3.

Syntactic Constructions
Peter Svenonius
CASTL, University of Tromsø– The Arctic University of Norway ∗

Table of contents
1. Introduction
2. Examples of constructions
3. Constructions and syntactic theory
4. Conclusion
5. References (selected)

Abstract
The term construction is widely used descriptively in discussing grammar, and is
still used informally in most theoretical work for characteristic formal patterns of
syntactic categories or features, usually associated with a meaning and/or function.
Modern linguistic theories employ a range of formal devices to produce or charac-
terize surface constructions; they may be rules, or schemata, or constraints. It is
usually assumed that competence in a language consists largely of these formal de-
vices together with a lexicon; the constructions themselves are epiphenomenal. As
such, constructions are an abstraction over the data which linguistic theory must
analyze; insight in syntax is achieved through discovering generalizations across
constructions.

1. Introduction
The term construction is ubiquitous in contemporary syntactic literature, being
used informally to refer to linguistic expressions in a variety of ways. The term
also has a technical sense in the theory of Construction Grammar, as detailed in
Chapter 28.
The term construction is widely used to characterize certain kinds of form–
meaning pairings, as when we refer to “possessive constructions” or “the verb-
particle construction” to refer to examples like those below.
∗
Published in 2015 in Syntax—Theory and Analysis: An International Handbook, Wal-
ter de Gruyter, Berlin, HSK Series 42.1: 15–23. Thanks to editors Tibor Kiss and Artemis
Alexiadou, and to an anonymous reviewer for comments on an earlier draft.

1
(1) Three examples of possessive constructions
a. Seymour’s new friend
b. a new friend of Seymour’s
c. Seymour has a new friend

(2) Three examples of the verb-particle construction

a. We picked up a lamp at the flea market.
b. We picked a lamp up at the flea market.
c. What did you pick up at the flea market?

In general, linguists would not refer to the three examples in (1) as comprising
a single possessive construction, because they are too different in their syntax;
in (1a) the possessor precedes the possessed noun, in (1b) the possessor follows
the possessed noun, and in (1c) the possessor is expressed as a distinct argument
outside the possessed noun phrase. These differences represent three different ways
of expressing the concept of possession, in English.
In (2), on the other hand, many linguists would be inclined to refer to all three
sentences as manifesting a single verb-particle construction, on the basis of the
perceived similarity of the syntax of the three cases. There is a very large class
of verb–particle pairings which allow the ordering alternation shown in (2a)–(2b),
where the order reflects no apparent difference in meaning (such as drop off, smash
up, fix up, turn on, leave out). In such cases, the object can systematically be the
focus of a question, as in (2c); so the general consensus would be that these three
sentences illustrate the verb-particle construction.
On this view, (1) illustrates three different form–meaning pairings, even though
one component of the meaning is shared across all three, while (2) illustrates one
form–meaning pairing, even though independent factors distort the shared form
(and the correct characterization of the meaning component may be elusive).
At the same time, (2c) illustrates a wh-question construction, in addition to
the verb-particle construction. Since the properties of the wh-question construc-
tion (e.g. wh-expression in clause-initial position, auxiliary in second position) are
independent of the verb-particle construction (e.g. the predicate includes a par-
ticle like up, down, out, etc.), there is no motivation for formulating a distinct
“verb-particle wh-question construction.”

1.1. Toward a definition

The term construction is not a technical term (outside of Construction Grammar),
and consequently it is difficult to define. As an approximation, it can be defined
as follows.

(3) A construction is a characteristic formal pattern of syntactic categories or

features, usually associated with some meaning and/or discourse function.

