Köhler's chimpanzee studied an out-of-reach banana, then suddenly stacked boxes and got it, with no gradual improvement. Why was this a problem for the learning theory of the day?

C. Because trial-and-error predicts a gradual curve, not a first-try solution

The shape of the curve is the evidence. Gradual error reduction means learning in the behaviour; a flat stretch and then a correct solution means the work happened somewhere unobservable. Like Tolman's rats, it forced an internal representation back into an account that had banned one.