Acquisition of Language
Acquisition of Language.
[This entry includes the following subentries:
The acquisition of the phonology of one's native language(s) includes mastery of the phonetic targets, phonemic contrasts, and morphophonemic alternations for the processing of language in both recognition and production (see Vihman 1996 and Jusczyk 1997 for reviews). A significant biological basis of this learning process is acknowledged even by those who reject nativist ideas about acquisition, and has been supported by crosslinguistic studies of the first two years of life. Most studies deal with Indo-European languages, notably Dutch, English, French, Swedish, and Welsh; Finnish, Chinese, Japanese, Thai, and Quiché (K'iche') Mayan are also represented.
1. Precursors of phonology: Pre-speech perception
Infants demonstrate preferential attention to the prosodic patterns of their own language from the first days of life. The fetus gains linguistic experience in the womb through hearing the sound of the mother's voice both “internally” and from the outside, filtered through the amniotic fluid. The prosodic information present in the lower frequency bands of the signal can reach fetal ears by the final trimester of pregnancy, when the auditory system is completely formed; segmental information, much of which is carried by higher frequencies, cannot. Prosody provides the initial entry into language, not only from establishment as an acoustic signal before birth but also from the intrinsic affective links between particular prosodic patterns and communicative meanings (Fernald 1992); these provide a foundation for the later-learned arbitrary (linguistic) form-meaning relations.
Young infants discriminate virtually all the segmental contrasts on which the phonological systems of the world's languages are based, and they do so whether or not a particular contrast occurs in the ambient language. For instance, American infants discriminate nasal from oral vowels, though these sounds do not contrast in English. Infant categorical response to consonantal contrasts was initially taken as evidence of innate “speech detectors,” but later research revealing a similar response in other animal species suggests that the structure of the mammalian auditory system itself may shape the response, and indeed that speech may have evolved to fit that system. (Categorical perception of non-speech sounds has been demonstrated as well; Jusczyk 1997.) Very young infants associate the visual appearance of degree of mouth opening with the corresponding vowel sound. Although this specific cross-modal link can be seen as “hard-wired” into the brain, an alternative interpretation assumes an intuitive or automatic infant production response—with or without overt imitation—to a within-repertoire phonetic gesture (arguably supported by “mirror neurons”; Stamenov and Gallese, 2002).
In the 1990s, experimental work on infant speech perception increasingly turned from studying discrimination (detection of changes) to probing longer-term representation for speech. These studies have revealed a clear developmental trend: for the first six months, it is primarily prosodic patterns that underlie a “familiarity response” to speech. Evidence of a holistic response to segmental patterns as early as four months is limited to stimuli that can be assumed to be imbued with strong affect for infants, such as the infant's own name. Aside from these exceptional words, preferential attention to a trained segmental pattern is not reported until seven months. Thereafter, steady gains in attention to native-language segmental patterning can be seen, along with a narrowing of attention to consonantal contrasts from broadly “universal” to native-language only by ten–twelve months. Cross-linguistic studies show infant preferential response to words familiar from everyday experience over unknown words by eleven months. Changes to the onset consonant of the unaccented syllable (initial in iambic French words [weak-strong], final in English trochaic words [strong-weak]) do not affect this preference, while changes to the accented syllable eliminate it. This suggests that, at this pre-speech-production stage, the accented syllable is more fully specified than other syllables in the child's representations for familiar words.
2. Precursors of phonology: Early vocal production
Infant non-crying vocal production develops steadily from a first period of reflexive sounds, through social comfort sounds (cooing and laughter, ca. two–five months), to phonatory expansion involving a range of prosodic effects—growling, yelling, and whispering. This increase in range of vocal production options accompanies anatomical changes in the infant vocal tract as it begins to assume the shape characteristic of adult humans, with lengthened pharynx and a right-angle bend. The major vocal landmark of the first year, readily recognized by an infant's caregivers, is the emergence of rhythmically produced “canonical syllables” with adult-like timing, often as a string of reduplicated babbling (six–eight months). Non-reduplicated or variegated babbling increases and differentiates gradually thereafter, culminating early in the second year, for some infants, in “expressive jargon,” lengthy babbled sequences that reproduce aspects of the ambient language intonation pattern, producing a striking impression of “conversation without words.”
The emergence of easily recognized babbled syllables in the middle of the first year appears to be maturationally based and fits into a broader framework of rhythmic motoric advances that occur around that age. One interpretation of the developmental match between the shift in the apparent basis for infant speech representation from prosody to segments (as outlined above) and the onset of CV production is the “articulatory filter” hypothesis. On this account, the experience of frequently producing CV syllables sensitizes infants to similar patterns in the input speech stream. (Deaf infants fail to persevere in CV syllable production at the typical age; compare the earlier production response to the sight and sound of vowels, noted above.)
The repertoire of sounds used in babbling is rather small, representing the core of sounds used in most languages. Within the limits of that core repertoire, babbling tends to reflect the phonetic profile of the ambient language (e.g. as regards vowel space). It also shows some tendency for consonant-vowel interactions: alveolar (“front”) consonants are more likely to be followed by front vowels, labials by central vowels, and velar (back) consonants by back vowels. Whereas the CV associations can be taken to reflect simple jaw oscillation, the statistical matching of ambient-language vowel space suggests pure (implicit) distributional learning, an effect that has been empirically demonstrated in both adults and children (Saffran et al. 1996).
The period of non-reduplicated babbling generally overlaps considerably with the production of the first recognizable words, and the repertoire of word forms preferred by a given child generally is closely related to the most frequent elements in that child's babbling repertoire (for late talkers, this may not be true). Contrary to the predictions of many theoretical approaches, many early words tend to be more accurate than slightly later ones, probably because of selection on phonological grounds. Analyses of later word forms indicate that a child's first well-practiced, consistent supraglottal production patterns provide the basis for the later development of “canonical forms” (Menn 1983) or “word templates” (Vihman and Velleman 2000). These word templates abstract from and extend the piecemeal learning evidenced by the selection patterns of first words. The apparent paradox of such early word selection—how do children know which words not to attempt, or which sounds they cannot yet produce?—does not arise if we assume that the first words result from infants' matching of their own vocal patterns to the input speech signal.
3. Emergence of phonology in early speech
A child's earliest meaningful forms may lack adult models, or they may be phonetically less controlled. Their semantic properties may also be somewhat anomalous: they may be social or action-accompanying rather than referential. For these reasons, terms such as “proto-word,” “phonetically consistent form,” “vocable,” and “context-limited word” have been used instead of “word” for the first recognizably recurrent meaningful units.
3.1. Characteristics of proto-words and early words
In contrast, some “earliest” words are accurate renditions of their adult models; if a few of them are conspicuously better approximations to the adult word than the rest of the child's output, these are termed “progressive phonological idioms.” Such early forms are likely to be produced before the child's own word-form production begins to cohere into a system. Typically, each child then develops a more or less systematic way of rendering adult words within the constraints of his/her limited output repertoire of sound-sequences by the time she is producing 50 words or so. The inaccuracies in these renditions appear to be due (with a few exceptions) to motor programming or representation problems, rather than to misperceptions of the adult sound pattern; children can usually discriminate between two words that they render identically, except for a few auditorily confusable phones like [f] and [θ].
3.2. Rules and processes
Once the child's output becomes systematic, mapping from the adult model to the child's form may be regular enough to be captured by writing phonological rules. This is not true for all children; some map entire multiword phrases onto loosely articulated sequences (e.g. I don't want it rendered [aoa:]); others reorder the sounds of adult words to fit their own restricted output templates. Output templates for words may include patterns like “second vowel higher than first vowel,” “medial consonant-glottal stop,” “first consonant labial, second alveolar,” and so on.
The shapes of such word templates are influenced by the particular language the child is learning. For example, initial consonants may be omitted in Finnish, French, or Welsh word templates, but this pattern is rare in English. The difference is probably in the rhythmic structure: French and Welsh tend to have very long final syllables, while Finnish commonly has medial long consonants. These perceptually salient features of the input lead some children to disregard initial consonants, especially if they are not yet within the child's repertoire (e.g. initial fricatives or liquids).
More systematic mapping patterns, when found, include a number of natural processes such as devoicing of final stops ([bip] for bib), substitution of stops for fricatives ([ti] for see), reduction of consonant clusters to singletons ([pat], less commonly [fat], for spot), deletion of initial [h], and deletion of unstressed syllables ([næ]. [ˈnænæ] for banana). However, since a child may fail to use such processes on her earliest words and then start to use them somewhat later, many researchers regard these processes as natural failure modes rather than as innate rules that must be unlearned by the child. Other common processes for English-acquiring children that are not prominent (sometimes not attested) in adult language include voicing of initial stop consonants and various assimilations at a distance that ensure whole-word consonant harmonies, especially place assimilation and nasal assimilation. For several languages, including Catalan, English, Estonian, and French, it has been observed that an unstressed syllable may be replaced by a “dummy syllable” (either a stereotyped near-constant shape or a copy of the stressed syllable) instead of being deleted. Cluster simplification, usually accomplished by preservation of the most obstruent element, may instead be achieved by combining features from the several segments (e.g./s + nasal/ rendered as voiceless nasal;/sk/ rendered as [x]), even when the adult language does not possess the resulting feature combination. Constraint-based approaches such as Optimality Theory (Kager et al., to appear) can provide a formal description of such mappings.
In ascribing rules or processes to a child, the investigator must take into account the surface forms of the words the child hears, including dialect variants and the reduced forms of rapid casual speech. Ambient dialect forms are targets, not “errors,” for the child, regardless of their social valuation; so are the normal assimilations and reductions of fluent speech (e.g., Eng. [hæftə] for have to. They cannot be analyzed as outputs of phonological processes for the child who is too young to be aware of the underlying forms.
