MODALS

AND MODALITY
IN ENGLISH

Carlos INCHAURRALDE BESGA

Universidad de Zaragoza

1. INTRODUCTION

Modality in English has traditionally been interpreted in terms of the use of

modals, and although this is not the only resource available for the expres-
sion of this notion,1 there is no doubt that it is the most important one. Still,
modality as a whole is not a clear area of study, for several reasons, the
most important being the fact that we can identify two different kinds of
modality: "root" (deontic) modality (dealing with obligation, permission,
ability, etc.) and epistemic modality (dealing with probability, possibility,
certainty, etc.). This distinction has been studied in the past, and an
interesting suggestion is that epistemic uses are dependent on, and derive
from, deontic ones (cf. Sweetser 1982, 1990, who gives a unified
treatment in terms of force dynamics and causality).2 From a diachronic
point of view, it is clear that epistemic modality derives from "root"
modality, and we shall elaborate on this below. Diachrony can also be
understood from a language learning perspective: children acquire the
deontic senses of modal verbs earlier than the epistemic ones (as
mentioned by Sweetser 1982: 485, who refers to Kuczaj and Daly 1979
and Shepherd 1981). Synchronically, "root" and epistemic modalities are
related by means of a subsumption relation which will also be discussed
here.
However, modality as a notion also needs to be examined in
connection with tense and aspect. As will be shown, epistemic modality
can be represented in a compact way together with tense in a graph which
has time as one axis and possible worlds as the other. There is strong
evidence that tense and modality are related: both are categories that are
encoded in predications at the same level of depth, and both clearly interact
with each other. This will also be looked into. If we consider tense, we
shall also need to consider aspect, which deals with the internal
configuration of time as it is expressed in verbs. The three categories tense,
aspect, and modality are expressed mainly by auxiliaries; there is a great
deal of crosslinguistic evidence that the three of them are closely
interrelated (cf. Givón 1984: 269-318).
In order to analyse the way in which this interrelation takes place, we
are going to use two axes: the diachronic one (relations through historical
time) and the synchronic one (relations in the system at a given moment).

2. THE DIACHRONIC AXIS

2.1. Heine's account of the grammaticalization process of

auxiliaries

Modals in English (can, could, may, might, must, will, would, shall,
should, ought and need) are considered to be auxiliary verbs with a high
degree of grammaticalization which is reflected in some morphosyntactic
characteristics, such as the fact that they have no -s endings for the third
person singular, and no infinitives, or past forms (except the forms could
and would in some instances). In addition, many of them also have weak
and contracted forms.
However simple their morphosyntactic functioning has become, a
corresponding simplification and reduction of their meaning has not taken
place. Let us take for instance the modal can. With this verb we can form
sentences like the following:

[1] If you want you can do it (theoretical possibility)

[2] Zaragoza can be very warm in Summer (characteristic behaviour of
reality)
[3] You can leave the classroom now (permission)
[4] I can swim (ability)
[5] We can meet tomorrow (suggestion)
[6] Can I help you? (offer)
[7] Can you come here a minute, please? (order)

We can see that there are many different possible senses for this verb
([5], [6], and [7] are pragmatically motivated, but [1], [2], [3] and [4] are
clearly different meanings of the same modal). It could be argued that this
is not a good example, since there are other modals which are more
grammaticalized. Such is the case of will or would. But, still, we can
distinguish different senses, as we can see in these examples for will:

[8] It will rain tomorrow (Prediction in the future)

[9] There is somebody coming. That'll be Peter (Prediction in the pre-
sent: possibility)
[10] He will keep forgetting everything (habitual behaviour)
[11] I won't open the door (refusal; willingness not to do something)
[12] You'll leave the classroom at six (order)
[13] Will you open the door, please? (request; order)

