Is retrieval fluency a heuristic in audience design?

BF Psychology

O'Shea, Kieran John ; Economic and Social Research Council (ESRC) ; Barr, Dale J.

University:University of Glasgow

Department:School of Psychology

关键词: Audience design, common ground, referential communication, language production.;   

Others  :  http://theses.gla.ac.uk/8414/1/2017OSheaPhD.pdf

来源: University of Glasgow

【 摘 要 】

Across three experiments, we sought to test the key assumption of Horton and Gerrig’s (2005a) memory-based model of common ground and audience design. Horton and Gerrig (2005a) argue that ordinary memory processes can serve as a proxy for more complex computations about common ground. Their key claim is that conversational partners act as memory cues for the retrieval of potentially relevant information through a process of resonance in episodic memory. Although studies have demonstrated effects in reference generation that are consistent with ordinary memory processes (Horton & Gerrig, 2005a), there has been no direct test to date of the key claim, which would require experimentally dissociating the effects of episodic memory from effects of common ground.Similarly to Horton and Gerrig (2005a), we hypothesised that memory plays a crucial role in audience design. Influenced by the work of Gann and Barr (2014), we formed an alternative retrieval fluency hypothesis for audience design. We predicted that the fluency with which a speaker’s expressions are retrieved would be dependent upon the degree that the referent and the retrieval context match the original encoding context (Tulving & Thomson, 1973). Our hypothesis proposed that expressions that were more fluent and had stronger memory signals would more likely be deemed contextually appropriate by the speaker – resulting in less consideration of context relative to expressions yielding weaker memory signals.To test this we used a referential communication game, with participants playing as ‘Director’ providing descriptions to the ‘Matcher’ experimenter. In our first two experiments, we manipulated the visual context that target items appeared in. This was a communicatively irrelevant feature of the stimuli display that was presented to participants. Whilst these manipulations were salient to the Director they were not relevant to the actual description of the target objects. This enabled us to test whether visual features in the environment (that were irrespective of common ground) cued memory during audience design performance. In our third experiment, we manipulated the Director’s perceptual experience – a communicatively relevant cue that is normally strongly correlated with common ground. In this study, we de-confounded the visual appearance of a potential addressee from the speaker’s pragmatic knowledge of whom they were interacting with. Crucially, this enabled us to directly test the assumption that episodic effects are a key source of partner specificity in reference production (Horton, 2007; Horton & Gerrig, 2005a).In Experiment 1, participants were shown a grid containing letters of various sizes and colours. We altered the appearance of the “competitor” and “foil” items, which alternated between training and test trials, so that participants had to adapt their descriptions at the test phase in order to avoid misspecifying descriptions. We expected speakers to experience greater retrieval fluency when the visual context in test trials was highly similar to the training trial configuration. It was predicted that this would result in them continuing to use the same description as before - making more descriptive errors than when presented with configurations that were dissimilar between the training and test phase. However, we found a lack of support for our hypothesis, as there was no main effect of visual context on reference production.In Experiment 2 minor adjustments were made to the configuration and sequencing of objects and the stimuli presented to participants. In this experiment, Directors described pictures of everyday objects to the Matcher. Experiment 2 provided weak statistical support in favour of the retrieval fluency hypothesis for audience design and suggested that visual context impacted upon reference generation. More specifically, participants appeared to rely on the strength of the memory signal present when designing descriptions for the listener.In Experiment 3, participants described target items to one of two Matchers using an interactive webcam design. At the test phase, the visual experience of the Director (participant) was controlled independently of the pragmatic situation, meaning that who the Director saw and whom they were speaking to did not always coincide. To the extent speakers use memory as a proxy for common ground, we expected misspecifications to be higher when participants viewed the same Matcher as they saw when they originally entrained on descriptions during training (effect of visual consistency). Furthermore, to the extent they use common ground, we expected misspecification to depend on their knowledge of who hears the description (effect of pragmatic consistency). Contrary to the memory-based model, there was no evidence that speakers misspecified more when viewing the same Matcher than when viewing a different Matcher. We also found no significant difference in misspecification rate when speakers believed that they had addressed the same Matcher versus a different Matcher.In all three experiments we found a high misspecification rate in referential descriptions, indicating clear evidence of reliance on episodic memory. However, we did not find evidence in support of the retrieval fluency hypothesis for audience design. Our results also failed to support the key claims outlined in Horton and Gerrig’s (2005a) memory-based model. In particular, the results of Experiment 3 cast doubt on the assumption of partner specificity in audience design.