Several measures of delay discounting have been shown to be reliable over periods of up to 3 months. In the present study, 115 participants completed a fill-in-the-blank (FITB) delay-discounting task on sets of 5 different commodities, 12 weeks apart. Results showed that discounting rates were not well described by a hyperbolic function but were significantly correlated across the 12 weeks for all 10 commodities. Discounting, when measured by area under the curve, was significantly correlated across the 12 weeks for 9 of the 10 commodities. Absolute values of both measures of discounting sometimes differed across the two administrations of the task. These results support the reliability of the FITB method but raise a number of issues to consider when choosing a method to study delay discounting. (Contains 4 tables, 1 figure, and 1 footnote.)