This thesis describes the elements that form part of a comprehensive neuromus-cular simulation system centered around control of the human arm. The re-sulting computational system necessarily covers many ¯elds of study, and spans several orders of magnitude: From the molecular level where muscle proteins generate forces, to the macroscopic levels where overt arm movements are vol- untarily controlled within an unpredictable environment by legions of neurons¯ring in orderly fashion. An extensive computer simulation system has been developed for this thesis, which at present contains a neural network scripting language for specifying arbitrary neural architectures, de¯nition ¯les for detailed spinal networks, various biologically realistic models of neurons, and dynamic synapses. Also included are structurally accurate models of intrafusal and extra-fusal muscle ¯bers and a general body-centered mechanical physics simulation system in which a realistically scaled human arm with accurate muscle origin- insertion points was modelled. At the molecular level, a novel hypothesis regarding the origin of muscle force is proposed. It is concluded that within the framework laid out by the sliding¯lament theory, the conformational entropy of the individual myosin molecules has a central role to play in the total force production of the sarcomere. All in all, much emphasis has been given in this thesis to develop a highly detailed model of human muscle. The ¯nal muscle ¯ber model accounts for a variety of phenomena, ranging from the force-velocity and force-length relationships, to tetanic fusion, "catch-like" e®ects and the distinctions between fast and slow muscle ¯ber types. Furthermore the model incorporates su±cient neuromus-cular information as to permit orderly recruitment of motor units, exponential motor-unit size distributions and gradual force increases. Also included in the computational model was a mathematical model of an important class of sensory receptors known as muscle spindles which report to the central nervous system on the length and contraction velocity of the inner- vated muscles. From the simulations it was concluded that the dynamic range of the modelled spindles, as they responded to fusimotor input, was such that it was possible to maintain constant activity levels in the primary and secondary a®erents. Further theoretical analysis of this spindle model revealed that an explicit function may be derived which expresses the force that the spindle contractile elements must produce to exactly counter spindle unloading during muscle shortening. This information was used to calculate the corresponding "optimal" °-motoneuronal activity level. For some simple arm movement tasks,this permits the derivation of a signature activation pattern which in principle may be used to identify such cells in vivo by monitoring spinal activity during tasks that are similar.