The Step Function: Against the Universality of Graded Forgetting

Hermann Ebbinghaus’s forgetting curve (1885) is one of the most replicated findings in ex- perimental psychology: retention of learned material decays as a power function of time, rapidly at first and then more slowly, approaching but never reaching zero. The curve is treated as a universal law of memory. We argue it is not a law.

It is a property of a specific memory architecture—the dual-store system with graded consolidation—and its universality is an artifact of sampling from a population in which that architecture pre- dominates. We present the theoretical case for a step-function consolidation architecture: a single-store system with a binary admission criterion in which material either consoli- dates with full fidelity or does not consolidate at all. In this architecture, the forgetting curve is not a smooth power function but a step function at the boundary of working memory duration: material that passes the admission criterion is retained indefinitely; material that fails is lost completely within minutes to hours.

The step function is not a pathological variant of the Ebbinghaus curve. The step function is an alternative archi- tecture that the Ebbinghaus paradigm was never designed to detect because the paradigm assumes graded retention and measures graded retention and therefore finds graded reten- tion. We specify the experimental design that would detect step-function consolidation, derive the theoretical conditions under which it arises, and connect the result to the single- store simplicial memory architecture.

Five falsifiers posted. Keywords: forgetting curve, Ebbinghaus, step function, consolidation, single-store, bi- nary admission, memory architecture, schema-dependent encoding, graded retention