The present experiment investigates the effects of practice at retrieval of examples of categories in semantic memory. Subjects were engaged in one of three kinds of practice; over a series of trials, practice involved recall of: a single category (called specific practice); all different categories (called general practice); or repeated trials for each of three different categories (called blocked practice). One day after practice, subjects performed a transfer task involving a series of trials with a new category. Over the practice trials, recall increased substantially for specific and blocked practice, but only slightly for general practice. On the first transfer trial, a small amount of transfer was found for all three kinds of practice, indicating that all three induced a small but general improvement in the ability to recall any category. From the first through the fifth transfer trial, recall increased by an amount that was comparable to the first five trials of specific practice on the first day but in a manner that was the same for these three forms of practice. Apparently the kind of practice on the first day did not facilitate the increases in recall from specific practice on the second day. Overall, the results indicate that the greatest gains in the amount retrieved from a category in semantic memory come from specific practice.