Cognitive-Functional Approaches

to Syntax
Constructions in a Usage-Based
Model of Language

DeLancey (2001): Functionalism and Syntax

... there is no such thing as "functionalist syntax" in the
sense that there is "generative syntax", since a generativist
assumes ex-hypothesi that there is a distinct syntactic
component in Universal Grammar for "syntax" to be the
study of.
Functionalists ... find explanations in function, and in
recurrent diachronic processes which are for the most part
function-driven. ... they see language as ... a set of tools,
whose forms are adapted to their functions, and thus can
be explained only in terms of those functions.
Formal principles can be no more than generalizations over
data, so that most Generative explanation seems to
functionalists to proceed on the dormitive principle.
Why does opium makes you sleep? Because it
contains a substance with a dormitive principle
Why do nouns move? Because they need to get case.

Introducing Constructions as the

Objects of Grammar
Constructions are linguistic units that necessarily

have some non-compositional semantics:

They are learned pairings of form with semantic or

discourse function.

In other words, constructions have some aspect

of meaning that is not reducible or predictable

from its component parts or other constructions
Constructions are argued to be the central notion
in any theory of grammar.

Examples of English Constructions

Morphemes
Word
Complex word
Complex word (partially filled)
Idiom (filled)
Idiom (partially filled)
Covariational conditional
Passive

e.g. pre-, -ing

e.g. avocado, anaconda
e.g. daredevil, shoo-in
e.g. [N-s] (for regular plurals)
e.g. going great guns, give the Devil
his due
e.g. jog <someones> memory,
send <someone> to the cleaners
The Xer the Yer (e.g. the more you
think about it, the less you understand)
Subj aux VPpp (PPby)
(e.g. the armadillo was hit by a car)
Constructions are patterns that occur
frequently and are predictable and patterns
that are infrequent and unpredictable.

Construction Grammar and Generative

Grammar commonalities
The idea that language is a cognitive (mental)
system
The idea that there must be a way to combine
structures to create novel utterances
(generative power)

Construction Grammar and Generative

Grammar differences
Constructionist approaches say when you study
formal structures, you must take their semantic and
discourse functions into consideration.
Constructionist approaches do not dismiss semiregular and cross-linguistically unusual patterns as
peripheral
Generative approaches look only at the core language
(AKA I-language)

Constructionist approaches say people are born with

general cognitive processes that can be used to help
them learn languagenot knowledge that is specific
to language (rejection of UG hypothesis)

Generative Grammar View of Constructions

Constructions are epiphenomenal
They are the result of an interacting set of universal, fixed
principles with parameters selected on a languageparticular basis (Chomsky, 2000).

Grammatical constructions are taxonomic artifacts,

useful for informal description perhaps but with no
theoretical standing" (Chomsky, 2000).
According to Chomsky, "the search for explanatory
adequacy requires that language structure must be
invariant, except at the margins" (Chomsky, 2000,
emphasis added).
What does Generative Grammar want to explain?
What does Construction Grammar want to explain?

Constructions and Coercion Effect

Constructions explain how certain expected semantic
anomalies do not materialize.
Give me some laptop!
Give me some pillow!

How can you possibly read the

semantic contribution of one token
within a construction?

Elements in some sentences are expected to clash by

virtue of their incompatible semantics, based on their
distribution outside the construction.
Notice that in Generative Syntax, by virtue of the
sharp distinction between syntax and semantics,
clash in derivation is, well, clash in derivation (i.e.
movement and structural constraints)

Constructions Override Lexical Meaning

Constructions fill in semantic substance and overcome
semantic incompatibility of component parts through
coercion
I slept my way across the Atlantic.
Sleep: lack of motion specification
Sentence as a whole: describes motion with concomitant
sleeping

pit the cherries, dust the furniture, bone the filet

conventionalized semantic elements added: motion,
directionality

Verbs and Argument Structure

Constructions
Generative Assumption: The form and
general interpretation of basic sentence
patterns are determined by semantic
and/or syntactic information specified by
the main verb.
Give requires 3 arguments: agent, recipient,
theme
Chris gave Pat a ball
Put requires 3 arguments: agent, theme,
location
Pat put the ball on the table

Problems With That Assumption

Verbs can have more than one
argument structure construction.
He sneezed.
He sneezed his tooth right across town.
We laughed.
We laughed our conversation to an end.
Cognitive Linguistics assumes that every
semantic element is organize in a network of
knowkledge, and that produces massive
polysemy effects.

Polysemy again: to slice

He sliced the bread (transitive)
Pat sliced the carrots into the salad. (caused
motion)
Pat sliced Chris a piece of pie. (ditransitive)
Emeril sliced and diced his way to stardom. (way
construction)
Pat sliced the box open. (resultative)
In all of these, the meaning of the verb slice does not
change. The argument structure constructions
provide the direct link between surface form and
general aspects of the interpretation.

