What Determines Contrast Sensitivity: An External Noise Study Across Spatial Frequencies?
62.16, Wednesday, May 15, 10:45 am - 12:45 pm, Royal Ballroom 1-3
CHANG-BING HUANG1, Ge Chen1, Fang Hou2, Fangfang Yan1, Pan Zhang1, Zhong-Lin Lu2; 1Key Laboratory of Behavioral Science, Institute of Psychology, Chinese Academy of Sciences 16, Lincui Rd, Chaoyang Dist., Beijing, 100101, China., 2Laboratory of Brain Processes, Department of Psychology, Ohio State University
Although it has greatly improved our understanding of a variety of sensory inefficiencies, pathological degradation, and system state changes (Ahumada & Lovell, 1971; Lu & Dosher, 2008; Ahumada, 1987; Geisler, 1989; Legge, Kersten, & Burgess, 1987; Lu & Dosher, 1999, 2001; Pelli, 1981, 1990; Huang, Tao, Zhou, & Lu. 2007), application of the external noise paradigm and observer models is usually implemented in a single spatial frequency condition. It is not clear how the various efficiencies vary across spatial frequencies and how they are related to the contrast sensitivity function (CSF). In the current study, we applied the external noise method and perceptual template model to stimulus conditions over a wide range of spatial frequencies. We evaluated how the various sources of observer inefficiency change with spatial frequency and determined the limiting factors underlying the contrast sensitivity function. The method of constant stimuli was used to measure contrast psychometric functions for a 2-AFC grating detection task at five spatial frequencies (0.5, 1, 2, 4 and 8 cpd) and eight external noise levels in four subjects. Each psychometric function was sampled at five contrast levels, based on pilot tests for each subject. There were a total of 200 combinations of spatial frequency, external noise and contrast levels (5*8*5), and 18,000 trials (200 conditions*90 trails/condition). We found that only internal additive noise (by more than 1000 times) and gain of the perceptual template (less than 2 times) changed significantly with spatial frequency, while transducer non-linearity and coefficient for multiplicative noise were constant. The 12-parameter model provided a very good account of all the data in the 200 conditions (87.5%, 86.0%, 90.0% and 96.4% for the four subjects, respectively). Our results contradict models that filter their inputs with CSF, and may shed new light on contrast sensitivity deficits in clinical populations.