2.1 The Ethnographic Context

There are various types of research in the course of which videotapes are produced. Our own practice has been to do videotaping in conjunction with ethnographic fieldwork. We rely on participant observation, in-situ interviewing, historical reconstruction, and the analysis of artifacts, documents, and networks for providing the framing context. In the course of this ethnographic work, we attempt to identify interactional "hot spots" -- sites of activity for which videotaping promises to be productive. Ethnographic information then furnishes the background against which video analysis is carried out while the detailed understanding provided by the micro-analysis of interaction, in turn, informs our general ethnographic understanding.

2.2 Content Logs

Once a video or audio tape has been made in the field, our first step towards analysis is likely to be a "content log" or "content listing." Content logs are best made as soon as possible after the tape is recorded because then the researcher's memory is still fresh, allowing annotations and explications of events that may not be possible later. Content listings are indexed by tape counter number for audio tape and by time stamp for video tape. They consist of a heading that gives identifying information, followed by a very rough summary listing of events as they occur on the tape. The level of detail is determined by the interests of the researcher and the available time. No attempt is made at this stage to provide either consistency or evenness in coverage. Content listings are useful for providing a quick overview of the data corpus, for locating particular sequences and issues, and as a basis for doing full transcripts of particularly interesting segments (see section 2.4). For an example of a content log see, Appendix D. Depending on the stage of progress towards transcription, either a content log or a transcript may be available for what often constitutes the next step: group work.

2.3 Group Work

Interaction Analysis, as practiced in our labs, makes prominent use of multi-disciplinary collaborative work groups that are assembled for particular projects. While in most video-based research investigators approach the tape with a preconceived coding scheme (for example Bales' categories or task analysis), we attempt to keep our work, to the largest extent possible, free from predetermined analytic categories. We expect such categories to emerge out of our deepening understanding of the orderliness of the interaction as participants on the tape make this orderliness visible to each other. In the course of multiple replayings, finer and finer levels of participants' social competence and their resources for mutual construction of meanings become apparent. At the same time, our working in groups reveals and challenges idiosyncratic biases on the part of individual analysts.

Group work is also essential for incorporating novices since Interaction Analysis is difficult to describe and is best learned by doing. Much in the manner of apprentices, newcomers are gradually socialized into an ongoing community of practice where they increasingly participate in the work of analysis, of theorizing, and of constructing appropriate representations of the activities studied.

Apart from working groups assembled for particular projects, a further resource available to us is the Interaction Analysis Laboratory (IAL), an ongoing permanent forum where researchers from different projects present tapes. The IAL meets for two or three hours each week. Researchers ("owners") bring tapes, often with transcripts or content listings, from their respective projects and after a brief introduction to the setting of the recorded activities and any special interests the researcher may have, the group works together to analyze the tape.

The tape is played with one person, usually the owner, at the controls. It is stopped whenever a participant finds something worthy of remark. Group members propose observations and hypotheses about the activity on the tape, searching for specific distinguishing practices within a particular domain or for identifiable regularities in the interactions observed. Proposed hypotheses must be of the kind for which the tape in question (or some related tape) could provide confirming or disconfirming evidence. The idea is to ground assertions about what is happening on the tape in the materials at hand. To escape the ever-present temptation to engage in ungrounded speculation, some groups have imposed a rule that a tape cannot be stopped for more than five minutes. This means in practice that rambling group discussions are discouraged and that no single participant can speculate for very long without being called upon to ground her or his argument in the empirical evidence, that is to say, in renewed recourse to the tape.

During group working sessions a large number of observations and hypotheses are produced and audio recorded for later review and more extensive (or intensive) analysis by the owner. Just as often, questions are generated which can only be answered by returning to the field and doing further, now more focused, ethnographic fieldwork or more targeted videotaping.

Collaborative viewing is particularly powerful for neutralizing preconceived notions on the part of researchers, and discourages the tendency to see in the interaction what one is conditioned to see or even wants to see. For example, in certain circumstances we expect people who are smiling at each other and who are in close physical proximity, also to touch each other. Observers have frequently reported that they have seen such touches even though on replay it is clear that none occurred. Similarly, on a tape of interaction between a mother and her small child baking muffins together, analysts were impressed with the mother's easy-going ways that left much of the activity up to the child. After viewing a particular segment, the question arose whether the child had accomplished the entire task of filling the muffin tins with batter by herself. The IAL group of professional observers were quite convinced that the mother, while assuming a supportive and appreciative position, had left the actual task to the child. On replaying the tape it turned out, however, that this was not the case (CAN.90.02.19.IAL).⁷

Errors of this sort are invisible in a paper-and-pencil record because there is no opportunity to go back and re-examine what happened. In contrast, a tape segment can be played over and over again, and questions of what is actually on the tape versus what observers think they saw, can be resolved by recourse to the tape as the final authority. This repeated and always jolting experience of having one's confidence in what one thinks one saw shaken, instills a healthy skepticism regarding the validity of observations that were made without the possibility of rechecking the primary record.⁸ Hutchins (1991) refers to this tendency as "confirmation bias," and explains it as the propensity to affirm prior interpretations while discounting or even ignoring counter-evidence. Group work dedicated to collaborative, interdisciplinary analysis is an effective antidote to this problem.

