We have previously reported that a classical conditioning paradigm involving stimulation of a row of facial vibrissae produced an expansion of the cortical representation of the "trained row", labeled with 2-deoxyglucose (2DG), in layer IV of the barrel field. The present study has examined the pattern of GABA immunoreactivity (GABA-IR) in the cortical representation of row B of the facial vibrissae after (i) 3 days of aversive training, and (ii) 2 months of appetitive training, where stimulation of row B of vibrissae on one side of the snout was used as a conditioned stimulus. The most notable observation was a greater density of GABA-IR cells concentrated in the hollows of the "trained row" B barrels compared to the hollows in the barrel field of the opposite hemisphere in the same mouse. After aversive training, we noted a 2-fold increase in the density of GABA-IR neurons in the hollows of row B; after reward training, the increase amounted to 49%. In contrast, GABA-IR was unchanged in the control groups, which received only stimulation of vibrissae without the unconditioned stimulus. The classification of labeled neurons according to size revealed that the increase in density of GABA-IR neurons was confined to the small (12-15 microm) diameter group. We concluded that the GABAergic system undergoes up-regulation, after both associative learning paradigms, and that the population of small, GABAergic neurons plays an active role in use-dependent plasticity.