The paper reports a Token Economy project involving 18 long stay male patients (diagnosis chronic schozophrenia), average length of hospitalisation twenty-five years. The project comprised a one month baseline period, a six month Token Economy phase, a two month experimental phase and a second three months Token Economy. Conditions were the same for all patients except during the experimental phase where they were randomly assigned to one of three experimental groups to investigate the relative importance of social reinforcement and other variables involved in token systems. The groups were (1) a social reinforcement only group; (2) a social reinforcement and non-contingent tokens; (3) a social reinforcement and contingent, non-spendable tokens. Each groups' performance was compared both with the other groups and with their own performance over all phases of the study. Assessment measures comprised standardised psychiatric rating scales - the Nurses Observation Scale for In-patient Evaluation (Honigfeld and Klett, 1965; Honigfeld et al 1966), and the Psychotic Reaction Profile (Lorr et al, 1960); other rating checklists, and time-sampling data. Analysis of Variance was carried out on all data over the four phases of the programme. The total token economy package as represented by the first token economy period was shown to be effective in promoting improvements in most areas of patients' functioning as compared to baseline. Results during the experimental phase suggest that contrary to our initial hypothesis social factors involved in exchanging tokens are not demonstrably important sources of reinforcement in Token Economies. There were no inter-group differences during the experimental phase, suggesting that none of the variables studied were critical factors. After the experimental phase the return to the complete token package produced an unclear picture where there were significant improvements on some scales, but significant declines on others.