While few would disagree that token economies are effective in bringing certain positive and negative target behaviors under contingent control, the recent past has witnessed increased concern over choice of targets as well as back-ups, voluntary versus involuntary client participation, as well as the effectiveness of such programs in meeting client versus staff needs. Within the latter category, reviewers of the token economy literature (e.g., Kazdin & Bootzin, 1972) have suggested that behaviors instigated within token economy mechanisms rarely generalize to alternative conditions, and this issue of generalization is perhaps one of the major criticisms of the token format. The present paper attempts to define and examine generalization in the light of recent research. Within the presentation, the author integrates certain of his own findings and personal speculations as the various methods and issues for examining generalization are presented. (Author)