1996 - mclau - hak - biling y l2

8/12/2019 1996 - McLau - Hak - Biling y L2

1/73

1SEVEN TENSIONS THAT DEFINE RESEARCH

IN BILINGUALISM AND SECOND LANGUAGE ACQUISITIONi

Kenji HakutaStanford Universityand

Barry McLaughlinUniversity of California, Santa Cruz

NTRODUCTIONT h e roots of the study of bilingualism and second language acquisition can be traced

to activities in multiple disciplines a s well as to catalytic social forces and needs that havearisen from the practice of language educa tion. The primary disciplines involved are

Piinguistics, psychology, anthropology and sociology, with additional contributions frombiology. The catalytic educational forces might be differentiated into foreign languageeduc ation (e.g., teaching of G er m an in the U nited States), second language ed ucation (e.g.,teachin g of English second language to immigrants to the U nited States, or in areas ofthe world wh ere English serves impo rtant societal functions such as commerce) and bilingualeducation use of both languages as a medium of instruction).

At least five major moments during this century can be identified when attention,either through disciplinary changes or practical needs, came to be intensely focused onbilingualism and second language acquisition. T he first phase (to be called thePsychometrics Phase) came in the 1920s during the height of interest in psychometricsand concern about the intellectual character of the new immigration (see Hakuta, 1986).


2/73

Since most of the new immigrants came from non-English backgrounds, questions aroseabout the possible effects of bilingualism on intelligence test performance as well as relatedquestions about whether immigrants were learning English and becoming American rapidlyenough. Although m uch of the research from this period is discredited because of unsoundsampling and measurem ent practices, the debate reflected many of the themes that continueto the present day about the nature of language and intelligence, whether they (and otherhuman abilities) are biologically based, and the nature of their relationship.

The second phase to be called the Foreign Language Phase) began in the late1950s, stimulated by the Soviet launching of the satellite Sputnik and growing Americananxiety abo ut national security and the poo r preparedness of the nation in foreign languages.Th e field of linguistics provided the technical backbone to this movem ent, an d gave rise toa field known a s contrastive analysis, in which careful compa risons were m ade between thegramm atical struc tures of the native language of the learners and the foreign language, withthe goal of targeting instruction to those are as where difficulties ar e predicted (Lado, 1964).This movement resulted in the proliferation of the audio-lingual method for teachingforeign languages focused on drilling problematic grammatical patterns, often aided bylanguage laboratory exercises. Th e audio-lingual method, as Rivers (1964) noted in herincisive review of the field, was thoroughly grounded in the dominant psychological theoryof learning which emphasized the formation and interaction of learned habits. As thistheory of learning became less acceptable with the advent of cognitive theories starting inthe 1960s, the audiolingual movement ground to a halt.


3/73

3T h e third phase (to be called the Language Acquisition Phase) cam e in the 1960s

on the heels of a revolution in our und erstand ing of child language acquisition, which in turnwas influenced by a revolution in theor etical linguistics that sta rted with Chomskys (1957)Syntactic Structures. Applied linguists such as Pit Corder (1967) who were trying tound erstand th e sources of learn er er rors in second language learning foun d many similaritieswith those being reported in the child language literature both seemed to be driven by anat tempt to make sense of the target language, rather than being slaves to the nativelanguage as contrastive analysis might lead one to believe. Since this time, although theresearch in first and second language acquisition have ten ded to be conducted by differentgroups of researchers in different academic departments, the questions have come to beintertwined: wh ether the capacity to learn language is best defined as specific to languageor reflects general learning mechanisms, and whether there are maturational constraints onlanguage learning capacity.

Th e fourth phase to be called the C anadian Imm ersion Ph ase) also started in the1960s, but stem med from innovations in French imm ersion education in Canada. Frenchimmersion programs were a radical way of responding to the needs of the English-speakersof a bilingual Montreal who wanted to ensure that their children had access to the benefitsof bilingualism. In thes e program s, native English-speaking child ren were providedinstruction exclusively in French from their first day of school (Lambert Tucker, 1972).This innovation has beco me very popular in Canada, even in its English-dominant areas, andhas gen erat ed considerable research o n its effectiveness and the conditions under which the


4/73

4native language and academ ic achievement ar e maintained, but with the benefit of becominghighly proficient in a second language.

Finally, beginning in the 1980s, there has been another wave (to be called theLanguag e M inority Phase) of interest in bilingualism and secon d language acquisition dueto educational needs. This time, however, the needs stem fro m changes in the immigrantpop ulation, especially in industrialized nations OECD, 1989; Padilla, 1990). T his movem enthas resulted in intense scrutiny of language and educational policies in many developednations of the world, including the the qu estion of how much em phasis is to be placed onthe native language of the students. In comparison to the additive bilingual policiespursued in French immersion programs whereby second language learning is an incremen tto native language development, the policies pursued in general for immigrant languageminority students ar e subtractive in nature. In the United States, for example, most of theofficial program evaluation research has focused on whether the students in bilingualeduca tion program s ar e learning English fast enough, and u nder what conditions this processcan be optimized; little concern has been shown to their developing bilingual competence.Many basic researchers, on the other hand, have tended to focus on what happens to thenative language an d e thnic com munity, often within disciplinary frameworks (e.g., Fishman

Ge rtn er, 1985; Ex tra Verhoeven, 1993).Given this rather complex history of the field of bilingualism and second language

acquisition, we have decided to address t he task o f this review by decomposing th e field intokey intellectual tensions that we believe capture its character. The se tensions reflecthistorical developments and provide a window into the dynamic process of the research


5/73

enterprise as it changes over time. Thes e key tensions are: empiricism vs. nativism;linguistics vs. psychology; psycholinguistics vs. sociolinguistics; cognitive skills vs. wholelanguage; elite vs. folk bilingualism; basic vs. applied re search; an d theory vs. data . Byoffering this list of tensions, we do not pretend to exhaust the issues that drive the field, nordo we claim that they a re independent of on e anothe r. It is from the perspective of theseintellectual tensions tha t one gains the distance to appreciate what is innovative an d excitingin empirical research on second language acquisition and bilingualism.

Emuiricism vs. NativismThis classic tension between whether learning is driven by experience or innate

knowledge is obviously not unique to the field of second language acquisition. But it is adeeply engrained issue in the field, underscoring how closely the study of bilingualism andsecond language acquisition is to a very central problem in human learning anddevelopment.Bilingualism and Intelligence

The earliest manifestation of the tension between empiricism and nativism cameduring the Psychometrics Phase, when there was concern that immigrant children werehandicapped in their language growth, as measured by standardized tests of languagedevelopment, because of their bilingualism (e.g., Smith, 1931). Such a handicap would bepredicted, as the empiricists did, if one assumes that there is a d irect relationship betweentime spent on learning and the degre e of learning accomplished, for bilinguals would haveto divide their learning energy between two languages. This led to the advice commonly


6/73

6given to immigrant parents not to use the native language at ho me b ecause it might lead tolinguistic retardation (Thompson, 1952).

Radical nativists during this early period went even further and concluded that thepoor performance of immigrant children on standardized language tests was not caused bytheir diffused learning experience, but by genetic inferiority (e.g., Goodenough, 1926; seeH a h t a , 1986 for a review).

This line of controversy between proponents of empiricist and nativist views oflearning has continued to the prese nt day, but in two diffe rent forms. On the on e hand, thenature-nurture question over intelligence continues with minimal attention to the questionof the role of bilingualism (Jensen, 1980). On the other hand, the issue of the relationshipof bilingualism and cognitive development continues as an empirical question, but withoutmaking strong assumptions about the nature of either intelligence or bilingualism. Thisliterature generally indicates a mild, positive effect of bilingualism, especially in the areasrelated to metalinguistic awareness (e.g., Bialystok, 1988).The Linguistic Abstractness Arrmment

Another instance of the empiricism-nativism tension can be found in assumptionsabo ut what it is that is learned in second language acquisition. This problem has receivedattention especially in the research during the Language Acquisition phase which wentthrough a period of intense appreciation for the uniqueness of language. This view oflanguage (a characteristic that Chomsky [1965] called task-specificity, and the p hilosopherJerry Fodor [1983] calls modularity) is based on logical arguments based on theoretical


7/73

7linguistics, as well as behavioral an d neurological ev idence suggesting that h um ans processlanguage in specialized ways.

