Purchase this article with an account.
Eugene Switkes, J. A. Wilson, Russell L. Valois; Glass pattern studies of contrast effects in feature integration and segmentation. Journal of Vision 2003;3(12):4. doi: https://doi.org/10.1167/3.12.4.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Glass patterns are ideal stimuli to study both local visual processing, where individual dots must be combined to form a perceptual pair, and global processes, where local oriented-pairs are integrated into patterns. At FVM-2002§ we described effects of chromatic variations on the detection of Glass correlations. We now report how luminance contrast variations alter an observer's ability to extract local features and to globally integrate dot-pairs.
We measured thresholds for pattern detection of rotational, translational, and hyperbolic correlations. Variation of the relative contrast of the dots within a dot-pair was used to investigate extraction of local oriented elements. Manipulation of contrast between dot-pairs allowed us to study the effects of contrast differences on the global integration of local features. Contrast variations included differences in contrast magnitude, where all dot-contrasts had the same polarity (increments or decrements), as well as differing contrast polarities.
For within-dot-pair contrast variations, we find: i) dots of opposite contrast polarity cannot be integrated to form local oriented features (as previously reported, Glass and Switkes, 1976); ii) for same-polarity contrast differences, integration is possible but thresholds increase as the ratio of the within-pair dot contrast increases; and iii) translational and rotational patterns give similar results. For between-dot-pair contrast variations, we find: i) correlated dot-pairs become more easily segregated from noise-pairs as the contrast difference between the signal and noise components increases; however ii) detection of translational correlations is reduced by noise of opposite-polarity while thresholds for rotational and hyperbolic patterns are unaffected by opposite-polarity noise of any contrast.
We present models of local and global processing to explain these observations. The contributions of the third-co-author to the study, as well as three decades of his provocative colleagueship and friendship, make it a special pleasure for me to present our work at this symposium honoring Professor Russell DeValois.
GlassL.SwitkesE.(1976). Pattern recognition in humans: correlations which cannot be perceived. Perception, 5, 67–72.
This PDF is available to Subscribers Only