Dr. James 1970
PSYCHOLINGUISTIC IMPLICATIONS FOR THE TEACHING OF FOREIGN LANGUAGES
This chapter attempts to summarize some recently developed notions about the language acquisition process and makes some preliminary suggestions about the implications of these ideas for the problem of teaching a second language. The original impetus in demonstrating the shortcomings of traditional psychological and linguistic theories in the understanding of the processes of language structure and language acquisition must be credited to Chomsky (1957; 1959) who also developed new theories to cope with the problem. Subsequent writers have elaborated upon this new outlook pointing out the various specific inadequacies of the earlier notions and making concrete suggestions for new approaches (see Miller, 1965; Katz, 1966; McNeill, 1966; Lenneberg, 1967; Slobin, 1966; and the contributions in Bellugi and Brown, 1964). To appreciate fully these new developments it is necessary to consider briefly the nature of the inadequacy of the earlier notions on the language acquisition process.
The traditional psychological approach to the language acquisition process was to view it within the framework of learning theory. The acquisition of phonology was viewed as a process of shaping the elementary sounds produced by the infant through reinforcement of successive approximations to the adult pattern. Imitation of adult speech patterns was thought to be a source of reward to the babbling infant and repeated practice on these novel motor habits was thought to serve the function of stamping in and automatizing them.
From these elementary phonological habits the words of the language were thought to emerge through parental reinforcement. It was said that the child could better control his environment by uttering words to which the parents responded by giving the child what he wanted. The child learned the meaning of words through a conditioning process whereby the referents which the word signalled appeared in contiguity with the symbol thus establishing an association. The acquisition of grammar was conceptualized as learning the proper order of words in sentences. Generalization carried a heavy theoretical burden in attempts to explain novel uses of words and novel arrangements of sentences. Perceptual similarity of physical objects and relations, and functional equivalence of responses were thought to serve as the basis for generalizing the meaning of previously learned words. Similarly, generalization of the grammatical function of words was thought to account for the understanding and production of novel sentences.
Two aspects of this approach are noteworthy. One is that the burden of language acquisition was placed on the environment: the parents were the source of input and reinforcement was the necessary condition for establishing the habits. The child was merely a passive organism responsive to the reinforcement conditions arranged by agencies in the environment. The second aspect to be noted was the relatively simplistic conception of the knowledge to be acquired: sentences were conceived as orderings of words, arranged in sequential probabilities that could be learned then generalized to novel combinations. A general characterization of this overall approach would be to say that the process of acquisition was from surface to base; that is, the knowledge represented by language learning at all levels-phonological, semantic, syntactic-was entirely based on the relations contained in the overt speech of the parents. The new approach to be discussed below can be characterized by saying that it reverses this order; that is, the burden of acquisition is now placed on the child with relatively minor importance attached to the environment as a reinforcing agency. Furthermore, the approach minimizes the relations contained in the surface of language, attributing the significant information to be acquired to the underlying structure of language which is not contained in the surface input. However, before taking up this new approach I shall point out the specific inadequacies of the earlier approach.
1.1.1 The Acquisition of Phonology
The notion that the child first learns the constituent elements of the
adult phonemic structure and then produces speech by associating these elements
appears to be contrary to fact. In the first place it is doubtful that speech
is made up of a concatenation of physically unique sound elements. A sound
typewriter which would convert each physically different sound into a different
orthographic type would not produce a very readable record (Lenneberg, 1967),
because speech recognition is not simply a process of identifying physical
differences in sounds. In fact it requires overlooking certain acoustic
differences as unimportant and paying attention to certain other features
in relation to the acoustic context in which the sound is imbedded. In other
words, the "cracking" of the phonological code of a natural language
involves a process of pattern recognition and equation, not simply learning
the identity of constituent elements. The first recognizable words of a
child are not composed of acoustically invariant speech sounds (see Lenneberg,
1967). Therefore, a description of phonological acquisition in terms of
learning individual speech sounds which are then combined into words must
be false. Furthermore, it is not clear how a notion of shaping by successive
approximation can ever account for the acquisition of sound pattern recognition
and the discovery by the child of phonological structure of a hierarchical
nature.
1.1.2 The Acquisition of Meaning
It is an indication of the simplistic character of previous behavioristic views of language that they have concerned themselves with the problem of reference to the almost total exclusion of the semantic interpretation of utterances. Reference deals with the relation between words and objects or aspects of the environment. Psychological theories of meaning (or reference) were based on a philosophical system of conceptualization which now appears to be false; namely the notion that "words tag things" in the physical environment. The adoption of such a view led to elementary descriptions whereby a particular combination of sounds (a word) was conditioned to an object or set of objects. When a new object having certain physical similarities to the one previously conditioned was encountered, the learned verbal response was said to have generalized to this new instance. More elaborate versions of this form of theorizing were developed to account for the obvious fact that familiar words would be used in connection with objects or situations which had no physical similarity to the originally conditioned object. However, due to the requirements imposed by viewing meaning as a conditioned response to a stimulus, these later elaborations merely pushed back the locus of the similarity from the external physical object to an internal (even though functional) representation of that object. Thus an individual's capacity to understand the extension of the word eye in the eye of the needle was thought of as arising from the fact that the internal conditioned responses elicited by the word eye in the above phrase are similar in some (unspecified) manner to the responses originally conditioned to the word eye in such instances as this is your eye, these are my eyes, this is the doggy's eye, etc. The total inadequacy of this kind of approach as an explanatory device is this: it leaves obscure the specific nature of the similarity of the conditioned response from the original to the extension, and it is incapable of specifying the nature of the extension and cannot predict it until after it has occurred. Thus the view of reference as a conditioning process has the same shortcomings for semantics as the view of conditioning of sequential probabilities of parts of speech has for syntax. That is, the creative and novel use of words which is so characteristic of language remains completely beyond its explanatory range.
