Two experiments were designed to study retrieval from a well-ordered set of overlearned information in long-term autobiographical memory. In Experiment 1, each of 161 university students was asked to recall the name of one teacher from each of the 12 preuniversity school years. Recall was cued in forward (Grades 1-12), backward (Grades 12-1), or random order. Backward-ordered search proved most efficient: Fewer students failed to complete the task :in that search order, and those who did complete the task were faster than the successful students in the other search orders. Experiment 2 was similar to Experiment 1, except that the 148 students tested were asked to “think aloud” as they attempted recall. From the resulting protocols, 25 categories of verbal reports were identified, including four that reflected retrieval strategies. Omission errors (failures to recall teachers’ names) and, with one exception, frequencies of verbal reports decreased as recency increased. Taken together, the results of both experiments show that the recall of one item can indirectly aid the recall of contiguous items and that the probability of recalling an item from autobiographical memory is primarily a function of recency.