Bias in postdiction from projective tests.

Abstract

IN A preceding investigation (1, 2) an effort was made to study the rater bias or set found in personality appraisals based on projective tests and other types of personal documents by comparing such appraisals with others based on more extensive information and observation. In this earlier study the criterion consisted of a series of trait ratings arrived at by a three-man team after five days of study of the individual through objective and projective tests, interviews, various personal documents, and observations of behavior in standard situations. However, inasmuch as the criterion ratings consisted of a series of judgments of traits, i.e., of certain abstractions currently familiar in personality studies, the validity of the appraisals which were investigated was pointed out to be a "relative validity" contingent upon the validity of the judgments of the team making the criterion ratings. Since the consensus of "experts" may or may not correlate satisfactorily with an external criterion of validity, an effort was made to conduct an analogous investigation in which the appraisals could be evaluated against a more rigorous criterion, viz., a criterion test dealing more directly with specific behaviors of the subject which had occurred in the past rather than with the appraisal of so-called traits. The present report deals with this effort. METHODCriterion test. As the criterion for this study a postdiction test was constructed consisting of an array of multiple-choice items which were built primarily around incidents known to have happened in tie recent past history of a single subject (5), hereafter called "Linda." Although the majority of items concerned specific situational behaviors, some items were developed from commonly shared characterizations of 5

The Scientific Status of Projective Techniques

“…Soskin (1954) found that the addition of TAT protocols did not add significantly to the validity of clinicians' personality ratings of normal participants above and beyond basic demographic information. In contrast, Golden (1964) found that clinicians' judgments concerning the personality traits of participants (both psychiatric and nonpsychiatric patients) increased significantly when the TAT was added to demographic information.…”

Clinical Intuition and Inferential Accuracy

“…evidence should be aimed at the assumed mode if the weighting of wrong predictions is comparable to that for correct ones If one is himself a fairly well-adjusted person, modal in response, then he is more likely to be a good predictor because, aiming at another person, he can hit the average of other's responses We would infer that good predictors will be relatively normal, and persons who could be best predicted also would be normal or modal, however we may refer to it One does not need to assume any empathetic process to expect this finding on uncorrected data Consider, for example, the items of the experiment described above Here, a person who is acting as target may choose an easy pattern or a hard pattern If he chooses an easy set of items, he would choose the modal item according to what the group tended to assume likely to be chosen by any such person standing in front of them If he chooses a hard pattern, he would choose low frequency responses as such groups marked them, and it is to be expected that the prediction of his responses will be poor DISCUSSION Despite the use of a psychometric device that should be sensitive for the purpose, these experiments yielded little evidence for the general existence of intuition at a clinically useful level of accuracy The level of contact between percipient and target was purposely kept relatively low, but the situation and implied subject matter, which seem related to motives and feelings, do not appear to be greatly different in form from common situations that lead to chnical prediction statements As was expected, positive evidence did develop that the predictions were not random and were different from zero in accuracy when analyzed for individual targets Some of these accuracy values were positive and some were negative The consistent accuracy or error was observed in both individual percipients and in the groups The classificatory generahzations from knowledge that the target person was a college student of given sex were often more accurate than those complicated by an opportunity to watch each target This finding comes from the fact that the averaged item predictions of a group of outside targets (gamma scores) were often better than the predictions of groups or persons who were watching the given target (alpha scores) Predictions as made from gamma data could apply to the given target only in being derived from like sex sources and from knowledge of the stu-dent status of the target It would seem that the consistent error of both individual percipients and of groups somehow derives from having too much information-at the least too much information disturbs the predictive process The trouble may be that seeing a target person can spoil the predictions through encouraging false dassifications Gage (9), Soskin (25), and others found and recognized a comparable phenomenon Of course, the argument would be absurd if the additional information from the target were directly informative about the items to be predicted This might be the case m using therapeutic interviews as material The extra information from seemg the target, as in these experiments, is not dearly informative It probably provides behavioral cues that are not simply and popularly translatable to suggest the predicted item No difficulty might be found in predicting "shy" on a Q-sort for a dient who had told of incidents where he had acted shyly, but it might be quite another thing to predict this from his behavior as he held a burning match as long as he could This was illustrated m the present data in the sporadic tendency of percipients to assume that a given target would select items that were more like the average of the opposite sex Some targets did this, especially some males chose rather female type patterns The difficulty was that the percipients could not accurately tell who would do this The effect was to increase error variance Confidence and S3anpathy felt by the percipient were not major variables in these errors…”

“…dent status of the target It would seem that the consistent error of both individual percipients and of groups somehow derives from having too much information-at the least too much information disturbs the predictive process The trouble may be that seeing a target person can spoil the predictions through encouraging false dassifications Gage (9), Soskin (25), and others found and recognized a comparable phenomenon Of course, the argument would be absurd if the additional information from the target were directly informative about the items to be predicted This might be the case m using therapeutic interviews as material The extra information from seemg the target, as in these experiments, is not dearly informative It probably provides behavioral cues that are not simply and popularly translatable to suggest the predicted item No difficulty might be found in predicting "shy" on a Q-sort for a dient who had told of incidents where he had acted shyly, but it might be quite another thing to predict this from his behavior as he held a burning match as long as he could This was illustrated m the present data in the sporadic tendency of percipients to assume that a given target would select items that were more like the average of the opposite sex Some targets did this, especially some males chose rather female type patterns The difficulty was that the percipients could not accurately tell who would do this The effect was to increase error variance Confidence and S3anpathy felt by the percipient were not major variables in these errors…”

To measure attributed mental illness1

“…The work of Soskin (1954) provided the basic conception for the Mental Illness Behavior Prediction Scale (MIBPS). Soskin assessed the rater bias tbat was generated through providing tbe rater with the projective test responses of the target person.…”