CHAPTER ONE
Second Language
Acquisition Theories
Second language acquisition (SLA), like applied linguistics, is an interdisciplinary field with widely varying interpretations placed on it, and varying demands made of it. For these reasons, and because it continues to be developed by individuals from diverse disciplinary backgrounds and with varied epistemological allegiances, it has become increasingly fragmented during the past 15 years, and is characterized by a multiplicity of theories (broadly construed). The theories differ substantively, and in at least five other ways: source, scope (or domain), content, type, and form.
Until recently, with only a few exceptions, research of very different kinds by numerous individuals, groups, and even whole âschoolsâ working under the rubric of one of these theories, or with no theory, has tended to continue in relatively untroubled isolation, without this being seen as problematic. Of late, however, circumstances have changed, portending a period of change in SLA theory, and in the way research is conducted: There is growing public pressure for accountability from SLA researchers. Increasingly robust empirical findings on several issues have lent support to some theories, but cast doubt on others. And, because all theories are at some level interim understandings of what theorists ultimately seek to explainâin this case, how people learn second languagesâidentifying faulty understandings, and amending or culling the theories concerned, constitutes progress, whereas persistence of a plethora of theories, especially oppositional ones, obstructs progress.
Acceptance of the need for absolute or comparative theory evaluation involves coming to terms with what are sometimes controversial and conflicting evaluation criteria used for theory assessment in other fields and/or developing (sometimes equally controversial and conflicting) new ones. Of particular, and long-running, concern in this regard is the unwarranted assumption in some quarters that SLA theories should be judged by their relevance for language teaching.
How SLA Theories Differ
SLA is a broad, expanding, and diverse field. It encompasses, at the very least, the simultaneous and sequential learning and loss of second (third, fourth, etc.) languages (L2) and dialects, by children and adults, with differing motivations, abilities, and purposes, as individuals or whole communities, with varying access to the L2, in formal, informal, mixed, foreign, second, and lingua franca settings. Researchers come to SLA with varied training in a variety of disciplines, includingâbut not limited toâlinguistics, applied linguistics, psychology, anthropology, education, and SLA. SLA theory and research findings are utilized in many fields, including first language acquisition, theoretical linguistics, neurolinguistics, language learning in abnormal populations, language teaching, education, and psychology.
Not surprisingly, given all this heterogeneity, the past three decades have witnessed the production of numerous theories of âtheâ SLA process. As discussed here, there is a tendency for researchers whose original graduate training was in a discipline outside SLA to bring with them into the field expertise and interests in issues and research methods that reflect this different training. Naturally enough, they then work on those issues, and as the pioneers doing the early research on the same topics in SLA, they import theories from other fields or develop new ones. The results are often little more (or less) than retrospective explanations of their own initial L2 findings. Whatever their merits, the theories developed apply only to a particular behavioral domain in many cases, and sometimes in a domain that others in the field would view as falling outside second language acquisition altogether (e.g., as treating one or another aspect of second language use).
As a result of these and other factorsâby some accounts, and depending on what one countsâthe literature offers as many as 60 theories, models, hypotheses, and theoretical frameworks, many of which are sketched here. These terms are often used nontechnically in the SLA literature (for review, see Crookes, 1992) and in much of what follows. Some view this situation as one of healthy, even inevitable, theoretical pluralism. Certainly, the proliferation is made more likely by the absence of one or more widely (cf. universally) accepted theories governing work in the field, a situation that in itself is indicative of the fieldâs immaturity, and by some accounts, signals prescientific chaos likely to obstruct progress as long as it lasts.
