While a study of high frequency verbs makes it possible to access a wide range of phraseological units, it is not the ideal way of accessing the more routine aspects of phraseology. These are best accessed via automatic extraction and analysis of recurrent combinations. i.e. continuous strings of words occurring more than once in identical form (Altenberg 1998).

We will use this method to extract continuous recurrent n-grams (i.e. sequences of n orthographic words) from native and learner corpora of English essays. This method has very strong heuristic power as it is not based on a pre-established list of phraseological units. The downside is that some of the units brought up by the method are not interesting to analyse from a phraseological point of view. We will disregard fragments, such as because of the, one of the, there is a, which are highly frequent recurrent combinations simply by virtue of the fact that the words that compose them are highly frequent, and focus on formulae, or formulaic expressions, i.e. «multi-word units that perform pragmatic and/or discourse structuring functions» (De Cock 1998).

Using this approach, we will be able to establish whether, as claimed by Kjellmer (1991: 124) learners’ «building material is individualized bricks rather than prefabricated sections». A number of studies that have used this approach prove that this statement needs to be qualified. Milton (1998) shows that learners seem to rely on a smaller repertoire of prefabs (prefab types), some of which are shared with native speakers, others learner idiosyncratic, which are used with very high frequencies of occurrence (prefab tokens). Among the prefabs which are overused by Chinese-speaking learners of English, he notes a high frequency of connectors (first of all, on the other hand), especially metaphorical ones such as in a nutshell, a phenomenon which he interprets as being teaching-induced (‘NNSs are using those expressions which they have been instructed to use’) rather than transfer-related. De Cock’s (2000) results, on the other hand, show that French-speaking learners overuse the least salient (i.e. most transparent) word combinations, such as we can say that, I think that, a phenomenon which is clearly not teaching-induced. Therefore results here are also highly inconclusive.

[back]