This study reports a comparison of the whole and part methods of serial learning. There were 4 lists of 10 CVC trigrams each representing all possible combinations of 2 levels of meaningfulnes s and 2 degrees of intralist similarity. The lists were divided into halves for part learning. Acquisition of the individual parts was followed by a combination stage in which the entire list was learned. Learning times varied inversely with the level of meaningfulness and directly with the degree of intralist similarity. This relationship held true for whole learning and throughout part learning. Much of the gain derived from the reduction of list length during the acquisition of the parts was offset by the duration of the combination stage. However, total learning times showed a relatively small but consistent advantage of the part method over the whole method. An analysis of the conditions of whole and part learning by McGeoch and Irion (1952, pp. 500 f.) points to three sources of differences between the two methods: (a) Since the difficulty of a list increases as a function of length, the sum of the times spent in the acquisition of the individual parts will be less than the time required for mastery of the intact list. (b) The time used in combining the parts into a whole will reduce the advantage which the part method derives from the length-difficulty relationship, (c) The acquisition and retention of successive parts may be influenced systematically by transfer and interference effects. The amount and direction of such effects may be expected to vary with the degree of similarity between the successive