Understanding Second Language Acquisition (2nd ed)

by Rod Ellis

  Linguistic difference and learning difficulty

  The key assumption of the Contrastive Analysis Hypothesis is that linguistic differences between the learner’s native language and the target language can cause learning difficulty and thus will result in errors. However, contrastive analysts acknowledged that linguistic differences are of different kinds and that some differences were likely to cause greater difficulty than others. Based on a contrastive analysis of Spanish and English, Stockwell, Bowen, and Martin (1965) proposed the hierarchy of difficulty shown in Table 6.1NOTE 2.

  Type of difficulty L1: English L2: Spanish Example

  1 Split

  x xy ‘for’ is either ‘por’ or ‘para’

  2 New

  o x grammatical gender

  3 Absent

  x o ‘do’ as a tense carrier

  4 Coalesced

  x x ‘his/her’ is realized as a single form

  y ‘su’

  5 Correspondence

  x x -ing, -ndo as the complement of verbs of perception, e.g. ‘I saw the men running’; ‘vi a los hombres corriendo’.

  Key: x = feature present; o = feature not present; x/y = two equivalent features

  Table 6.1 Simplified version of the hierarchy of difficulty (based on information given in Stockwell, Bowen, and Martin 1965)

  Table 6.1 suggests that the influence of the L1 is strongest when the difference between the native and target language is greatest and weakest when the difference is least. However, this need not be the case. Learners have been shown to find it more difficult to acquire sounds that are similar (but not exactly the same) to sounds in their native language than sounds that are dissimilar. In other words, ‘correspondence’ between the native and target languages does not guarantee ease of acquisition while entirely ‘new’ features do not always pose learning difficulty.

  Major and Kim (1996) proposed the Similarity Differential Rate Hypothesis, according to which L2 sounds dissimilar to the L1 may be difficult to acquire initially, but—over time—improvement takes place more rapidly than with similar sounds. They commented ‘gross differences are more often noticed, due to their perceptual saliency, whereas minimal differences are more likely to be overlooked and to result in confusion or nonlearning’ (p. 367). Major and Kim’s study of the rate of acquisition of /j/ and /z/ by ten adult Korean learners of English supported this hypothesis. They noted that—although all the evidence they cited concerned pronunciation—there was every reason to believe that the hypothesis would also apply to other aspects of language.

  However, where grammar is concerned, there is clear evidence that linguistic difference affects learning difficulty in both the short term and in the later stages of L2 learning. The Morphological Congruency Hypothesis (Jiang, Novokshanova, Masuda, and Wang 2011)—as its name suggests—claims that, when a grammatical morpheme with the same function is present in both the native and target language, learning will be easier than when it is present in the target language, but not in the native language. Jiang et al. investigated this claim by comparing Russian and Japanese ESL learners’ ultimate attainment of English plural -s. Whereas Russian marks nouns for plurality, Japanese typically does not. A feature of this study is that it sought to examine the effects of L1 transfer on implicit L2 knowledge by measuring learning in terms of the extent to which the two groups of learners noticed plural -s errors in a self-paced reading taskNOTE 3. The results showed that—whereas the Russian learners were generally sensitive to the plural errors—the Japanese learners showed very little sensitivity. As the learners in both groups were near-native, Jiang et al. concluded that the effects of L1 transfer were long-lasting and possibly non-remediable without formal instruction.

  Nevertheless, it would be wrong to conclude that all grammatical differences between the native and target languages result in learning difficulty. There is, in fact, no simple correlation between linguistic difference and learning difficulty as claimed by the Contrastive Analysis Hypothesis. For example, differences in word order between the native and target language do not typically result in learning difficulty. Japanese is a verb final language whereas in English the verb typically follows the subject; however Japanese learners of English do not appear to transfer their L1 word order and, if they do so, only in the very initial stages of L2 acquisition. Odlin (1989) in his extensive survey of transfer studies admitted that—although there were cases of basic word order transfer in the literature—such cases were rare. The obvious explanation for this is the saliency of word order in the L2 input: whereas a morpheme, such as plural -s, is not easily noticed in the input learners are exposed to, the basic order in which words appear in the target language is.

  Language distance

  Language distance also has more complex effects on language transfer than those predicted by the Contrastive Analysis Hypothesis. A common assumption is that if the native and target languages are proximate (for example, Dutch and English) learning will be easier and more rapid than if they are distant (for example, Chinese and English). However, the similarity of two languages can result in negative as well as positive transfer. Sjöholm (1976) compared the acquisition of English by Swedish-speaking and Finnish-speaking Finns. Whereas 93 per cent of the population speak Finnish as their mother tongue, a language distant from English, six to seven per cent speak Swedish, a language much closer to English. Sjöholm found that Swedish-speaking Finns enjoyed a substantial learning advantage over Finnish-speaking Finns—i.e. positive transfer occurred. However, the Swedish-speaking Finns also made more errors, i.e. they manifested more negative transfer. Ringbom (1978) reported that both the Swedish- and Finnish-speaking groups (both of whom were bilingual in the two languages) were much more likely to transfer word morphology from Swedish into English than from Finnish. Whereas the Finnish speakers avoided transferring elements from their L1, preferring instead to fall back on their first L2 (Swedish), the Swedish speakers readily transferred elements from their L1 but avoided doing so from their first L2 (Finnish). In other words, when the distance between source and target language is small, both positive and negative transfer are likely to occur. In general, however, the greater the language distance, the more difficult a language is to acquire.

