Frontiers in Psychology (May 2015)
Verb-based anticipatory processing in aphasia
Abstract
Healthy listeners use verb and agent+verb constraints to anticipate likely arguments (Altmann & Kamide, 1999; Kamide et al., 2003). Listeners also show early looks to unlikely but possible arguments (Boland, 2005; Borovsky, et al., 2012), suggesting that coarse-grained verb-based semantic constraints may influence early processing (Kuperberg, 2013). This study investigated the roles of verb and agent+verb constraints on expectations about likely and unlikely arguments in people with aphasia (PWA) and healthy age-matched controls. In two visual-world experiments, PWA (n=9) and older adults (n=27) heard sentences truncated at the determiner and clicked on the image that “best finished” the sentence (Mack, et al, 2013). In Experiment 1, sentences with a semantically-constraining vs. unconstraining verb (Someone will eat/move the …) were accompanied by images of: a likely target (cake), an unlikely but possible competitor (branch), and two unrelated impossible distractors (pail, van). In Experiment 2, sentences with a semantically-constraining vs. unconstraining agent+verb combination (The dog/Someone will drink the …) were accompanied by: a likely target (water), an unlikely but possible competitor (coffee), a semantically-related impossible distractor (cat – semantic associate of constraining agent), and an unrelated impossible distractor (rocks). By-participants and by-items ANOVAs showed that likelihood guided looks for both participant groups in both experiments. In the constrained condition in Experiment 1, gaze proportion was reliably higher to the target (cake) than the two impossible distractors (pail, van), with controls showing this effect in a 400-ms bin starting 400 ms after verb onset (Fig1a) and PWA approximately 400ms later, in a 400-ms bin starting 800 ms after verb onset (Fig1b). During the same time windows in the constrained condition in Experiment 2, both groups were more likely to gaze at the target (water) than the unlikely competitor (coffee) (controls, Fig1c; PWA, Fig1d). This confirms that verb constraints guide argument expectations in PWA (Mack et al., 2013), and provides novel evidence that agent+verb constraints do the same. These findings suggest that predictive processing may be preserved among PWA (Hanne, et al., 2015), particularly when grounded in event-related likelihood (McRae & Matsuki, 2009). Both participant groups also looked at unlikely but possible arguments, though for PWA this was reliable only in Expt 2. In a 400-ms bin beginning 800 ms after verb onset, gazes to the unlikely competitor (Exp. 1: branch; Exp. 2: coffee) became more likely than gazes to impossible distractors (Exp. 1: controls: ps<.01; PWA: n.s.; Exp. 2: controls: ps<.05; PWA: p1<.025, p2=.09). For the PWA only in Experiment 2, gazes to the unlikely competitor and semantically-related impossible argument patterned together. Facilitation for unlikely potential arguments is consistent with findings for young healthy adults (Kamide, et al., 2003; see Kukona, et al., 2011) and suggests that participants were unable to ignore bottom-up verb constraints (see Kukona, et al., 2014). The fact that PWA showed this interference, as well as interference from the agent-related distractor, suggests that they experience outsize interference from salient but grammatically-unlicensed semantic information (Dickey, et al., 2007; see also Milberg & Blumstein, 1981).
Keywords
