You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster D66, Thursday, November 9, 6:15 – 7:30 pm, Harborview and Loch Raven Ballrooms

Prosodic lengthening and boundary prediction in nominal compounds: An ERP study

Alicia Parrish1, Patrick Kelley1, Kaylin Smith1, Yan Cong1, Alan Beretta1;1Michigan State University

Recent studies suggest prosody plays an important role in auditory processing of compound nouns (CNs) (Koester, 2004; 2014). Good (2008) reports a reliable difference in the length of nouns, whereby a noun is shorter when used as the first constituent in a compound (N1) rather than as a single noun (SN). This lengthening may be due to a SN’s likelihood of occurring at a phrase boundary rather than a word boundary, as with an N1 (Turk & Shattuck-Hufnagel, 2007). Here, we test whether this lengthening of SNs is a sufficient cue for structural prediction. Stimuli varying in prosody (SN-prosody vs. CN-prosody) and phrase type (SN vs. CN) were constructed. For all compounds, both N1 and the second compound constituent (N2) were disyllabic nouns with word-initial stress (e.g. desert [N1] island [N2]). These compounds were situated within sentence frames where either the whole compound or just N1 was present (e.g. Cooper travelled to the [desert] / [desert island] in search of solitude). Splicing was completed to replace each N1 with an N1 that had CN- or SN-prosody. Thus, the sentence frames contained spliced nouns with prosody that was either congruent or incongruent with the frame. Our stimuli reliably differed in duration such that the average length of CNs (first syllable 178ms; second syllable 168ms) was significantly shorter than SNs (first syllable 200ms; second syllable 194ms), both ps<0.01. Twenty-five right-handed native English speakers participated in an auditory ERP study in which each participant heard 100 experimental sentences with 100 fillers in pseudorandomized order. Signals were recorded with a 32-channel cap, average referenced, and analyzed using two ROIs (Inferior & Superior) following Hestvik (2013). 200-700ms post N1, CN phrase types showed a significant interaction of ROI x prosody, p=0.017; SN phrase types were nonsignificant, p=0.459. Pairwise comparisons indicate a positivity of the incongruent SN-prosody within CN phrase types within the Superior region, p<0.01; while the Inferior region was nonsignificant, p=0.12. A Holm-corrected pairwise t-test within the Superior region revealed a difference of CN- vs. SN-prosody in the CN phrase type, p=0.02, but not the SN phrase type, p=0.97. This difference suggests SN-prosody is predictive, but CN prosody is not. As SNs exhibit lengthening, this feature’s presence cues the parser to predict a phrase boundary, whereas an absence provides no cue. With CN frames, when SN-prosody is heard, a boundary is predicted, requiring reanalysis when N2 (e.g., “island”) appears. When CN-prosody is heard, no prosody-based prediction is made, thus there is no need to reanalyze. In SN frames, SN-prosody elicits a prediction that there is a boundary, which is borne out, so no reanalysis occurs; while CN-prosody, again, elicits no prediction. In line with Kotzor et al. (2017) and Chladkova et al. (2015), we conclude that this asymmetry in processing supports the view of an underspecification of lengthening in its mental representation as a prosodic feature, extending their perception findings into the realm of syntactic predictions.

Topic Area: Perception: Auditory

Back to Poster Schedule