We cannot say for sure why the correlations for the U.S. sample were better than those for the Spanish sample. Some possible interpretations, based on what we know of the data, are (a) differential quality of supervisory ratings, (b) differential compositions of the two respective samples, and (c) relatively greater homogeneity of the Spanish sample relative to the U.S. sample. In particular, Spanish supervisors seemed less comfortable providing ratings of quality of performance than did U.S. supervisors, but because we did not anticipate this result, we collected no quantitative data with respect to it.

Self-ratings. The distributions of self-ratings proved not to be amenable to the validity analyses due to restriction of range in these variables. As has been found in other studies (e.g., Sternberg et al., 1981), participants tend to have almost a uniformly high opinion of themselves. With the item on planning career change deleted (because this item does not measure self-evaluation of competence), mean self-ratings on the 9-point scale ranged from 6.5 to 8.3, with an overall mean of 7.7 (sd = .47) for the United States sample and from 6.7 to 7.8 for the Spanish sample, with an overall mean of 7.1 (sd = .39). Corresponding standard deviations ranged from 1.0 to 2.4, with an overall mean standard deviation of 1.4 (sd = .40) for the United States sample. For the Spanish sample, standard deviations ranged from 1.1 to 1.9 with an overall mean of 1.5 (sd = .30). Ratings also were highly leptokurtic. Given the problem of restriction of range, we did not pursue further analyses of the self-ratings.

In sum, our main findings were of (a) satisfactory psychometric properties for the ESJI, (b) satisfactory to excellent internal-consistency reliabilities of the inventory, (c) excellent consistency of the data across cultures, (d) satisfactory internal validity using five-factor model, and (e) moderate concurrent validity in the U.S. sample and modest concurrent validity in the Spanish sample. We believe that the study showed the feasibility of our approach in measuring practical cognition.