na

Comparing the questions -proposed for this discourse panel with those identified for the TINLAP-2 panel eight years ago, it becomes clear that some progress has been made in discourse studies. While TINLAP-2's questions seem concerned with identifying an ideal model of discourse (e.g., what constitutes an appropriate model of discourse, how domain-dependent must such models be, what makes discourse coherent, and how can relationships between utterances best be characterized), TINLAP-3's questions reflect the debates which have attended individual attempts to define such models (e.g., are current discourse theories testable, are plan-based theories really feasible, how are processing theories defined and what have their authors failed to take into account, and what aspects of discourse study are really subsumed by more general problems of general communication or knowledge representation). However, in one area -the role of speech in discourse research -the questions posed at TINLAP-3 appear remarkably similar to those posed at TINLAP-2: essentially, do we need to study speech as well as text to study discourse? While a simple 'yes' is probably no more controversial now than it would have been in 1978, I think we can make an even stronger claim today: that while one may still study text in silence, one cannot study discourse without taking speech into account.