The difficulties attached to these behavioristic explanations of meaning can be resolved by abandoning the notion that "words tag things" in favor of the view that "words tag the processes by which the species deals cognitively with its environment" (Lenneberg, 1967, p. 334). This view reverses the order between the object stimulus and its conditioned response-process. That is, rather than saying that the concept-meaning involved in the use of the word eye is a conditioned process (external, internal, or cortical) developed as a result of tagging various objects having certain characteristics and experiences relating to them, this view says that the word eye tags a class of cognitive processes developed through a categorization and differentiation process which is independent of verbal labeling. When a child (or adult for that matter) is confronted with a new word, the new word acquires meaning only in the sense that it comes to refer to a class of cognitive processes already possessed by the individual. Novel uses of words, such as metaphoric extensions, are understandable to others by virtue of the fact that human categorization and differentiation processes are similar across the species, the word merely serving as a convenient tag whereby these processes can be labeled. The language of stimulus-response theory does not seem to 8 offer any particular advantages when conceptualizing the problem in this fashion.
A conception of meaning such as the one just outlined has certain implications for a theory of semantics which it might be important to state explicitly. Meaning becomes a purely cognitive concept (as linguists of a generation ago used to believe) and semantics represents the linguistic expression of these cognitive operations. The problem of the development of meaning becomes the problem of cognitive development, which is to say that the dimensions of meaning-how the human species categorizes and differentiates the universe-antedate the dimensions of semantics-how cognitive categories and relations find expression in linguistic terms. An adequate theory of meaning must be able to characterize the nature of this relation, namely the mapping of cognitive to linguistic processes. Note that this theory includes not only lexical (vocabulary) items but also the morphophonemic and inflectional system of language, since the latter contain cognitive differentiations such as present vs.: past, animate vs. inanimate, definite vs. indefinite, mass vs. count, male vs. female, plural vs. singular, and so on. It follows that an adequate theory of semantics must concern itself not only with the vocabulary of a language and the relation between words and things (reference) but also with the manner in which the syntactic component of a language allows the expression of cognitive relations (meaning). While the first aspect may be conceptualized as a closed system such as that represented by a dictionary of a language, the second aspect is an open system that cannot be described by taxonomy of properties or relations. In other words, while it is possible to make an inventory of all the words in a language, it is impossible to make an inventory of all the possible usages of any single word (with the exception perhaps of most function words). An adequate semantic theory must therefore contain at least the following two things: (a) a model of human cognition specifying a finite set of dimensions or features, probably in the form of a generic hierarchy of increasing inclusiveness as we move up the tree, and (b) a set of finite rules (or transformations) specifying the possibilities of manipulations of the elements in the tree. The description of (a) must be a general psychological theory and is made up of "psychological or cognitive universals" as defined by the biological capacity of the human species. The description of (b) must be a cultural and individual psychological theory as defined by individual differences in general intelligence and in personal experiences.
1.1.3 The Acquisition of Syntax
The failure of behavior theory to account in any significant manner for the problem of the acquisition of syntax can be interpreted as stemming from a failure to recognize the complexity of the syntax of language. As long as sentences are viewed as a sequential ordering of words or categories of words and the phenomenon to be explained as a problem in the learning of sequential probabilities of items or classes of items, no meaningful progress can be made. The relations among the following eight sentences taken from Lenneberg (1967, pp. 273-275) illustrate the complexities of the problem to be dealt with
(1) colorless green ideas sleep furiously
(2) furiously sleep ideas green colorless
(3) occasionally call warfare useless
(4) useless warfare call occasionally
(S) friendly young dogs seem harmless
(6) the fox chases the dog
(7) the dog chases the fox
(8) the dog is chased by the fox
If one compares sentence (1) and (2) it is evident that (l) is grammatical while (2) is not. The difference cannot be entirely in their meaning for, although sentence (1) is more likely to have some meaning than sentence (2), nevertheless sentence (1) will be judged more grammatical than sentence (2) even by the most prosaically inclined person. Nor can it be said that the reason sentence (1) is more grammatical than sentence (2) is that it is more familiar, since both sentences had a frequency of zero until linguists began to use them a short while ago to make the kind of point that is being made here. The ungrammatical string (4) has the same order of parts of speech as the grammatical string (1), namely (adjective + noun + verb + adverb). Similarly, the grammatical and semantically interpretable sentence (3) has the same order of parts of speech as the ungrammatical and semantically uninterpretable string (2), namely (adverb + verb + noun + adjective) [Sentence (3) might occur, as Lenneberg points out, "in an instruction booklet on pacifistic rhetoric" 1967, p. 274]. Consequently, the transitional probability of parts of speech in a sentence cannot account for either their grammaticality or their susceptibility to semantic interpretation. The same is true for the order of morphemes in the sentence as shown by the fact that sentence (5) which is both grammatical and meaningful uses the same order of bound morphemes (-ly, -s, -less) as sentence (2) which is neither grammatical nor meaningful. Sentences (6) and (7) demonstrate that the particular words used offer no clue to the meaning of the sentence. Sentence (8) can be recognized as having the same meaning as sentence (6) even though the order of subject and object is the same as that of sentence (7) showing that directional associations between the ordered elements are irrelevant to the understanding of the sentence.