2
The use of the word pattern here is an attempt to be as theory-neutral as possible;
a pattern might be a structure, or a template, or the output of a rule. The notion
formal is meant to include aspects of form which are of significance to grammar.
In some theories word order is a primitive of grammar, while in other theories word
order is derived from structure, such that structure, but not linear order, would
count as a formal property (see Chapters 18 and 40 on Word Order).
The notion syntactic categories is intended to include major parts of speech
but also minor or functional categories such as the class of English verb-particles,
or the class of determiners. The notion syntactic features in the definition is meant
to include morphosyntactic features such as the past participle or dative case but
also semantico-syntactic features such as negation or “wh” (borne by interrogative
expression like what and who). Together syntactic categories or features includes
function words such as infinitival to and bound formatives such as possessive ’s,
on any analysis.
The definition in (3) is meant to exclude phonology and surface exponence,
which do not characterize constructions as the term is ordinarily used in main-
stream syntax. For example, we would not expect to find a construction which
necessarily involved words beginning with the phoneme /w/, even if we speak
loosely of various kinds of wh-constructions. Similarly, if there is more than one
formal category in English which is spelled out by the suffix -ing, then we expect
a construction to be identified by the underlying features which are being spelled
out (e.g. progressive, or gerund), not by the phonological form of the exponent
doing the spelling out. Though we might descriptively call something an Acc-ing
construction, for example, in a more careful statement of its characteristics, we
would distinguish the feature or category that -ing manifests.
Thus, the definition offered above is intended to stress syntactic form, not
phonological form. This is in accord with the usual use of the word construction
in syntax. An idiom like kick the bucket meaning ‘die’ requires the lexical items
kick and bucket, and hence makes direct reference to exponents with phonological
content (see Chapter 41 on Idioms). As such, ordinary usage would not make
reference to a kick the bucket construction (Construction Grammar, however, is
different here: idioms are considered to be constructions).
In this way, more or less functional elements like the interrogative pronoun
what and the light verb do are treated together with syntax as opposed to lexical
items like kick and bucket. Thus is it not controversial to speak of a construction
of the form What’s X doing Y? meaning roughly ‘Why is X Y?,’ where X is a
subject and Y is a predicate. For example: What’s the newspaper doing in the
bushes? or What are you doing leaving without your shoes? This construction
requires what and do as well as the progressive with an appropriate form of be.
The definition offered in (3) also suggests that a construction is usually associ-
ated with some meaning and/or discourse function. The importance of meaning is

3
somewhat loosely applied in practice. Thus, it is not usually considered necessary
to have a rigorous statement of the meaning of a possessive construction like the
one in (1a) in order to call it a construction, if it has a clearly defined syntactic
form. But if there are two disjoint meanings involved, then it is common to think
of them as involving two distinct constructions. For example, in English, the aux-
iliary inverts with the subject when a wh-item is fronted, but also when a negative
element is fronted, as in (4) (cf. Chapter 13, on V2).

(4) a. Which of them would he recommend?

b. None of them would he recommend.

Even if the syntax of the inversion is identical in the two cases, it would be most
natural here to speak of subject-auxiliary inversion constructions in the plural,
rather than a single subject-auxiliary inversion construction which was indifferent
to whether the initial element was a negative or an interrogative phrase—though
practice varies somewhat here (and in Construction Grammar, there is no limit to
how abstract a construction can be).
The notion of discourse function in (3) is intended broadly, to include various
pragmatic inferences and affect. For example, the What’s X doing Y? construction
is only used when there is some sense that it is incongruous or inappropriate for X
to be Y (as discussed by Kay and Fillmore 1999). Thus the question Why are men
rebelling? can be asked in a range of contexts, but What are men doing rebelling?
can only be asked if there is some salient sense (perhaps the speaker’s opinion, but
not necessarily) that it is inappropriate, incongruous, or outrageous for them to
be doing so.

1.2. Applying the definition

Returning to the examples in (1)–(2), we can apply the definition offered in (3)
to show that it is harmonious with the common intuition that (1) illustrates three
different possessive constructions while (2) represents three different manifestations
of a single construction.
According to (3), a class of phrases or sentences must share a characteristic
formal pattern in order to belong to a single construction. The formal differences
among the three examples in (1) are fairly clear; (1a) lacks of and the possessor
precedes the possessum, while (1a) contains the function word of, and the posses-
sor follows the possessum. Furthermore, in (1a), the possessum is understood as
definite, while in (1b), the possessum is indefinite. The example in (1c) is predica-
tive, and contains the verb have. So the fact that the three expressions describe
the same semantic relation is not normally taken to imply the existence of a single
possessive construction. Thus it seems that the definition appropriately picks out
each of the three as a construction.
Turning to (2), we can first address the question of whether (2c) represents
a different construction from the other two. Of course it does, as it involves

