LANGUAGE LEARNING: HUMANS

An understanding of language learning presupposes clarification of what knowledge of language consists of, what mechanisms are available for language learning, and what the course of language development is. There are two broad types of knowledge of language: grammatical knowledge, the nature of which is largely biologically determined; and experiential knowledge, which arises mostly from the learner's encounter with the world (O'Grady, Archibald, Aronoff, and Rees-Miller, 2001).

Grammatical knowledge bears on rule-governed phenomena susceptible of a precise formal description. A formal description of a language is called a grammar of the language. Typically, grammar consists of phonology (sound patterns), morphology (structure of words), syntax (structure of phrases and sentences), and semantics (meanings of words, phrases, and sentences). Experiential knowledge of language consists in part of the specific meanings and pronunciations of the individual words of a language. Such knowledge is more or less idiosyncratic; it must be acquired item by item and cannot be predicted by general rules. For example, the fact that the word pig is pronounced a certain way and refers to a certain type of animal is something that we must learn through experience. The totality of this knowledge constitutes the lexicon of a language.

Pragmatics concerns the relationships among the language, the speaker, and the speaker's knowledge of the real world. People employ knowledge of the world to infer the communicative intentions and expectations of other speakers and hearers (Bach and Harnish, 1979). For example, when someone asks, "What are you eating?" we might use our knowledge of the world to infer that they might be asking to share what we are eating. Our language may also encode aspects of the social relationships between speakers, knowledge of which is acquired through experience. For example, both an imperative sentence (e.g., "Give me a piece of pizza") and a question (e.g., "Could you give me a piece of pizza?") can be used to ask the hearer to do something. But it is part of the knowledge of how English is used that the imperative is much more direct and in many circumstances will be viewed as insufficiently polite.

Paralleling the distinction between grammatical and experiential knowledge of language are two broad areas of study: development of knowledge of formal structure of language (development of grammar) and language development (development of word meanings and of the use of language to communicate ideas and intentions and to interact socially in other ways). This article focuses on the study of the relationship between the theory of grammar of natural language and the development of grammar, since it is this aspect of language learning that is least likely to prove to be merely a special case of a more general theory of knowledge acquisition.

Universal Grammar

The most influential (and controversial) view of the development of grammar is called universal grammar, or UG (Chomsky, 1975). The UG theory claims that linguistic knowledge consists of an inborn, universal, skeletal protolanguage, the details of which are elaborated in the course of learning. (A similar view has been proposed for the development of certain birdsong systems.). On this view, innate knowledge of language has two basic components: a catalog of the fixed, universal set of possible grammatical categories along with a set of universal constraints on how these categories may combine to form phrases and sentences, and a fixed set of specific, universal grammatical principles that determine in detail the shapeliness of linguistic phrasal structures. Typically, each constraint and principle may allow for some very limited range of variation.

The grammatical categories are assumed to be the familiar lexical categories of noun (e.g., pig), verb (e.g., imagine), adjective (e.g., tall), and preposition (e.g., about), and so on, and a set of functional categories whose members have formal grammatical functions and limited meaning (such as that in the sentence, "I think that it is raining") and grammatical inflection such as tense, case, and agreement.

Here is an example of a constrained phrasal structure that allows for limited variation between languages. It is known that in natural languages there is a privileged relationship between the verb and its direct object such that the two form, or are constituents of, a verb phrase (VP) (e.g., "read the book"). The verb and the direct object can thus be said to be sisters within VP. A native speaker of English knows that in a VP, the verb (V) normally precedes its sister, the direct object noun phrase (NP). On the basis of this knowledge, the native speaker judges that "*the book read" is not a valid VP of English. (The asterisk indicates that a string of words does not constitute a grammatical sequence in the language.) But in a language such as Japanese, the proper order is the reverse of that in English: [VP=NP V]. (Constituents within square brackets are sisters.)

On the UG approach to this phenomenon, language learners do not have to learn what the sisters in VP may be; they already know that the sisters in VP are V and NP. What learners do learn about their language is the proper order of the sisters inside VP. Learners of English learn that V precedes NP in VP, while learners of Japanese learn that V follows NP in VP.

Let us turn now to an example of a universal grammatical principle. Consider first the following sentences:

1. [_S [_NP George] [_VP loves himself]].
2. [_S [_NP *George's mother] [_VP loves himself]].
3. [_S [_NP *Himself] [_VP loves George]].

If two constituents, A and B, are sisters, then A is said to c-command B and all of the constituents of B. The subject NP George in (1a) is a sister of VP, and hence the NP George c-commands loves and himself. The c-command relation is illustrated in the tree diagram in Figure 1, corresponding to the sentence (1a), where the NP George c-commands the circled constituents. Sisters are represented as branches from the same node.

