Abstract
How does written and spoken feedback influence communicators' effectiveness in a shared task? Groups of two to four participants engaged in a referential communication task. The director described an array of shapes to the rest of the group via streaming video chat. In each group, one to three matchers attempted to arrange cards depicting those shapes into the director's arrangement. In one condition, matchers could speak to each other and the director through the video chat. In the other condition, matchers could only type contributions into a shared text chat. Spoken feedback dyads successfully arranged more cards than written feedback dyads. However, in groups with more than one matcher, feedback modalities were equally effective. Across group sizes and feedback types, when any matcher contributed more to the discourse, all matchers benefited. Spoken and written feedback groups differed in the amount and types of contributions matchers provided. Spoken feedback groups had more total discourse than written groups. Matchers in written groups made more bare requests for clarification, whereas matchers in spoken groups made more contributions of new information. These differences are explained in terms of constraints on grounding across the modalities.
Acknowledgments
This research was supported by faculty research funds granted by the University of California Santa Cruz to Jean E. Fox Tree and by a graduate research fellowship funded by the Perlino Foundation, Santa Cruz, California, to Nathaniel B. Clark. We thank our many research assistants who aided in data collection and coding. We also thank Jeffrey Hancock and three anonymous reviewers for comments on this manuscript.
Notes
1 Like Pearson's chi square, log linear analysis also requires that all expected values are greater than 1 and that 80% of expected values are greater than 5. Given that no alternative descriptions were observed in written feedback groups of two or four people, this assumption has been violated, which may lead to a loss of power. Accordingly, the same analysis was repeated after collapsing elaborations and alternative descriptions into a single category. Because this 2 × 3 × 2 hierarchical log linear analysis converged on the same most parsimonious model (including the three main effects, the two-way interaction of feedback modality and group size, and the two-way interaction of feedback modality and behavior type), we report on the 3 × 3 × 2 model.