
30
‘When they had wholly arrayed the goddess in splendor, / They led her to the
immortals, who, seeing her, welcomed her kindly / and offered their right hands in
greeting . . .’ (translation Sargent).
31
‘But luck-bringing Hermes, receiving his son, at once / Took him into his arms, his
heart filled with joy beyond measure, / And swiftly he went to the seat of the undying
254 albert rijksbaron

In this case, too, the newborn god is not active himself, cp. lines 42 and
45: it is Hermes who brings Pan to the assembly of the gods.
If, therefore, Th. 60–8 would be about the Muses joining the gods
immediately after their birth, one would expect them to be brought
there as well, rather than to go ( ἴσαν, 68) there all by themselves, let
alone ‘singing and dancing, all the while’.32
Now to return to τότ’ ἴσαν, what does τότε refer to, then? I take it that
the imperfect ἴσαν, the first imperfect having the Muses as its subject
after στεῖχον at line 10, is the sign that the poet resumes the narrative
about his encounter with the Muses, which he had broken off for a
hymn to them, and that τότε refers back to the form which concluded
the first part of the narrative, i.e. ἐδίδαξαν (line 22).33 After they have
bestowed ἀοιδή on Hesiod their mission on Helicon is completed,
and, there being no more need for their στείχειν on its slopes, they
now go (= return) to the Olympus, where their δώματα are (line 63).
For in principle νίσομαι (νισομένων 72), a variant of νέομαι, does not
mean ‘go’ but ‘return’. Cp. Chantraine Dict. Etym. s.v. νέομαι: ‘«reve-
nir, retourner» (notamment chez soi), rarement aus sens de «venir»’,
Lexikon des frühgr. Epos s.v. νίσομαι: ‘ “zurückkehren”, s.v. νέομαι:
“heimkehren.” ’34 (LSJ: ‘go, come’, wrongly).

gods / With the child warmly wrapped in the skin of mountain-bred hares, / And he
sat down beside Zeus and the other immortals, / And presented his son;’ (translation
Sargent, with corrections in lines 44–5).
32
Only newborn Hermes does not behave like a baby, see the elaborate description
of the first day of his life at h. 4.20–102, which was explicitly announced in lines 13–8:
καὶ τότ’ ἐγείνατο παῖδα πολύτροπον, αἱμυλομήτην, / ληϊστῆρ’, ἐλατῆρα βοῶν, ἡγήτορ’
ὀνείρων, / νυκτὸς ὀπωπητῆρα, πυληδόκον, ὃς τάχ’ ἔμελλεν / ἀμφανέειν κλυτὰ ἔργα μετ’
ἀθανάτοισι θεοῖσιν. / ἠῷος γεγονὼς μέσῳ ἤματι ἐγκιθάριζεν, / ἑσπέριος βοῦς κλέψεν
ἑκηβόλου ᾿Απόλλωνος, / τετράδι τῇ προτέρῃ τῇ μιν τέκε πότνια Μαῖα. Nothing of
the kind is found in our passage from the Theogony. Incidentally, ‘the first thing a
newborn god does’ is in Hermes’ case definitely not to go to Olympus. As for the other
passage mentioned by West (Th. 201–2), this is not about a first voyage to Olympus
either: τῇ δ’ ῎Ερος ὡμάρτησε καὶ ῞Ιμερος ἔσπετο καλὸς / γεινομένῃ τὰ πρῶτα θεῶν τ’
ἐς φῦλον ἰούσῃ. Here we are just told that Eros accompanied, and Himeros followed,
Aphrodite ‘as soon as she was born and when she went to the gods’, not necessarily,
then, immediately after her birth; τὰ πρῶτα must be connected with γεινομένῃ only,
cp. Il. 6.489, Od. 8.553 ἐπὴν τὰ πρῶτα γένηται. In fact, we had already been told that
the first thing Aphrodite did after she had grown in the foam of the sea (lines 191–2)
was to go to Cythera (lines 192–3), whereafter she went to Cyprus (line 193).
33
According to Verdenius τότε refers to στεῖχον (line 10), but this view is less likely,
since it ignores the presence of ἐδίδαξαν.
34
The lemma in the LfgrE lacks precision, for it is not clear whether νισομένων at
Th. 71 is viewed as an instance of zurückkehren, or of hin(und her)gehen/fahren. Also,
the lemma wrongly claims (top of col. 407) that the Muses go ‘durch die Luft’.
 discourse cohesion in the proem of hesiod’s theogony 255

A similar use of τότε is found at Od. 15.256–7 τοῦ μὲν ἄρ’ υἱὸς
ἐπῆλθε, Θεοκλύμενος δ’ ὄνομ’ ἦεν, / ὃς τότε Τηλεμάχου πέλας ἵστατο,
which comes after a long genealogical digression about the descent of
Theoclymenus. Here, τότε ultimately refers back, via ἐπῆλθε at 256,
to ἤλυθεν at 223, a distance of 34 lines.35 Naturally, our case differs
from Od. 15.256–7 in that in the Theogony, if τότε is meant to refer
back to ἐδίδαξαν at line 22, it has to cross the barrier, so to speak, of
line 35 ἀλλὰ τίη μοι ταῦτα περὶ δρῦν ἢ περὶ πέτρην. Irene de Jong
(p.c.) considers this barrier unsurmountable. She points out that, in
the Iliad, the comparable formula ἀλλὰ τίη μοι ταῦτα φίλος διελέξατο
θυμός (. . .) is used five times to put an end to some line of thought
for good (11.407, 17.97, 21.562, 22.122, 22.385), and that, in Pindar,
too, after a so-called Abbruchsformel (e.g. N. 3.26) the poet never
returns to the matter presented before this formula. This may be true,
but ignores the fact that the proem of the Theogony, presenting as it
does a mixture of hymn and personal narrative, is quite different from
both the soliloquies and speeches in the Iliad and the Pindaric odes.
As a matter of fact, as I argued already above, the proem is entirely
sui generis.

The movements of the Muses at Th. 68–70 are still witnessed by the poet,
cp. the presence of ὀπί (line 68), ἴαχε (line 69), ὑμνεύσαις and ἐρατὸς
δὲ ποδῶν ὕπο δοῦπος ὀρώρει (70), which presuppose the presence of
a human perceiver, and continue the situation of lines 10–1 στεῖχον
περικαλλέα ὄσσαν ἱεῖσαι, / ὑμνεῦσαι Δία τ’ αἰγίοχον καὶ πότνιαν ῞Ηρην
etc. In the same vein ἄειδον (line 75) refers to an action that is perceived
by the poet as well. This entails that ταῦτ’ (. . .) ἄειδον turns lines 71–4
into songs heard by Hesiod, while ultimately ταῦτα also comprises the
song of the Muses at lines 11–21.36

35
ἐπῆλθε at line 256 merely repeats (σχεδόθεν (. . .)) ἤλυθεν (line 223), there is no
temporal progression. This entails that to know the precise reference point of τότε the
hearer/reader has to go back 34 lines. One may also compare Od. 7.133 (quoted in note
11), where ἔνθα resumes the narrative after 49 lines of focalised and omnitemporal
descriptions. For a still greater distance spanned by τότε see αἷς ἐχειμάσθην τότε, spoken
by the Guard, at S. Ant. 391, which spans a distance of some 80 lines, since it refers
back to lines 305–14, where Creon speaks threateningly to the Guard.
36
For a similar wide-ranging ταῦτα with a form of ἀείδειν see Od. 8.367, where
ταῦτ’ ἄρ’ ἀοιδὸς ἄειδε summarizes the whole of lines 267–366, that were introduced
by αὐτὰρ ὁ φορμίζων ἀνεβάλλετο καλὸν ἀείδειν at 266.
256 albert rijksbaron

71 ἐμβασιλεύει
The singing of the Muses about the kingship of Zeus on Olympus men-
tioned in this brief attributive section prepares us for the special bond
of the Muses with earthly kings, who come to the fore in lines 80–92.

83–4 τῷ μὲν ἐπὶ γλώσσῃ γλυκερὴν χείουσιν ἐέρσην,

τοῦ δ’ ἔπε’ ἐκ στόματος ῥεῖ μείλιχα·
These words echo lines 39–40: (τῶν = the Muses) τῶν δ’ ἀκάματος ῥέει
αὐδὴ / ἐκ στομάτων ἡδεῖα, naturally without ἀκάματος: mortal kings
(and singers, see line 97) are not infatigable.

93 τοίη Μουσάων ἱερὴ δόσις ἀνθρώποισιν

Here τοίη (. . .) δόσις summarizes the preceding attributive section (lines
79–92) on Calliope as a whole, while δόσις refers back specifically to
line 83 τῷ μὲν ἐπὶ γλώσσῃ γλυκερὴν χείουσιν ἐέρσην—and ultimately
echoes ἔδον, line 30—and generic ἀνθρώποισιν expresses the idea that
eventually the gift of the Muses was not just meant for the kings, but for
all mankind, through the services of the kings as well as of the singers.
The latter are introduced through a paratactic analogon (lines 94–6):
while the kings are the offspring of Zeus, as was implied by lines 81–2,
the singers are the offspring of the Muses and Apollo. For this rhetorical
figure, which mostly involves the use of μέν (. . .) δέ, see the discussion
in Kühner-Gerth (1898/1904: 2.232–3), and Denniston (1954: 370–1).
A typical example is Il. 1.163–4, mentioned by Kühner-Gerth: (ὡς ἔμ’
ἀφαιρεῖται Χρυσηΐδα Φοῖβος ᾿Απόλλων, / ) τὴν μὲν ἐγὼ σὺν νηΐ τ’
ἐμῇ καὶ ἐμοῖς ἑτάροισι / πέμψω, ἐγὼ δέ κ’ ἄγω Βρισηΐδα καλλιπάρῃον
‘so werde ich, während ich diese (. . .) schicken werde, die Bris. in das
Zelt führen.’37 At Th. 94–6 we are dealing with the rather rare variant
in which it is the first clause that ‘carries the weight’ (Denniston 1954:
370); Denniston mentions as an example E. Cy. 199, where while μέν
is lacking τοι is present.

94 ἐκ γάρ τοι (. . .)
τοι appeals to Hesiod’s addressee in the human world, who was already
implied by ἀνθρώποισιν (line 93). Cp. Denniston (1954: 537): ‘(. . .) τοι
implies (. . .) an audience (. . .).’ See also note 4.

37
Μέν may also be omitted, for which see Kühner-Gerth (1898/1904: 2. 233) and
Denniston (1954: 165): ‘[μέν is sometimes omitted] (. . .) even when the close connexion
between coordinated clauses implies logical subordination.’
 discourse cohesion in the proem of hesiod’s theogony 257

96–7 (. . .) ὁ δ’ ὄλβιος, ὅντινα Μοῦσαι

φίλωνται· γλυκερή οἱ ἀπὸ στόματος ῥέει αὐδή.
The position of this sentence after ἐκ δὲ Διὸς βασιλῆες might suggest
that the poet is speaking about the kings. Actually, however, he must
be speaking about the singers (or rather singer, see below), who were
the most important element of the preceding sentence. And, in fact,
the ‘rulers [have] already been treated’ (Thalmann 228–229, fn. 16),
viz. at lines 81–92. Observe that the phrasing here differs from that
at line 81: there we read, in connection with kings: ὅντινα τιμήσουσι
Διὸς κοῦραι μεγάλοιο, but here ὅντινα Μοῦσαι φίλωνται, where Μοῦσαι
instead of Διὸς κοῦραι points to a different group, viz. the singers, being
meant. Compare Od. 8.480 Μοῦσ’ (. . .) φίλησε δὲ φῦλον ἀοιδῶν. Observe
also that at line 81, in the context of διοτρεφέων βασιλήων, the Muses
are referred to as Διὸς κοῦραι, but here as Μοῦσαι. Finally, it is worth
observing that the singer who is loved by the Muses is called ὄλβιος
(line 96), while the king honoured by the Διὸς κοῦραι (line 81) is not
provided with a similar laudatory predicative adjective: singers are just
a bit more valued than kings.

97 γλυκερή οἱ ἀπὸ στόματος ῥέει αὐδή

These words echo lines 83–4 (quo vide), and through them lines
39–40.

