Escolar Documentos
Profissional Documentos
Cultura Documentos
A new theoretical approach to language has emerged in the ‘core’ of language. Mainstream generative theory
the past 10 –15 years that allows linguistic observations argues further that the complexity of core language cannot
about form –meaning pairings, known as ‘construc- be learned inductively by general cognitive mechanisms
tions’, to be stated directly. Constructionist approaches and therefore learners must be hard-wired with principles
aim to account for the full range of facts about that are specific to language (‘universal grammar’).
language, without assuming that a particular subset of
the data is part of a privileged ‘core’. Researchers in this Tenets of constructionist approaches
field argue that unusual constructions shed light on Each basic tenet outlined below is shared by most
more general issues, and can illuminate what is constructionist approaches. Each represents a major
required for a complete account of language. divergence from the mainstream generative approach
and, in many ways, a return to a more traditional view of
Constructions – form and meaning pairings – have been language.
the basis of major advances in the study of grammar since Tenet 1. All levels of description are understood to
the days of Aristotle. Observations about specific linguistic involve pairings of form with semantic or discourse
constructions have shaped our understanding of both function, including morphemes or words, idioms,
particular languages and the nature of language itself. But partially lexically filled and fully abstract phrasal
only recently has a new theoretical approach emerged that patterns. (See Table 1)
allows observations about constructions to be stated Tenet 2. An emphasis is placed on subtle aspects of the
directly, providing long-standing traditions with a frame- way we conceive of events and states of affairs.
work that allows both broad generalizations and more Tenet 3. A ‘what you see is what you get’ approach to
limited patterns to be analyzed and accounted for fully. syntactic form is adopted: no underlying levels of syntax
This is in contrast to the mainstream ‘generative’ approach or any phonologically empty elements are posited.
to language, which has held sway for the past several Tenet 4. Constructions are understood to be learned on
decades, beginning with Chomsky in 1957 [1]. the basis of the input and general cognitive mechanisms
Many linguists with varying backgrounds have recently (they are constructed), and are expected to vary cross-
converged on several key insights that have given rise to a linguistically.
new family of approaches, here referred to as ‘construc- Tenet 5. Cross-linguistic generalizations are explained
tionist’ approaches [2– 23]. Constructionist approaches by appeal to general cognitive constraints together with
share certain foundational ideas with the mainstream the functions of the constructions involved.
generative approach. Both approaches agree that it is Tenet 6. Language-specific generalizations across con-
essential to consider language as a cognitive (mental) structions are captured via inheritance networks much
system; both approaches acknowledge that there must be a like those that have long been posited to capture our
way to combine structures to create novel utterances, and non-linguistic knowledge.
both approaches recognize that a non-trivial theory of Tenet 7. The totality of our knowledge of language is
language learning is needed. captured by a network of constructions: a ‘construct-i-con.’
In other ways, constructionist approaches contrast
sharply with the mainstream generative approach. The Each of these tenets is explained in a subsequent
latter has held that the nature of language can best be section below.
revealed by studying formal structures independently of
their semantic or discourse functions. Ever increasing Constructions: what they are
layers of abstractness have characterized the formal Constructions are stored pairings of form and function,
representations. Meaning is claimed to derive from the including morphemes, words, idioms, partially lexically
mental dictionary of words, with functional differences filled and fully general linguistic patterns. Examples are
between formal patterns being largely ignored. Semi- given in Table 1.
regular patterns and unusual patterns are viewed as Any linguistic pattern is recognized as a construction as
‘peripheral,’ with a narrow band of data seen as relevant to long as some aspect of its form or function is not strictly
predictable from its component parts or from other
Corresponding author: Adele E. Goldberg (agoldbrg@uiuc.edu). constructions recognized to exist. In addition, many
http://tics.trends.com 1364-6613/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S1364-6613(03)00080-9
220 Review TRENDS in Cognitive Sciences Vol.7 No.5 May 2003
Table 1. Examples of constructions, varying in size and complexity; form and function are specified if not readily transparent
Construction Form/Example Function
Morpheme e.g. anti-, pre-, -ing
Word e.g. Avocado, anaconda, and
Complex word e.g. Daredevil, shoo-in
Idiom (filled) e.g. Going great guns
Idiom (partially filled) e.g. Jog ksomeone’sl memory
Covariational-Conditional Form: The Xer the Meaning: linked independent and dependent variables
construction [10] Yer (e.g. The more you think about it,
the less you understand)
Ditransitive Form: Subj [V Obj1 Meaning: transfer (intended or actual)
(double-object) Obj2] (e.g. He gave her a Coke; He baked her a muffin)
construction
Passive Form: Subj aux VPpp (PPby) (e.g. The armadillo was hit Discourse function: to make undergoer topical and/or
by a car) actor non-topical
constructionist approaches argue that patterns are stored and relative clauses are understood to be learned pairings
even if they are fully predictable as long as they occur with of form and (semantic or discourse) function – that is, they
sufficient frequency [24– 29]. are also constructions. Each pairs certain formal proper-
Unlike mainstream generative grammar, the construc- ties with a certain communicative function.
