This study has aimed to describe the extent to which non-native writers make use of word combinations, and particularly strong collocations, in comparison to native speaker norms, by using methodologies which take advantage of frequency
information, and which take account of individual variability between texts. Three main findings have emerged. Firstly, native writers use more low frequency
combinations than non-natives. This trend appears to be fairly consistent across texts, even though it was statistically significant only in the comparison of longer texts. Secondly, non-native writers make at least as much use of collocations with very high t-scores as do natives. Since non-natives also tend to repeat certain favoured
collocations, if we consider collocation tokens, rather than types, they show a
significant overuse of these strong collocations in comparison to native norms. Thirdly, non-native writers significantly underuse collocations with high mutual information scores in comparison with native norms. Again, the repetition of favoured items bolsters the non-native count somewhat, so the difference is more marked on an
analysis of collocation types. All of these regularities were less marked in shorter texts, but even here we found sufficient consistency of usage for the same tendencies to emerge, if not always with statistical significance.
I argued above that, if Wray’s model were correct, we would expect non-native writers to make much less use of high frequency collocations than natives. I also argued that, if, in contrast to Wray’s thesis, second language learners retain and use those sequences of language which are frequent in their input, we would expect non- native writers to rely especially on the most high frequency collocations. The pattern of results presented here supports the latter position, suggesting that the retention of
the normal, long-term second language acquisition process. Taken together, then, this and the previous study suggest that Wray’s thesis that adult second language learners fail to acquire appropriate collocations because they take a more word-oriented approach to learning than child L1 learners is not correct.
At the same time, these results enable us to account for Kjellmer’s sense that there is something inauthentic about the phraseology of second language learners. The problem is not that they fail to use formulaic language altogether, but rather that they avoid those items which are of relatively low frequency, but which are strongly
associated. Since these items are probably highly salient for natives, their absence may give a strong impression in unnaturalness. However, it is not necessary to posit any radically different L2 learning mechanism to explain this absence; their
characteristically low frequency of occurrence simply means that such collocations are likely to be acquired later than other parts of nativelike phraseology.
5.5 Summary and conclusions: collocation learning from
input
This chapter set out to examine Wray’s (2002) claim that adult language learners do not acquire the collocations to which they are exposed because their mature, literate cognitive systems, together with various situational pressures, push them to focus their attention on individual words, rather than meaningful chunks. I argued that this claim cannot be properly evaluated by simply focusing on the end results of adult learning, since it is not clear whether any shortcomings in collocational knowledge are a product of an alternative learning approach or of insufficient input. The studies in this chapter have therefore attempted to link adults’ knowledge of collocation with the input they have received.
Linking input with knowledge presents the researcher with an observer’s paradox: the more tightly we control the input learners receive, the more likely we are to distort the natural learning situation. With this in mind, I approached the question using two complementary experimental paradigms. In the first, lab-based, approach, it was possible to exercise very tight control over input, but the learning process may not have been entirely natural, and only short-term learning could be traced. In the second,
corpus-based, approach, a more natural and longer-term learning situation was studied, but it was only possible to make a rough estimate of learners’ likely input.
The results of both studies suggest that adult learners do acquire at least some of the collocations to which they are exposed. The lab-based study (Section 5.3) showed that learners who were asked to perform an entirely formal task (reading a sentence aloud) with no knowledge that they were expected to learn anything from this exposure, retained information about which words appeared together. Limited repetition of the task increased this retention dramatically. This suggests that collocation learning may be an automatic process, which will continue regardless of any strategies adopted by the learner, but that techniques such as fluency-oriented re-reading may hasten this learning. The corpus-based study (Section 5.4) found that advanced non-native speakers of English make at least as much use of high frequency collocations as natives. However, their failure to use lower-frequency, strongly associated word pairs may create a superficial impression that they are avoiding formulaic language. This pattern of results is, I argued, compatible with a model whereby learners extract the most frequent collocations from the input they meet. It does not seem to be consistent with the idea that learners fail to remember the collocations they encounter.
Taken together, these results suggest that adult second language learners are capable of learning collocations implicitly from input. However, this does not mean that they typically do so. The distinctive pattern of collocation use found in Section 5.4
suggests that these learners had some way to go in their collocation learning. The most likely reason for the problem seems likely to be a lack of sufficient input. Adult
second language learners typically have far less exposure to the target language than native speakers. This relative sparsity of input may mean that for lower frequency collocations the gap between repeated exposures is too great for the necessary representations to become entrenched. If this is the case, then learners will require special instruction in collocation, either through explicit teaching or artificially enriched input. If such input is to be provided, of course, teachers will need to know what collocations their learners need to learn. The next chapter will explore the possibility of identifying key collocations for learning in one particular area of language – English for academic purposes.