Several different issues in the temporal coding of words were subjected to experimental analysis. Two experiments evaluated three response measures (recency judgments, position judgments, lag judgments) used to index temporal coding. Lag judgments were found to be of little use; subjects could make valid position and recency judgments without being able to make valid lag judgments. Practicing lag judgments produced heavy positive transfer to the other two measures. Experiment 3 showed that correct recency judgments were a direct linear function of rate of presentation through three seconds. Experiment 4 demonstrated that recency learning and two-category classification learning were substantially correlated, but a direct test (Experiment 5) indicated that the two-category classification task cannot serve as a paradigm for recency learning. Experiment 5 also showed that word frequency had no influence on either recency learning or two-category classification learning. Experiment 6 suggested that a recency principle may govern knowledge of temporal order for very short intervals of time. (Author)