Presented long lists of concrete and abstract nouns which appeared varying numbers of times to 2 groups of 10 university students. Group 1 judged how many times each word occurred during presentation (on-line) and again after the whole list had been presented (delayed). Group 2 only judged the words after the list had been presented (discrete). All Ss subsequently recalled the words. Frequency estimates were above true frequency for relatively low frequencies, and below for relatively high frequencies in both the delayed and discrete judgments, but more so in the latter. The interaction was more pronounced in abstract than in concrete nouns. Recall was a positive function of concreteness, true frequency, and subjective frequency. Results are discussed in terms of multiple-trace and multiple-process views of how frequency is represented in memory. (19 ref) (PsycINFO Database Record (c) 2006 APA, all rights reserved)