Ungrounded speculation about what individuals on the tape might be thinking or intending is discouraged in Interaction Analysis working groups. However, evidence about thinking and intending, such as errors in verbal production or certain gestures and movements, can be cited. For example, on a tape of a working session at a whiteboard, one of the participants stretches way up to reach the upper corner of the board as he begins to write.⁹ For an Interaction Analysis group, this would probably constitute evidence that he is projecting and orienting to further writing during this working session, for which he wants to leave space. An observation of this sort might be formulated as he "intends" to write a lot on the board.

IAL participants consistently attempt to talk about "mental states" and "mental events" in ways that are grounded in what happens on the tape. For example, in one working session a videotape from a school setting was examined for evidence of differential understandings on the part of students. On the tape, a group of four high school students are checking their math home work in class. One reads out her answer in a doubting, self-deprecatory way. Two others say that they also had trouble with this particular problem. The fourth gives the correct answer, a lengthy statement ending with: "... the two lines contact the same point." Is there anything that can be said about what is going on "in the heads" of these students from the tape? It turns out that there is considerable evidence for the quality of understanding present in three of them in their audible and visible reactions to the correct answer given by the fourth. The first student briefly writes an annotation to her own response and is immediately ready to go on to the next problem. She evidently sees in what ways her answer is deficient and knows how to remedy that deficiency. The other two, however, write extensively. It appears that they are copying the answer verbatim. One asks: "the two lines ... ?" soliciting completion of the statement, apparently unable to rephrase the correct answer for her own purposes (CAN91.07.23.IAL.VPP).¹⁰ Though all four students end up with the correct answer in their notebooks, we would argue from evidence on the tape that the degree to which they "own" these answers differs dramatically.

It is not the case, then, that intentions, motivations, understandings, and other internal states cannot be talked about in Interaction Analysis. Rather that they can be talked about only by reference to evidence on the tape.

2.4 The Individual Researcher's Work

After the IAL or similar working group has worked for one or several sessions on a particular tape, we are left with some hours of audio tapes documenting the analytic thinking of the group. A researcher can then extract interesting materials from these audio tapes by partially transcribing them, a process we refer to as "cannibalizing the audio tape." He or she is then left with a number -- often very large -- of potentially significant observations about the phenomena he or she is interested in. In the process of cannibalizing, some initial observations are thrown out, others are reclassified, and the significance of still others may be understood quite differently from what was originally proposed by the group.

It goes without saying that the very process of looking is informed by some notions of what one is interested in looking for, notions which, in turn, are modified by what it is that one finds as one gets deeper into the analysis. Nevertheless, at any one point when promising hypotheses have been formulated, it is incumbent on the researcher to assess which observations are indicative of general patterns, which are idiosyncratic or random perturbations, and which are due to some as yet unexplained (or unexplainable) cause. This is done by finding other instances of the event in question in the data corpus and checking whether the proposed generalization holds. In practice, Interaction Analysts often make use of "collection tapes," edited videotapes on which instances of the event in question are assembled. Played one after another, these provide a convincing demonstration that the phenomenon identified is or is not robust across instances.¹¹

For example, on one videotape of an American hospital birth it was noted that at the beginning of a particular uterine contraction the eyes of all those present -- nurse, husband, medical student -- go to the electronic fetal monitor by the woman's bedside. A behavioral pattern was proposed that when a monitor is present, birth attendants' eyes will move to the equipment when a contraction begins. The pattern was checked against other contractions in this particular labor, against tapes of other American hospital births where monitors were present, and against monitored contractions in European hospitals. It was found that the pattern held overwhelmingly. In the few cases where it did not, evidence for some competing local activity was available to explain the discrepancy. (For example, the woman was very distraught or the doctor was just at that time doing an examination and therefore had his eyes on the patient rather than the monitor.) The next step was to examine videotapes of births where monitors are not used. It was found that cross culturally in the absence of monitoring equipment the focus of attention almost always shifts to the woman when a contraction begins. Finally, a generalization was proposed which states that in the presence of high-technology equipment the attentional focus of medical personnel as well as of non-staff attendants moves from the patient to the machinery. This shift has important consequences not only for how the woman experiences the birth but also for the organization of care giving as it shifts from relying on data generated by the woman's experience to data generated by the machine.¹²

Interaction Analysis thus tends to proceed inductively, attempting to generate statements about general patterns from multiple sets of empirical observations. It heeds Geertzâs (1973) warning that theory should stay "rather closer to the ground than tends to be the case in sciences more able to give themselves over to imaginative abstraction" (p. 24).

