One of the more surprising (or at least, counter-intuitive) findings in current sentence processing is the fact that fully ambiguous structures like (1) seem to be processed more easily than their disambiguated counterparts like (2): The brother of the colonel who shot himself on the balcony had been very depressed. The daughter of the colonel who shot himself on the balcony had been very depressed. This so-called “ambiguity advantage” in sentence processing (Traxler, Pickering, & Clifton 1998) poses a serious challenge to attempts at unnifying syntactic and lexical ambiguity resolution strategies (MacDonald, Pearlmutter, & Seidenberg 1994, i.m.a.), and there remains widespread disagreement over the preferred explanation. In this talk, we focus on addressing one prominent account of the ambiguity advantage: strategic underspecification. This hypothesis suggests that syntactic ambiguity is easier to process because comprehenders may not fully parse material when the task context doesn’t demand it (Swets, Desmet, Clifton, & Ferreira, 2008). We address this hypothesis by attempting to replicate the initial findings of Swets et al. in three different task contexts: self-paced reading, eye-tracking while reading, and the maze task. Our findings consistently show that the task context does not (straightforwardly) impact the ambiguity advantage, favouring alternative explanations of the phenomenon (Van Gompel, Pickering, & Traxler, 2001; Levy, 2008; Logačev & Vasishth, 2016; Dillon, Andrews, Rotello, & Wagers, 2019). We conclude by suggesting (i) our findings reinforce the difference between syntactic and lexical processing (ii) we need to be cautious when applying “good enough” explanations to effects of interest.
Shayne Sloggett has been the Experimental Officer in Psycholinguistics at the University of York’s Department of Language and Linguistic Science since September 2019. His research interests are sentence processing, syntax, and the interaction of grammatical knowledge and sentence comprehension routines. This research draws on insights from linguistic theory to better inform psycholinguistic models (and vice versa), using a range of methods from formal acceptability rating tasks, to eye-tracking while reading and corpus linguistics.