Abstract : Electronic corpora can be used not only for language description but also as a resource for teaching and learning in what has come to be known as "data-driven learning" or DDL (Johns 1990). The arguments for and against have accumulated over two decades and more: in favour are numerous theoretical pedagogical advantages, while the objections tend to revolve around practical and logistical barriers to implementation. Certainly in its more extreme forms, it seems difficult to reconcile DDL with the reality of ordinary language classrooms (e.g. Mukherjee 2006). One of the most apparent obstacles here is the use of the technology itself - the computer with its query software and interfaces for accessing electronic corpora - which has repeatedly been found to pose substantial problems for many learners (e.g. Yoon & Hirvela 2004) as well as teachers (e.g. Farr 2008). Where this is the case, the obvious question is whether the computer can be successfully removed from the equation without losing the benefits of the overall approach (cf. Boulton 2010). It seems obvious that paper-based materials will not afford all the advantages of the fully autonomous, hands-on concordancing DDL is commonly associated with. However, the earliest instantiations of DDL relied almost exclusively on concordance print-outs, and prepared materials can be argued to allow a gentle lead-in, with teacher and learner roles changing gradually as they concentrate first on the new approach (DDL) and the new materials (corpora) without also having to worry about the new technology at the start. Indeed, such has been common practice in many DDL studies (see Boulton forthcoming for a review). Each approach has its adherents, but to date there has been little if any empirical research attempting a direct comparison of the use of paper-based and hands-on corpus consultation. This study draws on past experience with two different groups of students learning English at an architecture college in France. In the first case, the learners were highly receptive to an encounter with printed DDL materials (Boulton 2010); in the second, a hands-on approach to corpus consultation was far less well received (Boulton 2009). However, the varying conditions mean that direct comparisons are difficult to make. In the present study, two groups of lower-intermediate level students at the same college were introduced to English language corpora in the last 15-20 minutes of their regular weekly class. Each group covered the same points using essentially the same processes, but alternated between 'hands-on' consultation (i.e. on computer) and paper-based materials (i.e. 'hands-off'). In this way, after 10 weeks all the students (n=30) should have been at least moderately familiar with both approaches. The learning outcomes of the hands-on and hands-off approaches were compared by testing the final two items covered - language points which had previously been seen to pose problems for many of these learners. The results were compared against feedback obtained at the start and again at the end of the course on the participants' preferences and affective reactions. Previous experience suggests that those who do best are not necessarily most receptive, and vice versa. In an attempt to identify some of the variables that may underlie the individual differences, two instruments were used to relate the outcomes and reactions to motivation and learning styles. The overall conclusion seems to be that there is unlikely to be a single best version of DDL for all learners in all situations at all times, and that variety and sensitivity to learner variation and local conditions is crucial.