The Role of the First Language in Second Language Acquisition By Stephan Krashen
The topic of "first language interference" has had an unusual history in second language acquisition research and practice. For many years, it had been presumed that the only major source of syntactic errors in adult second language performance was the performer's first language (Lado, 1957), and a great deal of materials preparation was done with this assumption in mind (Banathy, Trager, and Waddle, 1966). Subsequent empirical studies of errors made by second language students led to the discovery, however, that many errors are not traceable to the structure of the first language, but are common to second language performers of different linguistic backgrounds (e.g. Richards, 1971; Buteau, 1970). These findings have led several scholars to question the value of contrastive analysis and to argue instead for error analysis. The first language, it is maintained, is but one of several sources of error, and other sources need to be considered.
The issue now, as I see it, is not whether first-language-influenced errors exist in second language performance (they clearly do), or even what percentage of errors can be traced to the first language in the adult, but, rather, where first language influence fits in the theoretical model for second language performance.
In this chapter, I will attempt to show that findings on first language influence on second language performance are quite consistent with findings and hypotheses from other apparently nonrelated areas, and that they contribute to a clear theoretical picture of second language acquisition and performance. First, the relevant findings on first language influence are summarized. Following this, the role of the first language as a substitute utterance initiator is discussed.
1. First language influence appears to be strongest in complex word order and in word-for-word translations of phrases.
Evidence for this generalization comes from several sources. Duskova (1969), for example, studied written errors in the compositions of Czech "postgraduate students" and concluded that "interference from the mother tongue... was plainly obvious in errors of word order and sentence construction" (p.18), a common example being the placement of the direct object after an adverbial, as in
I met there some Germans.
Also present in the compositions were many word-for-word translations of Czech expressions into English, such as "another my friend" instead of "another friend of mine".
LoCoco (1975), in a study of American college students learning Spanish and German in the US, a foreign language situation, reported that the "high incidence of interlingual (L1 interference) errors in German was due to word order errors..." (p. 101). Typical examples include
Hoffentlich du bist gesund
Hopefully you are healthy
correct: Hoffentlich bist du gesund
Ich bin glücklich sein hier
I ** happy to be here
correct: Ich bin glücklich hier zu sein.
First language-based errors in Spanish were less numerous and "pertained primarily to adjective position". The greater word order differences between English and German as compared to English and Spanish accounts for the differences in frequencies in interference word order errors. Spanish students were more often correct in using English surface structures in utterance initiation due to the greater surface similarity between English and Spanish. This also accounts for Chan's (1975) finding that English to Spanish interference errors occurred mainly "on grammatical categories absent in either the NL or TL" and not in word order.
LoCoco also found that second level Spanish students showed an increase in interference type errors that LoCoco calls "whole expression terms", or word-for-word translations of an L1 expression, which is similar to what Duskova reported.
2. First language influence is weaker in bound morphology.
Duskova (1969) notes that errors in bound morphology (e.g. omission of plurals on nouns, lack of subject-verb agreement, adjective-noun agreement) are not due to first language influence in her Czech students of EFL: Czech nouns do not distinguish singular and plural and in Czech "the finite verb agrees with its subject in person and number". These errors are, rather, "interference between the other terms of the English subsystem in question" (p. 21). Moreover, these errors "occur even in cases where the English form is quite analogous to the corresponding Czech form" (p. 21). Of 166 morphological errors, only nineteen were judged as due to Czech interference. (interestingly, of these nineteen, several were free morphemes; see discussion in Chapter 4.)
Also consistent is Kellerman's (forthcoming) suggestion that inflectional morphology ("except in very closely related languages") belongs to the category of structure that performers generally do not transfer in second language performance.
3. First language influence seems to be strongest in "acquisition poor" environments.
Dulay and Burt (1974b) and Gillis and Weber (1976) have demonstrated that first language influence is rare in child second language acquisition (but see below). On the other hand, studies that report a high amount of first language influence, such as those cited above, are mostly foreign and not second language studies, situations in which natural appropriate intake is scarce and where translation exercises are frequent. In this regard, it is interesting to note that we can find signs of first language influence in immersion bilingual programs where input is often primarily from the teacher and not from peers.
First-language-influenced errors here are also in the domain of word order (Selinker, Swain, and Dumas, 1975; Plann and Ramirez, 1976).
