Learning welding #4 - Deconstructing video data

We are now in the process of transcribing relevant video data. At present, doing this by watching and then coding snippets which we have identified as of relevance to our study. We are following steps suggested in Erickson’s (2006) recommendations for the steps to take when analysing video data and Jae Major’s advise. Snippets are tracked by recording time code. The columns in our transcript table are to record the time code, transcribe any conversations / interactions / verbal outputs, record activity being undertaken and record any non-verbal actions.

The next step is to study the above for patterns. Time codes can be matched to verbal transcripts, activities and non-verbal actions. Frequency of activities / non-verbal actions may provide some patterns. Verbal outputs are studied for themes. We have started to match activities and non-verbal actions to verbal themes. This provides for a better representation of the ‘whole’ picture of how interactions take place.


There will be a need at some stage to break the above down into more precise categories. In non-verbal activities, there is a need to code for the various forms of kinesics and haptics which are the predominant non-verbal activities.

Transcription of the video data takes an inordinate amount of time!! My speed is around 10 minutes for each minute of video. Flip, who is learning the process, takes double my time. In comparison, I usually take about 2 to 3 minutes to transcribe each minute of interview audio recording. This warns us about the need to be very precise as to our overall purpose for using video for gathering research evidence. Not that serendipity will be ignored :) but focus will be important in order not to be bogged down with large amounts of evidence and too many disparate themes to follow through. With the welding class in recess, we will have time over the Easter to re-evaluate our research question / purpose and begin next term with a better focus on the types of activities we will record.