At lines 98–104 at last the singers get their due:

εἰ γάρ τις καὶ πένθος ἔχων νεοκηδέι θυμῷ
ἄζηται κραδίην ἀκαχήμενος, αὐτὰρ ἀοιδὸς
100 Μουσάων θεράπων κλεῖα προτέρων ἀνθρώπων
ὑμνήσει μάκαράς τε θεοὺς οἳ ῎Ολυμπον ἔχουσιν·
αἶψ’ ὅ γε δυσφροσυνέων ἐπιλήθεται οὐδέ τι κηδέων
μέμνηται· ταχέως δὲ παρέτραπε δῶρα θεάων.
With Mazon,38 I take it that αὐτάρ introduces the main clause to the
conditional clause εἰ (. . .) ἅζηται; ὑμνήσει is a future indicative and
the main verb. ‘Even if someone, feeling sorrow in a newly anguished
spirit, is parched in his heart with grieving, yet a singer will sing of the

38
‘Un homme porte-t-il le deuil dans son cœur (. . .)? qu’un chanteur (. . .) célèbre
les hauts faits des hommes d’autrefois ou les dieux bienheureux, habitants de l’Olympe:
vite, il oublie ses déplaisirs (. . .).’ Mazon prefers the variant reading ὑμνήσῃ, which he
apparently takes as a hortative subjunctive. For this—extremely rare—use in the third
person cp. Chantraine (1953: 207). I should add that Mazon’s punctuation of the Greek
text—he has a comma before αἶψ(α), line 102—does not match that of his translation,
which has a colon before ‘vite’. For a different view of this sentence see below.
258 albert rijksbaron

glorious deeds (. . .)’, i.e., however sad the circumstances, there will
always be a singer to oppose the glorious, comforting, deeds of earlier
people, and the blessed gods, to the sorrow of the grieving man. For
apodotic αὐτάρ + future indicative after εἰ see Il. 3.288–91 εἰ δ’ ἂν ἐμοὶ
τιμὴν Πρίαμος Πριάμοιό τε παῖδες / τίνειν οὐκ ἐθέλωσιν ᾿Αλεξάνδροιο
πεσόντος, / αὐτὰρ ἐγὼ καὶ ἔπειτα μαχήσομαι (. . .); Il. 22.389–90 εἰ δὲ
θανόντων περ καταλήθοντ’ εἰν ᾿Αΐδαο / αὐτὰρ ἐγὼ καὶ κεῖθι φίλου
μεμνήσομ’ ἑταίρου, and cp. Denniston (1954: 55). For the use of the future
indicative in a habitual context cp. Kühner-Gerth (1898/1904: 1.171) and
especially Pfeijffer (1999: 45–52 ‘Futures expressing a general thought’).
Furthermore, the change from the plural ἀοιδοί at line 95 to the singular
ἀοιδός here should be seen as a sign that Hesiod is implicitly referring to
himself. Finally, note that I have put a high dot after ἔχουσιν; the last two
lines follow therefore asyndetically: they describe the effect of ὑμνήσει.
For similar effect-describing clauses introduced by αἶψα cp. Hes. Op.
43–6 ῥηιδίως γάρ κεν καὶ ἐπ’ ἤματι ἐργάσσαιο, / ὥστε σε κεἰς ἐνιαυτὸν
ἔχειν καὶ ἀεργὸν ἐόντα· / αἶψά κε πηδάλιον μὲν ὑπὲρ καπνοῦ καταθεῖο,
Od. 1.392 οὐ μὲν γάρ τι κακὸν βασιλευέμεν· αἶψά τέ οἱ δῶ / ἀφνειὸν
πέλεται καὶ τιμηέστερος αὐτός, 15.313–7 καί κ’ ἐλθὼν (. . .) / ἀγγελίην
εἴποιμι (. . .) / καί κε μνηστήρεσσιν (. . .) μιγείην, / εἴ μοι δεῖπνον δοῖεν
ὀνείατα μυρί’ ἔχοντες. / αἶψά κεν εὖ δρώοιμι μετὰ σφίσιν, ὅττι θέλοιεν.
Others take a different view of the syntax of our sentence. In their
view, αὐτάρ connects ἅζηται and ὑμνήσει, which is taken as a sub-
junctive, and ἐπιλήθεται at line 101 is taken as the main verb. This
is less likely, for although αὐτάρ can, indeed, connect two verbs in
a conditional clause (Od. 18.376–7 εἰ δ’ αὖ (. . .) ὁρμήσειε Κρονίων /
σήμερον, αὐτὰρ ἐμοὶ σάκος εἴη καὶ δύο δοῦρε / (. . .) / τῶ κέ μ’ ἴδοις
(. . .)) the result at Th. 99 is very lame. If the ἀοιδός is the subject of the
conditional clause, on a par with τις, he makes a most inconspicuous
entrance, while we might expect him to be the most important ele-
ment of the passage, and to contrast with the grieving man.39 Observe
that at Il. 3.288–91 and 22.389–90, quoted above, αὐτάρ is followed by
strongly contrastive ἐγώ.

39
Stretching the meaning of αὐτάρ West notes: ‘we would say “and then”.’ But ‘then’
would not seem to be part of the meaning of αὐτάρ. Observe that at Od. 18.376–7,
quoted in the main text, αὐτάρ cannot possibly be = ‘and then’. Stretching the meaning
of both εἰ and αὐτάρ, Most translates ‘even if (. . .), yet when (. . .)’. This will not do for
‘when’ is not in the Greek: if ὑμνήσει, like ἅζηται, is taken as a subjunctive and still
depends on εἰ γὰρ καί, the translation should be ‘even if someone is parched with griev-
ing, and a singer sings about the deeds of (. . .)’, which gives an impossible meaning.
 discourse cohesion in the proem of hesiod’s theogony 259

100–1 κλεῖα προτέρων ἀνθρώπων

ὑμνήσει μάκαράς τε θεοὺς οἳ ῎Ολυμπον ἔχουσιν·
Just as the Muses Διὶ πατρὶ / ὑμνεῦσαι τέρπουσι μέγαν νόον ἐντὸς
᾿Ολύμπου (lines 36–7) by glorifying in their song the venerated race
of the gods (θεῶν γένος αἰδοῖον, line 44) and Zeus himself (Ζῆνα,
line 47), and by singing of the race of human beings and the Gigants
(ἀνθρώπων τε γένος κρατερῶν τε Γιγάντων, line 50), so the mortal
singer, i.e. Hesiod, will comfort the human beings by singing about the
κλεῖα προτέρων ἀνδρῶν (. . .) μάκαράς τε θεοὺς οἳ ῎Ολυμπον ἔχουσιν,
lines 100–1. See also at line 114.

102 αἶψα echoes αἶψα at line 87: both a king and a singer have suc-
cess quickly.

102–3 οὐδέ τι κηδέων / μέμνηται·

Εchoes lines 60–1: ᾗσιν ἀοιδὴ / μέμβλεται, ἐν στήθεσσιν ἀκηδέα θυμὸν
ἐχούσαις. Just as ἀοιδή for the Muses is connected with the absence of
κήδεα, so the ἀοιδός (line 99) is connected with dissipating the κήδεα
of human beings.

103 ἐπιλήθεται οὐδέ τι κηδέων

μέμνηται· ταχέως δὲ παρέτραπε δῶρα θεάων.
(Gnomic) παρέτραπε at line 104 probably express an activity that is
anterior to (omnitemporal) ἐπιλήθεται and μέμνηται.
Παρ- in παρέτραπε echoes παρ- in παραιφάμενοι (line 90)—both
kings and singers ‘divert’ their audience, from bad plans and sorrows,
respectively—, while δῶρα echoes δόσις (line 93) as well as ἔδον (line
30) and announces δότε (line 104).

104 χαίρετε τέκνα Διός, δότε δ’ ἱμερόεσσαν ἀοιδήν·

κλείετε δ’ ἀθανάτων ἱερὸν γένος αἰὲν ἐόντων,
οἳ Γῆς ἐξεγένοντο (. . .)
Since Hesiod has now sung extensively of the Muses, and has thereby
executed the second part of the order given to him at lines 33–4, he
apparently feels free, being truly their θεράπων, to salute the Muses
and ask them to give him ἀοιδή, so that he can execute the first part of
their order, viz. to sing of the race of the gods (ὑμνεῖν μακάρων γένος
αἰὲν ἐόντων, line 33). Note that this is the first time he has to ask for
their assistance: at line 31 the Muses had of their own accord provided
him with the power to sing. His request basically amounts to ‘breath
260 albert rijksbaron

again a divine voice into me, so as to enable me now to glorify the

history of the gods’, as part of τὰ πρὸ ἐόντα, cp. line 32 ἵνα κλείοιμι
τά τ’ ἐσσόμενα πρό τ’ ἐόντα.

114 ταῦτά μοι ἔσπετε Μοῦσαι ᾿Ολύμπια δώματ’ ἔχουσαι

ἐξ ἀρχῆς, καὶ εἴπαθ’, ὅτι πρῶτον γένετ’ αὐτῶν.
Like the Muses among the gods (cp. lines 43–6 αἱ δ’ ἄμβροτον ὄσσαν
ἱεῖσαι / θεῶν γένος αἰδοῖον πρῶτον κλείουσιν ἀοιδῇ / ἐξ ἀρχῆς, οὓς
Γαῖα καὶ Οὐρανὸς εὐρὺς ἔτικτεν, οἵ τ’ ἐκ τῶν ἐγένοντο), Hesiod is
going to sing among mortals of the gods ἐξ ἀρχῆς. He in fact appeals
to the Muses to transfer their activities on Olympus to an audience of
mortals, using Hesiod as their θεράπων. Compare Thalmann (1984:
139): ‘The Muses’ gift to the mortal poet, the Theogony itself, is the
human realization of the divine song.’

Conclusion

Verdenius’ view, that the ‘(. . .) unity of the proem [lies] in the con-
tinuity of its progress’ could be shown to be correct. The progress is
established by the sequence στεῖχον (line 10), ποτ’ ἐδίδαξαν (line 22),
viz. during the στείχειν of the Muses, and αἳ τότ’ ἴσαν (line 68) (. . .)
νισομένων πατέρ’ εἰς ὅν (line 71), τότε referring back to ἐδίδαξαν. This
course of the actions strongly suggests that the Muses seized the occa-
sion of one of their regular visits to Helicon to dance there (ὀρχεῦνται,
line 4), to bestow the gift of poetry on Hesiod. But, contrary to what
Verdenius believed, the unity is also brought about by a number of
interdependencies. Thus, the imperfect verb forms which express a
sound (ἄειδον, line 75), or are accompanied by verb forms expressing a
sound (στεῖχον περικαλλέα ὄσσαν ἱεῖσαι, / ὑμνεῦσαι, line 10; αἳ τότ’ ἴσαν
(. . .), ἀγαλλόμεναι ὀπὶ καλῇ, / (. . .) μολπῇ· (. . .) ἴαχε / ὑμνεύσαις, (. . .)
δοῦπος, lines 68–70) presupppose the presence of a human perceiver:
the poet himself, who is present, so to speak, during the whole proem;
they all refer to the ‘same’ past of the poet. Also, there is a recurrent
reference to Hesiod acting as the servant (θεράπων, line 100) of the
Muses, to do among mortals what they themselves do among the gods,
to sing both of the gods and of mankind:

– line 33: the Muses to Hesiod: καί μ’ ἐκέλονθ’ ὑμνεῖν μακάρων γένος
αἰὲν ἐόντων;
 discourse cohesion in the proem of hesiod’s theogony 261

– lines 43–5: the Muses on Olympus: αἱ δ’ ἄμβροτον ὄσσαν ἱεῖσαι /

θεῶν γένος αἰδοῖον πρῶτον κλείουσιν ἀοιδῇ / ἐξ ἀρχῆς, οὓς Γαῖα καὶ
Οὐρανὸς εὐρὺς ἔτικτεν, οἵ τ’ ἐκ τῶν ἐγένοντο;
– lines 50–1: the Muses on Olympus: (. . .) ἀνθρώπων τε γένος κρατερῶν
τε Γιγάντων / ὑμνεῦσαι τέρπουσι Διὸς νόον ἐντὸς ᾿Ολύμπου
– lines 99–101: a mortal singer among mortals: ἀοιδὸς / Μουσάων
θεράπων κλεῖα προτέρων ἀνθρώπων / ὑμνήσει μάκαράς τε θεοὺς οἳ
῎Ολυμπον ἔχουσιν;
– lines 104–5: Hesiod among mortals, to the Muses: δότε δ’ ἱμερόεσσαν
ἀοιδήν· / κλείετε δ’ ἀθανάτων ἱερὸν γένος αἰὲν ἐόντων, 108 εἴπατε δ’
ὡς τὰ πρῶτα θεοὶ καὶ γαῖα γένοντο, 114–5 ταῦτά μοι ἔσπετε Μοῦσαι
᾿Ολύμπια δώματ’ ἔχουσαι / ἐξ ἀρχῆς, καὶ εἴπαθ’, ὅτι πρῶτον γένετ’
αὐτῶν.40

