Because cognition is inherently inaccessible to observers, researchers interested in cognitive approaches to argumentation frequently rely on self-reports for data. Opponents to self-reports argue that such data have very limited value. They claim that when subjects are put in a position where they have to tell more than they can know, they construct plausible answers that are not based on their specific cognitive experiences. An evaluation of the validity of self-reports does reveal certain conditions under which self-reports about cognition are least suspect. First, people do sometimes judge plausibility even when instructed to remember. A second pertinent chunk of evidence shows that people often behave mindlessly, finding that they can follow the appropriate script without paying conscious attention to what they are doing. Third, the effects of consciousness and memory on retrospective verbal reports are crucially affected by the passage of time. Therefore, questions should be asked as soon as possible. The worth of self-reports, therefore, depends almost entirely upon the procedures used to elicit them. Argumentation scholars should be wary of certain potential problems: letting even a few minutes elapse between a thought process and a report of it, seeking data that were never in consciousness, allowing a plausibility search to be a tempting alternative to retrieval efforts, asking for a general description of a process when subjects only experience a specific episode, and allowing evaluation apprehension or demand characteristics to influence responses. (HOD)