A brain-mapped neural network that combines attentional and configural mechanisms is able to characterize the attributes of multiple classical conditioning paradigms and to describe the effects of many neurophysiological manipulations. As shown in Buhusi and Schmajuk (1996), the attentional-configural model describes neural activity in several brain regions. As also shown in Buhusi and Schmajuk (1996), the attentional-configural model has been applied to the description of the effects of lesions. Schmajuk and Buhusi (1997) illustrated how the configural model correctly describes the effect of these lesions on discrimination paradigms in which stimuli can act as a simple conditioned stimulus (CS) or an occasion setter. Buhusi, et al. (1998) showed that the model can offer a resolution for the apparently conflicting results of hippocampal selective and nonselective lesions on latent inhibition. Schmajuk, et al. (1998) showed that the attentional model offers a description of the interaction between the procedural design and administration of dopaminergic drugs on latent inhibition. Many times, the effect of brain manipulations seems to be specific to the parametric conditions of the experiment, duration of the CS in the case of hippocampal lesions and trace conditioning, procedure and total time of preexposure in the case of hippocampal lesions and latent inhibition, and CS and unconditioned stimulus (US) intensity and duration in the case of dopaminergic agents and latent inhibition. The specificity of these results is well captured by the neural network approaches described in this article. We can now describe, in terms of the model, the functional anatomy of eyeblink conditioning, in animals and humans presented. Although the combination of traditional and advanced technologies can bring an enormous amount of exciting information about how different regions in the human brain participate in eyeblink conditioning, our understanding of the functionality of these regions can only be achieved with the help of formal neural network models.

In recent years, neural networks have been proposed that portray many of the complexities of adaptive behavior. The networks describe how agents learn to predict future events by: 1) building models of the would, 2) inferring new predictions from past experiences, 3) combining elementary environmental stimuli into complex internal representations, 4) attending to stimuli associated with environmental novelty, and 5) attending to stimuli that are good predictors of other environmental events. When a predictive network is attached to a goal seeking system, the resulting architecture is able to describe spatial and maze navigation, as well as problem solving and planning. When the predictions of future events are based on the combination of environmental stimuli and the animal's own responses the networks provide the information necessary to choose between alternative behaviors. When the agent's own responses can be identified with the responses of other agents, the networks can describe learning by imitation. It is suggested that these principles might be applied to the design of adaptive, communicating autonomous robots.

A model of the vestibulo-ocular reflex was built using neural network techniques. The network has three levels, corresponding to the three cell types which compose the basic reflex. The model's performance was tested by supplying single-frequency sinusoidal stimuli as input and measuring the gain and phase of the output waveforms generated. Network parameters were tuned for optimum performance at two different input frequencies and the model was tested across a range of frequencies. Gain and phase relationships generated by the model correspond in general to those observed in the normally functioning reflex, but the frequency tuning of the network is much narrower. The significance of these results in relationship to other models of the vestibulo-ocular reflex is discussed