Sentences (1a) to (1c) show that a reflexive must be c-commanded by its antecedent. (This principle is part of what is called the binding theory.) In (1a), the antecedent George and the VP loves himself are sisters in S; thus George c-commands himself. In contrast, George does not c-command himself in (1b) or (1c). Again, the UG claim is that the knowledge of this principle is not acquired through experience but is part of the linguistic endowment of the language learner.

Support for the Theory of Universal Grammar

Why is the UG view plausible? There are several reasons. Language apparently is learned almost entirely through exposure to examples of speech in a natural setting. That is, there is little or no explicit language instruction, and what instruction does take place is not in general perceived as such by learners. The spoken language to which the learner is exposed is not structured in the form of organized "language lessons" designed to reveal certain important properties of the language. Hence the learner is not systematically provided with information about what is grammatical and what is ungrammatical. In fact, while much of the speech encountered by learners is grammatical, much is not, yet all learners acquire the correct grammar of the language or languages that they are exposed to. Many utterances are incomplete, which also constitutes a potentially confounding input to the learner (Newport, Gleitman, and Gleitman, 1977). There is little if any explicit correction by adults of learner's errors, especially grammatical errors (Brown and Hanlon, 1970; Wexler and Culicover, 1980). In fact, it is not clear that the learner would know what the correction is about, even if such correction occurred (Pinker, 1989).

In spite of all this, children make remarkably few errors in their acquisition of a grammar, and they have essentially completed the task within the first three years of life. The implication is that in order for learning to be accomplished so rapidly and under such unfavorable circumstances, much of what the learner knows in the end must be known from the outset. There is also experimental evidence that certain aspects of children's knowledge of language are in place at the earliest stages of language use, before experience could have determined the form of this knowledge. For example, Crain and McKee (1985) have shown that children exhibit correct knowledge of aspects of the binding theory as early as age two (Crain, 1991).

Further, there is the "argument from the poverty of the stimulus," which says that much of our linguistic knowledge could not be based on experience, since the crucial evidence for acquiring this knowledge does not exist in the discourse around us. One example is the binding theory outlined above. A very different example involves sentences that contain relative clauses such as

2. John knows [_NP the man [_S that loves Susan]].

When a noun phrase is questioned in English, it must be located at the beginning of the sentence. Thus, if we replace Susan in "John said that the man loves Susan" with a question word like who, we have a sentence like "Who does John say that the man loves—." The dash indicates the original position of the moved who.

Strikingly, it is impossible to perform the same sort of replacement on a noun phrase when the noun phrase is situated within a relative clause, as in

3. *Who does John know [_NP the man [_S that loves—]].

The question naturally arises as to how we know that this sentence is ungrammatical. The literature on language development provides no evidence that there is any stage at which children learning English treat (3) as grammatical. There is no evidence that children produce systematic errors of the form given in (3), and hence no evidence that children are explicitly corrected for producing such ungrammatical sentences. There is no evidence that children are instructed as to the ungrammaticality of (3). Thus, the claim of the UG approach is that the environment of the learner simply does not provide evidence, in the form of either instruction or correction, on the basis of which the ungrammaticality of (3) could be determined. The UG approach claims that what is going on here is that there is a universal, inborn principle of language that prevents interrogatives (and other types of phrases) from moving out of a relative clause, among other structures. This principle is called subjacency (Chomsky, 1973).

Finally, there are formal learnability considerations that support the UG view. Without severe constraints on rules, formal systems that have the essential properties of natural languages are not learnable under plausible definitions of what the input to the learner is, and of what constitutes "learning" (Wexler and Culicover, 1980). It is posited that the learner is exposed to examples of utterances in context from some language L and must acquire the grammar of L, call it G (L). Only if the learner has available a highly restricted set of possible hypotheses consistent with experience can the learner's hypothesis about the grammar to be learned converge rapidly to G (L). On this approach, universal principles such as the binding theory and subjacency exist precisely because they restrict the range of hypotheses available to the learner so that such rapid convergence is possible.

At the same time, there is substantial evidence that not all grammatical knowledge is available to the learner at the earliest stages of development. Whether this is because the learner simply lacks this knowledge and acquires it gradually (in contrast to the strong UG view) or whether it is wired into the learner but not fully accessible is an open question. This issue is often referred to as continuity (Pinker, 1984). In contrast to continuity, Tomasello (2000), among others, has argued that learners first acquire specific forms and sequences and generalize to rules relatively late in their linguistic development.