These various examples should suffice to convince one that the process of acquiring language must involve a much more complex analysis procedure than that offered by such surface relations of sentences as order of elements and word-associations. As if this were not enough, we are confronted with the added complication that the child is continuously exposed to both well-formed and semiformed and semigrammatical sentences in the ordinary speech of adult speakers. Out of this confused input he has to be able to separate the false clues from the correct ones, yet he demonstrates this ability and succeeds in the relatively short period of 24 months (roughly from age one-and-a-half to three-and-a-half). Let us now turn to these newer formulations of child language acquisition.
If we discard earlier theories of language acquisition as unproductive, it is necessary to start anew right from the beginning. The study of the acquisition of grammar usually begins when the child is at about a-year-and-a-half, the time when he begins to use two word combinations. Prior to that it is difficult to study the child's grammatical competence since he uses single words, and techniques have not as yet been developed to study the child's grammatical comprehension at that early age. Speech records of a child over successive periods offer a picture of a changing grammar which the psycholinguist attempts to characterize in formal terms by giving a description of its structure at each period. This approach is necessarily limited since an inference of grammatical competence must be made from the child's speech performance, the latter being affected by a number of variables that are not directly relevant to grammatical competence (e.g., memory span, temporal integration, inattention, etc.). Given this limitation we can nevertheless inquire as to the kind of developmental picture that emerges.
1.2.1 Differentiation of General Classes
Children's earliest utterances of two words (or more) exhibit nonrandom combinations of words. Some examples from the speech of three children reported in the literature are the following (McNeill, 1966, Table 1): big boy, allgone shoe, two boot, that baby, here pretty. Distributional analysis of these two-word combinations reveals that the words the child uses at this earliest stage fall into two categories in terms of their privileges of occurrence. One of the two classes contains a small number of words each having a relatively high frequency of occurrence. Examples of this class include allgone, big, my, see in one child's speech, my, two, a, green, in a second child's speech, and in a third, this, a, here. The second class contains a larger number of words and additions of new words to this class occur at a higher rate (some examples are: beat, Mommy, tinker-toy, come, doed). Words in this second class occur by themselves or in combination with words from the first class, whereas words in the first class never occur alone. For these reasons the first class was named the pivot class (P) while the second class was named the open class (O). A shorthand expression of these facts can be represented by the following notation.
S --> (P)+O
This notation implies that the child's competence includes a rule which says that a sentence, S, can be produced by combining any two words from class P and class O (in that order) or, alternately, by using any single word from class O. The rule excludes such sentences made up of two words from the same class, or a sentence made up of a single word from the P class.
It is to be noted that the rule' for constructing this earliest sentence cannot have been developed as a result of direct mimicking of adult sentences. Many of the two-word combinations that this rule generates are in the wrong order from the point of view of adult speech (e.g., allgone shoe vs. the likely adult model of the shoe is allgone). In addition, it permits combinations that are unlikely to occur in adult speech at all (e.g., big milk). Such novel (and nonadult) combinations and the ready substitutability of words within each category are convincing arguments that these word combinations could not be memorized imitations of adult speech.
Distributional analysis of successive speech records of the children that have been studied shows that the words in the original pivot class begin to subdivide into progressively more differentiated categories in a hierarchical manner that can be represented as follows:
(based on McNeill )
This representation shows that the original pivot class (P1) subdivided into three classes of words: Articles, Demonstrative Pronouns, and all the rest (P2). Subsequently P2 subdivided into three further classes: Adjectives, Possessive Pronouns, and all the rest (P3).
The implications of this picture are extremely important. Note that there
is no logical necessity for the development of grammatical distinctions
to assume this particular form of development. The child could have made
up categories of words on a trial and error basis, continually rearranging
them on the basis of evidence contained in adult speech. He could thus isolate
a category of words that correspond to adjectives, or articles, or possessives,
until he gradually homes in on the full fledged adult pattern. However,
instead of making, as it were, a distributional analysis of adult speech,
he seems to have come up with a progressive differentiation strategy that
has the peculiar property of being made up of a generic class at each point:
that is, the original pivot class must already honor in a generic form all
the future distinctions at level 2; the undifferentiated pivot class at
level 2 (P2) must contain in a generic form all the future distinctions
at level 3, and so on. In other words, the child seems to honor grammatical
distinctions in advance of the time they actually develop. How is this possible?