4. Strategy variations
Initial avoidance of adult words outside a child's pre-speech production repertoire can be accounted for by positing that the child has a better capacity for representing practiced than unpracticed sound sequences, although some sounds produced in babble apparently cannot be used in early speech. Individual children vary in the extent to which they prefer to avoid adult words that they cannot yet render accurately. Experimental work concurs with diary studies in showing that such avoidance is especially likely in children who have produced fewer than about 75 different adult words. As a system develops, some children continue to avoid out-of-repertoire words; others freely adapt such words to fit within their templates.
5. Linguistic perception vs. production: Development after the onset of speech
Investigating the relation of perception to production during the period of acquisition of the sound system has required devising experimental techniques appropriate to the rapidly changing cognitive and motor abilities of the infant and young child. Although infants have phonetic perceptual and categorization abilities similar to, or indeed better than, those of adults, children as old as two often confuse minimal pairs and have a heavy bias to respond to the less familiar member of such a pair of stimulus words as though it were the more familiar one (e.g., pointing to a coat instead of a goat after the word goat has been especially taught for the purpose of testing the k/g discrimination). Clearly, different mechanisms are involved in discriminating repeated meaningless syllables as opposed to words; for example, differential response to words invokes long-term memory for the sound-meaning correspondence (even if the experimental paradigm does not require this), while the discrimination paradigms used with infants do not require such representations.
Children are not always aware of the degree of accuracy of their attempts. Numerous studies have reported interchanges of this kind: a child of two or three mispronounces a word (e.g. [fis] for fish), an adult requests the child to say [fiʃ] (emphasizing the final [ʃ]), and the child responds I did say [fis] (with similar emphasis). Since very young infants can, in general, make such discriminations, a possible explanation for such responses is that the toddler is listening “through” the surface form of the word to access the meaning rather than attending to the phonetic details. Experimental studies show such a shift in attention, from auditory discrimination of novel speech forms to sound-meaning mappings with discrimination of holistically dissimilar forms only, between eight and fourteen months (Stager and Werker 1997). After the child becomes aware of the difference between the target and her production, she may still be unable to manage any change in production (and may be able to comment on this). Later, correction can be approximated in imitation but not maintained in spontaneous speech; still later, a more correct form and the earlier form appear variably in production; and, finally, the more correct form is maintained. Progress through these phases may be rapid or slow, and a child may be at an advanced stage with respect to an early-acquired sound or sound-sequence while at a beginning stage with respect to a more difficult target.
6. Order of acquisition of phones and contrasts
The order of mastery of phones and phonemic contrasts and the age of mastery are variable across children. Phones that are dependent on precise relative timing of glottal and supra-laryngeal events (e.g. aspirated or glottalized stops) are generally acquired later than those that are not so dependent, other things being equal; also, phones that seem to require more precise positioning without tactile feedback (fricatives and liquids as opposed to stops) are acquired relatively late. The order of contrast acquisition proposed by Jakobson 1968 on the basis of general markedness is a good fit probabilistically, but many exceptions have been published, so his description in terms of “laws of irreversible solidarity” can no longer be considered appropriate.
Phonotactic constraints (evidenced by canonical forms and word templates) play a critical role in acquisition; as a result, the “order of acquisition” question is often unanswerable, even for an individual child. A particular phoneme might, for example, be generally correct for its intervocalic allophone, fully correct word-initially, absent in particular clusters, and rendered incorrectly in word-final position—all this apart from the possible interference of consonant assimilation in particular words. Although new consonant phones often are first mastered as initial singletons (at least in English), there is much individual variation; for English-speaking children, it is not uncommon that some or all of the fricatives first appear word-finally, while stops first appear word-initially. No general theory is yet able to account both for the degree of observed variation and the degree of observed commonality of acquisition of phones and phonemic contrasts across children.
7. Later phonological development
Well before children finish mastering the phonetic details of the ambient language, they have begun to acquire its morphology and morphophonemics. Phonological rules with a heavy natural component (e.g. voicing assimilation) appear to be acquired early, with the correct output form always present. Frequency and other factors interact in the acquisition of rules that apply to particular classes of morphemes (MacWhinney 1978), including the transparency and reliability of any semantic and/or phonological conditioning factors. Forms in which an ending happens to be very similar to a final stem syllable or consonant (e.g. the Eng. [-Iz] plural allomorph) tend to be acquired late. Rote application of a rule to a few frequent words precedes general application of the rule. Overgeneralization of a rule to cases where it should not apply is taken to be the hallmark of actual rule learning, and this interacts with the learning of subregularities conditioned by phonological factors or by membership in gender classes and conjugational or declensional classes.
Some nonproductive but well-evidenced morphophonological rules—such as the stem-final palatalizations in the Romance stratum of English (correct/correction; invade/invasion)—are apparently not learned until mid or late adolescence. Morphophonemic development in languages with complex morphology has not been well studied. Little is known about the acquisition of segmental or of tonal sandhi rules that apply across word boundaries, or of rules that apply in rapid speech, in spite of their evident productivity. The ability to understand and produce many aspects of English stress and intonation patterns develops gradually during the elementary-school years. English-speaking school-aged children may produce over-long vowels in lieu of the reduced vowel of unstressed syllables; such hypercorrection suggests awareness of reduction.
Fernald, Anne. 1992. Human maternal vocalizations to infants as biologically relevant signals. In The adapted mind, edited by Jerome H. Barkow et al. pp. 391–428. Oxford: Oxford University Press.Find this resource:
Jakobson, Roman. 1968. Child language, aphasia, and phonological universals. The Hague: Mouton.Find this resource:
Jusczyk, Peter W. 1997. The discovery of spoken language. Cambridge, Mass.: MIT Press.Find this resource:
Kager, René, Joe Pater, and Wim Zonneveld, eds. To appear. Fixing priorities: Constraints in phonological acquisition. Cambridge: Cambridge University Press.Find this resource:
MacWhinney, Brian. 1978. The acquisition of morphophonology. (Monographs of the Society for Research on Child Development, 43:1/2). Chicago: University of Chicago Press.Find this resource:
Menn, Lise. 1983. Development of articulatory, phonetic, and phonological capabilities. In Language production, vol. 2, edited by Brian Butterworth, pp. 3–49. London: Academic Press.Find this resource:
Saffran, Jenny R., Elissa L. Newport, and Richard N. Aslin. 1996. Word segmentation: The role of distributional cues. Journal of Memory and Language 35.606–621.Find this resource:
Stamenov, Maxim, and Victor Gallese. 2002. Mirror neurons and the evolution of brain and language. Amsterdam: Benjamins.Find this resource:
Stager, Christine L., and Janet F. Werker. 1997. Infants listen for more phonetic detail in speech perception than in word-learning tasks. Nature 388.381–382.Find this resource:
Vihman, Marilyn M. 1996. Phonological development: The origins of language in the child. Oxford: Basil Blackwell.Find this resource:
Vihman, Marilyn M., and Shelley L. Velleman. 2000. Phonetics and the origins of phonology. In Phonological knowledge: Conceptual and empirical issues, edited by Noel Burton-Roberts et al. pp. 305–339. Oxford: Oxford University Press.Find this resource:
Werker, Janet F., and Christine L. Stager. 2000. Developmental changes in infant speech perception and early word learning: Is there a link? In Papers in laboratory phonology V: Acquisition and the lexicon, edited by Michael B. Broe and Janet B. Pierrehumbert pp. 181–193. Cambridge: Cambridge University Press.Find this resource:
In learning a language, a child cracks a code constructed of arbitrary symbols (words) combined in patterns (syntax) that convey information about each word's role. To accomplish this, children must collect samples of the code, analyze them, and work out how the code's structure maps onto messages. Based on sentences they encounter, children form an idea of the structure of their language that is abstract enough that they can then produce and understand entirely new sentences, and detailed enough that they can retain information about the quirks of particular lexical items.
Productive language begins at about one year of age. Single-word utterances are typical at the start, and early productive vocabularies are dominated by nouns. Word combinations begin to appear at about 1.5 to 2.5 years. At first very short, typically only two or three words long, combinations gradually lengthen. Early sentences are often called “telegraphic” because they lack function morphology (Throw ball rather than I'm throwing the ball). Even these early sentences have structure, however, often exhibiting the correct word order for the language being learned. By age five, children's language is essentially adult-like, with sentences like I bet you don't know how to play this routine accomplishments. These familiar patterns are based partly on the growth of syntactic and lexical knowledge, and partly on the development of an efficient system for language production. Language comprehension typically shows earlier competence: 18-month-olds speak mostly in one-word utterances yet understand the significance of word order in longer sentences.
We generally assume that this feat of learning depends on species-specific capacities, since no other species learn our languages, and that it is a feat of learning, since children acquire the language used around them. The nature of learning mechanisms for syntax acquisition is controversial. Many theories propose that some constraints on acquisition are specific to language, constituting a universal grammar (UG) that permits children to create a highly structured system based on noisy and incomplete data. Other theories assume that syntax acquisition depends only on domain-general learning mechanisms and constraints. Regardless of its stance on this question, any theory of syntax acquisition must specify what elements the child can identify in the input before learning a grammar, and what learning mechanisms the child uses to construct a grammar based on those elements. These mechanisms must be constrained enough to explain the ways in which all human languages are alike, yet flexible enough to permit observed variation across languages.
