We compared the effect of locking the event-related potentials (ERPs) on the stimulus vs. the response in a simple (SRT) and in a discriminative (DRT, equally probable Go/No-go) visuo-motor task. The accurate alignment of stimulus- and response-locked ERPs on the same time scale was obtained selecting a group of 27 participants with low inter-individual response time (RT) variation in the two tasks to reduce the jitter of RTs. Two-second epochs were defined for the analyses based on averaged RTs in the two tasks. Results show that the preparatory pre-stimulus activities (i.e., the pN and the BP components) were not affected by the different locking in both SRT and DRT. As expected, the exogenous post-stimulus P1 and N1 components (and the P2, present only in SRT) were larger in stimulus- than response-locked averaging; the same trend was observed for the less-known prefrontal N1 and P1 components (pN1 and pP1) that were larger in stimulus-locked averaging (the effect was not significant in SRT). The prefrontal pP2 component was only present in DRT peaking around 370 ms, and did not show an effect of locking. The frontal-central N2 component was enhanced by response-locked averaging in SRT, while it did not show effects of locking in DRT. The P3 component peaked at about 350 ms over central sites in SRT, and at about 500 ms over parietal sites in DRT; in both cases its amplitude was larger in response- than in stimulus-locked averaging. Overall, the amplitude of the preparatory components was independent from locking, the exogenous components were enhanced by stimulus locking, and the late components were more related to the response than to the stimulus. Concluding, to investigate action preparation, perception, and perceptual-decisional activity ERP studies should adopt stimulus-locked averaging with an appropriate baseline and longer pre-stimulus interval, or use both types of locking.