Kelly Hoogervorst1, Leah Banellis1, Francesca Fardo1,2, Kai Xue3, Dobromir Rahnev3, Micah Allen1,4
1Center of Functionally Integrative Neuroscience, Aarhus University, Aarhus, Denmark
2Danish Pain Research Center, Aarhus University, Aarhus, Denmark
3School of Psychology, Georgia Institute of Technology, Atlanta, GA, USA
4Cambridge Psychiatry, Cambridge University, Cambridge, UK
Metacognition refers to the ability to monitor our cognitive and perceptual processes. Previous research has investigated gender differences in self-confidence and self-esteem, but comparatively little is known about gender differences in metacognition. To address this gap, we conducted a study of 317 individuals, who completed global and local measures indexing recognition memory, semantic memory, and visual sensitivity. We found both domain-specific and domain-general effects spanning global and local metacognition measures. Women exhibit a systematic negative bias in metacognitive self-beliefs and were also found to update these beliefs more negatively than men. A similar metacognition bias was found for women, which was driven by greater sensitivity to errors made during testing. Using computational modelling, we could show that this more conservative strategy is accompanied by an enhanced global metacognitive efficiency in women. These results highlight previously unknown but robust gender differences in metacognitive self-monitoring.
https://osf.io/preprints/psyarxiv/2gwky
- Kelly Hoogervorst - Lead Author: Data Wrangling, Analysis, Writing
- Leah Banellis - Analysis and Writing
- Francesca Fardo - Analysis and Writing
- Kai Xue - Analysis and Writing
- Dobromir Rahnev - Analysis and Writing
- Micah Allen - Senior Author: Project Design, Supervision, Analysis, Writing
- Center of Functionally Integrative Neuroscience, Aarhus University
- Danish Pain Research Center, Aarhus University
- School of Psychology, Georgia Institute of Technology
- Department of Psychiatry, Cambridge University
This work was supported by a Lundbeck Foundation Fellowship (R272-2017-4345), an ERC Starting Grant (ERC-2020-StG-948788), as well as generous support from the Aarhus Institute of Advance Studies. Francesca Fardo was supported by another ERC Starting Grant (ERC-2020-StG-948838). We are deeply grateful the COBELab participants and staff who contributed their time and effort to make this study possible.
To reference this work in your research, please cite as follows:
Hoogervorst, K., Banellis, L., Fardo, F., Xue, K., Rahnev, D., & Allen, M. (2024, January 31). Gender Differences in Metacognition: Global and Local Contrasts in Bias and Efficiency. Retrieved from osf.io/preprints/psyarxiv/2gwky
Top left: recognition memory, top right: visual discrimination, bottom left: general socioeconomic knowledge, bottom right: general nutritional knowledge. Each trial of each of the tested domains was followed by a retrospective discrete confidence rating ranging from 1 to 7 to measure metacognition. All tasks followed a 2AFC design to maximise comparability between domains. The timeline at the bottom of the figure highlights the global metacognition measures before and after completing the testing battery in randomised order.
Panel A) presents a comparison of metacognitive self-beliefs before and after cognitive testing (Pre vs. Post) in females and males, averaged across the four task modalities. Each line represents an individual participant, with the colour indicating the degree of change in belief. Box Plots depict median and interquartile range for each level of gender and time. A significant gender by time interaction effect (p = 0.003) was found, indicating that females exhibit a greater negative update in self-belief post-testing. Panel B) illustrates the average self-belief for memory, vision, general domain performance (GDP), and calories estimation tasks, separated by gender. Error bars represent 95% confidence intervals. A significant interaction effect (p < 0.001) was observed, indicating that gender differences in metacognitive self-belief are domain-specific. Notably, females exhibited a consistent negative bias across all domains with statistically significant decreases in the GDP (p < 0.001) and vision domains (p = 0.012).
A) Illustrates gender differences in local confidence across trial accuracy, showing that females exhibit greater error sensitivity than males, F(3, 315) = 7.18, p = 0.008. The boxplots convey median confidence levels and the spread of scores within each group, with connecting lines denoting individual shifts from correct to incorrect trials. The gradient of connecting lines indicates the degree of confidence change per participant. B) Depicts the interaction of task modality with confidence and gender, with solid lines representing females and dashed for males, revealing systematic variations in confidence across different cognitive tasks. Green and orange hues differentiate between correct and incorrect trials, respectively, highlighting the robust main effects of Domain and Accuracy. Notably, the gender disparity in average confidence is modality-dependent, suggesting task-specific metacognitive biases (F(3, 945) = 15, p < 0.001 for Domain × Gender). Error bars represent 95% confidence intervals.
Each of the presented distributions represents an individual model fit. The y-axis represents the sample count and the x-axis represents the M-ratio. Each distribution was calculated by subtracting the M-ratio estimates of the male group from those of the female group, resulting in a distribution displaying the difference between the groups. The top panel shows the domain-general model without distinguishing between tasks. The four panels below correspond individually to one of the tested domains: memory, vision, GDP and calories. The bars below the distributions indicate the HDI for each distribution. In the total and calories models, the HDI does not contain zero, indicating a significant difference in metacognitive efficiency.