1. The “bootstrapping problem.”
Syntactic regularities involve dependencies among classes of words. A child learning English learns that determiners precede their nouns; and that transitive verbs are followed by their objects. Learning syntactic dependencies requires identifying the categories over which they are defined—nouns and verbs, determiners and tense markers; but these categories, in turn, are defined by their positions in phrase structure. Nouns and verbs cannot be identified based on sound or meaning; ultimately, these categories are defined by the roles they play in a grammar. This reveals the fundamental circularity of syntax acquisition: the child needs phrase structure to create grammatical categories, yet needs grammatical categories to learn phrase structure. The question of how the child breaks into this circle is the “bootstrapping problem” of language acquisition (from the expression “Pull yourself up by your own bootstraps”).
Research in this area explores what relevant cues children can detect in the linguistic environment, and how they might use these cues to create grammatical categories. Three kinds of cues are distributional patterns, phonological cues, and semantic information.
2. Distributional learning
Because grammatical categories are distributionally defined, children could begin by grouping words based on their occurrence in similar environments. Nouns and verbs occur with different function morphemes (the, a, -s vs. is, can, -ing) and in different sentence positions. Children could use intercorrelations among these contexts to create grammatical categories (Maratsos 1982). Formal analysis along these lines is assumed in all theories of syntax acquisition, and it has the last word in syntactic decisions even in theories that give semantics a starring role in early identification of syntactic categories. One example that makes clear the need for distributional analysis is that of gender categories of nouns. The distinction between masculine and feminine nouns, in languages that have it, is notorious for its lack of semantic basis, yet children learn it as they do other grammatical categories.
This procedure has its problems, of course. The contingencies so useful in differentiating grammatical categories are relationships between open-class categories like nouns and verbs, and the closed-class or function morphemes that co-occur with them. Without already knowing that contexts like the and a and affixed -s are important, children would have to look for these helpful correlations among an extremely large set of unhelpful ones. The computational intractability of this problem has prompted researchers to seek sources of constraint on the detection of distributional patterns. These include phonological and semantic cues to grammatical structure, as well as the architecture of learning mechanisms for distributional analysis.
Function morphemes, central to identifying grammatical categories, appear late in children's speech, but they influence children's comprehension much earlier. Even one-word speakers understand familiar nouns better when they are preceded by the determiner the rather than a misplaced function word (Gerken and McIntosh 1993). Computational analyses of speech to children suggest that distributional similarity can be used to sort words into major grammatical categories (Cartwright and Brent 1997, Mintz et al. 1995).
There have been some notable attempts to study the learning mechanisms involved in distributional analysis in toddlers or infants, and to explore the constraints on these mechanisms. Slobin 1985 proposes perceptual operating principles that bias children's hypotheses, including a bias to attend to the ends of words or utterances. Some studies focus on infants' ability to detect structure in small artificial languages (Gomez and Gerken 1999, Saffran et al. 1996). These artificial language-learning experiments are not presumed to approach the complexity of distributional analysis required to learn natural-language syntax; however, their findings establish that infants readily pick up distributional patterns in materials that have not yet been assigned any meaning.
3. Phonological cues
Sentences have prosodic or intonational structure as well as syntactic structure; prosody defines domains within which phonological processes operate. Major prosodic boundaries in speech tend to align with syntactic boundaries, so that syllables that end major constituents are often lengthened, undergo more pitch change, and are followed by longer pauses relative to syllables within phrases. The view known as “prosodic bootstrapping” suggests that acoustic cues associated with large-scale prosodic boundaries in speech constrain the child's syntactic hypotheses (Morgan and Demuth 1995). Considerable evidence suggests that prosody structures speech perception and memory for infants as well as adults.
Sound also participates in grammatical categorization via probabilistic phonological similarity within grammatical categories. One of the best examples is the enormous difference in duration, pitch accent, inventory of consonants, and token frequency between open-class words and closed-class or function words. A first-pass distinction between open- and closed-class words could be made based on this perceptible difference and could constrain distributional analysis. Grammatical categories like noun and verb also tend to share within-class phonological similarity. For example, English nouns and verbs differ in their distributions of syllable numbers (nouns tend to be longer) and in their stress patterns; gender classes of nouns share stable and useful phonological similarity.
4. Semantic cues
Another fundamental source of information for syntax acquisition is meaning. The value of semantics in grammatical categorization and phrase structure acquisition follows from the intimate relationship between form and meaning in language. Children have strong expectations about mappings between meaning and form. First, they assume that different forms have different meanings (Clark 1987); this applies both to individual words and to phrase-structure patterns. This simple assumption permits children to use within-class similarity in meaning to help sort words into grammatical categories and to draw semantic conclusions from distributional evidence for category membership. Second, there are more substantive correspondences between syntax and semantics. Across languages, nouns refer to objects or more abstract entities, while verbs and other terms have relational or predicative semantics. Such abstract links between form and meaning, to the extent that they are universal, could be part of the child's endowment for language acquisition: given knowledge of meaning, the child could draw syntactic inferences; given syntactic information, she could draw semantic conclusions.
One account based on this hypothesis is “semantic bootstrapping” (Pinker 1987), which proposes a detailed set of innate semantic triggers that link words with the elements of UG. On this view, object meanings trigger an internal category noun, linking particular object names with the role that nouns play in grammars. Certain predicate meanings trigger an internal category verb, spatial meanings trigger the category ‘preposition’ (or ‘postposition’), and so on, for every grammatical category and phrase-structure configuration. On this theory, meaning constitutes a privileged source of information. Other views on the syntax-semantics interface suppose less detailed built-in links between meaning and syntax, to permit variation across languages. Not all languages have the same set of grammatical categories; thus, a set of semantic triggers that specifies all the grammatical categories needed for English is probably too specific.
5. The central role of verb learning
Learning large-scale syntactic categories like ‘noun’ and ‘verb’ is only part of the problem. The syntax of a clause depends on smaller subcategories within these abstract categories. In particular, the verb in a clause determines what other phrases can occur in the sentence, and what their roles in the sentence will be. To some degree, this is predictable from verb meaning. Across languages, verbs that describe action on an object tend to be transitive, with actor subjects; those that describe object motion along a path are often intransitive, with the moving theme as subject. Cross-linguistically, robust patterns in how verbs' arguments are linked with syntactic functions are reflected in various proposals for systematic linking of thematic roles (like agent and theme) and syntactic functions (like subject and object). Such links have played a major role in theories of syntax acquisition (Bloom 1991, Gleitman and Gleitman 1997, Pinker 1987).
The traditional approach to the role of verb learning in syntax acquisition has been essentially the semantic bootstrapping approach: if we assume that children can sometimes retrieve the semantics of sentences from extralinguistic context before learning the grammar, they could then use built-in links between semantics and syntax to impose syntactic structure on linguistic input. Knowing that a verb has an agent role in its semantic argument structure, for example, the child will conclude that the noun naming the agent is the subject of the sentence. By assuming that meaning is independently accessible and that linguistic forms are semantically predictable, many acquisition theories make meaning the primary “bootstrap” into syntax.
The syntactic bootstrapping theory also relies on tight links between verb syntax and semantics to drive language acquisition, but it questions the primacy of semantic information. Lexical-semantic organization varies across languages; therefore, semantic representations are language-particular and cannot be assumed to be directly predictable from observations of events. Some words, in practice, are easy to grasp from context—when we hear the word dog, dogs are likely to be relevant. Verb and sentence meanings are more problematic. Even adults are poor at guessing what verb a parent is saying to a child from observation of the context alone, though they are much better at guessing the nouns, given the same kind of evidence (Gleitman and Gleitman 1997). The child's growing knowledge of syntax is proposed to fill this informational gap. Children assign different meanings to verbs presented in different syntactic contexts (Naigles 1990, Fisher 1996), even though they are used to describe the same scene. Simple aspects of sentence structure (the number of familiar nouns in the sentence, and their order) may provide some very early constraint on verb and sentence interpretation.
6. Integrating the cues
Which of the three classes of cues (distributional, phonological, semantic) is the primary source of data from which children create syntactic categories and structures? None is without difficulties. Distributional patterns are the sine qua non of syntax, and they are generally given the status of final arbiter in theories of syntax learning, but strings of morphemes are full of ambiguity. The English present-tense -s occurs with verbs (she walks), but the plural -s occurs with nouns (two shoes). Distributional analysis must be a protracted process of sorting out context-sensitive distributional patterns, heavily influenced by progress in morpheme segmentation and constrained by additional sources of information. Sound cues to grammatical phrase boundaries and categories are weak cues: prosodic boundaries are related to syntactic boundaries in complex ways, and the tendency for words within a grammatical category to have similar sound patterns is also limited. Meaning seems like a reliable cue, but gender categories of nouns must be acquired despite their lack of semantic predictability, and languages vary in how they map meanings onto grammatical categories. The picture that emerges from this is of a process of constrained distributional analysis over a multi-dimensional data set in which no single source of information can be relied on absolutely, and none can be ignored.
Commentators disagree on just how protracted this analysis process is. Many theories hold that abstraction is an early and robust property of children's linguistic analyses (Gomez and Gerken 1999, Pinker 1987), while others point out that children's skill at learning distributional facts about particular words makes it difficult to tell what level of knowledge underlies early word combinations (Pine et al. 1998, Tomasello 2000).
The process of learning which verbs occur with which sentence structures, like the acquisition of broader grammatical categories, shows the influence of multiple cues. Distributional analysis is important: observations of particular verbs in sentences provide information about their syntactic structures. Children also use semantic cues to predict that verbs whose meanings they know can occur in unattested sentence structures. These innovations are sometimes ungrammatical because they violate subcategorization facts for the verb, but they are semantically reasonable nonetheless. Even phonological information plays a role. For example, verbs that accept the dative alternation (Mary gave a book to John; Mary gave John a book) share semantic similarity, but they also tend to share the phonological property of having at most one metrical foot. This includes monosyllabic verbs like give and stress-initial bisyllabic verbs like offer, but not stress-final bisyllabic verbs like explain. These multiple sources of information, some obviously probabilistic and none clearly having the upper hand, give language acquisition its opportunistic, “bootstrapping” feel.