  Markedness

  Marked linguistic features are those that are ‘special’ in relation to others; unmarked features are those that are, in some sense, ‘basic’. Markedness has been defined differently in different traditions in linguistics. Here, I will focus on how markedness has been handled in a functional-typological approach. This involves comparing different groups of languages in order to distinguish those properties that are universal or specific to particular languages. The general claim is that those features that are common in the world’s languages are less marked than those features that are only found rarely in the world’s language. For example, pied piping—i.e. the process by which one element in a clause drags other words along with it—as in ‘To whom did you speak?’—is much more common that preposition stranding, where the wh- element and the preposition are not attached—as in ‘Who did you speak to?’ The former is considered unmarked in relation to the latter.

  Two general hypotheses have been investigated: (1) learners will transfer unmarked forms when the corresponding target form is marked, and (2) learners will resist transferring marked forms, especially when the corresponding target language form is unmarked. As Hyltenstam (1984) put it:

  Unmarked categories from the native language are substituted for corresponding marked categories in the target language … Marked structures are seldom transferred, and if they are transferred, they are much more easily eradicated from the target language. (p. 43)

  A general assumption is ‘marked structures tend to require more attention, more mental effort, and time to be processed’ (Callies 2013: 406) and thus learners resist transferring L1 marked structures.

  Perhaps the most convincing evidence of markedness effects on transfer can be found in studies th
at have examined asymmetrical patterns (i.e. involving a kind of type 2 comparison). A straight contrastive analysis is unable to cope with evidence that shows that a given feature (Z) is transferred in one direction—i.e. transfer of Z occurs from language X to language Y—but not in the other—i.e. transfer of Z does not occur from language Y to language X. A theory of transfer that incorporates markedness, however, can provide an explanation for such phenomena.

  The study that is most commonly cited to illustrate asymmetrical patterns is Eckman (1977). This provides support for both of Hyltenstam’s hypotheses. Eckman investigated transfer in English learners of L2 German and German learners of L2 English, focusing on voice contrast in pairs of phonemes such as /t/ and /d/. In British English, this contrast exists word initially (for example, ‘tin’ versus ‘din’), word medially (for example, ‘betting’ versus ‘bedding’), and word finally (for example, ‘wed’ versus ‘wet’). In German, however, the distinction is only found word initially and word medially; in word-final position, only voiceless stops occur. Both the German and the English L2 learners, therefore, are faced with learning to make a known distinction (i.e. voiced/voiceless stops) in a new position. Eckman argued that typologically, voice contrast in word-final position is more marked than in the other two positions. That is, the markedness relationship was established on the basis of a cross-linguistic implicational generalization. He provided evidence to show that English learners have no difficulty in learning that German has no voicing in word-final stops, but that German learners experience considerable problems in learning that English does. In other words, no transfer effects are evident when the L1 position is marked and the L2 position unmarked, but they appear when the L1 position is unmarked and the L2 marked.

  In order to explain how markedness affects transfer, Eckman advanced the Markedness Differential Hypothesis (MDH):

  The areas of difficulty that a second language learner will have can be predicted on the basis of a systematic comparison of the grammars of the native language (NL), the target language (TL) and the markedness relations stated in universal grammar, such that:

  (a) Those areas of the TL which differ from the NL and are more marked than in the NL will be difficult;

  (b) The relative degree of difficulty of the areas of the TL which are more marked than in the NL will correspond to the relative degree of markedness;

  (c) Those areas of the TL which are different from the NL, but are not more marked than the NL, will not be difficult.

  (Eckman 1977: 321)

  The Markedness Differential Hypothesis constituted an attempt to reformulate the Contrastive Analysis Hypothesis to take account of markedness factors by explaining when differences between the native and target languages do or do not result in difficulty.

  However, learners have also been shown to transfer marked features from their first language even though the corresponding feature in the target language is unmarked. For example, English permits pied piping (which is typologically unmarked) and preposition stranding (which is typologically marked):

  To whom did you speak? (pied piping)

  The man to whom I spoke is now in hospital. (pied piping)

  Who(m) did you speak to? (preposition stranding)

  The man who(m) I spoke to is now in hospital. (preposition stranding).

  Other languages, however—for example, Spanish and French—do not permit preposition stranding (i.e. pied piping is obligatory). In accordance with the markedness hypothesis, one might expect English learners of L2 French or Spanish to opt for unmarked pied piping (i.e. resist transferring the marked preposition stranding). However, Liceras (1985) reported that 43 per cent of the English-speaking beginner learners she investigated accepted stranding in Spanish sentences such as *Es la mujer que Pedro vive con (= ‘This is the woman that Pedro lives with’) providing evidence of the transfer of the marked variant.

