Keywords: Computational Neurosciences; Performative Capacity of Neural Circuits; Neuromodulation; Computational Models of Neuromodulation.
It is increasingly evident that an understanding of both the functional repertoire and the behavioral plasticity of neural circuits - here referred to as the circuit's perfor-mative capacity - cannot be solely based on the knowledge of neural connectivity diagrams and their variations. The performative capacity of a neural circuit is in-deed qualitatively altered in a rapid and reversible way by various substances, called neuromodulators, in the extracellular fluid. This paper focuses on two properties of neuromodulatory action that are relatively neglected in current computational models of neuromodulation: the functional soundness of neuromodulated circuits and the robustness of neuromodulatory action. Both properties are examined here as sources of functional specifications for computational models of neuromodula-tory action. In particular, it is pointed out that in a computational model based on CTRNNs (Continuous Time Recurrent Neural Networks) robustness can be mod-eled through a hysteresis process and functional soundness through a multiplicity of stable fixed points of the dynamic system.