The degree to which researcher generated visual records (for example video texts) may be used to collect valid information about the social world is subject to considerable academic debate (cf. Feld and Williams, 1975; Gottdiener, 1979 and Grimshaw, 1982). On the one hand the method is assumed, by implication, to have limited impact on the data, the taped image being treated as a replica of the unrecorded event (Vihman and Greenlee, 1987; Vuchinich, 1986). On the other, it is suggested that the video camera has a uniquely distorting effect on the researched phenomenon (Gottdiener, 1979: p. 61; Heider, 1976: p. 49). Research participants, it is argued, demonstrate a reactive effect to the video process such that data is meaningful only if special precautions are taken to validate it. Strategies suggested include a covert approach to the data collection itself (cf. Eibl-Eibesfeldt and Haass, 1974; Gottdiener, 1979; Albrecht, 1985) or the application of triangulative techniques such as respondent validation (Gottdiener, 1979; Albrecht, 1985 and Arborelius and Timpka, 1990). In this paper we suggest that both these views are problematic. The insistence of one on marginalising the role of the research process and the other on attempting to separate the process from the research data is at the expense of exploring the degree to which the process helps socially and interactionally produce the data. As we demonstrate, the activity of data collection is constitutive of the very interaction which is then subsequently available for investigation. A reflexive analysis of this relationship is therefore essential. Video generated data is an ideal resource in as far as it can provide a faithful record of the process as an aspect of the naturally occurring interaction which comprises the research topic.