4
wh-movement, but this is irrelevant to the verb–particle construction. The inter-
rogative construction simply applies to a clause that has a verbal particle in the
predicate, just as it does in an ordinary transitive clause.
The second question is whether there is motivation to treat examples like (2a)
and (2b) as distinct constructions. This cannot be conclusively determined without
formal analysis. By and large, the two are distinguished only by word order, not
by meaning, nor by functional categories or features. There are some differences,
for example (2b) allows the object to be pronominal (We picked it up), but (2a)
does not (*We picked up it). If such differences can be independently explained,
then an analysis can be motivated in which there is a single construction with
some flexibility of order, that is, a single ‘characteristic formal pattern of syntactic
categories or features’ in which whatever determines the placement of the particle
before or after the internal argument is not characteristic, or is not a syntactic
feature. This is the usual consensus (see Ramchand and Svenonius 2002 for one
such account), though alternative analyses exist in which the two represent distinct
constructions (see Farrell 2005).

2. Examples of constructions
To further illustrate the notion of construction, several examples of constructions
are listed in this section. No detailed characterization or analysis is attempted.
The examples serve simply to give an impression of a small part of the range of
syntactic constructions.

2.1. Argument structure, grammatical functions

A number of constructions involve valency, argument structure, and grammatical
functions (see Chapter 9 on Grammatical Relations). For example, a passive
construction involves the demotion of the external argument, compared to the
active use of the same verb (see Chapter 23 on Voice). In English, the demoted
external argument can be expressed in a by-phrase or left implicit, and the verb
appears in a past participle form, with a form of the auxiliary be. The implicit
argument can control a purpose clause, as illustrated in (5b), just as with the
active construction in (5a). English also has a middle construction, in which the
external argument cannot appear in a by-phrase, and is not syntactically active as
diagosed by a purpose clause, as illustrated in (5c).

(5) a. The owner sold the house (to pay off debts).
b. The house was sold (by the owner) (to pay off debts).
c. Houses sell easily (*by the owner) (*to pay off debts).

In a conative construction, illustrated in (6b), the internal argument is oblique

and is not as fully affected as in a regular transitive construction, compare (6a).

5
In a benefactive construction, illustrated in (6c), an indirect object derives some
benefit from the action, or comes into possession of the direct object (see Chapters
10 and 37 on Arguments and Adjuncts).

(6) a. The baker cut the bread.

b. The baker cut at the bread.
c. The baker cut me a piece of bread.

Resultative and depictive constructions involve secondary predicates, as illustrated

in (7a) and (7b) respectively.

(7) a. They drank the bar dry.

b. They ate the bread dry.

Control, raising, and so-called Exceptional Case Marking (ECM, or accusative-

with-infinitive, or subject-to-object raising) constructions usually involve infinitive
complements in which arguments are, descriptively speaking, shared across the two
clauses (see Chapters 15 and 38 on Control).

(8) a. Ian wants to stay at the house. (control)

b. Ian seemed to stay at the house. (raising)
c. We believed Ian to stay at the house. (ECM)

English provides many more examples of constructions involving various configu-

rations of arguments, and other languages provide yet more. In some cases, the
availability of a construction may be tied to the availability of a lexical item; for
example, it may be that if a language has an ECM construction if and only if it has
a verb with the selectional properties exhibited by English believe in (8c). How-
ever, in other cases, the availability of a construction in a given language does not
seem to be connected to lexical items. An example is the resultative construction
illustrated in (7a), which many languages lack, despite having verbs and adjec-
tives which are otherwise like the ones used in the English resultative construction
(Snyder 2001).

2.2. Unbounded dependencies

There is a range of constructions which have been analyzed as involving displaced
constituents, or filler-gap dependencies. These include (in the order in which
they appear in [9]): wh-questions, embedded wh-questions, clefts, pseudoclefts,
relative clauses, comparative constructions, and (contrastive) topicalization (for
some discussion of some of these and related constructions, see Chapter 13 on V2,
Chapter 14 on Discourse Configurationality, and especially Chapter 22 on Relative
Clauses and Correlatives).

(9) a. Which book did you read?

6
 b. I asked which book you read.
c. It was this book that I read.
d. What I read was this book.
e. The book that I read was long.
f. She read a longer book than I read.
g. This book, I read.

These constructions have in common that the dependency can cross finite clause
boundaries.

