Using structural priming to test links between constructions: English caused-motion and resultative sentences inhibit each other

Abstract

Cognitive-linguistic theories commonly model speakers’ grammatical knowledge as a network of constructions related by a variety of associative links. The present study proposes that structural priming can provide psycholinguistic evidence of such links, and crucially, that the method can be extended to non-alternating constructions (i.e., constructions that differ in both form and meaning). In a comprehension priming experiment using the “maze” variant of self-paced reading, English caused-motion sentences were found to have an inhibitory effect by slowing down participants’ subsequent processing of resultatives, and vice versa, providing evidence that speakers store distinct but related representations for the constructions. Priming effects of a similar magnitude emerged in both directions, suggesting that the constructions are bidirectionally related, while not supporting previous claims about a metaphorical asymmetry between the patterns. Moreover, priming was only marginally affected when prime and target contained the same rather than different verbs, demonstrating that cross-constructional priming in comprehension can be observed in the absence of a “lexical boost”. The results raise questions for follow-up research on the role of inhibition in the grammatical network and the extension of structural priming to other types of constructional links.

Constructionist approaches: Past, present, future

“…between benefactive and dative sentences; Ziegler & Snedeker 2018). As a result, priming effects are regarded as one of the strongest sources of evidence for the network model (Diessel 2019: 204;Ungerer 2021Ungerer , 2022.…”

“…priming) and the time course of language acquisition, but that there is no evidence that speakers store a separate schema for each group of similar constructions. In contrast to this position, Ungerer (2021Ungerer ( , 2022 suggests that priming effects can be equally interpreted as evidence for horizontal links and vertically related schemas, especially if it assumed that both links and schemas can vary in "strength", i.e. in their degree of entrenchment (Hilpert 2015;Langacker 2017; see also Schmid [2020: 234] for a gradient concept of the "likelihood of schematic representations").…”

“…In addition, distributional semantic methods such as semantic vector space analysis (see Lenci 2018 for an overview) yield quantitative measures of the semantic similarity between lexemes -or, if averaging over those lexemes, of the abstract constructions in which they occur -based on their collocational profiles (Hilpert & Perek 2022;Percillier 2020). On the experimental side, priming effects, in particular, are regarded as an important indicator of constructional similarity (Perek 2015;Ungerer 2021Ungerer , 2022, given that priming tends to be stronger the more similar prime and target are (Branigan & Pickering 2017; see also Section 4.1).…”

A-maze of Natural Stories: Comprehension and surprisal in the Maze task

“…Boyce et al (2020) introduced several implementational innovations that made it easier for researchers to use Maze, and showed that for several controlled syntactic processing contrasts (Witzel et al, 2012) Maze offers better statistical power than self-paced reading, the other word-by-word response time method easy to use remotely. Maze has since had rapid uptake in the language processing community (Chacón et al, 2021;Levinson, 2022;Lieburg et al, 2022;Orth & Yoshida, 2022;Ungerer, 2021).…”

A-maze of Natural Stories: Comprehension and surprisal in the Maze task

“…Boyce, Futrell, and Levy (2020) introduced several implementational innovations that made it easier for researchers to use Maze, and showed that for several controlled syntactic processing contrasts (Witzel, Witzel, and Forster 2012) Maze offers better statistical power than self-paced reading, the other word-by-word response time method easy to use remotely. Maze has since had rapid uptake in the language processing com-1 munity (Chacón et al 2021;Ungerer 2021;Orth and Yoshida 2022;Lieburg, Hartsuiker, and Bernolet 2022;Levinson 2022). However, there is increasing interest in collecting data during comprehension of more naturalistic materials such as stories and news articles (Demberg and Keller 2008;Luke and Christianson 2016;Futrell et al 2020), which offer potentially improved ecological validity and larger scale data in comparison with repeated presentation of isolated sentences out of context.…”