The present study demonstrated that emotionally toned words can impact upon young and older adults' estimates of frequency of occurrence. Young and older adults saw positive, negative, and neutral words that varied in number of presentations (one, three, or five times). Young adults were expected to give higher frequency estimates to negative words, while older adults were expected to give higher estimates to positive words. In Experiment 1, both young and older adults gave higher frequency estimates to negative words than neutral words. In Experiment 2, we included a 10 minute filled interval between study and test. Older adults continued giving higher estimates to negative words, but younger adults' estimates no longer differed between the three valences. In Experiment 3, participants performed a short-term memory task in which the distracter words varied in frequency. Under these truly incidental encoding conditions, older adults gave higher frequency estimates to the positive words, while younger adults gave similar estimates across the valences. These results are discussed in relation to the age-related "positivity bias" in memory (Master & Carstensen, 2005).