This suggests that it is not simply the case that adults show first language influence while children do not. We would expect to see first language influence in situations where child second language acquirers obtain less intake or where affective conditions prevent or inhibit acquisition (where the affective filter "filters" the input; see Chapter 2).
We now attempt to integrate these findings and fit them into the Monitor Model for performance. First, let us reconsider Newmark's (1966) proposal for a mechanism for first language influence. According to Newmark, first language influence is not proactive inhibition, but is simply the result of the performer being "called on to perform before he has leaned the new behaviour". The result is "padding", using old knowledge, supplying what is known to make up for what is not known. Newmark suggests that the "cure for interference is simply the cure for ignorance: learning" (in terms of Monitor Theory, this would read "acquisition").
What can be concluded from the above is that the L1 may "substitute" for the acquired L2 as an utterance initiator when the performer has to produce in the target language but has not acquired enough of the L2 to do this.
First language influence may therefore be an indication of low acquisition. If so, it can be eliminated or at least reduced by natural intake and language use. This is what apparently occurred in Taylor's ESL subjects, who showed less first language influence with more proficiency (Taylor, 1975).
Perhaps the "silent period" observed in natural child second language acquisition (Hakuta, 1974; Huang and Hatch, 1978) corresponds to the period in which the first language is heavily used in "unnatural" adult second language performance. The children may be building up acquired competence via input, and several recent studies (Gary, 1974; Postovsky, 1977) imply that less insistence on early oral performance may be profitable for children and adults studying second languages in formal settings.
The L1 plus Monitor Mode
First language influence can thus be considered as unnatural. One could theoretically produce sentences in a second language without any acquisition: the first language surface structure can be used with second language yasaklı kelime lexicon inserted. The Monitor may then be used to add some morphology and do its best to repair word order where it differs form the L2. One can only go so far with this mode, as one is limited by the competence of the conscious grammar and one must appeal to it with every utterance. The adult can, however, produce sentences right away in the target language using this mode, and this may help to account for reports of more rapid progress in early stages for adults than for children in second language performance (Snow and Hoefnagel-Hohle, 1978). It is a temporary advantage, however. Acquisition may be slow, but it is, in the long run, much more useful when language is used for the purpose of communication.1
1Several recent papers point out that the hypothesis presented here, that we "fall back" on the first language when we have not acquired aspects of the second language, as stated here, is inadequate to account for all of the data. As Wode (1978) has pointed out, first-language-influenced errors may only occur at certain stages in development. Wode's example is quite clear, and is reviewed here.
In English, the negative particle appears after the auxiliary, as in
(1) I can not go,
but before main verbs, with do-support, as in
(2) I don't know.
In German, however, the negative particle appears after both auxiliaries and main verbs, as in
(3) Ich kann nicht gehen,
I can not go,
and (4) Ich weiss nicht,
I know not,
Wode's children, German-speakers acquiring English as a second language in the United States, produced some sentences showing apparent first language influence, such as
(5) John go not to the school.
What is interesting, Wode points out, is that they did not produce such sentences early on. Their first attempts at negation were similar to what one sees in first language acquisition, such as
(6) no, you
(7) no play baseball.
They only produced sentences such as (5) when they had begun to acquire the aux. + neg. rule, i.e. when they had begun to produce sentences such as
(8) lunch is no ready.
Only then did they "fall back" on the more general German rule. Wode (1978, 1979) suggests that there is, therefore, a structural prerequisite for first language influence: the performer's interlinguistic structural description, his idea of the target language rule, must be similar to the structural description of the rule in his first language. Wode's children's English negation rule was not at all similar to the German rule in early stages, but it became similar when they progressed to the aux.-neg. stage. Hence, first language influence appeared later, but not earlier.
Also, consistent with his argument is the fact that sentences such as (5) are only found in child second language acquisition of English when the first language has post-verbal negation, as in Ravem (1968), in which the first language was Norwegian. Such sentences were not observed in other ESL studies utilizing different first languages (Milon, 1974, L1 = Japanese) and in studies of negation in English as a first language.
Kellerman (1978) suggests another condition for "transfer" to occur. The acquirer must perceive a similarity between items in the first and second language. Items that appear to be language specific (e.g. idioms) will be less prone to transfer.
These conditions are not contrary to the generalization presented here. It can still be maintained that we "fall back" on the first language when we have not acquired aspects of the second language. They show, however, that ignorance is not a sufficient condition for the occurrence of first language influence.