Even though response latency can vary greatly among subjects and from trial to trial for the same subject, latency as a measure of associate strength might be useful in connection with response frequency to measure degree of learning. The effects on latency and learning of two task variables--training procedure (anticipation vs. recall) and information transmission (amount of information required for errorless responding)--were investigated using a factorial design. Several aspects of latency as a possible basis for instructional decisions were studied: the "TLE," or trial of last error, i.e., the trial immediately preceding the correct criterion response of N successive errorless trials; the correctness of response; item difficulty, individual differences, and the TLE; maximum response latency on the TLE; and errors made subsequent to the first correct response. A series of eight stimuli that transmitted different amounts of information were presented to the 96 subjects on the display screen of an on-line computer. Subjects responded via a keyboard, which also gave them feedback. The main conclusion from many interesting results was that latency is not a sensitive measure of associative strength during the pre-TLE period, but it may well be an accurate measure during overlearning, after the TLE. (LH)