  As Jarvis and Pavlenko (2008) concluded, ‘there is ample evidence to show that transfer of both marked and unmarked syntactic structures is quite common’ (p. 187). In general, though, L2 learners do seem more ready to transfer unmarked L1 features irrespective of whether the corresponding L2 feature is marked or unmarked. In other words, they act in accordance with the general principle ‘If a feature is unmarked in the L1, transfer it.’ Overall, then, although linguistic markedness has been shown to account for some instances of transfer, it has failed to provide a convincing explanation for many instances of L1 transfer.

  Problems with linguistic accounts of L1 transfer

  The main problem with the difference = difficulty approach to investigating language transfer is that whereas ‘difference’ is a linguistic concept, ‘difficulty’ is a psychological concept. It cannot be assumed that what constitutes ‘difficulty’ according to a linguistic theory (and, in fact, this itself is problematic as there is no agreement among linguists about which theory to apply) will have any psychological significance for the learner. A similar problem exists with markedness accounts of transfer. Markedness is a purely linguistic concept and although—as we have seen—it does seem to have some kind of psychological reality for learners, this cannot explain what goes on in the learner’s mind. Also—as Callies (2013) noted—there is no agreed way of determining which features are marked in relation to others. Eckman (1985) also pointed out that ‘more research is needed in defining markedness relations’ (p. 306).

  One way of addressing both these problems is by examining ‘native speakers’ own perceptions of the structure of their language—for example, by asking native speakers whether they perceive specific features as ‘infrequent, irregular, semantically or structurally opaque, or in any other way exceptional’ (Kellerman 1983: 117). In the next section I turn to research that has adopted a psycholinguistic perspective on first language transfer.

  Psycholinguistic factors

  Prototypicality

  Kellerman (1978) hypothesized that learners have intuitions about which features in their first lanaguage are potentially transferable or non-transferable. In his ‘breken’ study he first investigated Dutch L1 speakers’ perceptions of the ‘coreness’ of seventeen sentences containing the verb ‘breken’. He then asked 81 Dutch students of English in their first and third years at university to say which of the sentences containing ‘breken’ they would translate using the English verb ‘break’. He reported clear differences in the percentage of students prepared to translate each sentence. For example, whereas 81 per cent considered ‘hij brak zijn been’ (= he broke his leg) translatable only 9 per cent identified ‘sommige arbeiders hebben de staking gebroken’ (= some workers have broken the strike) as translatable. Kellerman found that the students were much more likely to consider the core meanings of ‘breken’ translatable than the non-core meanings.

  In another study, Kellerman (1979) examined whether learners’ perceptions of transferability differed according to their L2 proficiency. He asked 291 learners of English, who ranged from 12-year-olds in their second year of English to third-year university students, to assess the translatability of nine of the 17 ‘breken’ sentences. He found a high level of similarity in their responses and concluded that ‘the effects of teaching, learning, and growing older do not significantly alter learners’ beliefs about the relative transferability of the “brekens”‘ (1979: 52).

  These studies suggest that L2 learners have an intuitive understanding of the prototypical meaning of a lexical item such as ‘breken’ and that this matched the primary meanings of the word found in a dictionary. In a later study, Kellerman (1989) extended the notion of prototypicality to grammar. He provided evidence to show that advanced Dutch learners of L2 English are likely to produce errors of this kind:

  *If it would rain, they would cancel the concert in Damrosch Park.

  (= If it rained they would cancel the concert in Damrosch Park.)

  despite the fact that Dutch makes use of equivalent verb forms to English in both main and subordinate clauses. He suggested that
positive transfer does not occur because learners perceived it as more natural to say ‘would rain’ than ‘rained’ because the verb is explicitly marked for future time.

  The key point emerging from these studies is that learners—irrespective of their L2 proficiency—have perceptions about what is transferable from their first language and act in accordance with these perceptions. They prize ‘reasonableness in language’ and ‘attempt to keep their L2s transparent’ (Kellerman 1983: 129). Thus, L1 features that they perceive to be working against this principle—such as idioms that are highly metaphorical or grammatical structures where meanings are not overtly encoded—are not transferred. This claim has been supported in a number of more recent studies (for example, Kato 2006).

  Psychotypology

  We noted above that the distance between the native and target languages can influence the degree and nature of transfer. Distance, however, can be viewed as a psycholinguistic phenomenon as well as a linguistic phenomenon and, arguably, what matters most is not the actual linguistic differences between two languages, but whether learners think they are different. Kellerman (1979) used the term ‘psychotypology’ to refer to learners’ perceptions about language distance. He claimed that the decisions that learners make about whether or not to draw on their first langauge are based on their beliefs as to whether the native and target languages are the ‘same’ or ‘different’.