Individual differences in vision science: The case for a paradigm shift, what it buys, and how to do it well

Nearly a decade ago, Mollon, Bosten, Peterzell, and Webster (2017) asked what individual differences can teach visual science and what counts as good experimental practice. Studies since then have changed both answers. The study of individual differences is a paradigm, not a technique, with its own objects, standards of evidence, and products. Long practiced in the shadows by a small group, it should move into the mainstream and integrate with existing paradigms, treating differences among observers as evidence rather than error. Its central object is the latent factor, which can earn roles from latent mechanism to established mechanism, computational phenotype, heritable phenotype, endophenotype, or factor-targeted marker, each requiring stricter evidence. Twelve pillars organize what the paradigm buys, including links across domains and levels and tests of whether constructs are unitary. Several 2017 conclusions need correction, above all the definition of true individual differences as stable, long-term traits, and the 2017 prescriptions (rank statistics, required test-retest reliability, fixed test order, mandatory blocking, blanket correction for multiple comparisons) each trade one error for another. Because global artifacts produce general factors, tuned factor structure is itself evidence against them. The long-term goal is a computational atlas of human visual variation: its mechanisms, their sources, and optimal markers for the clinic and for design.