The effects of attention during encoding and rehearsal after initial encoding on frequency estimates were investigated in three experiments. Varying the level of processing affected the linear increase in frequency estimates as a function of actual frequency, but varying processing after encoding with remember or forget cues had the greatest effects on the intercept of the function relating judged to actual frequency. Deeper levels of processing improved performance in a frequency discrimination task, whereas remember and forget cues had only very small effects on performance. Materials that are easy to rehearse were compared with materials that are difficult to rehearse in Experiment 2. The results were interpreted as evidence against a covert rehearsal explanation of slope effects in frequency estimation tasks because materials that are difficult to rehearse tended to produce larger interactions between remember versus forget cues and frequency than materials that are easier to rehearse. In Experiment 3, an arithmetic task that was performed during word encoding affected the slope of the function relating judged to actual frequency, but the same task performed immediately after word presentation had no effect on frequency estimates. It was concluded that frequency is not stored automatically because attention during the initial stages of encoding affects it; however, attention devoted to processing after initial encoding does not affect the rate with which subjective frequency increases with repetitions.