sign in sign up
<<
Home , Cognition, Cognitive grammar, Cognitive linguistics, Cognitive psychology, Cognitive revolution, Elizabeth Bates

Galley: Article - 00559

Level 1 and Catherine L Harris, Boston University, Massachusetts, USA

CONTENTS Introduction Cognitive of cognition and language The Cognitive movement Conclusion

Cognitive have long debated whether lan- The second way ofconceptualizing cog- 0559.003 guage and cognition are separate mental faculties, nition emphasizes the differences between lan- or whether language emerges from general cogni- guage and other abilities. A key is that many tive abilities. distinct domains ofcognition exist and must be learned separately, using different mental mechan- INTRODUCTION isms. This approach is referred to as the `modular- ity ofcognition' or `mental modules' approach. At 0559.001 What is the relationship between language and first glance it may seem contrary to the interdiscip- cognition? Do people who speak different lan- linary spirit ofcognitive science and to the possibil- guages think differently? Is a certain level of ity ofa unified theory ofcognition. However, the required for language ac- unifying theory is the thesis of distinct mental quisition? These questions were ofkeen interest to modules, which are believed to have evolved to thinkers in the early twentieth century and remain accomplish specific tasks relevant to mammalian important in , linguistics and psych- , such as visual exploration, or relevant ology. However, the cognitive ofthe to , such as language use. Much of 1950s brought a new question about the relation- the appeal of this approach comes from findings in ship between language and cognition: is language showing that distinct areas ofthe the same type ofmental entity as other cognitive serve distinct functions such as vision, lan- abilities, or is it fundamentally different? guage processing, motor coordination, , 0559.002 A hallmark ofmodern is the and face recognition. The interdisciplinary spirit is goal ofdeveloping a theory ofcognition powerful maintained because advocates ofthis approach enough to encompass all human mental abilities, reach out to biological scientists and evolutionary including language abilities. A long-standing con- theorists. Those favoring modularity embrace the troversy concerns two ways ofconceptualizing the principle ofconverging methodologies: a theory architecture (or basic design) ofcognition. One ap- must have explanatory power in the distinct aca- proach proposes that general-purpose processes demic disciplines that compose the cognitive sci- and mechanisms provide a foundation for all var- ences. These two approaches to the architecture of ieties ofhuman . We can referto this as cognition developed out ofdifferentphilosophical `general purpose' cognition. Examples ofpossible traditions, and have evolved considerably during universal processes are the ability to induce a the half-century of cognitive science. category from exposure to examples (category in- duction), and the ability to mentally complete a known pattern when confronted with a piece of it CONCEPTS OF COGNITION AND (pattern completion). Cognitive scientists fre- LANGUAGE quently attempt to precisely specify their proposed Why are there two different views on the relation- mechanisms by implementing them as computer 0559.004 ship between language and cognition? At the dawn algorithms which can be tested in artificial intelli- ofthe , in the late 1950s, there gence (AI) programs. Researchers have tried to use were two distinct about the nature ofmind. AI programs to show that the same principles that Discussed first are the views of linguist Noam can explain general problem-solving can also Chomsky and the field of generative linguistics explain aspects oflanguage acquisition and - cessing. Galley: Article - 00559 2 Language and Cognition

which he developed, because the underlying phil- or speakers' communicative goals came to be called osophy has remained fairly constant over the inter- the `autonomy ofsyntax' hypothesis.

vening years. Chomsky's innovations developed in tandem 0559.007

0559.005 Chomsky's major innovation was to conceive of with the dawn ofthe cognitive revolution. With language abilities as akin to a mental organ. the birth ofcomputer science a new way ofconcep- According to this view, children are born with a tualizing human cognition arose, using the - ` device' and with specific lin- phor ofthe brain as a computer and the as guistic . This knowledge is to software. Much of Chomsky's early work include the concepts ofnoun, verb, grammatical depended on this : he conceived ofgram- , and structures that constrain possible mar as a ofrules forgenerating novel combin- grammatical rules. In contrast to the views ofthe ations ofwords, just as a computer program could dominant psychological theory ofthe 1950s, behav- generate a string ofsymbols according to a formula. iorism, Chomsky argued that children do not learn Thinking ofmental operations as akin to steps in a to speak by imitating . His key evidence was computer program allowed and that children spontaneously use incorrect forms workers in the new field of to they could not have heard, like `goed' and begin to subdivide mental tasks, such as arithmetic `breaked'. Linguistic overregularizations like these or language comprehension, into a series ofsteps in suggest that children are extracting rules from the a computer program. Their research evolved in a language they hear, not merely imitating. Theorists different direction from Chomsky's. Computer sci- at that time found it noteworthy that parents do not entists and psychologists such as Alan Newell, generally tell children that their utterances are un- John Anderson, and Patrick Winston grammatical. Because the language input to chil- began to describe a range ofhuman abilities ± from dren is full of mistakes, stops and restarts, visual recognition and general problem solv- Chomsky felt that children could not learn lan- ing, to metaphor use and story ± in guage using general purpose problem-solving or terms ofa set ofinternal representations and pro- regularity-extraction skills. They needed to come cesses that transform those representations. A key to the task with a rich set ofexpectations about aspect ofthis approach to cognition was that the the nature oflanguage. These expectations were theorist could write computer programs that believed to be specific to language, and thus did would provide a formalizable theory of mental op- not share commonalities with other aspects ofcog- erations. A rigorous test ofthe theory could be nition. This set oflanguage-specificabilities has performed by running the program and seeing if been variously called the `language acquisition output matched human output.

