DDL Is in the Details... and in the Big Themes

Much traditional language teaching focuses on the learning and application of overt rules. While rules are intended to simplify learning, it often seems that learners have difficulty in applying them even if they understand and retain them. This is perhaps partly because the very concept of rules is to an extent artificial. However, one thing humans have evolved to be naturally good at is the detection of patterns, and generalising from them to new cases. This capacity is exploited in data-driven learning (DDL), where learners are confronted directly with language data and encouraged to discover patterns of language use for themselves. DDL may therefore provide a useful supplement for some learners and for some language items. One criticism aimed at DDL is that it ignores the “big themes” of language such as tenses, although there seems little reason for this to be the case beyond the difficulty in collecting appropriate data. To test this, we chose a traditionally popular topic, the formally easy but semantically difficult pair of items will and going to—the commonest expressions of future meaning in English. 100 French engineering students of roughly intermediate level were divided into two groups: one was provided with edited concordance print-outs, the other with traditional rules from their usual grammar book. Tests were administered before and after the experimental condition to see if the first group was able to detect appropriate patterns from the concordances and to apply these to new situations, and compare results against the group with the familiar rule-based approach. We show that such empirical research can be conducted quite easily, and argue that it is essential if any form of DDL is to appeal to a wider audience.