Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Corinne Stevens. P5577126 TMA05 Evaluate the arguments that challenge nativist theories of language development In 1965 Noam Chomsky set the nativist perspective on language development by proposing the existence of an innate Language Acquisition Device (LAD), whereby humans are born with the pre-programmed ability to verbalize and comprehend all manner of sentences, even if they have never been heard before. He argued that although language content is provided by the environment, it is the principles of transformational grammar that are biologically inherent, claiming that a domain specific, modular structure exists in the brain, thus allowing children to “effortlessly” identify and utilize intricate language rules, to which he termed as “Universal Grammar”. According to Chomsky (1981), “Universal Grammar” is visualized as a “two-tier system”; a “hard-wired basic layer” equipped with “principles” significant to all languages, and a “partially-wired” second layer containing “pre-set options” or switches, reliant on for the setting of “parameters” (properties of a language), in order to distinguish appropriate rules, relevant to the native language that the child is exposed to. He stated that “Experience is required to set the switches. When they are set the system functions” (Chomsky, 1986), and “a few changes in parameters yield typologically different languages” (Chomsky, 1986), therefore acknowledging that the environment’s role is to help the infant “set the switches”. Nativists such as Fodor (1983) and Pinker (1994) advocated and elaborated on Chomsky’s original viewpoint, adding their own interpretations regarding language development. Fodor (1983) claimed that innate domain-specific, modular mechanisms exist in the brain to take on specific tasks such as language. Pinker (1994) supports this in his belief that language must arise from an innate mechanism in human brains, since it is far too complex to be learned, maintaining that there is “an identifiable seat [for language] in the brain, and perhaps even a special set of genes that help wire it into place” (Pinker, 1994). While nativists acknowledge that the environment has a role to play in developing these mechanisms, they believe it is over an evolutionary period that spans generations (Phylogenetic). It is the empiricist belief however, that it is the environment that plays the major role in language development and this takes place via interaction on an individual level (Ontogenetic). Fodor’s “model of the mind” consists of three main mechanisms; a transducer, for receiving external, sensory information; modular, cognitive input systems, for perceiving and processing sensation before sending the information to a central processor which deals with higher cognitive tasks (problem solving, abstract thought etc) by deciding how to utilize the information. He states that it is the input systems that are genetically predetermined, modular, encapsulated (working independently of other areas), domain specific regions that control specific processing tasks (eg. Language), and it is his conception on “innately specified modularity” has been of great interest to cognitive development theorists. Karmiloff-Smith (1992), contested his theory on “innate modularity”, hypothesizing instead, that the brain becomes “modularized” during the development process. Drawing on research carried out by Neville (1991) and Johnson (1993) on “equipotentiality” (both left and right hemispheres of the brain having “equal potential” for language development), KarmiloffSmith refers to the inherent property of plasticity that occurs during early brain development, where other areas of the brain can take over a function if a specialized area becomes damaged. Although language is associated with the left hemisphere of the brain in adults, Reilly and colleagues (1998), who studied children with focal lesions, found that language delay occurred regardless of which side of the brain the damage occurred. The most notable delays in word knowledge occurred in the right hemisphere group, therefore suggesting that the areas responsible for language acquisition in a child and are different to the region controlling language processing in adults, in addition to exemplifying that other areas of the brain can accomplish language if required. Karmiloff-Smith (1992) theorized that inherent, domain-specified predispositions exist, which eventually select brain circuits for domainspecific tasks, resulting in the formation of modules; however, she states that development of these modules occurs as a result of the brain “self-organising” due epigenetic activity (genetic predispositions interacting with environmental input) rather than via innate modular mechanisms proposed by Fodor. Johnson reinforces this view, stating “While regions of the left temporal lobe may be best suited for language processing, they are not critical since language can develop in a close to normal way without them. (Johnson, 1997). A “self-organising” brain is a viable concept according to Connectionist, neural network modelling. Plunkett & Marchman (1993) created neural networks to investigate language acquisition, providing evidence on