Marine environments require monitoring to determine the effects of impacts such as climate change, coastal development and pollution and also to assess the effectiveness of conservation measures. Marine protected areas (MPAs) are being established globally and require periodic monitoring to determine whether their objectives are being met. Baited underwater video systems are becoming a popular method for monitoring change within protected fish populations, because they are less damaging to habitats than bottom trawling and allow for more statistical powerful comparisons to determine spatial and temporal patterns in the relative abundances, lengths and biomass of demersal and pelagic fishes. However, much remains uncertain about how interactions between the fish and bait and between the fish themselves affect the results obtained. Agonistic behaviours are frequently observed around the bait of the camera and potentially bias fish density estimates by altering the number and size classes seen at cameras. Here we counted the number of agonistic behaviours between pink snappers (Pagrus auratus), the size of fish involved and whether the fish left the field of view following such behaviours. The study consisted of 20 baited underwater video deployments inside a New Zealand marine reserve and 20 in adjacent open areas. We observed a significant relationship between the peak number of fish observed at the camera and the total number of agonistic behaviours, as well as the number of both aggressor and subordinate fish leaving the camera field of view following interactions. The slope of the latter relationship and thus the absolute numbers of fish leaving were higher for subordinate fish. As subordinates were significantly smaller than aggressors, the apparent size frequency distribution is likely skewed away from smaller size classes. The staying time of the fish and thus the maximum number of fish present at the camera will be reduced by agonistic behaviours and the absolute magnitude of this effect appears to be greater at high fish densities. Our results suggest that an overall effect of these phenomena is to underestimate the differences in abundance between MPAs and open areas, but also to overestimate differences in average size.