We investigated the regulation of sequential action using a new paradigm. Participants learned a sequence of seven stimulus categories and then monitored for them during successive displays. All displays were instances of these categories, presented in pseudorandom order. On each trial, participants monitored for an instance of Category 1, pressed a key on a computer keyboard, then monitored for an instance of Category 2, pressed a key on the keyboard, and so on for all seven categories. Thus, a perfect trial contained exactly seven responses. Intrusion errors were classified as a function of ordinal distance from the current serial position (n). Fewer intrusion errors were made at near serial positions than at far ones, suggesting a gradient of lateral inhibition. In addition, more intrusions were made on n + 1 categories than n - 1 categories, suggesting greater availability of intended than completed goals. In accord with current models of sequential action, the results indicate lateral and self-inhibition as important mechanisms in regulation of sequential action.