vocabulary spurt, morphological inaccuracy and U-shaped development. They discovered that the consistent vocabulary spurts observed when researching early language acquisition in infants could be demonstrated by computer models, and their results demonstrated that learning became easier as a child’s vocabulary increased, hence resulting in a sudden increase in vocabulary. Plunket & Marchman (1993) proposed that a single mechanism was responsible for this, therefore giving support for empicist, single route theories, and challenging dual route theories proposed in nativist explanations of language development. Pinker & Prince (1988) hypothesized that a dual route, or two innate, parallel cognitive mechanisms are responsible for early language acquisition and mistakes occurring during the process. According to them, the mechanisms encompass rule systems regulating the correct word endings, and memory systems for learning inflections (words indicating past tense verbs or plurals, eg. –ed, s). Their theory predicts that vocabulary inflections are reliant on the memory system storing the required word and the rule system for providing the appropriate, “regular” word ending. For example, if the correct irregular past tense verb “threw” was already stored in the memory system, it would not be necessary to engage with the rule system and it would be blocked. If, however, the word “threw” was not stored in the memory system, the rule mechanism would be consulted, resulting in the application of the regular verb ending (-ed), thus leading to the morphological error of “throw” being inflected as “throwed”. Pinker & Prince (1988) argue that this over-regularization provides an explanation for morphological errors occurring, in addition to presenting their view on why a U-shaped pattern of development occurs during language production. Dual route theorists explain this pattern by claiming that a child’s initial language production is perfect and that the U-shaped dip transpires as more vocabulary is learned and more irregular verbs are encountered, therefore leading to over-regularization since the child is relying on a regular rule system that had accepted either “-ed” or “s” as default rules. The rise in the pattern materializes as a result of the child experiencing and storing correct irregular inflections, building up their memory system and minimising future errors. According to nativists such as Pinker (1991), “double dissociation” which demonstrates a failure in inflection usage provides evidence for dual route theory. Focusing on two developmental disorders, Pinker (1991) argues that research has shown that children with Williams Syndrome experience problems producing irregular, past tense verbs (Bellugi et al, 1990), while those with Specific Language Impairment (SLI) have difficulty producing regular past tense verbs (Gopnik & Crago, 1991), therefore indicative of two systems being required for inflection production. Conversely, Thomas et al (2001) observed no discrepancies in regular/irregular past tense verb production in children with Williams Syndrome and Bishop (1997) also discovered no clear-cut dissociation in SLI children, therefore rendering Pinker’s evidence for dual route theory as questionable. Dual route and Universal Grammar principles are contested by empirists such as Rumelhart & McClelland (1987) who argue that all inflections are memorised within a single route procedure. This theory gives consideration to huge memory capabilities, which opponents allege is unfeasible; however, using connectionist models to support their claims, Rumelhart & McClelland (1987) found that large numbers of inflections could be memorised by simple neural networks. Also supporting this theory, Plunkett & Juola (1999) found that a trained network of 5000 different inflections exhibited similar U-shaped development during training to that found in children developing language. Results were also consistent with Brown’s (1973) conclusions on children’s grammatical development, demonstrating that plural inflections preceded past tense inflections. Single route theories claim that as more inflections are stored to memory, over-regularization errors become more common. This is due to rising competition between words for memory storage, resulting in “interference effects”, where similar sounding words and their inflections easily present problems. For example, similar sounding words such “man” and “fan” have different plural inflections; “fans” is regular and “men” is irregular. If many regular words of similar phonologic structure are stored, irregular words such as “man” may be over-regularized; however like dual route theory as more irregular words are stored, less errors are made. Single route theory predicts that irregular verbs of similar sound to regular verbs will be susceptible to having “-ed” added whereas dual route theory states incorrect inflections are uninfluenced by similar words. Marchman (1997) demonstrated in her investigation involving 74 children that an irregular verb (eg. blow) with more regular verb “enemies” (eg. mow, show, stow) than irregular verb “friends” (eg. throw, know) would be more vulnerable to having “-ed” added incorrectly and this is consistent with