subtracting each judge's rating of the typical American from his or her rating of the typical compatriot for each NCS item. Assuming that cultures agree on the typical American, this procedure in effect subtracts the bias plus a constant and leaves a potentially better estimate of national character. We standardized the differences as T scores, using difference score normative values from the worldwide sample, excluding the United States. The difference scores were highly correlated with NCS

more »

... res (rs 0 0.65 to 0.91, P G 0.001) and provided essentially the same results. ICCs between difference scores and NEO-PI-R observer ratings ranged from -0.44 for England to 0.48 for Lebanon (median, 0.03). ICCs between differences scores and NEO-PI-R self-reports ranged from -0.47 for Russia to 0.53 for Poland (median, 0.01). For the five factors, correlations with observer ratings across cultures ranged from 0.08 to 0.23, and those with self-reports ranged from -0.37 to 0.23. These results suggest that the lack of correspondence between NEO-PI-R and NCS profiles is not simply due to different standards of evaluation in different cultures. A different issue concerns the referencegroup effect (28), according to which self-reports and observer ratings of individuals are implicitly made by reference to the distribution of scores in the rater's culture. Such an effect would tend to make aggregate personality scores uniform for all cultures, and the failure to find correlations with NCS factors would be due to a lack of variation in aggregate NEO-PI-R means. However, NEO-PI-R means in fact vary systematically across cultures and show strong correlations across methods and with other culturelevel variables (12, 14) . Thus, the reference-group effect cannot explain the failure to find correlations with NCS scales. 28.