Let us turn briefly to the problem of acquiring knowledge of the lexicon. For every word, the learner must learn at least the following:

• The meaning of the word—For a noun (e.g., book, grandfather, happiness) the meaning is tied intimately to its reference; for a verb (e.g., read, eat, think, see) the meaning is tied to a class of events or states of affairs (Dowty, 1979; Jackendoff, 1990);
• The category of the word (e.g., noun, verb, adjective);
• The conceptual structure of the word—Word meanings are composed of subtle and effectively arbitrary components (e.g., the difference between hop and skip, between kill, murder, and assassinate, or the difference between broil and roast). But some meaning components appear to have a privileged status in being intimately tied to the structure of sentences in which they appear. Whether the subject of a verb is a causal agent (John in "John read the book" versus the bomb in "The bomb exploded") is a more fundamental component of its meaning than is the manner of locomotion (e.g., hop and skip) (Jackendoff, 1990).

In acquiring the category and the conceptual structure of a word, the learner learns to link the word appropriately to the syntactic structures in which the word is used. Experiments on lexical development suggest that acquisition of the linguistic properties of a word proceeds very rapidly; between the ages of one and a half and six years, a learner acquires about 14,000 words, which works out to about nine words a day (Carey, 1977). Learning is "bootstrapped" by generalization of the conceptual structure of the word with that of words with similar meanings and grammatical functions (Pinker, 1989; Gleitman, 1990).

On the other hand, to acquire the meaning of a word, the learner must be able to link the word with the world. Full understanding of the word in a strict sense thus depends crucially on the learner's ability to interpret the fine detail of perceptual, cognitive, and social experience correctly. Experiments on the acquisition of word meanings show that this aspect of lexical development, not unexpectedly, is tightly yoked to perceptual, cognitive, and social development (Slobin, 1985; De Villiers and De Villiers, 1978; Bates, Bretherton, and Snyder, 1988).

In summary, then, two major types of linguistic knowledge are acquired in the course of language learning. The grammar of the language is acquired through a process in which the learner starts with a universal set of structures and constraints, and through experience fixes aspects of the grammar that the theory of grammar allows to vary from language to language. There is considerable debate as to how much knowledge of language the learner begins with and how much is acquired through experience. In addition to linguistic structure, the learner acquires a vast array of knowledge of the semantic and phonological properties of individual words and expressions, and learns how to use language to convey complex meanings.

Bibliography

Bach, K., and Harnish, R. M. (1979). Linguistic communication and speech acts. Cambridge, MA: MIT Press.

Bates, E., Bretherton, I., and Snyder, L. (1988). From first words to grammar. Cambridge. UK: Cambridge University Press.

Brown, R., and Hanlon, C. (1970). Derivational complexity and the order of acquisition in child speech. In J. R. Hayes, ed., Cognition and the development of language. New York: Wiley.

Carey, S. (1977). The child as word learner. In M. Halle, J. Bresnan, and G. A. Miller, eds., Linguistic theory and psychological reality. Cambridge, MA: MIT Press.

Chomsky, N. (1973). Conditions on transformations. In S. Anderson and P. Kiparsky, eds., Festschrift for Morris Halle. New York: Holt, Rinehart, and Winston.

—— (1975). Reflections on language. New York: Pantheon.

Crain, S. R. (1991). Language acquisition in the absence of experience. Brain and Behavioral Sciences 14, 597-612.

Crain, S. R., and McKee, C. (1985). Acquisition of structural restrictions on anaphora. In Proceedings of the Northeastern Linguistics Society 16, 94-110.

De Villiers, J. G., and De Villiers, P. A. (1978). Language acquisition. Cambridge, MA: Harvard University Press.

Dowty, D. R. (1979). Word meaning and Montague grammar. Dordrecht, The Netherlands: D. Reidel.

Gleitman, L. R. (1990). The structural sources of verb meanings. Language Acquisition 1, 3-55.

Jackendoff, R. S. (1990). Semantic structures. Cambridge, MA: MIT Press.

Marler, P., and Sherman, V. (1983). Song structure without auditory feedback: Emendations of the auditory template hypothesis. Journal of Neuroscience 3, 517-531.

Newport, E. L., Gleitman, H., and Gleitman, L. R. (1977). Mother I'd rather do it myself: Some effects and non-effects of maternal speech style. In C. E. Snow and C. A. Ferguson, eds., Talking to children. Cambridge, UK: Cambridge University Press.

O'Grady, W., Archibald, J., Aronoff, M., and Rees-Miller, J. (2001). Contemporary linguistics. Boston/New York: Bedford/St. Martin's.

Pinker, S. (1984). Language learnability and language development. Cambridge, MA.: Harvard University Press

—— (1989). Learnability and cognition: The acquisition of argument structure. Cambridge, MA: MIT Press.

Slobin, D. I., ed. (1985). The crosslinguistic study of language acquisition. Hillsdale, NJ: Erlbaum.

Tomasello, M. J. (2000). Do young children have adult syntax. Cognition 74, 209-253.

Wexler, K., and Culicover, P. W. (1980). Formal principles of language acquisition. Cambridge, MA: MIT Press.

Peter W.Culicover