A central question in online human sentence comprehension is: how are linguistic relations established between different parts of a sentence? Previous work has shown that this dependency resolution process can be computationally expensive, but the underlying reasons for this are still unclear. We argue that dependency resolution is mediated by cue-based retrieval, constrained by independently motivated working memory principles defined in a cognitive architecture (ACT-R). To demonstrate this, we investigate an unusual instance of dependency resolution, the processing of negative and positive polarity items, and confirm a surprising prediction of the cue-based retrieval model: partial cue-matches—which constitute a kind of similarity-based interference—can give rise to the intrusion of ungrammatical retrieval candidates, leading to both processing slow-downs and even errors of judgment that take the form of illusions of grammaticality in patently ungrammatical structures. A notable achievement is that good quantitative fits are achieved without adjusting the key model parameters.

on the grammatical

Abstract Similarity-based interference (SBI) has recently gained more attention in the domain of sentence processing (e.g. Gordon et al., 2007). In this paper we demonstrate that similarity can also have facilitative effects on processing, a find-ing that interference theories such as Gordon et al's cannot explain. We offer an explanation for such interference effects as well as the facilitative effects in terms of independently motivated assumptions about the structure of memory represen-tations (Hommel, 1998; inter alia). An attractive aspect of this explanation of simi-larity-based interference and facilitation effects is that so-called case-matching phenomena can also be accounted for. To this end we present two experiments: In Experiment 1 we demonstrate that case matching can occur even with non-coreferent NPs, given a sufficient level of similarity. In Experiment 2 we show that case matching is really driven by abstract case proper as opposed to other properties canonically associated with it. In sum, we provide a unified explanation for interference, facilitation and case-matching effects. A broader implication of this account is that case ambiguities are not resolved immediately but rather the multiple representations are maintained in parallel---a mechanism that is clearly not compatible with serial parsing strategies. Acknowledgments We would like to thank Gisbert Fanselow for very helpful suggestions concern-ing the design of Experiments 1 and 2, as well as Felix Engelmann, Esther Sommerfeld, and Titus von der Malsburg for extensive discussions of the ideas presented in this paper. Furthermore we thank Monique Lamers for very helpful comments on the first draft of this paper. The process of comprehending a sentence, just like any other cognitive process, requires us to use our memory. As we incrementally parse a sentence, linguistic objects have to be stored in working memory (WM) for later usage. Storage and

A central question in online human sentence comprehension is: how are linguistic relations established between different parts of a sentence? Previous work has shown that this dependency resolution process can be computationally expensive, but the underlying reasons for this are still unclear. We argue that dependency resolution is mediated by cue-based retrieval, constrained by independently motivated working memory principles defined in a cognitive architecture (ACT-R). To demonstrate this, we investigate an unusual instance of dependency resolution, the processing of negative and positive polarity items, and confirm a surprising prediction of the cue-based retrieval model: partial cue-matches—which constitute a kind of similarity-based interference—can give rise to the intrusion of ungrammatical retrieval candidates, leading to both processing slow-downs and even errors of judgment that take the form of illusions of grammaticality in patently ungrammatical structures. A notable achievement is that good quantitative fits are achieved without adjusting the key model parameters. Processing polarity items 3 Processing Polarity: How the ungrammatical intrudes on the grammatical 1.