Simulation of cognitive disturbances by a dynamic threshold semantic neural network AMIR B. GEVA 1 and AVI PELED 2 1 Electrical and Computer Engineering Department, Ben-Gurion University of the Negev, Beer Sheva, Israel 2 Department of Forensic Psychiatry, Shaa’r Menashe Mental Health Center, Israel; Technion-Israel Institute of Technology, Haifa, Israel (Received May 15, 1999; Revised October 1, 1999; Accepted October 8, 1999) Abstract A neural network model with dynamic thresholds, asymmetric connections, and clustered memories simulates spread activation that is hypothesized for semantic networks in the brain. By altering the parameters of the dynamic threshold a large range of disturbances can be generated in the model. These disturbances show metaphorical resemblance to certain general clinical descriptions of mental disturbances found in psychiatric patients engaged in various cognitive tasks. Even though the model is highly theoretical and metaphoric, it may help to gain certain insights into the relation between alterations of certain neural parameters, for example, thresholds and connectivity, and clinical symptoms in patients. (JINS, 2000, 6, 608– 619.) Keywords: Dynamic neural networks, Semantic networks, Cognitive task, Working memory, Concrete and abstract concepts representation, Schizophrenia, Thought disorders, Fuzzy clustering INTRODUCTION Recently neural network models have been used to simu- late many normal and pathological mental functions. The simulation of normal cognitive functions such as learning, memory, recognition, and categorization has provided im- pressive results (Hinton, 1981, 1986; Rumelhart & McClel- land, 1986). Simulations of cognitive disorders such as disturbances of memory activation and associations have also shown considerable achievements (Cohen & Servan- Schreiber, 1992, 1993; Hermann et al., 1993; Hinton & Shal- lice, 1991; Hoffman, 1987, 1992; Hoffman & Dobscha, 1989; Hoffman et al., 1994; Servan-Schreiber et al., 1996). Typi- cally simulations of cognitive disturbances involve a de- cline in the performance of a neural computation task as a direct consequence of a change in one of the networks’ pa- rameters, usually the threshold function. However, both the neuroscience literature and clinical experience suggest that brain functions are more complicated (Tucker, 1998; Van- Praag, 1997; Wilson, 1993). Neuroscience teaches us that many neuronal parameters (e.g., threshold, connectivity, and inputs) continuously change over time, and that a “bal- anced” interaction among many dynamic changes occurs dur- ing the normal functioning of the brain system (Globus, 1992; King, 1991). Clinical experience indicates an extraordinary variability (or spectrum) in the manifestation of cognitive disturbances (Spitzer & Williams, 1995; Tucker, 1998; Wil- son, 1993). It seems that while simple neural network models (i.e., models of fixed inputs, connections, and threshold alter- ations) are sufficient for simulating circumscribed mental disturbances, more complex models are required for simu- lating the complex variety (or spectrum manifestations) of mental disorders. Thus, increasing the intricacy of the model to approximate some of the complexities in the brain may serve to illustrate certain spectrum manifestations in men- tal disturbances not explained otherwise. In this work, the application of (1) dynamic threshold function, (2) asymmet- ric connections, (3) clustering of memory patterns, and (4) “internal inputs” offer the necessary complexity to sim- ulate variability and spectrum phenomena in mental dis- turbances. To demonstrate that certain common neural mechanisms can generate a wide variability in different cog- nitive functions, four different mental functions and their relevant tasks are chosen. The model simulates the distur- bances typically described in the psychiatric literature for Reprint requests to: Dr. Amir B. Geva, Electrical and Computer Engi- neering Department, Ben-Gurion University of the Negev, P.O.B. 653, Beer- Sheva 84105, Israel. E-mail: geva@ee.bgu.ac.il Journal of the International Neuropsychological Society (2000), 6, 608–619. Copyright © 2000 INS. Published by Cambridge University Press. Printed in the USA. 608