• No results found

The contribution of input to early L2 language acquisition

Taken together, the papers in this thesis shed some light on exactly what we can expect the role of input in SLA to be, especially at the very early stages. The debate on the role of input versus explicit instruction was outlined in section 3.1.2 and in section 3.1.2.2 in particular, but

47

this was in general terms, and not really addressing the question psycholinguistically in terms of processing (cf. VanPatten, 2012). The question thus remains of exactly how input contributes to acquisition. Any current theory of language acquisition will agree that acquisition happens through the interaction of input with specific or general mechanisms. However, what those exact mechanisms are, and how input works on different aspects of language competence, is debated.

For example, it is common to assume a distinction between input and intake, where the latter depends on attention or “noticingˮ (e.g., Corder, 1967; Long, 1996; Robinson, 1995; Schmidt, 2001). However, studies have found that acquisition of various aspects of language can take place in the absence of attention to the input in both children and adults (Saffran, Newport, Aslin, Tunick, & Barrueco, 1997; Schachter, 1998; Truscott, 1998). S. E. Carroll (1999) points out that language learning is not the same as speech processing, and the term input is thus ambiguous in that input to language learning mechanisms is not the same as input to speech processing mechanisms. She views speech processing as a chain of representations beginning with the auditory signal and ending with a conceptual representation, where the output of each stage of processing serves as the input to the next stage. Within levels of processing, she argues, the conversion of input to intake may be a process to which it is not possible to pay attention. In her model, the role of learning mechanisms is specifically to resolve processing problems;

(attempted) learning is automatically triggered when processing fails10. Input to the learning mechanisms thus consists of four elements: the partially analyzed parse representation, the unanalyzable item, the current set of parsing procedures, and whatever information external to the relevant processing level can be utilized to resolve the problem. What specific information this is depends on our theory of language; it may, for example, come from UG, from analogy or generalization mechanisms, or from inferences drawn from negative evidence or other mechanisms such as “negotiation of meaningˮ (S. E. Carroll, 1999, p. 365). Thus, in Carroll’s view, attention cannot simply be seen as a function which selects information from the speech signal to feed to the learning mechanisms.

The children both in the bilingually-based group discussed in papers 1-3 and in the immersion group discussed in papers 2 and 4 had all acquired both receptive vocabulary and sentence processing and comprehension skills from naturalistic input, and without any effort on

10 The qualification that such learning is "attempted" means that there is no guarantee that the restructuring resulting from the application of these learning mechanisms, allowing a successful parse, is "target-like" from the perspective of the L2.

48

the part of the teachers to draw their attention to form in the input. Input may have been adapted to the learners in being simplified, with emphasized intonation, and speech being particularly slow. Yet, there is nothing unnaturalistic about such adapted input; it resembles child-directed speech in L1 acquisition. Just like it is possible that such speech modification may be beneficial to young children acquiring their native language (Yurovsky, Yu, & Smith, 2012), it may facilitate SLA.

There is little evidence in the present studies to answer the question of whether also more detailed grammar competence can be achieved based on such naturalistic input, and without focus on form. However, the results of the immersion group in the repetition test reported in Paper 4 may be an indication. Four out of seven children score within the range of the native speaker control group on the repetition test, an indication that the necessary skills can indeed be acquired from input. The study reported in this paper found intra-subject differences between the three measures of English (vocabulary, sentence comprehension, and sentence repetition) after eight months, and there is thus no straightforward relationship between e.g., vocabulary size and accuracy in the repetition test. However, it must be assumed that the high score of most

participants in this group has come about based on their substantial exposure to the target language. They were “pushedˮ to produce language in the sense that the language of communication in their classroom was English and they could not expect all classmates to understand any of their other languages, but there was no focus on grammatical accuracy in these classrooms, and naturally no relevant explicit instruction, given that most of the students in each class were already fluent English speakers. The exception is one participant who attended EAL classes for a time. However, the fact that also three other children scored within the native speaker range on this test may indicate that also grammatical accuracy can in principle be acquired from naturalistic input in young L2 learners.

The role of input in SLA is likely to be connected to frequency. Frequency effects have been found for the acquisition of all areas of language, including phonology, vocabulary, morphology and syntax in children and adults. We know that the frequency with which items are encountered influences their recognition and recall. With increased input, both type and token frequency increases, which may facilitate statistical learning. While there is disagreement about the degree to which attention to the input is necessary for such learning to take place, it is clear that the mechanisms operate on mere exposure, and that attention to the input in this context does

49

not equal attention to specific forms in the input (N. C. Ellis, 2002, 2006a; Lieven, 2010;

Mirman, Magnuson, Estes, & Dixon, 2008; Pelucchi, Hay, & Saffran, 2009; Saffran, 2001;

Saffran, Newport, & Aslin, 1996; Thompson & Newport, 2007; Toro, Sinnett, & Soto-Faraco, 2005; Vouloumanos, 2008; Yu & Ballard, 2007; Yurovsky et al., 2012).