(10) a. Which book did your mother think you read?

b. I asked which book your mother thought you read.
c. It was this book that my mother thought I read.
d. What my mother thought I read was this book.
e. The book that my mother thought I read was long.
f. She read a longer book than my mother thought I read.
g. This book, my mother thought I read.

In this respect they contrast with passive, as illustrated in (11a), and for example
raising, as illustrated in (11b) (compare [12]).

(11) a. *The house was thought (by his mother) the owner sold.
b. *Ian seemed stayed at the house.

(12) a. It was thought the owner sold the house.

b. It seemed Ian stayed at the house.

In fact, the constructions in (9)–(10) can cross indefinitely many finite clause
boundaries, and for this reason are known as unbounded dependencies.

(13) a. Which book did you think your mother thought you read?
b. I asked which book you thought your mother thought you read.
c. It was this book that I thought my mother thought I read.
d. What I thought my mother thought I read was this book.
e. The book that I thought my mother thought I read was long.
f. She read a longer book than I thought my mother thought I read.
g. This book, I thought my mother thought I read.

The availability and properties of unbounded dependency constructions can vary

somewhat from one language to the next. For example, sometimes there is a re-
sumptive pronoun in the gap position, and in other cases there is no displacement
on the surface, with the filler element remaining in situ. Unlike the case with ar-
gument structure alternations, this kind of variation tends not to be dependent on
lexical items, though it may be connected to the properties of functional elements
such as that and which.

7
 In fact, it has been proposed that properties of constructions are largely de-
termined by the properties of their heads. This is explicit in the name of the
theory Head-Driven Phrase Structure Grammar (see Chapter 27 on HPSG), but
is also a common assumption in other theories (e.g. Borer 1984). For example,
the properties of a relative clause could conceivably be entirely determined by the
cluster of features contained on its (possibly abstract) head. The head would be
a kind of C[omplementizer] taking a finite TP complement, attracting a suitable
nominal element to its specifier, and projecting a category which could be used as
a nominal modifier.

2.3. Complex constructions

In this regard, constructions like the What’s X doing Y? construction mentioned
above are complex, as they appear to involve an interdependency of several heads.
Mainstream theory would probably treat a construction of this kind as a special
kind of idiom.
Construction Grammar, in contrast, holds that there is no principled difference
between a fully general construction and a highly idiomatic one, or even between
a fully general abstract construction and a lexical item; each is a pairing of form
with function, broadly construed.

3. Constructions and syntactic theory

Traditional descriptive grammars may characterize and exemplify a list of con-
structions in a given language. They may organize the constructions according to
perceived similarities, and may attempt to state generalizations which transcend
the individual constructions. This much is extremely useful in a reference grammar
(see Chapter 61 on Reference Grammars).
Modern syntactic theory necessarily goes further, and is based on the assump-
tion that higher-level generalizations are necessary in order to achieve what Chom-
sky (1964) called explanatory adequacy, a model of language which accounts for
how individual languages are learnable by children.
Traditional grammar posited rules to characterize or generate constructions,
such as passive and relative clauses. Generative grammar took this as a starting
point and went on to abstract properties from classes of rules, such as elementary
transformations and different kinds of formal conditions constraining them. Gen-
erative grammar in the 1970’s explored the ways in which the properties interacted
in rule systems, and sought to discover constraints on them.
For example, Emonds (1976) observed that transformational rules did not pro-
duce structures unlike those which had to be base-generated (“structure preserva-
tion”), suggesting that the full power of transformations was unneeded. Chomsky

