In a project-based approach to teaching a foreign language at the university level, students are often required to participate in several task-based writing activities. In doing so, language learners not only write incorrect forms, but also correct forms of the same structures, both of which provide useful information on their strengths and weaknesses. There is variability in language use and such variance may be due to a large number of factors such as the task type or the conditions under which the task is carried out. However, information drawn from direct observations of learners' texts only enables inferences to be made about learners' performance. This paper argues that making a distinction between errors and mistakes, where errors represent gaps in the knowledge and mistakes occasional lapses in performance, provides a better insight into learner variability in terms of the learners' ability to use their knowledge. Following an overview of the instruments used to discriminate between errors and mistakes, this paper investigates, through a case study approach conducted over a 3-month period of time, whether errors and/or mistakes are produced in one written task but not in another. A preliminary focus on a variety of task-based writing activities produced by one learner of French language is analyzed and presented. (Contains 5 figures and 2 tables.)