Eleven groups run under classical, instrumental, pseudoconditioning, CS-only, US-only, and time-lapse procedures, with the ITI illuminated and a color change CS, showed that true classically trained animals do not increase CAR performance with training, whereas instru- mentally trained goldfish do. This is consistent with a phototaxic interpretation suggested in earlier work. Additionally, the finding that CS-only and time-lapse controls show high “acquisi- tion” rates, whereas pseudoconditioning controls do not, not only questions the use of the pseudoconditioning procedure as the sole learning control in this situation, but also questions a learning interpretation itself. Conclusions of what and how, or even if, goldfish learn in this apparatus seem premature.