Prospective expectations for retirement and retrospective accounts of past work experiences are typically used in cross-sectional empirical studies of retirement. Unfortunately, little is known about the accuracy of such accounts. Therefore, in the present longitudinal study, we built on the work of Beehr and Nielson (1995) by examining the consistency of these accounts over time using data from 672 military retirees over a four- to five-year period. Mixed evidence for the consistency of these attitudes across time was found. Specifically, our results are much weaker than Beehr and Nielson's results, thus suggesting that longer time intervals between measurements may erode the accuracy of prospective estimates and retrospective accounts. Results are discussed in terms of the uniqueness of the military retirement and how civilian retirement and military retirement are becoming similar. (Contains 3 tables and 1 figure.)