A Cognitive Grammar Analysis of

RAISING AND TRANSPARENCY

Types of Raising Constructions

Subject to Object Raising (SOR)
I expect [David criticize this plan].
I expect David [to criticize this plan].

Subject to Subject Raising (SSR)

[David criticize this plan] is likely.
David is likely [to criticize this plan].

Object to Subject Raising (OSR)

[David criticize this plan] is easy.
This plan is easy [for David to criticize].

Constituency in Cognitive Grammar

Constituency is observed in symbolic assemblies
(constructions) when a composite structure at
one level of organization functions in turn as
component structure with respect to a higher
level.
Syntactic trees represent:

Constituency (hierarchical grouping)

Grammatical categories
Linear expression
Levels of grammatical representation (through
movement operations)

Relational predications
(V, ADJ, P, ADV) are
inherently hierarchical.
Grounding relations are
different, but they are
represented through the
same kind of syntactic
constituency.

Symbolic Assemblies are is more accurate and

incorporate the same information as tree
structures:
(i) Hierarchical relations (via trajector/landmark
pairing)
(ii) Grammatical categories (via domain profiling)
(iii) Linear expression (via phonological pole)

Construal
It is the way in which the same objective basis is
portrayed.
The glass is half full or half empty.

Different degrees of subjective involvement in

profiling a perceptual experience:
I find that this chair is uncomfortable.
I find this chair to be uncomfortable.
I fin this chair uncomfortable.

Metonymy is one construal operation widely

active in grammatical coding.

Active Zones
This form of metonymical construal is used when a
more cognitively salient entity is referred instead of
the one more objectively involved in the situation

Active Zone Analysis of SSR

The key is that the profile/Active
Zone discrepancy is that both
conceptual representations need to
be present in order to understand
the sentence.

Parallel with control verbs

1. The printer is fast (to print).
2. The surgeon is fast (to give
stitches).
3. She began (to read) the novel.

Subjective and Objective Perception

Subjective perception assumes the subject of
conceptualization guiding perception
Objective perception assumes the object of
conceptualization as independent from the subject
I heard Melvin enter the building.
The perception of Melvin is elaborate by the subordinate.
But Melvin is also metonymically construed as the sound
(its source).

Melvin is likely to enter the building

There is less subjective involvement in the perception of
Melvins characteristics.
Instead, they are objectively construed as depending on
Melvin himself.
Notice that this is not the case in It is likely that Melvin
enters the building.

Raising or Control?
Causative verbs
She caused a specialist to examine her mother
She caused her mother to be examined by a specialist.

Raising verbs that form a tight conceptual unit

with the raised object

I believe the report to be true.

? I believe the rain to be falling.
I believe Jennifer to be telling {the truth / ? A lie}
We confirmed the rumor to be essentially {true/?false}
The key difference between raising and
control is the way in which participants
relate to the main event.

Participant Involvement is a Continuum

Prototypical raising verbs (to be likely):
1. They are predicates for which the raised participant
has a low degree of involvement.
2. The subordinate construction (to + verb) is required in
order to characterize the raised participants
involvement, and also the main predicate.

Prototypical control verbs (To be reluctant):

1. They are predicates for which the controlling
participant has a high degree of involvement.
2. The subordinate construction (to + verb) is required in
order to, first, characterize the main predicate, and then
the controlling participant.
In both cases to + verb gives supplementary
information. However, there is no need for it to
be always present (adjunct nature).

Examples
1.
2.
3.
4.
5.
6.
7.
8.
9.

A war is likely.
There is no possible way
to clearly delimit the
* Carlos is likely.
behavior of control and
Carlos is likely to sing today. raising verbs: each
predicate has specific
? Carlos is reluctant.
requirements in order to
be properly construed.
Carlos is happy.
Carlos is happy to teach syntax.
Carlos is reluctant to teach syntax.
The happy syntactician taught the class.
The reluctant syntactician taught the class.

Explanation for Raising and Control

Formal Explanation

Functional Explanation

Raising takes place because of

arguments needing to fulfill
very precise grammatical
constraints (case filter)
Control is required to satisfy a
thematic restriction in any
NP/DP (they can have only one
semantic role).
Explanation is opportunistic
in nature and theory driven.

Raising and control share the same

formal structure because both are
cases of participants construed as
central in a predicate (adjective) for
which a process that also involves
them is assumed (active zone).
The difference between verbs in
that construction becomes evident
when the participant is actively
engaged in the main predicate
(control) and when it is only
enabling it (raising).
Explanation assumes linear order as
meaningful, while differences are
not derivational, but functional (i.e.
conceptual).

