Secondary data analyses of large-scale international student assessment at the country level often reveal that educational systems which are high in academic achievement are found low in attitudinal outcomes of schooling, and vice versa. This is contrary to the non-negative attitude-achievement relationship frequently found at the student within-country level. Using attitudinal outcome "mathematics self-concept," the afore-mentioned "attitude-achievement paradox" is hypothesized to stem from the incomparable student responses to Likert-type response scale of the attitude items. In this study, anchoring vignettes are used to examine whether it is feasible to tackle the effect of response style so as to improve the comparability of the self-concept measures across cultures and societies. Empirical evidences from PISA 2012 demonstrate that the paradox can be resolved satisfactorily, so that the attitude-achievement relationships at both student within-country and country levels are now in line with that predicted by contemporary self-concept theories in the literature. Using the anchored self-concept measure, Singapore's and Shanghai-China's 15-year-olds are found to perform pretty well cognitively and affectively in PISA 2012.