The time-course of feature interference in agreement comprehension: Multiple mechanisms and asymmetrical attraction

Abstract

Attraction interference in language comprehension and production may be as a result of common or different processes. In the present paper, we investigate attraction interference during language comprehension, focusing on the contexts in which interference arises and the time-course of these effects. Using evidence from event-related brain potentials (ERPs) and sentence judgment times, we show that agreement attraction in comprehension is best explained as morphosyntactic interference during memory retrieval. This stands in contrast to attraction as a message-level process involving the representation of the subject NP's number features, which is a strong contributor to attraction in production. We thus argue that the cognitive antecedents of agreement attraction in comprehension are non-identical with those of attraction in production, and moreover, that attraction in comprehension is primarily a consequence of similarity-based interference in cue-based memory retrieval processes. We suggest that mechanisms responsible for attraction during language comprehension are a subset of those involved in language production.

Cues, quantification, and agreement in language comprehension

“…Both behavioral (sentence acceptability judgment accuracy) and electrophysiological measures (P600 amplitude) converged on the finding that sensitivity to agreement-violating verbs increases when the verb is preceded by a subject NP with overt plural quantification. Our results align most closely with the prediction/retrieval account of agreement comprehension outlined in the introduction (Dillon et al, 2013;Tanner et al, 2014;Wagers et al, 2009). According to this approach, morphosyntactic sensitivity-indexed here by P600 amplitude and acceptability judgment accuracy-is a function of both predictive and retrieval cue quality.…”

“…Note that our findings go beyond previous accounts of prediction and retrieval in agreement comprehension, which argued that prediction was primarily responsible for the asymmetric effects of agreement attraction with respect to sentence grammaticality (Dillon et al, 2013;Tanner et al, 2014;Wagers et al, 2009). In those studies, prediction was implicated to explain the lack of a difference in reading times or ERP effects at grammatical verbs following singular versus plural Significant and near-significant results are bolded.…”

“…Gram=grammaticality; Quant=quantification status; Ant=anteriority; Hem=hemisphere attractors (e.g., when predicted features are encountered, no retrievals are necessary, such that no attraction effects ensue), whereas retrieval was implicated in accounting for differences in the processing of ungrammatical verbs following singular versus plural attractors (e.g., determining the relative degree of processing difficulty for ungrammatical verbs). Our results show not only that prediction in agreement comprehension can be responsible for determining whether or not retrieval processes are initiated (Dillon et al, 2013;Tanner et al, 2014;Wagers et al, 2009), but also that the strength of predictions can determine the processing costs when these predictions are violated. Degree of predictability and prediction accuracy have been shown to be strong determinants of brain responses in prior ERP work on both lexical and syntactic processing, in which earlier and stronger ERP responses were associated in a gradient fashion with increased predictability and prediction accuracy of words in context (e.g., Brothers, Swaab, & Traxler, 2015;DeLong, Urbach, & Kutas, 2005;Wlotko & Federmeier, 2012) and with violations of probabilistic syntactic subcategorization biases (e.g., Osterhout et al, 1994).…”

“…Recent research has also highlighted the occasionally errorprone nature of agreement in comprehension, and specifically shown that agreement processing in comprehension can also be somewhat fragile and show gradient effects of grammaticality when a subject NP contains potentially ambiguous or misleading cues to the grammatical number of the subject (e.g., Tanner, Nicol, & Brehm, 2014;Wagers, Lau, & Phillips, 2009). In light of this, some important research goals are to better understand (1) the range of cues used in the processing of agreement dependencies (e.g., those that modulate error rates and sensitivity to agreement relations), and (2) the potential convergence or divergence in the cues and mechanisms used during language production and comprehension.…”

“…Previous research using ERPs to study agreement comprehension has shown that violations elicit a large positive-going brainwave, beginning approximately 500 ms after stimulus onset with a parietal maximum (the P600 effect), which is sometimes (but not always) preceded by a negative-going wave, sometimes with a left anterior distribution (the LAN effect; e.g., Hagoort, Brown, & Groothusen, 1993;Osterhout & Mobley, 1995;see Molinaro, Barber, & Carreiras, 2011, for a review). Importantly, the P600 is very sensitive to subtle processing differences, as the amplitude of the positivity has been shown to vary in a graded fashion with perceived ungrammaticality and general processing difficulty, including in agreement studies (e.g., Nevins, Dillon, Malhotra, & Phillips, 2007;Tanner et al, 2014; see also Kaan & Swaab, 2003;Osterhout, Holcomb, & Swinney, 1994). Thus, for quantified versus unquantified sentences, the MM/markedness account predicts no differences in P600 effect size, the prediction/retrieval account predicts larger P600 effect sizes, and the redundancy account predicts smaller P600 effect sizes.…”