Distinguishing discrete and gradient category structure in language: Insights from verb-particle constructions

Laurel Brehm*, Matthew Goldrick

*Corresponding author for this work

Research output: Contribution to journalArticle

2 Scopus citations

Abstract

The current work uses memory errors to examine the mental representation of verb-particle constructions (VPCs; e.g., make up the story, cut up the meat). Some evidence suggests that VPCs are represented by a cline in which the relationship between the VPC and its component elements ranges from highly transparent (cut up) to highly idiosyncratic (make up). Other evidence supports a multiple class representation, characterizing VPCs as belonging to discretely separated classes differing in semantic and syntactic structure. We outline a novel paradigm to investigate the representation of VPCs in which we elicit illusory conjunctions, or memory errors sensitive to syntactic structure. We then use a novel application of piecewise regression to demonstrate that the resulting error pattern follows a cline rather than discrete classes. A preregistered replication verifies these findings, and a final preregistered study verifies that these errors reflect syntactic structure. This provides evidence for gradient rather than discrete representations across levels of representation in language processing.

Original languageEnglish (US)
Pages (from-to)1537-1556
Number of pages20
JournalJournal of Experimental Psychology: Learning Memory and Cognition
Volume43
Issue number10
DOIs
StatePublished - Oct 2017

Keywords

  • Break point modeling
  • Gradient symbolic computation
  • Illusory conjunction
  • Mental representation
  • Sentence processing

ASJC Scopus subject areas

  • Language and Linguistics
  • Experimental and Cognitive Psychology
  • Linguistics and Language

Fingerprint Dive into the research topics of 'Distinguishing discrete and gradient category structure in language: Insights from verb-particle constructions'. Together they form a unique fingerprint.

  • Cite this