Академический Документы
Профессиональный Документы
Культура Документы
11 Age
of Acquisition and
Success with a Second Language
11.1 Introduction
Are those two- and three-year-olds who speak two languages really budding geniuses?
$QGZK\DIWHU\HDUVRIVWXG\FDQW\RXVSHDN\RXUVHFRQGODQJXage
OLNH D QDWLYH QR PDWWHr how hard you try? This chapter considers questions and
answers that have to do with when a person becomes bilingual. First, we consider
children who are raised as bilinguals, basically from infancy. Then we consider others
who become bilingual at a later age and in generally a formal way through classroom
LQVWUXFWLRQLQDVHFRQGODQJXDJH$VZHOOVHHWKHWZRZD\VWREHFRPHELOLQJXDOKDYH
very different results.
7KHUHDUHWZRJHQHUDOTXHVWLRQVDQGWKHLUDQVZHUVWKDW ZHOOEHFRQVLGHULQJ)LUVW
why is the degree of bilingualism of child bilinguals different from that of speakers who
acquire a second language (L2) at a later age? That is, is there a cut-off point in language
acquisition so that second-language learning is less successful after a certain age?
Second, what are the factors that account for the difficulties and degree of success that
later learners have in acquiring an L2?
Bilingual child language acquisition generally refers to acquiring two or more
languages when exposed to them as a very yRXQJFKLOG)RUVXFKFKLOGUHQDFTXLVLWLRQ
means spontaneous learning with little or no obvious effort or instruction. However,
UHVHDUFKHUVZKRVWXG\ODWHUELOLQJXDOLVPDOVRXVHWKHWHUPDFTXLVLWLRQEXWWKH\XVHLW
to cover both bilingualism as a later event with no formal instruction and learning as an
event involving classroom teaching. In fact, the name given to the study of theories to
explain degrees of successful bilingualism beyond early childhood is second language
acquisition (SLA). One of the PDLQILJXUHVLQ6/$GHILQHV6/$DVWKHZD\LQZKLFK
SHRSOHOHDUQDODQJXDJHRWKHUWKDQWKHLUPRWKHUWRQJXHLQVLGHRURXWVLGHRIDFODVVURRP
5(OOLV,QWKLVFKDSWHUZHZLOOXVHDFTXLVLWLRQDQGOHDUQLQJVRPHZKDW
interchangeably when referring to this later bilingualism. Keep in mind, though, that
generally more explicit learning than unconscious acquisition seems to be involved in
ODWH ELOLQJXDOLVP :LWKRXW SLQQLQJ GRZQ WKH H[DFW DJH ZHOO XVH WKH WHUP late
bilingualism for a second langXDJHDGGHGWRRQHVUHSHUWRLUHDIWHUHDUO\FKLOGKRRG
We end this introduction by beginning to answer two questions many people have
about the relation between childhood language acquisition and later L2 acquisition.
First, is it possible to acquire these linguistic systems in two languages simultaneously
with native-like competence DVD\RXQJFKLOG"7KHDQVZHULV\HV&KLOGELOLQJXDOV
can acquire two (or even more) languages. They can do this as long as they are exposed
to them, although they tend eventually to develop dominance in one of them because
LWVXVHGPRUH1RZDQRWKHULQWURGXFWRU\TXHVWLRQ:K\GRHVQDWLYH-like acquisition of
DQ / VHHP VR GLIILFXOW DW D ODWHU DJH" 6RUU\ EXW WKHUH LVQW D FOHDU DQVZHU WR WKLV
question. True, near-native acquisition at a later age can occur for a few learners. But
QRWH WKDW ZH VD\ RQO\ QHDU-QDWLYH 9HU\ YHU\ IHZ OHDUQHUV ZKR VWDUW D ODQJXDJH
beyond early childhood achieve native-like competence. Whether there is a clear cut-off
age for true native-like acquisition and what that age is still remain controversial; but
evidence clearly shows there is a decline in the ability to achieve the same control of an
L2 that L1 speakers of that language have. We all know bright and motivated teenagers
or adults
who have studied a second language conscientiously and who may speak the studied
ODQJXDJHZHOOEXWFHUWDLQO\QRWDVQDWLYHVSHDNHUVGR,QWKHIROORZLQJVHFWLRQVZHOO
elaborate on the answers just given by presenting relevant research findings.
KRZ PXFK SDUHQWV RU RWKHUV WKLQN WKHLU WHDFKLQJ LV LQYROYHG LQ WKH VXFFHVV VWRU\
Understand that we are referring largely to pronunciation and grammar and a limited
vocabulary as child language acquisition. To learn anything like an adult vocabulary
takes longer and requires the experience that comes with age and possibly schooling.
Also, learning certain syntactic structures found in formal styles almost certainly
requires schooling.
Some readers may mistakenly think that children (or adults) who speak a non-standard
GLDOHFWRIDODQJXDJHKDYHQRWDFTXLUHGWKHODQJXDJHSURSHUO\7UXHWKHVHSHRSOHPD\
QRWKDYHDFTXLUHGWKHVWDQGDUGGLDOHFWRIWKDWODQJXDJHEXWWKDWVQRWWKHSRLQW&KLOGUHQ
acquire whatever variety of a language they are exposed to. That is, they are able to
produce and understand sentences that their community considers well-formed (i.e.
JUDPPDWLFDOLQWKDWFRPPXQLW\VYLHZ6RPH VSHDNHUVQHYHUDFTXLUHWKHVWDQGDUG
dialect, simply because they are not sufficiently exposed to it, or because they have no
desire to do so for many possible reasons. Many learn the standard dialect in school,
possibly replacing certain structures in their home dialect with these structures. All of
us slide along something like our own dialect continuum, meaning that we
unconsciously use certain pronunciations and grammatical forms in one situation, but
use them less or not at all in other situations.
a bilingual uses his or her two languages and possibly in how being bilingual affects
thinking abstractly. Also, even though the left hemisphere of the brain is the site of
language for both monolinguals and bilinguals, how WKHELOLQJXDOVGLIIHUHQWODQJXDJHV
are situated in that hemisphere seems to show variation across individual bilinguals.
