![Sample our Language & Literature journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5941/TOC-Subject-Sample-2021-Language-Literature.jpg"})
ABSTRACT
This study attempts to reveal the mechanisms behind the online formation of Wh-Filler-Gap Dependencies (WhFGD). Specifically, we aim to uncover the way in which maintenance and retrieval work in WhFGD processing, by paying special attention to the information that is retrieved when the gap is recognized. We use the agreement attraction phenomenon (Wagers, M. W., Lau, E. F., & Phillips, C. (2009). Agreement attraction in comprehension: Representations and processes. Journal of Memory and Language, 61(2), 206–237) as a probe. The first and second experiments examined the type of information that is maintained and how maintenance is motivated, investigating the retrieved information at the gap for reactivated fillers and definite NPs. The third experiment examined the role of the retrieval, comparing reactivated and active fillers. We contend that the information being accessed reflects the extent to which the filler is maintained, where the reader is able to access fine-grained information including category information as well as a representation of both the head and the modifier at the verb.
Acknowledgements
We are grateful to the anonymous reviewers for their invaluable suggestions and comments. We would like to also thank Brian Dillon, Yunyan Duan, Matt Goldrick, Shota Momma, Colin Phillips, Jeff Runner, Shayne Sloggett, Julie Van Dyke, Matt Wagers, Alexis Wellwood, members of the Syntax, Semantics and the Sentence Processing Lab at Northwestern, audiences at Haskins Laboratories, Center for Research in Language, the Language and Literacy group at Beckman Institute, the audiences of CUNY2018, CUNY2019, and the 93th annual meeting of the Linguistics Society of America for their invaluable discussions and comments. Parts of the earlier versions were presented at the CUNY 2017, 2018, 2019, and the 93th annual meeting of the LSA.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 Note that we do not commit to a specific analysis of WhFGD constructions. Specifically, we are agnostic about whether it involves a phonetically empty gap or not. We customarily call the controlling element as gap, but our conclusions do not necessarily require a gap-based analysis.
2 As an anonymous reviewer points out, some previous studies have suggested that decay is no longer a useful explanatory concept in the retrieval literature (see Berman, Jonides, & Lewis, Citation2009; Lewandowsky, Geiger, & Oberauer, Citation2008; McElree, Citation2006; Nairne, Citation2002). Wagers and Phillips (Citation2009, Citation2014) pointed out that not all the features of the elements that are retrieved at the head of the dependency or are fully reactivated at the verb position (e.g. semantic features of the wh-filler in Wagers and Phillips’s (Citation2014) study). Such findings can be accounted for by memory decay. Thus, for present purposes, we hypothesize that some of the information associated with the filler is subject to decay or interference. We assume that the success of retrieval is related to the amount of material intervening between the filler and the gap.
3 As an anonymous reviewer pointed out, Lewis and Vasishth (Citation2005) suggests that retrieval could occur in such a way that the parser can re-instate information into comprehenders’ focus of attention, in order to process that information. In this sense, if information were already in comprehenders’ focus of attention due to maintenance, there is no need for it to be retrieved. However, following Wagers and Phillips (Citation2014), we argue that comprehenders discharge some components associated with the features from focal attention and this information must be retrieved when the verb is processed.
4 Gaps are indicated by an underscore “__” in a sentence.
5 Note, the relative head needs to be linked to the embedded verb, but this is not relevant to the wh-gap dependency formation in terms of wh-question formation.
6 As an anonymous reviewer pointed out, decay should have less impact on the offline experiments as readers can look back at the left context anytime, to remember the content of the antecedent. Our purpose of the offline experiments was to understand how the availability of the contexts can influence the retrieval of different kinds of information.
7 As an anonymous reviewer has pointed out, it is possible that the absence of an agreement attraction at the verb might be due to the nature of the self-paced reading experiment. It has been well known that in self-paced reading experiments, the expected effect can be observed in one or two regions after the critical region (the spill-over effect; Vasishth & Lewis, Citation2006). Therefore, it is possible that, even if the agreement attraction effect is caused at the verb region, it would not be observed right on the verb region but in spill-over regions.
8 Following an anonymous reviewer’s suggestion, we also examined the region immediately preceding the verb (i.e. the pre-critical region). The results showed a main effect of Grammaticality (β = −0.03, SE = 0.01, t = −2.49, p < 0.05) but no main effect of Local noun (β = 0.00, SE = 0.01, t = 0.07, p > 0.05) as well as no interaction between Local noun and Grammaticality (β = 0.04, SE = 0.02, t = 1.49, p > 0.05). This further suggests that the effects we observe are not due to spillover effects from the prior regions.
9 An anonymous reviewer suggested that the recognition of the gap is not due to grammatical constraints such as the CSC and the ATB restriction. It could be the case that the readers recognize the presence of the gap due to the combination of the coordinating connective, and, and an adverb. If the combination of the coordinating connective and an adverb (… and certainly … ) helps reactivate the filler, then our assumption must be weakened, i.e. the reactivation of the filler is not due to the grammatical constraints. However, as Wagers & Phillips (Citation2009) showed, the gap in the coordinated structure and parasitic gap within an adjunct clause, which is optional, show different reactivation profiles. Therefore, it is still plausible that ATB/CSC plays a role in the reactivation of the wh-filler. As we do not have any evidence to distinguish the two hypotheses, we would like to leave this point open at this point.
10 As an anonymous reviewer pointed out, there is an alternative explanation for Experiment 3 that would not rely on the reactivated vs. active distinction, but rather, on differences in cue-based retrieval. In (5b), the attachment site of the RC is actually ambiguous, such that “that will be disastrous” could modify either “mistake” or “program(s)”. If readers prefer to attach the RC low, to “program(s)”, then according to cue-based retrieval this noun phrase will be reactivated, rendering it more active in memory. This would yield stronger attraction rates at the main verb (“is/are”), since the local noun will have higher activation (and thus interfere more) in (5b) than in (5a). However, if the attachment of RC modulates the accessibility of the lower noun, we also predict a similarity-based interference effect. In other words, the local noun should be more accessible across-the-board and thus should give rise to an interference effect whether the agreement is grammatical or ungrammatical. This should not predict the illusion of grammaticality we observed, but rather an agreement attraction effect in both grammatical and ungrammatical conditions.
11 We thank an anonymous reviewer for pointing this out.
12 Note that, under the cue-based retrieval model, if the gap increases activation specifically for the head of the wh-phrase but not the local noun (Nicenboim et al., Citation2015), then the weaker agreement attraction is predicted for the reactivated filler relative to the active filler. We would like to note this as a possible alternative hypothesis.
