Despite sparse and sometimes contradictory evidence, eXtreme Programming (XP) is currently growing in popularity as an alternative strategy for developing software. This paper proposes a methodology for observational studies of one of XPs twelve practices, pair programming, including the analysis of verbalisations, interactions and artefacts in order to improve our understanding of its cognitive implications. In addition, a protocol for analysing pair programming behaviours is included as a means of identifying the behavioural differences of those who are more expert or novice at pair programming and understanding the effect of pairing programmers of differing expertise. Initial results from a recent study using this methodology are then presented.