The goal of acquisition theories is to explain how it is that any normal child, born into any linguistic community, learns the language (or languages) of that community. For many theorists, the challenge is also to explain what appears to be the relatively short time period in which acquisition is achieved, the fact that it appears to be done without either overt teaching or sufficient information from the input (what the child hears [or sees, in the case of sign language]), and to follow a path that seems remarkably similar in all children, despite variation in early childhood experiences and in the types of languages they are exposed to. There is also a consensus that language acquisition is largely independent of cognitive development, despite the fact that some deficits in cognitive development can have an effect on certain aspects of language development. Whether the language is a spoken language or a sign language, whether the language is highly inflected like Finnish or uninflected like Mandarin, whether the child is raised in poverty or luxury, by highly educated or illiterate adults, or even other children, it seems that normally developing children pass through roughly the same stages in the same sequence, and achieve the steady state of acquired language by about the same age.
We know from unfortunate natural experiments in which children are raised in isolation (or near isolation) from language-using older members of our species (i.e. they are severely neglected, or are raised by other animals) that language does not emerge without at least some linguistic exposure—input—during the first few years of life. This strongly suggests that there is a critical or sensitive period during which the mechanism or mechanisms responsible for language development is/are primed to receive input. However, the resilience of language development to quite wide variations in input within any given language community, as well as the similarities among children learning quite different languages, suggest that these mechanisms, whatever they are, must either be quite tolerant of such variation, or be primed in such a way that the crucial input for language acquisition is always made available.
There are important roles for both ‘experience dependent’ (nurture) and ‘experience-expectant’ (nature) learning in language development, and theories are distinguishable in terms of the relative contributions they see for these two types. On the one hand, there are those researchers who see a large didactic role for input (experience-dependent learning), and on the other, there are those who see a much smaller role for input and a much larger role for genetic predispositions that are triggered by linguistic experience (experience expectant learning). Theories are also distinguishable in terms of whether they are trying to explain how language emerges on a day-by-day basis in any given child or whether they are trying to explain how what is perceived as a gulf between experience as a child and adult knowledge of language could be bridged in principle. The latter are engaged in trying to solve the ‘logical problem of language acquisition’.
Another dimension of difference between theories concerns the nature of the experience-expectant (innate) aspects of language acquisition. There are those researchers, most notably the generative linguists in a broadly Chomskyan paradigm, who argue for a dedicated language acquisition device (LAD), that has evolved to serve the precise purpose of language acquisition. This device is primed specifically to receive linguistic input, and requires a minimal amount of it to set to work building mental representations for language in the mind of the child. The fact that the required triggering input is so minimal provides an explanation for the consistency of language acquisition paths across otherwise fairly widely varying life experiences.
The most elaborated version of the LAD account – the Principles and Parameters approach—suggests that children are born with a Universal Grammar (UG), which means they are (unconsciously) anticipating those features that are common to all languages (the principles), as well as limited options for those things that differ among languages (the parameters). Upon exposure to actual input from a given language, children are able to ‘decide’ which sort of language they have encountered. So, for example, some languages have basic subject–verb–object organization in which complements are attached to the right of the heads of phrases (thus objects follow verbs, relative clauses follow noun heads, and noun phrases follow prepositions), while other languages are subject–object–verb where the reverse order of complements is found. A child exposed to a language of the first type need only process a simple structure (say one with a verb followed by an object) and it will trigger the expectation that all the other head-complement structures will be in the same order. When all the open parameters have been set, the child possesses the ‘core’ grammar of the specific language he or she is exposed to. At the same time, however, the child has also been acquiring those aspects of the language that are not anticipated by UG, using experience- dependent learning. These aspects are often together referred to as ‘peripheral’ grammar. Some researchers in this paradigm have assumed that all principles and parameters are operational or sensitive to the input from the beginning of life. Others have suggested that some may at least emerge with maturation.
In the course of acquisition, generalization (and overgeneralization) allows new knowledge to permeate across-the-board, and ‘bootstrapping’ allows learning in one area of language to trigger new learning in another. Semantic bootstrapping involves understanding an utterance in context and using it in conjunction with innate expectations of language principles to crack the code of the syntax. (For example, a child who does not yet know the required complements of the verb ‘put’, can work these out from understanding utterances such as ‘Put the cup on the table’ in context.) Syntactic bootstrapping involves working from an already understood structure to fill in meanings, semantic information, by deduction. (For example, if you hear ‘John glopped his friend on the head and he fell down’, you may not know exactly what ‘glopped’ means, but you can work out a lot of what it must mean.)