device' (the historically early term) and `universal Many ofthe early successes ofthis fieldinvolved 0559.008 ' (a more recent term). Chomsky's ap- language, including sentence comprehension, story proach to linguistics is called `generative linguis- understanding and metaphor use. An important tics' because its early goal was to describe mental aspect ofthe language and cognition relationship structures that can generate all the grammatically is that the AI models oflanguage did not draw on valid sentences ofa language. language-special algorithms or knowledge struc-

0559.006 Chomsky felt that the aspect of language that is tures. The new AI tradition and the information unique is syntactic ability. An example ofspecific- processing movement within empha- ally syntactic knowledge is illustrated by the sen- sized , particularly general-purpose learn- tence, `Colorless green ideas sleep furiously'. ing, and was thus opposed to the emphasis on Although the in this sentence contradict innate knowledge structures that was part of each other and do not correspond to a possible Chomsky's new linguistics. Psychologists drawn reality (green is not colorless, ideas cannot sleep to the movement were fre- furiously), speakers nevertheless recognize the sen- quently inspired by the work ofSwiss tence as having a correct grammatical structure. , whose works began to be translated Chomsky used this sentence as an example of into English in the 1960s. Piaget also emphasized how syntactic structure represents information in- the commonalities between language and cogni- dependently from the of the words in the tion, and proposed that language emerged out of sentence. He argued that is a unique, inde- the same broad cognitive changes that transform pendent human capacity and not derivative from the sensorimotor processing ofinfantsinto the other abilities. The proposal that syntax is not influ- formal and logical mind of adults.

enced by the meaning ofthe words in the sentence In the and 1980s, among those researchers 0559.009 who aligned themselves with the interdisciplinary Galley: Article - 00559 Language and Cognition 3

field of cognitive science, there was a tendency for units (), linked together in a linguists to emphasize the specialness oflanguage network of fiber tracts, with many units firing sim- and cognition, and for psychologists to emphasize ultaneously. This movement was labeled `connec- commonalities between language and cognition. tionism' because intelligent behavior was posited However, there was considerable variation in to emerge from large numbers of -like pro- viewpoint within psychology and linguistics, cessing units, connected together into networks in which continues to this day. Three scientific devel- ways that fostered parallel processing. One early opments during the 1980s and the 1990s had impli- success story was a connectionist model that could cations for theories of the language±cognition learn English forms given the present relationship: connectionism, , tense ofa verb. The domain ofEnglish past tense and the movement. The was provocative because it contained both rule-like development ofthese movements is summarized behavior (past tense forms are generated by adding in Table 1. `-ed' to a present tense form), and exceptions which themselves congregated into patterns, such as that CONNECTIONISM emerging from considering `grow/grew', `blow/ blew', `know/knew'. Both connectionist networks

0559.010 The `connectionist revolution' was launched in the and second-language learners are likely to mid 1980s. A new computational metaphor make the error ofgenerating `glew' forthe past emerged to explain both language and cognition, tense of`glow', presumably through with based not on the Von Neuman computer, but on this subregularity within the past tense system. the idea that sophisticated computations emerge Other connection language models explored how from massive networks of simple processing ambiguous words are understood in their sentence units, in which the units are akin to idealized context (such as `bat' in `The boy hit the bat'), and neurons. how abstract categories such as noun and verb can

