Abstract
<p>Experimental psychological studies consistently report small-to-medium effect sizes, with meta-analyses showing modal effects around Cohen’s d = 0.35 and the majority falling below d = 0.50. While these “smallish” effects are traditionally blamed on methodological deficits—such as low statistical power, publication bias, unreliable measures, and weak manipulations—this paper argues for a theoretically grounded constraint on the maximum expectable effect size. If human experimental participants are identity-preserving systems that naturally resist experimental task demands, these stable, modest effects can be understood as features of a system designed to preserve self-coherence. Drawing on the dynamics of assimilation and accommodation, we propose two axioms: a limited adaptation budget and the constraint that accommodation cannot exceed assimilation without destabilizing identity. Combined with a simple linear averaging assumption, these axioms yield a maximum standardized effect size of d = 0.50 as an upper bound for experimentally induced behavioral changes—on the assumption that ethically permissible manipulations in identity-relevant domains shift the average participant no more than one-half (z-standardized) standard deviation from the control baseline. This theoretically motivated explanation fits remarkably well with empirical meta-analytic evidence, yields falsifiable predictions, highlights the role of shared meaning spaces, and problematizes the conditions under which experimental and control groups can be meaningfully compared. We provide practical guidelines for designing studies that respect the bounded nature of identity-relevant change and for interpreting effect sizes as proportions of a domain-specific maximum.</p>