Performing a nonfocal prospective memory (PM) task results in a cost to ongoing task processing, but the precise nature of the monitoring processes involved remains unclear. We investigated whether target context specification (i.e., explicitly associating the PM target with a subset of ongoing stimuli) can trigger trial-by-trial changes in task interference according to stimulus relevance for the nonfocal PM task. Participants performed a lexical decision task in which a PM task (press "F6" when a target syllable appeared) was embedded. The target syllable always occurred in word trials, but we manipulated participants' expectations regarding the target context by instructing them that targets would occur in words only (specific condition) or in both words and nonwords (nonspecific condition). A control condition with no PM demands was also included. Although having a PM task led to noticeable slowing on the ongoing task, specifying the PM target context reduced cost to items irrelevant to the intention (nonwords) while leaving PM performance intact. Moreover, higher cost for nonwords in the nonspecific than specific condition was persistent across the ongoing task even though the target syllable was repeatedly presented in words. These results suggest that stimulus processing can be modulated according to participants' expectations about the lexical properties of the target, with trial-by-trial changes in task interference as a function of stimulus relevance to a nonfocal intention observed as a consequence.