McNeill's conclusion is as bold as it is inevitable: the hierarchy of progressive
differentiation of grammatical categories "represents linguistic universals
that are part of the child's innate endowment. The role of a universal hierarchy
of categories would be to direct the child's discovery of the classes of
English. It is as if he were equipped with a set of templates against which
he can compare the speech he happens to hear from his parents.... We can
imagine, then, that a child classifies the random specimens of adult speech
he encounters according to universal categories that the speech exemplifies.
Since these distinctions are at the top of a hierarchy that has the grammatical
classes of English at its bottom, the child is prepared to discover the
appropriate set of distinctions" (1966, pp. 35-36).
The assumption of innate language universals is sure to be unacceptable to current behaviorist theories. Someone is bound to point out that one does not explain the why of a complex phenomenon by saying it is innate. The fact of the matter is, however, that the complex behavior system of any organism is bound to be dependent upon the structural and functional properties of its nervous system. Language is a product of man's cognition, and, as Lenneberg (1967, p. 334) points out, "man's cognition functions within biologically given limits." Granting the innateness of language universals, we are still left with the task of explaining the how of language acquisition. The scientific investigation of language, both from the linguist's and the psycholinguist's point of view, is to give an adequate characterization of the structure of the child's innately endowed "language acquisition device," the nature of its universal categories and their interrelations.
1.2.2 The Development of Transformations
The ability to manipulate transformations constitutes an essential part of linguistic competence according to the linguistic theory developed by Chomsky, and Lenneberg (1967) argues convincingly that transformations are an essential aspect of categorization processes of all biological organisms. An insight into the nature of linguistic transformations can be gained by considering the manner by which the following two sentences are understood by an adult speaker (based on Lenneberg, 1967, pp. 286-292):
(1) they are boring students
(2) the shooting of the hunters was terrible
Both sentences are semantically ambiguous. The ambiguity in sentence (1) can be resolved by a process of "bracketing" which reveals that its constituent elements can be broken up into two different "phrase markers,"2 as follows:
This phrase marker shows that the ambiguity of the sentence lies in the fact that the word boring functions in one case as an inflected verb-form, and in the other case, as an adjective modifying the word students. Now consider sentence (2): it is ambiguous in at least two ways (one could say that either the hunters need more practice or they need a funeral!). Only one phrase marker description is possible for this sentence, so we need some other process to explain its ambiguity. One interpretation is related to the sentence hunters shoot inaccurately, the other, to the sentence hunters are shot. The reason we understand the ambiguity of sentence (2) may thus be attributed to the fact that we are able to recognize the relation between it and two other sentences each of which has its own distinct phrase marker. This type of relationship is the essence of transformations: they are laws that control the relations between sentences that have "grammatical affinity."
The early stages of child language competence does not apparently include the ability to perform transformations, according to McNeill (1966) who relates the impetus for acquiring transformations to the cumbersomeness of having to manipulate the elementary forms of sentences in the underlying structure of language ("base strings"). (Since more extensive discussion on the development of transformations is not possible here, the reader is referred to McNeill, 1966, pp. 53-65.)
1.3 IMPLICATIONS FOR SECOND LANGUAGE TEACHING
The view on language acquisition that has been outlined may at first appear frustrating to those whose inclination and business it is to teach language. The claim that a child has achieved linguistic competence by age three-and-a-half is likely to be scoffed at by the elementary school teacher in composition. At the claim that grammatical rules are discovered by the child through linguistic universals, the foreign language teacher is likely to wonder what happened to this marvelous capacity in the foreign language laboratory. In this section I would like to examine the implications for language teaching of the views outlined earlier on the language acquisition process. I shall discuss a number of topics including the role of practice and imitation, the distinction between competence and performance, and the nature of skills involved in foreign language acquisition.
1.3.1 The Role of Practice and Imitation
The assumption that practice plays a crucial role in language acquisition has been central to earlier speculations. To behaviorists it is almost an axiom not to be questioned. This view rests on the basic assertion that there exists a fundamental continuity between language acquisition and the forms of learning studied in the psychological laboratory. Chomsky (l959), Miller (l965), Lenneberg (1967), and others have questioned this view on general grounds and McNeill (1966) questions it on more specific and reduced grounds. If we grant that the language acquisition process is guided by the child's innate knowledge of language universals, does practice theory explain how children go about finding out the locally appropriate expression of the linguistic universals?
Practice theory leads to two possible hypotheses about language acquisition: one is that when the child is exposed to a novel grammatical form he imitates it; the other is that by practicing this novel form, he stamps it in. The evidence available indicates that both hypotheses are false. A direct test of children's tendency (or ability) to imitate adult forms of speech shows that children almost never repeat the adult sentence as it is presented. A child does not readily "mimic" a grammatical form that is not already in his repertoire as evidenced by his own spontaneous utterances. Direct attempts by the child at imitation of adult sentences end up as recodings, as the following examples taken from Lenneberg (1967, p. 316) illustrate:
Model Sentence Child 's Repetition
Johnny is a good boy. Johnny is good boy.
He takes them for a walk. He take them to the walk.
Lassie does not like the water. He no like the water.
Does Johnny want a cat? Johnny wants a cat?
It has been estimated that only about ten percent of a child's imitations of adult speech are grammatically progressive, that is, embody a form novel to the child.