Appendix to στεῖχον, line 10: a focalising imperfect in the myth

of Plato’s Phaedrus

In the myth of Plato’s Phaedrus, more specifically in the passage 253e5–

255a1, the omnitemporal, habitual behaviour of the horses of the soul
is described by means of present and (gnomic) aorist indicatives. But
in between there is, at 254d6, just one imperfect form, ἦσαν, in a tem-
poral clause: (. . .) κατέχει (. . .) ἐντρέπεται, (. . .) φέρεται, (. . .) ἀναγκάζει
(. . .) ἀντιτείνετον (254b1) (. . .) πορεύεσθον (. . .), καὶ (254b4) πρὸς αὐτῷ
τ’ ἐγένοντο καὶ εἶδον (. . .) ἠνέχθη, καὶ πάλιν εἶδεν (. . .) ἔδεισε (. . .).
ἀνέπεσεν (. . .) ἠναγκάσθη (. . .) (254c5) ἔβρεξε (. . .) ἐλοιδόρησεν (. . .)
(254d2) μόγις συνεχώρησεν δεομένων εἰς αὖθις ὑπερβαλέσθαι. ἐλθόντος
δὲ τοῦ συντεθέντος χρόνου (. . .) βιαζόμενος, χρεμετίζων, ἕλκων (254d5)
ἠνάγκασεν αὖ προσελθεῖν τοῖς παιδικοῖς ἐπὶ τοὺς αὐτοὺς λόγους, καὶ
ἐπειδὴ ἐγγὺς ἦσαν, ἐγκύψας καὶ ἐκτείνας τὴν κέρκον, ἐνδακὼν τὸν χαλινόν,
μετ’ ἀναιδείας ἕλκει· ὁ δ’ (254e1) ἡνίοχος (. . .) τήν τε κακηγόρον γλῶτταν
καὶ τὰς γνάθους καθῇμαξεν καὶ τὰ σκέλη (. . .) ὀδύναις ἔδωκεν. ὅταν δὲ
(. . .) λήξῃ, ταπεινωθεὶς ἕπεται ἤδη τῇ τοῦ ἡνιόχου προνοίᾳ, καὶ ὅταν ἴδῃ
τὸν καλόν, φόβῳ διόλλυται· etc. This imperfect is often misunderstood,
and has been translated as if it were a present indicative or a gnomic

40
I am indebted to the members of the Amsterdamse Hellenistenclub, especially
Niek van der Ben, Jan Maarten Bremer and Irene de Jong, and to the editors of this
volume, for their critical comments.
262 albert rijksbaron

aorist. Here follow some faulty English translations: Jowett (18923): (the
unruly horse) ‘(. . .) forces them [i.e. the other horse and the charioteer]
to draw near again. And when they are near he stoops his head (. . .),
and (. . .) pulls shamelessly’, Fowler (Loeb, 1914): ‘(. . .) he forces them
again (. . .) to approach the beloved one, and when they are near him,
he lowers his head (. . .) and pulls shamelessly’, Rowe (1986; he has no
note on ἦσαν): ‘(. . .) it forces them to approach the beloved again (. . .),
and when they are nearby (. . .) it pulls shamelessly.’ ‘When (. . .) are’ is
of course impossible: to have ‘when’ in English, the Greek should have
had ὅταν or ἐπειδάν + subjunctive, as indeed later in the myth. Robin,
however, correctly—and brilliantly—translates (Budé, 1961), avoiding
the use of a verb in his translation: ‘(. . .) une fois de plus il les a con-
traints d’approcher du bien-aimé (. . .). Enfin, maintenant que les voilà
à proximité, il se penche en avant sur lui (. . .) il tire sans vergogne.’
The imperfect all of a sudden, dramatically, transfers the reader from
the general description of the behaviour of the ‘good’ horse and the
charioteer to the point of view of the ‘bad’ horse. Having forced the
other two to approach the beloved one (ἠνάγκασε προσελθεῖν: they
have come close (aorist) to the beloved one) the ‘bad’ horse sees that
they are there: ‘now/seeing that they were close, it lowers its head and
(. . .) pulls shamelessly.’
 discourse cohesion in the proem of hesiod’s theogony 263

The structure of the proem of Hesiod’s Theogony

Μουσάων ῾Ελικωνιάδων ἀρχώμεθ’ ἀείδειν, 1 Metatextual formula

αἵ θ’ ῾Ελικῶνος ἔχουσιν ὄρος μέγα τε ζάθεόν τε, 2–8 Hymn to the Muses part I
καί τε περὶ κρήνην ἰοειδέα πόσσ’ ἁπαλοῖσιν attributive section: description of some
ὀρχεῦνται καὶ βωμὸν ἐρισθενέος Κρονίωνος characteristic activities of the Muses on
5 καί τε λοεσσάμεναι τέρενα χρόα Περμησσοῖο Mount Helicon (omnitemporal presents:
ἠ’ ῞Ιππου κρήνης ἠ’ ᾿Ολμειοῦ ζαθέοιο 2 ἔχουσιν, 4 ὀρχεῦνται, gnomic aorists: 7
ἀκροτάτῳ ῾Ελικῶνι χοροὺς ἐνεποιήσαντο, ἐνεποιήσαντο, 8 ἐπερρώσαντο) passing
καλοὺς ἱμερόεντας, ἐπερρώσαντο δὲ ποσσίν. into a narrative
ἔνθεν ἀπορνύμεναι κεκαλυμμέναι ἠέρι πολλῷ 9–34 Personal narrative: Hesiod hears
10 ἐννύχιαι στεῖχον περικαλλέα ὄσσαν ἱεῖσαι, the Muses singing on the slopes of Mount
ὑμνεῦσαι Δία τ’ αἰγίοχον καὶ πότνιαν ῞Ηρην Helicon
᾿Αργείην, χρυσέοισι πεδίλοις ἐμβεβαυῖαν, (imperfect + verb of sound: 10 στεῖχον (. . .)
κούρην τ’ αἰγιόχοιο Διὸς γλαυκῶπιν ᾿Αθήνην ὑμνεῦσαι)
Φοῖβόν τ’ ᾿Απόλλωνα καὶ ῎Αρτεμιν ἰοχέαιραν
15 ἠδὲ Ποσειδάωνα γαιήοχον ἐννοσίγαιον
καὶ Θέμιν αἰδοίην ἑλικοβλέφαρόν τ’ ᾿Αφροδίτην
῞Ηβην τε χρυσοστέφανον καλήν τε Διώνην
Λητώ τ’ ᾿Ιαπετόν τε ἰδὲ Κρόνον ἀγκυλομήτην
᾿Ηῶ τ’ ᾿Ηέλιόν τε μέγαν λαμπράν τε Σελήνην
20 Γαῖάν τ’ ᾿Ωκεανόν τε μέγαν καὶ Νύκτα μέλαιναν
ἄλλων τ’ ἀθανάτων ἱερὸν γένος αἰὲν ἐόντων.
αἵ νύ ποθ’ ῾Ησίοδον καλὴν ἐδίδαξαν ἀοιδήν, (. . .) they bestow on him the ability to
ἄρνας ποιμαίνονθ’ ῾Ελικῶνος ὕπο ζαθέοιο. sing (Dichterweihe) (aorist: 22 ἐδίδαξαν
τόνδε δέ με πρώτιστα θεαὶ πρὸς μῦθον ἔειπον, ἀοιδήν). . . .
25 Μοῦσαι ᾿Ολυμπιάδες, κοῦραι Διὸς αἰγιόχοιο·
‘ποιμένες ἄγραυλοι, κάκ’ ἐλέγχεα, γαστέρες οἶον,
ἴδμεν ψεύδεα πολλὰ λέγειν ἐτύμοισιν ὁμοῖα,
ἴδμεν δ’ εὖτ’ ἐθέλωμεν ἀληθέα γηρύσασθαι.’
ὣς ἔφασαν κοῦραι μεγάλου Διὸς ἀρτιέπειαι,
30 καί μοι σκῆπτρον ἔδον δάφνης ἐριθηλέος ὄζον
δρέψασαι, θηητόν· ἐνέπνευσαν δέ μοι αὐδὴν
θέσπιν, ἵνα κλείοιμι τά τ’ ἐσσόμενα πρό τ’ ἐόντα,
καί μ’ ἐκέλονθ’ ὑμνεῖν μακάρων γένος αἰὲν 33 (. . .) and urge him to use his newly
ἐόντων, σφᾶς δ’ αὐτὰς πρῶτόν τε καὶ ὕστατον acquired ability to sing of the gods and first
αἰὲν ἀείδειν. of all of themselves (imperfect: 33 ἐκέλοντο)
35 ἀλλὰ τίη μοι ταῦτα περὶ δρῦν ἢ περὶ πέτρην; 35–6 Metatextual formulas
τύνη, Μουσάων ἀρχώμεθα, ταὶ Διὶ πατρὶ 36 ταί –67 Hymn to the Muses part II
ὑμνεῦσαι τέρπουσι μέγαν νόον ἐντὸς ᾿Ολύμπου, attributive section: description of the hymnic
εἴρουσαι τά τ’ ἐόντα τά τ’ ἐσσόμενα πρό τ’ ἐόντα, activities of the Muses on Olympus, who sing
φωνῇ ὁμηρεῦσαι, τῶν δ’ ἀκάματος ῥέει αὐδὴ of the gods and of mankind
(omnitemporal presents: 37 τέρπουσι,
39 ῥέει, 40 γελᾷ, 42 ἠχεῖ, 44 κλείουσιν,
51 τέρπουσι, 66 μέλπονται, 67 κλείουσιν)
 264 albert rijksbaron

40 ἐκ στομάτων ἡδεῖα· γελᾷ δέ τε δώματα πατρὸς

Ζηνὸς ἐριγδούποιο θεᾶν ὀπὶ λειριοέσσῃ
σκιδναμένῃ, ἠχεῖ δὲ κάρη νιφόεντος ᾿Ολύμπου
δώματά τ’ ἀθανάτων· αἱ δ’ ἄμβροτον ὄσσαν ἱεῖσαι
θεῶν γένος αἰδοῖον πρῶτον κλείουσιν ἀοιδῇ
45 ἐξ ἀρχῆς, οὓς Γαῖα καὶ Οὐρανὸς εὐρὺς ἔτικτεν,
οἵ τ’ ἐκ τῶν ἐγένοντο, θεοὶ δωτῆρες ἐάων·
δεύτερον αὖτε Ζῆνα θεῶν πατέρ’ ἠδὲ καὶ ἀνδρῶν,
[ἀρχόμεναί θ’ ὑμνεῦσι θεαὶ † λήγουσαί τ’ ἀοιδῆς,]
ὅσσον φέρτατός ἐστι θεῶν κάρτει τε μέγιστος·
50 αὖτις δ’ ἀνθρώπων τε γένος κρατερῶν τε Γιγάντων
ὑμνεῦσαι τέρπουσι Διὸς νόον ἐντὸς ᾿Ολύμπου
Μοῦσαι ᾿Ολυμπιάδες, κοῦραι Διὸς αἰγιόχοιο.
τὰς ἐν Πιερίῃ Κρονίδῃ τέκε πατρὶ μιγεῖσα 53–62 Mousogony: narrative digression
Μνημοσύνη, γουνοῖσιν ᾿Ελευθῆρος μεδέουσα, within the attributive section
55 λησμοσύνην τε κακῶν ἄμπαυμά τε μερμηράων. (non-gnomic aorists: 53 τέκε, 60 ἔτεκε,
ἐννέα γάρ οἱ νύκτας ἐμίσγετο μητίετα Ζεὺς iterative imperfect: 56 μίσγετο)
νόσφιν ἀπ’ ἀθανάτων ἱερὸν λέχος εἰσαναβαίνων·
ἀλλ’ ὅτε δή ῥ’ ἐνιαυτὸς ἔην, περὶ δ’ ἔτραπον ὧραι
μηνῶν φθινόντων, περὶ δ’ ἤματα πόλλ’ ἐτελέσθη,
60 ἡ δ’ ἔτεκ’ ἐννέα κούρας, ὁμόφρονας, ᾗσιν ἀοιδὴ
μέμβλεται ἐν στήθεσσιν, ἀκηδέα θυμὸν ἐχούσαις,
τυτθὸν ἀπ’ ἀκροτάτης κορυφῆς νιφόεντος
᾿Ολύμπου·
ἔνθά σφιν λιπαροί τε χοροὶ καὶ δώματα καλά,
πὰρ δ’ αὐτῇς Χάριτές τε καὶ ῞Ιμερος οἰκί’ ἔχουσιν
65 ἐν θαλίῃς· ἐρατὴν δὲ διὰ στόμα ὄσσαν ἰεῖσαι
μέλπονται, πάντων τε νόμους καὶ ἤθεα κεδνὰ
ἀθανάτων κλείουσιν, ἐπήρατον ὄσσαν ἱεῖσαι. 67 End of attributive section on the Muses
αἳ τότ’ ἴσαν πρὸς ῎Ολυμπον, ἀγαλλόμεναι ὀπὶ 68–79 Narrative about Hesiod and the
καλῇ, Muses, continued: τότε refers back to 22
ἀμβροσίῃ μολπῇ· περὶ δ’ ἴαχε γαῖα μέλαινα ἐδίδαξαν, return of the Muses to Olympus
70 ὑμνεύσαις, ἐρατὸς δὲ ποδῶν ὕπο δοῦπος ὀρώρει (their abode, 63) as perceived by the poet
νισομένων πατέρ’ εἰς ὅν· ὁ δ’ οὐρανῷ ἐμβασιλεύει, (reappearance (cp. 10) of imperfects and
αὐτὸς ἔχων βροντὴν ἠδ’ αἰθαλόεντα κεραυνόν, verbs of sound: 68 ἴσαν (. . .) ἀγαλλόμεναι
κάρτει νικήσας πατέρα Κρόνον· εὖ δὲ ἕκαστα ὀπὶ καλῇ, 69–70 ἴαχε γαῖα (. . .) ὑμνεύσαις, 70
ἀθανάτοις διέταξε νόμους καὶ ἐπέφραδε τιμάς. οῦπος ὀρώρει, 75 ἄειδον)
75 ταῦτ’ ἄρα Μοῦσαι ἄειδον ᾿Ολύμπια δώματ’ 71–4 Interrupted by brief attributive section
ἔχουσαι, on Zeus (omnitemporal present: ἐμβασιλεύει)
ἐννέα θυγατέρες μεγάλου Διὸς ἐκγεγαυῖαι, + brief narrative digression (non-gnomic
Κλειώ τ’ Εὐτέρπη τε Θάλειά τε Μελπομένη τε aorists: διέταξε, ἐπέφραδε)
Τερψιχόρη τ’ ᾿Ερατώ τε Πολύμνιά τ’ Οὐρανίη τε
Καλλιόπη θ’· ἡ δὲ προφερεστάτη ἐστὶν
ἁπασέων.
 discourse cohesion in the proem of hesiod’s theogony 265

