(2001) - A Critical Period for Second Language Acquisition

8/3/2019 (2001) - A Critical Period for Second Language Acquisition

1/8

,Crit ical TMzine

About Critical Periodsedited by

Donald B. Bailey,Jr., Ph.D.Frank Porter Graham Child Development CenterUniversityof North Carolina at Chapel W

John T. Bruer, Ph.D.St. Louis, MissouriJames S. McDonnell Foundation

FrankJ. Symons, Ph.D.University of MinnesotaMinneapolis, Minnesotaand

Jeff W. Lichman, M.D., Ph.D.Washington University School of MedicineSt. Louis,Missouri

8Na&onal Center fo-EarlyDevelopment&Lear#ng. P A U L - H .

Baltimore * London * Toronto. Sydney


2/8

10A Critical Period forSecond Language Acquisition?Thecrit ic: )eriod hmoi :sis

Kenji H&ta

.. r second language &2) learning has foundresonance in a variety of policy positions regarding when a n L2 shouldbe introduced in the curriculum. For immigrant students, in a 1998 courtdeclaration urging that such students be exposed to English as early aspos-sible, an advocate wrote, The optimal time to learn a second language isbetween age three and five or as soon thereafter as possible, and certainlybefore the onset of puberty (Forter, 1998,p. 1).Porter is a leading advo-cate for English-only approaches to the education of language minoritystudents and an opponent of bilingual education programs because theydelay intensive instruction in English.She has found the critical periodhypothesis to be in suppor t of her position.

Advocates for the early introduction of foreign language in the ele-mentary schools, dating back to the Foreign Languages in the ElementarySchools (FLES) programs of the 1960s,also have found an important allyin the critical period hypothesis. For example, the New Jersey State De-partment of Educations World Languages Curriculum Framework citedcritical period research to make the following point: With each year ofgrowth, children are less ableto filter out b e istinctions amongthe soundsof other languages. After early childhood, the language acquisition mecha-nism becomes highly structured creating an interference effect that mayaccount for the difficulty in learning languages at a later time (1999, p. 7).Such statements draw on the critical period hypothesis for L2 a c q ~ i -tion, the origins ofwhich are amibuted to Penfield i d oberts (1959)andmore prominently perhaps to Eric Lenneberg (1967), who amassed 4-

193


3/8

194 Hshutadence in support of the view that first language (L1) acquisition is a bio-logically constrained process, with a specific timetable ending at puberty.In a single paragraph, Lenneberg speculated about the implications for L2acquisition, noting that after puberty, second languages are acquired con-sciously and with great effort, and often not successfully (see Chapter 9).The purpose of this chapter is to make assumptions underlying this hy-pothesis and to highlight what is known and not known about its empiricalstatus.

FIRST LANGUAGE (L l) ACQUISITIONA brief foray of the standard version of the critical period hypothesis for L1acquisition is in order. Th e clearest account can be found in Pinker (1994).This view is based on Chomskys account of linguistic competence-anabstract set of rules and representations that is highly specific to language(Le., organized differently from other mental capacities, e.g., visual cogni-tion) and an innate component of the human mind. Th e standard argumentis that i t is logically impossible for a child to acquire linguistic competenceof this complexity through the types of exposure to language that childrenreceive in their home environment. Th e argument states that a specializedbiological program must exist for language acquisition, similar to the pro-grammed course of development of physical systems such as vision, diges-tion, and respiration. As ong as children are exposed to a threshold amountof linguistic exposure during the critical period, they will all uniformlyacquire linguistic competence, much as children develop similar physicalorgans despite considerable variation in nutrition. And, if they are deprivedof this exposure during the critical period, no amount of exposure after itcan compensate for it.

