P300 and probability: comparison of oddball and single-stimulus paradigms