This paper provides a framework for understanding preferences over utility streams across different time periods. We analyze preferences for the close future, for the distant future, and a synthesis of both, establishing a representation involving weights over time periods. Examining scenarios where two utility streams cannot be robustly compared to each other, we introduce notions in which one has more “potential” to be preferred over another, which lead to MaxMin, MaxMax, and -MaxMin representations. Finally, we consider temporal bias in the form of violations of stationarity. For close future preferences, we obtain a generalization of quasi-hyperbolic discounting. For distant future preferences, we obtain Banach limits and discuss the relationship with exponential discounting.