A CAUTION AGAINST USE OF THE SHOSTROM PERSONAL ORIENTATION INVENTORY
J. J. RAY
Sociology Department, University of New South Wales, P.O. Box 1, Kensington, Sydney, N.S.W 2033, Australia.
(Received 3 February 1984)
Since the Shostrom (1964) Personal Orientation Inventory (POI) is one of the more widely used omnibus personality inventories in clinical practice, one would expect it to have generally good psychometric characteristics and validity. Unfortunately, there is almost no evidence that this is so. What evidence there is tends in fact to suggest the opposite. Silverstein and Fisher (1973) appear to be the only authors who have reported internal reliabilities (alpha) for each of the POI scales and the levels they found for this coefficient were surprising for how low they were rather than for how high they were. The reliabilities of 0.27, 0.30 etc. that they found for the various scales represent the most unambiguous testimony to an almost total lack of internal consistency in the scales concerned. Shaw and Wnght (1967) regard reliabilities of more than 0.90 as necessary in an instrument used to assess individuals.
It might however be argued that the problem lies with the Silverstein and Fisher data rather than with the POI. The client group used by Silverstein and Fisher was a sample of prisoners and it seems possible that prisoners might tend to give quite deviant and atypical responses. There is clearly a need for any further evidence on the question that can be obtained.
It seemed therefore useful to look again at a body of data originally gathered by Sherell de Florance (1972) and made available by her to the present author in connection with his studies into the relationship between authoritarianism and behaviour in the Milgram experiment (Ray, 1976). This data included one of the POI scales -- the Acceptance of Aggression scale. The Ss were 47 students at the University of Sydney taking part in a Psychology Department experiment. As such, they represent a fairly conventional sample in terms of what is reported in published psychological research and they do at least offer a strong contrast to the sample used by Silverstein and Fisher (1973).
The internal reliability of the Acceptance of Aggression scale was found to be only 0.55. This accords remarkably well with the result reported by Silverstein and Fisher (i.e. 0.54) and hence does suggest considerable generalizability across sample type for the Silverstein and Fisher results. The indications are, then, that further use of the POI for clinical assessment purposes would be extremely incautious.
What are we to put in place of the POI, then? There is, of course, the POD (Shostrom, 1977) but the failure by Shostrom to release the scoring key to this inventory renders psychometric examination of it impossible. Given the experience with the POI, the POD must be regarded as suspect until proven otherwise. One approach may be to abandon the multiscale battery approach entirely and simply seek out individual scales that measure precisely what one has in mind in a given situation. The Acceptance of Aggression scale used above, for instance, could have been replaced by the more reliable Ray (1972) scale of the same name.