“The data of child language acquisition can be adequately explained without postulating the existence of an LAD or specialised language ‘module’”.
Psycholinguists such as Skinner account for first-language acquisition in terms of processes such as conditioning. Children may come to associate the word “water” with water by repeatedly hearing the word spoken when water is present, in the same way that Pavlov’s dogs associated the sound of a bell with food. Correction and strengthening of children’s early output play an important part in this process. A baby saying “wawa” will be enthusiastically rewarded with water, but as the child gets older it will have to say “water” and later “Please may I have some water?” to produce the same result. Saying “Please may I have some water?” is arguably a complex behaviour, but operant conditioning can be shown to produce remarkably complex behaviour in animals as well. A small hungry animal can be placed in a cage with a button to press, and will at some stage press the button in the course of its random movements, at which point food will appear. Eventually the animal will learn that pressing the button causes the arrival of food, and then the experimenter can change the rules, for example, requiring that the animal press a series of buttons in a certain order. Similarly, small birds can be trained, step by step, to walk in a figure of eight. So, as parents correct children’s incorrect utterances and encourage their correct ones, their linguistic output can be honed into something quite specific. Even in the absence of explicit corrective influences, children participate in social learning, copying the way appropriate role models speak.
The initial emergence of words can be interpreted as the result of the reinforcement and shaping of babies’ natural babbling. Although the earliest babbling is universal, it then undergoes a drift, converging to the sounds of the language with which the baby is brought up. After that, anything which sounds like a recognisable word will be reinforced and encouraged.
Semantic knowledge can also be acquired by a process of trial and error. When a young child learns a new word, they hypothesise its meaning, and then external correction can help them to refine this hypothesis. The data of early semantic mistakes supports this account. Some children use “dog” to mean all animals of a similar size and shape, or alternatively all furry objects, animate or otherwise. Others use “ball” to mean all round objects, such as the moon. Under-extension of meaning is less well documented than over-extension, but this could be because it is usually undetectable – if a child thought “round” referred only to the moon, and called the moon “round”, they would not be corrected. The recorded instances of over-extension and hypothesised instances of under-extension suggest that children are indeed formulating and refining hypotheses of word meaning.
These arguments all suggest that language can be learned by universal cognitive processes, negating the need for an LAD. Another argument against the innateness of any predisposition to language comes from children brought up in isolation, who do not learn language. Contrastingly, songbirds brought up in isolation from other songbirds do learn to sing, albeit abnormally.
There are many flaws in the argument which claims language is learned by conditioning. Firstly, in an experiment such as Pavlov’s, there is one unique and predictable response to the conditioned stimulus, which happens independently of any decision on the part of the dogs. In human speech, however, the same stimulus can provoke an unlimited variety of linguistic responses. On seeing an object, people have the option of whether, and in what way, to remark on it, showing that the response is of a different nature to that of Pavlov’s dogs. Also, people can both utter and understand completely new sentences, and easily discard new ungrammatical ones, whereas the small bird in the earlier example cannot spontaneously trace out other numbers with its steps. The theory does not account for a person listening quietly to an utterance, not making any observable response, nor for a person understanding what they hear but not believing it. If the analogous “response” is something unobservable which occurs in the mind, then the theory has added an extra layer of complexity and not explained anything.
The remarkable speed at which young children learn language also testifies against the behaviourist account. At one stage children are learning nine new vocabulary items a day, which makes it unlikely that they are relying on reinforcement. The behaviourist account also relies on the tendency of parents to correct ungrammatical utterances, but in practice this turns out not to be the case. Parents show approval or disapproval of their children’s utterances based on truth, not on grammaticality. A child who presents a teddy bear wearing a sock and says “Teddy sock on” will meet with approval, while one who falsely states “Look, Teddy is wearing a sock” will be corrected. If conditioning were so important, children would grow up speaking ungrammatically but always telling the truth.
The syntactic and semantic simplicity of “Motherese” has been cited as evidence on both sides of the argument. Empiricists claim that the removal of unnecessary complexities makes it easier for children to extract a rule system from what they hear. Rationalists counter that an impoverished version of the language is actually harder, not easier, to learn. Children do learn to use and understand complex structures, which are certainly absent from Motherese, and often absent from ordinary day-to-day communication. Typical child-directed speech also includes expansion and correction of a child’s utterances, for example, if a child said “Teddy sock on” its parent would reply “Yes, Teddy has a sock on.” It has been argued that children can then learn the correct form of what they meant to say, and hence use it next time. However, an experiment was conducted in which care workers were instructed to respond to one half of a group of children with such expansions, and to the other half with conversationally sequential replies, such as “Yes; what colour is the sock?” The children in the second group were then shown to outperform the others linguistically.