Whatever the means by which novel forms enter the child's speech, does practice strengthen these responses? The evolution of the child's command of the past tense of verbs provides negative evidence to this question. In the child's early language the past tense of the irregular strong verbs in English (came, went, sat) appears with high frequency relative to the regularized /d/ and /t/ forms of the weak verbs. Thus, we would expect that these much practiced irregular forms would be highly stable, more so than the regular forms. Yet evidence shows that they are in fact less stable than the less practiced regular form, because at a certain point in the child's development he suddenly abandons the irregular form in favor of the regularized form and produces comed, goed, sitted. This kind of discontinuity shows that the practice model is not applicable here; rules that the child discovers are more important and carry greater weight than practice. Concept attainment and hypothesis testing are more likely paradigms in language development than response strength through rote memory and repetition.
This realization ought not to lead us to pessimism about the potential usefulness of language teaching. There is strong evidence that the attainment of grammatical rules can be facilitated by proper presentation of speech materials. Observation of children's speech during play interaction with an adult (usually the mother) shows that up to half of their imitations of adult expansions of children's speech are grammatically progressive (McNeill, citing data by Slobin, 1966, p. 75). An expansion is an adult's correction of the child's utterance. The advantage expansions seem to hold over other samples of adult speech may be attributable to the fact that expansions exemplify a locally appropriate expression of a linguistic universal at a time when the child is most ready to notice such a distinction. For example, if the child says Adam cry, and the mother expands this by saying, Yes, Adam cried (or Yes, Adam is crying-depending on her understanding of what the child intends), the child is thereby given the opportunity to discover the specific manner in which the past tense form (or progressive form) is expressed in English at a time when this distinction is maximally salient to him. The faster development of language in children of middle-class educated parents may be attributable to a tendency on the part of these mothers to expand to a greater extent than other parents. However, this hypothesis needs further investigation
1.3.2 On the Distinction between Competence and Performance
This distinction has been recognized by all psychological theories, including behavioristic ones (see Hull's, 1943, distinction between SHR and SER). A confusion that may arise in language behavior comes from the fact that understanding is usually (if not always) superior to speaking and one might want to equate understanding with competence and speaking with performance. However, this can not be the case. Both understanding and speaking must be viewed as instances of performance since the nonlinguistic factors that affect speaking (e.g., memory span, temporal integration, inattention, etc.) are equally likely to affect understanding. We are thus confronted with the fact that one type of performance, understanding, appears to develop before another type of performance, speaking. What may be responsible for this?
McNeill ( 1966) examines the specific claim that every grammatical feature appears first in understanding and second in speaking and is led to the conclusion that the overall parameters of conversion from competence to performance are simpler, easier, and less complex in the case of understanding. In order to account for this fact he postulates three kinds of memory span of different size or length, in the following order of decreasing magnitude: phonological production, grammatical comprehension, and grammatical production. He postulates these kinds to account for some data by Fraser, Bellugi, and Brown (1963) showing that a child can repeat a longer sentence than it can either understand or produce spontaneously, and also that it can understand a longer sentence than it can produce spontaneously. The difficulty with McNeill's hypothesis is that it equates sentence length with sentence complexity. It would seem that it is easier to understand a long but simple sentence than a short but involved one. It would also appear that one can understand a sentence too long to be repeated. Children show evidence of having understood sentences they cannot (or will not) repeat (see Lenneberg, 1967, p. 316). The problem may be conceptualized in a different way, as illustrated by the following diagram:
The asymmetry between the capacity to perform the understanding conversion as opposed to the speaking conversion may be related to the fact that the former requires an analytic approach while the latter demands a synthetic capability. It may be that for humans analytic processes are easier than synthetic ones. One might say that it is easier to learn the art critic's job than the artist's.
1.3.3 The Acquisition of Foreign Language Skills
Let us raise the question of the specific relevance of our discussion on first language acquisition for an understanding of second language learning and teaching. What are the parallels to be considered? First, let us look at the argument for the differences. Assuming second language acquisition which takes place after the age of four, one may point out the following: (a) the individual's cognitive development is at a later and more advanced stage; (b) he is already in possession of the grammatical structure of a language which may seem to facilitate the acquisition of a second one through transfer; (c) he already possesses concepts and meanings, the problem now being one of expressing them through a new vocabulary.
The importance of the first argument would seem to depend on the relevance of cognitive development for the acquisition of language. The view outlined in this paper is that the necessary knowledge for language acquisition cannot be gained from experience with the outside world and that language acquisition is dependent on an innate endowment which constitutes the knowledge of language universals. Hence the imputed advantage of advanced age and cognitive development is a dubious proposition. The two other arguments are based on the assumption of the operation of transfer in grammatical structure and in reference (vocabulary). What is the evidence in support of this assumption? It is necessary to distinguish between two claims about transfer theory. One refers to the general expectation that new forms of learning do not go on independently of what the organism has learned before. The truth of this statement would seem fairly obvious and need not concern us further. The second and specific claim expresses the expectation that the learning of certain specific and identifiable elements in Task B is facilitated (or hindered) by the previous learning of certain specific and identifiable elements in Task A. The status of this strong claim for any type of complex learning outside the laboratory is unknown. A serious test of it in second language acquisition would require the prior analysis of the two languages in a form which would identify the specific elements to be transferred at the grammatical and lexical levels. On a priori grounds we would expect negative transfer as much as positive transfer, assuming that transfer is relevant to the problem. Carroll ( I 966b) claims that the Modern Language Aptitude Test designed for English speakers predicts success in a foreign language equally well regardless of the particular language involved. This fact is difficult to explain if transfer has any overall relevance to the language acquisition process. Nevertheless, some phonological studies on contrastive analysis reviewed by Carroll would seem to indicate the operation of negative transfer effects. He cites Suppes et al. ( 1962) who "claim to be able to predict quite precisely from mathematical learning theory what [phoneme] discrimination problems will arise" (Carroll, 1966a, p. 16).
