Many scientific and mathematical concepts are counterintuitive because they conflict with misleading perceptual cues or incorrect naive theories that we build from our everyday experiences of the world. Executive functions (EFs) influence mathematics and science achievement, and inhibitory control (IC), in particular, might facilitate counterintuitive reasoning. Stop & Think (S&T) is a computerised learning activity that trains IC skills. It has been found effective in improving primary children’s mathematics and science academic performance in a large scale RCT trial (Palak et al., 2019; Wilkinson et al., Journal of Cognitive Enhancement, 4, 296–314, 2020). The current study aimed to investigate the role of EFs and the moderating effects of S&T training on counterintuitive mathematics and science reasoning. A sample of 372 children in school Years 3 (7- to 8-year-olds) and 5 (9- to 10-year-olds) were allocated to S&T, active control or teaching as usual conditions, and completed tasks assessing verbal and visuospatial working memory (WM), IC, IQ, and counterintuitive reasoning, before and after training. Cross-sectional associations between counterintuitive reasoning and EF were found in Year 5 children, with evidence of a specific role of verbal WM. The intervention benefited counterintuitive reasoning in Year 3 children only and EF measures were not found to predict which children would most benefit from the intervention. Combined with previous research, these results suggest that individual differences in EF play a lesser role in counterintuitive reasoning in younger children, while older children show a greater association between EFs and counterintuitive reasoning and are able to apply the strategies developed during the S&T training to mathematics and science subjects. This work contributes to understanding why specifically the S&T intervention is effective. This work was preregistered with the ISRCTN registry (TRN: 54726482) on 10/10/2017.