7. The contribution of the learner
The most striking evidence for the contribution of the learner to syntax acquisition comes from situations in which children receive impoverished linguistic input yet invent a syntax more systematic than the input should support. Children do not simply duplicate the input they receive. Instead, they regularize, imposing new structure on noisy data, or they invent from scratch communicative systems that exhibit properties of conventional languages.
Children whose profound hearing loss makes learning a spoken language impossible, and who are exposed to no sign-language model, invent “Home Sign” systems (Goldin-Meadow and Mylander 1998). Home signers create gestures that they combine into sentences. These sentences show signs of structure as found in established languages: signs glossed as verbs occur with predictable sets of noun-like arguments (eat has two arguments, sleep only one), and arguments appear in consistent positions. The parents' gestures do not exhibit the same structure. Apparently, children need not learn that there exists a fundamental distinction between nouns (argument terms) and verbs (predicate terms)—for example, that a verb meaning ‘eat’ presupposes an eater and a thing eaten—or that the logical arguments of a verb can be specified by nouns in sentences. Such basic aspects of linguistic structure may follow from the structure of human knowledge and the pressures of human communication (Goldin-Meadow et al. 1996, Haiman 1985).
Children are also better syntax learners than adults. Adult learners of a first or second language honor morphological and syntactic regularities only probabilistically, but children who learn their native language from parents with a non-native command of it repair or “creolize” it, rather than accurately reproducing all the idiomatic irregularities of the input. One study examined the progress of a deaf child, Simon, whose only exposure to American Sign Language (ASL) was the inconsistent usage of his parents, late learners of ASL (Singleton and Newport, to appear). Simon regularized his parents' language, achieving a system that was more native-like than the input. Children and adults do not learn language in the same way: in failing to learn all the quirks in the input (though they certainly learn some), children create the analytic, regular syntactic, and morphological systems of human languages. The resilience of linguistic structure, despite variation in the input, tells us that fundamental aspects of syntax follow directly from the biases and constraints children bring to the task of learning and of communication.
Bloom, Lois. 1991. Language development from two to three. New York: Cambridge University Press.Find this resource:
Cartwright, Timothy A., and Michael R. Brent. 1997. Syntactic categorization in early language acquisition: Formalizing the role of distributional analysis. Cognition 63. 121–170.Find this resource:
Clark, Eve V. 1987. The principle of contrast: A constraint on language acquisition. In Mechanisms of language acquisition, edited by B. MacWhinney, pp. 1–33. Hillsdale, N.J.: Erlbaum.Find this resource:
Fisher, Cynthia. 1996. Structural limits on verb mapping: The role of analogy in children's interpretations of sentences. Cognitive Psychology 31.41–81.Find this resource:
Gerken, LouAnn, and Bonnie J. McIntosh. 1993. Interplay of function morphemes and prosody in early language. Developmental Psychology 29.448–457.Find this resource:
Gleitman, Lila R., and Henry Gleitman. 1997. What is language made out of? Lingua 100.29–55.Find this resource:
Goldin-Meadow, Susan, and Carolyn Mylander. 1998. Spontaneous sign systems created by deaf children in two cultures. Nature 291.279–281.Find this resource:
Goldin-Meadow, Susan, David McNeill, and Jenny Singleton. 1996. Silence is liberating: Removing the handcuffs on grammatical expression in the manual modality. Psychological Review 103.34–55.Find this resource:
Gomez, Rebecca, and LouAnn Gerken. 1999. Artificial grammar learning by 1-year-olds leads to specific and abstract knowledge. Cognition 70.109–135.Find this resource:
Haiman, John. 1985. Iconicity in syntax. Amsterdam: Benjamins.Find this resource:
Maratsos, Michael. 1982. The child's construction of grammatical categories. In Language acquisition: The state of the art, edited by E. Wanner and L. R. Gleitman, pp. 240–266. New York: Cambridge University Press.Find this resource:
Mintz, Toben H., Elissa L. Newport, and Thomas G. Bever. 1995. Distributional regularities of form class in speech to young children. In Proceedings of NELS 25. Amherst, Mass.: GLSA.Find this resource:
Morgan, James L., and Catherine Demuth. 1995. Signal to syntax. Hillsdale, N.J.: Erlbaum.Find this resource:
Naigles, Letitia. 1990. Children use syntax to learn verb meanings. Journal of Child Language 17.357–374.Find this resource:
Pine, Julian, Elena Lieven, and Caroline Rowland. 1998. Comparing different models of the development of the English verb category. Linguistics 36.807–830.Find this resource:
Pinker, Steven. 1987. The bootstrapping problem in language acquisition. In Mechanisms of language acquisition, edited by B. MacWhinney, pp. 399–441. Hillsdale, N.J.: Erlbaum.Find this resource:
Saffran, Jenny R., Richard N. Aslin, and Elissa L. Newport. 1996. Statistical learning by 8-month-old infants. Science 274.1926–1928.Find this resource:
Singleton, Jenny L., and Elissa L. Newport. To appear. When learners surpass their models: The acquisition of American Sign Language from inconsistent input. Cognitive Psychology.Find this resource:
Slobin, Daniel I. 1985. Crosslinguistic evidence for the language-making capacity. In The crosslinguistic study of language acquisition, edited by D. I. Slobin, vol. 1, pp. 1157–1256. Hillsdale, N.J.: Erlbaum.Find this resource:
Tomasello, Michael. 2000. Do young children have adult syntactic competence? Cognition 74.209–253.Find this resource:
The term “communicative competence” was coined by Dell Hymes in 1966 in a proposal to broaden the scope of knowledge and skills embodied in Noam Chomsky's definition of “linguistic competence” (Chomsky 1965). Hymes argued that speakers who were able to produce all the grammatical sentences of a language would be institutionalized if they went about trying to do so without consideration of appropriate contexts of use, and of the socially and culturally determined norms for production and interpretation. He augmented Chomsky's criterion of systematic potential (whether or not an utterance is a possible grammatical structure in a language) with knowledge of appropriateness (whether and to what extent a potential communicative form is suitable), occurrence (whether it is really enacted), and feasibility (whether it is possible under particular circumstances) (1974, 1987). Hymes's proposal was quickly adopted both by sociolinguists and by applied linguists in the field of foreign/second-language instruction: the requisite knowledge of speakers which must be accounted for includes not only rules for communication (both linguistic and sociolinguistic) and shared rules for interaction, but also the cultural roles and knowledge that are the basis for the context and content of communicative events and interaction processes. Communicative competence extends to both knowledge and expectation of who may or may not speak in certain settings, when to speak, how one may talk to persons of different statuses and rules, what nonverbal behaviors are appropriate in various contexts, what the routines for turn-taking are in conversation, how to ask for and give information, how to request, how to offer or decline assistance or cooperation, how to exercise power, and the like. For reference, see Gumperz 1984, Saville-Troike 1989.
Both the communicative knowledge and skills which may be attributed to a speech community and the communicative competence of individual speakers are highly variable constructs. First, as a collectivity, a speech community includes a range of language varieties and registers (and frequently even different languages), and it is very unlikely in a complex community that any single individual would be able to produce the full range of its available linguistic repertoire. Second, any one speaker has a variety of codes, styles, and registers from which to choose, and is not infrequently a member of more than one speech community—often to different degrees. For individuals who are members of multiple speech communities, which ones they orient themselves to at any given moment—which set of social and communicative rules they use—is part of the strategy of communication. Accounting for the nature of communicative competence thus “requires going beyond a concern with Language (capital L) or a language. It requires a focus on the ways in which communities, and persons, make use of a repertoire of languages (a focus pioneered by Charles Ferguson). It requires a focus on the ways in which people do use language […]” (Hymes 1993:13). Multilingual speakers' communicative competence includes knowledge of rules for the appropriate choice of language and for switching between languages, given a particular social context and communicative intent. Central to the concept is the notion that language is a social and cultural practice (Kramsch 1997). An extension has been made to “intercultural communicative competence,” which requires an additional level of metacompetence involving explicit awareness of differential usages and ability to adapt communicative strategies to a variety of cultural situations (Kim 1991).
Among the most productive areas for application of this perspective has been the study of child language development, where it includes “discovery of how cultures themselves shape acquisition” (Hymes 1987:224; cf. Heath 1983, Ochs and Schieffelin 1995). In the field of foreign/second-language teaching, the concept has been applied to the development of “communicative approaches” to language teaching and testing (e.g. Savignon 1983); these have been widely accepted in most parts of the world.
Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, Mass.: MIT Press.Find this resource:
Gumperz, John J. 1984. Communicative competence revisited. In Meaning, form, and use in context: Linguistic applications (Georgetown University Round Table on Languages and Linguistics, 1984), edited by Deborah Schiffrin, pp. 278–289. Washington, D.C.: Georgetown University Press.Find this resource:
Heath, Shirley Brice. 1983. Ways with words: Language, life, and work in communities and classrooms. Cambridge and New York: Cambridge University Press.Find this resource:
Hymes, Dell. 1974. Foundations in sociolinguistics: An ethnographic approach. Philadelphia: University of Pennsylvania Press.Find this resource:
Hymes, Dell. 1987. Communicative competence. In Sociolinguistics: An international handbook of the science of language and society, edited by Ulrich Ammon et al., pp. 219–229. Berlin: Mouton de Gruyter.Find this resource:
Hymes, Dell. 1993. Anthropological linguistics: A retrospective. Anthropological Linguistics 35.9–14.Find this resource:
Kim, Young Yun. 1991. Intercultural communicative competence: A systems-theoretic view. In Cross-cultural interpersonal communication, edited by Stella Ting-Toomey and Felipe Korzenny, pp. 259–275. Newbury Park, Calif.: Sage.Find this resource:
Kramsch, Claire. 1997. The privilege of the nonnative speaker. PMLA 112(3).359–369.Find this resource:
Ochs, Elinor, and Bambi B. Schieffelin. 1995. The impact of language socialization on grammatical development. In The Handbook of Child Language, edited by Paul Fletcher and Brian MacWhinney, pp. 73–94. Oxford: Blackwell.Find this resource:
Savignon, Sandra. 1983. Communicative competence: Theory and classroom practice. Reading, Mass.: Addison-Wesley.Find this resource:
Saville-Troike, Muriel. 1989. The ethnography of communication. 2d ed. (Language in society, 3.) Oxford: Blackwell.Find this resource:
Meanings and Forms
Language in children first emerges at around one year of age, when they begin to understand and produce their first words. At this point, they often express a range of meanings for which they have no words. They rely on gestures (pointing, reaching, miming actions) and combine these with vocalizations to indicate what they want or are interested in. As they come to segment the speech stream and identify word- and phrase-sized units, they begin to make use of some linguistic forms themselves, but their meanings for words and phrases, like their early pronunciations, often diverge from those of adults. Working out which meanings go with which forms is a lengthy task; the acquisition of meaning is intimately linked to the acquisition of morphological and syntactic forms as well. The major problem children face is how to assign to each form the meaning it conventionally carries in the speech community. They have to make inferences in context, relying on joint attention, physical co-presence, and conversational co-presence, to assign a preliminary meaning. They then test their hypotheses about meanings and revise them when necessary. Research has focused on the cognitive and social sources of children's hypotheses about meanings, and on the stages they go through as they learn more about the adult meanings.
The meanings children adopt are offered by adults in child-directed speech. Parents and caregivers offer words (This is an owl, Look at the otter) and relations between words (An otter is a kind of animal, Ducks are birds). They also offer children different perspectives on the same objects or events (the spaniel vs. the dog vs. your pet; Give Jay the ball vs. Let Jay have the ball vs. Now Jay wants the ball). Children take up these adult offers that license specific inferences in context about possible meanings. With joint attention, physical co-presence, and conversational co-presence as general guides, children can assign initial meanings that they can refine with subsequent exposures to adult usage. Along the way, however, they may make errors that reveal which inferences they have made so far, and how close their meanings are to the adult ones. Children also make errors in the forms of their first language. They over-regularize inflectional and derivational endings; they over-extend syntactic patterns; and they impose regularity in the shape of paradigms, where the language around them is irregular. Here research has focused on two problems: first, how children master conventional but irregular forms of the language, and second, how children get rid of regularized forms which they themselves have constructed. The first problem has received more attention than the second, and the kinds of over-regularization that children come up with are well-documented for a variety of languages. Less is known about how and why children get rid of early regularizations and other errors.
In acquisition, meaning and form must be taken together; one cannot be acquired without the other. But researchers have sometimes focused on a single aspect, without considering the necessary coordination between the two. Data from a variety of languages (Slobin 1995) suggest that all children start their acquisition of meaning and form in much the same way. Whether they continue on the same route, as they learn more complex meanings and forms, depends on how similar languages are—and on how many possible routes children may find, as they move toward an adult-like mastery of a language.
1. Lexical meanings
Children's earliest lexical meanings emerge with their first words, and often diverge from adult meanings for the same words (Clark, 2002). Children may over-extend a word beyond its adult meaning, e.g. using dog for horses and sheep as well as dogs; or they may under-extend a word, e.g. using shoe only for laced shoes being worn on the feet, but not for shoes of other types or in other places. They may use a word so that its meaning overlaps with, but does not coincide with, the adult's meaning—as when horse is used for riding horses and is extended to donkeys and zebras, but is not used for cart-horses. Finally, children's earliest meanings may be complete mismatches and fail to overlap at all with adult meanings.
However, production does not necessarily match comprehension. Children who produce dog for four-legged mammals other than dogs typically understand that dog, when heard from others, denotes only dogs and not other kinds of animals. Over-extensions in production, therefore, probably reflect a communicative strategy for talking about things, prior to mastery of the relevant labels. As children learn to produce their first hundred words, they may over-extend up to 40% of them. Under-extensions and overlaps are harder to document; however, they may be even more pervasive in children's meanings, during a longer period, than over-extensions. The latter become rare by the age of 2;0 (i.e. 2 years, 0 months) to 2;6. Most observations about early meanings have come from diary studies of language production; however, researchers have also examined some sources of children's hypotheses about word-meanings by looking systematically at how children understand words, as well as when they produce them.
The stages in children's acquisition of word-meanings, in specific semantic domains, have been studied in some detail. Children's earliest hypotheses about word-meanings are often based on general conceptual knowledge about relevant domains. Thus, in the acquisition of locative terms, children appear to rely on two general (non-linguistic) strategies for coping with instructions using words like in, on, and under. The first, “Put X inside,” applies to any locations that are containers; the second, “Put X on top,” to places that are not containers. Children's attention to containers and supporting surfaces, in other words, provides their first hypotheses about the locative relations that words encode. Later, these interpretations become more specific, as they work out the contrasts in meaning offered by the words they hear in context: in and on, or on and above or under.
The coping strategies children rely on in the earliest stages of assigning meanings to unfamiliar words have been traced for a variety of locative terms (e.g. in, on, up, down, under), kinship terms (mother, uncle, cousin, grandfather, daughter), verbs of possession (give, take, trade, buy, sell), verbs of speaking (ask, tell), verbs of motion, and deictic terms (I, here, that, go, bring). As children add to their initial meanings, the errors they make change until their usage parallels that of adults. There has also been some research on how children build up taxonomies of terms in a hierarchy, for such domains as animals (e.g. animal, dog, collie); here we find little consistency in whether a superordinate like animal appears before or after a subordinate like collie, although terms at the generic or basic level (e.g. dog) typically enter first. Otherwise, the level of the earliest terms acquired appears to depend on their immediate usefulness to the young child. For instance, the lower-level apple and orange will be more useful than fruit; but higher-level bug will be more useful than stag-beetle or aphid, and tree more useful than oak or elm.
2. Word combinations
In their earliest word combinations, children talk about roles (e.g. agent, location, recipient) and actions in events. However, these early constructions appear to be lexically specific: they produce definite articles only with certain nouns; they combine certain verbs and nouns, but use only a small set of such combinations for several weeks or months. They take time to realize that one verb can occur with a large range of different nouns to denote objects-affected, that an article can precede almost every noun, or that an auxiliary verb can occur with many main verbs. They appear to build up constructions very slowly, word by word, as they hear more instances in child-directed speech.
Children just beginning to combine words show little evidence of marking such grammatical relations as “subject of” or “direct object of.” The two-word combinations observable in children acquiring different languages appear similar in meaning; however, not all children try the same kinds of combinations prior to acquisition of morphological marking or of word order to signal grammatical relations (Brown 1973). A number of studies suggest that the earliest uses of word order in production are actually to signal new information, as opposed to information already given in the conversation or the non-linguistic context. Only later, in languages where word order has a grammatical function, do children learn that too. Meanings attributable to structures emerge in production somewhat later than those attached to individual words in noun phrases, verb phrases, or prepositional phrases.
As their word combinations become longer, children fill out more of their phrases—adding articles, quantifiers, adjectives, prepositions, and other indicators of grammatical functions. They also begin to produce relative clauses, adverbial phrases and clauses, and complements; all these may require use of both word order and grammatical morphemes to mark structural relations among linguistic units (Clark, 2002, Maratsos 1983, Wanner and Gleitman 1982).
3. Grammatical morphemes
The affixes or free morphemes that mark grammatical relations among words begin to emerge during the first year of speech. In English, the first grammatical morphemes produced by children include aspectual -ing, plural -s, and past tense -ed. Later acquisitions are the articles a and the, the copula verb be, prepositions, and complementizers (e.g. to, for, that). In general, regular systems with few exceptions are mastered faster than systems with numerous small paradigms (cf. Turkish and Russian), or several genders marked by different affixes (cf. German and English). It also takes time to master the affixes for agreement in gender and number for noun and article, noun and adjective, or noun and verb combinations.
When children acquire inflections—whether for case, number, gender, person, tense, or aspect—they regularize the system. They typically choose the paradigm with the largest number of members (types, not tokens) and use it as a model for other nouns or verbs. In English, children regularize the past tense of irregular verbs like see, buy, go to produce seed, buyed, goed. In French, they regularize irregular verbs in -ir or -re to the -er paradigm, as in the past participles couré (for couru ‘ran’ < courir ‘to run’), metté (for mis < mettre ‘to put’), or mordé (for mordu < mordre ‘to bite’). Where children have to choose among different affixes that, for instance, all mark the same case but are used with nouns of different gender, they often choose a single form for that meaning. Thus, in Russian, children at this stage produce -om for the instrumental on all nouns—masculine, feminine, and neuter.
The overall pattern of acquisition for grammatical morphemes falls into four stages. After an initial period of no use, one finds sporadic use of irregular forms, followed by early uses of regular forms; then comes a flood of regularizations that are gradually replaced by appropriate irregular forms. A major theoretical question here is how children come to replace bringed, say, by brought, or foots by feet. The same question applies to children's over-regularizations of syntactic patterns, as in such forms as Don't say me that (for Don't say that to me), or Can I fill some salt into the bear? (for fill the bear with salt) (MacWhinney 1987).
