AbstractRandom noise has been shown to improve the detection of suboptimal signals in humans and machines. Based on that, transcranial random noise stimulation (tRNS) has aimed to improve human behaviour by targeting neuronal activity. To uncover the poorly understood mechanistic underpinnings of tRNS, we recorded electroencephalography data during arithmetic training while delivering active or sham tRNS above the dorsolateral prefrontal cortex (dlPFC). By successfully removing the tRNS artefact in the time and frequency domains, we examined the mechanisms that underlie its behavioural improvement. We found that active tRNS improved arithmetic performance and impacts specific ERPs components that are associated with attentional mechanisms. Furthermore, the tRNS effect was maximal in individuals with suboptimal arithmetic ability and neurophysiological measures of top-down control and excitation/inhibition ratio. These results providing a novel mechanistic explanation for the effect of tRNS on human behaviour and highlight how suboptimal task-specific behaviour and neurophysiology predicts its effect.