“…As a first test of whether Roma/non-Roma differences are more pronounced on the more g loaded items, we carried out Jensen's (1980) method of correlated vectors. To do this, we followed the procedure used by Rushton and Skuy (2000), Rushton, Skuy, and Fridjohn (2002), Rushton, Skuy, and Fridjohn (2003), Rushton, Skuy, and Bons (2004) and correlated the item-total correlations shown in Table 2 (which estimate g), with the standardized differences between Roma and non-Roma in proportion passing each item shown in Table 1 (which estimates the population-group effect size). The correlation between the g loadings and the absolute magnitude of the five sets of Roma/non-Roma differences averaged 0.54 (P b 0.01; Pearson's r) and 0.53 (P b 0.01, Spearman's rho) using the item-total correlations for the Roma group, and 0.59 (P b 0.01; Pearson's r) and 0.55 (P b 0.01, Spearman's rho) using the item-total correlations for all the non-Roma groups.…”