Direct evidence in support of the critical period for L1 acquisition isthinand based on theoretical arguments and analogy to othe r well-exploreddevelopmental processes, such as visual development in the cat (Hubel,1988). Most children are exposed to language early in life and acquire it suc-cessfully. Indeed, the first argument in favor of a critical period is its uni-formity in spite of considerable environmental variation in the ways thatparents talk to children. For ethical reasons, experiments in which infantsare deprived of exposure to language during the putative critical period arenot conducted. Other evidence in support of the hypothesis comes fromunusual, tragic cases of language deprivation resulting from child abuse, andfrom studies of deaf children who are born to hearing parents, but who areexposed to American Sign Language (ASL) a t a later age. Nevertheless, thehypothesis is commonly accepted. As Pinker s tated, Acquisition of a nor-mal language is guaranteed for children up to the age of six, s steadily com-

A Critical P d c d forSecondLa&+ Ac~isition7 195promised from then until shortly after puberty, and is rare thereafter (1994,p. 293).ELEMENTS OF THE CRITICAL PERIOD HYPOTHESISFOR SECO ND LANGUAG E (LZ) ACQUISITIONIn theorizing about a putative critical period forL2 acquisition, a key fram-ing question is whether L2 acquisition recapitulates the L1 acquisitionprocess (a hypothesis known in literature as the L2=L1 hypothesis), or al-ternatively, whether L2 acquisition is a cumulative process that builds onthe competence already developed in L1. If the L1=L2 hypothesis is cor-rect, then the evidence for or against a critical period for L1 acquisition isrelevant to L2 acquisition. However, if the cumulative model is correct, theevidence from L1 acquisition is irrelevant to answering the question aboutL2 acquisition.The research evidence on the nature of L2 acquisition is clear on twopoints, but they are contradictory First, with respect to rate of acquisition,there is evidence that linguistic similarity between the L1 and L2 matters(Odlin, 1989). Anative speakerofSpanish will acquire English more rapidlythan would a native speaker of Chinese, all other things being equal, be-cause of the linguistic similarity between Spanish and English. This evi-dence woula imply that the cumulative model is correct. Second, withtespect to error patterns and the overall qualitative course of L2 acquisi-tion, there is a similarity across speakers of different languages learning agivenL2,indicatingthatmorethanasimpletransferfromLl oL2 isoccur-ring (Bialystok& Hakuta, 1994). Indeed there is some sort of reenactmentof the L1 acquisition process at work. for Lenneberg, the originator ofthe critical period hypothesis for L2, it appears that he favored the cumu-lative model when he wrote tha t we may assume that the cerebral organi-zation for language learning as such has taken place during childhood, andbecause natural languages tend to resemble one another in many funda-mental aspects, the matrix of language skills is present (1967, p. 176). Inany event, the jury is still out as to whether L2 is a recapitulation of L1acquisitionor an add-on process. Wha t would be the key elementsof a crit-ical period in LZ acquisition?1. Clearly specified beginning and endpointrfw he period: Lenneberg sug-gested puberty, and others have followed suit. Johnson and Newport(1989) considered age 15 to he the end of the critical period. As notedpreviously, Pinker considered it to begin at age 6 and end at puberty.

For present purposes, assume that the critical period hypothesis isset by puberty and ends at age 15. In any event, any claim to a criticalperiod for L2 acquisition should be specific about an end point.


4/8

196 n&&2. Well-dejineddeclinein L2 acquisitim at the end of theperiod:Th e ability tolearn things declines with age, such as learning to ride a bicycle, yet itwould not be'stated that there is a critical period for cycling.A general

