A test of the dimensionality assumptions of Rotter's Internal-External scale

Summary: Examined two assumptions about the dimensionality of Rotter's I-E scale: First, the bipolarity of the two statements within each item pair; second, the unidimensionality of the overall construct. A revised I-E scale was constructed in which each of the statements from the original I-E scale was presented separately and the 367 subjects could accept or reject them independently. Correlations between statements within item pairs were low anti negative with only seven exceeding a value of -.20. The responses to the statements were factor analyzed resulting in six interpretable factors when rotated. Of the 23 item pairs, there were 11 pairs which loaded on the same factor in opposite directions while the other 12 loaded on different factors or on the same factor but in the same direction. A subsample consisting of 152 subjects also took the original I-E scale. The correlations between factor scores on the 6 factors and the original I-E score varied from .15 to .52. Both assumptions regarding Rotter's I-E scale were found untenable. Instead, it is suggested that the forced choice format be abandoned and the construct be thought of as a second order factor.