Adults with ASD More Rigid When Establishing Common Ground during a Referential Communication Task

Background:  A speaker’s ability to manage shared knowledge, or common ground, with a listener is a fundamental aspect of pragmatic language, which in turn is an area of deficit common to autism spectrum disorder (ASD). Common ground management problems observed in individuals with ASD include verbosity (de Marchena & Eigsti 2016), overspecification, and ambiguity caused by pronoun or article-definiteness reversal (Graf & Davies 2014). But little research has examined how interlocutors come to agree on the terms used to describe shared referents. The current study examines whether speakers with ASD choose their own or their interlocutors’ word labels for novel figures in a dyadic referential communication task.

Objectives:  To gauge flexibility in forming common ground.

Methods:  Adults with ASD (n=12) and age-, gender, and IQ-matched controls with typical development (TDC; n=8) completed a collaborative referential communication task designed to elicit spontaneous back-and-forth conversation in a controlled setting. (Data from 12 additional participants with ASD, and 16 additional TDCs will be available by IMFAR2017.) Participants were shown a grid of eight abstract figures and tasked with describing them to a confederate—blind to the figures’ arrangement—with the shared goal of ordering them correctly. Each participant-confederate dyad completed 5 trials (or grids), each presenting the same set of eight figures in a unique order, so that participants described each figure 5 times. Speech was transcribed, and participant’s descriptions for each figure in trials 2-5 were rated with respect to the descriptions used in immediately preceding trials. Descriptions in subsequent trials either reflected each participant’s own descriptions from the previous trial, the confederate’s, both, or neither.

Results:  Participants in both groups referred to figures largely using their own (52%) or their interlocutors’ (40%) terms from immediately preceding trials, and rarely use both (5.3%) or new terms (2.4%). There were group differences in the proportions with which participants used their own, their interlocutors’, both, or neither’s terms (p<.05), and trending differences particularly in the second trial (p=.077). Over trials 2-4, odds ratios indicate that the odds of a participant’s using their own terms were 1.148 times higher for the ASD than the TD group. For trial 2, the odds ratio was 1.62.

Conclusions: For both groups of adults with typical development and with ASD, there was variation in usage of participants’ own, versus their interlocutors’ terms when referring to previously described novel figures. These variations differed between groups. Adults with ASD were more likely to use their own terms than their interlocutor’s, potentially reflecting rigidity known to be associated with ASD. Use of one’s own description implies a refusal of one’s interlocutor’s terms, and lead to difficulty achieving or maintaining common ground in conversation.