Purpose: Investigative interviews are complex, dyadic, activities typically studied by focusing on developing techniques to improve witness performance. However, in field settings, interviewers are known to vary in their presentation of good interviewing practice. Thus, it is important to conduct research reflective of complex idiosyncrasies in witnesses, interviewers and unique pairings of both. This study explores such sources of variation in an by making use of a 'round-robin' design. Such methodology allows the statistical demonstration of individual difference and unique partner-generated variance in interview performance. Methods: In our study, a total of 45 witnesses were questioned about five real crime videos. After witnessing each event, witnesses were interviewed by a different interviewer (or a computer self-administered interview). In total, nine 'rounds' of interviews occurred, with five new witnesses being interviewed in the same five interview settings (resulting in 225 interviews). After each interview both interviewers and witnesses were asked to complete subjective interview experience ratings. The quality (grain size) and quantity of information in the statements was coded to index witness report accuracy. Results: Principally, the results demonstrate the degree to which witnesses and interviewers affect statement quality, and highlight the influence of unique interviewer-witness pairs. (Participant personality and interviewer behaviour were also analysed, but with limited effects). Conclusion: This study presents the useful round-robin methodology for studying interviewer-witness behaviour. Whilst the current personality and behavioural measures do not account for these individual differences, we advocate this methodology be adopted more widely to better understand interview performance.