<p>In this paper, I develop a constructionist account of anxiety by integrating constructionism about emotion with the predictive processing framework. On this view, anxiety is not a biologically hardwired state but a high-level inferential state that arises when the brain uses learned emotion concepts to interpret and regulate unresolved threats. Emotion concepts are conceived as structured probabilistic models that guide perception, bodily regulation, and action through hierarchical prediction, shaped by developmental and social contexts. This account explains the forward-looking, epistemically charged character of anxiety as a form of stalled inference, while also accounting for its adaptive and maladaptive variants within a unified predictive architecture. By addressing Charlie Kurth’s objections to constructionism concerning unconscious emotions, the distinction between moods and emotions, and the role of concepts in behavior, the constructionist predictive processing account avoids reliance on separate fixed subsystems characteristic of Kurth’s biocognitive account. It offers a dynamic and neurocomputational framework that bridges philosophical and scientific approaches to emotion and clarifies the role of anxiety in human cognition and behavior.</p>