3 experiments examined serial pattern learning in younger and older adults. Unlike the usual repeating pattern, the sequences alternated between events from a repeating pattern and those determined randomly. The results indicated that no one was able to describe the regularity, but with practice every individual in all 3 age groups (including old old) became faster, more accurate, or both, on pattern trials than on random trials. Although this indicates that adults of all ages are able to learn second-order statistical dependencies in a sequence, age-related deficits were obtained in the magnitude of pattern learning. There were also age differences in what was learned, with only younger people revealing sensitivity to higher order statistical dependencies in the sequence. In addition, whereas younger people revealed evidence of their pattern learning in a subsequent conceptually driven production test, young-old and old-old people did not.