Genetic factors underlie the association between anxiety, attitudes and performance in mathematics

Mathematics anxiety (MA) is common and linked to reduced engagement, motivation, and poorer mathematics performance. Around 30% of OECD 15-year-olds report anxiety or feeling incapable when solving math problems. Understanding why MA correlates with attitudes toward mathematics (self-efficacy and interest), cognition, and achievement is important for improving learning outcomes and guiding interventions. This study investigates: (1) the pattern of associations between MA and mathematics attitudes, abilities (understanding numbers, problem solving, approximate number sense), and achievement (GCSE); (2) the extent to which genetic and environmental factors explain these associations; and (3) whether general anxiety accounts for these links. The work aims to inform genetic discovery and developmental research on how genetically influenced predispositions shape math-related experiences.

- MA and attitudes (self-efficacy and interest): Prior research shows moderate-to-strong negative associations between MA and motivation/attitudes (perseverance, self-efficacy, interest) across student and pre-service teacher samples and cultures. Self-efficacy often mediates links between performance and MA, and between self-regulatory behavior and MA. Expectancy-value theory highlights the importance of subjective task value; limited studies suggest MA relates similarly to self-efficacy, interest, and perceived importance/usefulness of mathematics. It is unclear whether genetic/environmental influences on MA overlap similarly with self-efficacy and interest. - MA and achievement: Although high MA occurs across the ability distribution, higher MA is on average associated with lower achievement from primary school onward, beyond general cognitive ability. Longitudinal findings indicate reciprocal links between achievement and MA, with stronger effects from earlier achievement to later MA in some studies, and a role for self-evaluations/self-efficacy in linking MA and achievement. Evidence on basic numerical processing is mixed: some studies associate high MA with deficits in counting and simple enumeration, whereas links to basic numerosity discrimination are weak or absent. - Genetic evidence: One prior twin study reported that the association between MA and math problem solving is largely explained by common genetic influences. - MA and general anxiety: MA and general anxiety share physiological and neural features but are only moderately correlated (≈0.35). MA-performance associations often persist after accounting for general anxiety. Twin work suggests partial but not complete etiological overlap between MA and general anxiety, and that this overlap is independent from the etiology of MA–performance links.

Design and sample: Data come from the Twins Early Development Study (TEDS), a representative UK twin cohort (born 1994–1996). Ethical approval was granted by King’s College London; informed consent obtained. Analyses use data from two waves: age 16 and age 18–21. Exclusions applied for major medical, genetic, or neurodevelopmental disorders. - Age 16: Mathematics ability and achievement (N=3410 twin pairs; MZ=2612 individuals; DZ=4508; 56% female); mathematics self-efficacy and interest (N=2505 pairs; 61.2% female). - Age 18–21: Mathematics anxiety and general anxiety (N=1509 pairs; 63.9% female). Measures: - Mathematics anxiety (MA): Modified Abbreviated Math Anxiety Scale (AMAS), 9 items (5-point response), excellent internal consistency (α=0.94) and test–retest reliability (r=0.85). Two items adapted for age appropriateness. - Mathematics attitudes: Self-efficacy (8 items; 0–3 scale; α=0.90) and interest (3 items; 1–4 scale; α=0.93), adapted from OECD PISA. - Mathematics achievement: GCSE mathematics grade (A*–G recoded 11 to 4). Self-/parent-reported grades validated against National Pupil Database with high concordance (r≈0.99 for mathematics). - Mathematics performance tests (online): Understanding numbers (18 items; α=0.90); Problem Verification Test (48 timed items; α=0.85); Approximate number sense (150 trials; accuracy score; strongly correlated with Weber fraction). - General anxiety: GAD-7 (7 items; α=0.89; test–retest r=0.64). Analytic approach: - Phenotypic analyses: Descriptive statistics and ANOVAs on one randomly selected twin per pair; residualized for age and sex and standardized. - Twin modeling: Univariate ACE or ADE models to partition variance into additive genetic (A), shared/common environmental (C) or non-additive genetic (D), and non-shared environmental (E) influences based on intraclass correlations in MZ and DZ twins. Model choice (ACE vs ADE) guided by whether DZ correlations were more than or less than half of MZ correlations. • ACE models: GCSE, understanding numbers, problem verification (DZ > 0.5× MZ). • ADE models: MA, general anxiety, math interest, math self-efficacy, number sense (DZ < 0.5× MZ), indicating non-additive genetic effects. - Multivariate genetic analyses: Cholesky decompositions to decompose covariation among MA, attitudes, abilities, and achievement into A, C/D, and E components using cross-twin cross-trait covariances. This approach estimates shared and unique genetic/environmental influences, analogous to hierarchical regression. - Sex differences: Tested via ANOVAs; examined DZ same- vs opposite-sex correlations to assess qualitative/quantitative sex differences (none indicated); thus sexes modeled together in genetic analyses.

