This paper reviews existing methods used to address questions about interactivity, cognition and learning in multimedia learning environments. Existing behavioural and self-report methods identified include observations, audit trails, questionnaires, interviews, video-stimulated recall, and think-aloud protocols. The limitations of these methods are examined, highlighting the problems with using behavioural information to speculate about the cognitive activity that is occurring; the difficulties posed by relying on the correct interpretation of questions or accurate recall of past events; and the potential for the data collection methods to have an impact on cognition as it is occurring. Functional brain imaging techniques offer a complement to the existing methods, but possess a number of inherent methodological constraints. The specific constraints emerging from a recent pilot study carried out by the authors are discussed in detail, including: limitations on learner-computer interaction because of the need to use MRI compatible input devices; activation from motor tasks potentially confounding the results of comparisons between cognition occurring during interactive and non-interactive resources; visual differences between multimedia conditions potentially confounding the results due to activation associated with the visual processing; and problems with the provision of "baseline" or "rest" conditions for comparison purposes during holistic problem-based tasks. Despite these constraints it is argued that the use of functional imaging techniques within this field of research has significant potential. (Contains 4 figures.)