Past studies have shown that distinct yet highly correlated sub-constructs of three broad mathematics affective variables: (a) motivation, (b) attitudes and (c) anxiety, have varying degree of correlation with mathematics achievement. The sub-constructs of these three affective constructs are as follows: (a) (i) amotivation, (ii) external regulation, (iii) introjection, (iv) identification, (v) intrinsic motivation; (b) (i) enjoyment of mathematics, (ii) self-confidence in mathematics, (iii) perceived value of mathematics; (c) (i) anxiety with mathematics and (ii) ease with mathematics. This study identifies, both within and across these three affective variables, the key sub-constructs that educators should focus upon when selecting learning process variables for mathematics achievement. Results were analysed using a series of stepwise regression analyses using data from 1018 Grade 12 students enrolled in a top pre-tertiary institution in Singapore, both for concurrent and predictive relationships, and also for both genders. Results of this study showed that after taking into account self-confidence in mathematics and ease with mathematics, all other sub-constructs of mathematics motivation, attitudes and anxiety were not significantly correlated with mathematics achievement. This is true for both concurrent and predictive relationships, and for both genders. Implications of the results of this study are twofold. First, results of this study provide educators with priority affective variables to focus upon in their efforts to enhance performance in mathematics via the affective domain. Second, in situations where administration time is limited, it may be possible to consider only self-confidence and ease with mathematics when investigating relationships between affect and mathematics achievement.