80 ἡ γὰρ καὶ βασιλεῦσιν ἅμ’ αἰδοίοισιν ὀπηδεῖ. 79–80 Brief attributive section on
ὅντινα τιμήσουσι Διὸς κοῦραι μεγάλοιο Calliope, the most prominent of the Muses
γεινόμενόν τε ἴδωσι διοτρεφέων βασιλήων, (omnitemporal presents: 79 ἐστίν, 80
τῷ μὲν ἐπὶ γλώσσῃ γλυκερὴν χείουσιν ἐέρσην, ὀπηδεῖ), which triggers an attributive section-
τοῦ δ’ ἔπε’ ἐκ στόματος ῥεῖ μείλιχα· οἱ δέ νυ λαοὶ like description of the Muses’ activities
85 πάντες ἐς αὐτὸν ὁρῶσι διακρίνοντα θέμιστας among mortals through kings (81–92) and
ἰθείῃσι δίκῃσιν· ὁ δ’ ἀσφαλέως ἀγορεύων singers (94–103); like the Muses, the latter
αἶψά τι καὶ μέγα νεῖκος ἐπισταμένως κατέπαυσε· sing of mankind and the gods (verbless
τούνεκα γὰρ βασιλῆες ἐχέφρονες, οὕνεκα λαοῖς clauses: 88 ἐχέφρονες, 93 τοίη, 96 ὄλβιος;
βλαπτομένοις ἀγορῆφι μετάτροπα ἔργα τελεῦσι omnitemporal presents: 83 χείουσιν, 84 ῥεῖ,
90 ῥηιδίως, μαλακοῖσι παραιφάμενοι ἐπέεσσιν· 85 ὁρῶσι, 89 τελεῦσι, 91 ἱλάσκονται, 92
ἐρχόμενον δ’ ἀν’ ἀγῶνα θεὸν ὣς ἱλάσκονται πρέπει, 95 ἔασιν, 97 ῥέει, 102 ἐπιλήθεται,
αἰδοῖ μειλιχίῃ, μετὰ δὲ πρέπει ἀγρομένοισι. 103 μέμνηται; gnomic aorists: 87 κατέπαυσε,
τοίη Μουσάων ἱερὴ δόσις ἀνθρώποισιν. 103 παρέτραπε; generic future: 101 ὑμνήσει)
ἐκ γάρ τοι Μουσέων καὶ ἑκηβόλου ᾿Απόλλωνος
95 ἄνδρες ἀοιδοὶ ἔασιν ἐπὶ χθόνα καὶ κιθαρισταί,
ἐκ δὲ Διὸς βασιλῆες· ὁ δ’ ὄλβιος, ὅντινα Μοῦσαι
φίλωνται· γλυκερή οἱ ἀπὸ στόματος ῥέει αὐδή.
εἰ γάρ τις καὶ πένθος ἔχων νεοκηδέι θυμῷ
ἄζηται κραδίην ἀκαχήμενος, αὐτὰρ ἀοιδὸς
100 Μουσάων θεράπων κλεῖα προτέρων ἀνθρώπων
ὑμνήσει μάκαράς τε θεοὺς οἳ ῎Ολυμπον ἔχουσιν·
αἶψ’ ὅ γε δυσφροσυνέων ἐπιλήθεται οὐδέ τι κηδέων
μέμνηται· ταχέως δὲ παρέτραπε δῶρα θεάων.
χαίρετε τέκνα Διός, δότε δ’ ἱμερόεσσαν ἀοιδήν· 104–15 Salutation and prayer: having
105 κλείετε δ’ ἀθανάτων ἱερὸν γένος αἰὲν ἐόντων, executed the second part of the order of lines
οἳ Γῆς ἐξεγένοντο καὶ Οὐρανοῦ ἀστερόεντος, 33–4, viz. to sing of the Muses, Hesiod now
Νυκτός τε δνοφερῆς, οὕς θ’ ἁλμυρὸς ἔτρεφε salutes the Muses and asks them to give him
Πόντος. ἀοιδή so that he can execute the first part of
εἴπατε δ’ ὡς τὰ πρῶτα θεοὶ καὶ γαῖα γένοντο their order, viz. to sing of the gods and their
καὶ ποταμοὶ καὶ πόντος ἀπείριτος οἴδματι θυίων birth
110 ἄστρά τε λαμπετόωντα καὶ οὐρανὸς εὐρὺς
ὕπερθεν·
[οἵ τ’ ἐκ τῶν ἐγένοντο, θεοὶ δωτῆρες ἐάων·]
ὥς τ’ ἄφενος δάσσαντο καὶ ὡς τιμὰς διέλοντο,
ἠδὲ καὶ ὡς τὰ πρῶτα πολύπτυχον ἔσχον
῎Ολυμπον.
ταῦτά μοι ἔσπετε Μοῦσαι ᾿Ολύμπια δώματ’
ἔχουσαι
115 ἐξ ἀρχῆς, καὶ εἴπαθ’, ὅτι πρῶτον γένετ’ αὐτῶν.
 BIBLIOGRAPHY

Abelen, E., G. Redeker, S. Thompson, (1993), ‘The rhetorical structure of US-American

and Dutch fund-raising letters’, Text 13, 323–50.
Abraham, W., (ed.), (1986), Tijdschrift voor tekst- en taalwetenschap (TTT) 6 no. 2
(special issue on particles).
Adam, J.-M. & A. Petitjean, (1989), Le texte descriptif, Paris.
——, (2005), Les textes: types et prototypes, Paris.
Adema, S., (2007), ‘Discourse modes and bases in Vergil’s Aeneid’, in R.J. Allan &
M. Buijs (eds), 42–64.
Adrados, F.R., (1992), Nueva sintaxis del griego antiguo, Madrid.
Allan, R.J. & M. Buijs (eds), (2007), The language of literature: linguistic approaches to
classical texts (Amsterdam studies in Classical philology 13), Leiden.
Allan, R.J., (2003), The middle voice in Ancient Greek: a study in polysemy (Amsterdam
studies in Classical philology 11), Amsterdam.
——, (2007), ‘Sense and sentence complexity: sentence structure, sentence connection,
and tense-aspect as indicators of narrative mode in Thucydides’ Histories’, in R.J.
Allan & M. Buijs (eds), 93–121.
Austin, J.L., (1962), How to do things with words, Cambridge, MA.
Autenrieth, G., (2002), A Homeric dictionary for use in schools and colleges. (From the
German of Georg Autenrieth; translated, with additions and corrections by R.P.
Keep), Piscataway.
Baker, C.L., (1995), ‘Contrast, discourse prominence, and intensification, with special
reference to locally free reflexives in British English’, Language 71, 63–101.
Bakker, E.J., (1986), ‘ ὅσπερ and εἴπερ: een aspect van Attische conversatie’, Lampas
19, 142–58.
——, (1988), Linguistics and formulas in Homer. Scalarity and the description of the
particle per, Amsterdam/Philadelphia.
——, (1993), ‘Boundaries, topics and the structure of discourse: an investigation of the
Ancient Greek particle δέ’, Studies in language 17, 275–311.
——, (ed.), (1997a), Grammar as interpretation. Greek literature in its linguistic contexts
(Mnemosyne supplement 171), Leiden.
——, (1997b), ‘Verbal aspect and mimetic description in Thucydides’, in E.J. Bakker
(ed.), 7–54.
——, (1997c), Poetry in speech: orality and Homeric discourse, Ithaca.
——, (1999), ‘Homeric οὗτος and the poetics of deixis’, Classical philology 94, 1–19.
——, (2002), ‘Remembering the god’s arrival’, Arethusa 35, 63–81.
——, (2005), Pointing to the past: from formula to performance in Homeric poetics,
Washington.
Bakker, S.J., (2002), ‘Futura zonder toekomst’, Lampas 35, 199–214.
Bal, M., (19972), Narratology: introduction to the theory of narrative, Toronto/Buffalo/
London.
Barlow, S.A., (1996), Euripides Heracles, Oxford.
Barrett, W.S., (1964), Euripides Hippolytos, Oxford.
Basset, L., (1988), ‘L’infinitif en grec ancien: du sens à la référence’, in S. Rémi-Giraud
(ed.), L’infinitif. Une approche comparative, Lyon, 127–36.
——, (1989a), La syntaxe de l’imaginaire. Étude des modes et des négations dans l’Iliade
et l’Odyssée, Lyon.
——, (1989b), ‘L’augment et la distinction Discours/Récit dans l’Iliade et l’Odyssée’, in
M. Casevitz (ed.), Études homériques, Lyon, 9–16.
268 bibliography

——, (1997), ‘ Ἀλλ’ ἐξόλοισθ’ αὐτῷ κοαξ. Réexamen des emplois de ἀλλά à la lumière de
l’énonciation dans Les Grenouilles d’Aristophane’, in A. Rijksbaron (ed.), 75–99.
——, (1999), ‘Des participiales parmi les complétives’, in B. Jacquinod (ed.), Les com-
plétives en grec ancien. Actes du colloque international de Saint-Étienne, septembre
1998, Saint-Étienne, 33–44.
——, (2000a), ‘«Peux-tu le dire?» Étude d’une paire minimale PR/AO’, in B. Jacquinod
et al. (eds), 305–16.
——, (2000b), ‘Corrélations aspectuelles PR/AO entre verbes régissants et verbes régis’,
in B. Jacquinod et al. (eds), 233–45.
——, (2003a), ‘Les verbes «dire» en grec ancien d’après l’œuvre d’Isocrate’, in S. Rémi-
Giraud & L. Panier (eds), La polysémie ou l’empire des sens. Lexique, discours,
représentations, Lyon, 173–87.
——, (2003b), ‘Le prévisible et l’imprévu. Oppositions de l’imparfait et de l’aoriste
indicatif dans Thuc. VI, 50–52’, Syntaktika 26, 1–7.
Beekes, R.S.P., (1995), Comparative Indo-European linguistics, Amsterdam.
Benveniste, É., (1959), ‘Les relations de temps dans le verbe français’, BSL 54, fasc. 1,
69–82. (= Benveniste 1966, 237–250).
Benveniste, É., (1966), Problèmes de linguistique générale, Vol. I, Paris.
Bertocchi, A., (2000), ‘Ipse as an intensifier’, in G. Calboli (ed.), Papers on grammar
V, Bologna, 15–30.
Biraud, M., (1990), ‘Étude sémantique du determinant d’ipséité αὐτός’, Lalies 9,
95–8.
Blakemore, D., (1989), ‘Denial and contrast: a relevance theoretic analysis of but’,
Linguistics and philosophy 12, 15–37.
——, (1992), Understanding utterances, Oxford.
Blass, F. (1889), ‘Demosthenische Studien. Aorist und Imperfekt’, RhM 44, 406–30.
Blomqvist, J., (1969), Greek particles in Hellenistic prose, Lund.
Boel, G. de, (1980), ‘Towards a theory of the meaning of complementizers in Classical
Attic’, Lingua 52, 285–304.
Boisacq, E., (1916), Dictionnaire étymologique de la langue grecque, Heidelberg/Paris.
Bolkestein, A.M. & M. van de Grift, (1994), ‘Participant tracking in Latin discourse’, in
J. Herman (ed.), Linguistic studies on Latin. Selected papers from the 6th international
colloquium on Latin linguistics (Budapest, 23–27 March 1991), Amsterdam/Phila-
delphia, 283–302.
Bonheim, H., (1982), The narrative modes: techniques of the short story, Cambridge.
Bonifazi, A., (2001), Mescolare un cratere di canti. Pragmatica della poesia epinicia in
Pindaro, Alessandria.
——, (2004a), ‘Communication in Pindar’s deictic acts’, Arethusa 37, 391–414.
——, (2004b), ‘κεῖνος in Pindar: between grammar and poetic intention’, CPh 99,
283–99.
——, (2008), ‘Memory and visualization in Homeric discourse markers’, in E.A.
Mackay (ed.), Orality, literacy, memory in the Ancient Graeco-Roman world, Leiden,
35–64.
——, (1993), Hésiode. Théogonie. Texte et traduction, Paris.
Bosch, P., (1983), Agreement and anaphora. A study of the role of pronouns in syntax
and discourse, London/New York.
Bremond, C., (1973), Logique du récit, Paris.
Brown, G., & G. Yule, (1983), Discourse analysis, Cambridge.
Brugmann, K., (1904), Die Demonstrativpronomina der indogermanischen Sprachen
(Abhandlungen der philol.-hist. Klasse der sächsischen Gesellschaft der Wissen-
schaften, Band 22 n. 6), Leipzig.
Buijs, M., (2005), Clause combining in Ancient Greek narrative discourse: the distri-
bution of subclauses and participial clauses in Xenophon’s Hellenica and Anabasis,
Leiden.
 bibliography 269

