ABSTRACT
The underlying cognitive mechanisms explaining why speakers sometimes make communication errors are not well understood. Some scholars have theorized that audience design engages automatic processes when a listener is present; others argue that it relies on effortful resources, even if a listener is present. We hypothesized that working memory is engaged during communicative audience design and that the extent to which working memory is engaged relies on individual differences in cognitive abilities and concurrent amount of resources available. In Experiment 1, participants completed a referential task under high, low, or no cognitive load with a present listener whose perspective differed from the speaker’s. Speakers made few referential errors under no and low load, but errors increased when cognitive load was highest. In Experiment 2, the listener was absent. Speakers made few referential errors under no and low load, but errors increased when cognitive load was highest, suggesting that audience design is still effortful under high cognitive load, both when the listener was present and when the listener was absent. Experiment 3 tested whether cognitive abilities predicted communication performance. Participants with higher fluid intelligence and working memory capacity made fewer communication errors. Our findings suggest that communication relies on available cognitive resources, and therefore errors occur as a function of factors like cognitive load and individual differences.
Acknowledgments
We thank Mark Thornton and Erin Sherman for providing data collection assistance and Anthony Ryan for designing and creating the apparatus used in the study. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Notes
1. Inter-rater reliability of a subset of responses in Experiment 3 was examined to ensure the reliability of the coding system. The second author coded each of the four trials for a random subset (10%) of responses from Experiment 3 (a total of 40 responses). The results were then compared to the original responses coded by the first author. Pearson correlations showed an inter-rater reliability of r = .82 for egocentric errors and .75 for descriptors.
2. Generally, time describing was significantly correlated with descriptors (r = .87 in Experiment 1 and r = .60 in Experiment 2), but not with egocentric errors (r = .14 in Experiment 1 and r = −.00 in Experiment 2), except in Experiment 3 in which there was a moderate-to-low significant correlation with errors (r = .29) and a strong correlation with descriptors (r = .76). This indicates that, in our study, time describing was beneficial for the conversation (i.e., it was positively correlated with descriptors, but not with egocentric errors).