Biomedical subjects
Publications and source records attributed to B Farell.
When we look at Anderson's Fig. 1a, or at just about anything else, we see surfaces and boundaries, contours and intersections, edges and angles. These complex, interpreted image features might well be the stimulus elements that stereo mechanisms analyse to recover depth. But are they?
An isoluminant color grating usually appears to move more slowly than a luminance grating that has the same physical speed. Yet a grating defined by both color and luminance is seen as perceptually unified and moving at a single intermediate speed. In experiments measuring perceived speed and direction, it was found that color- and luminance-based motion signals are combined differently in the perception of 1-D motion than they are in the perception of 2-D motion. Adding color to a moving 1-D luminance pattern, a grating, slows its perceived speed. Adding color to a moving 2-D luminance pattern, a plaid made of orthogonal gratings, leaves its perceived speed unchanged. Analogous results occur for the perception of the direction of 2-D motion. The visual system appears to discount color when analyzing the motion of luminance-bearing 2-D patterns. This strategy has adaptive advantages, making the sensing of object motion more veridical without sacrificing the ability to see motion at isoluminance.
Measuring the dependence of visual sensitivity on parameters of the visual stimulus is a mainstay of vision science. However, it is not widely appreciated that visual sensitivity is a product of two factors that are each invariant with respect to many properties of the stimulus and task. By estimating these two factors, one can isolate visual processes more easily than by using sensitivity measures alone. The underlying idea is that noise limits all forms of communication, including vision. As an empirical matter, it is often useful to measure the human observer's threshold with and without a noise background added to the display, to disentangle the observer's ability from the observer's intrinsic noise. And when we know how much noise there is, it is often useful to calculate ideal performance of the task at hand, as a benchmark for human performance. This strips away the intrinsic difficulty of the task to reveal a pure measure of human ability. Here we show how to do the factoring of sensitivity into efficiency and equivalent noise, and we document the invariances of the two factors.
Three-dimensional visual scenes project onto the retina of the eye as two-dimensional images. The third dimension, depth, is projected as subtle differences between left and right retinal images. As early as the 1830s, stereoscopic depth perception was shown to depend on horizontal disparities between these images. To detect disparity, the visual system must match corresponding parts of the two retinal images. To identify the stimulus elements used in stereo matching, I applied a disparity-adaptation technique to visual patterns whose one-dimensional components and two-dimensional features have very different disparities. Surprisingly, the adaptors that are effective in altering depth perception appear widely separated in depth from the patterns they adapt. I conclude that stereo matching occurs in all directions of two-dimensional space and that one-dimensional components are the stimulus primitives, the fundamental elements of stereo matching. This is a reversal of the classical view of stereo correspondence as a one-dimensional (horizontal) matching of monocular two-dimensional features.
A binocular pair of fiberscopes relays high-resolution images of CRT displays from an adjacent room to an observer lying in a scanner in functional Magnetic Resonance Imaging (fMRI) studies of visual function. We review the problems that must be overcome by any visual display for use in fMRI, present the specific solution we developed, and discuss its merits. Together, the fiberscope and CRT conveniently display accurately controlled high- and low-contrast wide-field images to an observer in an fMRI scanner.
Our initial purpose was to develop a quantitative method of estimating the cardinal directions of color space. The method is based on the finding that patterns consisting of pairs of drifting gratings modulated along different cardinal axes appear to slip with respect to one another, while the same patterns appear as a single coherent plaid if the modulation directions of the patterns are rotated by 45 deg in color space [Krauskopf & Farell (1990). Nature, 348,328-331]. A forced-choice procedure was used in which observers were asked to choose which of two successively presented patterns appeared less coherent. The patterns consisted of pairs of drifting gratings; the direction of modulation of one of the gratings was fixed and that of the other varied. For example, an estimate of an individual's isoluminant plane could be obtained by fixing the modulation of one grating in the luminance direction and finding the elevation of the modulation of the other grating that resulted in minimum perceived coherence. We found it important to take into consideration individual differences in the tilt of the isoluminant plane in color space and in the detectability of targets in the nominal cardinal directions. When this was done we found that reliable measurements could be made. The method effectively provided quantitative estimates of the cardinal directions. However, the most important result was the inadequacy of the generalization that patterns appear coherent when they share similar components along cardinal directions (Krauskopf & Farell, 1990) to account for the new results. The present results suggest that patterns appear not to cohere to the extent that they fail to stimulate common chromatic mechanisms, but the assumption that these mechanisms are tuned only along cardinal axes can be rejected. Along with other data the results point to the existence of higher-order mechanisms tuned to different isoluminant chromatic directions.
