Schrodinger's mouse: human observation erases a drug-induced behavioral phenotype

Abstract

Animal studies primarily rely on human monitoring for health and behavioral outcomes. The presence of a human, however, can induce stress and potentially alter experimental results. We sought to answer whether human observation alone is enough to change experimental outcomes and obscure the effect of interest. Our automated home-cage behavioral phenotyping system provided the ideal platform to test this hypothesis. Two cohorts received semaglutide (0.03 mg/kg, intraperitoneally) on separate consecutive days. The first cohort was recorded without inspection after dosing; the second underwent a brief visual health check with no handling. Recordings from the checked cohort on the preceding day, without injection or inspection, provided a matched baseline. A single brief check produced large, time-locked changes across the behavioral repertoire. The changes began before the cage door was opened, indicating a response to experimenter proximity. In addition to adding variance, the semaglutide-treated animals' behavior during observation was more similar to the no-injection baseline than to the semaglutide/no-check reference, transiently erasing the drug phenotype. Brief human observation is therefore not a neutral measurement operation; over the minutes surrounding it, it is a perturbation in itself that is large enough to mask the drug effect it is meant to record.