Center for Management and Organizational ResearchUniversity of South CarolinaA role choice model, which included attraction, expectancy, and intention in-dexes for both civilian and military roles, was used to analyze the recruittraining turnover behavior of 1,521 male Marine Corps recruits. Demographic,expected leadership, and expected job content were also measured at thebeginning of recruit training. It was found that on the way into recruit training,subsequent graduates and dropouts differed significantly on 20 of 29 variables,including intention to complete their enlistments, expectancy of completingtheir enlistments, attraction to the Marine role, and a number of other expectedorganizational and demographic variables. When the variables were subjectedto stepwise multiple regression, a multiple R of .30 was observed for 11-weekrecruit training attrition, with expectancy of completing, education, Marinerole outcome expectancies, expectancy of finding an acceptable civilian role,and intention to complete being the first five variables to enter the equation.The results support the usefulness of moving beyond demographic predictionof attrition, including perception and evaluation of alternative roles, and ex-ploring more closely the organizational entry process.Recent reviewers of the literature on em-ployee turnover (Forrest, Cummings, & John-son, 1977; Locke, 1976; Mobley, Griffeth,Hand, & Meglino, in press; Porter & Steers,1973; Price, 1977) have observed the needfor better developed conceptual models of theturnover process. Macro-level studies of therelationships between general economic indi-cators, such as level of unemployment, andaggregate turnover rates add little to thepsychology of individual withdrawal behavior.At the micro level, the frequently replicatedcorrelations between job dissatisfaction andturnover and between turnover and demo-graphic and other univariate correlates donot account for sizeable amounts of varianceand provide a conceptually incomplete basis