The logical linguistic argument is covered in greater detail in the next section onlinguistics and psychology. Briefly, the gist is tha t all ma tur e speak ers of a language haveknowledge abo ut their language that is highly abstract. This can be proven by showing tha tpeople are able to distinguish between gram matical and ungramm atical sentences that differonly along this abstra ct dimension. Th e logical argume nt is that this ability could not havebee n induced from simple exposure to the surface patterns of the language. Th e only wayin which they could have gotten to the present state is if they had a critical a prioriknowledge a bou t language. Put together with the fact that children display ma tureknowledge of most aspects of language by age 5, the conclusion is that many aspects oflanguage m ust be inn ate. Fo r second language acquisition, the extension is that if learnerssuccessfully m ak e similar distinctions, they must also do so following their inn ate know ledge.Laneuaee as Specialized Behavior

T he re is a variety of evidence that strongly suggests that language is a spec ial activityunrelated to other human abilities. One such such piece of evidence comes from researchon the perception of speech sounds. Studies of infant speech perception since Eimas1971) have suggested that very young infants actively segment sounds into phonemic

categories even when the acoustic properties of these sounds vary along continuousdimensions. Re cen t comparative research of infants exposed to Swedish and English showe dthat these infants had already segmented the vowel continuum in ways that correspondedto the language of exposure (Kuhl etal.992).


8/73

8For second language learners, the evidence suggests that th e phon emic categories of

the native language serve as a starting point for speech perception of the second language,but that adjustments are made in the course of second language acquisition. Williams1974) m ad e good use of a difference between Spanish and English in the voice on set timeVOT) speech par am eter that distinguishes, for example, between the sounds /ba / and /pa/.

The VOT is the time between the initial release of air from the lips and vibration of thevocal cords. For English, native speakers categorize sounds at VOT less than 25milliseconds a s /b a/ and anything longer as /pa/. For Spanish, the boundary is at about10 milliseconds. Williams found that P uerto R ican native speake rs of Spanish who werelearning English shifted from the Spanish boundary to the English boundary both as afunction of length of exposure to English an d to the initial ag e a t which they were exposedto the second language. Eve n though the boundary shifted, i t is important to note that thesubjects preserved the categorical natur e of their perception.

O th er clear evidence for the special nature of language can he found in the variablesthat seem to affect the course of language development. Brown (1973) found no effects ofparental frequency, reinforcement, or correction on the course of grammatical development.Recently, Marcus 1993) found that acquisition of regular and irregular past tensemarking on verbs is remarkably similar between children learning English and German,despite the fact that irregular verbs are far more common in English, while regular verbsar e far more comm on in Ge rma n. Such evidence indicates that humans are highly prepare dto learn language, and that learning is relatively immune to variations in input language.


9/73

9Interestingly, there is some evidence to indicate that input frequency is more

important in second language acquisition than might be supposed from first languageacquisition. Larsen-Freeman 1976) found that the relative frequencies of grammaticalmorphemes (such as noun and verb inflections, prepositions, the verb tobend articles)successfully predicted the overall order in which they were mastered by second languagelearners across a wide range of ages and native language backgrounds. First languagelearners of English master these same structures in a different order, which is not relatedto the input frequency, but rather is predicted by their syntactic and semantic complexity(Brown, 1973). T h e greater sensitivity of second language learn ers to input frequency,however, do es not explain persistent differences between second language learning by nativespeake rs of differe nt languages, such as the great difficulty that native sp eakers of Japanesehave of the English article system, despite the very frequent occurrence of these forms inthe English language ( H a h t a , 1983).Ape Constraints on Second Lanpuaee Acquisition

One area that has seen considerable empirical activity is the question of whetherthere are age constraints on language acquisition. This question is usually viewed from theperspective of neurological bases of language, a perspective that was raised to prominenceby Penfield and Ro ber ts (1959) when they reporte d dram atic results of studies in which theystimulated specific areas of the brain in patients during surgery, and found correlationsbetwee n ar eas that were stimulated and language-specific functions. Thes e hardwareexplanations of language, put together with t h e obvious plasticity of the brain in childhood,were used to argue for foreign language education in the elementary grades during the


10/73

10Foreig n Lan guag e phas e of the field. The hypothesis about the age constraints on secondlangu age acqu isition really to ok sh ape with the publication of Lenneb ergs (1967) BiologicalFou ndatio ns of Langu age. Len nebe rg brought togeth er the Chomsky-inspired logicallinguistic arguments about the necessary abstractness and complexity of language with areview of the evidence on recovery from childhood traumatic aphasia and other disordersthat affect language developm ent. T he amassed evidence indicated that the potential forlanguage learning existed through childhood but disappeared at aroun d puberty. Lennebergsuggested that the period between birth and puberty constituted a critical period perha psresembling other well-documented cases of critical periods in learning such as imprintingin greylag geese, which is time-bounded and highly prepared learning for specificinformation (Lorenz, 1958).

The idea of extending the critical period to second language learning has beensubjected to empirical test by a nu mb er of researchers. T he earliest convincingdemonstration of an age effect was reported by Oyama (1976) who rated the pronunciationof Italians who had im migrated to the United States at various ages. She found a strongnegative effect of age of arrival, and no effect for length of exposure once age of arrival wascontrolled. Patkowski (1980) examined the syntactic ability of adult learners who hadlearned English before or after puberty, and found differences in favor of pre-pubescentlearner s. M ore recently, John son and Newport (1989) looked at the ability to ma kegrammaticality judgments of English sentences by native speakers of Chinese and Koreanwho had lear ned English at ages ranging from 3 to 39. The ir da ta suggest the following: (1)the re is a decline in accuracy across the age span that begins as early as age 5and continues


11/73

11through adulthood; 2) ther e is greater individual variation am ong subjects who had arrivedafter puberty; and 3) the decline in performance is steeper among those who had arrivedbefore puberty than those who had arrived after puberty.

It should be noted that all of these studies have considerable difficulty in controllingfor length of exposure to English when they look at age of arrival, especially because theage of subject at the time of testing can become a factor for younger and older subjects.Since these th ree factors are necessarily related (i.e., cu rrent ag e is the sum of age of arrivaland length of exposure), the designs of these studies a re never fully satisfactory.

Aside from the inherent empirical blemishes that mark these studies, the resultsindicate tha t the question is much more complicated than it app ear ed a t first blush. O necomplication is that the a ge-related decline is bette r cha racterized as a monotonic in nature,rath er than categorical. Th e ability to learn a second language does not seem to suddenlydisapp ear as might be expected of a n ability that is bounded by a critical period, such as th edevelopment of the visual system (Hubel, 1988). Even proponents of the biological viewreadily concede that second language learning might be better described as being affectedby a sensitive period . Pe rhaps it is not the withering away of a specific inna te capacity.However, if the constraint on learning is really just a decreasing ability that is evidently notspecific to language, then the explanation loses much of its original appe al about revealingsomething about the nature of language.

A second complication arises from the fact that the re a re many similarities betweenchild an d adul t second language acquisition. For example, the types of grammatical errorsas well as the o rde r of acquisition of gram matical morphe me s are not different with respect


12/73

12to age. Indee d, we ar e not awa re of any repor ted qualitative differences in the process ofsecond language acquisition in adults and children. To paraph rase the sage observation: ifthey look like one another, quack like one another, and walk like one another, well thenthey ar e probably learning to talk in the second language in the same way. W hat is lackingis any specifica tion of what might be lost in terms of specific language le arning ability as on egets older. Indeed, White and Genesee 1992) recently demonstrated that many advancedadult second language learners master highly abstract gramm atical patterns that a re thoug htto be innate properties of the human language capacity.

A final theoretical complication arises from a consideration of whether the processof second language acquisition is at all relevant to the question of a critical or sensitiveperiod for primaly language acquisition. All second language learners, by definition, havealready acquired a first language. So the notion of a critical period may be applied tosecond language in different ways, depending on ones theory of what it is that is acquiredduring second language acquisition.

Under one view, one might suppose that second language acquisition is like re-inventing the w heel, and thus it is a re-run of what happens during first language acquisition.Such a view might be held by theorists who see language acquisition as an essentiallyconc rete, close-to-the-surface event. This inte rpretatio n of language acquisition would alsobe g enerally com patible with empiricist accounts of language acquisition (e.& Mo erk, 1983).Extending this scenario, one might argue that the tools necessary to invent the wheel mightonly be available during a critical period, and thus second language acquisition woulddecline accordingly.