single route theory. Further evidence favouring a single route explanation of German inflectional morphology was illustrated in Szagun’s (2001) longitudinal study of 20 German children, where it was found that very young children made errors immediately when using plural inflections, as predicted. Nativist critic, Tomasello (2000) refers to the “piecemeal” nature in which infants build up their grammatical knowledge when acquiring syntax, claiming that this is because they have not yet learned what constitutes as “verbs” or “nouns”, fundamental to the nativist Universal Grammar theory. Tomasello’s “distributional approach”, suggests that verb meanings play a major role since they provide vital clues to rules regarding sentence structure. For example, infants in Naigles (2000) preferential looking task that used transitive (needs object) and intransitive (no object) verbs to examine whether meaning could be attached to grammatical context found that they were able to link phrases to the images, therefore demonstrating that their sensitivity to grammatical terms could enable them to learn about verb endings. Tomasello argues that early categorisation of “verbs” and “nouns” are “emergent properties of a learning process, rather than prerequisites for it” and that creativity in building sentences arises from growing “pockets of knowledge” that stimulate each other, leading to “systematic expression of grammatical regularities” His approach, however, offers no explanation for how children overcome the over-generalization errors found in dative alternations (sentences with different structures that have the same meaning) or question formation. Using a neural network model trained to predict the next word in a sentence; Elman (1990, 1993) found that the computer was able to use the inbuilt features within sentence structures to detect verbs and nouns and work out the correct sequence of those categories. He also discovered that the computer’s ability to learn grammar was dependent on the sentence type it was exposed to. He discovered that if it was exposed initially to simplified grammatical sequences, it would perform better in learning more complex grammatical sequences later in training. Some researchers refer to this simplified dialogue as “motherese” or “child directed speech” (Snow, 1972) and argue that it teaches language that assists grammatical development. (Furrow et al, 1979). Bruner (1975) argued that “true language” arose from scaffolding a child’s learning using simplified speech to talk about objects. Despite uncertainty regarding the motherese hypothesis, Elman’s experiment is congruent with it, as it is with Tomasello’s “distributional approach”, therefore demonstrating that “grammatical development proceeds in a piecemeal fashion”, in addition to demonstrating that interaction is necessary for language development. To conclude, this essay has attempted to focus on the viability of the empiricist challenge to nativist theories on language development. Nativists argue that language is innate and controlled by a domain specific, modular mechanism that is pre-programmed with grammatical knowledge exclusive to any native language. Empiricists reason that cognitive mechanisms develop as a result of direct interaction with the environment or via epigenetic principles where brain is able to “self-organise” to correspond with environmental input. Cortical lesion research has been inconclusive in finding a specialized area for language acquisition in children; however research on “equipotentiality” using ERP scanning has provided evidence for brain plasticity and Connectionist modelling has confirmed “selforganisation” does take place. Connectionist modelling has also provided substantial information that favours empiricist, single-route and “distributional approach” theories, whereas “double-disassociation” has not yet been conclusive for dual route theory. Some aspects of nativist theory are conceivable and useful for understanding phenomena such as language recognition; however, empiricist theories are reinforced by more convincing evidence, with epigenetic research being the more rational approach to future language development exploration. 2,092 words References: First Relationships in John Oates, Psychological Development and Early Childhood, Oxford, Blackwell/ The Open University. Brain and Cognitive Development in Denis Mareschal, Mark H Johnson and Andrew Grayson, Cognitive and Language Development in Children, Oxford, Blackwell/ The Open University Reading A: “The fragmented mind” from what infants know in Jacques Mehler and Emmanuel Dupoux, Cognitive and Language Development in Children, Oxford, Blackwell/ The Open University Reading B: Beyond Modularity in Annette Karmiloff-Smith, Cognitive and Language Development in Children, Oxford, Blackwell/ The Open University The development of children’s understanding of grammar in Kim Plunkett and Clare Wood, Cognitive and Language Development in Children, Oxford, Blackwell/ The Open University Reading A: An extract from The Articulate Mammal in Jean Aitchison, Cognitive and Language Development in Children, Oxford, Blackwell/ The Open University Neural Networks Study Booklet, ED209 Child Development, The Open University