In Experiment 1, adults estimated the frequency of typical and atypical actions presented in stories about scripted routines. Judgments of frequency were more accurate for atypical than for typical actions. In Experiments 2 and 3, children and adults estimated the frequency of atypical actions that were presented in lists, or embedded in stories that described activities that are scripted, or in stories that described unfamiliar activities. The results supported Hasher and Zacks's (1984) proposal that encoding of frequency information is automatic and invariant across a wide range of ages. However, both children's and adults' frequency estimates were influenced by manipulations that varied the difficulty of retrieving the representation of each occurrence of an event. The studies also provide novel support for the Script-Pointer-Plus-Tag theory of script memory.