The problem is complicated still further by the possibility that transfer effects might affect performance and competence factors in different ways. Or the various performance factors themselves (understanding, speaking, reading, writing) may be affected to different degrees. The same comment might be made for different levels of performance, that is phonology, vocabulary, and syntax. A further aspect to this problem is the consideration of whether transfer effects are necessary processes or whether the extent of their operation is dependent on the strategy with which the learner attacks the new task. An individual who tries to "fit in" the dimensions of the new task into the old structure may encounter different problems from the individual who inhibits the interaction of the two tasks, if we assume that the latter strategy is possible. Finally, the fact that it is possible to predict errors of confusion, as in contrastive analysis of phonology, is not necessarily an indication that transfer effects will operate in the acquisition of the new task. Thus, the fact that the [I] and [r] sounds are predictable areas of confusion for a Japanese learning English says nothing about the way in which he will eventually learn the distinction. It is unlikely that this distinction is learned in isolation. Instead, it is more likely that the confusion will disappear when the overall structure of English phonology is internalized. (For a more extensive discussion on transfer see Chapter 4 Section 4-7.)
The above considerations lead to a number of implications for the teaching of a second language which I shall now take up.
1.3.4 Teaching the Knowledge of Structure
Since it is clear that knowledge of language at all levels consists of knowing patterns of relations rather than constituent elements, the usefulness of efforts to teach the latter is in doubt. Examples of such efforts include teaching specific sound discriminations, shaping phonological production, increasing vocabulary through association of translation equivalents, and practicing specific morphological and inflectional examples. Pointing to individuals who successfully acquired a foreign language in a course using these methods has no force of argument, for it is quite possible that their success occurred despite these methods rather than because of them.
1.3.5 Teaching Successful Strategies of Acquisition
Carroll (1962) has isolated a number of factors which are predictive of success in a foreign language. These factors may offer clues about the strategies that a successful learner uses with the possibility that such strategies may be taught to those who normally make no use of them. One of the abilities Carroll has identified deals with verbalization of grammatical relations in sentences. The successful foreign language learner is apparently capable of the following task: given a word italicized in one sentence (e.g., "The man went into the house.") he can identify that word in another sentence which has the same grammatical function (e.g., picking one of the italicized words of the following sentence: "The church next to the bowling alley will be built in a new location next year. "). We know of course that the individual is capable of recognizing the grammatical relations in the second sentence (otherwise he could not give it a semantic interpretation), so the ability must be one of explicit verbalization of implicitly known rules and relations. Verbalizing a grammatical relation can take two forms; one refers to the type of statement that can be found in a grammar book that includes technical terms (relative clause, head noun, modifier, predicate phrase, etc.); the second refers to a statement of equivalence or relation expressed in any convenient way using whatever terms are available to the individual, whether technically correct or not. The teaching of such verbalizations therefore ought to facilitate foreign language acquisition.
Another variable identified by Carroll "is the ability to 'code' auditory phonetic material in such a way that this material can be recognized, identified and remembered over something longer than a few seconds" (1962, p. 128). We do not know at present the specific strategy that may be employed in facilitating this kind of coding. Whatever the strategy may be, it seems unlikely that the superior person in this task derives his advantage from a special innate capacity. In the first place the strategy is not related to the ability to perceive phonetic distinctions, and second, given the biological foundations of language capacity (see Lenneberg, 1967), we would not expect innate differences in the general capacity of coding phonological material.
Contrastive analysis of grammatical structure would not seem to offer particular advantages beyond those provided by verbalization of grammatical relations and by attention to a grammatical distinction at a time of saliency (see the effects of expansion, discussed above). The expectation that the advantage of contrastive analysis lies in making the contrast per se is based on an assumption of transfer for which evidence is lacking. At any rate the pointing up of the contrast may just as well lead to negative transfer by facilitating the assimilation (or "fitting in") strategy. I know of no evidence that emphasizing distinctions of incompatible responses, especially those that are automatized, leads to a decrease in incompatibility.
1.3.6 Teaching Habit Integration and Automaticity
Temporal integration of phonological skills, both of understanding and production, is a problem independent of the knowledge of the phonological structure and transformations of a language. It would seem likely that sensory and motor integrations of this type can be automatized through practice and repetition. The more interesting problem would relate to the time at which automaticity practice is likely to be valuable and to the form it is to take. Reading represents a different aspect of phonological production skill than speaking, as is well known, and practice in reading does not represent a sufficient or necessary condition for achieving automaticity of phonological production in speaking.