decline in learning isnot sirong evidencefor a m'ticalperiad fir L2 acquisi-tion. The appeal of a critical period hypothesis liesin its specificity, thatis, its ability to'target specific learning mechanisms that get turned offat a given age.(Birdsong, 1999). Thus, one important piece of evidencewould be ifa rapid decline could be found around the end of the criti-cal perioti, rather than a general monotonic and continuous declinewith age thatcontinues throughout the life span.Evidenceof p l i t a t i v e d if fwen ces in learning between aquisition w ithin andoutside the criticalperiod:A critical period is assumed to he caused by theshutting down of a specific language learning mechanism. Therefore,any learning that happens outside of the critical period mwt be theresult of alternative learning mechanisms. If that were the case, thenthere should be clear qualitative differences in the patterns of acquisi-tion between child and adult L2 learners. For example, if certain gram-matical errors could he found among adult learners that are neverfound in child learners, or if child learners were able to leam specificaspects of the language that adults could not learn, then this would hestrong evidence for a critical period.Robustness to environmental variation inside the m'ticalpm'o d:Another at-traction of the critical period hypothesis is that there is a thresholdlevel of exposure with uniformed outcomes, even with considerableenvironmental variation. The environment might play a larger role be-yond that period and the outcomes would become more variable.

3.

4.

JOHNSON AN D NEWPORT'S STUDYA study hy Johnson and Newport (1989) reported results consistent withthe critical period hypothesis. Th e study's results are cited as authoritativeevidence for a critical period in L2 acquisition. In their s tudy of native speak-ers of Chinese and Korean who came to the United States at ages rangingfrom 3 to 39 years old, they asked individualsto identify grammatical andungrammatical sentences that were presented auditorily. They reportedthat prior to age 15 there was a negative correlation with age, hut after age15 there was no correlation with age (satisfymg conditions 1 and 2). In addi-tion, the adult learners showed great variability in learning outcomes,whereas the child learners did not (condition 4).A reanalysis of the data by Bialystok and Hakuta (19941, however, re-vealed some problems with the original interpretation. Bialystok and Ha-kuta argued that the data showed a discontinuity not at puberty hut rather

A Citicd P&od forSecond hnguage Acquieition? 197at age 20, and that there was statistically significant evidence for a contir-ued decline in L2 acquisition well into adulthood. It is likely that the pecu-liarities of the sample (students and faculty from the University of Illinoisat Urbana-Champaign) could have further complicated their results. A pic-ture of the reanalysis is shown in Figure 10.1.The data show a continuousdecline with age of arrival, which is not consistent with condition 2. Fur-thermore, the data patterns suggest two distinct groups of subjects, thosebefore and after age 20, both of which show declining performance withage. T he study should not he considered definitive in light of its samplinglimitations.

All Subjects300

crnCW.-El 100

WW

r=-.87 r= -.49

00 10 20 30 40Age of ArrivalFigure 10.1. Reanalysisof Johnsonand Newpolt (1989)study showing discontinuity at age20 , andcontinueddeclineinadultrubjects. (From Bialystok, E., & Hakuta. K. 119941. Inotherwords: The scienceand psychologyof second-language acquisition. NewYork Basic Books:reprinted by permirsion.1


5/8

198 H&taIhl

Age of Acquisition Age of AcquisitionFigures lO.2aan d 10.2b. Sometheoretical predictio nsofthe criticalperiod hypothesisshow-ing disruption at predicted end of the critical period.

CONDITIONS 1 AND 2:END POINT FOR THECRITICAL PERIOD AND DISCONTINUITY AT THAT POINTTheoretically, the critical period hypothesis generates a prediction thatshould look like Figure 10.2a or 10.2b, with a disruption occurring at thepredicted age point. The difference may be in slope breaking at the agepoint, as in Figure 10.2a, or the slopes may be the same on either ends ofthe age point, but the re could be a sharp drop-off at the age point, as in Fig-ure 10.2b. A test of conditions 1 and 2 can be found in a study reported byBialystok and Hakuta (1999) using the United States census data from1990. Th e study looked a t a large sample of immigrants whose native lan-guages were Chinese and Spanish and who had immigrated to the UnitedStates at ages ranging from birth to 70 years old. Th e census bureau askedfor a self-report of their English ability, which was converted to a four-point scale. The scale was validated by the census bureau against actualmeasures of English proficiency in a separate study. The data from thisstudy showed continuous decline with age and no evidence of a disconti-nuity or sharp break at puberty as would be expected by conditions 1 and2. The data are shown in Figure 10.3-it is essentially a straight line andthere is no evidence for conditions 1 or 2.CONDITION 3: QUALITATIVE DIFFERENCESBETWEEN CHILD AN D ADULT LEARNERSIt is important for a critical period hypothesis to demonstrate that a specificlearning mechanism is present during the period but not outside of theperiod, and one way to do sowould be to show different patterns of acqui-sition in adults and children. Studies that compare the errors and perfor-mance patterns of child and adult L2 learners are informative in testing forthe viability of condition 3 .

