Two Korsakoff amnesics (A.G. and G.S.) and two control subjects were t
aught six new concepts. Each concept was composed of three parts: the
name of the concept, the context in which the concept originated and i
ts definition. The learning procedure consisted of two phases: (1) lea
rning the concept names and definitions by means of the vanishing-cues
method; (2) practice on examples of the concepts through a classifica
tion task: examples were either set in the same context as that given
in the original definition or in mixed contexts (same and new contexts
). Subjects were then tested after 24 hours, a week and a month on the
ir ability to identify new examples as belonging to one of the concept
ual rules studied (transfer tests). Both patients showed substantial l
earning. Patient A.G. was slow and dependent of the first letter cues
in the vanishing-cues learning phase but nevertheless, she acquired a
large and flexible conceptual knowledge and this was especially true f
or concepts that were practised by means of mixed-context examples. Pa
tient G.S. easily learned to associate the definitions with the concep
t names but her conceptual knowledge remained more limited. These resu
lts confirm the existence of a semantic learning ability in amnesic pa
tients. They also suggest that under appropriate learning conditions,
amnesics may eventually acquire a new flexible conceptual knowledge.