Why are there limits on theory of mind use? Evidence from adults' ability to follow instructions from an ignorant speaker

Abstract

Keysar et al. (Keysar, Barr, Balin, & Brauner, 2000; Keysar, Lin, & Barr, 2003) report that adults frequently failed to use their conceptual competence for theory of mind (ToM) in an online communication game where they needed to take account of a speaker's perspective. The current research reports 3 experiments investigating the cognitive processes contributing to adults' errors. In Experiments 1 and 2 the frequency of adults' failure to use ToM was unaffected by perspective switching. In Experiment 3 adults made more errors when interpreting instructions according to the speaker's perspective than according to an arbitrary rule. We suggest that adults are efficient at switching perspectives, but that actually using what another person knows to interpret what they say is relatively inefficient, giving rise to egocentric errors during communication.

Keywords:

• Theory of mind
• Adult
• Cognition
• Executive function
• False belief

View correction statement:

Corrigendum

Acknowledgments

The research reported in this paper was supported by a grant from the Leverhulme Trust, F/0094/AH.

Notes

¹ We only considered this measure an estimate of response time because it included time taken to identify the correct referent and execute the motor response of clicking the mouse on the referent. Nonetheless, this measure does afford a check that participants are not merely trading off accuracy against speed.

² Confusingly, the illustrative example used in the text of Keysar et al. (2003) is actually ambiguous rather than relational, but the stimuli used in these experiments and in Keysar et al. (2000) were all relational.

³ Here and throughout, control trials were not included when comparing error rates across conditions because they showed zero variance (there were no errors).

⁴ An alternative analysis in which the missing data were replaced with the group mean values from the relevant condition yielded exactly the same pattern of significant and nonsignificant results.

⁵ This interaction (but not the main effects) was significant in an alternative analysis in which the missing data were replaced with the group mean values from the relevant condition. Separate comparisons of experimental and control trials for informed and ignorant perspectives showed neither difference to be significant, but the trend was for slower responses to experimental trials than to control trials when the speaker was ignorant and the opposite trend when the speaker was informed.