More on these subjects follows.
provided some important results about how language is organized and accessed in
language processing.
spends more time with one caregiver than others. Second, even very young children
rarely hear only their own parents speak; that is, they soon become aware that one
language is more used in their community than the other and this can influence the
FKLOGVown use patterns.
<HWHYHQWKRXJKDFKLOGVGRPLQDQFHLQRQHRIKLVRUKHUODQJXDJHVVHHPVFOHDUWR
observers, dominance is extremely hard to measure in any objective way. And in
writings on bilingual children the term dominance has been used in different ways. Some
researchers use it to refer to more frequent use of one language over the other. Others
VD\ LW LV WKH ODQJXDJH LQ ZKLFK WKH FKLOG LV PRVW SURILFLHQW EXW KRZ LV WKDW WR EH
measured? One possible measure is to note which language is the source of the
morphosyntactic frame when the child produces utterances with words from both
languages (cf. chapter 9 on the MLF model of codeswitching). But to our knowledge,
no one has used this measure. Problems with assessing dominance also plague studies
of adult bilinguals; often the researcher just asks the participants what they think their
dominant language is.
10
from one lexical system, to two such systems but one set of syntactic rules, to finally
WZRVHSDUDWHODQJXDJHV5HVHDUFKHUVVSRNHRIIXVLRQRIWKHFKLOGVWZRJUDPPDWLcal
V\VWHPV%XWDV'H+RXZHUSRLQWVRXWWKHIXVLRQRUK\EULGYLHZGLGQRW
offer an explanation of how children eventually manage to differentiate their languages.
7KRVHUHVHDUFKHUVZKRDUJXHWKDWWKHFKLOGVWZRODQJXDJHV\VWHPVDUHIXVHGKDYH to
come up with an explanation for how they eventually become divided.
As the 1990s began, a number of studies had just appeared or were appearing with
H[WHQVLYH GDWD VXSSRUWLQJ 'H +RXZHUV UHVHDUFK HJ 'HXFKDU DQG 4XD\
Genesee et al., 1995; Lanza, 1997; Meisel, 1989). Her Separate Development
H ypothesis basically states that very young bilinguals employ two separate grammatical
systems, one for each language, when they speak their two languages.
11
(codeswitching) is systematic. In this case, to say it is systematic means that within the
VDPHFKLOGFODXVHDVLQJOHXWWHUDQFHWKHJUDPPDWLFDOSDWWHUQVRIRQO\RQHODQJXDJH
DUH DOORZHG $V ZHYH QRWHG EHIRUH HDFK ODQJXDJH KDV LWV RZQ ZHOO-formedness
FRQVWUDLQWVRQKRZWRRUJDQL]HZRUGVLQDFODXVH
12
When French clitics (particles standing for pronouns, used to indicate subjects and
objects of the verb) were required for a well-formed French clause, the children
generally used them even when the clause had words from both languages. For example,
one child said fish, il mange pas ILVK KH GRHVQW HDW ,Q WKLV H[DPSOH fish can be
FRQVLGHUHG D WRSLFDOL]HU $V D FRQWHQW ZRUG LW FDQ FRPH IURP HLWKHU WKH 0DWUL[
Language or the other language (the Embedded Language). But the order of the verb
and the negative element (pas) follow French rules. (If this utterance were following
English order, then the negative element (pas) would come before the verb.) There are
some examples in which the clitic is not produced as it would be in French, but
YLRODWLRQVZLWKUHVSHFWWRSURQRPLQDOVFRPSrise just 13 percent of all mixed utterances
ZLWKSURQRPLQDOVS$OVRWKHUHVHDUFKHUVYLHZRIZKLFKPRUSKHPHVYLRODWH
the MLF model does not always correspond to that of the original model.
Still, the general implication of these findings is very important. The study indicates
that, even at a very young age, children follow the appropriate grammatical
constraints when they engage in codeswitching within the same clause. The data do not
support a grammatical deficiency hypothesis indicating that codeswitching was
unconstrained before the use of certain grammatical structures, as proposed by Meisel
(1994). In addition, the results of this Paradis et al. study also support the notion of a
dominant language as controlling the grammatical frame of a mixed constituent
(according to their definition). Further, this study supports the claim that children
possess and can use language-specific syntactic knowledge, even at this young age.
The issue of why children engage in codeswitching is still being debated. Genesee
(2001) cites support for a proficiency-based hypothesis (e.g. children are filling lexical
JDSVLQWKHLUNQRZOHGJHZKHQWKH\LQVHUWZRUGVIURPRQHODQJXDJHLQDQXWWHUDQFH
basically in the other language, or they are filling grammatical gaps when they use a
JUDPPDWLFDOIUDPH IURPWKHLUPRUHGRPLQDQWODQJXDJH%XWWKLVLVDFRQWURYHUVLDO
claim. Of course, if a child is not exposed to codeswitching (from parents and others),
he or she may not use two languages in the same clause. Even then, the child could
develop codeswitching spontaneously.
13
14
In addition, Siri also demonstrated an awareness that when all three (parents and
child) were together, speaking both languages is appropriate depending on the
interaction. Here is an example when Siri is at the age two years and three months and
she is eating dinner with her parents with the talk about a skiing trip when both parents
had fallen.
(1)
Siri Parents jet falt
I fell
F: Ja, du falt du g.
Yes, you fell, you too.
M: You went all the way back past our house And to the
garage. On skis today.
fall
M: Mm. You fell sometimes.
nesten/nesten/nesten almost/almost/almost
F: Nesten? Almost
ja
yes
M: Almost what?
almost fall
(Lanza, 1997: 298)
Another study that seems to show the effects of parental input dealt with four
Japanese-English pre-school bilingual boys whose production of the minority language
-DSDQHVHLQFUHDVHG.DVX\D7KHFKLOGUHQZHUHUDLVHGZLWKWKHRQHSDUHQWRQH
langXDJH SDWWHUQ EXW WKH\ OLYHG LQ DQ HQYLURQPHQW LQ ZKLFK WKH\ ZHUH H[SRVHG WR
English more than Japanese (they were raised in the United States). In fact, generally,
they were exposed to Japanese only at home. Still, the parents wanted the children to
retain Japanese past their early years so the Japanese-speaking parent always spoke
Japanese to the children, even though the parent was bilingual in English.
