0% 20% 40% 60% 80% 100% sing.-subj. sing.-N2 sing.-subj. pl.-N2 pl.-subj. sing.-N2 The consequences of number agreement on number interpretation Ellen Lau 1 • Matthew Wagers 2,3 • Colin Phillips 1 University of Maryland Linguistics (1) • NYU Psychology (2) • UC Santa Cruz Linguistics (3) Agreement attraction in comprehension: independence of combinatorial operations? • A plural noun can interfere with singular subject-verb agreement, even though it does not head the subject projection: The key to the cabinets were on the table. • These errors are common in production (Bock & Miller, 1991) and elicit an illusion of grammaticality in comprehension (Pearlmutter, Garnsey, & Bock, 1999). • We have argued that such ‘agreement attraction’ in comprehension is due to errors in the cue-based retrieval of number features from memory (Wagers, Lau, & Phillips, in press; Badecker & Lewis, 2007). Agreement attraction provides an opportunity to investigate questions about the independence of the multiple processes necessary to combine new material with the prior sentence context. Current Study: Is agreement independent from the conceptual interpretation of subject number? The key to the cabinets were … C Design & Predictions Complete complex-NP attraction sentences with plural predicates: Control: The defect in the appliance was surprisingly numerous. Attraction: The defect in the appliances were surprisingly numerous. Predictions: • If the erroneous retrieval of a plural feature in attraction also results in a revised interpretation of the subject as plural, readers should incorrectly accept a plural predicate like ‘numerous’ • If computing the formal agreement relationship is carried out independently of the subject’s conceptual interpretation, readers should correctly reject the plural predicate at the same rate. Experiment 1: Controlling for noun number Perhaps just the presence of an embedded plural noun will improve acceptability of plural predicates, regardless of whether a plural verb was present to induce agreement attraction: The defect in the appliances appeared to be surprisingly numerous If comprehenders are sloppy in this way, our manipulation won’t tell us much about agreement computation. Therefore, we began with a control experiment with non-number marked verbs. Methods: • Speeded acceptability task, following RSVP: 350 ms/word, 2 s for judgment. • 114 items: 30 experimental items, 84 fillers. 54% gram.; 46% ungram. • Participants: 24 native speakers; Analysis: Logistic mixed-effect regression % judged accurately Experiment 2: Agreement attraction • Experiment 1 showed that a plural embedded NP by itself does not improve acceptability. • Critical question: when combined in an attraction configuration with a plural verb, does the agreement attraction result in a plural subject interpretation? The defect in the appliances were surprisingly numerous. Methods: Same as Exp. 1, with number-marked verbs. % judged accurately 0% 20% 40% 60% 80% 100% singular subj. singular subj. +attraction plural subj. ..defect in the appliances…was…serious.. ..were…numerous.. ..were…serious.. ..was..numerous.. ..defects in the appliance…were…serious.. ..defects in the appliance…were…numerous.. (% Y) (% N) (% N) (% N) (% Y) (% Y) Conclusions References Bock & Miller, 1991. Cognitive Psychology. Badecker & Lewis, 2007. Talk presented at CUNY, San Diego. Pearlmutter, Garnsey, and Bock, 1999. Journal of Memory and Language. Wagers, Lau & Phillips, in press. Journal of Memory and Language. Lau, Wagers, Stroud, & Phillips, 2008. Talk presented at CUNY, Chapel Hill. Wagers, Lau, & Phillips, 2008. Poster presented at CUNY, Chapel Hill. Thornton & MacDonald, 2003. Journal of Memory and Language. Vigliocco, Butterworth, & Semenza, 1995. Journal of Memory and Language. • The grammatical illusion of the plural verb being acceptable is not linked to a plural interpretation of the subject. • Previous work suggests agreement attraction is also not linked to misinterpretation of the subject role. • Under the view that agreement attraction in comprehension is due to errors in cue-based retrieval at the verb, these findings converge in suggesting that single features of a constituent can be retrieved to satisfy licensing requirements without forcing realignment of tightly linked relationships. • However, thematic, conceptual, and agreement information likely can interact when they are simultaneously available, as in production. • Main accuracy effect of incongruent predicate (p < .01), apparently due to yes-bias in this task (Wagers et al., 2008). • Main effect of plural N2 (p < .01), perhaps due to inherent complexity of processing plurals (Wagers et al., in press). • However, no interaction; plural N2 does not selectively improve acceptability of plural predicates (p > .1). • Main effects of attraction (p < .01) and incongruent predicate (p < .05); since these items were unacceptable, lower accuracy consistent with yes- bias in task. • However, no interaction between predicate and attraction (p > .1); attraction does not selectively improve acceptability of plural predicates CUNY 2008: Is agreement independent from the analysis of grammatical roles? • Does the incorrect analysis of agreement result in an incorrect interpretation of the number of the subject? • CUNY 2008 (Lau, Wagers, Stroud & Phillips): Does the incorrect analysis of agreement result in an incorrect analysis of subjecthood? • Production studies show that thematic and conceptual-semantic factors can affect production of agreement (e.g. Thornton & MacDonald, 2003; Vigliocco et al., 1995) • If agreement attraction reflects a retrieval error, this retrieval could result in reassignment of the subject. • In fact, reading time results suggested that agreement attraction does NOT result in re-analysis of the subject. • This result suggests that the formal number features can be retrieved independently of the computation of grammatical roles. ..defect in the appliance…seemed…serious ..numerous… ..defect in the appliances..seemed serious.. ..numerous… ..defects in the appliance…seemed…serious.. ..defects in the appliance…seemed…numerous.. (% Y) (% N) (% Y) (% Y) (% Y) (% N)