# Mathematics performance predicts cognitive and affective math anxiety through mutual mediation pathways from adolescence onward with potential working memory moderations

## TL;DR

Math performance influences both emotional and thinking-related math anxiety through interconnected pathways, with working memory playing a moderating role in high school and university students.

## Contribution

This study reveals a dual-pathway mediation structure linking math performance to cognitive and affective math anxiety, with working memory subsystems moderating these relationships differently in high school and university students.

## Key findings

- Math performance is linked to affective math anxiety through cognitive math anxiety and vice versa.
- Working memory subsystems moderate these links differently in high school and university students.
- Simpler mediation models without working memory interactions showed better fit than complex moderated models.

## Abstract

This research investigated the pathways from math performance to the cognitive (worried thoughts) and affective (emotional tension) dimensions of math anxiety (MA) in high-school and university students. Using the PLS-SEM method, we identified a consistent dual-pathway mediation structure across both groups: Math performance is associated with affective MA via cognitive MA, and with cognitive MA via affective MA. We also explored whether working memory (WM) subsystems, based on Baddeley’s model, moderate these links and found differences between these two successive academic cohorts. In high-school students, phonological loop (verbal) WM moderated the link from math to affective MA, with this effect being evident only for those with medium-to-low capacity. In contrast, for university students, a dual-pathway moderated-mediation structure was found: the pathway to cognitive MA was contingent upon medium-to-high capacity in visuospatial sketchpad (representing image and location) WM, whereas the pathway to affective MA was contingent upon medium-to-high capacity in both phonological and visuospatial WM. However, simpler mediation models without WM interactions consistently showed superior model fit. These findings support a stable interplay between the MA components across the sampled academic stages, emphasizing the explanatory and predictive primacy of these core mediation mechanisms during the transition from adolescence to young adulthood.

The online version contains supplementary material available at 10.1038/s41598-026-45516-y.