This broad spectrum of possibilities for the different modals is a direct

consequence of a historical process of grammaticalization with several
stages. When the process stops at different points, different senses appear.
Some of these senses are closer to the etymological origin (root modality),
whereas others are nearer the pure epistemic modality scale in terms of
possibility, probability or certainty.
Heine (1993) has put forward a very clear explanation for this phe-
nomenon in cognitive terms. According to him, and drawing on extensive
crosslinguistic data, there are several basic event schemas which underlie
possible processes of grammaticalization for auxiliaries. They provide the
basic concrete representations from which all known languages make the
shift towards abstract concepts like tense, aspect, and modality distinctions.
As basic event schemas he proposes the following:

a. Location (i.e. where one is)

b. Motion (where one moves to, from, through, etc.)
c. Activity (what one does)
d. Desire (what one wants)
e. Posture (the way one's body is situated)
f. Relation (what one is like, is associated with, or belongs to), or
g. Possession (what one owns)

(Heine 1993: 28)

All these event schemas have a very clearly identifiable conceptual

form which helps us identify the constructions that use them:

Conceptual form Proposed label

a."X is at Y" Location

 b. "X moves to/from Y" Motion
c."X does Y" Action
d. "X wants Y" Volition
e."X becomes Y" Change-of-state
f. "X is (like) a Y" Equation
g."X is with Y" Accompaniment
h. "X has Y" Possession
i. "X stays in a Y manner" Manner

(Heine 1993: 31)

These basic schemas, together with a few derived ones (serial,

evaluative and purpose), can be the etymological source for auxiliaries
marking tense, aspect or modality. This is shown in the following table:

Location progressive, ingressive, continuous

Motion ingressive, future, perfect, past
Action progressive, continuous, ingressive, completive, perfect
Volition ingressive, future
Change-of-state ingressive, future
Equation resultative, progressive, perfect, future
Accompaniment progressive
Possession resultative, perfect, future
Manner progressive

(Heine 1993: 47)

The following are examples of how these schemas have evolved in dif-
ferent European languages:

[14] I am going to play golf tomorrow (motion schema > future)

[English]
[15] I do work very often with Mary (action schema) [English]
[16] You will go to the cinema (volition schema > future) [English]
[17] Anda diciendo por ahí que Juan es un incompetente (manner
schema > progressive) [Spanish]
[18] Hij is een boek aan het lezen (location schema > progressive)
[Dutch; example from Heine 1993]
[19] Bernd wird kommen (change-of-state schema > future) [German]
[20] Hoy comeré [comer-he] pronto (possession schema > future)
[Spanish]

Now, if we look at the way in which these schemas have given rise to
modals in English, it is interesting to note that verbs with full semantic
content have become emptied of their semantic load. This phenomenon has
also run parallel to a shift in grammatical category, word-class,
morphosyntactic properties and even phonetic form. In a schematic way,
we can say that modals have followed the shift from being in a structure
[X-verb-complement] to a structure of the form [X-grammatical concept-
main verb] (as postulated for all kinds of auxiliaries in Heine 1993: 47).
Heine calls this shift the Verb-to-TAM3 chain, in which several processes
work at the same time: Desemanticization, decategorialization,
cliticization, and (phonetical) erosion (Heine 1993: 54-58). This is well
conceptualized by means of 'overlap' models, such as the following:

Overlap model of conceptual shift:

Stage: I II III
---------------------------------------------------------------------------------
---
Type of
concept: Source Source Target
Target
 Overlap model of morphosyntactic shift:

Stage: I II III
---------------------------------------------------------------------------------
---
Morphosyntax: Verbal Verbal Grammatical
Grammatical

Overlap model of erosion:

Stage: I II III
---------------------------------------------------------------------------------
---
Phonological
form of Full Full Reduced
expression: Reduced

(Heine 1993: 49-51)

Heine's account is extremely interesting as a model for the formation

of current English modals. He gives even more details about the different
stages in the four processes mentioned above, and specifies up to five steps
in the process of decategorialization. Looking at the way in which these
different steps take place together he postulates the existence of seven main
stages in the Verb-to-TAM chain, which he denotes by the letters A to G:

