Two experiments analyzed the effectiveness of goal statements in aiding recall of self-generated as opposed to experimenter-imposed command names. Subjects were presented with a series of before-after pairs representing the computer states before and after a command was executed. In Experiment 1, during study, one group of subjects generated a command name in response to each pair; a second group generated a goal statement describing the goal to be accomplished in addition to generating a command name. During recall, half of each group was required to recall the name, whereas the other half was required to describe the goal before attempting to recall the name. In Experiment 2, during study, command names (and goals for those subjects in the goal condition) were imposed by the experimenter rather than generated by the subject. Subjects who generated goals and names recalled more command names than did those who generated only names or who received imposed goals and/or names. Generation of an appropriate goal at study improved encoding by helping subjects to select more appropriate command names; generation of an appropriate goal at test improved retrieval for appropriate names only, presumably by activating a relevant subset of names. Even in the relatively simple task of naming and remembering command names, having an appropriate model of the domain through the use of specific goal statements substantially improved performance.