Abstract
Serial dependence is a phenomenon in which people's perceptions of current stimuli values depend on previously seen stimuli. However, it is still poorly understood which aspects of a stimuli affect the amount of serial dependence observed. One theory is the visual system weights stimulus values adaptively depending on sensory noise. When stimuli are less noisy more weight is attached to new values and under more noisy conditions more weight is attached to past values. One problem is that previous studies have not quantified the amount of sensory noise present in stimuli and used a method of adjustment procedure that contains response noise. Here we measure sensory noise. We did this by using a two-alternative forced choice paradigm (2AFC). Participants viewed a Gabor stimuli presented at a random orientation. Next a second Gabor was presented at 7 different angles ranging 15° either clockwise or anticlockwise from the initial orientation. Sensory noise was manipulated by presenting the Gabors in two contrast conditions 5% contrast (high noise) and 20% (low noise). We fitted cumulative normal psychometric functions to the data and used the variance parameters as our estimate of sensory noise (5% = 115.85 deg2, 20% = 11.03 deg2). Data was compared to results from an experiment using a method of adjustment procedure with identical stimuli and contrast conditions. Serial dependence was quantified by determining the amount of influence previously shown stimuli have on the participants' response (5% contrast= 0.20, 20% =0.00). We found, consistent with the adaptive theory, the amount of sensory noise is important for predicting the amount of serial dependence observed. Our results indicate that quantifying the magnitude of sensory noise is important for interpreting the results of serial dependence and also that a 2AFC paradigm is an effective in method for quantifying the sensory noise in these experiments.
Meeting abstract presented at VSS 2018