The use of metalingual tasks in L2 acquisition research has recently been challenged by a number of critics on the grounds that they may constitute only an indirect and unreliable reflection of learner competence. Recent research involving metalingual judgements tests has revealed that learners' reactions to well-formed and ill-formed strings emerge largely as a result of the cognitive processes brought about by the particular characteristics of the linguistic stimulus. Moreover, the assumption that learner judgements tap into a unitary or stable knowledge source has been called into question, since a number of knowledge systems and levels of awareness have been found to contribute to learner performance in L2. Because disagreement continues about the usefulness of metalingual judgements and the information they can provide, this article aims to clarify currently established perspectives on this issue, particularly with respect to the question of how learners' tolerance of ill-formedness compares to their acceptance of well-formedness. It will be argued that specific aspects of the form and content of grammatical and ungrammatical test strings may strongly influence the saliency of certain malformations, and impact the decision-making process in systematic ways.Over the past two decades, metalingual tasks have been used frequently by numerous L2 researchers, many of whom have viewed the results of such tasks, at least to some extent, as a 'window into competence' (e.g.