Stages in the Verb-to-TAM chain:

Overall stage A B C D E F G
----------------------------------------------------------------------------------
-
Desemanticization I II III
Decategorialization I II III IV V
Cliticization I II III
Erosion I II III

(Heine 1993: 58)

2.2. English modals and the grammaticalization process

If we try to fit the evolution of English modals into this general schema and
look for the stages that correspond to some of them, we find that they are
in stage E in most of their uses, which means that, according to the
specifications provided by Heine (1993: 54-56) for every stage, they have
the following properties:

Desemanticization: "The subject is no longer associated with will-

ful/human referents, and the verb acquires a grammatical function" (III).
This is clearly seen with will, the most desemanticized of the English
modals. However, there are some instances of the use of this verb as a full
semantic form in fossilized expressions such as "do as you will".

Decategorialization: "The verb loses further verbal properties such as

its ability to be negated separately and to occur in other positions of the
clause, and the complement loses in nominal (and adverbial) properties,
such as its nominalizing and/or adverbial morphology" (IV). All modals in
English need to be attached to the main verb and have lost most of their
verbal properties. This is a very characteristic feature of English modals,
which all grammars identify as peculiar to them. Other languages retain
verbal morphology for some of these auxiliaries (e.g. Spanish poder and
French pouvoir, German können and Dutch kunnen, etc.)

Cliticization: "The verb loses its status as a separate word and

develops into a clitic. The verb and its complement are now likely to form
a simple phrase, which permits only one expression of tense, negation,
passivization, etc." (II). The rule for negation in English clearly takes into
account the relative position of the modal and the main verb, with only the
form not inserted between them. Tense itself is marked by the modal.
When this is not possible, the verb have is used (e.g. "he must be there
now" vs. "he must have been there") adding perfectivity in the process.

Erosion: "The phonological substance of the verb tends to be eroded"

(II). This is clear in written English with the modals will and would, where
contractions like 'll, 'd, are common. With other modals, unstressed forms
are not uncommon. The tendency to fuse with not in negations also has a
written manifestation: can't, won't, shouldn't, wouldn't, etc.

In general terms, English modals tend to be very homogeneous as far

as their decategorialization and cliticization stages are concerned. But they
are less so for desemanticization and erosion, which tend to go parallel.
Some senses are closer to etymology, and they are expressed phonetically
in a less eroded way. This means that they did not travel all the way to
stage E of Heine's table without leaving some instances of less
grammaticalized uses unharmed.

3. THE SYNCHRONIC AXIS

3.1. Introduction: The TAM system and layers in FG

In the preceding section, we have seen that there exists a clear, systematic
account of how different senses of the same modals are etymologically
related. However, their coexistence in a given moment is something that
needs a synchronic formulation. Owing to the fact that modality is closely
related with tense/aspect in most linguistic descriptions, it is fair to treat
them here together, as constitutive parts of the so-called TAM (Time-
Aspect-Modality) system. With this in mind we could start by referring to a
grammatical model in which tense, aspect and modality have provided
evidence for a complex, multilayered description of the clause: S. C. Dik's
Functional Grammar (from now on, FG).
In FG, following suggestions by Hengeveld (1989), which were in turn
based on the layered structure of the clause proposed in Foley and Van
Valin's (1984) RRG,4 utterances can be analysed in terms of a multilayered
structure with this form:

LAYERS
Clause: (E1 : [ILL (S) (A) (x1 : etc. (x1 ))] (E1))
Proposition: (x1 : [(e1: etc. (e1))] (x1 ))
Predication: (e1: [Pred (x 1)n ] (e1 ))
Term: (x1 : PredN (x1 ))

(Dik and Hengeveld 1990: 3)

