Constructing a Language: A Usage-Based Theory of Language Acquisition

  • Michael Tomasello
Harvard University Press: 2003. 388 pp. $45, £29.95
Credit: AL GRANT

In 1965, Noam Chomsky posited that linguistic theory should be able to account for how children acquire a first language. In so doing, he triggered debates that have lasted ever since. What is innate? What and how much language do children hear? Are children's errors corrected? Do children use two different mechanisms for learning grammar: one for regularities in syntax and morphology that can be described with rules, another for irregular forms?

But few linguists have spent much time looking at language acquisition itself, preferring to debate the logic of the enterprise within linguistic theory. They have tended to ignore findings from studies of acquisition that are inconsistent with their favourite theories. In short, they have not recognized that describing a language is not the same as describing the process by which people acquire it.

By contrast, in the past few decades, psycholinguists have documented many of the facts of acquisition. They have examined what speech children hear, studied the processes for learning complex systems, and identified factors that influence development. But they have also generally ignored changes in syntactic theories — descriptions of the rules that govern language — and the associated issue of just what is innate about language in humans.

Tomasello has added a new perspective to these debates from the psycholinguistic side, based on his work with primates as well as children. He has brought together a number of the topics that psycholinguists have worked on: language studied as a system for communication, the relationships between language, memory and attention, how inferences about meaning are made in context, choices of conceptual perspective — the decision to call a dog a “dog” rather than an “animal” — how common ground is built up in communication, and how a speaker's intentions are interpreted. He emphasizes that language is essentially social and that it relies not only on vocabulary and linguistic constructions, but also on non-linguistic elements such as gesture and gaze.

He starts from the premise that children acquire language by attending closely to the language they hear. To do that, they must analyse speakers' intentions and find any patterns in the language that speakers use. Tomasello argues that children acquire constructions in the same way as they do words: they have to learn both, and just as they slowly build up their vocabulary, they also slowly build up a repertoire of constructions. Words, in fact, are stepping-stones to constructions. For instance, children first use a verb like “want” only with “that” (“want that”), then with a following verb (“wanna go”), and only later still with a direct object and following verb (“want him [to] go out”). They build up larger constructions by combining smaller ones. Because Tomasello looks at speaker intentions as well as patterns of use, he integrates the cognitive and the social in language development from the start.

Researchers have always assumed that children acquire vocabulary by learning, but many have argued that learning alone can't explain children's acquisition of the regularities of language that can be described in rules for syntax and morphology. Acquisition of these, propose Steven Pinker and others, depends on innate language-specific capacities. These capacities, argue researchers, are shared by all languages and include innately given word classes.

But in Tomasello's view, acquiring a first language entails mastering more than its grammar. It means learning to use the language to communicate, using the same resources that adult speakers do. The child's abstraction of grammatical rules, as sketched out in Chomsky's proposal, remains an important part of this task but, as Tomasello points out, it is unclear how quickly children identify such rules.

Tomasello presents a wealth of observation and argument in support of his approach. He appeals to recent linguistics research on constructions in syntax, and to psychological research on children's understanding of intentions and beliefs in others, on joint attention in communication, and on function-based distributional analyses and analogy in learning. He makes a compelling case for his view of acquisition as an alternative both to those linguistic accounts that have focused on grammar and on how much is innate, and to current 'connectionist' accounts by Jeffrey Elman and his colleagues that focus on the forms learnt but not on their meanings. He argues against the idea that there are different mechanisms for learning rule-based and irregular forms of language, and in favour of a single mechanism for learning both words and constructions. And, like many biologists, he cautions against assuming innateness without examining the alternatives.

Tomasello should make us think more, and more carefully, about language in social as well as cognitive terms, and to consider the roles of attention, memory and learning in the process of acquisition. But he also leaves many questions unanswered. For example, what are the units of language being learnt? How should one define them — the notion of clause, for example? When do children learn to rely on conventions — that “dog” designates the category of dogs in English, but “chien” does so in French? Did languages evolve in the kinds of settings where adults and infants first establish joint attention? How do children learn the meanings of words and constructions? How is the choice of a conceptual perspective by the speaker — “that dog” or “that animal” — related to the build-up of common ground in conversation? How do children get rid of errors such as “comed” (for “came”) or “me throw” (for “I want to throw it”) in their speech? Are they attentive to corrections from adults, and if so, how? Are they, in fact, exposed to enough information about constructions to allow for learning?

These are all empirical questions that must be taken seriously. How children acquire language can no longer be presented as a thought-experiment about the language that is the product of acquisition: it demands concrete data and theory about the process of acquisition.