A Critical Period for Second hnguage Ac4;s;tid 199One area of research is in t he extent of naave language influence L2learning. The relevant question is whether children differ from adults inthe extent to which native language influence can he found. The theoreti-cal basis for this can be found in the late 1950s and 1960s when the pre-dominant view of L2 acquisition was that of language transfer, based on theprinciples of behaviorist psychology (Hakuta & Cancino, 1977). In this

view, the points of contrast between the native language and the target lan-guage determined the course of earning-positive transfer happenedwherethe two languages were similar and neptive transfer where they were dif-ferent. For example, native speakers of Japanese have difficulty with theEnglish determiner system (e.g., a, the, some) because there is no equivalentsystem in their native language. Native speakers of Spanish, however, donot have as much difficulty because a similar system exists in their nativelanguage. The question, then, is whether adult learners show more evidenceof transfer errors than children because, according to the critical periodhypothesis, children directly gain access to the target language whereasadults must go through their native language. This does not appear to bethe case. Children who are learning a second language show evidence oftransfer errors similar to adults and overall patterns of errors are not dis-tinguishable between children and adults (Bialystok& Hakuta, 1994).Another opportunity to look for differences between children andadults is in the pattern of development in the L2. In one study, Bailey, Mad-den, and Krashen (1974) compared the performance of adult and child

Chinese Spanish

4 4L.

'i- 1 ' t0 '0 20 40 60 80

Ag e o i arrivalge of arrivalFigure 10.3. Self-reported English proficie ncy for U.S. immigrants as a iunction of age ofarrival: Data from the 1990 US . ensus. (From Bialystok. E., & Hakuta, K. [19YY1. Con-founded age: Linguistic andcognitive factors in age differences in second language acquisi-tion. In D. Birdsong rEd.1, Second language acquisition and the critical period hypothesis[pp. 161-1811. Mahw ah, NJ:Lawrence Erlbaum Associates; reprinted by permission.)


6/8

,.200 H&hlearners of English as L2 on a test of English morphological srmctures.Specifically, they compared their ability to correctly use the present pro-gressive -ing, forms of th e verb to be, the plural -5, determiners (u, he), thepast tense, the third person indicative (he runs every day), and the posses-sive ? The results found a remarkable similarity in the rank ordered per-formance between children and adults, as can he seen in Figure 10.4. Thenative language background of students did not seem to affect the results.Overall, this study provides support for the fact that child and adult learn-ers progress along similar paths of development.A specific way to test the critical period hypothesis is by askingwhetheradult learners can demonstrate knowledge in the abstract aspects of lan-guage that are presumably accessible only through language-specific learn-ingmechanisms (what linguists have come to calluniversalgrammar).Whiteand Genesee (1996) conducted such a test to see whether adult L2 learnersof English had access to the following pattern of intuitions that all nativespeakers of English have:1. Who do you want to see?2. Who do you want to feed the dog?3 . Who do you wanna see?4. *Who do you wanna feed the dog?Number 4 (marked by *) is ungrammatical. Wby, despite surface similari-ties, is 3 considered okay, but 4 is not okay? If grammatical intuitions wereformed on the hasis of analogy, 4 should be okay. The logical argumentmade by linguists is that the underlying structure for the sentences can hehypothesized as,5 . Yon want to see who?6. You want who to feed the dog?According to the theoretical model of universal grammar, these underlyingforms of mho are moved to the front of the sentence, leaving behind a tracet i n the original location:7. ' Who, do you want to see t?8.Th e rule that reduces want t o to wanna for 8 is blocked hy the trace betweenwant and to . According to this analysis, this knowledge is needed in orderto find 3 to be okay hut 4 not t o be okay. The abstractness of this rule makesit hard to learn without preexisting knowledge. The critical period hypoth-esis says that the learning mechanism that allows for this knowledge to beacquired is no longer present in adults.

