Simulation-based training for enhancing psychiatric nurses' knowledge and clinical competencies in communication and psychiatric history taking: a quasi-experimental study.

Understanding a word such as requires linking a lexical form to its real-world referent and to a set of distinguishing semantic features (e.g., barks, has four legs, has a tail). Similarly, action-related words such as may derive their meaning from the hand action to which they refer, including the characteristic motor patterns involved in grasping and manipulating objects. Other categories, including colour words (red), sound-related words (whisper), or emotion words (fear), may likewise be grounded in perceptual, motor, or affective experiences. From this grounding perspective, linguistic symbols are meaningful because they are intrinsically tied to sensorimotor experiences, making word meanings difficult to explain without reference to such experiential representations (; ; ). By contrast, alternative accounts, the so-called amodal symbolic theories, argue that meaning is represented in abstract, modality-independent codes, whose computational efficiency derives precisely from being detached from perceptual and motor systems (; ; ). On this account, the semantic content of words such as dog, grasp, or red is encoded without requiring grounding in sensorimotor experience, which has been argued to provide computational advantages. This theoretical debate has shaped much of the research in neurosemantics, with the central question being whether word meaning is inherently grounded in perceptual and motor neural representations or instantiated by an independent, abstract symbolic module. Since then, much research has focused on identifying the neural correlates of semantic processing in the human brain in an attempt to resolve the debate. Neuroimaging studies show that symbolic processing engages a distributed cortical network extending beyond classical perisylvian language areas (; ), with modality-preferential visual, auditory, and motor regions exhibiting meaning-specific activation patterns for visually, auditory, and motor-related words, respectively (; ; ; ; ; ; ; ; ; ; ; ; ). Although some studies and theoretical accounts have questioned or failed to demonstrate consistent sensorimotor involvement in semantic processing (; ; ), converging evidence from lesion and patients studies shows that damage to these modality-preferential regions results in selective word comprehension deficits (; ; ; ; ; ; ; ). Beyond sensorimotor involvement, research also indicates the presence of category-general semantic regions, often referred to as semantic hubs, where different types of word meanings are processed in a similar manner. Although the anterior temporal lobe has long been proposed as an amodal semantic hub (; ; ), converging evidence suggests that hub-like semantic functions are not restricted to this region, but also involve the anterior inferior parietal and posterior inferior frontal cortices (; ; ; ; ; ; ). Taken together, these findings raise the question of modality-specific regions respond selectively to certain word types, whereas hub regions support general semantic processing, and biological mechanisms underlie these processes. Computational neural models provide a powerful tool for simulating symbolic learning and processing in the human cortex and for examining the emergence of functional roles of cortical regions. Although previous models successfully captured language and semantic patterns observed in human cognition (; ; ; ; ; ; ; ; ), most of these models remained relatively distant from the cortical structure and the white matter pathways at different levels. Some models did incorporate connectivity structure constraints into their architectures (; ; ), yet they often relied on biologically implausible learning mechanisms (backpropagation, ; ; ; ) or did not incorporate the range of multimodal hub regions documented in previous studies (see for a review, ; ). Towards a genuine neuromechanistic explanation of symbolic processing, several scholars have therefore emphasised the need to implement brain structural and physiological constraints into model architectures to directly link neural architecture to emergent cognitive phenomena (; ; ; ; ; ).