Whats the Fuss About Logical Grammatical

Relations?
If you remember well, thematic/semantic roles are not the
same as grammatical relations.
The term argument as a required participant in an event
comes from logical calculus.
Thus logical grammatical relation = Thematic role

The most amazing thing has been that no one has really asked
what does it mean agent, patient, experiencer, etc. etc.
It is just blindly assumed that those notions are well-defined primitives
and they are not.
Think of this: A 5k is easy to run. Is it a control or
raising verb? What is the role of 5k?

What is Transparency Doing?

Transparency is the property of doing things in
grammar without any semantic consequence.
Raising from one structural position to another is a
case of transparency.

Langackers article wants to show how there is no

real transparency in raising.
SSR (be likely to), SOR (expect to), and OSR (be hard
to) are all of them motivated through semantic
factors that make the raised position the most natural
one for presenting an argument and some
supplementary predication that precise how the
focused element will be construed.

The syntax-lexicon continuum

Construction type Traditional name

Example

Complex, and (mostly syntax

schematic)

[SBJ be-TNS VERBen by NP] passive

Complex, substantive subcategorization

verb
frame

[SBJ consume OBJ]

Complex, and (mostly) idiom

substantive

[kick-TNS the bucket]

Complex but bound

morphology

[NOUN-s] plural,
[VERB-TNS] tense

Atomic and
schematic

syntactic category

[DEM], [ADJ]

Atomic and
substantive

word/lexicon

[this], [green]

transitive

die

The Autonomy Issue

If something in grammar cant be fully
predictable from its meaning or other
independent factors, then grammar is defined
separate from meaning.
Type/Predictability Fallacy
What type of linguistic units are there available?
How can we predict their grammatical behavior?
If units are meaningful (semantics-based) then every
grammatical form needs to be explained through its
meaning.

The Symbolic Alternative

Every grammatical unit has semantic and
phonological content.
The only grammatical units allowed link a
phonological expression with a semantic
content.
Meaning is conceptual structure, and it allows
a number of subtle variation in how reality is
presented.

What is Language Like?

1. Different structures will make use of different grammatical
strategies (ranging from lexical to morpho-syntactic ones),
but there is no guarantee that every language is doing at
some deep level the same kind of structuring.
This assumes linguistic diversity not as a problem that needs to
be explained away, but as the result of the continous nature of
potential structures available and the semantic variation
inherent to conventional arrangements.

2. The difference between constructions is the level of

abstraction (i.e. generality) with which they can be used.
Syntax is nothing but the conventional pairing of the most
abstract conceptual relations with the broadest range of
participants
Cognitive and functional, opposed to formal
theories of syntax have their focus on
languages, not the so-called language faculty.

What is Syntax?
Constructional (sub-)schemas of higher level
of abstractness and correspondence between
meaning of the units.
Conventional pairing of increasingly complex
conceptual structures with phonological forms
(non-generative).
This means that every syntactic assembly is
meaningful, even if redundant (e.g.
agreement) or non-fully predictable (e.g.
paradigms, agreement, word order, case)

Conventionality and Variation

The symbolic nature of grammar looks for
semantic and external motivation for
structures.
Motivation is not the same as predictability!
Conventionality is gradient between pure
motivation and pure arbitrariness.

What a Great Linguist Once Said

Were a language ever completely
grammatical, it would be a perfect engine of
conceptual expression. Unfortunately, or
luckily, no language is tyrannically consistent.
All grammars leak (Sapir, Language, 1921,
p.38; italics added)

EXTRAS ON CONSTRUCTION GRAMMAR

AND USAGE-BASED MODELS

Bidirectional links
Langacker (2003) points out that there are many verbs that
have a strong associative link to a particular construction.
give is extremely frequent in the ditransitive compared to other verbs
The ditransitive construction is extremely frequent with give.
The usage-based model predicts, based on frequency, that there is a
highly conventionalized link to the ditransitive that is part of our
knowledge of give.
If so, give is an access point to the ditransitive construction and its
associated frame

Hypothesis: The links between lexical item and constructions

reach in both directions
We posit both nodes as units if both are conventionalized. Give may
activate the ditransitive just as the construction primes the word.

The usage-based model

In the usage-based model, links in a linguistic
knowledge network are viewed as activation
pathways with potentially bidirectional
activation flows (cf. Lamb 2000)
Predicts that strongly entrenched links could
potentially go in either direction.
Converges with findings from neurology
suggesting that links between neurons and
between cortical columns have physically distinct
pathways that can have differential activation
strength.

