Responses to the remaining eight items were subject to an exploratory factor analysis using principal components analysis without rotation, as suggested by Cortina (1993). This analysis resulted in three factors with eigenvalues greater than 1.0, accounting for 63% of the variance. However, the scree plot clearly indicated a one factor solution, with factor 1 having a much higher eigenvalue than the other factors. We decided to follow the recommendation of Floyd and Widaman (1995) who argued that the eigenvalue cutoff of 1.0 often overestimates the number of factors and that the scree plot is a more useful guide. Therefore a second factor analysis was conducted forcing one factor, again with principal components analysis with no rotation. This second analysis resulted in one factor with an eigenvalue of 2.85, accounting for 36% of the variance. We examined the factor loadings for items loading below .4, indicating a weak correlation with the other items (Clark and Watson, 1995). No items loaded below .4 and thus all items were retained (see Table 1). We concluded that this eight-item scale would be our final scale to examine for construct validity.
