20 women (M age = 30.4 yr.) were given the “Mental Dice Task” (randomization of the numbers from 1 to 6) once during the preovulatory and once during the premenstrual phases of their menstrual cycles. In addition, for both test sessions a premenstrual-symptom score was assessed reflecting self-rated severity of cognitive, affective, and somatic complaints during the preovular and the premenstrual phases. In comparison to series of real dice throws, the Mental Dice sequences of all the subjects showed a relative lack of repetitions and an excess of counting at both times of testing. Counting bias was significantly enhanced in the sequences generated during the premenstrual testing and the size of this bias was positively correlated with the self-rated severity of premenstrual symptoms. Pronounced counting despite the instruction to randomize was interpreted as reflecting decreased ability to suppress task-irrelevant cues and indicating a relative impairment of frontal lobe functioning. Resistance of the Mental Dice Task towards expectation biases renders a sociopsychological interpretation of these results improbable. We conjecture that, like other cognitive processes, frontal lobe functions also fluctuate with hormonal changes.