0559.011 In the early 1980s several groups ofcognitive emerge from distributional regularities in text scientists became dissatisfied with the AI systems (such as the fact that nouns tend to occur in similar being used to model cognition. These rule- sentence positions). Because the basic system of governed systems employed databases containing processing units under these connectionist lan- knowledge ofcommon situations, such as what one guage models was the same as those used to does at a restaurant; `if±then' rules specified the model visual and motor behavior, successes like action that an expert would take in a specific situ- the past tense model were taken as support for a ation. This expert knowledge thus constituted in- common computational architecture underlying telligent reasoning. Critics noted that these expert both language and cognition. systems were brittle and frequently failed to gener- alize beyond the bounds oftheir circumscribed COGNITIVE LINGUISTICS database. Real human cognition seemed intuitively to involve not the application offixedrules to a A subset oflinguists disagreed with Chomsky's 0559.013 specific situation, but the satisfaction of many soft emphasis on the uniqueness and specialness of constraints, some ofwhich could be ignored. Cat- language, then generally accepted. The field of cog- egories were not logical sets of necessary and suffi- nitive linguistics emerged in the late 1980s and cient conditions, but were graded groupings which helped initiate a flood of work connecting language shared `family resemblance', as had been argued and cognition. One source for the cognitive linguis- earlier in the century by Wittgenstein. To the math- tics movement was an older tradition within lin- ematical and cognitive psychologist David Rumel- guistics called `functionalist' linguistics. This held hart, flexible and creative reasoning seemed to that constraints on the form of language (where similar to the motor process involved in reaching `form' means the range of allowable grammatical for a cup behind a pencil-holder on one's desk ± or rules) derive from the function of language. The at least, more similar to this type ofplanning than goals of using language include serving efficient to rule-governed algebraic reasoning. . Thus the function of diverse syn-

0559.012 Rumelhart and other cognitive scientists such as tactic forms, such as subject versus object and main Terrence Sejnowski and Stephen Grossberg agreed clause versus subordinate clause distinctions, serve that mental functioning involves computation, but communicatory goals such as conveying which asked what type ofcomputation would be carried information is most important and which is out by an organic structure like the brain, com- background or context. However, the goals of posed ofmassive numbers ofsimple processing Galley: Article - 00559 4 Language and Cognition

conveying the relevance and background/fore- known by Mary' does not. To fully describe which ground structure ofthe message are outside the transitive sentences can undergo passivization, one language system. The proposal that the form of needs to invoke the notion that the subject and grammatical rules is influenced by communication direct object are in dynamic interaction with each is thus inconsistent with the autonomy ofsyntax other. Thus, one cannot passivize a sentence like hypothesis. That hypothesis specified that syntax `John left the auditorium' because John is not acting was its own system, not shaped by the need for on the auditorium in a way that has consequences efficient processing or other exigencies of commu- for it; but the sentence `John left the auditorium nication. unguarded' can transformed into `The auditor-

0559.014 Functionalist linguists as well as other linguists ium was left unguarded by John', presumably were dissatisfied with the range of linguistic phe- because John's action affects the status of the audi- nomena excluded by generative linguistics. They torium. rejected the Chomskyan view that the most import- By 2000 the cognitive linguistics movement had 0559.016 ant aspect oflanguage was a mechanical device for grown into an enduring subfield, but it has generating only legitimate grammatical sentences. remained outside the mainstream oflinguistics. They wanted to understand language in all its di- While some cognitive linguists have remained versity, including narrative, discourse, dialects, focused on specific linguistic questions, others sociocultural influences on language use, and have addressed questions in an interdisciplinary metaphor. Some functionalist linguists even pro- manner, drawing on , posed that rule use is just a minor aspect oflan- brain science, and category induction performed guage. Linguists such as Dwight Bolinger and by artificial neural networks. Charles Fillmore argued that utterances, not rules for generating utterances, are what is THE COGNITIVE NEUROSCIENCE mentally stored. They noted that every language MOVEMENT speaker has memorized huge numbers ofodd

coinages, colloquialisms, and collocations, The field of cognitive neuroscience emerged from 0559.017 many ofwhich share patterns. An example is the work in neuroscience and cognitive science. Cogni- phrase `know by heart' which has the variation tive neuroscience differs from basic neuroscience `learn by heart'. Rules, partial regularities and rule by having the goal ofexplaining complex cognitive exceptions appear to differ in degree, occupying a abilities, but rejects the tradition ofartificial intelli- continuum from fully idiosyncratic, to partially gence (and much ofcognitive science) that one can regular, to fully rule-governed. This idea made re- understand cognition abstractly, without reference searchers in this nascent linguistic movement sym- to its neural underpinnings.