Most construction grammars these days

are usage-based
knowing *idiomatic expressions+ is part of knowing
a language, and clearly their specifics are not
determined by universal principles but must be
learned on an item-by-item basis
Examples
1. English black eye = German blue eye
2. English sleep like a log = German sleep like
a woodchuck or marmot
3. English think of oneself as Gods gift to the world= French
believe oneself sprung from Jupiters thigh

Most construction grammars these days

are usage-based contd
Since every linguist agrees that the peripheral,
difficult cases must be learned inductively on the
basis of the input, constructionists point out that
there is no reason to assume that the more general,
regular, frequent cases cannot possibly be. (p. 14)

Constructions with no verbs

In many languages the construction cues the
argument interpretation even if there is no verb.
Russian
1. Kirill
v magazin
Kirill-NOM to store-ACC
Kirill goes/will go to the
store
2. Kirill iz magazina
Kirill-NOM from store-GEN
Kirill just got back from the
store
Taken from Goldberg, Adele. 2006.
Constructions at work, p. 8

German
1. Larry und Arzt?!
Larry and doctor
Larry, a doctor?!
French
1. Foc[tout le monde qui
part en weekend]
all the world who leaves
in weekend
Everyone is leaving for
the weekend.

Examples of argument structure

constructions
1. Transitive
2. Intransitive
3. Ditransitive (double object)
Subj
V
Obj1
He
gave
her
She
bought
him

Obj2
a flower.
a book.

The Power of Surface Generalizations

The recipient in the
ditransitive is not a good
argument to ask about.
But in the equivalent
paraphrases is just fine to
ask about.
No room for
adverbs to interrupt
the ditransitive
construction.
Construction Grammar assumption: No derivation is required, surface
structure is a specific coding of semantic /pragmatic function

Ditransitive Constructions: A Fully

Semantic Account
Semantic properties of
give
Complete transfer of an
object.
There is intention from
the giver.
Animate recipient.
Transfer of control over
object.
Notice that understanding is the conection of
different units of knowledge . Thus, certain
elements can be assumed to be valid
characterization of a category despite not
fulfilling each and every knowledge unit.

Extensions from the

prototype
Mina bought Mel a
book.
Mina sent Mel a book.
Mina
guaranteed/offered Mel
a book.
Mina refused Mel a
book.
Mina cost Mel his job.

Ditransitive Construction + hand

CAUSE-RECEIVE < agent

recipient theme>

HAND

< hander

handee

Verb

Subject

Object

Instance,
means

handed >
Object2

Pat handed Bill the keys.

cf. *Pat handed [] the keys. (not an instance of this
construction)

Ditransitive Construction + kick

CAUSE-RECEIVE

< agent

KICK

< kicker

Verb

Subject

recipient

theme>

Means

Joe kicked Bill the ball.

kicked>

Object

Object2

Role can be contributed by

construction (dashed line)

Ditransitive Construction + send

CAUSE-RECEIVE

< agent

SEND

< sender

recipient

theme>

Instance

Verb

Subject

send.goal

sent >

Object

Object2

Joe sent Leonard a letter.

"send.goal" role must be recipient and therefore animate
Closest object to verb is prototypically associated with
higher degree of affectation.

*Ditransitive Construction + anger

CAUSE-RECEIVE

< agent

ANGER

< angerer

recipient

theme>

Instance?
Means?

Verb

Subject

anger.goal >

Object

Object2

*Joe angered Bob the pink slip.

("Joe gave Bob a pink slip, causing Bob to become angry.")

Problematic cases: inanimate

participant in ditransitive construction
1.
2.
3.
4.

?? Mina sent that place a box

Mina sent a box to that place.
Mina sent Washington a box.
The paint job gave the car a higher
price sale.
5. Those cars sell like hotcakes.
Causation is not only a volitional notion (agentive), but also a logical one
(what causes something). The latter is found in the inherent properties of
an object.

Problematic cases: Ditransitive

Construction + Steal
CAUSE-RECEIVE

< agent

recipient

patient >

< thief

target

goods >

Object

Object2

Instance?
Means?

STEAL

Verb

Subject

*Robin Hood stole the rich their money.

But: You stole me my happiness; I stole her a kiss.
48

Different Construals in Syntax

Taken from
Goldberg
(1995: 47)

steal <thief target goods>

rob <thief target goods>

(1) a. Jesse robbed the rich (of all their money). Negative effects
b. *Jesse robbed a million dollars (from the are crucial in rob,
but not on steal.
rich).
(2) a. Jesse stole money (from the rich).
Constructional meaning
b. *Jesse stole the rich (of money).
requires profiling on the
(3) *Robin Hood stole the rich their money.
effects.
(4) You stole me my happiness.
(5) I stole her a kiss.

Blending Constructions
Active zone for
happiness and its
immediate domain.
Active zone for kiss
and its immediate
domain.

(4) You stole me my

happiness.
(5) I stole her a kiss.

(i) Input 1: Stealing

(ii) Input 2: Ditransitive construction
(iii) Conection of commonalities: Affected
participant and extracted good are equated
(iv) Meaning: Extracting something a non-harmful
way that necessarily affects the owner