The Principles and Parameters (P&P) model has been a highly influential linguistic approach in language acquisition research, even while researchers in psychology and anthropology have been pursuing significantly different lines of investigation. In linguistics, the P&P model lies at the intersection of generative (specifically Chomksyan) linguistic theorizing about the nature of adult mental representations for language and accounts of how children acquire language. It has evolved as an account of how language could actually develop across time, even while its roots are in the ‘logical problem of language acquisition’ because it assumes that what cannot be learned from the input must be genetically prespecified. It advances various arguments in support of the thesis that the input is in fact incapable of providing sufficient information for language to be learned entirely through experience-dependent mechanisms, and thus that there is an essential problem of the ‘poverty of the stimulus’. The account is also strongly modular in the sense that it sees language as being acquired by a specially dedicated mechanism (the LAD), independent of other types of mental representations or mental functioning.
Despite its power within linguistics, accounts predicated on some version of the UG story actually attract only a minority of adherents within the broad field of child language research. Most researchers are convinced that language in its entirety can be worked out by the child on the basis of the input, coupled with innate (nonspecific) predispositions for analyzing their environment. As a result of their analysis, children possess the capacity to produce and comprehend language. They may also have stored mental representations for how language works, but, unlike the P&P account, most of the accounts that rely on processing of the input as the way in which acquisition occurs also regard the endpoint as processes for production and comprehension, without independent mental representations.
Ann Peters and Dan Slobin, for example, have argued that children possess strategies, operating principles, for carving up the speech stream into repeating bits, storing those bits, recognizing the commonalities across them, and thereby building up a performance grammar for comprehension and production called Basic Child Grammar, which subsequently becomes elaborated into an adult grammar.
Another of the processing theories of language acquisition is Elizabeth Bates’ and Brian Mac Whinney’s Competition Model. Unlike the operating principles approach, this does not argue for a steady state in the form of a grammar at the end of acquisition, but rather for a permanently dynamic response to input throughout life. The impression that language has been definitively learned comes simply from the fact that new input changes the child’s system very little if a child remains in the same speech community (although it will usually change with exposure to a new dialect or to a new language). Acquisition takes place as children respond to the distribution of various cues to meaning (word order, inflectional information, intonation and stress, etc.) and respond probabilistically to conflict among them. Since each child processes the input independently, individual differences between children are expected, and advocates of this approach argue that the differences found among children support the model. As they learn more of their language, they pay attention to more and more of the cues and let the stronger ones win out over the weaker ones for their language. So, at an early stage, a child learning English may assume that nouns at the beginning of sentences are agents, but when they begin to pay attention to passive morphology, they will have to adjust their assumptions accordingly. As should be clear from this example, the ability to derive meanings for language from context, in advance of actually understanding how language is structured is crucial (as it is in semantic bootstrapping). Active application of this and other distributional models, aimed at demonstrating that the input is sufficient to account for language acquisition, is seen in the computer-modeled connectionist networks. Attracted by the architectural similarity between computer networks and neurons, these researchers argue that experience with language teaches the network so effectively and so quickly that it gives the impression of prior knowledge.
As already indicated, children frequently seem to rely on pragmatic expectations of what language ought to mean in a given situation as a basis for learning how it is structured. Some theories of language acquisition place an even greater emphasis on the role of the sociopragmatic environment. Jerome Bruner, for example, argues that the behavior of a child’s caretaker provides all the cues that the inherently social child needs to acquire language. The way in which language is used in the here and now, in conjunction with actions that match what is being said, ‘scaffolds’ the child’s understanding of the language. Others see difficulties with this approach, not only because not all successful language learners receive the kind of careful scaffolding it seems to require but also because the parental complexity of language seems to follow rather than lead increasing complexity in the child’s language. In these and other ways, the account seems too simple to account for the complexity of the task. A similar approach was advocated by Jean Piaget, who saw language development as the logical extension not of social behavior, but of cognitive development. In Piaget’s initial proposal, embedding of sentences was seen as analogous to nesting boxes, and the former dependent on the latter. Although most of his specific predictions have not been supported by subsequent research, it is clear that at least certain aspects of language are intimately related to cognitive development, although the direction of influence is not clear. One area of current research concerns the emergence of the capacity to make informed guesses about what other people know (‘theory of mind’), and its relationship to language development. There is considerable discussion about whether the capacity to embed clauses under main verbs such as ‘thinks’ and ‘know’ drives or is driven by the capacity to understand that others may not share the same assumptions as oneself.
Finally, it is worth noting that significant debate surrounds the issue of whether, when an individual learns a second language, they go about it in the same way as a first language learner, or whether it is a fundamentally different process. Evidence suggests that there is some kind of critical period for second language development, as children seem to be able to do it so much better than adolescents and adults. However, it is not yet known when bilingual first language development should be seen as having given way to second language development. Nor is it clear that adults, when given the kind of input and motivation of children, are always incapable of the same level of success. It is also unclear whether second language learners are able to reaccess the learning capacities they had as children learning their first language, or whether these are permanently overridden and made unavailable by the presence of the first language.