tionist framework emphasizes the semantics and distri- Even basic sentence patterns of a language can be
bution of particular words, grammatical morphemes, and understood to involve constructions. That is, the main verb
cross-linguistically unusual phrasal patterns. The hypoth- can be understood to combine with an argument-structure
esis behind this methodology is that an account of the rich construction (e.g. transitive, intransitive, ditransitive
semantic, pragmatic, and complex formal constraints on constructions, etc.) [7]. The alternative is to assume that
these patterns readily extends to more general, simple or the form and general interpretation of basic sentence
regular patterns. patterns are determined by semantic and/or syntactic
As an example of an unusual pattern, consider the information specified by the main verb. The sentence
Covariational Conditional construction in Table 1 (e.g. ‘The patterns given in (1) and (2) indeed appear to be
more you think about it, the less you understand’). The determined by the specifications of give and put respect-
construction is interpreted as involving an independent ively:
variable (identified by the first phrase) and a dependent (1) Chris gave Pat a ball.
variable (identified by the second phrase). The word the (2) Pat put the ball on the table.
normally occurs at the beginning of a phrase headed by a
noun. But in this construction it requires a comparative Give is a three-argument verb: an act of giving requires
phrase. The two major phrases of the construction resist three characters: a giver (or agent), a recipient, and
classification as either noun phrases or clauses. The something given (or ‘theme’). It is therefore expected to
requirement that two phrases of this type be juxtaposed appear with three phrases corresponding to these three
without conjunction is another non-predictable aspect of roles. In (1), for instances, Chris is agent, Pat is recipient,
the pattern. Because the pattern is not strictly predictable, and a ball is theme. Put, another three-argument verb,
a construction is posited that specifies the particular form requires an agent, a theme (object that undergoes the
and semantic function involved [10]. change of location) and a final location of the theme’s
Other unusual constructions include those in Table 2. motion. It appears with the corresponding three argu-
Although some of the patterns are primarily used ments in (2). However, whereas (1) and (2) represent
colloquially, they are part of every native speaker’s perhaps the prototypical case, in general the interpret-
repertoire of English. (The stranded preposition construc- ation and form of sentence patterns of a language are not
tion is unusual not by virtue of its being prescriptively reliably determined by independent specifications of the
dispreferred, but in that it is found only in a few Germanic main verb. For example, it is implausible to claim that
languages). sneeze has a three-argument sense, and yet it can appear
More common patterns such as passive, topicalization as such in (3). The patterns in (4)– (6) are likewise not
naturally attributed to the main verbs:
Table 2. Productive or semi-productive constructions that are (3) ‘He sneezed his tooth right across town.’ (Robert
unusual across languages and must be learned on the basis of
Munsch, Andrew’s Loose Tooth)
the input
(4) ‘She smiled herself an upgrade.’ (Douglas Adams,
time away construction Twistin the night away [13] Hitchhiker’s Guide to the Galaxy, Harmony Books)
What’s X doing Y? What’s that fly doing (5) ‘We laughed our conversation to an end.’ (J. Hart. Sin
in my soup? [30]
Nominal Extraposition construction It’s amazing the difference! [31]
Ivy Books, New York)
Mad Magazine construction Him, a doctor?! [32] (6) ‘They could easily co-pay a family to death.’ (New York
Noun –Pronoun –Noun (N P N) house by house; day Times, 1/14/02)
construction after day [12]
Stranded preposition construction Who did he give Examples need not be particularly novel to make the
that to?
