Code and data: separable encoding and execution layers carry a vector between honeybee brains.

Reanalysis of two publicly deposited honeybee datasets.1. Deprivation experiment** (Dong, Lin, Nieh & Tan 2023, Science 379:1015-1018; data doi:10.5281/zenodo.7301648). Competing accounts of the deprived-minus-tutored gap (no recovery, maturation delay, two layers, one shared rate, two free rates) are fitted jointly to log waggle-run duration, divergence angle and disorder by maximum likelihood (bivariate normal per bee and measure) and compared by AICc, with a colony-paired bootstrap, model-recovery simulations, sensitivity analyses, a gain-versus-offset test, flight-time and motivation checks, colony intraclass correlations, a persistence bound and an external calibration benchmark.2. Audience experiment (Lin et al. 2026, PNAS 123:e2518687123; data doi:10.5281/zenodo.10892587). Effects of audience size on run duration and within-dance variability, and the predicted effect of the deprivation experiment's follower deficit.