Animals hold many executable behavioral programs at once yet express one at a time. This review builds an account of that arbitration in three stages. First, we revisit six classical findings—stickleback territorial contests, white-crowned sparrow song learning, behavioral choice in Drosophila, rodent lordosis genetics, clownfish sex change and cichlid social ascent—and show that each poses a problem its original framing cannot resolve. Reanalysis of the published data from four of these systems supplies quantitative tests: winner-loser effects fit a metaplastic drift of the baseline threshold rather than a change in drive; the song-learning window behaves as a graded threshold rather than a categorical veto, with fifty percent adoption near sixty-five days; behavioral alternation in Drosophila follows from each program consuming the very terms that opened it, with no fatigue or noise term required; and lordosis knockouts separate into categorical and graded failure modes by whether the deleted element is an obligatory transducer. Second, we assemble these results into an explicit model of Dynamic Archetypal Coordination: conjunctive gating with a structural veto, a winner-take-all arbitration layer, margin consumption, and a slow metaplastic baseline. This model requires no gating field and stands on data generated by other laboratories. Third, we identify what it cannot explain—that a single physiological manipulation can scale several unrelated behaviors together—and propose a shared sterol-derived permissive field as the resolution, deriving a dissociation signature that distinguishes it from gain-control and additive accounts. This is a theoretical synthesis: no new experiments are reported and all data are previously published.