Who, d o you want t, o feed the dog?

A CriticalP&cd $orSecond*e Aquiition? 201

\ c .\, .......',, I' : ?... ...! ! : .: , ..\.I 1 ..: \. '4-- acramento ' y. '. \. \' 'EastHarlem

Figure 10.4. Comparison of performance on selected English grammatical structures i nadult and child learners. (From Bailey, N., Madden, C., & Krashen, S. I19741. Is there a "nat-ural sequence" in adult second language learning? Language Learning, 24, 235-243; re-printed by permission.)

Using sentences like these, White and Genesee asked adults who hadlearned English at different ages to discriminate between grammatical andungrammatical sentences based on abstract concepts. Although more adultlearners had difficulty in distinguishing between these sentences than childlearners, about one third of the adults who had acquired these rules showedequivalently high performance to child learners and native speakers ofEnglish. Thus, adults are capable of learning highly abstract rules that


7/8

202 H h atheory would say are accessible only with specialized language acquisitionmechanisms.There are no demonstrated differences between the process of L2acquisition in children and adults, with respect to condition 3 . As Bialystokand Hakuta concluded, The adult learning a second language behaves justlike a child learning a second language: he walks like a duck and talks like aduck, the only major difference being that, on average, he does not waddleas far (1994, p. 86).CONDITION 4 THE EFFECTS OFENVIRONMENTAL VARIATIONThe critical period assumes a minimal role for environment in learning,such that once the learner is exposed to a necessary and sufficient amountof stimulation, the learning is complete. An important variant in the envi-

0.9a, 0.8-8 0.7-05 0.6-LL5 0 . 5 -.O_ 0.4-c

0.3-2

L

-C

a 0.20.10.0 I I I I I 10 1 2 3 4 5 6 7Grade

Figure 10.5. English oral proficiency development in immigrant students from a northernCalifornia school district, separated by poverty level in schools. Th i s is a cross-sectional sam-ple, but al l subjects included n this analysis were enrolled n his school distric t since kinder-garten. (Key: Poverty level: A70, 50 . x 2 5, * 10.1

0 20 40 60Ageof Immigration 80

Figure 10.6.ageof arrival, separatedby educational attainment: Data from the 1990 U.S. censusSelf-reported English proficiency or native Chinese immigrantsas a function of

ronment is socioeconomic status of the learner. Figure 10.5shows oral pro-ficiency data for immigrant students from a school district in northern Cal-ifornia, varying by the socioeconomic environment of the school. Thisschool district does not provide bilingual education and students are ex-posed only to English during the school day The data show students whoare from lower socioeconomic schools attain English proficiency a yearslower than those students in higher socioeconomic schools.Strong socioeconomic effects can be found in the census data as well.Figure 10.6 shows the same data as Figure 10.2, hut they.are separated hyyears of education attained as a proxy for socioeconomic status. Ther e areeffects for years of education-a regression analysis revealed that educationaccounts for the same amount of variance as the effects of age of immigra-tion. In addition, the education effects are uniform across the life span andthere is no indication, as might be suggested by the critical period hypoth-esis, that it works differently in child and adult learners.

CONCLUSIONThe evidence for a critical period for L2 acquisition is scanty, especiallywhen analyzed in terms of its key assumptions. There is no empirically de-finable end point, there are no qualitative differences between child and


8/8

204 HA&adult learners, and there are large environmental effectson the outcomes.None of the conditions a re met hy the present research.

