Inactivation by GABA is a powerful tool for studying the function of specific cortical regions. It is especially useful in electrophysiology, because inactivation is reversible within short time periods, and because the extent of the inactivated region can be accurately controlled. Iontophoresis of GABA inactivates neurons up to 300 microm around the micropipette. Pressure injection of GABA inactivates neurons further away, but the spatial and temporal characteristics of inactivation by this method have been poorly studied. In order to address this question, we built devices made of micropipettes and microelectrodes glued at various distances. We experienced that repetition of small injections of 100 mM GABA inactivate cortex in a more homogenous way than bolus injections. Diffusion of GABA after pressure injection does not seem to follow a point spread diffusion model as in the case of iontophoresis: GABA probably goes up along the micropipette shaft, and the volume of inactivation has an ellipsoidal form. In order to precisely determine the extent of the inactivated region, we built a mathematical model to fit the experimental data of inactivations obtained above and below the pipette tip. The model provides estimates of the inactivated region for volumes smaller than 60 nl of GABA 100 mM. Limits of inactivation are between 250 and 500 microm lateral to the tip of the pipette. The geometry of inactivation is difficult to predict beyond 60 nl and it seems hazardous to try to inactivate neurons beyond 800 microm with pressure injections of GABA 100 mM.