Phraseological errors constitute another interesting window onto the learners’ phrasicon. There is unfortunately no reliable way of accessing these errors automatically. Fortunately, we will be able to make use of two corpora of learner writing, a corpus of French as a Foreign Language (300,000 words), the other of English as a Foreign Language (150,000 words), in which all errors have been tagged with a standardized annotation system (see Dagneaux et al 1998 & Granger et al 2001). The corpora contain a total of 1,870 authentic phraseological errors (670 for English and 1200 for French) which will be extracted from the annotated corpora with standard retrieval programs and subjected to detailed quantitative and qualitative analysis. As all texts have well-identified mother tongue backgrounds (English and Dutch for the French learner corpus; French for the English learner corpus), it will be possible to assess the influence played by transfer in the production of these errors.

[back]