Source
The first way in which SLA theories differ is by source, that is, in their (primary) origins inside and/or outside the field. A number of theories have emerged, at least in part, from empirical research findings on second language learning. Examples include the ZIZA Groupâs Multidimensional Model (Meisel, Clahsen, & Pienemann, 1981), Krashenâs Monitor Theory (Krashen, 1985), Schumannâs Acculturation Model (Schumann, 1986), Cumminsâ Linguistic Interdependence Hypothesis for bilingual proficiency (Cummins, 1991), and Ellisâ Integrated Theory of Instructed Second Language Acquisition (R. Ellis, 1990). Other theories have been imported ready-made from related areas of cognitive science, notably from linguistics and psychology. Linguistic models tested as theories of SLA, or as parts of same, include Chomskyâs and othersâ theories of Universal Grammar (UG; e.g., White, 1996, 2003a, 2003b), Prince and Smolenskyâs Optimality Theory (e.g., Eckman, 2004), Bickertonâs Bioprogram Hypothesis (e.g., Huebner, 1983), Givonâs Functional-Typological Model (e.g., Sato, 1990), OâGradyâs general nativist theory (e.g., Wolfe-Quintero, 1992), and Bresnanâs Lexical-Functional Grammar (e.g., Pienemann, 1998). Work introduced from psychology includes Gilesâ Accommodation Theory (e.g., Beebe & Giles, 1984), Batesâ and MacWhinneyâs Competition Model (e.g., Kilborn & Ito, 1989), various connectionist models (e.g., Gasser, 1990), and several models based on Andersonâs and othersâ distinction between declarative and procedural knowledge (e.g., K. Johnson, 1996).
Domain, or Scope
Theories also differ with respect to domain, or scope (i.e., as to what they purport to explain, or their âcoverageâ). Most nativist theories, for example, focus primarily, or thus far even exclusively, on phonology, syntax, and morphology, and at the level of form only, whereas some functionalist theories (e.g., Bates & MacWhinney, 1989; Givon, 1979) attempt to account for the acquisition of all levels of language, consider data from all levels when seeking explanations, and attribute a major role to communicative function in driving language acquisition (indeed, language change of all kinds). The acquisition type and context of interest (naturalistic or instructed, foreign or second, individual or community, etc.) also varies. The primary focus of Schumannâs Acculturation Model, for exampie, is naturalistic acquisition by learners as members of groups. The central claim is that the degree to which members of such identifiable communities as Koreans in Los Angeles, Turks in Germany, Arabs in Israel, or Americans in the Middle East acquire the surrounding L2(s) will be a function of the extent to which they adapt to the new culture. This will depend on the social distance between their group and that new culture; social distance, in turn, is determined by eight factors: social dominance (the relative status and power of the groups in question), integration pattern (assimilation, acculturation, or preservation), enclosure (self-sufficiency in terms of such things as first language (LI)-medium clubs, newspapers, employment opportunities, shops, and churches), cohesiveness, community size, cultural congruence (e.g., shared or conflicting values and religious beliefs), attitude (positive to hostile), and intended length of residence (projected stay in the target language environment). Whereas the Acculturation Model speaks to naturalistic learning only, the domain of Ellisâ theory, as its name implies, is the classroom. Krashenâs Monitor Theory sets out to handle both naturalistic and instructed acquisition, but other things being equal (e.g., if each of the three theories were successful in their respective domains, and if they were comparable in other ways), the broader scope of Krashenâs theory would make it the preferred choice.
Content
Theories differ, thirdly, with respect to content (i.e., in the variables, and kinds of variables, that make up their explanatory core) and at a broader level, the relative importance accorded internal or environmental factors. Theories such as Schumannâs Acculturation Model and Gardnerâs Socio-Educational Model (Gardner, 1988) draw primarily on social and social-psychological variables for their accounts, attempting to predict that SLA will or will not occur, and the degree of likely success, mostly as a function of group membership and intra- and inter-group relations. Whiteâs UG-based theory and Eckmanâs functional-typological approach (Eckman, 1996), on the other hand, invoke linguistic theory and related findings from studies of child first language acquisition (LIA), attempting to predict how acquisition will occur at the level of the individual, not the group, and as an internal cognitive process, not a social one, mostly as a function of prior linguistic knowledge and L1-L2 relationships. Whereas social-psychological models tend to make claims about general learning outcomes, assessed via global measures of L2 proficiency, researchers working within the UG and functional-typological frameworks test detailed claims about learning sequences, and about the ease, difficulty, and learnability of particular phonological and grammatical structures for speakers of certain types of LI learning certain types of L2.