Three results from this study about the effect of parental language choices are
noteworthy. First, there was a relation between what the parent spoke to the child and
the language in which the child responded. Second, the children whose overall use of
Japanese increased during the course of the study came from the families in which the
parents used the moVW -DSDQHVH WR WKH FKLOGUHQ 7KLUG 7KH FKLOG ZKR JRW WKH OHDVW
Japanese in both absolute and relative terms was showing the lowest relative use of
-DSDQHVHFRPSDUHGWRWKHRWKHUFKLOGUHQS
15
Very young children develop their two or more languages as separate grammatical
systems.
These young bilinguals are able to use their languages appropriately according to
their addressee.
They may develop discourse preferences and strategies based on the input they
receive, although not all results support how input matters.
$FKLOGVFRGeswitching does not reflect an incapacity in the language faculty.
When children use both of their languages in the same utterance, the evidence points
WR WKH FKLOGV XVH RI RQO\ RQH DV WKH VRXUFH RI WKH JUDPPDWLFDO IUDPH IRU WKDW
utterance.
16
17
an overall deficit in the vocabulary of bilingual children, but some note that bilinguals
displayed better verbal fluency.
Again, studies that compare development on well-formedness conditions on sentence
formation (syntax) in their two languages show mixed results. In this test, children were
asked if various strings of words could be well-formed sentences in language X or
language Y. Gathercole and Montes (1997) compared Spanish-English bilinguals with
English monolinguals; all were seven and nine years old. Different types of educational
programs and different social classes were represented, but apparently children and their
test scores were not grouped in terms of these variables.
For all the structures studied, the bilingual children lagged behind. But the older
bilinguals were doing better. The older and stronger bilinguals resembled
monolinguals in their syntactic sensitivity. Across both older monolingual and bilingual
children, the progression in mastering structures was the same and the structures were
learned in the same order. The difference was that children learning two languages
took longer to learn the subtleties of syntax than did children concentrating on only
one language. How much exposure the bilinguals had to English or Spanish seemed to
matter; at least the children with more exposure did better in recognizing acceptable
structures in the language being tested.
18
their awareness that in a game one symbol can be substituted for another. For example,
in the game, children were told that the word spaghetti was to be substituted for the
word we. Children were told this new meaning of spaghetti and then were asked to use
this word in a sentence to indicate we are good children. The response called for would
be spaghetti are good children. Bilinguals did significantly better in making such
substitutions than monolinguals.
Galambos and Hakuta (1988) compared bilinguals and monolinguals in making
JUDPPDWLFDOLW\MXGJPHQWVLHDQVZHULQJVXFKTXHVWLRQVDVFDQ\RXVD\WKLVVHQWHQFH
LQ(QJOLVK"DQGLQGHWHFWLQJDPELJXLWLHVLQVHQWHQFHV$PELJXRXVVHQWHQFHVFDQKDYH
two meanings; e.g. visiting relatives can be a nuisance , caQ PHDQ HLWKHU WKH DFW RI
YLVLWLQJUHODWLYHVLVDQXLVDQFHRUUHODWLYHVZKRYLVLWDUHDQXLVDQFH7KH\IRXQGWKDW
bilinguals had a consistent advantage over monolinguals in the syntax task, but only
older children were better than the monolinguals on the ambiguity task.
Finally, a study in Stockholm seemed to indicate bilinguals have advantages in similar
tests (Cromdal, 1999). Schoolchildren whose stronger language was English and who
had at least a basic knowledge of Swedish were compared with monolingual English
speakers to test a number of metalinguistic benefits between monolinguals and
bilinguals. A total of 40 bilinguals and 16 monolinguals were selected between the ages
of 6.4 and 7.3, plus two 8-year-olds. Among other results, the bilinguals did better in
recognizing ungrammatical sentences in either English or Swedish and in correcting
them. They also were more accepting than English monolinguals of marked, but
acceptable, ways of constructing a well-formed sentence.
Probably the most influential study of the cognitive effects of bilingualism was the
early study of Peal and Lambert (1962). The researchers gave monolinguals and
bilinguals a battery of tests to assess difference in the structure of intelligence. They
found that bilinguals were especially superior in the area of symbolic
reorganization, leading the researchers to say that bilinguals showed more mental
flexibility. There were problems with this research, as there often are with any study.
For one thing, the subjects came from a select group of francophone families in
Montreal, the type of residents who were also very proficient bilinguals in English.
Children from these families were compared with monolingual francophones who
probably had less exposure to education and other advantages. Still, this was an
important study because it awakened educators and other researchers to the notion that
in the important area of ways of processing information bilingualism might be an
advantage, not a disadvantage.
19
20
foreign features; these included participants who claimed to be balanced bilinguals and
both groups of late bilinguals (low or high proficiency in Italian) as well as early
bilinguals who were dominant in Italian (cf. an overview in Harley and Wang, 1997).
21
of late L2 learners than early bilinguals. Clearly, vocabulary learning for child bilinguals
differs a great deal because the acquisition of word form and the word meaning proceed
in parallel, just as they do for monolingual children. In later L2 learning, a word has to
be learned for a meaning that is already in place (with some adjustment).
Vocabulary has a critical role in L2 learning; after all, if they need to use their L2,
VSHDNHUV FDQt get these needs met unless they can talk about them in the L2.
)XUWKHUPRUH LWV EHHQ FODLPHG WKDW QDWLYH VSHDNHUV FDQ EHWWHU XQGHUVWDQG
ungrammatical utterances if they contain appropriate words than if the utterances have
flawless grammar but the wrong words.
But learning words in an L2 can be a daunting task. First, the majority of words in
DQ\ODQJXDJHKDYH PXOWLSOH PHDQLQJV )XUWKHUWKHVH PHDQLQJVDUHQW VWDEOHEXWFDQ
change over time. Second, to learn each word, a number of types of information have to
be learned, all in addition to its possible meanings. These include how the word is
pronounced, how it is spelled, how it can fit into a syntactic unit, how it can combine
with other words or morphemes to make a larger unit, and what its connotations are
when used in that unit.
