Participants viewed digit strings and typed them on a computer keyboard. Wh
en they used the same key configuration across training and test, they type
d test strings that adhered to the same sequence rule as training strings f
aster than test strings that adhered to the opposite rule (general-regulari
ty [GR] learning), and they typed test strings that were processed repeated
ly during training faster than test strings that were not (specific-sequenc
e [SS] learning). However, when they used different key configurations at t
raining and at test, GR learning, but not SS learning, was observed. Conver
sely, when they did not type but spoke the strings aloud during training, S
S learning, but not GR learning, was observed. Results suggest that in addi
tion to declarative memory for specific sequences, relatively independent s
ubsystems underlie procedural learning of perceptual-motor sequence compone
nts (producing GR effects) and sequence wholes (producing SS effects).