In two experiments, subjects studied a long series of words and pictures for recognition. Retention intervals varied from several minutes to a few months. The complicated testing procedures in Experiment I required the use of a traditional correction for guessing to obtain estimates of subjects' memory performance. A comparable, but simpler, design in Experiment II permitted the calculation of sensitivity and bias measures. In both studies, pictorial memory was superior to verbal memory at all retention intervals tested, and this advantage was essentially constant over time. In addition, the experiments identified an increasing tendency to call verbal test items "old" over time. Bias scores in Experiment H revealed that subjects adopted a more lenient criterion in responding to words than to pictures, and increased leniency was noted for both item types over time. Explanations of the results are offered in terms of differences in initial encoding and of a loss of discrimination between experimental and extraexperimental materials.