This paper is concerned with the amount of consistency displayed by subjects in simulated situations which vary in amount of similarity to one another. Odd-even reliabilities measured consistency of performance within a situation, while "alternate form" reliabilities measured consistency between administrative performance in which 232 elementary school principals took four in-basket tests. In three of the in-basket situations the subjects played the role of principal of a simulated elementary school. In the fourth in-basket they acted as Executive Officer of a division of the Bureau of Business, a simulated situation which differed considerably from the school situation. Each half of the in-basket, composed of odd- or even-number items, was scored by a different scorer. For each of the four in-baskets, scores on 34 categories of behavior were selected for use in the present study. The odd-even and "alternate form" reliabilities of these scores were compared, in order to investigate consistency in behavior with in-baskets, between pairs of school in-baskets, and between school and Bureau of Business in-baskets. Conclusions included: 1) No marked differences found in odd-even reliabilities of scores from the school and business in-baskets; 2) odd-even reliabilities were markedly higher than the "alternate form" reliabilities for all in-baskets' 3) "alternate form" reliabilities were generally higher for pairs of a school and a business in-basket; and 4) influence of the background situation (school or business) on consistency of behavior was small compared with the difference between odd-even and "alternate form" reliability. (JGL)