13/73

13Under another view, one might think of second language acquisition as something

only moderately incremental to what one already has accomplished in first languageacquisition. Ma ybe it is m ore like recycling -- learning new terminology for old concepts,like a Californian who moves to Boston and learns that regular coffee comes served withcream and sugar. Such a view can be derived from those who ad vocate language acquisitionto be a highly abstract process consistent with rationalist accou nts of the process. Un de rthis view, because the language acquisition ability has already be en capacitated in the co urseof first language acquisition, there may be no age implications for second languageacquisition, even if there were indeed a critical period for first language acquisition (seeNewport, 1991). Curiously, however, the critical period hypothesis has found its most arde ntsupp ort am on g research ers with strong rationalist orientation s (e.g., Johnson Newport,1989; Long, 1985; Patkowski, 1980).

In sum, there are a number of complications in interpreting the data on age effectson second language acquisition, especially as they might bear on the question of empiricismvs. rationalism. It may well be that until there is bette r elabo ration of the theoreticalpredictions, the primary reason for investigating the question is simple curiosity and perha psthe need to answer educational policy questions such as the optimum age to begin foreignlanguage instruction.

Linmistics vs PsychologyThe tension between linguistics and psychology has existed since the Language

Acquisition phase of research that began in the 1960s. Prior to the Chomskyan revolu-tion in linguistics, there was a fundamental compatibility between linguistics and


14/73

14psychology, bo th of which were solidly empiricist in their orientation . Ev en durin g theearly stages of paradigm change in linguistics, psychologists were enthralled with thepossibilities of the new an d mo re powerful linguistics. But efforts to test predictionsfrom linguistic theory failed miserably and psychologists an d psycholinguists becam e disil-lusioned with the new linguistics. As the distinction between competence andperformance became more clearly understood on each side, it became apparent thatlinguists and psychologists were in pursuit of two different Holy Grails. T he linguistswere concerned with the linguistic intuitions of an idealized speaker; psychologists wereconce rned with the behavior of their all-too-real subjects. Nonetheless, research ersinvestigating bilingualism and second language acquisition have drawn on both linguisticsand psychology.The Lineuistic Perspective

On the linguistic side, the dominant current influence is Chomskyan generativegram mar . Th is appro ach assumes that the first-language learner comes to the acquisitiontask with innate, specifically linguistic, knowledge, or Universal G ram mar . Th e claim isthat certain principles of the human mind are, to a degree, biologically determined andspecialized for language learning. As Chomsky put it: Universal gram ma r is taken tobe the set of properties, conditions, or whatever, that constitute the initial sta te of thelanguage learner, hence the basis on which knowledge of language develops (1980, 69).These abstract and linguistically significant principles are thought to underlie all naturallanguages and comprise the essential faculty for language with which all individuals arein general uniformly and equally endowed.


15/73

15According to this theory, the ability to acquire a human language is genetically

determined. T he theory postulates that th e child faces a projection problem i n thatthe language learning task must be accomplished with deficient input. The only way toexplain how children succeed is to assume that they approach the task endowed with aUniversal Gr am m ar that comprises a rich set of innate principles that govern theemergence of language. -Universal Grammar constrains the hypotheses that childrenmake and the childs language environment determines which principles of the UniversalGrammar will be accessed. Acquisition involves setting the pa ram eters in a specific wayto reflect the properties of the gra mm ar of a particular language.

As a linguistic theory, Universal Grammar does not concern itself withsecond-language acquisition. Nonetheless, a num ber of second language researchershave applied the theory in their work, motivated by the need for a sufficientlysophisticated linguistic theory to describe the complex structural characteristics of thelearners language. Universal Gram ma r, its propo nents argue , provides a detailedlinguistic theory to account for second-language phenomena.

Second-language learners are thought to face the same projection problem(White, 1985)--that is, they, like first-language learners, have to work out a complexgrammar on the basis of deficient data. The learners grammatical knowledge cannot beexplained by the input alone. Felix 1984) listed three limitations to such an explanation.First, some structures are so rare and marginal that it would not be possible for learnersto obta in sufficient exposure to them. Second, incorrect hypotheses require negativefeedback (correction, identification of errors, etc.) if they are to be discarded, but such


16/73

16feedb ack usually does not occur. Finally, the rules of any gra mm ar ar e highly abstractand so d o not reflect the surface properties of the language.

According to the theory, Universal G ram ma r involves a set of principles withcertain param eters. These param eters remain ope n until they are set by experiencewith the environment. For Chomsky, language acquisition is not so much a problem ofacquiring grammatical rules, but rather a process whereby the learner discovers how theprinciples operate in the target language and what parameter values apply. The grammarof a language is the set of values it assigns to various parameters. As Chomsky put it,Ex perienc e is require d to set the switches. O nc e they are set, the system functions(1984, 25).

An oft-cited example of such a parameter is the pro-drop parameter, whichspecifies that languages vary with regard to whether they allow the deletion of pronounsin subject position, together with related phenomena such as inversion of subject andverb. English does not have pro-drop because a subject is required for every sentencean d the subject canno t be inverted with the verb in declarative sentences. This is nottrue of Spanish, however, which, as a pro-drop language, allows empty subjects and sub-ject-verb inversion in declarative sentences.

Another example is the principle of adjacency, according to which noun phrasesmust be next to the verb or preposition that gives them case. Hen ce in English anadverb cannot intervene between a verb and its direct object. Sentences such as Maryat e quickly her dinner a re not allowed, whereas in French such sentences a re permitted:Marie a mange rapidement le diner (White, 1989). The French option is assumed to


17/73

17be set for the child learning French as a first language on the basis of positiveevidence in the form of such sentences.

The crucial issue in much linguistically-based second-language research is how theparameters that have been set in the first language need to be reset or readjusted for thesecond language. So me investigators argue that Universal G ram m ar principles a re fullyavailable to the learner and the task of second-language learning involves resetting thefirst-language parameters in line with those of th e second. T he relative similarity ordifference of specific parameters across the learners first and second languages will thenconstrain this resetting process (Flynn, 1984).

Ot he r researchers m aintain that Universal G ra m m ar principles ar e available, butthat they interact with, and are highly constrained by other factors--e.g., cognitivestrategies, proce ssing considerations (Bley-Vrom an, Felix, Ioup, 1988). Bley-Vroman(1990) has argued that the child language learner possess a language acquisition systemthat contains the following two subcomponents: 1) a definition of possible grammar: aUniversal Grammar, and (2) a way of arriving at a grammar based on available data: aLearning Procedure (or set of procedures). Th e adult second-language learner,Bley-Vroman argued, does not have access to Universal Grammar, but instead constructsa kind of surrogate for Universal G ram m ar from knowledge of the native language. Thisknowledge, plus general cognitive abilities that enable adult learners to deal withabstract formal systems, enables adult learners to acquire imperfect knowledge of targetlanguages.T he Psvchological Perspective


18/73

18The Bley-Vroman position accords well with what most psychologists and

psycholinguistics would maintain abou t second-language learning. T he con tention is thatther e may b e som e access to Universal G ram m ar through knowledge of the firstlanguage and adult second-language learning is the result of this knowledge and generalcognitive abilities. Fr om this perspective second-langu age learning, like all adultcognitive problem solving, is god-oriented, involves analysis, hypothesis formation andtesting, an d analogy. T he learner is thought to proceed with practice from atten-tion-demanding, controlled processing to more automatic processing.

Second-language theorists and researchers have drawn from cognitivepsychological work in general problem solving, schema theory, and production models.For example, OM alle y and Cha mo t (1989) used Andersons (1983) notions ofdeclarative and procedural knowledge to express the manner in which information aboutlangu age is represe nted in mem ory. Kenned y (1988) ma de a similar distinction based onGag nes (1985) information-processing model. Both approaches stress the differencebetween knowing concepts, propositions, and schemata (declarative knowledge) andknowing how to perceive and classify patterns and how to follow specific steps until anend goal is reached (procedural knowledge). A related distinction is made by Bialystok(1Y81), who uses the concepts analysis and control to distinguish between thecognitive skills involved in the learners linguistic knowledge from skills involved incontrol of processing.

Procedural knowledge is thought to be acquired through extensive practice andfeedback and, once learned, is more easily activated in memory than declarative


19/73

19knowledge. Initially, the learning of proce dures req uires conscious attention, but as thelearner becomes more and more skilled at a task, less conscious work is required.McLaughlin, Rossman, and McLeod (1983) used Shiffrin and Schneiders (1977)distinction between controlled and automatic processing to account for this progressionfrom a more cognitively demanding to an autonomous stage of learning.

As performance becomes m ore automatic, elements of the task bec ome unitized(Gag ne 1985). Th at is, there is an integration of skills into larger and more efficientunits. Th is unitization process involves a progressive reorganization of information as a nincreasing num ber of procedures be com e automatic and controlled processes are freedfor new tasks. In the case of second-language learning there is constant restructuring aslearners simplify, unify, and gain increasing control over the procedures involved inprocessing the language (McLaughlin 1990).

