Author reports a study in maze learning in which factors of recency and frequency were exactly balanced against themselves as far as positive and negative effects were concerned. Since the subject was sent back to the starting point at the commission of each error, the probability of a correct choice became increasingly less at each successive blind in the maze, and the impressions resulting from the frequency and recency of wrong choices militated, therefore, against the learning of the maze. From the fact that in learning the maze the subjects' errors were eliminated in the forward direction, whereas in other forms of maze learning where factors of recency and frequency are not balanced against themselves elimination of errors is in the opposite direction, author argues that, contrary to those doctrines coming down from early associationistic psychology, the factors of recency and frequency in the modified form of mental maze play no rôle in the act of learning. Learning must be attributed, rather, to a cumulative process in the afferent nerve impulses which, being reflected up through cortical synapses out to efferent nerve paths, exercises a directive influence on the nerve impulses which subsequent stimuli initiate. From Psych Bulletin 19:12:00926. (PsycINFO Database Record (c) 2006 APA, all rights reserved)