Chomsky believed that Universal Grammer is something that everyone is born with regardless of what language they speak. Universal grammar is thought to contain a set of constraints on language processing which can be switched on or off through exposure to spoken language (Plunkett and Woods, Pg.172). Aitchinson discussed Chomsky’s theory in Reading A an abstract from The Articulate Mammal which states that we can think of Universal Grammar as an intrical structure system, but one that is only partly wired up. The system can be associated with a finite set of switches, each of which has a finite number of positions. Chomsky suggests that that experience is required to set the switches and that once they have been set the system functions (Plunkett and Woods, Pg. 202).
Pinker, (1994) also a nativist further developed Chomsky’s ideas. Pinker used four main points to support the argument that language is innate. These were; that children who grew up in pidgin-speaking communities (make-shift jargon) grew up to speak Creole (full grammatical language). He believed this to be evidence that children are able to reinvent language because language is innate (Mareschal et al, Pg 138). Drawing on Chomsky’s ‘argument from the poverty input’, Pinker also suggested that children are able to produce language that they have never heard before, which implies that they are not just learning through imitation. He also claims that strong similarity in grammatical structure across different languages reflects the common structure of the brain and he also suggested that there is an area of the brain specifically for language (Mareschal et al, Pg 138). Pinker supported a dual route theory of language development which suggests that plural and past tense inflection is achieved through ‘two cognitive systems operating in parallel with each other’ (Plunkett and Wood, pg.182).
I will now discuss some of the recent challenges to the Nativist theories of language development. These challenges mainly fall under Empiricism. Empiricism is the theoretical position which argues that development is primarily driven by factors in the environment (Plunkett and Wood, Pg.165). Empiricists challenge the nativist theories of language development as they believe that language is not innate but learned within the environment. They disagree with the nativist’s theory of Universal grammar and suggest that language is built up in a ‘piecemeal’ fashion which is based on regularities in the way words are used within the language (Plunkett and Wood, Pg.194). This ‘piecemeal’ fashion of acquiring language is also known as the distributional approach. The distributional approach was proposed by Tomeasello.
Tomeasello discredits the nativists’ belief in Universal Grammar and suggests that infants build their grammatical knowledge through exposure to language in their environment and do not have innate knowledge of categories such as nouns and verbs. The distributional approach suggests that these categories are actually emergent properties of the learning process (Plunkett & Wood, Ch.4, Pg.94). Empiricists also propose that all inflections are produced by a single memory system in comparison to the dual route theory favoured by nativists. This single route theory is based on the belief that only one system, a memory system is necessary to produce grammatical utterances. Evidence of a single route system has been found with the use of connectionism.
Connectionism has provided Empiricist with substantial data which challenges the Nativist position on language development. Connectionist networks are which are designed to model some properties of biological neural networks. The computer simulates the ‘nodes’ and ‘connections’ that are similar in structure and function to neurons and axons in the brain. The development of connectionist models (neural networks) of language development enable theorists to better understand the cognitive process of language and also provides them with a means to test their theories and predictions. Elgman (1993) discovered that the order the computer was exposed to different sentence types had an effect on its ability to learn grammar (Plunkett & Wood, Ch.4, Pg.196). He found that by initially providing the program with simple grammatical sentences lead to greater success in later training. This idea of simple grammatical sentences enabling language development co-relates with the special style of speaking which is naturally adopted by parents called child directive speech or ‘motherese’. Child Directed Speech (CDS) or 'Motherese', is a special style of speech where parents speak in a higher pitch, stress important words, and talk more slowly to their infants. This theory of initial simple grammatical sentences aiding the progression to complex language also correlates with the distributional approach which suggests that children learn language in a piecemeal fashion and these findings therefore challenge the empiricists theory that language is innate.
The connectionist model or neural networks as it is also known has managed to demonstrate the vocabulary spurt. The vocabulary spurt is the period in which children ‘show a sudden increase in the number of name that they use for things’ (Quinn & Oates, Ch.1, Pg.94). Plunkett and Marchman (1993) found that when they developed a program on the neural networks programme which matched labels to images they noticed that the computers learning replicated the typical vocabulary spurt that occurs in young children. Rumelhart and McClelland (1986) discovered that neural networks could also demonstrate the U-shaped development pattern, where good early performance suddenly deteriorates but later improves (Plunkett & Wood, Ch.4, Pg.183).
Their findings correlated with the single route theory (the belief that only one system is necessary to produce grammatical utterances) as it suggests that learning starts to increase rapidly once a certain level of learning has been achieved which implies that learning is the function of learning itself (Harris, Ch.2, Pg.80). Neural networks (when damaged) have been found to react in similar ways as children with disabilities. Seidenberg and McClelland (1989) found that when a model trained to read words will, when lesioned, sometimes behave in a similar way to a child with dyslexia (Neural Networks Study Booklet, Pg.11). As with most theories connectism still has its flaws. One of these flaws is the fact that it is a computer being used to replicate the functions of the brain and many theorists question whether it is actually possible to really replicate the functions of a human brain. This would suggest that any evidence provided by this method could be considered unreliable and invalid. It is also argued that many of the processes that the computer is able to model have had to be simplified to such an extent that it has made them meaningless. These are just a few criticisms of neural networks.
In conclusion, it apears that the field of language development is a controversial one and this is probably due to the complex nature of language development and the importance it plays in the overall development of a child. Several theories have been put forward which attempt to account for basis of language development, however it seems that these theories have tended to focus on small areas within the broad spectrum of language development which seem to support their assumptions. There are many theroreticians who believe that both nature and nurture play a part in the aquisisition of language development, that language development is a combination between biological as well as environmental influences. I would have to agree with this assumption. Although after evaluating recent challenges to Nativist theories of language development, it is my opinion that Empiricism with the use of Connectionism appears to be a very credidble method of explaining the basis of language development. Connectionism has the ability to explain and demonstrate the single route theory and can test and provide evidence of distributional accounts of syntax, the vocabulary spurt, and the U-shaped pattern of language development. Although Connectionism fails to completely account for all areas of language acquisition and it certainly has some flaws, Connectionism still seems to provide a very firm basis on which future research can be carried out.
Word Count 1856
References
Harris, M. ‘First Words’ Cognitive and Language Development in Children. Blackwell Publishing in association with The Open University (ED209 Book 3)
Mareschal, D., Johnson, M. & Grayson, A. ‘Brain and Cognitive Development’ Cognitive and Language Development in Children. Blackwell Publishing in association with The Open University (ED209 Book 3)
Neural Networks Study Booklet, ED209 (2006)
Plunkett, K. & Woods, C. ‘The Development of Children’s Understanding of Grammar’ Cognitive and Language Development in Children. Blackwell Publishing in association with The Open University (ED209 Book 3)
Quinn, P. & Oates, J. ‘Early Category Representations and Concepts’ Cognitive and Language Development in Children. Blackwell Publishing in association with The Open University (ED209 Book 3)
Wikipedia Online Dictionary. (accessed on 12/7/06)