There are operators (p) for each of the four levels, so that the total
representation would be something like
 (E1 : [4 ILL (S) (A) (3 x1 : (
2 e1: [1 Pred(x1 : pred(x 1)) ...
(xn )] (e1 )) (x1 ))] (E1))

1 : Predicate operators 
3 : Proposition operators
2 : Predication operators 4 : Illocution
operators

(Dik and Hengeveld 1990: 2)

Van Valin's original proposal found justification for its tripartite divi-
sion of clause structure in evidence from expressions for the TAM system.
This is also the case with FG; we shall therefore look at how FG treats
TAM questions in its description, in order to see what can be used for an
adequate treatment of modality interactions.

3.2. Tense and aspect

Tense and aspect are traditionally considered to deal with the expression of
time. Tense concerns its "external" configuration, that is, the distribution of
events along a temporal line, whereas aspect concerns its "internal"
structure, that is, how time is organized inside a situation. Tense, according
to Hengeveld (1989: 132), is represented by level-2 2 operators. This is
so because it tends to be further away from the verb nucleus than aspect
(level-1 and level-2 operators; see below). Aspect, however, is a more
complicated matter. Aspectuality encompasses many distinctions that are
categorized under different labels in different linguistic traditions. In
Functional Grammar, according to Dik (1989: 186-187), who reserves the
term "aspect" only for those aspectuality distinctions which are
grammatically rather than lexically expressed, aspectuality covers the
following sub-areas:

(a) The type of SoA [State of Affairs] as designated by the

predicate frame . . . also called Aktionsart (Mode of action). . . .
(b) Perfectivity/Imperfectivity. . . .
(c) Phasal aspectuality distinctions serve to describe what can be
said at some reference point on the temporal dimension, in relation
to the occurrence of some SoA. . . .
(d) Quantificational aspectuality distinctions express different
forms of quantification over sets of occurrences of SoAs. . . .

(a) concerns aspectuality expressed lexically, which means that it is

not to be treated as aspect in FG. As for (b), (c) and (d), they have
operators at two different levels, because they function at different levels
of the clause. This can be contrasted empirically in different languages.
The opposition perfectivity/imperfectivity and phasal aspect are
treated as level-1 1 operators. Dik (1989) notes that there is very frequent
interaction of these operators with a certain SoA. Conflicts between
imperfectivity and telic SoAs, or between perfectivity and open-ended
SoAs tend to resolve by assigning certain interpretations (conative,
iterative or distributive for the former; ingressive or terminative for the
latter). As for phasal aspect, it is expressed through different
interpretations of the perfective and imperfective (cf. Dik 1989: 187 ff.;
Siewierska 1991: 118-120).
Quantificational aspect is treated as a level-2 2 operator for several
reasons: it can quantify over any element of a core predication5 and it can
be specified independently of the other kinds of aspect (cf. Dik 1989: 204
ff.; Siewierska 1991: 121-122).
3.3. Modality

Modality receives a very complex treatment in Functional Grammar. Dik

(1989: 205), following Hengeveld (1987, 1988), mentions the following
sub-areas of modality:

Level 1: Inherent modality, which defines the relation between a

participant and the realization of the SoA in which he is involved. It
concerns ability or willingness, obligation, and permission.

Level 2: Objective modality, which expresses the speaker's evaluation

of the likelihood of occurrence of the SoA. It can be divided into two sub-
areas with a gradation of the degree of actuality involved. These are:
Epistemic objective modalities (Certain-Probable-Possible-Improbable-
Impossible) and Deontic objective modalities (Obligatory-Acceptable-
Permissible-Unacceptable-Forbidden).

Level 3: Epistemological modality. Here we have modal distinctions

signalling the speaker's personal commitment to the truth of the
proposition. They are: Subjective modality, in which the speaker takes
personal responsibility for the content of the proposition, and signals how
certain he is about its truth; and evidential modalities, in which the speaker
assesses the quality of the proposition according to how he has obtained it,
be it through evidence, by personal experience, or by having heard it from
someone else.

