Attentional control over prepotent responses has previously been shown by manipulating the probability with which stimuli appear. Here, we examined whether prepotent responses to self-associated stimuli can be modulated by their frequency of occurrence. Participants were instructed to associate geometric shapes with the self, their mother, or a stranger before having to judge whether the sequential shape-label pairs matched or mismatched the instruction. The probability of the different shape-label pairs was varied. There was a robust advantage to self-related stimuli in all cases. Reducing the proportion of matched self pairs did not weaken performance with self-related stimuli, whereas reducing the frequency of either matched mother or stranger pairs hurt performance, relative to when the different match trials were equiprobable. In addition, while mother and stranger pairs jointly benefitted when they both occurred frequently, there were benefits only to self pairs when the frequency of self trials increased along with either mother or stranger trials. The results suggest that biases favoring self-related stimuli occur automatically, even when self-related stimuli have a low probability of occurrence, and that expectations to frequent, self-related stimuli operate in a relatively exclusive manner, minimizing biases to high-probability stimuli related to other people. In contrast, biases to high-familiarity stimuli (mother pairs) can be reduced when the items occur infrequently and they do not dominate expectations over other high-frequency stimuli.