Abstract
Purpose is to present a method of verifying an analyser's characteristics which is free of the probability summation effect. Method was designed to verify Quick's detection model generalised then by Wilson et al that consists of a set of linear analysers φI,..., φn followed by non-linear transducer functions, the outputs of which are combined according to the p-norm decision rule. It is based on two predictions derived from this model (Logvinenko, Biol. Cybern., in press). First, there exists a dual basis, i.e. a set of stimuli xI,...,xn such that the stimulus xi activates only the analyser φi despite the considerable overlap of the receptive fields of the neurons constituting the analysers. Second, detectability of a stimulus xi is not affected by presenting a stimulus xj(j ≠ i) despite the probability summation that may occur between analysers as well as between neurons within an individual analyser. So, a critical feature of the model is independent detection of stimuli in the dual basis. What is meant by independency here is that the tangent to the contrast interrelation function (i.e. the contour of equal probability of detection) for any pair of stimuli from the dual basis at the point where it intersects one coordinate axes should be parallel to the other axes (Logvinenko, ARVO, 1995). Results. Depicted is the dual basis evaluated for Wilson and Bergen's model. Each stimulus is marked with the same label as that used for the channel to which it is selectively tuned. A testable prediction of this model is that these four stimuli should be detected independently. Conclusion. The method proposed allows us to discount the effect of probability summation when testing an analyser's characteristics. (Graph Presented).
Original language | English |
---|---|
Pages (from-to) | S1070 |
Journal | Investigative Ophthalmology and Visual Science |
Volume | 37 |
Issue number | 3 |
Publication status | Published - 15 Feb 1996 |
Externally published | Yes |
ASJC Scopus subject areas
- Ophthalmology
- Sensory Systems
- Cellular and Molecular Neuroscience