——, (2007), ‘Aspectual differences and narrative technique: Xenophon’s Hellenica and
Agesilaus’, in R.J. Allan & M. Buijs (eds), 122–53.
Burkhardt, P., (2002), ‘Logophors: looking outside of syntax’, in M. van Koppen,
E. Thrift, E.J. van der Torre & M. Zimmermann (eds), Proceedings of console IX,
14–27.
Carrell, P.L., (1982), ‘Cohesion is not coherence’, Tesol quarterly 16, 479–88.
Chafe, W., (1994), Discourse, consciousness, and time: the flow and displacement of
conscious experience in speaking and writing, Chicago/London.
Chanet, A.-M., (1988), ‘Objet propositionnel, prolepse et objet externe’, in A. Rijksbaron,
H.A. Mulder, G.C. Wakker (eds), In the footsteps of Raphaël Kühner, Amsterdam,
67–97.
Chantraine, P., (1926), Histoire du parfait grec, Paris.
——, (1968–80), Dictionnaire étymologique de la langue grecque. Histoire des mots,
Paris.
——, (1953) (reprinted several times), Grammaire homérique, 2 vols., Paris.
Chatman, S., (1978), Story and discourse: narrative structure in fiction and film,
Ithaca/London.
——, (1990), Coming to terms: the rhetoric of narrative in fiction and film, Ithaca, NY.
Clay, J.S., (1988), ‘What the muses sang: Theogony 1–115’, GRBS 29, 323–33.
——, (1989), The politics of Olympus, Princeton.
——, (2003), Hesiod’s cosmos, Cambridge.
Comrie, B., (1976), Aspect, Cambridge.
——, (1985), Tense, Cambridge.
Conacher, D.J., (1980), Aeschylus’ Prometheus Bound: a literary commentary, Toronto.
Conte, M.-E., (1999), Condizioni di coerenza. Ricerche di linguistica testuale. Nuova
edizione, Alessandria.
Cornish, F., (1999), Anaphora, discourse, and understanding. Evidence from English
and French, Oxford.
——, (2002), ‘Anaphora: lexico-textual structure, or means for utterance integration
within a discourse? A critique of the functional-grammar account’, Linguistics 40,
469–93.
Coulon, V. & H. van Daele, (1960), Aristophane (Tome 5), Paris.
Crespo, E., (2007), ‘Particles as markers of literary units in the Attic orators’, unpub-
lished paper.
Cristofaro, S., (1996), Aspetti sintattici e semantici delle frasi completive in greco antico,
Florence.
——, (1998), ‘Grammaticalization and clause linkage strategies: a typological approach
with particular reference to Ancient Greek’, in A.G. Ramat & P.J. Hopper (eds), The
limits of grammaticalization, Amsterdam/Philadelphia, 59–88.
Culioli, A., (2000), ‘Essai de bilan’, in B. Jacquinod et al. (eds), 19–26.
Culy, C., (1997), ‘Logophoric pronouns and point of view’, Linguistics 35, 845–60.
Daneš, F., (1990), ‘On the stylistic relevance of the choice of anaphoric expressions’,
Rivista di Linguistica 2, 121–39.
Danielewicz, J., (1976), Morfologia hymnu antyznego. Poznán.
Davies, M., (1991), Sophocles: Trachiniae, Oxford.
Denniston, J.D., (1952), Greek prose style, Oxford.
——, (19542), The Greek particles, Oxford.
Depew, M., (2004), ‘Gender, power, and poetics in Callimachus’ book of Hymns’, in
M.A. Harder, R.F. Regtuit, G.C. Wakker (eds), Callimachus II (Hellenistica Gro-
ningana 7), Leuven, 117–38.
Dickey, E., (1996), Greek forms of address: from Herodotus to Lucian, Oxford.
Dijk, T. van, (1982), ‘Episodes as units of discourse analysis’, in D. Tannen (ed.), Spoken
and written language: exploring orality and literacy (Advances in discourse processes
9), Norwood, NJ, 177–195.
270 bibliography

Dik, H., (1995), Word order in Ancient Greek. A pragmatic account of word order varia-
tion in Herodotus (Amsterdam studies in Classical philology 5), Amsterdam.
——, (2007), Word order in Greek tragic dialogue, Oxford.
Dik, S.C., M.E. Hoffmann, J.R. de Jong, S.I. Djiang, H. Stromer, L. de Vries, (1981), ‘On
the typology of focus phenomena’, in T. Hoekstra, H. van der Hulst, M. Moortgat
(eds), Perspectives on functional grammar, Dordrecht, 41–74.
Dik, S.C., K. Hengeveld, E. Vester, C. Vet, (1990), ‘The hierarchical structure of
the clause and the typology of adverbial satellites’, in J. Nuyts, A.M. Bolkestein,
C. Vet (eds), Layers and levels of representation in language theory. A functional view,
Amsterdam/Philadelphia, 25–70.
Dik, S.C. & K. Hengeveld, (1991), ‘The hierarchical structure of the clause and the
typology of perception verb complements’, Linguistics 29, 231–59.
Dik, S.C., (1968), Coordination: its implications for the theory of general linguistics,
Amsterdam.
——, (1997a), The theory of functional grammar, Part 1: The structure of the clause (2nd,
revised edition, ed. by K. Hengeveld), Berlin/New York.
——, (1997b), The theory of functional grammar, Part 2: Complex and derived construc-
tions (2nd, revised edition, ed. by K. Hengeveld), Berlin/New York.
Dodds, E.R., (1959), Plato. Gorgias: a revised text with introduction and commentary,
Oxford.
Dover, K., (1993), Aristophanes Frogs, Oxford.
Ducrot, O., (1984), Le dire et le dit, Paris.
Duhoux, Y., (1997a), ‘Grec écrit et grec parlé: une étude contrastive des particules aux
Ve–IVe siècles’, in A. Rijksbaron (ed.), 15–48.
——, (1997b), ‘Quelques idées reçues, et néanmoins fausses, sur les particules grecques’,
AC 66, 281–8.
——, (1998), ‘Les particules grecques: les situations homérique et mycénienne’, in
Y. Duhoux (ed.), Langue et langues. Hommage à Albert Maniet, Louvain-la-neuve,
13–42.
——, (20002), Le verbe grec ancien. Éléments de morphologie et de syntaxe historiques,
Leuven.
——, (2006), ‘Les particules: une classe de mots à supprimer en grec ancien?’, in
E. Crespo, J. de la Villa, & A.R. Revuelta (eds), Word classes and related topics in
Ancient Greek, Louvain-la-neuve, 519–36.
Dunbar, N., (1995), Aristophanes Birds, Oxford.
Duncan, S., (1972), ‘Some signals and rules for taking speaking turns in conversations’,
Journal of personality and social psychology 23, 283–92.
——, (1974), ‘On the structure of speaker-auditor interaction’, Language in society 2,
161–80.
Easterling, P.E., (1982), Sophocles: Trachiniae, Cambridge.
Ebeling, H., (1885), Lexicon Homericum, Leipzig.
Edwards, M.J., (1999), Lysias. Five speeches (Speeches 1,12,19,22,30), Bristol.
Edwards, M. & S. Usher, (1985), Greek orators—I. Antiphon & Lysias, Warminster.
Elsness, J., (1997), The perfect and the preterite in contemporary and earlier English,
Berlin.
Emmott, C., (1997), Narrative comprehension. A discourse perspective, Oxford.
Enkvist, N.E., (1978), ‘Coherence, pseudo-coherence, and non-coherence’, in J.-O.
Östman (ed.), Cohesion and semantics, Åbo, 109–128.
——, (1985), ‘Introduction: coherence, composition, and text linguistics’, in N.E. Enkvist
(ed.), Coherence and composition: A symposium’, Åbo, 11–26.
——, (1990), ‘Seven problems in the study of coherence and interpretability’, in
U. Connor & A.M. Johns (eds), Coherence in writing: research and pedagogical
perspectives, Alexandria, 9–28.
 bibliography 271

Evelyn-White, H.G., (1914), Hesiod, Homeric hymns, Homerica, Cambridge, MA

(Loeb).
Evers-Vermeul, J., (2005), The development of Dutch connectives. Change and acquisition
as windows on form-function relations, PhD dissertation, University of Utrecht.
Faulkner, A., (2005), ‘Aphrodite’s aorists: attributive sections in the Homeric hymns’,
Glotta 81, 60–79.
Fenn, P., (1987), A semantic and pragmatic examination of the English perfect, Tübin-
gen.
Fleischman, S., (1990), Tense and narrativity: from Medieval performance to modern
fiction, Austin.
Fludernik, M., (1991), ‘The historical present tense yet again: Tense switching and
narrative dynamics in oral and quasi-oral storytelling’, Text 11, 365–98.
——, (1996), Towards a ‘natural’ narratology, London/New York.
——, (2000), ‘Genres, text types, or discourse modes. Narrative modalities and generic
categorization’, Style 34, 274–92.
Fournier, H., (1946), Les verbes ‘dire’ en grec ancien, Paris.
Fraenkel, E., (1950), Aeschylus Agamemnon, Oxford.
Fraser, B., (1999), ‘What are discourse markers?’, Journal of Pragmatics 31, 931–52.
Furley, W.D. & J.M. Bremer, (2001), Greek hymns (2 vols.), Tübingen.
Garvie, A.F., (1998), Sophocles Ajax, Warminster.
Genette, G., (1972), Figures III, Paris.
Gildersleeve, B.L., (1900), Syntax of Classical Greek. From Homer to Demosthenes,
New York.
Givón, T., (1980), ‘The binding hierarchy and the typology of complements’, Studies
in language 4, 333–77.
——, (1983), ‘Topic continuity in discourse: an introduction’, in T. Givón (ed.), Topic
continuity in discourse: a quantitative cross-language study, Amsterdam/Philadelphia,
1–41.
——, (1992), ‘The grammar of referential coherence’, Linguistics 30, 5–55.
Goodwin, W.W., (1889), Syntax of the moods and tenses of the Greek verb, London.
Goutsos, D., (1997), Modeling discourse topic: sequential relations and strategies in
expository text, New Jersey.
Goward, B., (2005), Aeschylus: Agamemnon, London.
Grice, H.P., (1975), ‘Logic and conversation’, in P. Cole & J.L. Morgan (eds), Syntax
and semantics. Vol. III: Speech acts, New York, 41–58.
Groeneboom, P., (1944), Aeschylus’ Agamemnon, Groningen.
Grosz, B.J. & C.L. Sidner, (1986), ‘Attention, intentions, and the structure of discourse’,
Computational linguistics 12, 175–204.
Gundel, J.K., N. Hedberg, R. Zacharski, (1993), ‘Cognitive status and the form of refer-
ring expressions in discourse’, Language 69, 274–307.
Hagège, C., (1974), ‘Les pronoms logophoriques’, Bulletin de la Société de Linguistique
de Paris 69, 287–310.
Haiman, J., (1978), ‘Conditionals are topics’, Language 54, 564–90.
Halliday, M.A.K. & R. Hasan, (1976), Cohesion in English, London.
Halliday, M.A.K., (1985), An introduction to functional grammar, Amsterdam.
Hannay, M. & C.H.M. Kroon, (2005), ‘Acts and the relationship between discourse
and grammar’, Functions of language 12, 87–124.
Hannay, M., (1983), ‘The focus function in a functional grammar: questions of contrast
and context’, in S.C. Dik (ed.), Advances in functional grammar (Publications in
language sciences 11), Dordrecht/Cinnaminson, 207–23.
——, (1985), ‘Inferrability, discourse-boundness, and sub-topics’, in A.M. Bolkestein,
C. Groot, J.L. Mackenzie (eds), Syntax and pragmatics in functional grammar (Func-
tional grammar series 1), Dordrecht/Cinnaminson, 49–63.
272 bibliography