point. Verbs typically appear with a wide array of
http://tics.trends.com
Review TRENDS in Cognitive Sciences Vol.7 No.5 May 2003 221
http://tics.trends.com
222 Review TRENDS in Cognitive Sciences Vol.7 No.5 May 2003
requires an animate recipient argument conflicts with the same construction if and only if their form and function is
meaning of storage in (9c) resulting in unacceptability. The identical once other constructions in the language that
observation that language has an infinitely creative might differ are factored out. In fact, this rarely occurs
potential [1,36] is accounted for, then, by the free except in cases of shared diachronic history or language
combination of constructions. contact [20,45,46]. What is truly remarkable is the degree
to which human languages differ from one another, given
Learning constructions that all languages need to express roughly the same types
The fourth tenet states that constructions are understood of messages. Constructionist approaches anticipate such
to be learned on the basis of positive input and to vary fairly wide variability across languages [47,48].
across languages. This idea highlights a major difference We can understand what is actually intended by
between most constructional approaches and most main- references to the ‘same’ construction in unrelated
stream generative approaches, as the latter have argued languages as types of constructions. Two constructions
that learners must be hard-wired with principles specific might be, for example, of the passive type in that they
to a language faculty, that is, to possess a ‘universal share certain functional and formal characteristics even if
grammar’ ([37]; see also [21]). they are not identical. That is, two constructions in
Crucially, all linguists recognize that a wide range of different languages can be identified as instances of the
semi-idiosyncratic constructions exist in every language, same type of construction if and only if they serve a closely
constructions that cannot be accounted for by general, related function and form.
universal or innate principles or constraints (see examples
in Table 2). Mainstream generative theory has taken the Cross-linguistic generalizations
position that these constructions exist only on the A driving question behind much of linguistic research is
‘periphery’ of language, and that therefore they need not what is the typology of possible constructions and what
be the focus of linguistic or learning theorists [37]. constrains it? Constructionist approaches often turn to
Constructionist approaches, on the other hand, have grammar-external explanations such as universal func-
zeroed in on these constructions, arguing that whatever tional pressures, iconic principles, and processing and
means we use to learn these patterns can easily be learning constraints to explain such empirically observa-
extended to account for so-called ‘core’ phenomena. In fact, ble cross-linguistic generalizations. For example, certain
by definition, the core phenomena are more regular, and generalizations about how form and meaning tend to be
also tend to occur more frequently within a given linked across languages can be explained by appeal to
language. Therefore if anything, they are likely to be iconic and analogical processes [6,35,49 – 51]. Constraints
easier to learn. Because every linguist would presumably on long-distance dependency constructions (traditional
agree that the ‘peripheral’, difficult cases must be learned ‘island constraints’) appear to yield to processing expla-
inductively on the basis of the input, constructionist nations that take into account the function of the
theories propose that there is no reason to assume that constructions involved [19,52– 54]. Processing accounts
the more general, regular, frequent cases cannot possibly have also been suggested to account for certain alternative
be learned in this way. word-order options [55,56].
In fact, constructionist theories argue that language Even among generative linguists there has been a trend
must be learnable from positive input together with fairly towards the view that many constraints on language that
general cognitive abilities [18,29,38], because the diversity have traditionally been seen as requiring recourse to
and complexity witnessed does not yield to accounts that innate stipulations specific to language can actually be
assume that cross-linguistic variation can be character- explained by general cognitive mechanisms. For example,
ized in terms of a finite set of parameters [37]. Research in the fact that that all languages appear to have noun and
this area is quickly gaining momentum. Several construc- verb (and, possibly, adjective) categories may be explained
tionists have made good on the promise to explain how by the existence of corresponding basic semantic cat-
particular constructions are learned [26,27]. It turns out egories [57]. In a recent paper, Hauser, Chomsky and Fitch
that the input need not be nearly as impoverished as is go so far as to suggest that the only language-specific
sometimes assumed [39]; analogical processes can be seen innate ability that is absolutely required is recursion, and
to be viable once function as well as form is taken into they raise the point that even that might turn out not to be
account [40,41]; there is good reason to think that specific to language [58] (see also Box 1. Questions for
children’s early grammar is quite conservative, with Future Research).