2.5 Transcription

As particular tape segments emerge as significant, content logs are expanded into transcriptions. These may be more or less elaborate and detailed, depending on the nature of the researcher's analytic interests. Minimally, they contain a representation of participants' talk, since speech is always important in human interaction. They may also contain annotations for nonverbal behaviors, such as changes in body position, gaze, gesture, and the like, or for object manipulation, document processing, and employment of certain technologies. Sometimes it is important to transcribe what participants type on computers, particularly if the computers link interacting participants and provide an alternate or supplementary channel to voice communication. If the interest is specifically in human-computer interaction, then it may be screen-based activity itself that needs to be transcribed, such as cursor movement, highlighting, or mouse button depression. In general, whatever social or material features of the situation might be relevant to the analysis are candidates for inclusion in the transcript. It makes sense, then, for researchers to think very seriously about what kind of analysis they intend to do before launching into full-scale transcription, because the choice of what to transcribe determines what will be available for analysis. Thus, a transcript that does not contain information on pauses in talk, on gaze, or shifts in body position, obviously cannot support an analysis in which those elements play a role. Nonetheless, it is impossible to include all potentially relevant aspects of an interaction, so that, in practice, the transcript emerges as an iteratively modified document that increasingly reflects the categories the analyst has found relevant to her or his analysis. Kendon (personal communication) talks about this as "locally relevant transcription." ¹³

Some practitioners argue that "everything" should be transcribed because even if, say, pauses or overlaps are not germane to the current analysis, some other researcher might want to use the same materials for checking findings or for novel analytic purposes. In conversation analysis, transcription conventions (see for example, Jeffersonâs transcription, Appendix E), have often been considered the model for comprehensive transcription. But it is clear that these conventions, too, leave out certain phenomena that could be transcribed. Thus there is no ideal or complete transcript according to any abstract standard. Rather, the question must be: how adequate is this transcript for purposes of the analysis to be performed?¹⁴

Because there are substantive analytic insights to be gained during transcription, many researchers choose to do their own transcription. For those who delegate some part of the process the financial cost is substantial. A very rough transcript of a very simple conversation produced by a clerical typist requires on the order of three hours for every hour of tape. To produce a transcript usable for analytic purposes, each hour of audio tape requires something like 10 to 20 hours for transcription, with the ratio depending on sound quality and number of speakers. By the time nonverbal interaction is added, the investment in researchers' time becomes substantial indeed.¹⁵ What has emerged in our practice is that we allow the extent and detail of transcription to be driven by our analytic interests, so that those features of the interaction that emerge as significant in the course of tape analysis are more comprehensively and exhaustively transcribed, while others are set aside until shown to be relevant to the analysis.

At the present time, the form which transcripts take is changing rapidly, in part because of opportunities offered by emerging computer technologies. As technological support for different transcription schemes becomes more widely available, it becomes even more important to pay attention to the implications of committing to one type or another. (For more detail on transcription strategies, see Appendix A: Issues in Transcription and Representation.)

2.6 Video Review Sessions

Apart from the data generated by researchers' reviews of the tapes, both in group work and individually, video records are also useful for eliciting detailed information from participants themselves. Such data represent the participants' perspective, their view of the world, which may contrast substantially with the analyst's.¹⁶

In practice, the individuals whose interactions have been recorded are asked to come to a viewing session with the research team. Some researchers use the tape to elicit specific information from informants about actions and events whose significance is unclear to them. This often proves more productive than returning to the field for additional data collection, in particular where the informant's comfort is at stake or when publicly sensitive issues are under discussion. For example, Jordan, during fieldwork in Mexico, asked individual village women to look at videotapes of public gatherings in the privacy of her hut, a strategy which produced extensive accounts of interpersonal, economic, and ritual relationships in the village

Other researchers ask informants to stop the tape whenever anything strikes them as significant. This gives some idea of how participants parse the event -- where they see significant segments as beginning and ending. It also gives information on troubles that may be invisible to the analyst, on resources and methods used by participants to solve their problems, and on many other issues of importance to participants. For example, Frankel asked patients and physicians who had been videotaped during medical consultations to stop the tape when they saw something of interest. He found that there was a substantial overlap in where they stopped the tape; but the explanations patients and physicians gave for why they stopped the tape there were widely divergent, indicating substantially different views of what their interactions were about (Frankel, 1983).

Elicitation based on video tapes has the advantage of staying much closer to the actual events than if one were to ask questions removed from the activity of interest. Thus, instead of interviewing designers about their practices (or, even more removed, asking them to fill out a questionnaire) one might ask them to look at a videotape of themselves or of other designers at work and ask questions about that work as they arise from the activity being viewed. Data elicited in this manner are likely to have greater "ecological validity," that is to say, are more readily applicable to real conditions of work than data generated under more artificial circumstances.