Over three immediately succeeding trials using different randomized orders of stimuli, 45 young adults, mostly males, named 50 pictures of objects, each as rapidly as possible; latencies were measured for each item. Fifteen of the subjects were tested in a control condition; before each of 5 blocks of 10 items in a trial, subjects in a second group (N = 15) were given information as to the word frequency (WF) level of the items' names, and a third group (N = 15) were given information as to the typical age of acquisition (AOA) level of the names, the pictures and names having been selected such that the WF and AOA levels had a correlation as low as possible. Under all conditions, the previous finding (Carroll & White, 1973a) that AOA was a stronger predictor of latencies than WF was confirmed. Scores on a psychometric test of speed of picture naming were highly correlated (within‐group r = .69) with mean reciprocals of latencies in the experimental picture‐naming situation. “Priming” of responses by prior information on AOA was more facilitative than WF information, but as compared to the control condition such information had a retarding effect. Speed of naming improved over trials in all conditions, but AOA levels of the names, as well as the lags between successive presentations of an item, affected the amount of improvement in a complex manner. It is suggested that retrieval of an item from TLTM (“truly long‐term memory”) places it in an immediate memory buffer.