Another area of cognitive psychology that has impacted on second-languageresear ch c oncerns expert-novice systems. Th e literature suggests that experts usedifferent informationprocessing strategies than do more novice learners, Differencesbetween experts and novices have been found in research on learning mechanisms inphysics, arithmetic, algebra, geometry, computer programming, and chess. For the mostpart, research indicates that experts restructure the elements of a learning task intoabstract schemata that are not available to novices, who focus principally on the surfaceelements of a task. Thus experts replace complex sub-elements with single schemata thatallow more abstract processing.


20/73

20In th e realm of language learning, experts ar e those individuals who have learn ed

a number of languages. Th ere is considerable anecd otal evidence (though littleempirical research) that suggests that once a person has learned a few languages,subsequent language learning is greatly facilitated. In a study using a miniature artificiallinguistic system, Nayak, Hansen, Kreuger, and McLaughlin (1990) found thatmultilingual subjects showed more flexibility in switching strategies than did monolingualsubjects. This is consistent with the research of Nation and McLaughlin (1986), whofound that m ultilingual subjects were able to avoid perseveration errors m ore than wereot he r subjects in their experiment. Similarly, Ram sey (1980) reported that multilingualsubjects demonstrated greater flexibility in restructuring mental frameworks than didmonolingual subjects. Thus there is some evidence to suggest that more expert languagelearners show greater plasticity in restructuring their internal representations of the rulesgoverning linguistic input.

In oth er research within an expert systems framework, Faerch an d Kasper(1983), McG roarty (1989), Oxford (1986), and O M alle y and Chamot (1989) haveattem pted to specify strategies that good language learners use. Th e ultimate goal ofmuch of this research has been to expand and refine the repertoire of strategies of poorlearners so that they may benefit from strategies used to good effect by expertlearners. Wenden (1987) noted that intervention research on training learners to usecognitive strategies in o ther skill areas has dem onstrated that the appropriate use ofstrategies in a variety of situations requires metacognition. It is not enough for learners


21/73

21to be trained to use a particular strategy, they must also understand the significance ofthe strategy and b e able to mon itor an d evaluate its use.

Rese arch on learn er strategies and the cognitive ap proac h generally fit well withthe needs of classroom teachers. Con temp orary linguistic theory with its concern withabstract principles is arcane and inaccessible to teachers. Most teachers are m orecomfortable with an approach that sees language learning as a n active process ofinternalization, through practice, of various rules and representations. Teach ers are athome with a theory that is concerned with learning, production, and communicationstrategies. Nonetheless, the insights of contemporary linguistics have had a powerfuleffect on thinking ab out how languages are acquired. It remains to be seen whetherthese insights will be useful to classroom teachers.

Psvcholineuistics vs. SociolinguisticsThe psycholinguistic perspective dates back to the Language Acquisition phase of

the 1960s and 1970s, especially early work on error analysis and the morphemestudies (Ellis, 1985; Hatch, 1983). H er e we will focus on two contemporarymanifestations of the psycholinguistic approach--work on the Competition Model andwork using m iniature artificial languages. Th e sociolinguistic perspective stresses thesocial nature of language and its use in varying contexts, and in many ways defined itselfas a rea ctio n to the dom inanc e of psycholinguistics. According to this view, thepsycholinguistic experiment is only one of many possible contexts in which language isused, an d, consequently, does not tell the whole story. Sociolinguists argue th at how aperson uses the language will depend on what is understood to be appropriate in a given


22/73

22social setting. Sociolinguists see linguistic knowledge as situated not in the individualpsyche, but in a groups collective linguistic norms.T he Psvcholineuistic Perspective

One of the ways to study how individuals learn a second language is to examinehow the ways in which language is used affects how grammatical forms are acquired.This is the question of form-function mappings that is central to a popular researchparadig m d eveloped by Bates, MacW hinney, and their colleagues (Bates andMa cW hinne y 1982; MacW hinney, Bates, and Klieg1 1984). Th eir m odel is called th eCornuetition Model, and it assumes that the structural properties of language arerepresented not in terms of rules (explicit or otherwise), but rather by mappings betweensurface linguistic forms and underlying functions. In any given language, a particularinstantiation of a form-function mapping is assigned a weight depending on how oftenand how reliably a given form is used to perform a given function. Th e information alearner uses to decide which function is meant to be expressed by a particular form isreferred to as a cue, and cues vary in their reliability and availability.

In the second-language context, the task facing the learner is to discover howspecific form s ar e used fo r specific functions in the new language. Typical experimentstesting this mo del u se bilingual subjects in a within-subjects, cross-language design.Subjects a re given a sentence interpretation task designed to produce competitionsamong a restricted set of grammatical cues (e.g., word order, animacy relations,subject-v erb agreem en t, and case inflections). T he task is to say which noun is the agentof the action. For example, subjects may hear, The apple is eating the man. In this


23/73

23example the canonical SVO pattern of English is in competition with the animacy cue.Studies comparing different groups of bilingual subjects suggest that during the initialperiod of second-language acquisition, subjects rely on the transfer of first-languagestrategies to aid sentence competition. Thus Italian an d Japa nese subjects learningEnglish tend to rely on the animacy cue rather than using the word order cue, which isthe processing strategy employed by native English speakers (Gass 1987; Harrington1987).

From experiments using various combinations of competing cues advocates of thisapproach argue that it is possible to examine which cues are most important in sentencecomprehension in a language, and how certain cues come under strategic control asfluency increases. This would have imp ortant pedagogical implications, as it is these cuesthat would be critical for learners coming from particular background languages andteachers could attend to such cues in teaching these learners.

Recently, however, Gibson (1992) has raised the question of the adequacy of thedefinitions of key concepts in the competition model, especially the notion of cue andcue reliability and validity. Gib son argues that insufficient attentio n has been paid tohow cues ar e identified an d used by learners. H e also argues that the experimentalparadigm used to support the model may tap into different processing mechanisms--specifically, if a given stimulus string is grammatical in a language, it may be processeddifferently than are ungrammatical strings.

One wonders, moreover, about the ecological validity of an experimentalprocedure in which subjects have to make decisions about sentences that are as deviant


24/73

24as The apple is eating the man is in English (McLaughlin Harrington, 1990).Perhaps subjects are not processing such sentences as they would in actualcomm unicative situations, but are settling on a particular problem-solving strategy to getthem through the many judgments of this nature they have to make.disturbing finding from the competition model is that there is a great deal of L1 transfer,whereas most naturalistic research reveals surprisingly little L1 transfer (McLaughlin,1986). As MacWhinney and Bates (1989) noted, it is important for testing the adequacyof the Competition Model to develop more on-line measures of sentence processing,both with respect to comprehension and production.

O ne particularly

Another psycholinguistic research method that is widely used in current second-language research involves the use of miniature artificial languages (MALs). Becausenatural language learning takes place in a n environment where it is impossible to controlthe input the learner receives, researchers have not been able to specify as accuratelyand exhaustively as possible those features of the environment that causally influencelearning. By clearly specifying the input and output characteristics of the languageacquisition task through the use of MALs, it becomes possible to make systematicinferences about the structures and processes within the organism that make learning alanguage possible.

Artificial linguistic systems resemble natural languages in that they contain a setof verba l symbols and a set of rules for combining these symbols into sentences . Likenatural languages, the set of rules can specify class membership, order, andco-occurrence constraints on the linguistic structure of the artificial language. Unlike


25/73

25natural languages this set of rules is fairly limited in scope, thus making it possible toobserve a language-learning situation wherein various language features can be studiedin isolation from the complex interactions found in natural systems. It is this ability tomanipulate systematically all features that might influence language learning mechanismsthat makes the study of artificial linguistic systems an important tool in psycholinguisticresearch (Moeser, 1977; Morgan and Newport, 1981).

Subjects in an experiment using an artificial linguistic system are exposed to alimited subset of permissible strings. The question of interest is whether they can applygeneralizations derived from the learned subset to novel strings and if so, what is thenature of these generalizations. In a number of pap ers based on M A L research, Reberand his associates (e.g., Re be r, 1976; Re ber Allen, 1978 ) have argued for what theyhave called implicit learning. In this research, subjects were exposed to finite-stategrammars, made up of letter strings, and were found to be significantly accurate whenthey subsequently have an opportunity to judge th e grammaticality of novel grammaticaland non-g ramm atical strings. Because subjects seem ed to be learning these rules withoutbeing able to articulate their knowledge, Reber concluded that the learning was implicitand unconscious.