——, (1990), Pragmatic functions: ‘The view from the V.U. pragmatic function assign-
ment and word order variation in a functional grammar of English’, Working papers
in functional grammar 38, 1–23.
Hartung, J.A., (1832–3), Lehre von den Partikeln der griechischen Sprache, Erlangen.
Havers, W., (1906), ‘Das Pronom der Jener-Deixis im Griechischen’, IF 19, 1–98.
Henderson, J., (1985), Aristophanes Lysistrata, Oxford.
——, (1998), Aristophanes: Clouds, Wasps, Peace, London (Loeb).
——, (2002a), Aristophanes: Lysistrata, Oxford.
——, (2002b), Aristophanes: Frogs, Assemblywomen, Wealth, London (Loeb).
Hermann, G., (1827), ‘Dissertatio de pronomine αὐτός’, in G. Hermann, Opuscula, I,
Leipzig, 308–42.
Hilton, J., (1989), ‘Temporal connectors in the narrative discourse of Cicero’, in
M. Lavency & D. Longrée (eds), Proceedings of the Vth colloquium on Latin linguistics
(Cahiers de l’institut de linguistique de Louvain 15), Leuven, 173–84.
Hoey, M., (1991), Textual interaction: an introduction to written discourse analysis,
London/New York.
Hoffmann, M.E., (1989), ‘A typology of Latin theme constituents’, in M. Lavency &
D. Longrée (eds), Proceedings of the Vth colloquium on Latin linguistics (Cahiers de
l’institut de linguistique de Louvain 15), Leuven, 185–96.
Hoogeveen, F., (1782), Doctrina particularum Graecarum (Recensuit, breuiauit et auxit
Christian Godofr. Schütz), Dessauiae/Lipsiae.
Hooper, J.B., (1975), ‘On assertive predicates’, in J.P. Kimball (ed.), Syntax and seman-
tics, vol. 4, New York, 91–124.
Hopper, R., (1992), Telephone conversation, Bloomington.
Hude, C., (1912), Lysiae orationes, Oxford.
Humbert, J., (19453), Syntaxe grecque, Paris.
Jacquinod, B. et al. (eds), (2000), Études sur l’aspect chez Platon, Saint-Étienne.
——, (1997), ‘Sur le rôle pragmatique de καίτοι’, in A. Rijksbaron (ed.), 131–50.
Jary, M., (2008), ‘The relevance of complement choice: a corpus study of ‘believe’’,
Lingua 118, 1–18.
Jebb, R.C., (1885), Sophocles. The plays and fragments. Part II: The Oedipus Coloneus,
Cambridge.
——, (1892), Sophocles: The plays and fragments. Part V: The Trachiniae, Cambridge.
——, (1896), Sophocles. The plays and fragments. Part VII: The Ajax, Cambridge.
Jefferson, G., H. Sacks, E.A. Schegloff, (1974), ‘A simplest systematics for the organiza-
tion of turn-taking for conversation’, Language 50, 696–735.
Johanson, L., (2000), ‘Viewpoint operators in European languages’, in Ö. Dahl (ed.),
Tense and aspect in the languages of Europe, Berlin/New York, 27–187.
Jong, I.J.F. de, & A. Rijksbaron (eds), (2006), Sophocles and the Greek language,
Leiden.
——, (1987), Narrators and focalizers: the presentation of the story in the Iliad, Amster-
dam.
——, (1991), Narrative in drama: the art of the Euripidean messenger-speech, Leiden.
——, (1997), ‘Γάρ introducing embedded narratives’, in A. Rijksbaron (ed.), 175–85.
——, (2001), A narratological commentary on the Odyssey, Cambridge.
Jucker, A.H., (1993), ‘The discourse marker well: a relevance-theoretical account’,
Journal of pragmatics 19, 435–52.
Kamerbeek, J.C., (1953, 19632), The plays of Sophocles: commentaries. Part I: The Ajax,
Leiden.
——, (1959), The plays of Sophocles: commentaries. Part II: Trachiniae, Leiden.
——, (1967), The plays of Sophocles: commentaries. Part IV: The Oedipus Tyrannus,
Leiden.
——, (1984), The plays of Sophocles: commentaries. Part VII: The Oedipus Coloneus,
Leiden.
 bibliography 273

Karttunnen, L., (1974), ‘Presupposition and linguistic context’, Theoretical linguistics

1, 181–94.
Kiparsky, P. & C. Kiparsky, (1970), ‘Fact’, in M. Bierwisch & K.E. Heidolph (eds),
Progress in linguistics, Berlin/New York, 143–73.
Klein, J., (1988), ‘Homeric Greek αὖ: a synchronic, diachronic, and comparative study’,
Historische Sprachforschung (= Zeitschrift für Vergleichende Sprachforschung) 101,
249–88.
Knott, A., J. Oberlander, M. O’Donnell, C. Mellisch, (2001), ‘Beyond elaboration: the
interaction of relations and focus in coherent text’, in T. Sanders, J. Schilperoort,
W. Sporren (eds), Text representation: linguistic and psycholinguistic aspects. Amster-
dam, 181–90.
König, E., (1991), The meaning of focus particles. London/New York.
——, (1998), ‘Toward a typology of intensifiers (Emphatic Reflexives)’, in B. Caron
(ed.), XVI Congrès international des linguistes (Paris, Juillet 20–25, 1997), Amster-
dam, 1–16.
——, (2001), ‘Intensifiers and reflexive pronouns’, in M. Haspelmath, E. König,
W. Osterreicher, W. Raible, (eds), Language typology and language universals, Berlin/
New York, 747–60.
Kovacs, D., (1994–2002), Euripides (6 vols.), Cambridge, MA / London (Loeb).
Kroon, C.H.M., (1992), ‘Particula perplexa’, in G. Bakkum et al. (eds), Pentecostalia,
Amsterdam, 53–64.
——, (1994), ‘Discourse connectives and discourse type: the case of Latin at’, in J. Her-
man (ed.), Linguistic studies on Latin (Studies in Language Companion Series 28),
Amsterdam/Philadelphia, 303–17.
——, (1995), Discourse particles in Latin: a study of nam, enim, autem, vero and at
(Amsterdam studies in Classical philology 4), Amsterdam.
——, (1997), ‘Discourse markers, discourse structure and functional grammar’, in
C.S. Butler et al. (eds), Discourse and pragmatics in functional grammar, New York,
17–32.
——, (1998), ‘Discourse particles, tense, and the structure of Latin narrative texts’,
in R. Risselada (ed.), Latin in use. Amsterdam studies in the pragmatics of Latin
(Amsterdam studies in Classical philology 8), Amsterdam, 37–61.
——, (2002), ‘How to write a ghost story? A linguistic view on narrative modes in Pliny
EP. 7.27’, in L. Sawicki & D. Shalev (eds), Donum grammaticum. Studies in Latin
and Celtic linguistics in honour of Hannah Rosén, Leuven, 189–200.
——, (2007a), ‘Discourse modes and the use of tenses in Ovid’s Metamorphoses’, in
R.J. Allan & M. Buijs (eds), 65–92.
——, (2007b), ‘Anaphoric reference in Latin epic. Discourse functions of referential
choices in Ovid’s Metamorphoses’ (paper given at the XIV colloquium on Latin
linguistics, Erfurt).
Kühner, R. & B. Gerth, (1898/1904), Ausführliche Grammatik der griechischen Sprache.
Zweiter Teil: Satzlehre (2 vols.), Hannover.
Labov, W., (1972), Language in the inner city, Philadelphia.
Lamb, W.R.M., (1930), Lysias. Text with an English translation (Loeb series).
Lang, E., (1988), Syntax und Semantik der Adversativkonnektive. Einstieg und Überblick.
Manuscript, Berlin, Zentralinstitut für Sprachwissenschaft.
——, (2000), ‘Adversative connectors on distinct levels of discourse: a re-examination
of Eve Sweetser’s three level approach’, in E. Couper-Kuhlen & B. Kortmann (eds),
cause, condition, concession, contrast, Berlin/New York, 235–56.
Lambrecht, K., (1994), Information structure and sentence form: topic, focus, and the
mental representations of discourse referents, Cambridge.
Langacker, R.W., (1991), Foundations of cognitive grammar, Vol. II: Descriptive appli-
cation, Stanford.
Lattimore, R. (1951), The Iliad of Homer, Chicago.
274 bibliography

——. (1967), The Odyssey of Homer, New York.

Leclerc, M.-C., (1993), La parole chez Hésiode, Paris.
Le Guern, M., (1986), ‘Notes sur le verbe français’, in S. Rémi-Giraud & M. Le Guern
(eds), Sur le verbe, Lyon, 9–60.
Lenk, U., (1998), Marking discourse coherence: functions of discourse markers in spoken
English, Tübingen.
Levinsohn, S.H., (1987), Textual connections in acts, Atlanta, GA.
Levinson, S.C., (1983), Pragmatics, Cambridge.
Lewis, D., (1979), ‘Scorekeeping in a language game’, Journal of philosophical logic 8,
339–59.
Lightfoot, D., (1975), Natural logic and the Greek moods, The Hague/Paris.
Lloyd, M., (20052), Euripides Andromache, Warminster.
Lloyd-Jones, H. & N.G. Wilson, (1990), Sophoclis Fabulae, Oxford.
Lloyd-Jones, H., (1991–1994), Sophocles (2 vols.), Cambridge, MA / London (Loeb).
Lombardo, S., (1993), Hesiod. Works and days and Theogony, Indianapolis/Cam-
bridge.
Longacre, R.E., (19962), The grammar of discourse, New York/London.
Lyons, J., (1977), Semantics, Cambridge.
——, (1995), Linguistic semantics, Cambridge.
MacDowell, D.M., (1971), Aristophanes Wasps, Oxford.
Magnien, V., (1922), ‘Emploi des démonstratifs chez Homère’, Bulletin de la Société
de Linguistique de Paris 23, 156–83.
Mann, W.C. & Thompson, S.A., (1988), ‘Rhetorical structure theory: toward a functional
theory of text organization’, Text 8, 243–81.
Marg, W., (1970), Hesiod. Sämtliche Gedichte, Zürich/Stuttgart.
Martinéz Vázquez, R., (1993), ‘Una interpretación estructural del perfecto griego
antiguo’, Revista española de lingüística 23, 87–94.
Mastronarde, D.J., (2002), Euripides Medea, Cambridge.
Mazon, P., (1928), Hésiode, Paris (Budé).
McCoard, R.W., (1978), The English perfect: tense choice and pragmatic inferences,
Amsterdam.
McKay, K.L., (1965), ‘The use of the Ancient Greek perfect down to the end of the
second century’, Bulletin of the institute of Classical studies 12, 1–21.
Mey, J.L., (20012), Pragmatics: an introduction, Oxford.
Minchin, E., (2001), Homer and the resources of emory. Some applications of cognitive
theory to the Iliad and the Odyssey, Oxford/New York.
Minton, W.W., (1970), ‘The proem-hymn of Hesiod’s Theogony’, TAPA 101, 357–77.
Moeschler, J., (1985), Argumentation et conversation. Éléments pour une analyse prag-
matique du discours, Paris.
Monro, D.B., (1891), A grammar of the Homeric dialect, Oxford.
Monteil, P., (1963), La phrase relative en grec ancien: sa formation, son développement,
sa structure, des origines à la fin du Ve siècle A. C., Paris.
Moorhouse, A.C., (1982), The syntax of Sophocles, Leiden.
Morgan, J.L., (1978), ‘Towards a rational model of discourse comprehension’. Theoretical
issues in natural language processing. Proceedings of the 1978 workshop on theoretical
issues in natural language processing, 109–114.
Mosegaard Hansen, M.B., (1996), ‘Eh bien: marker of comparison and contrast’, in
E. Engberg-Pedersen et al. (eds), Content, expression and structure: studies in Danish
functional grammar, Amsterdam, 315–42.
Most, G.W., (2006), Hesiod. Theogony, Works and days, Testimonia, Cambridge, MA
(Loeb).
Murray, A.T., (1976–1978), Homer. The Iliad, Cambridge, MA (Loeb).
Noël, D., (1997), ‘The choice between infinitives and that-clauses after believe’, English
language and linguistics 1, 271–84.
 bibliography 275

Olson, S.D., (2002), Aristophanes: Acharnians, Oxford.

