The Autidolfijn project was undertaken in 1991 in Bruges, Belgium, to asses the effect of interactions with captive dolphins on learning in autistic children. The project lasted four years and ended with ambiguous results. Although one group of children did seem to gain learning benefits from working with the dolphins, difficulties arising from the experimental set-up itself seemed to play a role in how well the groups performed. The Autidolfijn project was intended as an experiment, but it was imperfectly controlled and consequently the data could not be used to answer directly the question of the therapeutic effect dolphins have on children with autism. Here the data generated from the project are presented as ethnographic data attached to a particular situation. It is hoped they will help researchers understand the difficulties involved in trying to assess experimentally the positive effects of animals on people (cf. Rowan, 1996; Marino and Lilienfeld 1998; Nathanson 1998).