Existing models of intertemporal choice normally assume that people are impatient, preferring valuable outcomes sooner rather than later, and that preferences satisfy the formal condition of independence, or separability, which states that the value of a sequence of outcomes equals the sum of the values of its component parts. The authors present empirical results that show both of these assumptions to be false when choices are framed as being explicitly defined sequences of outcomes. Without a proper sequential context, people may discount isolated outcomes in the conventional manner, but when the sequence context is highlighted, they claim to prefer utility levels that improve over time. The observed violations of additive separability follow, at least in part, from a desire to spread good outcomes evenly over time.