As far as English modals are concerned, only objective modality will

be considered here. In fact, this is the proper type of notion that
corresponds to modality, according to most authors. Inherent modality
covers modal distinctions such as ability, willingness, obligation,
permissibility, and volition. However, Siewierska remarks that

these distinctions are realised lexically . . . , not grammatically.

Hengeveld (1987: 11-12) suggests that there is also a semantic dif-
ference between inherent and objective modality in that by means
of the former speakers merely present their knowledge of a given
situation, while by means of the latter they offer an evaluation of
the situation in terms of this knowledge. For many linguists, this
difference excludes inherent modality from the proper domain of
modality. (1991: 124)

As for subjective modalities, they are statements of opinion rather

than fact, and Siewierska reminds us that, in English, "the modal
auxiliaries are open to an objective and a subjective reading" (1991: 126).
The distinction between "root" and epistemic modality (cf. Sweetser 1982),
made on etymological grounds, corresponds broadly to the distinction
between deontic and epistemic modalities. Root modality is closer to the
etymological meaning, which has to do with evaluative judgement in many
cases. Other authors also agree with this distinction, although they define
modality in a way sometimes misleadingly similar to FG modality varieties
for level 3 or 1. This is the case with Givón's (1993: 169) definition, in
which, if we assume that modality has to do with the speaker's attitude
towards a proposition, this attitude concerns two types of judgement:

(a) Epistemic judgements of truth, probability, certainty, belief or evi-

dence.
(b) Evaluative judgements of desirability, preference, intent, ability,
 obligation or manipulation.

Of the utmost interest is the fact that, regardless of their concrete value
as judgements, all these possible modalities can easily be explained in
logical terms by means of a possible worlds approach.6 Modality expressed
logically in terms of possible worlds (certainty and probability) has been
extended in order to interpret obligation/permission (deontic logic), and
knowledge/belief (epistemic logic); all these different logical models
account for the opposition realis/irrealis, which Givón (1993: 172 and ff.)
uses to denote a characteristic feature of modality. The logical approach
therefore provides us with an instrument which explains deontic modality
as a special variety of epistemic modality, the subsumption of the former
by the latter being plausible in the kind of explanatory schema suggested
below. Obligation implies certainty and permission possibility; in a similar
way, knowledge implies certainty and belief possibility.

3.4. Interactions

It has been mentioned in the previous sections that both FG objective

modality and tense are considered to be in level 2 of the multilayered
structure of the clause. We shall argue here that this is due to the fact that
both tense and modality interact in a coordinated way, which in turn
implies that a compact representation for both of them can be found. This
representation can be made in a simple way by means of a two-axis schema
in which the vertical axis represents time and the horizontal axis possible
worlds or possible states of affairs (let us call it the hypothesis/reality axis).
For a start, it has been suggested that the following correlations
between tense-aspect and epistemic modality show high predictability
(Givón 1993: 171), which is a proof of interaction between the time axis
and the hypothesis axis:

(a) Past ===> R-assertion (or presupposition)

(b) Perfect ===> R-assertion (or presupposition)
(c) Present ===> R-assertion
(d) Future ===> IRR-assertion
(e) Habitual ===> IRR-assertion