Vocabulary items are almost always learned as explicit learning (i.e. declarative
memory is involved). That is, even the young children who acquire their grammar and
pronunciation effortlessly have to at least pay overt attention to learn new words. Several
studies indicate that knowing a relatively small number of words is sufficient for
VSHDNLQJDQ/VHH+D]HQEHUJDQG+XOVWLMQIRUDQRYHUYLHZ,IRQO\DIHZ
thousand words are needed, then explicit vocabulary learning is a reasonable goal.
When the L2 vocabulary is taught in a formal setting, various methods are in use.
:HOORQO\ORRNDWRQHRIWKHPRVWVXFFHVVIXOPHWKRGV,WVFDOOHGWKHkeyword method.
The two-step process is this. The learners associate a new word with a keyword in their
/WKDWLW VRXQGV OLNHRU ORRNVOLNH7KHQWKH OHDUQHUFUHDWHV a mental image that
somehow includes both the new word and the keyword. So, for example, if the Englishspeaking learners are studying Chinese and the new word is ga oxng KDSS\ WKH
learner can associate it with glad and then create an image with the learners feeling
JODGEHFDXVHVRPHWKLQJKDSSHQHGWRPDNHWKHPJDR[LQJRUKDSS\
This may sound like a lot of effort, but the method seems to work. A number of studies
show that learners using the keyword method outperform learners using any other
strategy. For example, one study with striking results looked at a 47-year-old university
lecturer who learned Italian via this method (Beaton et al., 1995, cited by De Groot and
Van Hell, 2005). After ten years, he was tested on his retention of a vocabulary of 350
words even though he had used no Italian in the meantime. The lecturer remembered 35
per cent of the words (about 122 words) he had learned and after ten minutes of
relearning, he added an additional 93 words.
Theoretically, the keyword method is of interest because it shows the role of
imaging. The idea is that the keyword method promotes learning because it
22
uses both the language system and the image system in human memory (Paivio and
Desrochers, 1981). However, the few studies that have compared how long it takes a
subject to retrieve a word learned via the keyword method with learning via just simple
rote learning show that it takes the keyword learner longer to access a word (Van Hell
and Candia Mahn, 1997). So, if fluency is a goal the keyword method may not be the
best to follow.
The goal of vocabulary learning, obviously, is to retain the words learned. De Groot
and Van Hell (2005) report that studies show that more forgetting occurs for the types
of words that are more difficult to learn in the first place. Evidence shows that concrete
words are easier to learn than abstract words and also are easier to DQFKRUWLHWRRWKHU
words). Word forms that have familiar pronunciation patterns are easier to learn than
words that sound more foreign. This is important because it means that cognate words
(words that have similar forms and meanings in another language the speaker knows)
are easier to learn. After they are learned, concrete words and cognates are also easier
to retain than non-cognates and abstract words. Unfortunately, this also means that
learning words in a language that is totally unrelated to your L1 are much harder to learn.
When your author studied Swahili, learning that mti ZDVWKHZRUGIRUWUHHDQGWKDW
nyumba ZDVWKHZRUGIRUKRXVHZDVKDUGHUWKDQOHDUQLQJWKDW jardin in French meant
JDUGHQRUWKDW stuhl LQ*HUPDQPHDQWFKDLU:KHWKHUthe actual learning conditions
themselves have any effect is another issue and presents something of a puzzle. De Groot
and Keijzer (2000) showed that words learned under difficult learning procedures
actually may be better retained than those same words when learned under easy
learning conditions.
De Groot and Van Hell present evidence that suggests that for lowproficiency L2
learners, accessing the meaning of a word is YLD D URXWH WKURXJK WKH ZRUGV /
counterpart0RUHSURILFLHQWOHDUQHUVGRQWKDYHWRGo this, but make direct connections
between a word and its meaning.
23
24
universities. Birdsong and Molis also found a significant decline in performance on test
sentences as age of arrival to the US increased.
But the difference is that some of their late ar rivals scored high. Four participants
who arrived at or after age 17 scored high on the same type of acceptability judgment
test used in the Johnson and Newport study. These speakers were accurate on 94.8% of
the test sentences; on these same sentences, the mean accuracy of native speakers was
95.2%. Interestingly, Birdsong and Molis also point out that among all their subjects the
amount of current English use was a strong predictor of performance. Obviously, one
can claim that the four late arrivals showed native-like results even though they seem to
have had no or little exposure to English before any so-called critical age.
However, Birdsong (2005) makes the point that his argument with claims about a cutoff age is mainly with the strong version of a Critical Period Hypothesis. The strong
version is that there is a period when acquisition always happens and after which similar
SURILFLHQF\LQWKH/FDQWEHUHDFKHG%XWLIWKHUHZHUHDQDFWXDOFXW-off point, then
late learners ought not to show the variation that they do in how well they perform in an
L2.
25
Another study looked at quite a different aspect of age effects and bilingualism. Izura
and A. Ellis (2002) considered the relation of age of acquisition of an L2 and
performance on word recognition experiments. Many earlier studies show that, all
other things being equal, words learned early in life can be recognized and produced
faster than later-learned words. This effect is found in a number of different types of
tasks (e.g. object picture naming) and seems to be independent of other factors such as
frequency of use of words in adult language. Most previous studies with this format
have dealt with monolingual subjects.
Izura and Ellis devised studies to see if a second language acquired in late childhood
or adulthood showed the same effects. Their study included Spanish participants who
acquired Spanish as their L1 in Spain and who then learned English later. These
participants were introduced to English at the age of 8 to 10 in school and then they
studied English after that throughout their schooling. They were living in England as
university students when tested. In one experiment, the subjects named objects in
pictures whose names were learned early or late in both Spanish and English. The words
were matched for frequency, object familiarity, and word length.
This study showed two main results. First, response times were faster for words
acquired in their first language (Spanish) than in English (their L2). Second, response
times also were faster for earlier acquired words than later acquired ones, but overall
were faster for Spanish words. In a second experiment the participants had to carry out
a lexical decision task in which they were asked to distinguish words in both English
and Spanish from stimuli that were non-words in both languages. Again, results showed
that the order of acquisition was important for both Spanish and English words, but
responses were faster for Spanish.
