Players of Matching Pennies automatically imitate opponents' gestures against strong incentives

Abstract

There is a large body of evidence of apparently spontaneous mimicry in humans. This phenomenon has been described as "automatic imitation" and attributed to a mirror neuron system, but there is little direct evidence that it is involuntary rather than intentional. Cook et al. supplied the first such evidence in a unique strategic game design that gave all subjects a pecuniary incentive to avoid imitation [Cook R, Bird G, Lünser G, Huck S, Heyes C (2012) Proc Biol Sci 279(1729):780-786]. Subjects played Rock-Paper-Scissors repeatedly in matches between fixed pairs, sometimes with one and sometimes with both subjects blindfolded. The frequency of draws in the blind-blind condition was at chance, but in the blind-sighted condition it was significantly higher, suggesting automatic imitation had occurred. Automatic imitation would raise novel issues concerning how strategic interactions are modeled in game theory and social science; however, inferring automatic imitation requires significant incentives to avoid it, and subjects' incentives were less than 3 US cents per 60-game match. We replaced Cook et al.'s Rock-Paper-Scissors with a Matching Pennies game, which allows far stronger incentives to avoid imitation for some subjects, with equally strong incentives to imitate for others. Our results are important in providing evidence of automatic imitation against significant incentives. That some of our subjects had incentives to imitate also enables us clearly to distinguish intentional responding from automatic imitation, and we find evidence that both occur. Thus, our results strongly confirm the occurrence of automatic imitation, and illuminate the way that automatic and intentional processes interact in a strategic context.

Fig. 1.

The START (A) position and the OPEN HAND (B) and CLOSED HAND (C) gestures.

Fig. 2.

Estimated intentional and automatic responses, for early vs. late games. Probit regression coefficients ± 1 SE corresponding to probit regressions with player fixed-effects and SEs clustered at the pair level.