The factors that enter into the problem of automatizing grammatical habits are not very clear. Tests of speech comprehension under conditions of noise (see for example Spolsky et al., 1966) seem to be quite sensitive to the level of automaticity and degree Of integration achieved by a foreign language speaker. They show that the problem of integration goes deeper than high proficiency in understanding and speaking demonstrated under ordinary conditions. At the moment we do not have available a psychological theory of sentence understanding or production. The relevance to this problem of recent experiments on latency of various grammatical manipulations still remains to be shown. Many language teachers seem to be convinced that pattern drills serve to automatize grammatical habits. However, it is difficult to justify this expectation on theoretical grounds. I have already argued that the semantic interpretation of a sentence cannot be viewed as a process of sequential analysis of categories of words. Thus, pattern drills, at best, can serve only to automatize phonological production skills, and for this latter purpose other methods may prove equally, if not more effective. At any rate, if the pattern drill argument is taken literally, namely that the structure is automatized through practice of the specific pattern that is being repeated, then the learner could never achieve automatized speech. This consequence must follow since in ordinary speech we use an infinite variety of patterns, and therefore, since the second language learner could not possibly be drilled on an infinite variety of patterns, he could never develop automatized speech. Hence pattern drill cannot possibly do what it is supposed to do.
From a theoretical point of view the development of grammatical competence should be facilitated by getting the learner to perform a set of transformations on families of sentences (e.g., I cannot pay my rent because I am broke; if I weren't broke I could pay my rent; Given the fact that I have no money, I cannot pay my rent; How do you think I could possibly pay my rent if I am broke; Since I am broke, the rent cannot be paid; To pay the rent is impossible given the fact that I have no money; etc.). The distinction between this exercise, which we may refer to as perhaps a "transformation exercise," and "pattern drill" is that the first deals with the competence involved in deep structure while the second focuses on surface structure. Rutherford (1968) has shown that surface structure similarities are completely unenlightening as to the semantic interpretation of sentences.
The notion of transformation exercises is equally applicable to phonology and vocabulary. Exercises in vocabulary transformations are more difficult to specify at this stage of our knowledge, but from our earlier discussion of meaning we can perhaps anticipate giving the student a task of this kind: "Change the following list of words using the gender transformation: boy, father, bull,"-which might yield: "girl, mother, cow." Other examples might include asking the student to give opposites, similars, subordinates, superordinates, and so on, in a restricted word-association task. Semantic relations of this kind may be responsible for the well known psychological fact that in memory words are organized in clusters (see, for example, Deese, 1965).
1.3.7 On Semigrammatical Sentences
The fluent speech of most native speakers does not consist totally (or even in the majority of instances) of well formed sentences. One would imagine that the imposition of a requirement to utter exclusively well formed sentences would seriously hinder the fluency of most native speakers. The logical implication of this observation would be that no language teacher should ever force his pupils to use only well formed sentences in practice conversation whether it be in the classroom, laboratory or outside. This conclusion is not as odd as it might seem at first sight. After all, children seem to acquire the competence to produce well formed sentences despite the semigrammaticality of the adult speech to which they are continually exposed. It is important to note that semigrammaticality does not mean randomness. The reason that in most instances we are able to give a semantic interpretation to semigrammatical sentences lies in the fact that we have the capacity to relate these semisentences to their well formed equivalents. There must therefore exist lawful transformations between semisentences and well formed ones. We are able to understand the speech of children for the same reason: the grammar of their utterances is generic of the later grammar of well formed sentences. If it were not so, we would not be able to expand (hence understand) their utterances.
An important question poses itself at this juncture: should second language teaching take specific account of the developmental stages that are likely to mark the acquisition of a language? By ''specific account" I mean at least the following two propositions: First, to recognize and allow the production of semisentences on the part of the learner; and second, to expose the learner to utterances which are grammatically progressive at each stage but which fall short of having the full complexity of well-formed sentences. The first proposition may already be the policy in some modern and intensive courses which encourage active speech production "at any cost" [sic]. The second proposition is sure to be resisted by most teachers; yet the fact of the matter is that all natural language acquisition situations expose the learner to semigrammatical sentences more often than not. We do not know whether these are facilitative or retarding situations. Some parents tend to talk to their children by attempting to imitate their speech and it is sometimes said that this kind of "baby talk" retards acquisition. The evidence on this point is lacking. It may be of course that the fastest method of acquiring a second language need not be one that replicates the conditions existing under natural language acquisition. In fact various claims for highly intensive language courses followed by individuals with high foreign language aptitude put the time requirement for the acquisition of a foreign language at between 250 and 500 hours of study (Carroll, 1966a, b). Compare this figure with a minimum estimate of 3,000 hours for first language acquisition. This rough figure is arrived at by estimating the total waking hours of a child up to age three-and-a-half and taking thirty percent of that as an estimate of the amount of exposure to language. Of course the two situations are not directly comparable and the level of competence achieved may be different (especially by measures of automaticity and background noise, see Spolsky, et al, 1966); nevertheless, the comparison highlights the fact that certain aspects of the natural rate of language acquisition process can be greatly accelerated. It is important to note that although the language acquisition capacity per se must be viewed as an innate capability shared by all members of the species, the rate at which language is acquired, especially a second language, and the effectiveness with which language is used as a communicative process are performance factors that are affected by individual differences within the species (variations in general intelligence, in experiences, in physical health, in motivation, etc.). It is here that the concept of teaching may assume its full importance.