All this is congruent with a branching-path perspective at which we

also arrive using a combination of temporal and possible-worlds logic (cf.
van Benthem 1988: 32). The present is closed and comes from a single
path of events that are already fixed in the past; but the future is open and
there are parallel possible paths. Their number and "width," so to speak, is
limited only by a certain perceived speed of change of things in reality. A
good model for representing this would be a conceptual analog of
Minkowski's light cones.
Minkowski's light-cones have been used as a convenient
representation of what happens with the light travelling from distant
objects in relativistic physics. If we look at fig. 1, in which the horizontal
axis represents distance and the vertical one time, and we imagine
ourselves to be at the PRESENT point, the light that comes from objects
inside the PAST cone can reach us at this point. The light of objects
outside the cone will not reach us at this moment, but only later on in time,
because the speed of light is fixed and there is too much distance for too
little time. The same happens with our light, which will reach all objects
inside the FUTURE cone but none outside it. We only have to substitute
hypothetical distance for physical distance and perceived speed of changes
in reality for speed of light, to use this representation for the tense-modality
complex in natural language.7 In this case, both PAST and FUTURE cones
have limits that change according to the experienced speed of change in
reality (we need to be careful about this convention: personal experience is
highly subjective and it may be the case that speaker and hearer do not
agree on it, which leads to not-completely-shared knowledge about
possible paths of events).

High
Impossibility Low possibility FUTURE
possibility

PRESENT

Impossibility Possibility in Actuality PAST

the past
Fig. 1

The dynamics of this representation is similar to that of Langacker's

dynamic evolutionary model (1991: 277), an idealized cognitive model
with three main components: (1) the structured world model (the world is
structured in a certain way, which motivates the possibility of some
situations and events, but the impossibility of others), (2) the elaborated
epistemic model (in which reality evolves in a certain way, but only a
limited portion of reality is known to the conceptualizer), and (3) some
force-dynamic concepts necessary to account for the evolution of events.
However, the PAST and FUTURE cones we use here seem to us a more
realistic diagrammatical representation of the tense-modality complex than
Langacker's cylinders8 (1991: 242, 244, 277). They clearly show the fact
that, for a given point in time, the farther one goes into the past/future, the
larger the number of possible alternative preceding/following situations
one can find.
We can add to this representation a three-parameter general scheme
similar to Reichenbach's three-parameter system for tense. In his analysis
(Reichenbach 1947), verbal tenses can be classified according to the
following temporal points: time of event (E), time of reference (R), and
time of speech (S). If we follow an order relation (denoted by <; time
overlap is denoted by =), we have the following combinations:

Present (R = E = S)
Past (R = E) < S
Future (R = S) < E
Present Perfect E < (R = S)
Past Perfect E<R<S
 Future PerfectS < E < R

The three parameters can now be used not only as points in temporal
space, but also as points in hypothetical space. Thus, the expression of hy-
pothesis can be compactly explained together with time. If we look at fig. 1
again, we can see that in the PAST cone there is an actual path (certainty),
some other paths that could have been possible since they could have led
to the PRESENT point (possibility), and an area outside the cone in which
no path could have led to that point (impossibility). In the FUTURE cone
the picture is pretty much the same, although, as we are now in the irrealis
mode, there is no certainty path but only a highest possibility one. A
convenient way of stating three parameters for modality would be to
consider at least three degrees of "reality" (certainty, possibility, and
impossibility or uncertainty). Again, we can set three parameters: reality of
event (Er), reality of reference (Rr), and reality of speech (Sr). The reality
of event stands for the degree of reality that the main event has; the reality
of reference may be the certain condition or assumption that is used for
assigning realis or irrealis status to the event; finally, the reality of speech
has to do with the utterance itself.
Some examples of how this representation is to be applied are the fol-
lowing:

1. Certainty (Unmarked schema: [Rr = Er = Sr])

shall, shall not, will, will not, must, cannot, could not, would, would not

Future + Certainty Near future + Certainty

(Highest possibility): (Highest possibility)

[21] I shall see you tomorrow.

[22] I'll see you in a minute.

Present + Certainty Past + Certainty:
(Highest possibility):

[23] That can't be John - he's in Dublin.

[24] I knew it couldn't be John.

 Future of the past + Certainty:
[25] This child would one day rule all England.

2. Probability (Unmarked schema: [Rr = Er < S r])

should, should not, ought to, ought not to, may (not)

Proximal future + Probability (High possibility):

[26] He should be here soon.
 3. Weak probability (Unmarked schema: [Er < Rr < S r])
might, might not, could

Future + Low probability:

[27] I might see you again next year - who knows.

