IS SHORT-TERM ODOR RECOGNITION PREDICTABLE FROM ODOR PROFILE

Smelling an odour induces a pattern of sensations, images and memories which participate in identification. It was proposed that perceptual memory performances for odours could be inferred from the description of these olfactory representations. The subject was asked to elaborate an odour descriptor profile, and a short-term odour recognition memor y task was chosen to test the individual perceptual memory performance . Two pattern-recognition methods based on artificial neural networks and discriminant analysis were carried out and allowed odour profile a nd perceptual memory performance to be related. Insofar as the subject s gave dichotomic responses in the recognition memory task, each respo nse could be evaluated in terms of correct or incorrect responses. Sim ulations indicated that the olfactory recognition memory performance c an be predicted in man from odour-elicited semantic profiles by using artificial neural networks. It was also shown that all semantic descri ptors do not participate in olfactory recognition to the same degree. Low-level information, such as intensity, familiarity and hedonic judg ements, did not allow the artificial neural network to predict the olf actory performance. By contrast, high-level information, such as gusta tory, olfactory and visual evocations, allowed artificial networks to make such predictions.