You are here:
Detection of visual signals by rats: A computational model
Schmajuk, N. A. AND P. J. BUSHNELL. Detection of visual signals by rats: A computational model. BEHAVIOURAL PROCESSES. Elsevier Science Ltd, New York, NY, 82(3):340-351, (2009).
We applied a neural network model of classical conditioning proposed by Schmajuk, Lam, and Gray (1996) to visual signal detection and discrimination tasks designed to assess sustained attention in rats (Bushnell, 1999). The model describes the animals’ expectation of receiving food reward (the unconditioned stimulus, US), when pressing a ‘signal lever’, in terms of associations formed between the signal light (conditioned stimulus, CS) and contextual cues (CX) with the US. These classical associations determine the probability with which the animal presses the signal lever after presentation of lights of different intensity. The model explains the proportion of Hits, P(hit), and the proportion of False Alarms, P(fa), emitted by trained rats as functions of light intensity, event rate and task type. Competition between the CX and the CS to gain associations with the US explains the inverse relationships between P(hit) and P(fa) in both tasks. Lower P(hit) and higher P(fa) were also observed in the discrimination task, in which a dim light was presented on all blank trials. The model explains the lower accuracy in the discrimination task in terms of partial extinction of the CS-US association when the signal light is turned on at a low brightness level. Increased and overlapping P(fa) values in the discrimination task were the result of these values being mostly proportional to the dim light intensity and the light-US association. Thus the model describes signal lever pressing behavior in terms of (a) the associations between a compound stimulus of the signal and the signal lever with reinforcement, and (b) the attention paid to the compound stimulus. However, the model does not account for the response selection aspects of the tasks.
Toxicology in the 21st century will increasingly rely on modeled estimations of toxicity. Predictive models require information about the kinetics of the chemicals of concern and their effects at biological targets and how those effects are propagated dynamically into adverse effects on the function of the test animal or human. Kinetic models are relatively well developed in comparison to dynamic models that relate effects at the chemical-target interface to adverse outcomes, particularly in the nervous system. This manuscript describes the application of a neural network model to a behavioral signal-detection method that has been validated and used to assess the effects of many chemicals on sustained attention in test animals. The model accounts for the behavior of rats performing the task in response to three parametric manipulations, and provides quantitative estimates of the roles of associative and attentional processes in this performance. This report lays the foundation for further work to explain the effects of CNS-active drugs and toxic chemicals on this behavior. This work demonstrates how neural network models may be developed to relate psychological processes to behavior, and provides an approach for revealing mechanisms by which chemicals acting at target receptors in the brain can alter behavior.
Record Details:Record Type: DOCUMENT (JOURNAL/PEER REVIEWED JOURNAL)
Organization:U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF RESEARCH AND DEVELOPMENT
NATIONAL HEALTH AND ENVIRONMENTAL EFFECTS RESEARCH LAB
NEUROBEHAVIORAL TOXICOLOGY BRANCH