This conclusion has been challenged by Dulany and his associates (Dulany,Carlson, Dewey, 1984), who questioned the deg ree to which the knowledge of subjectsin these experiments is properly characterized as abstract and the degree to which it istruly unconsciously held. Th eir research indicated that subjects developed personal andidiosyncratic sets of rules and that these sets of rules correlated with the finite-state


26/73

26grammar in the sense that both sets of rules resulted in the same grammaticalclassifications. T h e subjects idiosyncratic rules were of im perfect validity and of limitedscope, but we re accessible to consciousness. This is an im portant pedago gical point, asthe re ar e som e (e.g., K rashen, 1982) who argu e that second-language learning is largelya n unconsciou s process, and others (e.g., Schm idt, 1990) who m aintain that wh at islearned has to be noticed.

T he critics of research with MALs question whether the abilities recruited in suchexperiments ar e the same as those engaged in natur al language learning, specificallythose recruited when a child is acquiring a first language. McL aughlin (1980) has arguedtha t b eca use subjects learning artificial linguistic systems are not linguistically naive,research on artificial linguistic systems, rather than being viewed as a method forund erstand ing the nature of first-language acquisition, is more suited to further ourknowledge of the process of second-language learning. In fact, as some research suggests(N atio n McLaughlin, 1986; Nayak, Han sen, Krueger, McLaughlin, 1990), theamount of exposure that subjects have had to various natural language systems may be acritical factor in how they go about learning a new linguistic system.

Rese arch on the Com petition M odel and on miniature artificial languages areinstanc es of the psycholinguistic approach , in that studies ar e typically tightly controlledand tak e place in a laboratory setting. Sociolinguists argue that this is not the way toobtain good data on the learners abilities. On e has to go beyond the lab oratory andexamine systematically how language is used in various settings.T h e Sociolinguistic Perspective


27/73

27O n e of the axioms of the sociolinguistic perspective is that speakers have several

styles that they use according to the demands of the social context. Th e vernacularstyle is associated with informal, everyday speech. It requires minimal atte ntion an d is atthe opposite end of the continuum from language used in forma l situations where speechis highly monitored.

T he classic researc h is Labovs (1970) study of th e sp eech pa tterns of NewYorkers. He sampled speech styles that ranged from casual speech to carefullymonitored speech. It was possible to characterize different styles of speaking in terms ofthe variable use of such sounds as the b n thing. In more casual speech, he found agreater use of non-prestige variants of b uch as /t/.

Similar researc h with second-language learne rs (Dickerson, 1975; Schmidt, 1977;Tarone, 1983) indicates that language learners also show contextual variability accordingto linguistic setting. Th ere is a continuum of usage with on e en d represen ted by theverna cular stvle, which is seen when the lea rner is not attending to speech. At the oth eren d is the care ful style, which involves close atten tion to spe ech. Ta ron e called this theinterlanguage continuum. It reflects the fact that differing degrees of attention arerequired for different performance tasks.

Tarone (1983) argued that second-language learners have variable capability andthat this is a better description theoretically than a Chomskyan com petenc e model. Thisvariable capability underlies all interlanguage behavior and is due, ultimately, to thedifferential atten tion to language in different tasks. Ellis (1 98 3, like Tar one , maintainedthat interlanguage output is best described by a continuum ranging from planned


28/73

28discourse to unplanned discourse, but he differs from Tarone in distinguishing non-systematic an d systematic variability. Th e first includes free variation orunpredictable variability; the second is similar to the variability described by Labov andTarone.

Th e views of Tarone and Ellis have been criticized on theoretical grounds byGregg 1990) who argued that variationkt models of second-language acquisition areinherently incapable of accounting for the phenomena they are invoked to explain. nadequate theory, such as Universal Grammar, is a model of linguistic competence basedon principles and parameters. For Gregg, variation exists, but it is not interesting;indeed, it is not something that a theory of acquisition nee d be conce rned with. It is apesky mosquito that is best ignored.

Preston 1993) took a different tact. He is uncomfortable wi th free orunpre dictable variation and argues that all variation is systematic. For him, the task forsociolinguistics is to determine the probabilistic weightings of influences o n varying formsthat oc cur in language usage. Prestons approach require s multivariate analysis of factorsthat affect the occurrence of one form or another. This work is especially promising as away of linking sociolinguistic and psycholinguistic concerns. As Preston argued, itprovides quantifiable features that make a storage-production psycholinguistic modelpossible--at least in principle. Thus, to some extent there is a convergence of interests,as both psycholinguists and sociolinguists such as Preston are concerned with determiningwhat are the important features influencing language use and how these features getadded in a predictive equation.


29/73

29Other lines of sociolinguistic research examine the social nature of language

acquisition and the roles that other individuals play in the language learning process.Many immigrant children come from communities where horizontal peer-to-peerinteraction pa tterns a re much m ore com mon than vertical adult-child patterns (Brent-Palmer, 1979). Furthermore, the experience of many immigrant children in the home isthat learning occurs through observation and nonverbal means, rather than throughlanguage. There is none of the information testing through questions that characterizesthe teaching-learning process in American middle-class homes (Edelsky, et al., 1983).

Such observations have important pedagogical implications for the childs schoolachievement (Padilla, 1990). Public education in Europe and the United States is gearedto the middle-class child, and most teachers incorporate middle-class values in theirlessons. M any of the problems of immigrant working-class children in the public schoolsystem derive from the discontinuity between their values and communicative norms andthose of the school (Brent-Pa lmer, 1979; Delp it, 1986; Extra Verhoeven, 1993).

Cognitive Skills vs. Whole LanguavePart of the Language Acquisition Phase of the 1960s and 1970s involved a

rejection of traditional behavioristic notions of human psychology and the adoption ofthe new cognitive psychology. This approach has become the dominant paradigm inpresent-day psychology, with implications to second-language research as well. It is notwithout its detractors, however, many of whom are part of what has become known asthe whole language movement.


30/73

30For researchers working within the framework of contemporary cognitive

psychology, second-language learning is one of many complex cognitive skills and islearned in a similar fashion. Initially the learner is overwhelm ed by the sheer num ber oftasks that have to be performed in speaking a second language--correct articulation ofsounds, correct lexical choice, correct grammar. But with practice, it become s eas ier topronounce the sounds of the language correctly and more attention can be given tocorrect word choice and grammar.

For advocates of the whole language approach, this view of learning is overlysimplistic an d lead s to a fragmentation of the learning process into discrete, isolatedtasks. It leads to a dead ening pedagogy that focuses on skills rather than engaging thelearner.minority childr en learning E nglish is Americas schools. In the whole language appro ach,meaning is essential and the learning of skills is subordinated to the task of makinglearning meaningful to the student.Th e Cognitive Skills Am roa ch

Focusing on skills is especially detrimen tal t o t h e education of language

In what follows, we will focus on a particular task, that of learning to read in asecond language. Rea ding can be viewed as a cognitive skill: indeed, as the mostcomplex and difficult of all the cognitive skills that the child must master in school. Thechild who accurately and efficiently translates a string of printed letters into meaningfulcommunication may appear to be accomplishing that task with little mental effort. Infact, from a cognitive skills perspective, the child is engaging in complex interactiveprocesses that are dependent on multiple subskills and an enormous amount of coded


31/73

31information. T he fluent read er must have autom ated language skills, intact visual andauditory memory, the ability to associate and integrate intra- and intermodal stimuli, andthe ability to abstract and generalize pattern ed or rule-generated information (Vellutino

Scanlon, 1982).More specifically, to become an accomplished reader, the child must have

mastered three important tasks that are developmentally linked to each other. Only afterthe child has automated word-decoding operations, is it possible to acquire moresophisticated reading and comprehension skills. Similarly, the automation ofword-decoding skills is dependent on mastery of symbol-sound correspondence rules.

Research by cognitive psychologists with good and poor readers has indicated thatcertain components of the reading process are more advanced in good than in poorreaders. Specifically, good read ers ar e distinguished from po or re aders by:

Bottom-up skills:superior ability to store information in short-termmemory.superiority in visual discrimination.superior phonological analysis skills.superior attentional abilities.

Top- dow n skills:superior ability to use syntactic knowledge.superior semantic knowledge and abiiity to use


32/73

32context.superior ability to go beyond the single sentencein drawing inferences about the story line.

It seems reasonable to argue that the cognitive skills required in reading aredifficult tasks for second-language learners to master and often lead to frustration andschool failure.