4. Word formation
The same question arises in the domain of word formation. Here too, children regularize irregular forms to fit their paradigms. They construct nouns for agents (e.g. cooker for a cook), for instruments (driller for a drill) and for states (longness for length). They often construct compound nouns; but in the early stage, they fail to do so appropriately (open-man, open-door, or opener-door, all for door-opener). And they also coin verbs—from nouns (e.g to broom for to sweep), from adjectives (to dark for to darken), and from prepositions (to up for to raise). In each case, some formations are pre-empted by existing words, and children have to learn that those meanings are conventionally expressed by other established forms.
5. Social meaning
Meaning in language derives from the senses associated with specific words and the constructions in which those words appear. Further meaning is added by the uses to which speakers put words and constructions—the conventions associated with particular forms within a social group. In most languages, speakers mark differences in status, power, and sex in their choices of address terms, words, and constructions. English-speaking children by age four or five know some of these conventions and differentiate among addressees by using imperatives (Give me that ball) or direct demands (I want that ball) for lower status, but hints (It is fun to play with balls, isn't it?) for higher status (Andersen 1990). Children may take several years to acquire the words and constructions involved and to know when to use which form, and to whom.
6. Meaning and form
Children make inferences about possible meanings and then build on these, adding or deleting details as they test their initial hypotheses by using words with the meanings they have assigned. When their uses match adult ones, their hypotheses are supported; when they do not, children must refine or change their hypotheses. One way to track these developments in meaning acquisition is to examine consistent errors, which provide clues to the meanings children have assigned to specific forms; changes in their errors mark the path being followed in development. Consistency in these errors also helps to identify general strategies of acquisition, whether for analyzing units of meaning and form in comprehension or for combining such elements in production. Similarities in children's initial coping strategies and in their uptake of pragmatic inferences about meanings (licensed by adult usage) help to identify the general learning mechanisms children bring to language. Tracking children's errors, along with what they tend to get right, is critical in finding what is universal across languages and in modeling the process of acquisition (Elman et al. 1996, MacWhinney 1999).
Research on the general process of acquisition—on what is general, and what is shaped by the structure of specific languages—contributes essential information to the puzzle of what to count as universal across languages. The study of language acquisition offers insight into the learning of complex systems and raises new questions about the kinds of models that will account adequately for both what and how children learn as they acquire a first language.
Andersen, Elaine S. 1990. Speaking with style. London: Routledge.Find this resource:
Brown, Roger. 1973. A first language: The early stages. Cambridge, Mass.: Harvard University Press.Find this resource:
Clark, Eve V. 2002. First language acquisition. Cambridge: Cambridge University Press.Find this resource:
Elman, Jeffrey L., et al. 1996. Rethinking innateness. Cambridge, Mass.: MIT Press.Find this resource:
MacWhinney, Brian, ed. 1987. Mechanisms of language acquisition. Hillsdale, N.J.: Erlbaum.Find this resource:
MacWhinney, Brian, ed. 1999. The emergence of language. Mahwah, N.J.: Erlbaum.Find this resource:
Maratsos, Michael. 1983. Some current issues in the study of the acquisition of grammar. In Handbook of child psychology, vol. 3, Cognitive development, edited by John H. Flavell and Ellen M. Markman, pp. 707–786. New York: Wiley.Find this resource:
Slobin, Dan I., ed. 1995. The cross-linguistic study of language acquisition. 2 vols. 3rd ed. Hillsdale, N.J.: Erlbaum.Find this resource:
Wanner, Eric, and Lila R. Gleitman, eds. 1982. Language acquisition: The state of the art. Cambridge and New York: Cambridge University Press.Find this resource:
In many domains and species, evolutionarily significant behaviors—for example, the acquisition of the speciestypical mating song in finches and sparrows—develop partly through innate predispositions and are partly shaped by experience. In many of these systems, the effects of experience are limited to a critical or sensitive period.
A critical or sensitive period for learning is present when there is a relationship between the age (more technically, the developmental state of the organism) at which the crucial experience is presented to the organism and the amount of learning that results. In most such cases, the privileged time for learning is during early development, but this is not necessary (bonding in sheep occurs immediately surrounding parturition). The important feature is that there is a peak period of plasticity, occurring at some maturationally defined time in development, followed by reduced plasticity later in life. This contrasts with the presence in other systems of plasticity uniformly throughout life (open-ended learning), or an increase with age as experience or higher-level cognitive skills increase. The term “critical period” is sometimes used when there is an abrupt decline in plasticity and no residual plasticity after this period is over, whereas the term “sensitive period” is used when there is a more gradual decline, with some reduced plasticity remaining throughout life. However, most critical periods show more gradual offsets and more complex interactions between maturation and experiential factors than was implied by the original concept of a critical period, so the two terms are often used interchangeably.
1. Evidence of a critical or sensitive period for language
Lenneberg 1967 suggested that there is a critical period for human language acquisition, extending from birth to puberty. Within this period, language acquisition occurs as a result of ordinary exposure to a linguistic community; after this period, Lenneberg suggested, it occurs only with difficulty, and perhaps through different mechanisms than those used early in life. His evidence included the better speech outcome for deaf children with early hearing, and the improved recovery from aphasia when brain damage occurs during childhood. He also proposed a mechanism for this age effect that involves the maturation of lateralization. Since then, some of his evidence and his hypothesized mechanism have been questioned, but an extensive behavioral and neurobiological literature has suggested that there is indeed a change in language learning over age for both first- and second-language acquisition.
Case studies of feral or abused children who were isolated from exposure to their first language until after puberty (Curtiss 1977) show that they have extreme deficits in phonology, morphology, and syntax; however, the physical and cognitive status of such children may be a concern. Studies of populations of normal individuals permit one systematically to examine proficiency in relation to age of language exposure without concern for the physical status of the learning brain. These studies also show a strong relationship between the age of exposure to a language and the ultimate proficiency achieved in that language (Johnson and Newport 1989, Krashen et al. 1982, Long 1990, Newport 1990), though adult learners exhibit much less extreme deficits than do isolated children. Adult language-learners have an advantage in learning during the first months or year of exposure, but long-term outcome is clearly better for those who start learning during childhood. Peak proficiency, in control over the sound system and of grammatical structure, is displayed by those whose exposure to the language began in infancy or very early childhood. Increasing age of exposure results in a decline in average proficiency, beginning as early as ages four to six and continuing until proficiency reaches a plateau for adult learners (Johnson and Newport 1989, Newport 1990). However, individual variation also increases with age (Johnson and Newport 1989), and some late-learning individuals may approach the proficiency of early learners (Birdsong 1992).
These effects have been shown for both first and second languages, for both spoken and signed languages, and for measures of proficiency including degree of accent, production and comprehension of morphology and syntax, grammaticality judgments for morphology and syntax, and syntactic processing speed and accuracy. For example. Johnson and Newport 1989 have shown that Chinese or Korean immigrants who move to the United States and become exposed to English as a second language show strong effects of their age of exposure to the language on their ability to judge its grammatical structure many years later, even when the subjects are matched in the number of years of exposure. These effects are not due merely to interference from the first language: deaf adults who acquired American Sign Language as their primary language show effects of age of exposure on their grammatical skills in ASL as much as fifty years later, even though they may not control any other language with great proficiency (Newport 1990).
Age of exposure does not affect all aspects of language learning equally: acquisition of vocabulary and semantic processing occurs relatively normally in late learners. Critical-period effects thus appear to focus on the formal properties of language. Even among these, however, late learners acquire the basic word order of a language relatively well, while more complex aspects of grammar show stronger effects of late acquisition (Johnson and Newport 1989, Newport 1990). Further research is needed to characterize the structures most affected by age of learning.
Age of exposure also affects how language is represented in the brain, with similarities between the behavioral and neural results. PET, fMRI, and ERP studies all indicate strong left hemisphere activation for the first language in bilinguals; however, when second languages have been learned late (after seven years), the regions and patterns of activation are partially or completely non-overlapping with those for the first language. Neural organization for late-learned languages tends to be less lateralized and, like proficiency, varies considerably across individuals. The few available studies of early bilinguals or highly proficient late bilinguals report congruent results for first and second languages (Kim et al. 1997, Perani et al. 1996, 1998). As in behavior, age of acquisition has more pronounced effects on grammatical processing and its representation in the brain than on semantic processing (Weber-Fox and Neville 1996).
Taken together, these results provide fairly strong evidence for a critical or sensitive period in acquiring the phonological and grammatical patterns of the language and in organizing the neural mechanisms for handling these structures efficiently.
2. Age effects arise from a critical period?
Do these age effects represent the outcome of a critical or sensitive period, or do they arise from variables correlated with age but not with maturation (Birdsong 1999)? Some investigators have argued that, to support a critical period hypothesis, age effects must coincide with puberty (though neural maturation continues through the late teens and does not cease at age twelve to thirteen). Others have suggested that, if there were a critical or sensitive period, no adult learners should achieve native proficiency (though sensitive periods in other species do show individual variation in adult plasticity). Finally, investigators have noted that it is difficult to distinguish a critical period from an interference effect (similar points have been made in discussions of interference in imprinting).
However, these strong or absolute characteristics are not true of critical or sensitive periods in other behavioral domains, most of which involve gradual declines in learning, with mature organisms retaining some (reduced, but not absent) ability to learn. If such complex phenomena are found within critical periods in other domains, they should also be expected for language learning.