4. Theoretical or habitual possibility

(Unmarked schema: [Rr = Er < S r])
can

[28] Zaragoza can be very warm in September.

This possibility can be well understood in terms of a structured world

model (Goldsmith and Woisetschlaeger 1982; Langacker 1991: 264). 9

5. Conditional certainty or possibility

(Unmarked schema: [Rr < Er = S r]; also [R r < (E r < S r)])
would, would not, could, could not, might, might not

[29] If I were a rich man I would buy a yacht.

[30] If John came we could all go home.

Conditionals, whether clearly marked by a suitable subordinator or

not, involve shifting the deictic center (cf. Langacker 1991: 266-269). In
the three-parameter schema proposed here the reality of reference has a
higher degree of uncertainty. If this condition were to become true, Er and
Sr should also become true (in a basic conditional; if we use modals like
"may," "might," "can," or "could," we change their degrees of reality
somewhat).

4. CONCLUSION

To conclude, we can say that epistemic modality in terms of a certain-

probable-possible-impossible-uncertain scale can conveniently be
considered to be the most basic kind of modality which subsumes others,
and this is clearly seen from both a diachronic and a synchronic
perspective. If we analyse modal auxiliaries as a characteristic word class
for the expression of modality in English, it is clear that from a diachronic
point of view there has been a process of grammaticalization in which their
semantics has evolved from the concrete to the abstract. Heine's account
(1993) is a good model which shows how constructions for the expression
of concrete event schemas have gradually been desemanticized and eroded
to end up as expressions for TAM notions in English, as well as in many
other languages. From a synchronic point of view, epistemic modality is
the basic reference kind of modality: it interacts directly with tense and can
therefore be combined as complementary notions in any expression using
modal auxiliaries. The schema proposed as a descriptive representation has
a very strong resemblance to Minkowski's time-space cones used in
relativistic physics and it is a simple and clear model which allows for
more detailed analysis using lineal algebra methods. These methods could
perhaps help in the future to compute relationships between tense and
modality. We all know that language analysis is not as straightforward and
direct as we would like it to be, but there is no doubt that this kind of study
could reveal interesting potential phenomena for further study.a

NOTES

1. Dirven points out that "there is a long tradition in the descriptive grammar writing
of English which concentrates solely or predominantly on the modal auxiliaries, excluding
the other expressions of modality" (1989: 60). We acknowledge the fact that this is true;
moreover, there are even expressions in which modals themselves use adverbials to grade
their meaning. Consider, for instance, the following sequence of expressions for epistemic
modality:

will certainly + V 100%

will almost certainly + V
will probably + V
may well + V
may/will possibly + V
might + V
probably will not + V
certainly will not + V 0%

Our consideration of only modal auxiliaries here does not imply that there are not other
resources for the expression of modality.

2. It is customary in research on modals and modality to refer to the work carried out by
authors like Palmer (1979, 1986), Twaddell (1963), and others. The interested reader may go
to these sources for more information. Basically, Sweetser (1990) and Heine (1993) give a
very up-to-date (although less extense) account of this topic, enriched with a more cogni-
tively-oriented perspective.

3. TAM is a widely accepted acronym for "Time-Aspect-Modality."

4. Foley and Van Valin propose three levels: the nucleus (predicate and predicate
operators - aspectual inflections and adverbials), the core (nucleus and verbal arguments, as
well as some modal operators), and the periphery (adjuncts, tense, subjective markers,
evidential modalities, and indicators of illocutionary force).

5. A core predication has the form [[1pred (arg)n ] (1)n ] in FG, corresponding to a
level-1 structure.

6. Basically, in a possible worlds semantics we have a model M=<W, R, V> in which

W stands for a set of "possible worlds," R for a relation of "accessibility," and V for a
valuation. It is not my intention to explain this kind of semantics in more detail, since it is a
very well-known area in logical linguistics. For references, vid. Van Benthem (1988: 15).