pathetic to the connectionist movement, although In the 1990s some cognitive neuroscientists 0559.018 the commonalities have not yet been fully explored. argued that basic aspects ofthe language±cognition 0559.015 Although trained in the relationship, such as the autonomy ofsyntax hy- tradition, linguists such as and pothesis and the innateness and modularity oflan- Ronald Langacker noted that one could not de- guage, could be evaluated from the neuroscientific scribe which sentences are syntactically valid and point ofview. Neurobiologists have noted that de- which are invalid without reference to nonlinguis- veloping neural tissue is very plastic. For example, tic concepts. They pointed to the tendency for the the auditory association areas ofthe fre- same words describing movement in space to be quently represent visual and gestural language in used to describe movement in time (`This meeting individuals who are born deaf. The regions of the runs until 3 o'clock'), and emphasized the necessity brain that mediate language use appear to be espe- ofincorporating the of cially malleable. Like other aspects ofcognition, human category formation into linguistics. This language acquisition is heavily dependent on ex- approach to linguistics came to be called `cognitive' perience.

linguistics because aspects ofgeneral cognition ± R-A Mueller has remarked that regional special- 0559.019 such as how we construe the meaning ofa gram- ization in the brain is beyond , but modular- matical construction ± were proposed to be import- ity ofcognitive functions, including language, is ant for describing linguistic structure. For example, highly debatable from the view of neurobiology one descriptive problem ofgrammar is to account and evolution. Functional specialization ofbrain for why some sentences, but not others, can under- areas most probably emerges because some brain go the `passivization' transformation. The sentence areas are near to the site ofsensory input, such as `John was hit by Mary' sounds fine, but `John was Galley: Article - 00559 Language and Cognition 5

sensory systems for vision and audition. Scientists mirroring the development within cognitive sci- such as Jeffrey Elman, and their ence of two fundamentally different conceptions colleagues note that however closely one looks at ofthe . The tradition ofartifi- the anatomy and physiology ofthe brain, there is cial intelligence emphasized general-purpose prob- no evidence ofcortical structures unique to lan- lem-solving abilities, while the tradition of guage or unique to . These researchers linguistics and led to an emphasis on argue that language has an `epigenetic' not a `gen- distinct mental modules.

etic' origin. Epigenetic development is the proposal The four different theoretical perspectives on the 0559.023 that behavior results from a complex dynamic evo- language±cognition relationship were summarized lution ofgenes and environmental forcesduring in Table 1. The view at the beginning ofthe twenty- prenatal and postnatal development. The first century appears to be best captured by the idea ofepigenesis dates back to psychologist Jean Pia- that cognition and language have complex similar- get, who argued that cognitive abilities emerge ities and differences, and both develop over the from a biological structure which evolves, both human life span from genetic factors constrained before and after birth, in tandem with environmen- by environmental input and cultural learning. New tal forces. Contemporary researchers who embrace possibilities for synthesis continue to emerge, espe- the epigenetic view point out that are too few genes cially as cognitive scientists pay more to in the human genome to directly for outcomes evolutionary, neurobiological and cultural factors. such as the ability to use language. Like other brain It may be possible to set aside the question of regions, the language areas in the adult brain are whether language is distinct from cognition and the end product ofcomplex chains ofinteractions whether the brain is composed ofdistinct mental with internal and external environments. These se- modules. The theorist has noted a quences ofevents are based probabilistically on growing consensus about the importance ofa new genes rather than being rigidly determined by the set ofquestions about how to divide up the grand genome. areas ofmind and brain. Scientists are emphasizing

0559.020 The neurobiological evidence thus may run the distinction between areas ofhuman ability that counter to what would be expected under the au- are available to all humans and played a part in the tonomy ofsyntax hypothesis. There is no known evolution ofour species (such as language and way that genes could encode for concepts like `sub- basic number use), and areas requiring cultural ject' and `verb'. Thus the most parsimonious per- elaboration (such as algebra and the ability to play spective is that language is similar to other aspects musical instruments). The era ofsimplistic state- ofcognition in terms ofemerging out ofa brain ments about the language±cognition relationship which evolved to have an oversized frontal cortex is drawing to a close, as cognitive scientists begin (relative to other primates) and an elongated period to deliver on the promise ofa truly interdisciplin- ofchildhood, which privilege the role oflearning. ary approach to understanding the mind±brain.