Birdsong, David, ed. 1999. Second language acquisition and the critical period hypothesis. Mahwah, N.J.: Lawrence Erlbaum.Find this resource:
Curtiss, Susan. 1977. Genie: A psycholinguistic study of a modern-day “wild child.” New York: Academic Press.Find this resource:
Johnson, Jacqueline S., and Elissa L. Newport. 1989. Critical period effects in second language learning: The influence of maturational state on the acquisition of English as a second language. Cognitive Psychology 21.60–99.Find this resource:
Kim, K. H. S., N. R. Relkin, K.-M. Lee, and J. Hirsch. 1997. Distinct cortical areas associated with native and second languages. Nature 388.171–174.Find this resource:
Krashen, S. D., Michael H. Long, and R. C. Scarcella. 1982. Age, rate, and eventual attainment in second language acquisition. In Child-adult differences in second language acquisition, edited by S. D. Krashen et al., pp. 161–172. Rowley, Mass.: Newbury House.Find this resource:
Lenneberg, Eric H. 1967. Biological foundations of language. New York: Wiley.Find this resource:
Long, Michael H. 1990. Maturational constraints on language development. Studies in Second Language Acquisition 12.251–285.Find this resource:
Newport, Elissa L. 1990. Maturational constraints on language learning. Cognitive Science 14.11–28.Find this resource:
Perani, D., S. Dehaene, F. Grassi, L. Cohen, S. F. Cappa, E. Dupoux, F. Fazio, and Jacques Mehler. 1996. Brain processing of native and foreign languages. Neuroreport 7.2439–2444.Find this resource:
Perani, D., E. Paulesu, N. S. Galles, E. Dupoux, S. Dehaene, V. Bettinardi, S. F. Cappa, F. Fazio, and Jacques Mehler. 1998. The bilingual brain: Proficiency and age of acquisition of the second language. Brain 121.1841–1852.Find this resource:
Weber-Fox, Christine, and Helen J. Neville. 1996. Maturational constraints on functional specializations for language processing: ERP and behavioral evidence in bilingual speakers. Journal of Cognitive Neuroscience 8.231–256.Find this resource:
The study of S[econd-]L[anguage] A[cquisition] is a broad, interdisciplinary field of inquiry which aims to describe and explain the development and non-development of languages and language varieties beyond the first language. SLA researchers study children and adults learning naturalistically or with the aid of formal instruction, as individuals or in groups, and in foreign, second-language, and lingua franca settings. The research draws upon and contributes to knowledge and procedures in a variety of disciplines, including theoretical linguistics, neurolinguistics, psycholinguistics, sociolinguistics, historical linguistics, pidgin/creole studies, applied linguistics, psychology, sociology, anthropology, and education. SLA research findings are used to test hypotheses and build theories in those areas, as well as for a variety of practical purposes such as the improvement of language teaching, language testing, teacher education, and the design of instructional programs delivered through the medium of a second language or dialect (Larsen-Freeman and Long 1990).
The conditions producing SLA are diverse. It can result, for example, from simultaneous or sequential bilingual or multilingual exposure in infancy—as in the case of children born to parents with different first languages (L1s) who use them in the home and who may also live in a country where a third language is spoken. It is sometimes the product of either forced or voluntary educational experience, including submersion, immersion, bilingual education, foreign-language teaching, and other school, university, or vocational programs in which a second language (L2) is the medium or object of instruction. It can follow informal exposure to languages later in life, as with millions of migrant workers, refugees, and tourists. Finally, it is a routine experience for a substantial part of the world's population who live in multilingual societies.
While the SLA phenomenon is widespread, many aspects of the process itself remain something of a mystery. Only severely subnormal children fail to develop a high degree of proficiency in their native language; but relatively few people, however intelligent and motivated, reach such high standards in a second or third language, especially if they are first exposed to the additional language as adults. Is it that unsuccessful older learners simply do not learn as much as cognitively less developed children? Or is it that they cannot learn as much, or in the same way? Part of their interlanguage often appears to stabilize, or cease to develop, far short of a communicatively adequate and/or socially acceptable level. Sometimes, it is claimed, the stabilization is permanent, in which case it is referred to as fossilization, but the evidence for fossilization is sparse (Long, 2003).
Various explanations have been proposed for the heterogeneous achievement of SL learners, for both success and failure. Some researchers believe that development of both L1 and L2 is maturationally constrained, and that biological (usually neurophysiological) changes progressively diminish the older learner's capacity to learn—i.e., that there are one or more sensitive periods for SLA (Hyltenstam and Abrahamsson, 2003; Scovel 1988). Others hold that psychological and social factors, such as attitude, motivation, and social distance—or, collectively, the degree to which someone acculturates to the target language group—determine success and failure (Schumann 1978, Andersen 1983, Robinson 2001). Still others claim that learning depends on the degree of access to universal grammar (White, 2003), to innate general cognitive mechanisms (O'Grady, 2003), or to input and/or conversational opportunities, particularly access to comprehensible L2 samples from which the grammatical rules of the L2 can be induced (Hatch 1983, Krashen 1985). While all these variables may be relevant in some cases, counter-evidence exists to each as a single-factor explanation, and no one theory enjoys wide acceptance.
While unresolved issues abound, a good deal has been discovered about SLA through a relatively recent but steady empirical tradition (Spolsky 1989). To begin with, learner language is quite variable. Part of the variability is systematic, i.e. rule-governed, such that a learner may use alternate forms according to linguistic environment, situation, task, degree of planning, attention to speech, etc. For example, reflecting variation in linguistic context, learners initially tend to use the English regular past tense marker -ed and plural -s only on certain verbs and nouns, and to omit the forms elsewhere. They do so in a fairly predictable manner. However, part of the variability is non-systematic, or free. Thus learners may temporarily alternate between two verb forms or two negators (e.g. No have and Don't have) in an apparently arbitrary fashion, before gradually assigning different functions to the two items; at that point the variability begins to become systematic (Ellis 1994, Huebner 1983).
As in other kinds of language change, variation in interlanguage (IL) at one point in time often reflects developmental change over time (Preston 1996). This is one reason why the documented variability of ILs is not inconsistent with another of their well established qualities, namely systematicity. Some of the early research demonstrating this was inspired by work on L1 development. Thus, in longitudinal and cross-sectional studies, L1 researchers in the early 1970s discovered a common order of appearance for a set of grammatical morphemes 90 percent accurately supplied, in the speech of children learning English as L1, in obligatory contexts (linguistic environments where omission of the morphemes would result in ungrammaticality). The finding was quickly replicated for E[nglish as a] S[econd] L[anguage]—where, although not invariant, a common accuracy order (slightly different from the L1 order) was established for the elicited and spontaneous speech of children and adults, with or without formal instruction, and most interestingly, from a variety of L1 backgrounds. The following nine items, for example, were repeatedly found to reach 80 or 90 percent accurate suppliance in ESL in approximately this order: -ing, plural, copula, auxiliary, article, irregular past, regular past, 3sg. -s, and possessive -'s. Small but clear effects were observed for L1 differences, such as the later development of accuracy in articles by Japanese learners (with no articles in their L1); however, these were too rare to alter the sequence to any statistically significant degree.
Explanations for the L1 and L2 orders were elusive. Researchers found no consistent relationship to the syntactic and/or semantic complexity of the grammatical items concerned, their markedness, their perceptual saliency or, in the L2 case, their position in an instructional sequence. A combination of factors was probably at work. Accuracy orders did often correlate significantly with input frequency; however, input alone clearly could not account for the data—given that, among other problems, articles were always the most frequent item in English input, but reached criterion (in the sense of accurate suppliance) relatively late in the learner output.
While the morpheme studies were criticized methodologically, and the orders themselves remain in need of explanation, both L1 and L2 findings were usually accepted as demonstrating a role for powerful internal factors in the acquisition process—or, as was sometimes said, of an internal learner syllabus. This was an interpretation consistent with nativist theories in linguistics. Noam Chomsky's claims about acquisition, for example, would predict that an order would exist, although not the specific orders observed. As critics pointed out, the L1 and L2 findings pertained to a theoretically unmotivated miscellany of linguistically unrelated items, which, because language-specific, also revealed little about SLA in general.
While the accuracy orders were perhaps marginal with respect to nativist claims, providing only language-specific evidence, they certainly posed problems for neo-behaviorist models of language learning, and in SLA also for the C[ontrastive] A[nalysis] H[ypothesis]. This had claimed that differences between L1 and L2 led to difficulty in SLA, and governed the course of acquisition. Coupled with neo-behaviorist learning theory, the CAH had motivated the audio-lingual method of language teaching and its many variants, whose practitioners set out to eradicate L1 language “habits” and to inculcate new ones in learners through intensive drill work in areas of contrast between the L1 and the L2.
Strong independent evidence against the CAH, for systematicity in IL, and for a major learner contribution to SLA, was provided by work on so-called developmental sequences in IL, such as those for ESL negation and interrogatives, ESL and Swedish SL relative clauses, and German word order. Developmental sequences are fixed series of overlapping stages, each identified by the relative frequency and/or order of emergence of an interim IL structure, which learners must traverse in the acquisition of a target construction or rule system. Numerous studies have shown, for example, that ESL negation has a four-stage sequence, as shown in Table 1.
Table 1. Emergence of Negation
No + X
No is cheap.
No you give him.
no/not/don't + V
He not living here.
They don't have.
aux + neg
I can't sing.
You mustn't go.
I didn't tell him.
She doesn't play.
If only, say, Spanish speakers (whose L1 has pre-verbal negation) produced pre-verbally negated constructions at stages (1) and (2), then L1 transfer could explain that aspect of the sequence. But in fact, all SL learners, whether naturalistic or instructed, initially produce pre-verbal negation—including Japanese speakers, whose L1 system is post-verbal. Turkish speakers begin Swedish SL negation that way, too, even though both L1 and L2 are post-verbal.
