<p>Behavioral measures of impulsivity and other traits often show weaker test–retest reliability than self-report measures. Weaker reliability impacts the assessment of individual differences in the trait or state being assessed. Behavioral tasks demonstrate greater sensitivity to state variables which may be a key reason for changes in ranked performance across time. The present study examines a single impulsivity task, the escalating interest task, and considers the design principles that may alter the within-session stability of the assessed behavior. A reanalysis of existing data is contrasted with new behavioral data to reveal that rapid changes in task contingencies produced more stable individual differences than prolonged exposure to each contingency. This outcome may be driven by expanding the number of contingencies experienced at each assessment or by keeping behavior in transition. An attempt to avoid floor or ceiling effects by increasing the ambiguity of the contingency, however, did not produce the desired result. The implications of these results for the escalating interest task as well as other behavioral tasks are considered.</p>