Traditional psychological theories about language acquisition emphasize the role of reinforcement provided by environmental agencies and view language as a set of vocal habits that are conditioned to stimuli in the environment. Imitation and practice of new forms are the processes by which language behavior develops and generalization of learned forms is supposed to account for the novel uses of language. Recent developments in linguistics have influenced our conception of the structure of language, hence the nature of the knowledge that the child has to acquire. A radically new psycholinguistic theory of language acquisition has been proposed which emphasizes the developmental nature of the language acquisition process and attributes to the child specific innate competencies which guide his discovery of the rules of the natural language to which he is exposed. Imitation, practice, reinforcement, and generalization are no longer considered theoretically productive conceptions in language acquisition. The implications of these new ideas for the teaching of a second language lie in the need for controlled exposure of the student to linguistic materials in a manner that will facilitate his discovery of the significant features of the language. "Shaping" of phonological skills, discrimination training on sound "units" and pattern drills are rejected in favor of "transformation exercises" at the phonological, syntactic and semantic levels.
Probably the most controversial aspect of this chapter is the argument that "pattern drill cannot possibly do what it is supposed
to do." The usual claim of the audiolinguist is that mere repetitive practice of patterns is sufficient for acquiring grammatical structure through a process of inductive reasoning subsequently followed by generalization to related patterns. This is essentially a one-factor theory of learning that proceeds from surface to base. This theory is rejected in favor of a two-factor acquisition theory that is composed of the following two steps: (a) the discovery of the underlying structure of the language by means of inductive and deductive inferences guided by (i) innate grammatical universals and (ii) sample linguistic data which are sentences and semisentences in the second language; (b) the automatization of the phonological surface transformations of this underlying knowledge through practice. Note that in step (a) exposure to the linguistic data is not a sufficient condition for acquiring the structure. Furthermore, at any one time in the acquisition process only those linguistic data will be relevant which are needed to test whatever particular hypothesis about underlying structure the learner is working on: the teacher cannot control the inference process directly by choosing the relevant linguistic data since he has no access to the learner's inferencing process (neither does the learner, since the process is unconscious). Note also that the automatization of speaking and understanding can come only after knowledge of the structure has been acquired. Thus, practicing sentence patterns whose structure is not yet understood would seem to be useless. In addition, the type of practice that leads to automatization in speaking is not repetition of sentence patterns given to the learner, but rather the phonological surface actualization of base sentences the individual produces himself.
This two-factor theory of language acquisition implies a very different teaching approach from that dictated by the audiolingual habit theory approach. First, it suggests that the learner should be exposed to the full range of linguistic data right from the beginning so as to give him maximum opportunity to test out his inferences about the underlying structure of the language. Second he should be encouraged to produce any sentence, even if "incorrect," to enable him to practice phonological surface transformations of base strings; "correction" of such semisentences by the teacher is helpful only when they represent "expansions," as discussed in this chapter in connection with first language acquisition. Third, drills and exercises are of dubious utility unless they represent attempts to communicate freely (as opposed to practicing a grammatical rule artificially).
It would be extremely difficult (in my opinion impossible) to demonstrate "experimentally" the correctness of the two-factor theory over the audiolingual habit theory, but there are both empirical and theoretical considerations which argue persuasively to that effect. First, there is the evidence (see Chapter 2) which shows that the audiolingual method has failed to produce meaningful foreign language achievement in the vast majority of students exposed to it. Second, the two-factor theory appears to be a more correct description of what goes on under natural conditions of language acquisition, whether first or second. Third, intensive language courses, which follow the implications of a two-factor theory more closely than those of the one-factor theory, have been consistently successful in their goals of teaching communicative skills. Fourth, the recent developments in linguistic theory and the empirical work on language acquisition in children, as reviewed in this chapter, are consistent with the two-factor theory while they specifically deny the assumptions underlying the audiolingual habit theory. In the last analysis it will not be a proof that will resolve this argument but the realization on the part of language teachers that teaching methods which do not work should be abandoned.
Attacks on the audiolingual method, such as this one and others as well, have generated heated counterattacks on the part of devotees who have committed a life time of work to the method and feel that without "definitive proof" it should not be abandoned. Unfortunately, in the heat of this polemical debate we have allowed ourselves to be drawn into a power struggle of clashing personalities and personal threat to established power structures. Surely the proponents of the audiolingual method are not to be personally blamed for the inefficacy of their approach to Foreign Language (FL) teaching. At the time it became entrenched in the profession, it did seem to be the best possible strategy, and the proponents deserve credit for the forthrightness and purity with which the approach was developed and applied. But educational practices, no less than scientific theories, continually evolve and there are no absolute "truths." It is thanks to the consistency and purity with which the audiolingual method has been steadfastly applied that it has been possible to test out its ultimate inefficacy and which made it possible now to move on to something potentially better. And when these newer ideas are eventually put into practice and found wanting, as surely they must be, we will be able to move on to still other, hopefully better, approaches. Let us not forget that our goals coincide in the highest interest of the pupils in our charge-the effective teaching of FLs.