Prediction in Sentence Processing

    • Focus in Sentence Comprehension: The notion of “focus” plays a prominent role in many areas of linguistic theory. In sentence processing, focused constituents seem to enjoy improved encoding and representation for subsequent recall (Birch & Garnsey 1995). However, the mechanisms which give rise to this benefit are still poorly understood. Studies of reading comprehension have variably found that focused constituents are read more quickly (Birch & Rayner 2010), more slowly (Birch & Rayner 1997), or no differently from non-focused constituents (Sturt et al. 2004). In this collaboration with Dr. Amanda Rysling (UCSC), we attempt to bring some clarity to these disparate findings by using a novel adaptation of the display-change paradigm to test the hypothesis that the focus benefit derives from comprehenders predictively allocating attention to likely focused constituents.
    • Active Morphosyntactic Prediction: Several studies in sentence processing suggest that comprehenders maintain expectations about likely continuations of partial inputs (Hale 2001, DeLong et al. 2005, Levy 2008, i.m.a.): material which is predictable in context seems to be processed more easily. However, researchers disagree on whether these effects should be attributed to eased integration of material in supportive contexts, or active prediction of that material in advance of bottom-up input. Collaborating with Drs. Masaya Yoshida (NU) and Dave Kush (NTNU), I am attempting to address this tension by showing that fronted reflexive pronouns lead comprehenders to predict (1) the morphological form of the matrix subject; (2) subsequent agreement morphology on the intervening auxiliary. This finding provides strong evidence of active morphosyntactic prediction which cannot be reduced to eased integration cost.

Eye-movements during Reading

    • Cumulative Progression Eye-movement Data: Traditional measures of reading difficulty require the researcher to divide the sentence into discrete, categorical “regions of interest” over which to calculate reading-time dependent measures. This technique has been immensely fruitful, but introduces at least two concerns: (1) there is considerable variability in how researchers may define regions of interest (i.e. researcher degrees of freedom); (2) effects which may be diffusely spread across several regions of text are unlikely to be uncovered by this technique. Cumulative progression analyses address these concerns by providing a continuous measure of reading difficulty (as indexed by rate of progression) after some critical inflection point. However, while this measure has the potential to be highly informative, there is not yet an agreed upon analysis technique, nor is its relation to more traditional measures well understood. Working with Dr. Klinton Bicknell, I have been engaged in redressing these gaps by replicating several well known effects from sentence processing for a direct comparison of their profile in traditional measures and cumulative progression data. In addition, we propose a non-parametric means of analyzing this continuous measure: the cluster-mass permutation test previously developed for continuous ERP data (Maris & Oostenveld 2007).

Long-distance Dependencies

    • Reflexives & Logophoricity: A large body of theoretical and psycholinguistic research is concerned with the formulation and application of principles governing anaphoric reference (Binding Theory). Recent experimental work seems to show that, in some cases, reflexive pronouns can be sensitive to antecedents which are predicted to be inaccessible on standard accounts. I investigate the possibility that a logophoric use of reflexives underlies this sensitivity using offline rating and interpretation surveys, as well as online eyetracking-while-reading experiments. These studies show that reflexives are more sensitive to non-local referents which are the subject of speech verbs, and less sensitive in the presence of indexical (first/second person) pronouns. Such findings are expected if reflexives attend to non-local referents which are, themselves, logophoric (i.e. persepctive) centers.
    • Inductive Agreement Learning: Agreement processes are often taken as primitive operations of syntactic theory, with relatively little focus on how these morphosyntactic dependencies might be induced by a learner. This project adapts the Minimal Generalization Learner developed for morpho-phonological learning by Albright and Hayes (2002) to syntactic environments in an attempt to gain traction on the agreement learning problem.
    • Case Licensing in Processing: Recent psycholinguistic investigations of verb agreement have revealed an important role for cue-based retrieval in agreement dependency processing. This project extends such work to case dependencies, showing that the same intrusion pattern observed for verb agreement can be observed in dative case dependencies in German.