Transverse patterning (TP) task is a relational memory task that depends on configural association, which is composed of a set of discriminations that conflict with each other (A+/B-, B+/C-, C+/A-). Recent studies in humans have suggested that the prefrontal cortex (PF) is important for configural learning. In the present study, to elucidate the functional roles of the PF subregions in solving TP task, neuronal activities were recorded from the medial prefrontal cortex (MPF), dorsal convexity (DC), and principal sulcus (PS), of two monkeys during the performance of a TP task. To control physical stimulation and motor responses, we used a delayed matching-to-sample (DMS) task as a control task. In both tasks, the monkeys chose a target from a pair of visual stimuli and moved the eyes to it, depending on the relation between the two stimuli in the TP, and depending on the sample stimulus in the DMS. Change in activity was analyzed in two task periods: one shortly after the onset of the choice stimuli (CUE period) and one immediately before the saccade onset (PRESAC period). In the TP task, the ratios of the neurons showing significant activity change to the recorded neurons were different among the three PF subregions; highest in the MPF (45% and 52% in CUE and PRESAC periods, respectively), and lowest in the DC (19% and 24%). Effects of the task type (TP/DMS), target location (left/right), and pattern of target stimulus on the neuronal activity were analyzed by 3-way ANOVA. Some neurons exhibited significant change in activity depending on the task type. In the CUE period, the ratio of the task type-dependent neurons was 46%, 13%, and 24% in the MPF, DC, and PS, respectively (p = 0.086, chi-square test). In the PRESAC period, the ratio was 46%, 37%, and 32%, respectively, in the MPF, DC, and PS. The present results suggest that the three PF subregions were involved in the performance of the TP task, and that among the three subregions the MPF may play the most important roles in decision based on configural learning.