Oréal, E., (1997), ‘Sur la fonction argumentative de quelques particules grecques’,
Lalies 17, 229–49.
Orriens, S.A.J., (2007a), Pasting past into present. A revised approach to the semantics
and pragmatics of the perfect in Classical Greek, (unpublished MA thesis. University
of Groningen).
——, (2007b), The Greek intensive perfect revisited, (unpublished MPhil essay. University
of Cambridge).
——, (2008), The Greek synthetic perfect in transition, (unpublished MPhil thesis.
University of Cambridge).
Page, D., (1972), Aeschyli septem quae supersunt tragoedias, Oxford.
Panhuis, D., (1984), ‘Prolepsis in Greek as a discourse strategy’, Glotta 62, 26–39.
Pfeijffer, I.L., (1999), First person futures in Pindar, Stuttgart.
Pinkster, H., (1972), On Latin adverbs, Amsterdam.
Places, É. des, (1929), Études sur quelques particules de liaison chez Platon. ΟΥΝ et ses
composés, ΑΡΑ, ΤΟΙΝΥΝ, Paris.
Polanyi, L. & R.J.H. Schwa, (1983), ‘The syntax of discourse’, Text 3, 261–70.
Prince, E.F., (1981), ‘Towards a taxonomy of given/new information’, in P. Cole (ed.),
Radical pragmatics, New York, 223–56.
Pucci, P., (2007), Inno alle Muse (Esiodo, Teogonia, 1–115), Testo, Introduzione, Tra-
duzione e Commento, Pisa-Roma.
Puddu, N., (2005), Riflessivi e intensificatori: greco, latino e le altre lingue indoeuropee,
Pisa.
Radermacher, L. & W. Kraus, (1954), Aristophanes’ ‘Frösche’, Vienna.
Redondo Moyano, E., (1996), Estudio sintáctico de las partículas en el periodo Helenístico,
Amsterdam.
Reichenbach, H., (1947), Elements of symbolic logic, New York/London.
Reinhart, T. & E. Reuland, (1993), ‘Reflexivity’, Linguistic inquiry 12, 657–720.
Revuelta Puigdollers, A.R., (1996), Estudio de las partículas griegas αὖ, αὖτε, αὖθις,
ἄψ, πάλιν y ὀπίσω en griego antiguo, PhD dissertation, Universidad Autónoma de
Madrid.
Rijksbaron, A., P. Stork, S.R. Slings, G.C. Wakker, (2000), Beknopte syntaxis van het
Klassiek Grieks, Lunteren.
——, (1976), ‘How does a messenger begin his speech? Some observations on the open-
ing-lines of Euripidean messenger-speeches’, in J.M. Bremer et al. (eds), Miscellanea
Tragica in Honorem J.C. Kamerbeek, Amsterdam, 293–309.
——, (1984), ‘Het Griekse perfectum: subject contra object’, Lampas 17, 403–19.
——, (1986), ‘Infinitivus en participium als complement in het Oudgrieks: het probleem
van ἄρχομαι en πειρῶμαι’, Lampas 29, 175–91.
——, (19942), The syntax and semantics of the verb in Classical Greek, Amsterdam.
——, (1995), ‘Euripides Bacchae 35–36’, Mnemosyne 48, 198–200.
——, (ed.), (1997), New approaches to Greek particles (Amsterdam studies in Classical
philology 7), Amsterdam.
——, (1997), ‘Introduction’, in A. Rijksbaron (ed.), 1–14.
——, (20023), The syntax and semantics of the verb in Classical Greek, Amsterdam.
——, (2006), ‘On false historic presents in sophocles (and Euripides)’, in I.J.F. de Jong
& A. Rijksbaron (eds), 127–49.
——, (20073), The syntax and semantics of the verb in Classical Greek: an introduction,
Chicago.
Rood, T., (2007), ‘Xenophon’, in I.J.F. de Jong & R. Nünlist (eds), Time in Ancient
Greek literature, Leiden, 147–63.
Roulet, E., et al., (1985), L’articulation du discours en français contemporain, Bern.
——, L. Fillietaz, A. Grobet, (2001), Un modèle et un instrument d’analyse de
l’organisation du discours, Bern.
276 bibliography

——, (1989), ‘Des dimensions argumentatives du récit et de la description dans le

discours’, Argumentation, 247–70.
Ruijgh, C.J., (1971), Autour de ‘τε épique’, Amsterdam.
——, (1985), ‘L’emploi ‘inceptif’ du thème du présent grec’, Mnemosyne 38, 1–61.
——, (1991), ‘Les valeurs temporelles des formes verbales en grec ancien’, in J. Gvoz-
danovic, & Th.A.J.M. Janssen (eds), The Function of Tense in Texts, Amsterdam,
197–217.
——, (1999), ‘Sur l’emploi complétif de l’infinitif ’, in B. Jacquinod (ed.), Les complétives
en grec ancien. Actes du colloque international de Saint-Étienne, septembre 1998,
Saint-Étienne, 215–31.
——, (2004), ‘Over de gebruikswijzen van het Griekse perfectum. Met speciale aandacht
voor Plato’s Politeia’, Lampas 37, 24–45.
——, (2006), ‘The use of the demonstratives ὅδε, οὗτος and (ἐ)κεῖνος in Sophocles’, in
I.J.F. de Jong & A. Rijksbaron (eds), 151–62.
Ruipérez, M.S., (1954), Estructura del sistema de aspectos y tiempos del verbo griego
antiguo. Analisis funcional sincrónico, Salamanca.
Sadoulet, P., (1984), ‘Les significations de αὐτός en grec’, Lalies 4, 61–5.
Sandt, R.A. van der, (1992), ‘Presupposition as anaphora resolution’, Journal of Seman-
tics 9, 333–77.
Sanford, A.J. & S. Garrod, (1981), Understanding written language: explorations in
comprehension beyond the sentence, Chichester.
Sargent, T., (1973), The Homeric hymns, New York.
Sauge, A., (2000), Les degrés du verbe. Sens et formation du parfait en grec ancien,
Bern.
Schegloff, E.A., (1987), ‘Recycled turn beginnings: a precise repair echanism in
conversation’s turn-taking organisation’, in G. Button & J.R.E. Lee (eds), Talk and
social organisation, Clevedon, 70–85.
Schiffrin, D., (1987), Discourse markers, Cambridge.
——, (2001), ‘Discourse markers: language, meaning and context’, in D. Schiffrin,
D. Tannen, & H.E. Hamilton (eds), The handbook of discourse analysis, Malden,
MA, 54–75.
Schwyzer, E. & A. Debrunner, (1950), Griechische Grammatik auf der Grundlage
von Karl Brugmanns griechischer Grammatik. Zweiter Band: Syntax und Stilistik,
Munich.
Seaford, R., (1984), Euripides Cyclops, Oxford.
Sells, P., (1987), ‘Aspects of logophoricity’, Linguistic inquiry 18, 445–79.
Shorey, P., (1928), ‘The pathos and humor of αὖ’, Classical philology 23, 285–7.
Sicking, C.M.J. & J.M. Van Ophuijsen, (1993), Two studies in Attic particle usage:
Lysias & Plato, Leiden.
Sicking, C.M.J., & P. Stork, (1996), ‘The synthetic perfect in Classical Greek’, in C.M.J.
Sicking, & P. Stork (eds), Two studies in the semantics of the verb in Classical Greek,
Leiden, 119–298.
——, (1997), ‘The grammar of the so-called historical present in Ancient Greek’, in
Bakker (ed.), (1997a), 131–68.
Sicking, C.M.J., (1991), ‘The distribution of aorist and present tense stem forms in
Greek, especially in the imperative’, Glotta 59, 14–43.
——, (1993), ‘Devices for text articulation in Lysias I and XII’, in C.M.J. Sicking & J.M.
Van Ophuijsen (eds), 1–66.
——, (1997), ‘Particles in questions in Plato’, in A. Rijksbaron (ed.), 157–74.
Simons, M., (2007), ‘Observations on embedding verbs, evidentiality and presupposi-
tion’, Lingua 117, 1034–56.
Slings, S.R., (1980), ‘KAI Adversativum: some thoughts on the semantics of coordina-
tion’, in D.J. van Alkemade et al. (eds), Linguistic studies offered to Berthe Siertsema,
Amsterdam, 101–25.
 bibliography 277

——, (1994), ‘Geschiedenis van het perfectum in het oud-Grieks’, in R. Boogaart &
J. Noordegraaf (eds), Nauwe betrekkingen. Voor Theo Janssen bij zijn vijftigste ver-
jaardag, Amsterdam, 239–47.
——, (1997a), ‘Adversative relators between PUSH and POP’, in A. Rijksbaron (ed.),
101–29.
——, (1997b), ‘Figures of speech and their lookalikes: two further exercises in the
pragmatics of the Greek sentence’, in Bakker (ed.), (1997a), 169–214.
——, (2002a), ‘Figures of speech in Aristophanes’, in A. Willi (ed.), The language of
Greek comedy, 99–109, Oxford.
——, (2002b), ‘Oral strategies in the language of Herodotus’, in E.J. Bakker, I.J.F. de
Jong, H. van Wees (eds), Brill’s companion to Herodotus, Leiden, 53–77.
Sluiter, I., (1997), ‘Parapleromatic lucubrations’, in A. Rijksbaron (ed.), 233–46.
Smith, C.S., (2002), ‘Perspective and point of view: accounting for subjectivity’, in
H. Hasselgård, S. Johansson, B. Behrens, C. Fabricius-Hansen (eds), Information
structure in a cross-linguistic perspective, Amsterdam, 63–79.
——, (2003), Modes of discourse. The logical structure of texts, Cambridge.
Smyth, H.W., (1920), Greek grammar, Cambridge, MA.
——, (1963), Aeschylus, London.
Sommerstein, A.H., (1982–98), The comedies of Aristophanes. (vols. 2–10), Warmin-
ster.
——, (2002a), ‘Comic elements in tragic language: The case of Aeschylus’ Oresteia’, in
A. Willi (ed.), The language of Greek comedy, Oxford, 151–68.
——, (2002b), Greek drama and dramatists, London.
Spenader, J., (2003), ‘Factive presupposition, accommodation and information’, Journal
of logic, language and information 12, 351–68.
Stahl, J.M., (1907), Kritisch-historische Syntax des griechischen Verbums der klassischen
Zeit, Heidelberg.
Stalnaker, R., (1974), ‘Pragmatic presuppositions’, in M. Munitz & P. Under (eds),
Semantics and philosophy, New York, 197–213.
——, (2002), ‘Common ground’, Linguistics and philosophy 25, 701–21.
Stanford, W.B., (1958), Aristophanes. The Frogs, London.
——, (1963, 19832), Sophocles. Ajax, Bristol.
Starkie, W.J.M., (1968), The Wasps of Aristophanes, Amsterdam.
Stede, M., (2004), ‘The Potsdam commentary corpus’, in B. Webber & D. Byron (eds),
ACL 2004 workshop on discourse annotation, Barcelona, 96–102.
Stevens, P.T., (1971), Euripides Andromache, Oxford.
Stoddard, K., (2004), The narrative voice in the Theogony of Hesiod, Leiden.
Stork, P., (1982), The aspectual usage of the dynamic infinitive in Herodotus, Groningen.
Taboada, M.T., (2004), Building coherence and cohesion, Amsterdam/Philadelphia.
Taillardat, J., (1987), ‘A propos de grec αὐτός: ipséité, identité, anaphore’, Lalies 5,
75–86.
Tanskanen, S.K., (2006), Collaborating towards coherence: lexical cohesion in English
discourse, Amsterdam.
Taplin, O., (1977), The stagecraft of Aeschylus, Oxford.
Thalmann, W.G., (1984), Conventions of form and thought in Early Greek epic poetry,
Baltimore/London.
Thrall, M.E., (1962), Greek particles in the New Testament, Leiden.
Tomachevski, B., (2001 [1965]), ‘Thématique’, in T. Todorov (ed.), Théorie de la lit-
térature: textes des Formalistes russes, Paris.
Towle, J.D., (1978), Parataxis and organization in the ‘Histories’, Ph.D. dissertation,
University of Washington.
Ussher, R.G., (1973), Aristophanes Ecclesiazusae, Oxford.
——, (1978), Euripides Cyclops, Rome.
278 bibliography

