Portal de Periódicos da CAPES

From Minor Mishap to Major Catastrophe: Lexical Choice in Miscommunication

2013; Wiley; Volume: 35; Issue: 35 Linguagem: Inglês

1551-6709

Jennifer M. Roche, Alexandra Paxton, Alyssa Ibarra, Michael K. Tanenhaus,

Language, Metaphor, and Cognition

From Minor Mishap to Major Catastrophe: Lexical Choice in Miscommunication Jennifer M. Roche (jroche@bcs.rochester.edu) a Alexandra Paxton (paxton.alexandra@gmail.com) b Alyssa Ibarra (aibarra@bcs.rochester.edu) a Michael K. Tanenhaus (mtan@bcs.rochester.edu) a a b Brain and Cognitive Sciences, University of Rochester Rochester, NY 14627 USA Cognitive and Information Sciences, University of California, Merced Merced, CA 95343 USA Abstract Miscommunication is often regarded as noise or uninformative in psycholinguistic research. However, Coupland et al. (1991) suggest that miscommunication can provide rich information about how interlocutors come to communicate successfully. Successful communication necessarily needs the individuals involved to coordinate and update their mutual knowledge, experiences, beliefs, and assumptions. However, the process of updating this information may be ridden with unsuccessful attempts that eventually help interlocutors reach a common goal. This study evaluates the relative contribution of linguistic factors to communicative success, based on verbal grounding (e.g., mutual agreement on a referent) and visual congruency (e.g., interlocutor’s visual environments match or mismatch) during a collaborative task. We show that varying levels of communicative success are laden with rich linguistic information that may uncover interesting aspects of successful and less successful communication. Keywords: Joint action; grounding; successful communication; miscommunication; psycholinguistics. Introduction Interactive language, in particular face-to-face interactive conversation, is the most canonical form of language use (Clark, 1992; Goodwin & Duranti, 1992). In interactive conversation, interlocutors are typically both speakers and listeners (addressees) and they often are conversing to achieve joint goals. Nonetheless, most research on human language processing focuses on the speaker and the listener as individual cognitive agents in non-interactive tasks. There are important exceptions. For example, a large body of work has used the Edinburgh Map Task (Brown et al., 1983) to address a range of psycholinguistic issues. In this task two interlocutors collaborate, with the director guiding the matcher to reproduce a route printed on the director’s map. Aist and colleagues developed a “fruit cart” domain as a vehicle for eliciting human language production for (a) dialogue system research and development and (b) psycholinguistic research (Aist, Campana Allen, Swift & Tanenhaus, 2012). Senft (2002, 2007) developed a number of domains to evaluate lexical choice in spatial terms during a space game and tinker toy task for cross-cultural analysis. Brown-Schmidt, Tanenhaus and colleagues have adopted a complementary strategy, using targeted language games to produce trial-like structure in unrestricted interactive conversation to address specific psycholinguistic issues with real-time response measures, such as visual world eye- tracking (e.g., Brown-Schmidt, Gunlogson & Tanenhaus, 2008; Brown-Schmidt & Tanenhaus, 2008). In this paper we provide a preliminary report on a project using a new domain intended to examine how referential domains are constructed, updated, accepted and rejected during a goal-driven task, with naive participants and unrestricted speech. Here we examine how the language used in grounding might be diagnostic of, and contribute to, miscommunication. The domain is similar to those discussed by Sentf (2002) and is designed to allow a face-to-face interaction through a barrier separating the two participants. This task involves a collaborative dyadic interaction that required participants to instruct each other in building a bŸlocŸo™ animal figure from abstract three-dimensional puzzle-like pieces (see Figure 1 for an image of the animals; Methods for full task description). The bŸlocŸo™ paradigm was created to serve a number of purposes. First, we wanted a domain that would lend itself to investigating both generation and interpretation of referring expressions. Secondly, we wanted to observe how referring expressions change when the goals change. For example, during the build stage pieces that were initially referred to using conceptual pacts, such as “the Christmas tree” would eventually assume a different identity, “the body” (see the green item on the left of Figure 2). This domain offers a rich domain for investigating grounding. The domain creates a corpus that contained frequent communication failures (e.g., confusions and misunderstandings) that had to be resolved. These failures in communication are often regarded as noise and therefore uninformative in psycholinguistic research (Coupland, Giles, & Weimann, 1991; Keysar, 2007). However, as