Kent-Rosanoff stimulus words were separated into 3 categories having different cultural probabilities of producing R-sub(1) (primary), R-sub(2) (secondary) or R-sub(3-n) (sum of R-sub(3) to R-sub(n)). The average cultural probabilities of R-sub(1) for the 3 categories were .14, .32, and .73. A proaction design (Training A, Training B, and Test B) with 60 Ss/category was used. Training consisted of a list of 5 R-sub(1) words followed by a list of 5 corresponding R-sub(2) words. After 120 sec. of vowel cancellation, Ss were tested for recall of the R-sub(2) words in the presence of the 5 stimulus words corresponding to the R-sub(1) and R-sub(2) words of training. Intrusions of R-sub(1) words were found to be strongly related to the cultural probability of R-sub(1). This finding was interpreted as the most solid empirical demonstration we have that language habits can interfere greatly in short-term retention. (PsycINFO Database Record (c) 2006 APA, all rights reserved)