Why developmental sequences look the way they do is still unclear. It is difficult to untangle the effects of several forces which often converge on the same construction (Johnston 1997). In the early stages of many sequences, when the learner's linguistic resources in the L2 are limited, processing constraints are presumably at least partially responsible (Clahsen et al. 1982, Pienemann 1998). The need to reduce redundancy and discontinuity for ease of production and comprehension favors one simple, fixed word order, rather than more complex, variable orders (e.g. with the verb phrase interrupted by the negator). Language universals and typological markedness also seem to be at work (Rutherford 1984), since pre-verbal position is the most preferred (least marked) for negators in natural languages—including early child language, pidgins, creoles, ungrammatical foreigner talk (No drink water!), and highly conventionalized formulas (No can do, Long time no see).
There is clearly a danger of circularity in attempting to explain developmental IL constructions and sequences by appeals to universals and markedness, while simultaneously using the IL data to establish the universals and markedness relationships. Further, cognitive factors like processing constraints may themselves be causal where universals are concerned, in that some linguistic universals may be a function of universal processing constraints. However, the fact that several studies have found six stages in the development of Swedish SL and ESL relative clauses, corresponding to those predictable from Keenan and Comrie's 1977 noun phrase accessibility hierarchy, suggests that ILs tend to develop in ways consistent with typological universals, whatever gives rise to the universals in the first place (Eckman 1996). Some studies even find learners accepting and producing pronominal copies on grammaticality judgment and elicited production tasks (e.g. Number seven is the woman who she is holding the child), when neither their L1 nor the L2 (e.g. Italian and English) permit copies in any kind of relative clause. Results like these not only support the idea that ILs are responsive to language universals, but also that they are to some degree autonomous linguistic systems, not simply relexified versions of the L1 or poor approximations to the L2 (Davies et al. 1984).
Despite the striking commonality of the stages in developmental sequences, as with morpheme accuracy orders, local effects for L1 can be seen (Gass and Selinker 1992, Kellerman and Sharwood Smith 1986). Studies show that, while the order of basic stages in sequences appears to be immutable, L1 differences sometimes result in additional sub-stages, and also in swifter or slower passage through stages. L1 influence seems most likely at points in a sequence when an IL form is similar to an L1 structure. German learners of ESL, for example, sometimes follow their L1 pattern in producing utterances with the negator after the main verb (* David plays not soccer very good), when they correctly begin to place it after the auxiliary. Likewise, speakers of languages like Spanish with preverbal negation in their L1 are slower to relinquish this as an IL strategy than speakers of languages like Japanese which do not have pre-verbal systems—presumably they perceive a similarity between their L1 systems and the preverbal negation of IL stages (1) and (2).
As indicated by these findings, the role of the L1 in SLA is far more complex than originally believed. To begin with, not just differences but also similarities between languages can cause learners difficulty. However, structural identity between L1 and L2 idioms, lexis, and syntax (but probably not phonology) does not necessarily result in transfer, as shown for example by adult Dutch learners' reluctance to transfer correct Dutch uses of the verb break into English (e.g. His voice broke and She broke the world record) if their L1 usages seem to them to be too idiomatic to be likely to occur in the L2. When the structure of the L1 does influence SLA, it generally operates in harmony with what appear to be natural developmental processes—as revealed by the findings on developmental sequences, which show them to be modified but not fundamentally altered by transfer. L1 influences also appear to be constrained by various kinds of linguistic markedness. In general, typologically unmarked L1 forms are more likely to be transferred than marked ones, unless the corresponding L2 form is also marked; however, beginners seem more willing to transfer both marked and unmarked forms. There is some evidence that learning difficulty arises only from L1/L2 differences involving greater L2 markedness, with degree of difficulty reflecting degree of markedness.
Many of the findings described briefly here suggest a more important role for classroom language learners than they have traditionally been accorded (Hyltenstam and Pienemann 1985, Chaudron 1988). Errors are largely beyond the teacher's control; they are inevitable and often a sign of progress, indicating formation of interim IL rules. They also show, if evidence were still needed, that SLA is not simply a process of habit formation. Acquisition sequences do not reflect instructional sequences; learning difficulty is a function of several factors, not just L1/L2 differences, some of them as yet rather poorly understood by researchers and teachers. Finally, learners do not pass from zero to full knowledge of a target construction in one step, although many syllabuses, textbooks, and teachers (and some SLA researchers) implicitly assume that they do—presenting one native-speaker structure at a time and practicing it, followed by another, in building-block fashion.
Formal instruction is still very valuable, however. Other SLA research is beginning to show that, while developmental sequences are unaffected, at least a periodic focus on language as object, or form, does have important benefits (Doughty and Williams 1998, Norris and Ortega 2000). It increases the rate of development. It appears to sensitize learners to communicatively redundant language forms—as evidenced by instructed learners' initial overuse of certain grammatical morphology, compared with naturalistic acquirers' greater tendency to delete those items. Finally, it may also raise the level of ultimate attainment where marked, low-frequency, or perceptually non-salient L2 structures are concerned. These are three of the areas otherwise particularly susceptible to premature stabilization.
Andersen, Roger W., ed. 1983. Pidginization and creolization as language acquisition. Rowley, Mass.: Newbury House.Find this resource:
Chaudron, Craig. 1988. Second language classrooms: Research on teaching and learning. Cambridge and New York: Cambridge University Press.Find this resource:
Clahsen, Harald, Jürgen M. Meisel, and Manfred Pienemann. 1982. Deutsch als Zweitsprache: Der Spracherwerb ausländischer Arbeiter. Tübingen: Narr.Find this resource:
Davies, Alan, et al., eds. 1984. Interlanguage. Edinburgh: Edinburgh University Press.Find this resource:
Doughty, Catherine J., and Jessica Williams, eds. 1998. Focus on form in classroom second language acquisition. Cambridge: Cambridge University Press.Find this resource:
Eckman, Fred. 1996. A functional typological approach to second language acquisition in theory. In Handbook of second language acquisition, edited by William C. Ritchie and Tej K. Bhatia, pp. 195–211. San Diego, Calif.: Academic Press.Find this resource:
Ellis, Rod. 1994. The study of second language acquisition. Oxford and New York: Oxford University Press.Find this resource:
Gass, Susan, and Larry Selinker, eds. 1992. Language transfer in language learning. 2d ed. Amsterdam and Philadelphia: Benjamins.Find this resource:
Hatch, Evelyn Marcussen. 1983. Psycholinguistics: A second language perspective. Rowley, Mass.: Newbury House.Find this resource:
Huebner, Thom. 1983. A longitudinal analysis of the acquisition of English. Ann Arbor: Karoma.Find this resource:
Hyltenstam, Kenneth, and Nicolas Abrahamsson. 2003. Maturational constraints in second language acquisition. In Handbook of second language acquisition, edited by Catherine J. Doughty and Michael H. Long. Oxford: Blackwell.Find this resource:
Hyltenstam, Kenneth, and Manfred Pienemann, eds. 1985. Modelling and assessing second language acquisition. Clevedon, Avon, England: Multilingual Matters. San Diego, Calif.: College-Hill.Find this resource:
Johnston, M. 1997. Development and variation in learner language. Canberra: Australian National University dissertation.Find this resource:
Keenan, Edward L., and Bernard Comrie. 1977. Noun phrase accessibility and universal grammar. Linguistic Inquiry 8.63–99.Find this resource:
Kellerman, Eric, and Michael Sharwood Smith, eds. 1986. Crosslinguistic influence in second language acquisition. Oxford: Pergamon.Find this resource:
Krashen, Stephen D. 1985. The input hypothesis. London: Longman.Find this resource:
Larsen-Freeman, Diane, and Michael H. Long. 1990. An introduction to second language acquisition research. London: Longman.Find this resource:
Long, Michael H. 2003. Stabilization and fossilization in interlanguage development. In Handbook of second language acquisition, edited by Catherine J. Doughty and Michael H. Long. Oxford: Blackwell.Find this resource:
Norris, John, and Lourdes Ortega. 2000. Effectiveness in L2 instruction: A research synthesis and quantitative metaanalysis. Language Learning 50.417–528.Find this resource:
O'Grady, William. 2003. The radical middle: Nativism without Universal Grammar. In Handbook of second language acquisition, edited by Catherine J. Doughty and Michael H. Long. Oxford: Blackwell.Find this resource:
Pienemann, Manfred. 1998. Processability theory. Amsterdam and Philadelphia: Benjamins.Find this resource:
Preston, Dennis. 1996. Variationist linguistics and second language acquisition. In Handbook of second language acquisition, edited by William C. Ritchie and Tej K. Bhatia, pp. 229–265. San Diego, Calif.: Academic Press.Find this resource:
Robinson, Peter, ed. 2001. Cognition and second language instruction. Cambridge: Cambridge University Press.Find this resource:
Rutherford, William E., ed. 1984. Language universals and second language acquisition. (Typological studies in language, 5.) Amsterdam: Benjamins.Find this resource:
Schumann, John H. 1978. The pidginization process: A model for second language acquisition. Rowley, Mass.: Newbury House.Find this resource:
Scovel, Thomas. 1988. A time to speak: A psycholinguistic inquiry into the critical period for human speech. New York: Newbury House.Find this resource:
Spolsky, Bernard. 1989. Conditions for second language learning. Oxford: Oxford University Press.Find this resource:
White, Lydia. 2003. On the nature of interlanguage representation: Universal Grammar in the second language. In Handbook of second language acquisition, edited by Catherine J. Doughty and Michael H. Long. Oxford: Blackwell.Find this resource: