‘Statistical significance’ is more than just a label. Cognitive psychological theories suggest it may represent a mental concept of a category of p-values due to the pervasive practice of dichotomous interpretation of p-values. This paper builds on previous research identifying categorical boundary effects in the initial information processing of p-values by examining the encoding and retrieval of p-values embedded in the context of scientific abstracts. A sample of 30 U.S. graduate students in the psychological sciences read blocks of abstracts, then were prompted to recall certain details, including p-values. Results show that memory for p-values was skewed away from the .05 boundary, suggesting that training in dichotomous ‘p < .05’ thinking may lead to categorical biases in memory for p-values. These results set up experiments to test mechanistic hypotheses of boundary effects on statistical cognition as well as the efficacy of teaching interventions to address and ameliorate these categorical biases.