26
GLIIHUHQFHV7UXHZHKDYHGRFXPHQWDWLRQDV\RXOOVHHLQRQKRZVRPHDVSHFWV
of language seem to be represented in the brain differently in individuals, depending on
age of acquisition of an L2.
Even though we now have imaging studies, DeKeyser and Larson-Hall point out that
there is no evidence to date to link the learning problems of late L2 speakers to actual
changes LQQHXURSK\VLRORJLFDOPHFKDQLVPVLQWKHEUDLQ7KLVGRHVQWPHDQDOLQNFDQW
EHIRXQGMXVWWKDWLWKDVQWEHHQIRXQG.
27
8OOPDQV K\SRWKHVLV LV EDVHG RQ H[WHQVLYH HYLGHQFH IURP LPDJLQJ WHFKQLTXHV DQG
other sources that the two types of memory are located in different areas of the brain.
For example, one recent study of language impairment cited by Ullman supports this
claim (Ku, Lachmann, and Nagler, 1996). The person studied was a right-handed 16year-old Chinese-English bilingual who developed a type of encephalitis that involved
his left temporal lobe. His L1 was Chinese but he had been living in the United States
since age 10. He lost his ability to speak and comprehend English, but retained his ability
in Chinese, although he produced only relatively short sentences.
Imaging techniques produce accurate images of the brain region involved in this
disease (the region just mentioned above the left temporal lobe). Other imaging
evidence indicates that procedural memory is located in the left frontal lobes and the
basil ganglia areas, but declarative memory is located in the left temporalparietal/medial-temporal lobes. So it was the site of declarative memory that appeared
to be most affected in this patient. The implication is that his problems with English,
while retaining Chinese, meant that English (his L2) was situated in declarative memory.
After all, it was the left temporal lobe that seemed affected by his disease. In contrast,
the fact that he could still produce Chinese is implicational evidence that the
grammatical aspects of Chinese were situated in procedural memory (since the supposed
sites of procedural memory were not affected by the disease).
The relevance to the issue of L2 acquisition of this discussion about areas of the brain,
selective brain damage, and types of memory is that possibly procedural memory is
QRORQJHUDYDLODEOH for language learning past eDUO\FKLOGKRRG:K\LWVQRORQJHU
available is another issue.
In normal L2 learners (with intact brains), the areas of the brain that seem to support
SURFHGXUDOPHPRU\DUHVWLOOWKHUHEXWLWVSRVVLEOHWKDWWKHVHDUHDVKDYHXQGHUJRQH
various changes or that certain relevant neural procedures are no longer available. But
if Ullman (and others) are correct in assuming that the problem with achieving a measure
of success with an L2 is a problem regarding the type of memory (declarative memory)
that is still available to the L2 learner, this finding would be very important to instructors
of L2 in designing their curriculum. Obviously, it would be important for instructors to
emphasize explicit learning of structures that are assembled implicitly by the L1 learner.
The need for SLA practitioners and theorists to consider possible differences in language
processing is raised again in 11.8.
28
29
different results in L2 learning found across studies. We will only report their six general
conclusions. First, they make it very clear that such studies indicate that the bilingual
brain is definitely not the sum of two monolingual language systems. Second, there
are individual differences across bilinguals in the organization of the neural system.
For example, imaging studies do show some activation in the right hemisphere for
some individuals for some tasks. Third, of the variables that seem to affect the bilingual
V\VWHPWKHVHUHVHDUFKHUVDWOHDVWFRQFOXGHWKDWRQWKHEDVLVRIWKHDYDLODEOHHYLGHQFH
[degree of] proficiency VHHPVWREHWKHPRVWLPSRUWDQWRQHS1RWHWKDWWKLVLV
a different conclusion from that of DeKeyser and Larson-Hall (2005) who place most
importance on age of acquisition. Further, other researchers give more credit to other
factors. Fourth, it seems that comprehension and production involve different levels of
FHUHEUDODFWLYDWLRQLQWKHELOLQJXDOVOHVVSURILFLHQWODQJXDge; there is more activation
associated with production. Bear in mind that the majority of the neuroimaging
experiments on language production in bilinguals are based on single word processing
(producing a word according to a cue). Their survey also shows that language
comprehension in bilinguals (as studied via neuroimaging) appears to give more
consistent results than production studies.
Fifth, the researchers consider the role of age of acquisition and conclude that it is
less important than other factors. But they qualify this statement by saying that this
DSSOLHV WR ELOLQJXDOV ZKR VKRZ WKH VDPH SURILFLHQF\ +HUH LV ZKDW WKH\ VD\ :KDW
functional imaging has shown is that, when proficiency is kept constant, age of
acquisition per se does not seem to have a major impact on brain representations of L2,
DWOHDVWDWWKHPDFURVFRSLFEUDLQDUHDOHYHOS
It seems as if Abutalebi et al. partly base this conclusion on the fact that a study by
Chee et al. (1999) of early and late bilinguals found little difference when the subjects
were producing words. The participants came from Singapore. Abutalebi et al. reasoned
that this was a society in which bilingual speakers are often highly proficient in both
WKHLUODQJXDJHV7KXVWKHSDUWLFLSDQWVEDFNJURXQGVuggests that proficiency, not age of
acquisition, is important. But these results contrasted with those in a study by Kim et al.
(1997) of 12 proficient bilinguals, some early and some late (with their language pairs
including diverse languages). The task the participants performed was to tell what they
had done on the previous day. Kim and his associates found that in late learners, the L1
DQGWKH/ZHUHUHSUHVHQWHGLQGLIIHUHQWSDUWVRIWKHOHIWLQIHULRUIURQWDOFRUWH[%URFDV
area). In contrast, overlaSSLQJSDUWVRI%URFDVDUHDZHUHDFWLYDWHGIRUERWKODQJXDJHV
in early learners.
Finally Abutalebi et al. concluded that any evidence of the role of degree of exposure
to a language is too limited to draw any conclusions. They recognize that there are
important questions that such research does not answer, such as how people learn and
acquire a second language.