A crucial period is the late elementary grades. It is at this time that childrentypically read reasonably smoothly in units larger than individual words, but are not yetfully m atu re and skilled readers (G ibson Levin, 1975). Th e jum p to mastery in readingrequires that the child learn how to extract meaning quickly from text--a task thatassumes that words are decoded quickly enough to allow space in working memory forretaining the evolving meanings (L aBerge Samuels, 1974; Perfetti Ho gab oam ,1975). Poor readers may be hampered in achieving comprehension by their inability toachieve automatic word-decoding or even by non-automatic symbol-sound matching.

Reading a second language requires all these bottom-up skills. Furthermore,children who are learning to read in a second language may have more problems thanmo nolingual children because of their lack of familiarity with the semanti c and syntacticconstraints of the target language. If children are not able spontaneously to identify andexploit syntactic relations and are not flexible in their use of semantic context as a guideto prediction, their reading comprehension and speed decline (Carr, 1981).The Whole Lan-guage Approach


33/73

33Focus on such bottom-up skills is anathema to advocates of the whole language

appro ach. The y view such efforts as fragmentary and reductionistic. Th e emphasis inthe whole language approach is on making reading meaningful and on involving studen tspersonally.entire curriculum. The ma tic instruction mak es reading an integral part of instruction,not a subject matter of its own.

Language should not be taught piecemeal, but as the essential focus of the

Whereas the traditional cognitive approach views the teacher as an expert and thestudents as apprentices, the whole language approach sees the teacher as a facilitatorand the student as defining the task of making meanings. Th e traditional approach tendsto view the skills involved in reading as developmentally sequenced, whereas in thewhole language approach a skill is taught when a particular child needs it or somethingthat t h e child is working on. Literacy skills are seen as interrelated in the wholelanguage approach; oral skills need not be fully developed before reading, nor doesreading necessarily precede writing.

T h e whole language mov ement is mo re than a theory of language learning; itrepresents a philosophical stance on education and makes a political statement regardingthe distribution of power (Edelsky, 1990). It sees education as a socially and culturallyshared activity and asks how literacy is socially constru cted in the classroom. Stude ntsneed to be empowered so that they value their own experiences, communities, andcultures.

The whole language movement has impacted more traditional views of literacyinstruction. For example, Means and Knapp (1991), in a discussion of how the cognitive


34/73

34appro ach ap plies to children from culturally diverse backgrounds, argue that curricularchanges need to have a focus on complex, meaningful problems and that connectionsshould be m ade with students out-of-school experience and culture. Wh ile stressing theimportance of modelling powerful thinnking strategies and providing scaffolding to enablestuden ts to accomplish complex tasks, these authors also n ote the im portan ce ofencouraging multiple appro aches and solutions and making dialogue the central m ediumfor teaching and learning.comprehension strategies, Harris and Graham (1992) noted that such instruction musttake place in appro priately meaningful contexts and environmen ts.

Similarly, in a recent discussion of m etho ds of teaching

In the ar ea of writing similar efforts have been made to place the construction ofmeaning at th e center of the cu rriculum and to make writing integral to all instruction.This movement views writing as a process, and has been brought into the class-room bythe National Writing Project and the Writing Project of the University of California.Th e writing process appro ach is used widely with m ainstream children and has bee napplied in some contexts to language minority children Cutierrez, 1992).

In a review of the research base of the whole language appr oach, Pearson andRap hae l (1991) noted that several features of the whole language model have beenpositively associated with successful literacy instruction. For example, there is consider-able evidence that reading literature results in better reading com prehension than doesisolated skill practice. Similarly, research has indicated that the quality and q uantity ofchildrens writing a re im proved when they ar e encouraged to participate in wide-ranging,unfettere d w riting activities from the outset of schooling. In addition, reliance on


35/73

35authen tic functional literacy tasks has bee n shown to develop a more realistic view of theuses of reading and writing.

There is also evidence that the whole language approach reduces the culturalmismatch that frequently occurs in classrooms with children from linguistically andculturally diverse backgrounds because the students and not th e tea cher define the con-text of the learning situation. However, there a re also unanswered questions about theeffectiveness of reform efforts in teaching literacy skills to ethnic and language minoritychildren. De lpit (1986) and others have bee n critical of the e ffects of writing processinstruction on minority children. Th e concern is that such metho ds do not allow studentsto learn and produce the type of discourse upon which assessment is based--Le., standardacade mic discourse.

A number of authors have recently attempted to reconcile whole language andmo re traditional cognitive appro ache s (e.g., Ga rcia Pea rson, 1990; McKenna,Rob inson, Miller, 1990). How ever, Edelsky (1990) and othe rs have argued that suchattempts are futile and that whole language represents a parad igm shift. Attempts, forexample, to use traditional assessment instruments as outcome measures to determineinstructional effectiveness are regarded by whole language advocates as instances ofparad igm blindness. Re liance on test score data is seen by whole language advocates asreinforcing mechanisms for stratifying society--i.e., test score-based tracking.

W he the r these conflicting views can be reconciled rem ains to be seen. However,regardless of whether researchers use the more qualitative methods of the wholelanguage paradigm or more traditional quantitative methods, it is important to determine


36/73

36under what conditions innovative instructions are effective with language minoritystudents. Especially in the la te elem entary grades, where literacy skills ar e central toacade mic success for the se children, there few more important ed ucational challenges.

Elite vs. Folk BilingualismFishrnan (1977) draws a key distinction between situations where bilingualism is a

goal for the eli te or, simply a p redicament of the common folk. Oth er relatedterminology that has been offered include additive vs. subtractive bilingualism (Lambert,1975) and elective vs. circum stantial bilingualism (V aldes, 1992). Th e distinctionbetw een diffe ren t situations of bilingualism, especially as it pertains to the status of thegroups, is useful in understanding t he orientation of researchers working in the differentphases, a s well as in sorting through conflicting findings and conclusions (Hakuta, 1986).

T h e teaching of a second language applied to the elite has been t he mainpreoccupation of the Foreign Language and the Ca nadian Im mersion phases of research.In these situations, the problem to be solved is how to most creatively or intensely teachthe second language because of the needs of the m iddle and uppe r classes. During theForeign Langu age phas e, the need was international competitiveness. In the case of theCanadian immersion programs, the main proponents were middle class Anglophoneparents who sensed opportunities for their children if they learned French in a societythat is officially bilingual. In both cases, the status of the native language is neverquestioned, and the desirable goal is bilingualism.

Durin g the Psychometric an d Language Minority phases, on the o ther hand,attention has been on the popu lation of studen ts who a re us~ ially mm igrants and of the


37/73

37"folk" variety. T he main social question is whether the im migrants ar e learning Englishfast enough; there is an attitude of benign neglect or outright hostility toward the nativelanguage, often accompanied by a fear of perm anen t ethnolinguistic ghettos that couldlead to social fragmentation. In the Psychometric phase during the early part of thiscentury, researchers used the yardstick of standardized intelligence tests to see whetherthe new immigrants were sizing up to the old (Hakuta, 1986). And in the currentLanguage Minority Phase, researchers have tended to measure it in the speed of Englishproficiency development and measures of school achievement. In these settings, thestatus of the native language is marginal, and while i t might be acceptable if thelanguage were maintained in t h e home, bilingualism is not a desirable goal Imhoff,1990).

As Ha ku ta (1986) has documented, the social status of the subject population aswell as the social values of the researcher conducting the studies have led to verydifferent conclusions about the effects of bilingualism on intelligence. During thePsychometric Phase, th e studies mainly concluded that w hen bilingual subjects werecompared with monolingual controls, the bilinguals performed worse than themonolinguals on various standard ized tests of intelligence. O n the othe r hand, much ofthe research with elite bilingualism indicate that the cognitive and linguistic outcomesare positive. Thus, bilingualism might be considered good for the elite, bad for thefolks. In reconciling con tradicto ry evidence, a key is to locate the exact explanatory levelwhere the differences might be reconciled.


38/73

38Cum mins (1976) surveyed research with both folk and elite bilinguals on the

effects of bilingualism and hypothesized that the key mediator was whether a thresholdlevel of bilingualism had be en attained . Th e threshold hypothesis states that positiveoutcomes result only when children have attained a high level of functioning in twolanguages. In contra st, in situations of folk bilingualism in which the first language iscompromised, such as that found in immigrants who replace their native language withth e second language, negative co nsequences would result. T he locus question is whe therthe appropriate explanatory factor is a cognitive one in which the level of bilingualismattained that is in turn determined by the sociological circumstances of language status,or whether it is alternatively a sociological on e in which statu s affects both linguistic a ndpsychological conclusions. The threshold hypothesis places the locus at the cognitivelevel.

