Mechanisms of Language Acquisition

The 20th Annual Carnegie Mellon Symposium on Cognition

Brian MacWhinney, Brian MacWhinney

• Publisher: Psychology Press
• Year: 2014
• ISBN:
  9781317757399
• Edition: 1
• Topic: Personal Development
• Subtopic: Writing & Presentation Skills

First published in 1987. Three decades of intensive study of language development have led to an enormous accumulation of descriptive data. But there is still no over-arching theory of language development that can make orderly sense of this huge stockpile of observations. Grand structuralist theories such as those of Chomsky, Jakobson, and Piaget have kept researchers asking the right questions, but they seldom allow us to make detailed experimental predictions or to formulate detailed accounts. The papers collected in this volume attempt to address this gap between data and theory by formulating a series of mechanistic accounts of the acquisition of language.

I THE PROBLEM OF
OVERGENERALIZATION

1 The Principle of Contrast: A Constraint on Language Acquisition

Eve V. Clark

Stanford University

Different words mean different things. That is, wherever there is a difference in form in a language, there is a difference in meaning. This is what, in 1980, I called the Principle of Contrast.¹ It is by virtue of this property that language maintains its usefulness as a medium of communication. As Bolinger put it, “any word which a language permits to survive must make its semantic contribution” (1977, p. ix). This applies as much to constructions as to words: “the same holds for any construction that is physically distinct from any other construction” (1977, p. ix–x).

In the present paper I focus on the acquisition of meaning in light of the Principle of Contrast. This principle makes specific predictions about acquisition that are supported by data from many different domains. It shapes the lexicon for immature and mature speakers alike. It also plays a role in establishing which forms are conventional and thus contributes a solution to why children give up over-regularizations in morphology and syntax. Finally, it helps account for individual variation during acquisition.

I begin in the first section with a statement of the Principle of Contrast together with its corollary, the Principle of Conventionality, and review its predictions about language use in general. In the next section I review the evidence for the Principle of Contrast in acquisition and show that children observe it in both expected and unexpected ways from the earliest stages in the acquisition of language. In the third section I look at the consequences of the Principle of Contrast for the acquisition of morphology and of syntax, and the role it plays in children's getting rid of over-regularizations. I then show how this principle helps account for variations in the courses children follow during acquisition. In the last section I argue that this principle subsumes several other proposals to constrain language development in that they each constitute special cases of the Principle of Contrast.

THE PRINCIPLE OF CONTRAST

The Principle of Contrast states that any difference in form in a language marks a difference in meaning. The term dog, for instance, which differs in form from horse also differs from it in meaning. This principle can be stated as:

The Principle of Contrast: Every two forms contrast in meaning.

This principle is a general one for speakers of a language. It is one that has been stated or assumed by virtually every linguist over the years.

The Principle of Contrast must be carefully distinguished from its converse, which I will call the Homonymy Assumption. This assumption is that every two meanings contrast in form. Under this view, one should never find two different meanings being carried by the same form, as in bank of a river versus a financial institution, or bat, a small flying mammal versus an instrument used in playing cricket or baseball. This assumption clearly doesn't hold in general for speakers of a language. But, within one level of a semantic field, where the words for two different meanings over time come to have the same form, the resultant hom-onymy may cause genuine confusion. Speakers then typically introduce another form to carry one of the meanings (see Orr, 1962). Aside from this special case, the Homonymy Assumption should be kept distinct from the Principle of Contrast because it may play little or no role in either adult language use or acquisition.

The Principle of Contrast is essential, though, because it helps maintain conventionality in language:

The Principle of Conventionality: For certain meanings, there is a conventional form that speakers expect to be used in the language community.

If one wishes to talk about an instance of the category ‘dog’, one had better use the conventional word dog (and not horse), or no one will understand. Conventional terms used conventionally work best to convey speakers’ intentions within the speech community. Conventional terms work in large part because speakers are consistent with the conventional meanings they assign to forms from one occasion to the next, and therefore maintain the same contrasts in meaning over time. Speakers of English use the word dog to denote dogs, not dogs one day, horses the next, and some other animal the day after. These two principles jointly constrain the choices speakers make in language use (e.g., Bolinger, 1977; Clark & Clark, 1979). Without them, languages simply wouldn't work.

If the Principle of Contrast (from now on, Contrast for short) is truly general in language, then a number of predictions follow:

1. Words contrast in meaning, so there are no true synonyms.
2. Established words have priority in the expression of meaning.
3. Innovative words fill lexical gaps and so may not be used in place of established words with the identical meanings.

The evidence for these predictions is extensive, so I will simply summarize some of the major findings before turning to the predictions Contrast makes about acquisition.

Contrast in Meaning

Evidence for the first prediction comes from the lexicon and from syntax. In both, differences in form make for contrasts in meaning. Meanings may overlap, of course, but they nonetheless contrast in at least some contexts. In the lexicon, many apparent synonyms are in fact not synonymous; they mark contrasts in dialect, in register, or in connotation. In syntax, differences in form mark differences in meaning, but some of these reflect subtle shifts in perspective or topicalization.