30
31
true, at least for the languages studied. The structural differences between Italian and
English are much greater than those between Spanish and Catalan, and the differences
GRQWVHHPWRPDWWHUUHJDUGLQJEUDLQDFWLYLW\+RZHYHULWLVWUXHWKDWWKHGLIIHrences
EHWZHHQ ,WDOLDQ DQG (QJOLVK DUHQW VR JUHDW WKDW WKH\ RIIHU DV JRRG D WHVW DV UHDOO\
different languages, such as Vietnamese or Yoruba, compared with Italian.) Still, Perani
HWDOFDQSRLQWWRVWULNLQJO\ VLPLODUSDWWHUQRIDFWLYDWLRQREVHUYHGIRU L1 and L2 in
Italian-English and in Catalan-6SDQLVKKLJKSURILFLHQF\ELOLQJXDOVS2YHUDOO
Perani et al. (1998) seem to conclude that the degree of mastery of the L2 is what makes
the difference between the low- and high-proficiency bilinguals in regard to their brain
scanning.
Other studies have similar results with similar implications. In light of such evidence
and other neuroimaging studies, Green (2005) suggests a convergence hypothesis. The
hypothesis states that as L2 proficiency increases, its representation (in the brain) and
its imaging profile converge with that of native speakers of the relevant language. For
an imaging profile, Green refers to PET and fMRI and ERPs (Event Related Potentials
that are recordings of cerebral electrical activity during cognitive tasks). Green makes it
clear that this is not a claim that an L2 speaker ends up with native-like speech, but
rather that neural processing can change. That is, he suggests that some reorganization
may take place if, for whatever external reasons, the speaker moves toward more
fluency in the L2.
32
33
foremost in the minds of students of second languages. You can imagine having no good
DQVZHUV IUXVWUDWHV ODQJXDJH FODVVURRP WHDFKHUV WRR ,Q WKLV VHFWLRQ ZHOO FRQVLGHU
FXUUHQW DWWHPSWV WR FRPH XS ZLWK VRPH DQVZHUV 7KDW LV ZHOO ORRN VSHFLILFDOO\ DW
theories to explain why SLA is as successful (or not) as it is. Certainly, there is a
tremendous amount of research going on today, but we warn you that more questions
than answers remain.
Recall that we included under second language acquisition any learning of a second
language at a later age than when the first language was acquired. But here we will focus
on studies of SLA as formal language learning, generally in the classroom and generally
past a very young age.
34
35
SXUH FRQQHFWLRQLVP WKH YLHZ WKDW JHQHUDO OHDUQLQJ PHFKDQLVPV HQDEOH KXPDQV WR
acquire language, given the necessary environment and motivations based on the
functions language can serve. Connectionism per se is a processing model. It emphasizes
that struFWXUDO UHJXODULWLHV HPHUJH IURP WKH ODQJXDJH OHDUQHUV DQDO\VLV RI OLQJXLVWLF
input, driven by the frequency and complexity of the formfunction relationships that
HPHUJH LQ WKH LQSXW 7KXV DV 1 (OOLV SXWV LW WKH NQRZOHGJH RI WKH VSHDNHUKHDUHU
cannot be understood as an innate grammar, but rather as a statistical ensemble of
language experiences WKDWFKDQJHVVOLJKWO\HYHU\WLPHDQHZXWWHUDQFHLVSURFHVVHG
(p. 64).
Basically, connectionists claim that, with computer-based models of language
learning, they can model the process that a language learner goes through to learn to
distinguish such forms as past tense version of English irregular and regular verbs.
Rumelhart and McClelland (1986) presented the first connectionist model of this
construction. TKH LGHD RI FRQQHFWLRQLVWV LV WKDW WKHUH LV QR UXOH EXW RQ-line
JHQHUDOL]DWLRQV0RVWZRUNWRGDWHKDVFRQFHUQHG/DFTXLVLWLRQ*UHJJSRLQWV
out that a problem with connectionism is that it has no theory of what it is the learner
has acquired, partly because no final state is posited.
,QWKHIROORZLQJVHFWLRQVZHOORQO\KDYHPRUHWRVD\DERXWWKHWZRPDLQJURXSVRI
6/$UHVHDUFKHUV)XUWKHUZHOOOLPLWGLVFXVVLRQWRPRUSKRORJ\DQGV\QWD[HYHQWKRXJK
there are many studies on phonology, too. For more on the entire field of SLA research
see these edited volumes: Archibald (2000), Doughty and Long (2003), Eubank,
Selinker, and Sharwood Smith (1995), Gass and Schachter (1989), and Mitchell and
Myles (1998), as well as the textbook by R. Ellis (1994). In addition, there are many
new texts every day on related issues, such as an edited volume by Cook (2003) titled
Effects of the Second Language on the F irst . Obviously, SLA research is a robust, fastmoving field of study.
36
backgrounds, indicate that the progression from single content words to more complex
clause structures and the use of inflections largely mirrors that of L1 acquisition. The
first verb form the immigrants used in their L2 was the present progressive form (the ing verb, as in boy singing). This is also the first grammatical morpheme that Roger
Brown (1973) found was acquired by three children from different backgrounds who
were acquiring English as their first language. Also, as Nick Ellis (2005) points out,
7KHIDFWWKDW/OHDUQLQJLVLQIOXHQFHGE\WUDQVIHU>RIVWUXFWXUHV@IURP/PHDQVWKDWD
PRGHORI6/$PXVWWDNHLQWRDFFRXQWWKHDFTXLVLWLRQDQGVWUXFWXUHRI/%XWDJDLQLI
there are resemblances between acquiring an L1 and an L2, why are the results so
different?
Thus, the subject under study here is the interlanguage that develops in L2 learners.
Selinker (1972) coined this term. Interlanguage refers to the linguistic systems (note the
plural) that an L2 learner constructs as he or she progresses toward the target variety. At
any stage, speakers have a system of systematic rules, but they are not the same thing as
WKRVHRIWKHOHDUQHUVWDUJHWZKLFKLVSUHVXPDEO\DVWDQGDUGGLDOHFWRIWKH/DQGZKLFK
most L2 learners never reach completely. As L2 learners progress toward their target,
their interlanguages change and become more like the target.
37
38
-DSDQHVH@ S :KLWH KHUVHOI KDV FRQGXFWHG D QXPEHU RI VWXGLHV HVSHFLDOO\ RQ
French-speaking learners of English and their abilities to correctly recognize the
impossibility of certain structures in English.