8
(1977) showed that the class of unbounded dependencies displayed highly uniform
properties, suggesting that they were not produced by distinct rules.
Subsequent work has increasingly focused on higher-level generalizations over
rule types, shedding much light on the nature of grammar.
Thus, generalizations about constructions involving valence and argument struc-
ture led to those being analyzed in Lexical-Functional Grammar as the output of
lexical rules (see Chapter 25 on LFG), accounting, for example, for their structure-
preserving properties and their relative sensitivity to lexically listed traits of indi-
vidual verbs. Similarly, generalizations about the class of unbounded dependen-
cies as a whole led to the development of Move-α in Government-Binding Theory
(Chomsky 1981), and the SLASH feature as used in Head-Driven Phrase Structure
Grammar (see Chapter 27 on HPSG).
The development of Principles and Parameters theory (Chomsky 1981) in-
volved rethinking the nature of rules entirely; once deconstructed into a system
of invariant universal principles interacting with parametric points of variations,
there are no rules per se. This is expressed in the following quote from Lectures
on Government and Binding: “The notions “passive,” “relativization,” etc., can
be reconstructed as processes of a more general nature, with a functional role in
grammar, but they are not “rules of grammar”” (Chomsky 1981: 7).
Since constructions were the output of rules in the traditional conception of
grammar, the elimination of rules from the theory means that in a Principles
and Parameters framework, constructions are epiphenomenal, as reflected in the
following quote, also from Chomsky but a decade later: “A language is not, then, a
system of rules, but a set of specifications for parameters in an invariant system of
principles of Universal Grammar (UG); and traditional grammatical constructions
are perhaps best regarded as taxonomic epiphenomena, collections of structures
with properties resulting from the interaction of fixed principles with parameters
set one or another way” (Chomsky 1991: 417).
Since that time, although the notion of parameter has been substantially
rethought, mainstream syntactic theories continue to regard the notion of con-
struction, like the notion of rule or transformation, as a descriptive stepping stone
on the path to greater understanding rather than as an analytic result in its own
right.
Work in Construction Grammar, too, recognizes that insight does not come
from simply listing the individual surface constructions in a language, and there-
fore, like other theories, seeks generalizations over constructions; the difference
between Construction Grammar and other theories mentioned here is that in Con-
struction Grammar, the generalizations are themselves modelled as abstract con-
structions. Nor is this just a terminological distinction: the claim in Construction
Grammar is that the generalizations have the same kinds of properties as the
constructions themselves, at a suitable level of abstraction.

9
4. Conclusion
I have characterized a construction as a characteristic formal pattern of syntactic
features, usually associated with some meaning or discourse function. The way
constructions are characterized in descriptive work tends to combine morphologi-
cal, syntactic, and semantic facts. Purely phonological facts, however, tend not to
be significant for the characterization of constructions. The precision of the syn-
tactic characterization is usually taken to be more important than the semantic
description of the meaning of a construction (e.g. even if two possessive expressions
had exactly the same meaning, substantially different syntaxes would usually lead
to them being classified as distinct constructions).
Some such notion is descriptively indispensible in syntactic work. Descriptive
grammars must contain detailed characterizations of surface properties of construc-
tions, and all careful empirical work makes reference to numerous construction
types, as will be seen in the other chapters in this work.
Linguistic theory advances through the careful examination of the properties
of constructions. Ultimately, principles are sought which transcend the individual
constructions themselves, and the construction itself can be taken to be a taxo-
nomic artifact.

5. References (selected)
Borer, Hagit. 1984. Parametric Syntax . Foris, Dordrecht.

Chomsky, Noam. 1964. Current Issues in Linguistic Theory. Mouton, The Hague.

Chomsky, Noam. 1977. On wh-movement. In Formal syntax , edited by Peter

Culicover, Thomas Wasow, and Adrian Akmajian, pp. 71–132. Academic Press,
New York.

Chomsky, Noam. 1981. Lectures on Government and Binding. Foris, Dordrecht.

Chomsky, Noam. 1991. Some notes on economy of derivation and representation. In

Principles and Parameters in Comparative Grammar , edited by Robert Freidin,
pp. 417–454. MIT Press, Cambridge, Ma.

Emonds, Joseph E. 1976. A Transformational Approach to English Syntax: Root,

Structure-Preserving, and Local Transformations. Academic Press, New York.

Farrell, Patrick. 2005. English verb-preposition constructions: Constituency and

order. Language 81 1: 96–137.

Kay, Paul and Charles J. Fillmore. 1999. Grammatical constructions and linguistic
generalizations: The What’s X doing Y? construction. Language 75 1: 1–33.

10
Ramchand, Gillian and Peter Svenonius. 2002. The lexical syntax and lexical
semantics of the verb-particle construction. In Proceedings of WCCFL 21 ,
edited by Line Mikkelsen and Christopher Potts, pp. 387–400. Cascadilla Press,
Somerville, Ma.

Snyder, William. 2001. On the nature of syntactic variation: Evidence from com-
plex predicates and complex word-formation. Language 77: 324–342.
Peter Svenonius, Tromsø (Norway).

11