respective factors and the inter-correlations between factors did not exceed .85
recommended as the cutoff point for discriminant validity purposes (Kline, 1994). The
internal Cronbach alpha-reliability of the help scale was .83, the reliability for innovative
behavior was .90; the competitive/high performance scale had a reliability of .73, and
finally, the scale for compliant behavior exhibited a reliability of .80.
The scales used to measure behaviors also exhibited good fit using the supervisor
perceptions (CFI= .96; SRMR= .10; RMSEA= .07). The items were loading on the
specified latent factors, and none of the inter-correlations between factors was higher
than .85. The reliabilities in the supervisor sample were as follows: .85 (help), .93
(innovative behavior), .84 (achievement), and .79 (compliant behavior).
Issues of aggregation with behaviors. In order to justify the aggregation of
coworker rated behaviors for each individual I selected a score of within-group inter-rater
agreement Rwg to establish if there is sufficient agreement between respondents (James,
Demaree, & Wolf, 1984, 1993). The median Rwg values for the help/cooperation,
innovation, achievement, and compliant behavior, were as follows: .88, .87, .67 and 0.61
(in the coworker sample). All of these values are above the recommended level of .60
(James, 1982), thus, justifying the aggregation of multiple sources for each individual.
In addition, I calculated the ICC statistics using the same procedures as with the
measures of culture. The F-statistics associated with organizational membership for three
of the behaviors rated by coworkers was significant at p<. 001; it was not significant for
achievement behavior (p< .18). The ICC (1) values in the coworker sample were as
follows: 0.10 for help, 0.08 for achievement behavior, 0.20 for innovative behavior, and
finally, 0.12 for compliant behavior, thus, suggesting that there are some discernible