39
Results did not support the notion of full access to UG in adult L2 learners; the data
from the Korean participants indirectly supported the notion of access only if the
OHDUQHUV ODQJXDJH KDG WKH IHDWXUH EHLQJ WHVWHG +HUH ZHUH WKH UHVXOWV )LUVW QDWLYH
speakers of English also were tested as a control group and their performance indicated
they knew the construction being tested and which test sentences violated constraints on
it. Second, only about one-third of the non-native English speakers showed knowledge
of the construction being tested and its violation. Korean does not have this type of
movement restriction and the Korean participants performed the least well. Thus, the
Korean data support the notion of partial access to U G at best. The Chinese and
Indonesian subjects performed better on the test sentences and both of their languages
have some movement rules that can be considered constraints on Subjacency. Still,
neither group performed well enough to give strong support to the position that UG is
available to adult learners.
40
VSHFLILFFODLPVRIWKHSURSRQHQWVRI8*2*UDG\FDOOVWKHDSSURDFKKHIDYRUVgeneral
nativismDOVRVHH&URIW /DQJDFNHU6XFKDQDSSURDFKLQFOXGHV VRPH
sort of innate acquisition system that guides the learner and interacts with experience.
But the important feature of this approach is that it rejects the specific, detailed
constraints as part of UG that proponents of the role of UG in SLA seem to see. That
is, theories of this type hold the view that the mind is geared toward processing
information, but linguistic information is just one type.
Disagreements with theories based on UG seem to center on the question of whether
specific grammatical structures and categories are innate. A shortcoming of general
QDWLYLVP DFFRUGLQJ WR 2*UDG\ KLPVHOI LV WKDW LW GRHVQW LQFOXGH DQ\ WKHRU\ RI
learnability and development. (The claim of access to UG is, after all, such a theory.)
Note that some connectionists claim that proponents of cognitive linguistics, such as
Langacker mentioned above belong in their camp, not with proponents of any form
of UG at all.
41
behaviorism7KLVDOOFKDQJHGE\WKHVLQLWLDWHGE\&KRPVN\VHPSKDVLVRQWKH
essential part of language as abstract systematicity that is, production governed by
DEVWUDFWUXOHVWKDWVSHDNHUs hold in their minds. But for the moment, instructors, along
with any researchers there were, believed that success in language learning, like in any
other kind of learning, came largely from forming habits and the related notion that
practice ma kes perfect.
In this vein, two main approaches were advocated to explain successful L2 learning.
The first was contrastive analysis. The idea was that by analyzing the structure of the
OHDUQHUV ILUVWODQJXDJH DQGFRPSDULQJLW ZLWKWKH WDUJHWVHFRQGODQJXDJHLQVWUuctors
could work out the problem areas for the learner. But you can imagine that the task of
comparing the structures of any two languages was a huge task. Further, because of the
prevailing view that the structure of language lay only in its surface, linear orders, a real
FRPSDULVRQ RI WZR ODQJXDJHV VLPSO\ ZDVQW PDGH EHFDXVH WKH DEVWUDFW QDWXUH RI
language was not considered).
The second approach that took over in the 1970s was called er ror analysis. Now, at
least, the interest had shifted to the language produced by the learners themselves. But
working out why learners make the errors they do is not easy. The underlying
assumption was still that errors occurred because learners were using structures from
their L1. The notion of interference or transfer from the L1 was part of this approach.
$V 5 (OOLV SRLQWV RXW ($ >(UURU $QDO\VLV@ SURYLGHG D PHWKRGRORJ\ IRU
LQYHVWLJDWLQJOHDUQHUODQJXDJH7KDWLVLQVWUXFWRUVFRXOGLGHQWLI\HUURUVDQGXVHWKHP
to see what areas needed to be emphasized in teaching) and researchers could use them
to attempt to explain them and the L2 learning process as a whole. However, both of
WKHVH DSSURDFKHV VWLOO HPSKDVL]HG FKDQJLQJ KDELWV )XUWKHU HUURU DQDO\VLV GRHVQW
necessarily mean that the source of errors is identified. As for the L1 as a source of
errors through transfer, various studies claimed that anywhere from 3 to 51 per cent of
errors could be attributed to the L1; but Mitchell and Myles (1998: 30) cite a survey by
Ellis of studies of errors that shows that only about a third of all errors are traceable to
the L1.
42
43
the learner learns one form, grammatical information can be held in a grammatical
memory storage, which is highly task-specific. Basically, then, what the learner (L1 or
L2) has to do is come up with a new conceptualization that puts together routines in that
memory storage.
For example, in order to handle subjectverb agreement (putting the -s on third person
singular present tense verbs), the English learner has to have acquired the notion of
feature unification. This is done in noun phrases where the learner recognizes the
IHDWXUHRIVLQJXODURUSOXUDOLQUHODWLRQWRQRXQV%XWWKHQIRUVXEMHFWverb agreement,
the learner has to match features across constituents (from the NP to the VP). This
interphrasal matching is not predicted before matching within the phrase (e.g. within the
NP). (Note some similarities between the divisions of the 4-M model discussed in
chapter 9 in relation to language contact and the divisions between affixes depending on
intra-phrase information (early system morphemes) and those outsider late system
morphemes that always depend on cross-phrasal information.)
