2018 Friday Session C 1200

Friday, November 2, 2018 | Session C, Terrace Lounge | 12pm

Incremental Interpretation in the First and Second Language
L. Brehm, K. Miller, C. Jackson

In conversation, listeners sometimes must interpret erroneous utterances. First- (L1) and second-language (L2) listeners can successfully interpret the intended meaning of utterances containing syntactic errors or semantic anomalies (e.g., Levy et al., 2009; Frazier & Clifton, 2015; Futrell & Gibson, 2017), but less is known about how interpretations unfold over time. The present study investigates how L1 English speakers and proficient L1 Spanish-L2 English speakers process and interpret anomalous utterances. In so doing, this study addresses open questions regarding how L1 and L2 speakers deal with conflicting linguistic cues when retrieving information from memory during language comprehension (Cunnings, 2017).

Thirty-two L1 English speakers and 28 L1 Spanish-L2 English speakers listened to sentences containing (a) possible syntactic anomalies, manipulated by subject-verb agreement (SVA) or (b) possible semantic anomalies, manipulated by a preposition (without-blend; see Figure 1). We tracked participants’ attention to images of sentence referents during comprehension using visual world eye-tracking. This was paired with an offline interpretation measure where participants selected the image matching the sentence subject.

For SVA sentences, participants most frequently selected the literal head noun (key) as sentence subject. Both groups were more likely to select the non-literal head (keys) when the sentence contained a local plural noun (cabinets; β=-1.98, SE=0.40, p<0.001) or plural verb (*were; β=-4.23, SE=0.48, p<0.001; see Figure 2). L2 speakers also selected the non-literal head more often than L1 speakers (β=1.46, SE=0.66, p=0.03). While listening to sentences, both groups were more likely to fixate an alternate version of the head (keys) in regions following ungrammatical verbs (*were; see Figure 3). This occurred in two phases: Participants fixated the head foil keys after hearing a local plural noun (cabinets) from 0 to 500 msec after verb onset (β=-0.05, SE=0.01, p<0.01), and after an ungrammatical plural verb (*were) from 500 to 1000 msec after verb onset (β=-0.10, SE=0.01, p<0.001). These results show L1 and L2 speakers process number information incrementally, attending to alternate repairing referents as needed.

For without-blend sentences, participants most frequently selected the literal subject (woman wearing hat), but were more likely to select the non-literal subject (woman not wearing hat) if the sentence contained the anomalous preposition off (β=-5.95, SE=1.12, p<0.001; see Figure 2). Again, L2 speakers were more likely to select the non-literal subject than L1 speakers (β=3.82, SE=1.01, p<0.001). While listening to the sentence, both groups were more likely to fixate an alternate version of the subject (woman not wearing hat) in a time-window 500 to 1000 msec after hearing the anomalous preposition (β=-0.09, SE=0.04, p=0.04; see Figure 3). Again, this shows that L1 and L2 speakers adapt their interpretations incrementally by considering repairing referents.

These data show that both L1 and L2 speakers understand anomalies by probabilistically revising interpretations over time. Differences in rates of non-literal interpretations for L2 versus L1 speakers also underscore the difficulty of memory retrieval during L2 processing: L2 speakers are more likely to rely on the most recently encountered cue to guide their interpretation of utterances.