This paper contests what has remained a core assumption in social psychological and general understandings of the Milgram experiments. Analysing the learner/victim’s rhetoric in experimental sessions across five conditions (N = 170), it demonstrates that what participants were exposed to was not the black-and-white scenario of being pushed towards continuation by the experimental authority and pulled towards discontinuation by the learner/victim. Instead, the traditionally posited explicit collision of ‘forces’ or ‘identities’ was at all points of the experiments undermined by an implicit collusion between them: rendering the learner/victim a divided and contradictory subject, and the experimental process a constantly shifting and paradoxical experiential-moral field. As a result, the paper concludes that evaluating the participants’ conduct requires an understanding of the experiments where morality and non-destructive agency were not simple givens to be applied to a transparent case, but had to be re-created anew – in the face not just of their explicit denial by the experimenter but also of their implicit denial by the victim.