7. Van Benthem (1988: 36) makes a connection with tense logic systems and also refers
to a partial tense logic system developed by Goldblatt (1980), but the use of these cones for
the study of modality in natural languages is unknown to me.

8. Langacker's representation is a tridimensional one. We follow a similar criterion: the

horizontal dimension has a central point of maximal probability, and the values decrease the
farther we move from this central point on two dimensions (this makes a total of three with
the vertical axis). As is clear in the figure, the result is two cones.This representation is the
same as the original by Minkowski for relativistic physics.

9. A structured world model assumes that the world is structured in a certain way and
that some events are incidental whereas others are regular and predictable. When it is the
case that we are referring to manifestations of the normal course of events, then we may do
so by using strong probability or habituality.

WORKS CITED

VAN BENTHEM, J. 1988. A Manual of Intensional Logic. Stanford: CSLI.

DIK, S. C. 1989. The Theory of Functional Grammar. Part I: The Structure of the
Clause. Dordrecht: Foris.
DIK, S. C., and K. HENGEVELD. 1990. "The Hierarchical Structure of the Clause
and the Typology of Perception Verb Complements." WPFG 37. Amsterdam:
U of Amsterdam.
DIRVEN, R. 1989. "Cognitive Linguistics and Pedagogic Grammar". In Linguistic
Theorizing and Grammar Writing. Ed. G. Leitner and G. Graustein.
Tübingen: Niemeyer. 56-75.
FOLEY, W. A., and R. D. VAN VALIN. 1984. Functional Syntax and Universal
Grammar. Cambridge: Cambridge UP.
GIVóN, T. 1984. Syntax. A functional-typological introduction. Vol. I.
Amsterdam: John Benjamins.
- - -. 1993. English Grammar: A Function-Based Introduction I. Amsterdam: John
Benjamins.
GOLDBLATT, R. 1980. "Diodorean Modality in Minkowski Space-Time." Studia
Logica 39: 219-236.
HEINE, B. 1993. Auxiliaries: Cognitive Forces and Grammaticalization. Oxford:
Oxford UP.
HENGEVELD, K. 1987. "Clause Structure and Modality in Functional
Grammar." In Ins and Outs of the Predication. Ed. Johan van der Auwera and
Louis Goossens. Dordrecht: Foris.
- - -. 1989. "Layers and Operators." Journal of Linguistics 25: 127-157.
KUCZAJ, S. A., and M. J. DALY. 1979. "The Development of Hypothetical
Reference in the Speech of Young Children." Journal of Child Language 6.
LANGACKER, R. W. 1987. Foundations of Cognitive Grammar. Vol. I.
Theoretical Prerequisites. Stanford: Stanford UP.
- - -. 1991. Foundations of Cognitive Grammar. Vol. II. Descriptive Application.
Stanford: Stanford UP.
PALMER, F. 1979. Modality and the English Modals. London: Longman.
- - -. 1986. Mood and Modality. Cambridge: Cambridge UP.
REICHENBACH, H. 1947. Elements of Symbolic Logic. New York: Academic
Press.
SHEPHERD, S. 1981. Modals in Antiguan Creole, Child Language Acquisition,
and History. PhD. dissertation. Stanford.
SIEWIERSKA, A. 1991. Functional Grammar. London: Routledge.
SWEETSER, E. E. 1982. "Root and epistemic modals: causality in two worlds."
Proceedings of the Annual Meeting of the Berkeley Linguistics Society 8: 484-
507.
- - -. 1990. From Etymology to Pragmatics: Metaphorical and Cultural Aspects of
Semantic Structure (Chapter 3: "Modality"). Cambridge: Cambridge UP.
TWADDELL, W. F. 1963. The English Verb Auxiliaries. Providence (RI): Brown
UP.