A self-based reference point
People's own preferences were linked to the way they worked out what a stranger would choose. In the model that fit the competing explanations best, larger gaps between self and stranger ratings, as well as greater distance between the inferred stranger preference and the available options, were associated with higher accuracy. The findings describe an association within a task, not proof that self-preferences caused better performance.
The question was whether this kind of egocentric anchoring, meaning a tendency to use oneself as a reference point, depends on the social entity being judged. The researchers compared tasks in which an individual or a group occupied different reference frames, asking whether the pattern was tied to the type of social agent or instead reflected a broader feature of absolute social reference-frame calculations.
How the comparison was built
The final analyses included 118 participants. Experiment 1 contributed 73 people after 25 of 98 recruits were excluded; experiment 2 contributed 45 after 11 of 56 were excluded. The second experiment had a preregistered target of 51, so its analyzed sample fell short after exclusions. The first experiment's design and hypotheses were set out in an a priori OSF protocol, and the second was preregistered on AsPredicted under no. 190261.
Participants completed a two-choice transformation task with 240 trials, split into two runs of 120. Experiment 1 put an individual in the relative reference frame and groups in the absolute frame; experiment 2 reversed that arrangement. On each trial, accuracy was based on selecting the option that was numerically closer to the inferred preference.
The anchoring analysis was reported as exploratory because its registered outcome and modelling approach changed. Within that analysis, longer time spent rating others was associated with a larger gap between self and other ratings. Individual targets were rated closer to participants' own preferences than group targets, while the reaction-time-by-entity interaction was not significant.
Accuracy was analyzed with logistic mixed-effects models, a method for right-or-wrong outcomes that can also account for repeated observations from the same participant. The models included absolute distance, self-other rating discrepancy and rating consistency for both reference frames, along with experiment and interactions. A separate preregistered model tested whether absolute rating discrepancy had both a straight-line and curved component.
Distance mattered most
The primary preregistered result was that greater absolute distance was associated with higher accuracy. The model estimate was 0.55, with a standard error of 0.045 and a z score of 12.30, and the association had p<0.001. In ordinary terms, participants tended to perform better when the inferred stranger preference sat farther from the available choices.
Adding other measures produced a more mixed picture. Both absolute and relative self-other discrepancy were positively associated with accuracy. Rating consistency in the absolute frame was negatively associated with accuracy, while consistency in the relative frame was not significant. The association with relative discrepancy was weaker when the relative frame was a group, and the experiment itself had no main effect.
A predicted curved relationship also needs caution. In the dedicated model, the positive quadratic term was significant, with a coefficient of 0.46, a z score of 8.82 and p<0.001. But after the model also accounted for absolute distance and the other predictors, the coefficient fell to 0.004, with a z score of 0.07 and p=0.95. The apparent curve therefore did not survive the fuller adjustment.
An additional comparison favored a model built around the gap between self and stranger ratings. Using AIC, a score for comparing how well competing models fit while accounting for their complexity, the stranger model outperformed models based on the relative frame, the absolute frame, or all discrepancy measures together.
In that best-fitting model, both absolute distance and self-stranger discrepancy were positively associated with accuracy, with estimates of 0.514 and 0.056. Absolute distance was also associated with faster responses, with an estimate of -0.01. Self-stranger discrepancy, by contrast, was not associated with reaction time, with an estimate of -0.002.
What the evidence can support
The ratings themselves also differed by social entity. Self-ratings were the most consistent, followed by group and individual ratings. Groups were judged more dissimilar from the self than individuals were, and neither consistency nor discrepancy showed an experiment-by-entity interaction. The entity effects were statistically significant, while the interaction tests were not.
The main design caveat is that the second experiment fell short of its preregistered target after exclusions, and the two frame assignments were tested in separate experiments. The results therefore show a pattern across these task configurations, but leave open whether the social entity, the frame position or the wider experiment context contributed to the differences.
Because the analysis relates ratings and task performance, the findings do not by themselves establish a causal effect of self-preferences on social judgments. They show how self-related discrepancies, reference-frame measures and task accuracy moved together in the reported experiments.
Pre-processed behavioral data, statistical-analysis and figure-generation scripts, and supplementary material are reported as available through the study's GitHub and Zenodo repositories. The authors declared no competing interests.
Paper data and sources
Original title: The emergence of egocentric biases when comparing others is independent of whether groups or individuals are being compared
Authors: Rodríguez Aramendía M, Kaplan R
Journal/Repository: Royal Society open science
Status: Peer-reviewed
First online: 2026-10-07
DOI: Not available
Original paper · Full text