In event-based prospective memory (PM) paradigms, participants are engaged in an ongoing task (e.g. lexical decision) while maintaining an intention to produce a special response if they encounter pre-defined targets (e.g. animal words). This leads to slowed response times even on nontarget trials, which might be caused by: A) a periodic or intermittent process that occurs transiently to check whether the current stimulus is a target, and/or B) a sustained monitoring process maintained throughout task performance rather than being time-locked to stimulus presentation. These processes are hard to distinguish, seeing as the key difference between them occurs in the gap between trials. Processes occurring in these gaps cannot be measured directly by behavioural methods. Here we measured pupil size as a continuous index of intention-related processing in an event-based prospective memory task. Participants performed a lexical decision task while remembering intentions based on either specific target words or categories (e.g. animal words). In two experiments, response times were slowed during PM conditions. Pupil size was significantly increased in the category but not the specificword condition. This effect was sustained throughout task performance rather than occurring transiently when stimuli were presented. Therefore there was no evidence for a transient pupillometric response associated with nontarget checking, although there was a strong transient response when targets were presented in either PM condition. These results provide evidence for a sustained PM monitoring process that occurrs even in the gaps between trials.