The processing of pronominal relative clauses: Evidence from eye movements

Relative clauses have played a key role in distinguishing between different theories of language comprehension. A reversal in processing costs between full NP and pronominal relative clauses reported by Reali and Christiansen (2007) has been used to argue for expectation-based theories of comprehension (e.g., Hale, 2001; Levy, 2008), and against memory-based theories of comprehension (e.g., Gibson, 1998, 2000; Gordon, Hendrick, & Johnson, 2001; Lewis, Vasishth, & Van Dyke, 2006). We present results relying on eye-movements during reading, in conjunction with modeling of differences between self-paced reading and eye movement data, to argue that the results observed by Reali and Christiansen and others are due to the self-paced reading paradigm, and do not reflect an actual reversal in processing costs. Overall, our results suggest that a combination of memory-based factors and spillover explains the pattern of reading times observed in various relative clause experiments such as those in Reali and Christiansen (2007), and that while comprehenders' expectations undeniably play a role in language comprehension, the role may be less dramatic than is suggested by previous studies.