Lexical contrasts. Meaning differences, large and small, are characteristic of the lexicon. The study of such differences has traditionally been carried out within semantic fields where linguists have analyzed and characterized patterns of contrasts (e.g., Bierwisch, 1967; Lehrer, 1974; Lyons, 1963). While different lexical domains may be organized in a variety of ways, the property they all display is that each term within a domain or semantic field contrasts in meaning with all the others. The precise pattern of lexical contrasts will vary with the internal organization of a semantic field (for discussion, see Fillmore, 1978; Kay, 1971; Lehrer, 1974; Lyons, 1977).

Possible relations in lexical domains include those among co-hyponyms (terms contrasting at the same level). For example, horse, dog, cat, and sheep are all co-hyponyms of terms above them, hierarchically, like mammal or animal. This relation of hyponymy may hold across two or more levels. Thus spaniel, a co-hyponym of boxer, Alsatian, and Labrador, is a hyponym of and contrasts with dog, and dog in turn is a hyponym of and contrasts with animal. Contrasts in meaning, then, may hold at the same level (among co-hyponyms) or across different levels.² Contrasts may also be orthogonal, between terms that potentially belong in more than one domain. Dog, for instance, is a hyponym of animal and also of pet (see further Fillmore, 1978; Lyons, 1977).

Analyses of specific lexical domains, then, have focussed on the contrasts inherent in the meaning relations within each domain. Many contrasts are obvious but others are more subtle. All languages contain numerous expressions whose meanings overlap. In many contexts these may be exchanged for each other, and it is this degree of overlap or partial synonymy that is exploited in dictionaries or thesauri, e.g., for the adjective mature, one finds adult, ripe, perfect, due; for the verb govern: direct, control, determine, require; or for the adjective loose: inexact, free, relaxed, vague, lax, unbound, inattentive, or slack. When the entry for each of these is inspected, one moves further and further away from the original word being “defined.” What this shows, clearly, is that overlaps are not equivalent to synonymy. While two terms may be interchangeable in many contexts, they are not so in all, and it is the contexts where they are not equivalent that reveal their often subtle contrasts in meaning.

Dialect, register, and connotation. English, like most other languages, contains many apparent synonyms, but these typically contrast in meaning according to dialect or register choices, or according to emotive coloring, connotation. Terms that differ only in one of these dimensions have the same extensions: their intensions are different. It is this that may mislead. Such pairs are then perceived as synonyms and their meaning differences ignored.

Choice of a term from one dialect over another in many settings identifies the speaker's membership in a particular societal group. Dialect differences account for pairs such as autumn (UK) versus fall (Western UK and US), as well as for differences between pairs like truck/lorry, pail/bucket, sack/bag, and cup/tassie (Palmer, 1981). They also account for multiple terms with the same denotation such as cowshed, cowhouse, and byre; haystack, hayrick, and haymow; tap, spigot, and faucet, and so on. The contrasts between dialects are really no different from translation equivalents across languages like French and Hungarian or English and Hebrew. In many communities speakers may be unfamiliar with the original dialects while being familiar with some of these pairs from written sources. This, though, simply makes the pairs similar to equivalent terms from two distinct languages, e.g., house and maison.

Other apparent synonyms mark different registers (speech styles). Registers may differ in formality, e.g., the contrasts among smell, effluvium, stink (straightforward, pretentious, colloquial) or, on a similar continuum, die, pass away, pop off. Speakers often opt for Latinate vocabulary in English to mark a more formal register: compare numerous and many, facilitate and ease, attempt and try, sufficient and enough (Joos, 1961). Choices of lexical items may signal solidarity or identification with a particular social group, formality or informality, or politeness. The dimensions along which lexical choices can mark register are not clearcut, and the same choices may have different consequences on different occasions (Lakoff, 1973; Nunberg, 1978).

Yet other apparent synonyms differ in the emotive coloring or connotation each carries. That is, the speaker's choice of term can convey his attitude towards the person or event being described. Compare the choice of politician versus statesman, where the latter is laudatory and the former not (see also Orwell, 1950). Much the same contrast appears to underlie choices of skinny versus slim, obstinate versus firm, and spendthrift versus generous. The first term typically carries a negative connotation, while the second carries a positive one. Many apparent synonyms contrast in connotation.

Syntactic contrasts. Differences in form at the syntactic level also mark contrasts in meaning (e.g., Bolinger, 1977; Chafe, 1971). Consider the following pairs of sentences:

(1a)	They pulled the ropes in.
(1b)	They pulled in the ropes.
(2a)	Jan taught Rob French.
(2b)	Jan taught French to Rob.
(3a)	Jo lit the fire.
(3b)	The fire was lit by Jo.
(4a)	Bees swarmed in the garden.
(4b)	The garden swarmed with bees.

In (1), as Bolinger (1977) pointed out, the contrast is one of completion or achievement marked by the first form (the ropes were in) compared to non-completion in the second. Much the same contrast appears in (2): in the first, one infers that Rob learned some French; in the ...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• Preface
• Part I The Problem of Overgeneralization
• Part II Competition
• Part III Constraints on the form of Grammar
• Commentary
• Author Index
• Subject Index