I've attached a copy of our .csv file in case you're interested in the nuts and bolts, but you've got it right: we want the participants to be able to make a voice response at any time from the moment that the word to be named appears. This could be during the subsequent mask, or after, or, I suppose, during the initial 180 ms presentation of the word itself... although as you noted, that seems highly unlikely. This is our problem, for the research assistants helping us test this have average voice RTs of about 150 ms, which makes no sense. We're replicating another study where the averages were in the much-more-reasonable 450-550 ms range, so I know something is amiss...