The tendency of children to over-generalise morphological rules, producing forms like “goed” and “foots”, means that they do not learn by repeating what they hear, because they will not have heard these forms anywhere, but will have generated them from their own immature grammars. This suggests that the simplicity of Motherese is indeed a hindrance. If they learned by reproducing what they heard, then simple input would make their task easier; however, if their progress depends on perfecting their internal grammar until it matches that of an adult, a diverse range of complex sentences would be of most help. The behaviourist theory also accounts for children’s rapid progress by claiming that after a child reaches a certain age, its parents will no longer be satisfied with “Water!” and will require “Please may I have some water?” to achieve the same result. However, children’s actual progress exceeds the demand for it, in that they often produce more complex forms at a stage when their parents would still be satisfied with the simpler form, and they rapidly learn to produce complex utterances in cases where they are not “rewarded” for them, even in intangible terms such as parental approval.
One of the most convincing arguments for an LAD is the large and complex set of rules which children acquire subconsciously. The task is to analytically reduce the enormous and heterogeneous set of observable correct English utterances to a list of generative rules. Young children with otherwise relatively underdeveloped cognitive skills can do this in a few years, from a very limited and simplified data set. Linguists, approaching the problem with all the resources of their adult intelligence and experience, and with access to vast corpora, have never succeeded in accomplishing the same task, even after a lifetime’s work.
It is possible that this observation is accountable for in terms of the difference between the abilities of the conscious and subconscious minds. Very little is known about the relative “intelligence” of our subconscious, or about how it performs the tasks it does. People can learn to play a game such as tennis, and their assessment of the path of the ball is calculated by their subconscious, whereas to solve the necessary equations consciously would require perhaps more intellectual ability and certainly more time.
There is a difference, though, between language and other abilities such as tennis. All children learn language at approximately the same rate with approximately the same degree of success, and the process is automatic, in that there is no record of anyone choosing not to learn a language. Abilities such as tennis, music, or mathematics do have some degree of innateness, in that some people seem to be much more genetically predisposed towards them than others; but, unlike language, these abilities are learned actively rather than acquired, the learning process itself counts for more than any natural predisposition, and there is vast variation between different people’s competence at them.
From a very young age people can reject ungrammatical sentences, and correct them; but young children, and many adults, are unable to explain why a sentence is incorrect, or what is wrong with it. Linguistic development is thus ahead of metalinguistic development. This is also shown by the inability of young children to distinguish between a word and its referent, as 3-year-olds are likely to say that their favourite word is “candy” and that an example of a long word is “river”. Metalinguistic competence appears to be something we learn with our conscious minds, just like any academic discipline; and it thus highlights the subconscious nature of language acquisition itself.
There are things which everyone knows about language which could not be accounted for if it were an unspecialised cognitive process. For example, children seem to know automatically that language features structure-dependent operations, even though this is not a simple idea, nor is it transparent to the objective observation of a computer, which would tend to treat sentences as linear strings. Structures are not physically marked in a sentence at all, and yet we perceive them, and indeed impose them when they are not there.
Aitchison lists six typical characteristics of biologically controlled behaviour, which can be applied to language acquisition. Biologically controlled behaviours emerge before they are necessary, which is true of the development of language, because in human society people don’t have a survival-based need to be able to talk until around the age of puberty, when they may begin having to fend for themselves. Secondly, such behaviours do not occur as the result of a conscious decision, which again contrasts language acquisition with the learning of a skill such as sport. Three more characteristics are the existence of milestones which correlate with age and other aspects of development, the emergence of the behaviour without being triggered by external events, and the negligible effect of teaching and practice. The final characteristic is the existence of a critical period for language learning. This compares to other biologically programmed phenomena in nature, such as the metamorphosis of chrysalises into butterflies. It is clear that there is a critical period for language acquisition, and that it ends at around the age of seven. After this, the brain loses its plasticity for language learning. This is shown when adults emigrate to a foreign country and try to learn the language, as compared with people who emigrated in early childhood. Another piece of evidence comes from comparing Genie, who lived in isolation until her teens and was later taught language, relatively unsuccessfully, with Isabelle, who was found at the age of six and a half, and went on to learn language normally. The critical period is strong evidence that language acquisition is somehow biologically programmed.
There is also evidence that there is a separate language module, independent of other cognitive elements and general intelligence. Linguistic savants such as Christopher have excellent linguistic competence even though their intelligence is well below average. Sufferers of isolation aphasia have their linguistic abilities intact but they are unable to relate them to the world. They cannot name objects, or recognise semantic oddity in sentences, but they can recognise and correct syntactic incorrectness. Fodor pointed out in 1983 that our beliefs and desires do not so much influence language processing, but operate after it, otherwise we would have difficulty understanding peculiar or false sentences. This all suggests that language processing happens in a separate part of the brain, which is independent of other mental functions; and this is tantamount to saying there is a specialised language module.
Aitchison, Jean. 1996. The articulate mammal.
Crain, S. & D. Lillo-Martin. 1999. An introduction to linguistic theory and language acquisition.
Gleason & Bernstein Ratner, 1993. Psycholinguistics.
Harley, T. 1995. The psychology of language.