The theoretical perspective on language implied by the threshold hypothesis hasbeen attacked, from the sociological perspective, for implicitly legitimizing the notion ofsemilingualism (Edelsky etal 1983; Cummins Swain, 1983), a condition in whichthe child develops full proficiency in neither language (Skutnabb-Kangas Tou kom aa,1976; see Ro ma ine, 1989). Wh ether semilingualism is a valid concept at the cognitivelevel is a matter of great controversy (Paulston, 1982).

The roots of this controversy over the ontological status of semilingualism referback to a fundamental issue of whether certain socially accepted linguistic codes can beinherently limited in their functioning. Th e controversy parallels the de bate o n BlackEnglish of the 1960s on whether the vernacular was simply a degraded version of


39/73

39standard English or possessed its own integrity as a linguistic system (La bov, 1972). Italso parallels the de ba te over code-switching in bilinguals and w hether the phenomen ondem ons trate s linguistic confusion or a controlled form of expression (Zentella, 1981). Inthe view of some critics of the concept of semilingualism, there is nothing deficient in thelanguage of folk bilinguals, and insisting on its existence merely reflects middle class bias(Brent-Palmer, 1979). Thus, the locus of explanation in their view is social bias againstlower class immigrants and the a ccep tance of linguistic and cognitive measures that ar enot valid.Bridges betwee n Elite and Folk B ilineualism

It is our view that basic psychological and linguistic processes are notfundamentally different between elite and folk bilinguals. Error an d performa nceanalysis of the acq uisition of second language gram mar, for example, do not indicate anysystematic differences between studies condu cted with elite or folk bilinguals. W esubscribe to what Jo hn Macnamara (1976) once said abo ut language acquisition: whenan infant, a ten-year-old child, and an adult learn Russian, the most remarkable outcomeis Russian (P. 175). The cognitive and linguistic mechanisms for learning language areuniversally available and ar e unlikely to be incapacitated in most circumstance ofbilingualism.

In addition to similarities in the process of second language acquisition, theliterature suggests other important commonalities between elite and folk bilinguals.There is evidence in the literature of positive correlations between bilingualism andmeasures of cognitive performa nce even among folk t)iIiiigual subjects when pro per


40/73

40methodological controls ar e employed (Duncan D e Avila, 1979; Hakuta, 1987). Th ereappears to be nothing about the sociological situation that causes the results to bedifferent fro m what has bee n found with elite bilinguals. Likewise, studies of n aturaltranslation among children not formally trained in the task show a high level offunctioning both among elite bilinguals in Gen eva (M alakoff, 1991) and low-SES PuertoRican bilinguals in New Hav en, Connecticut (Malakoff Hak uta, 1991).

A final example of a bridge between elite and folk bilinguals is found in two-waybilingual education programs that involve a m x of language majority and languageminority students, with the overall goal of developing bilingualism for both groups ofstuden ts beginning in elem entary school. Thes e programs ar e rapidly growing inpopularity in the Un ited States (Christian M ahrer, 1992). In effect, they combine thecharacteristics of traditional bilingual education programs for language minority studentswith immersion e duca tion for language majority students. They address one of the majorconcerns about these programs in that they address the sociolinguistic needs of languagedevelo pm ent. T he concern has been that language minority students in bilingualeducation are not sufficiently exposed to English models (which, however, is addressedby the fact that English is ubiquitous in the United States), and the more seriousproblem that traditional immersion has tended to create its own sociolinguistic situationbecause of the lack of native speak ers of the language (Selinker, Swain Dum as, 1975).This elite-folk combinant experiment deserves to be followed with special interestbecause it directly addresses the major sociological tension in the field of secondlanguage acquisition.


41/73

41Basic vs. Am lied Research

Bo th first an d second language acquisition a re exciting fields becaus e they holdpromise to help answer im portant fundamental theoretical questions on the nature oflanguage, mind and culture. Yet one striking characteristic abo ut research in secondlanguage acqu isition is the e xtent to which it is motivated by th e ne ed to addressproblems of the rea l world. The activity of second language acquisition researchers is farmo re clearly shaped by societal concerns tha n is the work of their first languageacquisition counterparts.

Fo r example, there is an impressive body of research on the Canadian Frenchimmersion experience in which A nglophone students a r c placed in French-only classesfrom kinder garten (e.g., Lam bert Tucker, 1972; Lambert , 1984; Genesee, 1978). T h emain question that is asked is whether they are able to maintain pace in English

language arts and subject matter with Anglophone sttrtlents schooled only in English.This is a primary question because while parents want their children to becomefunctional in French, they also want strong reassurance that they are not losing ground inthe do minant language of the country. T he research h a s also examined the effectivenessof program variations, using different configurations of grade and language mixtures (seeGenesee, 1984).

T h e q uestions that m otivate second language acquisition researchers, of course,fall along th e entire spectrum from basic to applied. Or1 the more theoretical end,que stion s ask ed by second language acquisition researc hers include: Are ther e anynegative or positive consequences, e ithe r in term s of lariguage or cognitive development,


42/73

42associated with bilingualism (e.g., Diaz, 1983)? Is there an optimum age for secondlanguage learning/teaching (e.g., Long, 1990)? W hat are the differences between thecognitive and social uses of language when it comes to second language acquisition (e.g.,Snow, 1987).

Somewhat more applied in nature are questions such as: What should be theexpected rate of second language acquisition so that we can build policy expectationsabout how long children might stay in bilingual education programs (e.g., Collier, 1987)?How can bilingual children best be assessed in terms of their language proficiency andacademic achievement (e.g., Cummins, 1981)?

Among the most applied of research questions concern program evaluation:What are the characteristics of effective bilingual education classes ( e g , Tikunoff,1983)? Do es immersion in French impede the academ ic and English languagedevelo pm ent of Can adian anglophone students (e.g., Lam bert Tucker, 1972)? Wh at isthe relative effectiveness of various approaches to the education of language minoritystudents (e.g., Ramirez, etal991).

Perhaps not surprisingly, it is the most applied, policy-oriented questions that tendto gene rate t he greatest am ount of political controversy. At times, the political heatobstructs the ability to conduct objective research, or unnecessarily constrains the way inwhich th e questions ar e framed. In our view, such a situation must be balanced by good,theoretica lly sound, basic research . This need is most dramatically demo nstrated in thecase of the pursuit of a single policy point: the ues tion.The Lau Ouestion


43/73

43T h e Lau question in the educa tion of limited-English-proficient studen ts in th e

U. S. refers t o whether it is justified to prescribe transitional bilingual educ ation(providing content instruction in th e native language until th e child beco me s proficient inEnglish) as a method to educate such students. Th e nameSupr eme Cou rt ruling, Lau v. Nichols, which ruled that the San Francisco school districthad violated Title VI of the Civil Rights Act of 1964 by failing to provide specializedprograms to m eet the n eeds of Chinese-Am erican studen ts who were limited in Englishproficiency. In response to this ruling, the Department of Health, Education andWelfare issued a set of proposed remedies (known as the Lau remedies) to be used bythe Office for Civil Rights to negotiate compliance plans with school districts that did notprovide special programs for limited-English proficient students, and thus were inviolation of Feder al law. The se prop osed remedies, and a proposed federal regulationissued in 1980, require d the provision of transitional bilingual ed ucation in mostinstances (see Baker de Kanter, 1983a, Appendices A B, and C).

comes from the 1974

In addition to the proposed remedies, the Bilingual Edu cation Act (T itle VI1 ofthe Elementary and Secondary Education Act) authorized competitive grants to localscho ol districts to develo p their capacity to provide bilingual education. This lawrequired that most funds be used to provide programs that used native languageinstruction. Combined with thethe affirmation and prescription of bilingual education for limited-English-proficientstudents (Birman Ginsburg, 1083).

remedies, these federal actions could be viewed as


44/73

44T he uestion has been controversial from the beginning, especially during th e

presidency of Ron ald Re aga n (1981-1988). Th e political chara cter of the pro blemstemm ed from the perception that this amounted to fe dera l sanction of ethnolinguisticdiversity as well as the intrusion of the fe deral go vernment in local governance (Epstein,1977). Th ese ar e questions that speak to the heart of Am erican identity. Th equestion has inevitably come to define the research in this area, especially work fundedby the government.

