Florian Schwarz, from the University of Pennsylvania, will present joint work with Sonja Tiemann, University of Tuebingen. Abstract below.
One prominent line of experimental research on meaning in the past decade or so has been concerned with the relative time course of different aspects of meaning unfolding in online processing, mostly with a focus on scalar implicatures vs. literal meaning. More recent work has begun to look at the processing of presuppositions as yet another layer of meaning, which is not part of the core proffered content, and which may or may not be lexically encoded according to different theoretical perspectives. Initial results in this area confirm the viability of experimentally investigating effects based on this aspect of meaning, both in terms of offline-judgment tasks and self paced reading times. In this talk, I'll discuss two experiments on German looking at the presupposition trigger 'wieder' ('again') using eye tracking during reading, which provide a more fine-grained look at the time-course of the online interpretation of presupposed content. The first part of the results speaks to the immediacy of interpreting presuppositions, as we find processing delays caused by mismatches of context and presupposed information in early reading measures. This seems more in line with theories of presuppositions which assume that they are conventionally encoded, rather than pragmatically derived, at least to the extent to which we assume that the pragmatic inferencing typically assumed to be involved would require extra processing time (as has been commonly argued for implicatures).
The second part addresses one of the central and most intriguing theoretical properties of presuppositions, namely the fact that they project out of many embedding environments, such as negation and conditionals. Descriptively speaking, we can see this as a case of a mismatch between the syntactic location of an expression and the level at which it is interpreted. One question to ask, then, is whether resolving such a mismatch and determining the level at which it is appropriate to interpret the presupposed content comes with any additional processing cost. Different theoretical accounts of projection suggest different answers to this. In particular, a representational approach such as Discourse Representation Theory (DRT) derives projection via explicit manipulations on representations, which might be costly. Non-representational theories such as dynamic semantics, on the other hand, do not provide any grounds for expecting extra costs of projection, as they do not distinguish different levels at which support for a presupposition is introduced. The experimental results based on projection out of conditionals and negation suggest that there indeed are costs associated with presupposition projection. Furthermore, the extent of these costs appears to be correlated with the distance between the location where the trigger is introduced and the level at which it is interpreted, as measured in DRT-terms, thus providing rather strong support for this type of representational perspective. Putting our results into a broader perspective, it is particularly interesting to contrast them with recent experimental findings by Chemla and Bott (2012), who argue that global (i.e., non-local) interpretations of presuppositions are faster than local ones. While there is some tension between these results, I will attempt to reconcile them, and also discuss some of the next steps in pushing this research program further.