Little is known about the accuracy of temporal codes for memories, and still less about the nature of the codes. This report addresses the central question of the mechanisms by which order information is attached to memories. The results of 16 experiments indicate the role of some independent variables on temporal coding in relatively short-term memory and in long-term memory. Several experiments in which changes in proactive inhibition are used as an index of temporal differentiation show that the nature of the words making up the lists is involved fundamentally in temporal coding. Other experiments demonstrate that in relatively short-term memory subjects cannot learn to improve their performance in estimating how far apart in time two events occurred. Still other experiments show that recency judgments for two events improve with practice, but the improvement is minimally influenced by the temporal separation of the events. The context in which memories are established is shown to influence temporal codes only if an ordering metric is a part of the context. Several theoretical propositions are advanced to account for the findings. (Author)