This paper presents an axiomatic treatment of dynamic choice behavior under uncertainty, emphasizing the timing of uncertainty resolution. The authors introduce a framework where uncertainty is "dated" by the time it resolves, distinguishing between lotteries based on when their uncertainty resolves. This approach allows for choice behavior that cannot be represented by a single cardinal utility function on payoff vectors. The paper provides both descriptive and normative treatments of the problem, showing they are equivalent. It extends the concept of "separable" utility and demonstrates how dynamic choice can be represented by a single cardinal utility function. The authors consider a dynamic choice problem where at each time t, an individual chooses an action d_t, constrained by the state x_t. A random event determines an immediate payoff z_t and the next state x_{t+1}. The probability distribution of (z_t, x_{t+1}) is determined by the action d_t. The standard approach assumes that the individual's choice behavior is representable by a von Neumann-Morgenstern utility function on the vector of payoffs (z_0, z_1, ..., z_T). However, the authors propose a more general approach that accounts for the temporal resolution of uncertainty. The paper introduces the concept of "temporal consistency," which ensures that the individual's preferences at different times are consistent. This leads to a representation theorem that shows how dynamic choice behavior can be represented by a continuous function U and auxiliary functions u_t. The authors also compare their approach to the payoff vector approach, showing that their framework allows for modeling the effects of temporal resolution of uncertainty without overburdening the model with details of primitive preferences. The paper also discusses the implications of assuming that the individual prefers earlier or later resolution of uncertainty. It shows that if the timing of resolution is inconsequential, the payoff vector approach is recovered. The authors also consider the consequences of assuming that the individual's choices at time t are independent of past payoffs, leading to a separable utility representation. The paper concludes with a discussion of the philosophical differences between the descriptive and normative approaches, emphasizing that the temporal aspect of uncertainty is central to their treatment. The authors argue that their approach provides a more accurate representation of dynamic choice behavior under uncertainty by explicitly modeling the timing of uncertainty resolution.This paper presents an axiomatic treatment of dynamic choice behavior under uncertainty, emphasizing the timing of uncertainty resolution. The authors introduce a framework where uncertainty is "dated" by the time it resolves, distinguishing between lotteries based on when their uncertainty resolves. This approach allows for choice behavior that cannot be represented by a single cardinal utility function on payoff vectors. The paper provides both descriptive and normative treatments of the problem, showing they are equivalent. It extends the concept of "separable" utility and demonstrates how dynamic choice can be represented by a single cardinal utility function. The authors consider a dynamic choice problem where at each time t, an individual chooses an action d_t, constrained by the state x_t. A random event determines an immediate payoff z_t and the next state x_{t+1}. The probability distribution of (z_t, x_{t+1}) is determined by the action d_t. The standard approach assumes that the individual's choice behavior is representable by a von Neumann-Morgenstern utility function on the vector of payoffs (z_0, z_1, ..., z_T). However, the authors propose a more general approach that accounts for the temporal resolution of uncertainty. The paper introduces the concept of "temporal consistency," which ensures that the individual's preferences at different times are consistent. This leads to a representation theorem that shows how dynamic choice behavior can be represented by a continuous function U and auxiliary functions u_t. The authors also compare their approach to the payoff vector approach, showing that their framework allows for modeling the effects of temporal resolution of uncertainty without overburdening the model with details of primitive preferences. The paper also discusses the implications of assuming that the individual prefers earlier or later resolution of uncertainty. It shows that if the timing of resolution is inconsequential, the payoff vector approach is recovered. The authors also consider the consequences of assuming that the individual's choices at time t are independent of past payoffs, leading to a separable utility representation. The paper concludes with a discussion of the philosophical differences between the descriptive and normative approaches, emphasizing that the temporal aspect of uncertainty is central to their treatment. The authors argue that their approach provides a more accurate representation of dynamic choice behavior under uncertainty by explicitly modeling the timing of uncertainty resolution.