| Abstract: | A Markov grammar was developed to account for the sequence of conversational moves in discourse about transgressions. This discourse interrupts ongoing activity and functions to repair the social fabric when inappropriate or awkward behavior has attracted the attention of interactants. A corpus of 1248 such interchanges, obtained in 90 hr of observation in kindergarten through fourth grade classrooms, was the basis of the grammar. Beginning with theoretical assumptions about the canonical form of these interchanges, the process of grammar development was explicitly described and the direction given by pragmatic considerations (such as attaining a balance between parsimony and accuracy) was discussed. The resultant grammar accounted for 83% of the corpus and served as a framework within which age and context effects on the discourse were explored. Loglinear analyses of multiway contingency tables at each transition point in the grammar revealed that the presence vs absence of an adult in the interchange affected the relative preferences for different alternatives at one juncture, but that subjects' ages did not influence the sequencing of discourse as described by the model. The overall approach was considered in relation to the more traditional uses of Markov models and its strengths, weaknesses, and future promise for discourse analysis were discussed. |
