The work of Interaction Analysis is time-consuming and labor-intensive. In addition to the time spent in group working sessions, Interaction Analysis requires a considerable investment in the effort of transcription and in the painstaking searching of tapes for instances of particular activities and events. Moreover, this work cannot easily be delegated to assistants since a deep understanding of the phenomenon of interest requires proceeding through successive approximations until the relevant analytic categories are identified.

If, then, the production and analysis of videotapes is so expensive in researcher time and effort, why would one want to opt for this type of data? In the following pages we argue that for certain kinds of research (though clearly not for all) video-based Interaction Analysis may be the optimal choice.

3.1 Reconstructing the Event

Selectively employed video analysis is a particularly valuable analytic tool for the study of learning activities and work practices in complex real world settings for a number of reasons. One of these is that by approximating direct observation, video provides a shared resource to overcome gaps between what people say they do and what they, in fact, do. Video provides optimal data when we are interested in what "really" happened rather than in accounts of what happened.

To make the importance of that distinction clearer, let us differentiate between two types of data available for the analysis of social interaction. One type is generated by direct observation, the other, by telling some sort of "story" about an event. The "story" may be expressed in words, in numbers, diagrams, check-marks, or other kinds of symbolic representations produced by informants as they answer researchers' questions, or by researchers themselves when they write down fieldnotes, make sketches, or tally observations. In each case the event of interest is not so much "present-ed" (in the sense of "drawn-into-the-present") but re-presented in processed form. It is reconstructed. As Bergmann (1985) has pointed out, the reconstruction occurs through a variety of methods and interpretive devices that, by transforming and reducing reality, invariably and unavoidably import meaning into events. He notes that retrospective representation assigns a secondary web of meanings to the original events as they happened, so that the past events that once played themselves out according to then-existing relevancies are interpretatively re-created and re-constructed during data collection. As Bergmann suggests, the crucial point here is that secondary interpretation has crept into what we think of as the primary data and the researcher has no chance of reversing this process. In a fundamental sense, the events themselves have disappeared; what passes as data is actually their reconstruction. Practically all of our data, from interviews to field observations to records of experiments, fall into this category.¹⁷

Video recordings replace the bias of the researcher with the bias of the machine. The recording process becomes, to some extent, automated, and thereby removed to a greater degree than other methods from the reconstructive bias of individual researchers. Videotaping is not shaped by the requirements of a successful story or the demands of a coding sheet -- rather it is subject to the limitations of the technology. It is a process that is essentially passive (but see section 4). Unlike fieldnotes or stories that highlight "important" aspects and pass over "unimportant" ones, video records social events as they occur and with a level of detail that is unattainable for methods that rely on reconstruction. The camera's bias is consistent. Thus we would argue that videotaping, the mechanical audiovisual fixation of an event, produces data much closer to the event itself than other kinds of re-presentation.

One reason for relying on video, then, as the preferred kind of data for our analyses is that we would like our theorizing to be responsive to the phenomenon itself rather than to the characteristics of the representational systems that reconstruct it and thereby constrain the direction of the analyst's thinking.

3.2 Permanence of the Primary Record

There are other reasons for favoring video recordings (and the output from other kinds of electronic recording devices) over data generated by non-electronic methods. A key characteristic of such data is the permanence of the primary record in all its richness. This makes possible an unlimited number of viewings and listenings, in IAL-like group settings as well as by individual analysts. It is in the course of repeated viewing that previously invisible phenomena become apparent and increasingly deeper orders of regularity in actors' behaviors reveal themselves. In addition, a videotape can be played in slow or accelerated motion, thereby exposing otherwise unseen patterns in the movements of persons or artifacts. Moreover, the audiovisual record itself can be made available to other investigators who may examine it to extend or refute the original analysis,¹⁸ or for quite different interests of their own. Video recording creates permanent primary records as resources that can be shared between researchers and practitioners whose activities are recorded, facilitating reflective review by both. Video records thus can support true collaborative group work.

3.3 Complexity of Interaction Data

But there are yet more fundamental reasons for favoring videotaping as a method for the collection of primary data. Even a trained observer cannot keep track of the overlapping activities of several persons with any accuracy or any hope of catching adequate detail. Consider an excerpt from fieldnotes, a paper-and-pencil snapshot of childbirth in a Maya village: "midwife bathing baby; mother in hammock; father out to bury placenta; grandmother rummaging in cardboard box." By contrast, the videotape provides an incomparably richer record. The kind of talk (or silence) going on at the time, the procedural details of the bath, the mother's eyes on the infant, the grandmother's rummaging for oil and baby wraps, the looks, the body orientations, all these are lost and are not recoverable from the researcher's memory. Nor are they recoverable from the memory of participants by interviewing after the fact (Jordan, 1992a). In classroom research, ethnographers are forever frustrated by the necessity to decide if they should focus on the teacher, a particular student, or a group of students. Similarly, in complex work settings such as an airlines operations room, a single investigator with paper and pencil is simply overwhelmed by the necessity to keep track of multiple operators interacting not only with each other but also with headquarters, pilots, ground crews, maintenance people, baggage handlers, gate agents, and so on.¹⁹

Furthermore, the essence of most manipulative procedures -- be they those of a traditional midwife turning a fetus that is in the wrong position for birth, those of a pair of air traffic controllers arranging flight-status strips on their work panels, or those of a secretary moving a cursor while text editing -- resists capture in words, both because of the density of behavioral details and because there is no ready descriptive vocabulary for bodily behavior which could capture such activity in notes. It is for these reasons that we opt for video data for many of our research questions.