So far, we have made a case for the many advantages of making audiovisual records of the activities we are interested in, and for using these records as our primary data for analysis. This stance requires, at the same time, a lively awareness of the constraints that this approach imports. We find it useful to think of the production of a video tape as a method that transforms the lived-in, real world of people's activities and joint experiences into data of a certain sort. In analyzing a tape, we are then dealing with a transformation of that world and not simply with an objective, faithful re-presentation. Transformations are always less rich than the original events that gave rise to them, so that every transformation involves some loss of information in relation to the event it captures. We would argue, however, that video loses less, and loses less seriously, than other kinds of data collection. The losses which video incurs are primarily of two kinds: the first stems from decisions made by the camera's operator; the second is inherent in characteristics of the technology.

4.1 Limits of the Operator

The person operating the camera, by pointing the equipment at one object and not another, by adjusting from zoom to wide-angle views, by setting the audio level and so on, determines who or what is visible and audible and what is not. The camera operator's notions of what is significant and what is not invariably influence the kind of record he or she produces. This points to the importance of supplemental information from concurrent fieldnotes to clarify such issues as who else is in the room but off-camera. It also points to the advantages of a fixed position for at least one of the cameras, to provide consistent coverage of the scene. Sometimes it is possible to run not only a camera or two but also several supplemental audio tape recorders placed at strategic points in the room to provide additional information. Nonetheless, much of the time, what the camera operator did not capture is subsequently unavailable to the analyst. Some of these deficiencies are due to personal or cultural bias and are remediable with foresight and experience. For example, in her first videotapes of childbirth, Jordan, at the time of the birth, zoomed in on the emerging baby, thus losing the interaction between the mother and her attendants -- a biomedical bias that became clear (and remediable at least for subsequent tapes) only during analysis. To counteract such unconscious tendencies, it is often best to leave the camera stationary.

4.2 Limits of the Technology

A second limitation of video is that video equipment is inherently more restricted in its information processing capacities than the human sensory apparatus. It produces, for example, no record of smell or of heat radiation. While human beings have available to them their full sensory capacities (color, full resolution, peripheral vision, etc.) standard video is not even capable of capturing the amount of detail visible on a medium-resolution workstation screen. More subtly, what for a human observer may be at the periphery of attention but still appreciable, may be altogether off screen in a video recording.

No matter how elaborate and sophisticated the recording setup is, the record will always be impoverished in some way or other and it is important for the analyst to be aware of that. Consider, for example, the experience of a helicopter crew who monitor up to twelve audio channels, process visual information from the ground and air, keep a number of instrument panels under surveillance, and, in addition, also talk to each other (Linde, 1988a, 1988b). No amount of instrumentation can fully capture the complexity of this situation and make it available to the analyst. In this example, the number of communication channels is unusually high. It is important to remember that this inherent partiality of the record is, in principle, no less of a problem for less complex interactions in less complex settings.

Another concern is the relationship between the record and the event as experienced by the individual. What the analyst may see or hear via the tape may or may not be what participants hear and see. For example, on a tape of teenagers playing a particularly exciting video game in an arcade, the camera, positioned behind the players, overheard comments made by the audience which were probably inaudible to the players. In the socially and technologically "simple" situations we videotaped in the past (for example a family at dinner, a couple discussing their finances, a doctor talking to her patient), we were reasonably confident that what the camera heard was also what the actors heard. It is not possible to have that kind of confidence when we videotape spatially strung out interactions with multiple actors. Work settings with complex technologies and multiple activities require multiple cameras or, at least, audio tracks. These give the analyst a privileged compound view of activities that is not necessarily available to individual participants.