Information Theory and Ellipsis Redundancy

Project B3

The overarching goal of project B3 is to investigate to what extent the usage of ellipsis is determined by information-theoretic concepts like surprisal and entropy. This perspective on ellipsis complements existing linguistic theories, since the latter only tackle the question under which syntactic and semantic conditions an ellipsis is, in principle, grammatically licensed. Focusing on discourse-initial fragments and constituent ellipsis in phases I and II, respectively, we found strong evidence that the speaker’s choice between alternative encodings is essentially guided by two pragmatic imperatives of rational communication: “Avoid peaks!” and “Avoid troughs!”. In phase III, we will shift our focus to the remaining major class of ellipses: coordination ellipsis. To tie in with phases I and II, we first aim at confirming our results on the avoidance of peaks and troughs also in the area of coordination ellipsis. At the same time, we aim at strengthening the audience design hypothesis, a crucial prerequisite for Uniform Information Density (UID), in more interactive settings. Furthermore, the dynamic nature of coordination opens a window into the way the processing of prosodic information affects the recipient’s common ground, changes her expectations, and modulates the predictability of ellipsis as a function of time. Especially with right node raising, we expect a trade-off between surprisal and memory here. Finally, B3 is a hub for bringing together the research conducted in the CRC, which is at the interface between information structure and information theory.


Keywords: theoretical linguistics, psycholingustics, ellipsis