In Whiteâs UG-motivated work, for example, speakers of languages like French, which allows adverbs to be placed between verb and direct object (Je bois toujours du cafĂ©), are expected to have difficulty with languages like English, which disallows this (*I drink every day coffee), but not vice versa (see White, 1991a, and elsewhere, and for an alternative analysis; Schwartz & Gubla-Ryzak, 1992). (The facts about adverb placement are manifestations of a broader underlying structural difference between these and other languages.) English speakers should be able to learn the new option that French permits by encountering instances of verb-adverb-direct object strings in the input (positive evidence), whereas French speakers face the very different task of noticing the absence of such strings in the language they are learning, and so may need, or at least benefit from, various kinds of âcorrectionâ or explicit grammar instruction (negative evidence). Such an asymmetrical prediction is different from, and more accurate than, one that would have been derived from the old Contrastive Analysis Hypothesis (CAH; i.e., difficulty predicted for adverb-placement for both French and English speakers), with any L1-L2 difference being seen as a potential difficulty for the learner.
Also illustrating the level of detail and focus on process, but operating within a very different linguistic framework, Eckmanâs functional-typological approach, with its related Markedness Differential Hypothesis (MDH) and Structural Conformity Hypothesis, draws upon work on typological linguistic universÄls, combined with knowledge of L1-L2 contrasts, to make precise predictions about SLA in or out of classrooms. For example, where the presence of one structure (e.g., voiced stops: /b/, /d/, /g/) in the worldâs languages always implies the presence of another (e.g., voiceless stops: /p/, /t/, /k/), in various positions (word-initial, medial, or final), but not the reverse, the first structure is said to be âunmarkedâ relative to the second, and the second âmarkedâ relative to the first. Unlike the discredited CAH, which predicted, wrongly, that (all) such differences between LI and L2 would be difficult to learn (and all L2 features similar to those in the LI simple), the MDH predicts, generally correctly, that learners will have problems with those areas of the L2 that differ from, and are more marked than, the LI. Again unlike the CAH, which could not predict difficulty if LI and L2 both allowed the âsameâ structure, the MDH predicts varying degrees of difficulty, generally correctly, in such grammatical domains as relative clauses. Thus, where LI and L2 both allow relativization, but from different levels in a six-tier noun phrase accessibility hierarchy, previously demonstrated to constitute an implicational markedness scale in the worldâs languages (Keenan & Comrie, 1976), ease or difficulty with L2 relative clause formation can be accounted for in part by whether the more marked structures occur in the LI or L2 (see, e.g., Hamilton, 1994, and for review, Eckman, 1996).
Type
Schumann, Gardner, and other theorists focus on SLA as a social process, and draw primarily on situational and social-psychological variables to explain success and failure at the level of whole communities. White, Eckman, and others, conversely, emphasize SLA as a mental process, with the individual as the unit of analysis, and rely primarily on different kinds of linguistic theory to account for interlanguage development. Most of these and many other models fall into one or other of two broad camps: nativist (special, general, or hybrid) and empiricist.
Special nativist SLA theories (e.g., those of White, Krashen, and others) assume continued access by L2 acquirers, including adults, to genetically transmitted abilities specific to language learningâused for that, and nothing elseâincluding innate knowledge of highly abstract syntactic principles and of the parameters along which languages can vary (UG), or of a set of universal semantic distinctions (Bickerton, 1984), which are held to govern child LIA and adult SLA alike, the latter sometimes with modifications brought about by prior learning of the LI.
