Language comprehension is remarkably efficient, in part because humans continuously generate predictions about upcoming words. Yet predictions can be wrong, raising a fundamental question: What happens when linguistic expectations are violated, and do the mechanisms that resolve these conflicts rely on language-specific or domain-general neural systems? In this talk, I will present a line of research examining how the brain responds to unexpected words during real-time language processing. Behavioral studies show that when predictions are violated, individuals actively inhibit previously predicted words, resulting in measurable costs to lexical access. Moreover, prediction violations impair simultaneous action stopping, suggesting that linguistic and motor inhibition draw on a common underlying process. Converging evidence from EEG, intracranial recordings, and deep-brain stimulation demonstrates overlapping neural signatures of motor and linguistic inhibition, including shared decoding patterns and increased beta-band activity within the subthalamic nucleus (STN). Together, these findings suggest that the inhibition of outdated linguistic representations is supported by a domain-general fronto-basal ganglia control mechanism, providing a new perspective on how the brain resolves conflict during language comprehension.