7KHUHDUHVLPLODULWLHVEHWZHHQ3LHQHPDQQVPRGHODQGVHYHUDORWKHUDSSURDFKHV6WLOO
EH VXUH WR QRWH WKDW QHLWKHU 3LHQHPDQQV QRU WKH RWKHUV LV WKH VDPH WKLQJ DV
connectionism, which does not see the learning of rules as underlying the construction
of linJXLVWLFNQRZOHGJH)RUH[DPSOH3LHQHPDQQVDSSURDFKHFKRHV.UDVKHQV,QSXW
+\SRWKHVLV.UDVKHQDUJXHGWKDWWKHOHDUQHUFDQRQO\DFTXLUHcomprehensible
inputWKDWFDPHDERXWWKURXJKLQIHUHQWLDOOHDUQLQJZKHQWKHOHDUQHUDFKLHYHGDFHUWDLQ
stage. WKLOH .UDVKHQV PRGHO ZDV YHU\ SRSXODU LQ WKH V LW ZDV DOVR FULWLFL]HG
because of its vagueness, it was not clear how it could be tested (and therefore supported
or rejected). Another complementary theory is the Autonomous Induction Theory of
Carroll (2000). Briefly, her idea is that what a learner can take in is limited by the
grammar (interlanguage) the speaker has already constructed of the target language. Van
Patten (1996) also distinguishes input and intake, arguing that a set of strategies (under
WKHUXEULFDWWHQWLRQLVQHFHVVDU\WRWXUQLQSXWLQWRLQWDNH5HFDOOERWK.UDVKHQVYLHZV
PRUHRUOHVVIDYRULQJLPSOLFLWOHDUQLQJDQG9DQ3DWWHQVYLHZVPRUHRUOHVVIDYRULQJ
both explicit instruction and implicit learning) when we discuss these two types of
OHDUQLQJLQVHFWLRQ3LHQHPDQQVSURSRVDOVVHHPHVSHFLDOO\RSHQWRWHVWLQJ,Q
fact, Pienemann (2003: 695) reports on a number of studies of L2 learners of English
whose progress in learning English structures follows the implicational table that he
predicts for English.
44
or classroom-EDVHGWKHRUHWLFDODSSURDFKHV6KH ZULWHV(YHQLID8*H[SODQDWLRQRI
SLA were to prevail, the elements of language that are governed solely by UG are
limited. Much more of the L2 remains which is potentially acquired more efficiently
SURYLGHG LQVWUXFWLRQ DSSURSULDWHO\ HQJDJHV OHDUQHUV FRJQLWLYH SURFHVVLQJ DELOLW\ S
258).
As she points out, a major division today in SLA revolves around the question of
whether implicit or explicit learning is possible through formal L2 instruction.
Doughty (2003: 268) reproduces details from a study of 250 programs by Norris and
Ortega (2000) that shows that 70 percent assume that learning occurs through explicit
instruction and only 30 percent assume that implicit learning is possible. (Remember
that implicit learning is associated with what can be stored as procedural memory while
explicit learning is associated with declarative memory.)
Even though such a study shows that many researchers seem to favor what appears to
be explicit instruction, almost as many shy away from referring to explicit instruction
when they discuss their approaches. For example, Long (1996) developed an
Interaction H ypothesis WKDWHPSKDVL]HVGUDZLQJOHDUQHUVDWWHQWLRQWRWDUJHWVWUXFWXUHV
yes, but doing this specifically in contexts where the attention is meaningful.
2WKHU UHVHDUFKHUV UHIHU WR QRWLFLQJ HJ 6KDUZRRG 6PLWK RU FRQVFLRXVUDLVLQJHJ3LFD2YHUDOOVXFKDWWHQWLRQLVFDOOHGfocus on form, but this is not
WKH VDPH DV IRFXV RQ IRUP-s ZKLFK LV VDLG WR VWUHVV PRUH H[SOLFLW LQVWUXFWLRQ RI
specific forms. R. Ellis, Basturkmen, and Loewen (2001: 407) define focus on form as
WKH LQFLGHQWDO DWWHQWLRQ WKDW WHDFKHUV DQG / OHDUQHUV SD\ WR IRUP LQ WKH FRQWH[W RI
meaning-IRFXVHG LQVWUXFWLRQ This focus can be either reactive (focus on corrective
feedback) or preemptive (when either the instructor or the student makes form a topic
to discuss). No matter what it is called, the assumption seems to be that learning occurs
through some attention to form.
Some approaches pay less direct attention to form, such as meaningbased approaches
and task-based approaches or interactive approaches (cf. Gass, 2003). The assumption
underlying these approaches is that students can learn the structure of the target L2
implicitly through various tasks that focus on meaning2IFRXUVHLWVXQOLNHO\WKDW
any instructional program follows any approach exclusively, but more researchers
currently seem to argue that successful later L2 learning is more likely to be explicit
than implicit learning.
11.9 Summing up
In an overview statement, Nick Ellis (2005) points out three themes that have been
present in research on language acquisition. These themes remind us what we know
about acquisition and what still needs more study.
45
The first theme is the age factor (YHQ WKRXJK DOO UHVHDUFKHUV GRQW DJUHH RQ D
critical age after which native-like acquisition is near impossible, all do agree that
6/$LVOHVVVXFFHVVIXOLQROGHUOHDUQHUV6RWKHTXHVWLRQLVZK\",QDGGLWLRQZH
have to deal with the fact that the success of different L2 learners varies.
A second theme is second language processing. If we accept the notion that SLA
is different from child language acquisition, we need a detailed theory of processing,
that includes both comprehension and production, that takes account of these two
different types of acquisition. We have only some glimmerings of what is involved
in SLA processing. For example, recall the views of Pienemann and others in
section 11.8.4.3; the views of connectionists mentioned in section 11.8.2; and
Ullman on the different roles of procedural memory and declarative memory
discussed in section 11.6.4. Also, other researchers not discussed here argue that not
all elements are processed in the same way and possibly not in the same area of the
brain (e.g. Clahsen, 1999 argues that certain irregular verb forms are processed
differently from regularly inflected ones). But, as Doughty (2003) points out, even
now the proportion of studies that investigate SLA processes is very small.
FLQDOO\(OOLVVWKLUGWKHPHLVYHU\VSHFLILFWR6/$,WLV transfer itself (cf. Odlin,
2003 for an overview). Research seems to show that there is little transfer between
the languages in child bilingualism. This is yet another reason to distinguish child
and late L2 acquisition because there clearly is almost always some transfer in SLA.
Transfer is a hard nut to crack. One intriguing question is this: What type of
transfer is even possible in SLA? Are there innate constraints on which morphemes
or structures are open to transfer?
In addition, a key theme throughout the chapter is the notion that language shows
systematicity. That is, linguistic knowledge is organized in a system of
representations. We see this in child bilingualism where different systems are
acquired and the differentiation of the two systems is accomplished implicitly,
without conscious effort. We assume that this should be possible in second language
learners and we do see at least some systematicity in indications of their underlying
competence and in their production.
T he issue remains of why young children acquire languages so easily and why
adults seem to be unable to reach the same final state as these children do.
46
47