General nativist proposals, such as those of OâGrady and Wolfe-Quintero, hold that SLA proceeds without UG or any such language-specific innate knowledge and abilities, and is instead accomplished through use of modularized general cognitive mechanisms. In OâGradyâs (1996) formulation, there are five (perceptual, propositional, conceptual, computational, and learning) innate mechanisms, which suffice both for language and other kinds of learning, although possibly supplemented by a few (nonsyntactic) concepts used only for language. In OâGradyâs account, the relative failure of most adult SLA (following the relatively uniform success of child LIA) is posited to be due to degradation of, or reduced access to, three of the five modules.
Hybrid nativist models, such as those of Clahsen and Muysken (1986), and Bley-Vroman (1990âhis Fundamental Difference Hypothesis), are special nativist for LIA, usually holding it to be governed by UG, but general nativist for SLA, proceeding via general problem-solving procedures of various kinds. Bley-Vroman (1997), for example, proposed that adult SLA is accomplished by a general human cognitive categorization ability, manifested in SLA as the learning of variously weighted, category-based, and linked âconstructions,â or âpatterns,â roughly equivalent to phrase-structure rules, the learning process mediated by such phenomena as ânoticingâ (Schmidt, 1995), input frequency, saliency, and prototype effects.
From a theory construction point of view, although the success of particular exponents of these three general positions in predicting the facts about child and adult first and second language acquisition will ultimately determine their fate (see, e.g., Schwartz, 1992), general nativist theories have the initial advantage of being less powerful than special nativist theories, because they set out to handle the same data without recourse to innate linguistic knowledge. Special nativist and general nativist theories, in turn, are less powerful than hybrid nativist theories, because they attempt to explain the data on both first and second language acquisition using only one set of innate abilities, whether or not language specific, whereas hybrid theories assume both language-specific knowledge and general learning mechanisms, and are discontinuous (i.e., in this case, posit very different processes for child and adult and/or LI and L2 acquisition). The discontinuity is usually motivated by the need to account for putative âsensitive period(s)â effects (i.e., maturational constraints). Young children typically become native-like in a first or second language if allowed continued access to the language concerned, and do so in about the same time period, despite wide variations in intelligence, linguistic exposure, or the kind of language(s) involved; adults, conversely, typically exhibit partial, often gross, failure under the same conditions, this despite what on the surface, at least, would seem to be the distinct advantage of cognitive maturity, improved memory, and prior success with their native language (for reviews of findings, see, e.g., DeKeyser, 2000; Hyltenstam & Abrahamsson, 2003; Long, 1990b, and chap. 3 in this volume). General nativist theories need to account for the seeming lack of decline in the ability to perform other general learning tasks at the same age(s) as the closure(s) of the one or more putative sensitive periods. Discontinuous (hybrid nativist) models would require an alternative motivation altogether should the alleged biological scheduling turn out to be illusory, or alternatively, should not one, but several sensitive periods be discovered for different linguistic systems.
In contrast to nativist theories, empiricist models are âdata-driven,â with linguistic input acting on universal cognitive, not linguistic, architecture. As conceived here, they include a greater variety of theories than those in the nativist camp, ranging from functionalist linguistic accounts, through social-psychological models, to connectionism. Some empiricist positions are referred to as âsocial-interactionistâ (e.g., Gass, 1997) or âcognitive-interactionistâ (e.g., Andersen, 1989). Most theorists stress that the environmental factors they consider important interact with internal mental abilitiesâhence, âinteractionistâ theoriesâand their views by no means signal a return to behaviorism. In all cases, however, in combination with internal factors, the learnersâ or learner groupsâ linguistic experience is said to play a stronger determining role in acquisition than anything countenanced by nativists of whatever stripe.
Experience can refer to the amount and quality of contact with the target language and its speakers, as in the case of the Acculturation and Socio-Edu-cational Models, and as in several âskill-buildingâ models of acquisition. The latter are usually based on variants of the idea from general cognitive psychology that learning (here, language learning) is chiefly a matter of converting declarative knowledge (knowledge that) into procedural knowledge (knowledge how) through a process of automatization (for review, see K. Johnson, 1996, pp. 77â15...