0559.021 Cognitive neuroscientists share a view oflan- guage that resonates with the cognitive linguists: Further they emphasize the jointdevelopment oflanguage and perceptuomotor processes, with language ac- Baars BJ (1986) The Cognitive Revolution in Psychology. New York: Guilford Press. quisition understood to be semantically driven and Dromi E, ed. (1993) Language and Cognition: A embodied. The neurological representation of Developmental Perspective. Norword, NJ: Ablex. grammar is continuous with the representation of Edelman GM (1992) Bright Air,Brilliant Fire . New York: other language `components' and the neural sub- Basic Books. strate. Elman JL, Bates E, Johnson MH et al. (1996) Rethinking Innateness: A Connectionist Perspective on Development. Cambridge, MA: MIT Press. CONCLUSION Gardner H (1985) The Mind's New Science: A History of the Cognitive Revolution. New York: Basic Books. 0559.022 In the first half of the twentieth century the main question about the relationship between language Gumperz JJ and Levinson SC (1996) Rethinking Linguistic and cognition was whether the grammatical struc- Relativity. Cambridge, UK: Cambridge University Press. ture or vocabulary ofour language influenced Harris CL (1990) Connectionism and cognitive thought processes. Cognitive science introduced a linguistics. Connection Science. new question: are language and cognition similar Lakoff G and Johnson M (1999) Philosophy in The Flesh. or distinct human abilities? The last 50 years have New York: Basic Books. seen considerable controversy on this question, Galley: Article - 00559 6 Language and Cognition

Langacker RW (1987) Foundations of , human psychology and the mind, including the use of vol. 1, Theoretical Prerequisites. Stanford: Stanford the computational metaphor of mind, and the idea that University Press. language consists of a mental grammar which is a set MacWhinney B, ed. (1999) The Emergence of Language. of unconscious rules. Mahwah, NJ: Erlbaum. Epigenetic Resulting from a complex dynamic evolution Morris RGM, ed. (1989) Parallel Distributed Processing: of genes and environmental forces during prenatal and Implications for Psychology and Neuroscience. Oxford: postnatal development. Oxford University Press. Generative linguistics The dominant paradigm in lin- Muller RA (1996) Innateness, autonomy, universality? guistics from the 1960s to the early twenty-first century. Neurobiological approaches to language. Behavioral and The refers to 's proposal that the Brain Sciences 19: 561±610. goal of linguistics should be to understand how novel Newell A (1990) Unified Theories of Cognition. Cambridge, sentences are generated. MA: Harvard University Press. Language acquisition device Linguistic relativityhy- Pinker S (1994) The Language . Cambridge, MA: pothesis The proposal that different are MIT Press. incommensurate systems of communication and thus Rumelhart D and McClelland J (1986) Parallel Distributed utterances in one language can only be understood Processing: Explorations in The Microstructure of relative to the broader language system. Related to Cognition, vols 1 and 2. Cambridge, MA: MIT Press. the Sapir±Whorf hypothesis, which is that the form of language influences thought. Named after the anthro- Glossary pological linguistics of the first half of the twentieth century who wrote the uniqueness of different language Architecture of cognition General term for the design and the necessity of evaluating a culture on its own of the human cognitive system. Specific proposed terms (cultural relativity). architectures may differ in their structure, such as the Regularityextraction General term for inducing the types of memory and processing structures. dominant statistical tendencies in a set of data. Artificial Intelligence The field of study dedicated to Information processing A view of human cognition em- understanding how intelligence emerges from compu- phasizing the analysis of large tasks into their subcom- tational systems such as computer programs. ponent processes and information structures, The dominant paradigm in psychology analogous to steps in a computer program. from the 1930s through the 1950s. Behaviorists Von Neuman computer The original computer architec- claimed that human and animal behavior could be ture consisting of registers in which arbitrary informa- understood by analyzing the relationshipbetween stim- tion bits could be stored, and a centralized processor uli presented to an animal and its response. which serially moved information between register lo- Cognitive revolution The two decades beginning in the cations. 1950s when the dominant behaviorist paradigm floun- dered and gave way to new ways of thinking about

Keywords: (Check) cognitive revolution; cognitive linguistics; Chomsky; language acquisition; connectionism

0559t001 Table 1. Theoretical perspectives on the language±cognition relationship

Timeline Movement Main source of constraints Language/cognition 1957±present Chomskyan linguistics Innate Language unique, unlike cognition 1960s±1990 Artificial intelligence Learned Subject to same principles 1980s±1990 Connectionism Learned Subject to same principles 1980s±present Innate Language unique, unlike cognition 1990s±present Cognitive neuroscience Dynamical interaction Complex similarities and differences ECS author queries

Article 559 [Harris]

The Cognitive Neuroscience Movement paragraph 3, sentence 1, ‘R-A Mueller’ - please give first name(s) in full to avoid starting the paragraph with initials.

Further Reading

Dromi - place of publication?

Harris - volume and page numbers?

Glossary

‘Language acquisition device hypothesis’ - is there a definition missing here? ‘Named after the anthropological linguistics...relativity).’ - sentence not clear. Is this actually a definition of the Sapir-Whorf hypothesis?