Twenty factors identified through a factor analysis of a 102-item biographical inventory were used as predictors in a multiple regression equation to predict on-the-job performance (supervisory ratings) of oil field employees. This yielded a multiple R of .41. A total of 295 subjects participated in the study. Cross-validation yielded a correlation coefficient of .06. The t-test analyses of the factor means of equipment operators and field mechanics proved that two factors could discriminate between the groups, Mechanical Experience (p<.01) and Social Orientation (p<.05). The results of this study indicate that conducting a factor analysis of unvalidated biographical items and attempting to predict performance would be less appropriate than factor analyzing predictive items to gain an understanding of their underlying dimensions.

A replication of an original study of managerial assessment centers performed by Sackett and Dreher (1982) is presented. Their major finding, indicating that assessment centers lack key tenets of internal construct validity, was corroborated in this study of a hotel managers' assessment center. This hotel managers' assessment center is also found to be externally valid using criterion-related validity. The argument is posed that assessment centers, as standardized tests of complex behavioral traits, appear to be operating outside the bounds of normal test construction principles. Five key explanations for this paradox are offered to guide much needed future research in this area. Additionally, a description of commonly utilized assessment center activities is offered the reader.

A biodata based personnel selection measure was created to improve the retention rate of Catalog Telemarketing Representatives at a major U.S. retail company. Five separate empirical biodata scoring methods were compared to examine their usefulness in predicting retention and reducing adverse impact. The Mean Standardized Criterion Method, the Option Criterion Correlation Method, Horizontal Percentage Method, Vertical Percentage Method, and Weighted Application Blank Method using England's (1971) Assigned Weights were employed. The study showed that when using generalizable biodata items, all methods, except the Weighted Application Blank Method, were similar in their ability to discriminate between low and high retention employees and produced similar low adverse impact effects. The Weighted Application Blank Method did not discriminate between the low and high retention employees.

An attempt was made to determine the presence of subjectivity in the appraisal system of a data processing company. Ninety-one clerks were given individual performance ratings by their supervisors, including an overall rating and ratings on seven performance dimensions. A multiple regression performed on these data resulted in a set of empirical weights. Supervisors were also asked to rank the relative importance of each of the seven dimensions to the clerk job. The mean rankings were regarded as apparent weights. A comparison of the empirical and apparent weights led to the conclusion that supervisors were not rating their employees according to what they said was important for successful performance, thus introducing an element of subjectivity into the system.

Empirical findings on Behaviorally Anchored Rating Scales (BARS) have been mixed, despite early researchers' claims that BARS were superior to trait ratings and in reliability and resistance to leniency, central tendency, and halo. The study presented compared a BARS format to an independently derived trait scale as measures of secretarial and clerical performance. Though the BARS showed slightly inflated mean ratings, the instruments showed nearly identical variability. Neither demonstrated sufficient resistance to halo. Thus, despite their intuitive appeal and the rigors involved in format development, it did not appear in this instance that BARS were an efficient and psychometrically superior alternative to the traditional trait rating format.

An "in-basket" test (representative sample of work usually found in the incoming mail basket of a person in a specific desk job) was designed to be used under limitations imposed by Civil Service Statute 1269m concerning merit examinations for the position of police sergeant. This test was used in conjunction with the traditional cognitive skills. test and performance evaluations. Subjects were 20 white male police officers. Peer and supervisory evaluations and predictions of who would make the best sergeant were correlated with total scores on the three-part test. Results indicate that the in-basket test contributes a unique and viable dimension to the traditional merit examination, and aids in the selection of those considered most qualified. Use of the in-basket test under Code 1269m was subsequently approved by the Civil Service Commission.

This study examined the validity of the Minnesota Multiphasic Personality Inventory (MMPI) as a predictor for police officer selection. The MMPI profiles of 212 police officer applicants selected to enter the training academy were compared to the standardized MMPI norms. Significant differences between the police officers and the normative population were found on all but two scales. When the average profile of officers still on the police force was compared with the average profile of terminated officers, two scales were significantly different. Significant correlations were obtained between four MMPI scales and the academy score criterion and two scales each for the commendation and supervisory rating criteria. A prediction equation was developed for academy score using multiple regression analysis.