Practising simple visual tasks leads to a dramatic improvement in perf
orming them. This learning is specific to the stimuli used for trainin
g. We show here that the degree of specificity depends on the difficul
ty of the training conditions. We find that the pattern of specificiti
es maps onto the pattern of receptive held selectivities along the vis
ual pathway. With easy conditions, learning generalizes across orienta
tion and retinal position, matching the spatial generalization of high
er visual areas. As task difficulty increases, learning becomes more s
pecific with respect to both orientation and position, matching the fi
ne spatial retinotopy exhibited by lower areas. Consequently, we enjoy
the benefits of learning generalization when possible, and of fine gr
ain but specific training when necessary. The dynamics of learning sho
w a corresponding feature. Improvement begins with easy cases (when th
e subject is allowed long processing times) and only subsequently proc
eeds to harder cases. This learning cascade implies that easy conditio
ns guide the learning of hard ones. Taken together, the specificity an
d dynamics suggest that learning proceeds as a countercurrent along th
e cortical hierarchy. Improvement begins at higher generalizing levels
, which, in turn, direct harder-condition learning to the subdomain of
their lower-level inputs. As predicted by this reverse hierarchy mode
l, learning can be effective using only difficult trials, but on condi
tion that learning onset has previously been enabled, A single prolong
ed presentation suffices to initiate learning. We call this single-enc
ounter enabling effect 'eureka'.