IDENTIFICATION OF SEQUENTIAL DEPENDENCIES IN CONVERSATIONS - A PASCALPROGRAM

A Pascal program that provides an analysis of the sequential dependenc ies between conversational events (eg, speech acts) generated by two s ources is described It produces z scores, chi-square values, and confi dence limits for the difference between estimated unconditional probab ilities of the occurrence of selected events (valid events) and estima ted conditional probabilities of these events following specified targ et events immediately (at Lag 1) and, if required, more distantly (at Lags 2, 3, etc.). Valid events and target events may be independently selected from the set of events generated by the first source, the sec ond source, or both. Events not selected are ignored The user may also elect to exclude the target event from the set of valid events (an ad justment for autocorrelation). The program is especially appropriate f or conversational events because it does not assume that two events fr om the same category can never follow one another.