Evidence from several sources suggests that visual attention is tuned to stimulus scale. This tuning impairs performance when the observer must attend to more than one scale at a time. In experiments originally designed to measure the bandwidth of attention to stimulus scale, we have found tasks in which observers show no attentional tuning for scale. In separate blocks, observers located or identified targets in arrays of elements, either numbers in arrays of letters or static squares in arrays of flashing squares. Each display contained two arrays, either of the same scale or of different scales. The accuracy of locating the target element was lower in mixed-scale than in single-scale displays, but the accuracy of identifying the target was unaffected by a mixing of scales within the same display. This holds for both high-level discriminations--numbers vs letters--and low-level discriminations--static vs flashing. Thus, at least for identifying, one can attend to large and small at the same time. The difference in bandwidth between "what" and "where" implies that stimulus identification is not dependent on prior localization.
Offset thresholds were measured for targets whose horizontal profiles were either Gaussian or odd-symmetric Gabor functions. The targets were defined either by variation along the constant B or the constant R & G axes of color space or by luminance variation. Blur was varied in the case of the Gaussian targets by varying the standard deviation of the distribution and in the case of the Gabor functions by varying the spatial frequency of the sinusoidal component. Detection thresholds for all the stimuli were measured. The contrast of the targets used in the measurement of offset thresholds was varied from just above detection threshold to the maximum that could be produced. The offset thresholds obtained with targets of different chromatic composition are nearly identical when blur and contrast relative to detection threshold are held constant. We attribute the slight advantage held by luminance targets over chromatic targets for narrow Gaussians to the detectability of low frequency components of the chromatic targets which are of little use in the assessment of offsets. This conjecture is supported by the complete absence of such an advantage in the case of Gabor targets.
We have colour vision because there are three types of cone photoreceptors which are maximally sensitive in the long (L), middle (M) and short (S) wavelength regions of the spectrum. Psychophysical experiments have, however, revealed mechanisms selectively responsive to light modulated in three 'cardinal directions' in colour space. The responses of these mechanisms are determined by algebraic sums of the excitations of the cones. One of these mechanisms is responsive to changes in luminance, its spectral sensitivity being that of the sum of the L and M cones. The other two respond best to isoluminant changes in light. The responses of one of these mechanisms are determined by the difference in the excitations of the L and M cones, and those of the other one determined by the difference between the excitation of the S cones on the one hand and the excitations of the L and M cones on the other. We have obtained quite surprising results concerning the role of these mechanisms in the perception of motion. Drifting gratings modulated along different cardinal directions appear to slip with respect to one another. In contrast, when the directions of the modulations are rotated by 45 degrees in colour space, the gratings cohere. Our results are consistent with the notion that information about movement is analysed within mechanisms maximally responsive along the cardinal directions.
Visual sensitivity to spatial direction has classically been associated with motion perception. Yet humans are adept at deriving directional information in the absence of motion, as when they read maps, or follow arrows or animal tracks. Experiments are reported on the perception of parallel arrow-like forms in which a specific visual sensitivity to static direction is demonstrated. Global processing is operationally defined in terms of the relative discriminability of sets and subsets of stimulus elements; a set of parallel elements and a set in which one element is antiparallel to the rest are shown to be processed globally. The result of this global processing is a static analog of unidirectional optic flow. Global spatial direction differs fundamentally from other perceptions derived from static image processing. It involves long-range interactions in texture arrays, it does not carry information about stimulus location, and it is not reducible to the perception of component stimulus elements. Its likely function is in the construction of the layout of visual space.
Perceptual matching data show several puzzling effects. Particularly problematic are the disparities between the processing rates for same and different stimuli--the fast-same effect--and between the processing rates for two same-different judgment tasks that are related as mirror images--the task effect. Current models have difficulty accounting simultaneously for both effects. Central to these models is a stimulus comparison process that derives relative judgments of sameness and difference from tests of the congruence of stimulus representations. A contrasting view holds that same-different judgments can be modeled as absolute, rather than relative, judgments. This latter view is shown to be supported by experimental data. Reaction times for judgments of identical letter strings increase with string length at the same rate whether judgments are based on all the information in the strings or just the information in a single pair of component letters. The data show that stimulus comparisons of the sort described by previous models are not involved in these judgments. An attentional model accounts for the data and for the fast-same and task effects as well.
Explore the source record for details and available documents.
The distinction between operations in visual processing that are parallel and preattentive and those that are serial and attentional receives both theoretical and empirical support. According to Treisman's feature-integration theory, independent features are available preattentively, but attention is required to veridically combine features into objects. Certain evidence supporting this theory is consistent with a different interpretation, which was tested in four experiments. The first experiment compared the detection of features and feature combinations while eliminating a factor that confounded earlier comparisons. The resulting priority of access to combinatorial information suggests that features and nonlocal combinations of features are not connected solely by a bottom-up hierarchical convergence. Causes of the disparity between the results of Experiment 1 and the results of previous research were investigated in three subsequent experiments. The results showed that of the two confounded factors, it was the difference in the mapping of alternatives onto responses, not the differing attentional demands of features and objects, that underlaid the results of the previous research. The present results are thus counterexamples to the feature-integration theory. Aspects of this theory are shown to be subsumed by more general principles, which are discussed in terms of attentional processes in the detection of features, objects, and stimulus alternatives.