measure of sampling adequacy (KMO) was used to test the appropriateness of using a PCA. The KMO value was found to be 0.955; this result strongly supports the use of a PCAandsuggeststhatthedatamaybegroupedintoasmallersetofunderlyingfactors. Two extraction criteria were used: eigenvalues greater than one and interpretability. The analysis of the items identified four factors and these accounted for approximately 69percentofthevarianceinthevariableset.Ofthe32items30wereusedtodelineatethe components. Four subscales were then derived from a grouping of the items as defined by the majorfactors. This derivation resultedby summingthe raw scores of each item loading on a factor and dividing by the number of items in the subscale. The subscales were labelled as follows: research subscale 1 (reporting and supervising research); research subscale2(skillsrelatedtotheconductandmanagementofresearch);researchsubscale 3 (writing major works and reviewing articles/books); and, research subscale 4 (having a broad view of a research area). The subscales had kurtosis and skewness values within or reasonably close to the range 21 toþ1 and thus deemed to be appropriately normally distributed and suitable for analysis using bivariate and multivariate techniques (Hair et al., 1998). All of these measures, as well as the reliability coefficients of the subscales, are presented in Table I. A measure was developed from the number of peer-reviewed publications the participants had published during their career. As directed by federal governmental guidelines, these publications record a particular score. That is, journal articles, conference papers, and book chapters each receive a point, whereas peer-reviewed books are scored as five points. Participants’ points were then added. This measure, publication total, was constructed by transforming the raw publication data to produce six divisions showing a symmetrical and close to normal distribution. The resulting measure had a skewness of 20.13 and a kurtosis of 21.06 and met the assumptions of a dependent variable required in a multiple regression analysis (Hair et al., 1998). A correlation matrix (n¼343) is presented in Table II. An inspection of the correlation coefficients revealed that all the subscales were positively and significantly related (p , 0.01), and each subscale was also significantly correlated with Publication Total (p , 0.01). A multiple regression analysis was undertaken to determine the extent to which the four research subscales predicted publication total. By entering all subscales in a single block, theresults showed that research subscales 1, 2, and 3 contributedsignificantly to the model at the 5 percent level and that the adjusted R 2 was 0.46 (Table III). However, itneedstobenotedthatifresearch subscale4wereentered firstinamultipleregression analysis it would account for in excess of 16 percent of the variance in publication total.