- Sex differences: Males showed higher math self-efficacy, interest, and performance, and lower math and general anxiety. Sex accounted for 0–7% of variance; no evidence for sex-specific etiologies. - Univariate genetic architecture: Best-fitting models were AE for most traits; GCSE showed significant shared environment (C=18%). Heritability estimates across measures ranged from 36% to 63%; remaining variance was non-shared environmental. - Phenotypic and genetic correlations: • MA with math attitudes (self-efficacy, interest): Moderate negative phenotypic associations ≈ −0.45; strong negative genetic correlations ≈ −0.70 (reported range −0.67 to −0.75 across math-related variables). • MA with math performance (understanding numbers, problem verification, GCSE): Moderate negative phenotypic associations ≈ −0.35; strong negative genetic correlations ≈ −0.70. • Exception—approximate number sense: Weaker associations with MA (phenotypic r ≈ −0.10; genetic r ≈ −0.31). - Multivariate genetic decomposition: The set of mathematics-related attitudes, abilities, and achievement jointly explained approximately 75% of the genetic variance in MA and about 20% of its environmental variance. - Role of general anxiety: The observed genetic overlap between MA and math attitudes/performance was not accounted for by general anxiety. - Overall pattern: Genetic effects were largely shared across MA, attitudes, abilities, and achievement, with approximate number sense showing comparatively distinct etiology relative to MA.

Findings demonstrate that the negative associations between mathematics anxiety and both mathematics attitudes (self-efficacy and interest) and performance measures are largely due to shared genetic influences. This suggests that individuals’ genetic predispositions simultaneously shape their experience of anxiety about mathematics and their confidence, interest, and performance in the subject, potentially via gene–environment correlation processes (e.g., selection or modification of math-related environments). The multivariate results, whereby three-quarters of the genetic variance in MA is captured by mathematics attitudes/abilities/achievement, underscore a common genetic architecture underlying math-related traits. Importantly, these overlaps were not explained by general anxiety, indicating that the etiology of MA is partly domain-specific rather than merely reflecting a general anxiety liability. The weaker association of approximate number sense with MA suggests that low-level nonsymbolic numerical acuity is less central to the MA network compared with higher-level math cognition and attitudes. These insights have implications for educational and clinical approaches: targeting math-specific attitudes (e.g., enhancing self-efficacy, interest) alongside performance may be effective, and genetic research can focus on shared variants influencing these correlated math traits.

This study shows that substantial shared genetic factors underpin the links between mathematics anxiety and mathematics-related attitudes and performance, with approximate number sense being a notable exception. Mathematics-related attitudes, abilities, and achievement jointly account for a large portion of the genetic variance in MA, and this overlap is not attributable to general anxiety. The results highlight a partially domain-specific genetic architecture of MA and suggest that students’ genetically influenced predispositions may guide their math-related experiences and outcomes. Future research should: (1) identify specific genetic variants contributing to the shared liability across MA, attitudes, and performance; (2) clarify developmental mechanisms and gene–environment interplay across schooling; (3) examine intervention targets focusing on self-efficacy and motivation alongside instruction; and (4) test generalizability across ages, cultures, and educational systems, and further probe the distinct role of approximate number sense.