This study investigates how the artificial grammars of timbres are lea
rned. In Experiment 1, the participants listened to sequences of timbr
es produced using an artificial grammar. They were then asked to diffe
rentiate between sequences which either did or did not violate the gra
mmar. The participants in the explicit condition were informed of the
existence of rules underlying the sequences. Those in the implicit con
dition received no such information (Reber, 1967). Experiment 2 addres
sed the influence of the learning mode on the content of the knowledge
acquired. At the end of the learning phase, the participants were ask
ed to judge the grammaticality of new sequences of timbres (test condi
tion) or letters (transfer condition). The results confirmed the advan
tage of the implicit condition over the explicit condition but suggest
that the knowledge acquired pertained more to surface regularities th
an to abstract rules. The results are discussed within the framework o
f current work on implicit learning and musical cognition.