In multiple-regression analysis of picture-naming latencies from an experiment modelled on Oldfield and Wingfield's (1965), with 94 stimuli and 37 adult subjects, two word frequency measures had insignificant beta weights, while two measures estimating age at which the word was learned had highly significant weights. Objects whose names were learned early were named faster. This result may have important implications for the interpretation of studies using word frequency as a critical variable. It is suggested that word retrieval may be a one-stage process that depends upon the age at which a word was learned.

To obtain data for the further evaluation of age‐of‐acquisition as a word attribute in studies of verbal behavior, learning, and memory, estimates were secured from 62 undergraduates (35 males, 27 females) of the age at which they believed they had learned each of 220 picturable nouns (divided into two lists assigned randomly to halves of the sample), according to a 9‐point scale. Reliabilities of these ratings were about .98. For comparative purposes, word frequency values for the words were secured from three large word‐count studies or, where necessary, from subjective estimates made by 20 adults. Use of these and other variables as predictors of previously obtained picture‐naming latencies (Carroll & White, in press) yielded results supporting the previous finding that age‐of‐acquisition is a more relevant predictor than word frequency. Some word frequency indices tend to reflect age‐of‐acquisition, but when this influence is minimized word frequency makes little contribution to the prediction.

The present study was undertaken to assess whether training that was known to produce transfer within the Cognition of Figural Relations (CFR) domain of Guilford's Structure‐of‐Intellect model would also produce transfer to other operations in Guilford's model. Fifty subjects, matched for pretest score on a double classification task, were randomly assigned to either training (with shape, color, shading and addition the logical relations to be learned) or control (no training) groups. Within the training group, subjects matched for pretest score were assigned at random to either of two trainers. Training was given both on CFR and NFR (Convergent Production of Figural Relations). Posttests were administered covering CFR, NFR, DFR (Divergent Production of Figural Relations) and EFR (Evaluation of Figural Relations). As in previous studies, training subjects significantly outperformed control subjects in CFR with both taught and untaught logical relations. The two trainers differed in effectiveness. For the better trainer, transfer was obtained within NFR and to EFR, but not to DFR or to Raven's Coloured Progressive Matrices as a measure of CFR. Transfer effects held up three months later.