Recency Effect in the Ensemble Perception of Average Facial Attractiveness

Introduction

This project investigates a fundamental question in ensemble perception: Do all group members contribute equally to the perceived average facial attractiveness, or do recent (later-presented) faces exert greater influence? While ensemble perception enables rapid extraction of a group’s statistical properties, the mechanisms underlying average attractiveness judgments—particularly the role of temporal order—remain unclear. The study addresses this gap by examining the recency effect in the perception of group attractiveness using rapid serial visual presentation (RSVP).

Methods

• Experimental Design:

  • Designed and conducted four behavioral experiments combining RSVP paradigms with both morphed and real facial images
  • Manipulated presentation order (ascending, random, descending attractiveness) and group size (4, 8, 12 faces)
  • Participants rated average group attractiveness, compared groups to probe faces, and completed memory tasks
  • Developed and tested competing models: arithmetic mean, weighted mean, arithmetic-average face, and weighted-average face hypotheses

• Key Analyses:

  • Quantified the influence (“weight”) of each face’s serial position on ensemble judgments
  • Compared judgments across different presentation orders to reveal temporal weighting
  • Used ANOVA, Bayesian statistics, and model fitting to determine the best account of behavioral results

Results & Conclusion

• Recency Effect:

  • Across all experiments, later-presented faces carried significantly greater weight in participants’ judgments of group attractiveness—a robust recency effect.
  • The recency effect was stronger for larger groups and was linked to memory retention, as shown in memory probe tasks.

• Weighted-Processing Model:

  • Results rejected the simple arithmetic mean model, demonstrating instead that ensemble perception of facial attractiveness is best explained by a weighted-mean process, where recent faces disproportionately influence the perceived average.
  • Analysis showed that participants’ ratings aligned more closely with the weighted mean of individual faces than with the attractiveness of an average (morphed) face.

• Implications:

  • The findings reveal that ensemble perception in social contexts is temporally sensitive and shaped by memory constraints.
  • This challenges traditional views of uniform averaging in ensemble perception, offering new perspectives for understanding group impression formation, memory, and decision-making in social perception.

My Contributions to the Project

• Experimental Design: Designed and conducted behavioral experiments to investigate group facial attractiveness perception. Developed protocols to quantify how each face’s serial position influenced perceived group attractiveness
• Statistical Analysis: Performed data analysis using regression models and Bayesian statistics to differentiate between arithmetic mean and weighted mean mechanisms underlying ensemble perception
• Reporting: Authored the research report, emphasizing the discovery that group attractiveness judgments rely on a weighted average process rather than simple averaging, and discussed the broader implications for social impression formation