This is not to say that there are not any age effects for L2 acquisition.All studies show that there is a monotonic decline in ultimate attainment inL2 with age. Failure to find support ing evidence for a critical period meanstha t the view of a biologically constrained and specialized language acqui-sition device that is turned off a t puberty is not correct. Th e gradual declineover age in the ultimate attainment of an L2 means that there are multiplefactors at work-physiological, cognitive, and social. Researchers who wishto pursue the critical period hypothesis would need to become more spe-cific in their predictions, such as identifying the linguistic processes that areputatively shut down at the end of the critical period. Yet, it is incumbenton those who wish to stress the cognitive and social factors in L2 acquisi-tion to he equally specific n their predictions. Given the harsh implicationsof a critical period for policy and practice (i.e., not only that exposure isneeded early, hut also that exposure later in life is less valuable), the stan-dards of evidence in this area must he held high and educators and policymakers who pay attention to this research should demand no less.Beyond urging for the highest standards in research that bear on thistheoretical question, policy makers and practi tioners should also seek addi-tional information related to age and L2 instruction, such as1 .2.3 .4.

What capabilities are there for staffing different program options a tdifferent grade levels?Are students in different grade levels differentially responsive to tech-nolog) -supported language learning environments?In what ways does the relationship between language and contentchange as students progress through school?What specific resources and needs exist in the community for L2s thatcan help motivate students at different ages?

Informed answers to such questions can help guide local communities andstates in making decisions about optimal times for L2 insmction.REFERENCESBailey, N., Madden, C., & Krashen,S. 1974). s there a ~ t u r a lequence in adultsecond language learning?Lnnguage Learning, 24,235-243.Bialystok, E,,& Hakuta,K. (1994). n otherwordc: Thescienceandpsychohgy ofsecmd-languageacguintion.New York: Basic Books.Bialystok,E., & Hakuta, K. (1999).Confounded age: Linguistic and cognitive fac-tors in age differences n second language acquisition.In D. Birdsong (Ed.), Sec-

m d language acquisition and the criticalperiod hypothenr (pp. 161-181). Mahwah,NJ:Lawrence Erlbaum Associates.

A Citicd P e d or Semnd hngv.ge A ~ i i t i o n ? 205Birdsong, D. (1999). Introduction:Whys and why noB of the critical period hy-pothesis for second language acquisition. In D. Birdsong (Ed.), Second hnguageacyuisitionand the miticalperiod hypotheis (pp. 1-22). Mahwah,NJ: LawrenceErl-baum Associates.

Hanard Edvcatianl Review, 47,294-3 6.Hakuta, K, & Cancino, H. (1977).Trends in second-language acquisition research.Hnbel, D. (1988).Eye, brain,and vision.New York: Scientific American Library.Johnson, J., & Newport,E. (1989).Critical period effects in second language le an-ing: The influence of maturational state on the acquisition of English as a secondlanguage.CognitiuePsychohgy, 21 , 60-99.Lenneherg, E.H. (1967).B i o l o ~ c a l ~ u n d a t o f ~ n g u a g e . N e w Y o r k :ohn Wiley &Sons.New Jersey State Depamnent of Education. (1999).New Jersey World Lan-guages Curriculum Framework. Available: http://wMv.state.nj.us/njde~~me-workdworldlanguagedOdli, T. (1989). L n n p g e tru+er: Crm-linpinic influence in language learning.Cambridge, England: Cambridge University Press.Penfield, W., & Roberts, L. (1959).Speech and brain mechaninnr. Princeton, NJ:Princeton University Press.Pinker,S.(1994).Thelanguage imim.New York William Morrow & Company.Porter, R. (1998).Defedant declaratim.Valeria G. et al., v. Pete Wilson et al., No.C-98-2252-CAL, .S. District Court, Northern District of California.White, L., & Genesee, E (1996).How native is near-native? The issue of ultimateattainment in adult second language acquisition. Second Lnnguoge Research, 12 ,233-265.

Documents

(2001) - A Critical Period for Second Language Acquisition