A num ber of major attempts have been u ndertak en to examine whether bilingualeducation is more effective than alternatives, such as the provision of ESL (English as aSecond Language) only. All studies were conducted at the elementary school level, inmost cases focusing on English proficiency and academic achievement measured inEnglish. A study by the American Institutes for Research (Danoff etal977, 1978)compared a large sample of students in Title VII-funded transitional bilingual educationprogr ams with those who were not. Baker and de Kante r (1983b) summ arized availableindividual evaluations of bilingual education projects that reported data from controlgroups (where bilingual education was not available). In the 198O s, the Department ofEduc ation commissioned a pair of longitudinal studies. O n e study attem pted to follow anationally representative sample of LE P students who varied in the types of services theyreceived, and to conduct causal modelling of the data to determine the effectiveness ofthe service types (Development Associates, 1986). Another (Ramirez et 991) used amore traditional comparison model to look at three existing models (transitional


45/73

45bilingual education, structured immersion in English only, and bilingual education witha native language maintenan ce orientation ) in a selected number of schools.

Th ese studies (many of them rath er expensive) have failed to provide conclusiveevidence on the uestion -- whe ther bilingual education is superior and thereforeshould be the me thod of choice. What should the conclusion be ? Is the null hypothesiscorrect? Or have we a case of Type I erro r? Many have speculated . Critics ofbilingual educ ation prefer the interpretation that the evidence is accurate, and ther e isindeed no effect (Rossell Ross, 1986). Supporters claim the evidence foul, pointingout the flaws (not a difficult task) in the studies (e.g, Gray, 1981), or becomephilosophical about whether the positivistic approach toward program evaluation isapp rop riate (Cziko, 1992). Oth er supporters look at the evidence and find solace in thefact that the more honest comparisons yield data in favor of bilingual education (Willig,1985).

It may well be the case that these evaluation studies point to the limits of anapproach that compares one program type with anothri-. A National Academy ofSciences pa ne l condu cted a thorough review of these I \V O major national longitudinalstudies (Fienberg 91 Meyer, 1992). Aside from d ocun ienting fatal design flaws in thestudies, th e pa ne l was critical of the general atheoretical orientation of the researchprogram, essentially arguing that large studies cannot serve as theoretical prosthesis.Rather, the panel recommended a model of knowledge development based on smallerscale, targeted studies that test and refine the basic theoretical premises of bilingualeducation (as in Can ada).


46/73

46Many of the theoretical questions, it turns out, have been asked by second

language acquisition researchers whose work lean toward the basic research end of thespectrum. For example, in answering the question of whether bilingual education iseffective, much of the fear is based on the belief that second language acquisition is azero-sum process such that instruction in the native language detracts from rapid andefficient learning of English. Yet the basic research her e suggests that if anything, thereis a positive correlation between first and second langiiage proficiency, and the cognitiveconsequences of bilingualism are probably positive (Cummins, 1976). This findingshould allay the concern that bilingual education comes at the expense of Englishdevelopment.

Another important finding is the rate of second language acquisition, whichsuggests that most children do not attain the asymptotic levels of English proficiency for

anywhere between 5 to 7 years, conbiderably less than t h e time frame (usually two years)required in federal and state legislation (Collier, 1987). Setting more realisticexpectations of the rate of English development is critical in ensuring that bilingualeducation programs not be evaluated solely on the speed with which students exit fromthe programs.

Perha ps the most important contribution of t he perspective provided by basicresearch has been its ability to offer insights into the processes involved in themaintenance and loss of bilingualism. Th is effor t is a rnultidisciplinary one, ranging fromsociology (e.g., Fishman etal966; Veltman, 1983), ethnography of communication (e.g.,Ga l, 1979; Gu mpe rz, 1982; Trueb a, 1989), linguistics (e.g., Ex tra Verhoeven, 1993),


47/73

41and psychology (e.g., Ha ku ta, Dia z Fer dm an, 1987; IHakuta DA ndrea, 1992).The se perspectives amply dem onstrate the complexity of the factors involved inbilingualism and point to the barren nature of the ways in which the questions have beenaddressed in the program-oriented, ap plied research studies. In a nutshell, this area o fbasic resera ch ex tends the perspective on bilingual proficiency from the psycholinguistict o the sociolinguistic, and from the individual to the speech community as the unit ofconcern. U nd er this view, what is learned is not just the ability to speak the secondlanguage and maintain the native language, but rather the ability and the social capacityto become active participants in two speech comm unities. In addition, main tenance ofbilingual proficiency is viewed not just as the question of an individual who, in the courseof the lifetime, might retain or lose proficiency in the cthnic language. Rathe r, theadditional question is whether the ethnic language gets transmitted to the nextgeneration of the speech community, or whether the speech community withers away.

question, it is likely that the policy community would have Iittles to thepatience for what they would see as the basic research on social science gibberish aboutbilingualism. From their perspective, they are interesiccl in the bottom line, whether theprograms that are funded work or do not work. They itre not interested in fantasizingabout what is possible to use Czikos (1992) words, iiiey are interested in what is the

probable outcome of the programs, not in what is possible. Basic and a ppliedresearch must meet somewhere in between if they arc to have any impact. Th e ideasgenerated from basic research need to be woven into [lie culture of policy and programs


48/73

48and become an integral part of evaluation studies.

Theorv vs Dat aO n e of the enduring tensions in any field of inquiry concerns theory and

methodology . This is no less the case in second-language research. Th er e are those whoargue th at theory should drive research and othe rs who feel that one should work fromthe bottom up, building theory piece by piece on the l m i s of research findings.Where to Begin

This issue has been addressed by Long (1985), who distinguished between atheory-then-research strategy and a research-then-lheory strategy. No research isentirely a-theoretical, but some research is more theory-driven than other research.Long noted that there are advantages and disadvantages to both the theory-then-researchand the research-then-theory orientations.

The theory-then-research strategy has the advaii tage of providing an approximateanswer until the final truth is known. Such theories scrve a useful heuristic, assumingthat they generate testable hypotheses that can confirm or disconfirm the theory. Thedisadvantage of a theory-driven approach is what soci:iI psychologists call confirmationbias (Gree nwa ld, Pratkanis, Leippe, Baum gardner, 1986): ones preliminaryhypotheses have a decided adva ntage in the judgment process.

The advantage of the research-then-theory approach is that one is closer to theempirical evidence at hand and makes only limited clxims. T h e likelihood of aconfirmation bias is not ruled out because all research tests implicit theory, but there is


49/73

49less investment in a theoretical point o f view. Th e disadvantage is that such an approachlacks the heuristic power of a more developed theoretical approach.

Long (1985) argued that th e theory-then-research strategy allows for moreefficient research . H e maintained that the theory governing the research at any point intime tells the investigator what the relevant data are and what is the critical experimentto run. Such a research strategy leads to explanatory xccounts of the processes at workin a given domain. In contrast, Greenwald and his associates (1986) have argued that th eresearcher who sets out to test a theory is likely to become ego-involved with atheoretical prediction, to select proce dures that lead eventually to prediction-confirmingdata, an d th ereby prod uce overgeneralized conclusions.

Th e deba te has a long history in the philosophy of science. Kuh n (1962) favoredthe theory-then-research strategy, and argued that o r d nary scientific activity thrives ontheory confirmation--solving puzzles within the existing paradigm. He pointed out,however, that theory conf irmatio n does not succeed indefinitely. Anom alous resultsaccumulate until only a major theoretical reorg anization (scientific revolution) canaccommodate them. Popper (1959), on the other hand. regarded exclusive use ofconfirmation-seeking methods as non-scientific. In his view empirical knowledge in ascientific domain grows only by the use of critical, falsification-seeking me thods .

The difficulty is that falsification-seeking is given more lip-service than practiced.If, as many have argued, all research has an implicit theory, it is impossible to escapeconf irmatio n bias. Even researc hers who stay within I limited dom ain and deal with


50/73

50only cer tain issues ar e likely to have definite expec tations abo ut their data. Nonetheless,many agree with Popper that

. . . we start our investigation with problems. W e always find ourselves in acertain problem situation; and we choose a p roblem which we hope wemay be able to solve. Th e solution, always tentative, consists in a theory, ahypothesis, a conjecture (1976, p. 86).

Because we approach problems with an implicit theory, the process is an interactive one:we test our conjectures, we modify our theory, and as the theory withstands tests we areless tentative in accepting the original hypothesis.

In second-language research, many investigators are currently working with aresearch-then-theory strategy, looking first at what the data tell us descriptively and thenmoving upward toward theoretical claims. Thus there w

1996 - mclau - hak - biling y l2

Documents