This article concerns the effect of context on people's judgments about sequences of chance outcomes. In Experiment 1, participants judged whether sequences were produced by random, mechanical processes (such as a roulette wheel) or skilled human action (such as basketball shots). Sequences with lower alternation rates were judged more likely to result from human action. However, this effect was highly context-dependent: A moderate alternation rate was judged more likely to indicate a random physical process when encountered among sequences with lower alternation rates than when embedded among sequences with higher alternation rates. Experiment 2 found the same effect for predictions of the next outcome following a streak: A streak of 3 at the end of the sequence was judged less likely to continue by participants who had encountered shorter terminal streaks in previous trials than by those who had encountered longer ones. These contrast effects (a) help to explain variability in the types of sequences that are judged to be random and that elicit the gambler's fallacy, and urge caution about attempts to establish universal parameterizations of these effects; (b) are congruent with theories of sequence judgment that emphasize the importance of people's actual experiences with sequences of different kinds; (c) provide a link between models of sequence judgment and broader accounts of psychophysical/economic judgment; and (d) may offer new insight into individual differences in randomness judgments and sequence predictions.