Why Do Test Scores Plateau?

Analyses medical-residency standardized-test scores by training year (internal medicine, psychiatry, surgery) and finds gains plateau sharply after ~3 years (Y1-Y2 +81, Y4-Y5 +5) despite continued teaching. Rules out ceiling effects, senioritis, and teaching-stops, and lands on a learning/forgetting-equilibrium model: knowledge as reinforced trees decaying via the forgetting curve, reaching an equilibrium where daily learning equals daily forgetting — modulated by individual memory/curiosity. Ties to the Benezet no-math-until-7th-grade experiment and the education-vs-maturation confound. The memorable '4% of graduating surgeons know less than a first-year' stat.

Quality 61 · Strong. 61 — low Strong. A genuinely interesting, original empirical exploration (a novel data cut) with a nice conceptual payoff, but modest in scope and honestly inconclusive ('I don't have a good answer'). Firm low-Strong.

Claude’s paradigm shift 44 · Moderate. 44 — Moderate. The forgetting-equilibrium framing of expertise plateaus is a mild, useful reframing of standard learning-science.

Real-world impact 2 · Minor. 2 — minor/within-blog. A data-curiosity post; no policy or practice change.