“Far transfer is rare” is among the most robust findings in educational psychology. But rare forwhom? This paper identifies an unstated boundary condition in the learning transfer literature:virtually all canonical transfer studies draw participants exclusively from neurotypical populations.Cognitive architecture is neither reported as a variable nor controlled as a moderator. We identifythis as Boundary Condition Dropout (BCD)—the systematic loss of validity conditions whenfindings are generalized beyond their original scope—and argue that it has produced a theoreticaldistortion that constrains the field’s central constructs. Evidence from four literatures converges on asingle conclusion: learning transfer is a property of cognitive architectures, not of domain pairs.Autistic individuals achieve equivalent abstract reasoning through fundamentally differentprocessing routes. Gifted learners spontaneously abstract structural features without explicitscaffolding. Twice-exceptional learners provide within-person natural experiments showing transferasymmetries that track cognitive subsystem integrity rather than domain distance. Even withinneurotypical populations, every major predictor of transfer success—fluid intelligence, workingmemory, metacognition, cognitive flexibility—is an architectural variable. The near/far distinction,the field’s central organizational construct, is cognition-dependent: “distance” between domains iscomputed by the cognitive architecture encountering them, and different architectures computedifferent distances. We propose that the near/far distinction be parameterized by architecture, thatcausal claims in transfer research specify their architectural preconditions, and that cognitive profilebe treated as an independent variable. Five testable predictions are derived that diverge from thecurrent consensus. The “transfer problem” may be partly an artifact of studying one cognitivearchitecture and generalizing to all.
Philos Sophia Franny (Tue,) studied this question.