generalizations emerging only slowly [29,42,43]; and the
ability to record transitional probabilities and statistical Intra-language generalizations
generalizations in the input has proven a powerful means Inheritance hierarchies have long been found useful for
by which to learn certain types of generalizations [44]. representing all types of knowledge, for example, our
This approach takes a somewhat different view from knowledge of concepts. The construction-based framework
mainstream generative theory of what is universal about captures linguistic generalizations within a particular
language. Linguists talk of certain constructions as language via the same type of inheritance hierarchies
existing in many languages, for example, the passive [2,59,60]. Broad generalizations are captured by con-
construction, relative clause construction, question con- structions that are inherited by many other constructions;
struction, and so forth. However, two constructions in more limited patterns are captured by positing construc-
different languages can be identified as instances of the tions at various midpoints of the hierarchical network.
http://tics.trends.com
Review TRENDS in Cognitive Sciences Vol.7 No.5 May 2003 223
35 Wasow, T. (2002) Postverbal Behavior, CSLI Publications 49 Haiman, J. (1985) Iconicity in Syntax, Cambridge University Press
36 Chomsky, N. (1965) Aspects of the Theory of Syntax, MIT Press 50 Givón, T. (1991) Isomorphism in the grammatical code: cognitive and
37 Chomsky, N. (1981) Lectures on Government and Binding, Foris, biological considerations. Studies Lang. 1, 85– 114
Dordrecht 51 Kemmer, S. and Verhagen, A. (2002) The grammar of causatives and
38 Elman, J. et al. (1996) Rethinking Innateness: A Connectionist the conceptual structure of events. Mouton Classics: From Syntax to
Perspective on Development, MIT Press Cognition, From Phonology to Text, pp. 451 – 491, Mouton de Gruyter
39 Pullum, G.K. and Scholz, B.C. (2002) Empirical assessment of 52 Kluender, R. (1998) On the distinction between strong and weak
stimulus poverty arguments. Linguist. Rev. 19, 9 – 50 islands: a processing perspective. Syntax Semantics 29, 241– 279
40 Goldberg, A.E. (1999) The Emergence of argument structure seman- 53 Kluender, R. and Kutas, M. (1993) Subjacency as a processing
tics. In The Emergence of Language (MacWhinney, B., ed.), phenomenon. Lang. Cogn. Process. 8, 573 – 633
pp. 197– 212, Erlbaum 54 Erteschik-Shir, N. (1998) The syntax-focus structure interface. In
41 Israel, M. (2002) Consistency and creativity in first language Syntax and Semantics Vol. 29: The Limits of Syntax (Culicover, P. and
acquisition. Proc. Berkeley Linguist. Soc., p. 29 McNally, L., eds.), pp. 211 – 240, Academic Press
42 Lieven, E.V.M. et al. (1997) Lexically-based learning and early 55 Hawkins, J. (1994) A Performance Theory of Order and Constituency,
grammatical development. J. Child Lang. 24, 187 – 219 Cambridge University Press
43 Tomasello, M. (2000) Do young children have adult syntactic
56 Yamashita, H. and Chang, F. (2001) ‘Long before short’ preference in
competence? Cognition 74, 209 – 253
the production of a head-final language. Cognition 81, B45– B55
44 Saffran, J.R. (2001) The use of predictive dependencies in language
57 Baker, M. (2003) Verbs, Nouns, and Adjectives: Their Universal
learning. J. Mem. Lang. 44, 493 – 515
Grammar. Cambridge University Press (in press)
45 Birner, B. and Ward, G. (1998) Information Status and Noncanonical
58 Hauser, M.D. et al. (2002) The faculty of language: what is it, who has
Word Order in English, John Benjamins
it, and how did it evolve? Science 298, 1569 – 1579
46 Zhang, N. (1998) The interactions between construction meaning and
59 Pollard, C.J. and Sag, I. (1994) Head-Driven Phrase Structure
lexical meaning. Linguistics 36, 957 – 980
Grammar, CSLI Publications/Cambridge University Press
47 Foley, W.A. and van Valin, R. Jr (1984) Functional Syntax and
60 Goldberg, A.E. Words by default: inheritance and the Persian Complex
Universal Grammar, Cambridge University Press
48 Garry, J. and Rubino, C., eds. (2001) Facts about the World’s Predicate Construction. In Mismatch: Form –Function Incongruity
Languages: An Encyclopedia of the World’s Major Languages Past and the Architecture of Grammar (Francis, E. and Michaelis, L., eds.),
and Present. H.W. Wilson CSLI Publications (in press)
http://tics.trends.com