Tracking learners' actual uses of corpora: guided vs non-guided corpus consultation

Much of the research into language learners' use of corpus resources has been conducted by means of indirect observation methodologies, like questionnaires or self-reports. While this type of study provides an excellent opportunity to reflect on the benefits and limitations of using corpora to teach and learn language, the use of indirect observation methodologies may confine the scope of research to learners' opinions about the benefits of using corpora for language learning and their self-perceived difficulties in consulting them. This article proposes and discusses the use of logs to research learners' actual use of corpus-based resources, analyzing the number of events or actions performed by each individual, the total number of different web services used, the number of activities completed, the number of searches performed on the British National Corpus (BNC) and, last, the number of words or wildcards per BNC search. Our research used these parameters to investigate whether learner interaction with corpus-based resources differed under different corpus consultation conditions: guided versus non-guided consultation. Our findings show that the individuals in the two research conditions behaved differently in two of the parameters analyzed: the number of different web services used during the completion of the tasks and the number of BNC searches. Our results corroborate empirically the suggestions found in the literature that skills and guidance are necessary when teachers take a corpus to the classroom. Similarly, we offer evidence that user tracking is essential to claim research and results validity.