Search Abstracts | Symposia | Slide Sessions | Poster Sessions
LAN effects signalling memory retrieval in wh-islands: An experiment proposal
Poster Session D, Saturday, September 13, 5:00 - 6:30 pm, Field House
This poster is part of the Sandbox Series.
Chi Hou Pau1, Robert Kluender1; 1UCSD
Introduction: There has been a debate whether island constraints are due to language processing constraints. D-linking effects in wh-islands provide an opportunity to examine this issue. It is observed that the use of D-linked wh-phrases ameliorates wh-island violations. To account for this phenomenon, it is proposed that wh-islands are the result of retrieval interference in working memory and D-linking reduces the degree of similarity-based interference (Atkinson et al., 2016 etc.). If that is the case, D-linking is expected to modulate the ERP component - Left Anterior Negativity (LAN) which is suggested to reflect working memory load during sentence processing (King & Kutas, 1995 etc.). Therefore, it is proposed to conduct an ERP experiment to examine if the use of D-linked wh-phrases in wh-islands modulates LAN effects elicited in filler-gap dependencies. Method: 25 UCSD students who are English native speakers will be recruited to join the ERP experiment. The stimuli consist of five conditions as shown below. Participants will be presented with the stimuli in pseudo-randomized order. For each stimuli, one word will be presented at a time for 300ms followed by 300ms of a blank screen (SOA=600ms). Baseline: Did you decide whether you should tell the boss about it before the meeting? Non-island with Bare wh-phrase: What did you decide that you should tell the boss about _ before the meeting? Non-island with D-linked wh-phrase: Which problem did you decide that you should tell the boss about _ before the meeting? Wh-island with Bare wh-phrase: What did you decide whether you should tell the boss about _ before the meeting? Wh-island with D-linked wh-phrase: Which problem did you decide whether you should tell the boss about _ before the meeting? Proposed data analysis: To make sure the proposed experiment design can properly elicit LAN, a linear-mixed model will be used to compare the 300-600ms time window after the onset of the preposition “before” in the Baseline condition vs. the Non-island with bare Wh-phrase condition. A significant interaction between Condition and Anteriority will indicate the current experimental design can replicate LAN effects reported in previous studies. To examine whether D-linking effects in wh-islands modulates LAN effects, another linear-mixed model with the factors Structure Type, Wh-phrase Type and Anteriority as main factors will be fitted with EPR responses during the 300-600ms time window after the onset of the preposition “before” across four conditions except the Baseline condition. Expected results and implications: A three-way interaction indicating that using D-linked wh-phrases additionally reduces LAN in sentences containing a wh-island is expected. This result is parallel to the finding that D-linking modulates the degradation in acceptability of wh-island violation sentences (Villata, 2017), supporting the idea that D-linking effects are the result of facilitation of working memory retrieval. This study will provide additional evidence for accounts ascribing island constraints to language processing constraints. Moreover, it will demonstrate that LAN signals working memory retrieval underlying language processing.
Topic Areas: Syntax and Combinatorial Semantics,