Van Ophuijsen, J.M., (1993), ‘οὖν, ἄρα, δή, τοίνυν: The linguistic articulation of argu-
ments in Plato’s Phaedo’, in C.M.J. Sicking & J.M. Van Ophuijsen (eds), 67–164.
Vendler, Z., (1957), ‘Verbs and time’, The philosophical review 66, 143–60.
Verdenius, W.J., (1972), ‘Notes on Hesiod’s Theogony’, Mnemosyne 25, 225–60.
Vuchinich, S., (1977), ‘Elements of cohesion between turns in ordinary conversation’,
Semiotica 20, 229–57.
Wace, A.J.B. & F.H. Stubbings, (1967 [1962]) A companion to Homer, London/Mel-
bourne.
Wackernagel, J., (1904), Studien zum griechischen Perfektum, Göttingen.
Wagnon, A., (1880), Le pronom d’identité et la formule du réfléchi dans Homère, dans
les poètes tragiques et chez les doriens, Genève.
Waltz, P., (1914), ‘Note sur la Théogonie’, REG 27, 229–35.
Wakker, G.C., (1994), Conditions and conditionals. An investigation of Ancient Greek
(Amsterdam studies in Classical philology 3), Amsterdam.
——, (1995), ‘Welaan, dan, dus, nu. Partikels in Sophocles’, Lampas 28, 250–70.
——, (1997), ‘Emphasis and affirmation: some aspects of μήν in Tragedy’, in A. Rijks-
baron (ed.), 209–31.
——, (2006), ‘ “You could have thought”: past potentials in Sophocles?’, in I.J.F. de Jong
& A. Rijksbaron (eds), 163–82.
Weinrich, H., (20016 [1964]), Tempus: besprochene und erzählte Welt, München (French
translation 1973: Le temps, Paris).
West, M.L., (1966), Hesiod. Theogony, Oxford.
——, (1987), Euripides Orestes, Warminster.
——, (1989), ‘An unrecognized injunctive usage in Greek’, Glotta 67, 135–8.
White, J.W. & M.H. Morgan, (1892), An illustrated dictionary of Xenophon’s Anabasis,
Boston.
Widdowson, H.G., (1978), Teaching language as communication, Oxford.
Willi, A., (2002a), ‘The language of Greek comedy: introduction and bibliographical
sketch’, in A. Willi (ed.), The language of Greek comedy, Oxford, 1–32.
——, (2002b), ‘Languages on stage: Aristophanic language, cultural history, and Athenian
identity’, in A. Willi (ed.), The language of Greek comedy, Oxford, 111–50.
Willink, C.W., (1986), Euripides Orestes, Oxford.
Ziv, Y., (1994), ‘Left and right dislocations: discourse functions and anaphora’, Journal
of pragmatics 22, 629–45.
 INDEX LOCORUM

aeschylus 893 121

A. 898–9 121
274–6 148 1152–3 147
1178 124 1324–5 119
Pers. 1482–3 119–20
226 146
233 146 euripides
Andr.
aristophanes 1073–1165 189–98
Av. 1166–7 116
893 121 Ba.
898–9 121 918 122
915–23 237 Cyc.
1461–3 123 488–94 115–6
Ec. El.
581–2 133 858 183n32
786–7 133 HF
Eq. 867 122
620–4 126 922–7 181
Lys. 956 193n58
1 142 995–9 178–9
65 118 Hipp.
76–7 119 1201–1208 178
909 129 1342 116
Nu. Ion
126 155 1146–58 180–1
738–9 224 1178–9 183
1063–7 125 1250–8 117
1351–4 125–6 1286–7 149
Pax Med.
178 118 386 132–3
665–9 236 1064–6 132
Ra. 1116–9 114
134 145 1160–6 193–4
256–9 145–6 1365–6 148
604 120 Or.
646–7a 129 132–9 96
927–8 146–7 1107–8 131
Th. 1214–15 131–2
164–5 45 1629–45 105
669–70 122 Ph.
885–6 223 1104–38 179n22
V. Supp.
164–5 149 726 182–3
169–73 152 Tr.
350–3 137 706–8 96n24
280 index locorum

herodotus 94 256
1.182.1–2 48 96–7 257
1.16.1 211 98–104 257–8
1.18.1–3 216–7 100–1 259
1.29.1 214 102 259
1.46.1 216 102–3 259
1.49 42 104 259–60
1.95.3 210 114 260
1.106.1 216 201–2 254n32
1.130.1 211
1.163.2–3 215 homer
2.30.3 212 Iliad
2.50.2–3 49 2.102–7 90
2.111.2 217 2.586–9 13
2.123.2 207 3.87–91 16n39
2.127.1 215 3.328–33 91n19
2.128.1 215 3.406–11 17–8
2.133.1 211 4.127–33 92n21
2.137.2 217 5.418–9 96n24
2.157.1 213–4 7.17–43 99
3.59.2 214 9.697–702 12
3.82.4–19 88 10.25–7 15
3.97.14–9 108 10.540 16
4.1.2 212 15.332–8 87n15
4.157.1 210 16.648 11n28
4.158.1 216 16.659–62 14
6.58.2 48 17.596–602 91
7.6.3 48–9 24.242–44 11
7.23.2–8 89n17 Od.
7.149.1 207 2.1–10 4–5
7.154.1 213 3.404–12 93
8.4.1–2 34 7.104–32 241–2
8.132.3 22 8.94–5 [= 533–4] 13–4
9.32.2 45 8.572–6 16n38
11.390 11
hesiod
Th. Hymni Homerici
1 242–3 4.20–102 250n32
2–4 243–4 6.14–6 249
5–9 244
9–10 245–7 lysias
22 248–9 1.5–6 63–4, 72
24 249 1.15–16 95
25 249 1.49 71n20
33–4 249–50 3.7 78
35 250 3.9 74
36 250–1 3.10 80
36–7 251–2 3.13 80
53 252n26 3.15 73
68–70 252–55 3.20 69
71 256 3.35 74–5
83–4 256 6.23 69
93 256 6.49 35n28
 index locorum 281

6.55 233 350c4–9 56

7.11 73–4 362d1–3 105
10.15 77 362e3–364a1 103n39
12.3–4 64 379b 155
12.7 67–8 388e2–4 53
12.47 71 397d4–9 56
12.100 233 421d1–2 101
13.1–3 77–8 471a1–3 57
13.43 33 508a9–b2 57
13.48 68–9 521e3–522a1 54
13.51 70n18 Smp.
13.55 68 201a3–7 56
14.47 69 202d1–4 54
17.5–6 238 218b3–c1 44
18.17–8 237–8 Sph.
19.2 70 220b4–6 53
19.60 73 227d4–11 46
22.1–2 64 255e11–256a6 55
24.4 70 261a6 43
30.24–25 22, 236 Ti.
21c2–4 45
plato 48a 52
Ap.
29b9–d2 60 sophocles
30c2–5 50–1 Aj.
30d 162n17 47–9 130
Chrm. 540–4 111
167e1–9 45–6 1142–58 234–5
Cra. 1168–9 113
385d5–9 54 Ant.
406e2 52 441–3 232
415a4–7 59 536–41 149–50
Cri. 1061–5 151–2
44a5–8 224 El.
Euthd. 78–9 120
279d7–8 34 OC
Euthphr. 31 114–5
5d7–9 38 549–50 113
Grg. 1248 114
470d8–e7 36 1579–80 39
517a1 35n28 OT
Lg. 343–6 124
647a8–c1 51–2 Tr.
688c1–5 43–4 227–9 150
792d7–e6 46–7 485–90 151n35
893e5 38
Phd. thucydides
116c4–8 37 1.3.3 162
Phdr. 2.97.6 162
253e5–255a1 257–8 3.82.6 162
R. 6.55.3 162
327b1–4 60 7.75.5–6 163
328b2–5 60 8.94.1–3 94
282 index locorum

xenophon Mem.
An. 1.4.18 30
1.5.3 242 1.6.13 31
1.5.6 242 1.2.24–6 87
1.8.27 35 4.2.16–17 100
1.9.31 35 4.25–5.1 102
1.10.16 35 Oec.
2.2.14–6 36 7–11 104
5.7.33 30 15 102
6.4.1–6 243 Smp.
6.6.17 22 1.15 159
HG. 2.14 159
2.1.21 243 4.38 160, 166n30
4.2.16–8 106 8.15 160–1
9.6 160
 GENERAL INDEX

adversativity, 140 general knowledge, 50–3, 57–9

anaphora, 2–4, 19, 24, 27n14, 34, 38 given information, see familiarity
aspect, 171n3, 172, 175, 186, 189, 198
imperfective/perfective, xix, 205, 207, historical present, see tense
209, 218 hymns, characteristics of, 241n5
continuative/discontinuative, xix,
218 imperfective/perfective, see aspect
perfect aspect, xix–xx, 221–39 implication, 140–2, 145–6, 148–53
asyndeton, 66–67, 71, 79–81, 165–6, 197 indirect reflexives, 15–6
attributive section, 244, 251, 252, 256 inferable information, see familiarity
infinitive, 21–3, 28, 29n17, 232–3
background, xii–xiii, 32, 174, 178, 192, information structure, 21, 23, 29n17, 39,
211,217–8, 228 208, 215, 216–7
intensifiers, 12, 14n33
cognitive verbs, 21, 23, 27–8, 31–3,
38n32, 39 levels of discourse, 66, 158n3, 167–8,
complementation, xv–xvi, 21–4, 26–40 218, 228–9
complexive aorist, see tense
continuative/discontinuative, see aspect meaning
correction, 140–2, 145, 148–9, 151–3 functional, 65
referential, 65
deictic pronouns, see pronouns mood, 185–6
dialogical, xviii, 155–69, 187n42, 231–2 ‘Mousogony’, 251–2
dialogue, viii, 26–7, 135, 138–9, 142n20, Muses, the
145, 147, 149, 151, 154, 184n33, 188, birth of–, 252
190, 231–38 encounter of Hesiod with–, 241, 248,
diaphony, xviii, 159–63, 184–5 250, 254
discourse marker, 61, 135–9, 152–4, movements of–, 241, 245, 247, 254
165n26
narrative mode, viii–xix, 162n22, 171–99
empathy, 15, 16n37 narrative past, xix, 174, 219
excursus, 85n10, 92–6, 107–9 narrative structure, 34–5, 171–99, 205
expectation, 50–8, 140, 146, 177, 184 negation, 141–2, 145, 147–8, 150, 176–7,
extratextual relationships, xix, 221–39 180, 184–5
new information, see familiarity
factive verbs, 22–7, 31, 32n22, 33n25
familiarity, xv–xvi omnitemporal present, see tense
given/old information, 24, 25n10, 27,
34, 40, 47–61 particle
inferable/contextually prepared connective, 66–81, 164–9, 175n15
information, 40, 47–61 pop particle, 41, 58, 66, 69, 72, 75,
new information, 24, 25n10, 28, 36, 40, 80–1, 95
47–8, 54, 57 push particle, 41, 66, 107n42, 191n52
flash-back, 212, 218 ‘situating’ particle, 228–30
‘focalising’ imperfect, see tense presupposition
foreground, 32, 174, 179, 217, 218, 228 pragmatic, xvi, 21–9, 32, 33n25, 34–9
284 general index

semantic, xvi, 21, 25, 28, 29n16, given/accesible topic, 85, 86–91, 92,
33n25, 39 96, 98, 108, 109
pronouns new topic, 85, 89, 92n20, 96–8, 100,
deictic, 4n8, 6n11, 8, 249 105, 109
third person, 1–8, 12, 17–9 promoted topic, 85n10, 89–90, 109
prototype theory, xvii, 81, 185 resumed topic, 85, 92–6, 109
(speaking) turn, xviii, 135–54, 158–9,
reference point, 206, 209, 211–2, 213, 161, 232–38
215–8
‘referent-in-the-mind’ model, 2–3, 18 visual field, 13–4
relevance, 70, 81, 183n32, 225–6
rhetorical question, 160, 185, 196, 197, ἀλλά, xviii, 135–54
198 αὐτός, xv, 1, 6–9, 12–17, 19
γάρ, xvi, 30, 41–61, 63, 173, 176, 184,
speech past, xix, 219 186, 191n52, 192
state of affairs, 205 γὰρ οὖν, xvi–xvii, 38, 41–61
non-transformative, 206 γε, 168, 175, 184, 186
γε μήν, xviii, 166, 168n36
tense, 172–5, 177, 179, 183, 186–90, 192, δέ, 66, 80–1, 173, 186
198 δή, 59n18, 127–8, 232
complexive aorist, 244, 248 ἤδη, 128–33
‘focalising’ imperfect, 245–7, 261–2 καί, 127–8, 173
historical present, 173–9, 186, 188, καὶ δή, xvii, 37, 111–33
192–7 καὶ μήν, xvii, 111–28
omnitemporal present, 243–4, 246, κεῖνος, 1, 6–8, 10–2, 18–9
247, 251, 259 μέντοι, 161, 176, 184
perfect tense, xix–xx, 221–39 μήν, xviii, 140, 145–6, 158–64, 176
third person pronouns, see pronouns ὅτι-clause, 21, 28–9, 31–2, 36, 38–40
topic(icality), 25n10, 30–1, 176n15, 179, οὖν, xviii, 41–61, 63–81, 173, 176, 184,
180, 181, 196 186, 198
closing topic, 85n10, 99, 102–5, 107, τοίνυν, xvi–xvii, 63–81, 176
109 ὡς, 21n2, 23, 31–3, 196