Research Article: Does social context impact metacognition? Evidence from stereotype threat in a visual search task

Date Published: April 15, 2019

Publisher: Public Library of Science

Author(s): Thibault Gajdos, Isabelle Régner, Pascal Huguet, Marine Hainguerlot, Jean-Christophe Vergnaud, Jérôme Sackur, Vincent de Gardelle, Mark Alexander Williams.


While recent studies have emphasized the role of metacognitive judgments in social interactions, whether social context might reciprocally impact individuals’ metacognition remains an open question. It has been proposed that such might be the case in situations involving stereotype threat. Here, we provide the first empirical test of this hypothesis. Using a visual search task, we asked participants, on a trial-by-trial basis, to monitor the unfolding and accuracy of their search processes, and we developed a computational model to measure the accuracy of their metacognition. Results indicated that stereotype threat enhanced metacognitive monitoring of both outcomes and processes. Our study thus shows that social context can actually affect metacognition.

Partial Text

Metacognition, i.e., the process of monitoring and controlling one’s own cognitive processes [1], plays a crucial role in the regulation of our behavior [2]. It might be either implicit, involving automatic cognitive processes, or explicit, relying on conscious reflection. Recent research demonstrated the importance of both forms of metacognition in social interactions (see [3], for a review). For instance, it has been shown that, even in a simple visual task, dyads perform better than each member separately, when their members can share their confidence about their visual perceptions [4].

All outcomes were primarily analyzed through generalized hierarchical linear mixed-effects regressions with target type (either X or L), the number of displayed items (set size) and their interactions and treatment (i.e. threat vs. no-threat), sex, frame (gain vs. loss) and identification (high vs. low) and their interactions as fixed effects. The model thus contains the intercept, the effect of target type, set-size and their interactions, as well as the frame, as random-effects. We focus on the effects of interest in the main text of the manuscript, and in particular on the four-way interaction involving sex, stereotype threat treatment, frame and identification. The full tables of the regression results are presented in the supplementary material.

The present study aimed at quantifying the effects of stereotype threat on metacognition during visual search. As expected from previous research on visual search tasks, we first found that performance decreased with the number of distractors, an effect that was more pronounced for “Ls” targets than for “Xs” [52]. Replicating these classic results enabled to build a model for the number of inspected items [40,41], and to evaluate whether participants had a good metacognitive access to this variable, by defining a new measure of metacognitive monitoring we called SNSI error. In addition, we used Brier scores to measure metacognitive monitoring based on confidence judgments. We found that stereotype threat did not affect performance, but did affect both measures of metacognitive monitoring. We will now discuss these two aspects in turn.