Visual search and attention: a signal detection theory approach.
Explore the source record for details and available documents.
Biomedical subjects
Publications and source records attributed to P Verghese.
Explore the source record for details and available documents.
Spinal cord injury leads to many forms of autonomic dysfunction including autonomic dysreflexia, a condition involving recurrent episodes of paroxysmal hypertension and associated bradycardia. This hypertension may reach intensities that are life-threatening. We investigated autonomic dysreflexia and the sprouting of central processes of primary afferent neurons (a potential mechanism for autonomic dysreflexia) in a clinically-relevant calibrated clip-compression model of spinal cord injury in the rat. Autonomic dysreflexia was induced by colon distension in the conscious rats 2 weeks after severe (50-g) clip compression injury of the spinal cord at the 4th thoracic segment. The central arbor of small-diameter primary afferent fibers in laminae III-VII of the spinal cord dorsal horn was also assessed at 2 weeks after cord injury by quantitative morphometry, using calcitonin gene-related peptide as a marker. In response to colon distension, arterial pressure increased by 41 +/- 3 mmHg from a resting value of 109 +/- 4 mmHg, and heart rate decreased by 124 +/- 13 beats/min from a value of 515 +/- 16 beats/min (n = 7). Minimal locomotor function was recovered by these rats: by 2 weeks after injury they attained scores of only 3.1 +/- 1.3 on the Basso, Beattie and Bresnahan scale. Histopathology of the clip-compression lesion site in the cord consisted of extensive central necrosis extending several segments rostral and caudal to the lesion. Quantitative measures of the small-diameter afferent arbors revealed significant increases in area ranging from 20-27% in thoracolumbar segments caudal to the injury (n = 5) in comparison to sham-injured rats (n = 6). A second study was done to assess the impact of severity of injury on the relationship between the size of the primary afferent arbors and autonomic dysreflexia. At 2 weeks after milder (20-g) clip injury at T4, rats exhibited responses to colon distension that were not those associated with autonomic dysreflexia (n = 5). Arterial pressure increased by only 16 +/- 3 mmHg and heart rate tended to increase (+19 +/- 12 beats/min). These rats attained a locomotor score of 7.1 +/- 0.4 by 2 weeks. The lesions at the injury site also contained necrosis and mild cavitation within the gray matter. No change in the small-diameter afferent arbor was detected at 2 weeks after the 20-g clip injury at T4 (n = 6 rats). These findings suggest that after severe but not mild clip compression injury of the spinal cord, sprouting of the afferent component of the spinal reflex are contributes to the development of autonomic dysreflexia. Neither dysreflexia, nor changes in the afferent arbor size occurred after mild cord injury. This clinically relevant clip compression cord injury model, studied more frequently for locomotor function, is excellent for investigating mechanisms for the development of autonomic dysreflexia and strategies for its prevention.
Most theories of visual search emphasize issues of limited versus unlimited capacity and serial versus parallel processing. In the present article, we suggest a broader framework based on two principles, one empirical and one theoretical. The empirical principle is to focus on conditions at the intersection of visual search and the simple detection and discrimination paradigms of spatial vision. Such simple search conditions avoid artifacts and phenomena specific to more complex stimuli and tasks. The theoretical principle is to focus on the distinction between high and low threshold theory. While high threshold theory is largely discredited for simple detection and discrimination, it persists in the search literature. Furthermore, a low threshold theory such as signal detection theory can account for some of the phenomena attributed to limited capacity or serial processing. In the body of this article, we compare the predictions of high threshold theory and three versions of signal detection theory to the observed effects of manipulating set size, discriminability, number of targets, response bias, external noise, and distractor heterogeneity. For almost all cases, the results are inconsistent with high threshold theory and are consistent with all three versions of signal detection theory. In the Discussion, these simple theories are generalized to a larger domain that includes search asymmetry, multidimensional judgements including conjunction search, response time, search with multiple eye fixations and more general stimulus conditions. We conclude that low threshold theories can account for simple visual search without invoking mechanisms such as limited capacity or serial processing.
We measured the detectability of moving signal dots in dynamic noise to determine whether local motion signals are preferentially combined along an axis parallel to the direction of motion. Observers were asked to detect a signal composed of three dots moving in a linear trajectory among dynamic noise dots. The signal dots were collinear and equally spaced in a configuration that was either parallel to or perpendicular to their trajectory. The probability of detecting the signal was measured as a function of noise density, over a range of signal dot spacings from 0.5 degrees to 5.0 degrees. At any given noise density, the signal in the parallel configuration was more detectable than that in the perpendicular configuration. Our four observers could tolerate 1.5-2.5 times more noise in the parallel configuration. This improvement is not due merely to temporal summation between consecutive dots in the parallel trajectory. Temporal summation functions measured on our observers indicate that the benefit from spatial coincidence of the dots lasts for no more than 50 ms, whereas the increased detectability of the parallel configuration is observed up to the largest temporal separations tested (210 ms). These results demonstrate that dots arranged parallel to the signal trajectory are more easily detected than those arranged perpendicularly. Moreover, this enhancement points to the existence of visual mechanisms that preferentially organize motion information parallel to the direction of motion.
Previous work has shown that a single dot moving in a consistent direction is easily detected among noise dots in Brownian motion (Watamaniuk et al., Vis Res 1995;35:65-77). In this study we calculated the predictions of a commonly-used psychophysical motion model for a motion trajectory in noise. This model assumes local motion energy detectors optimally tuned to the signal, followed by a decision stage that implements the maximum rule. We first show that local motion detectors do indeed explain the detectability of brief trajectories (100 ms) that fall within a single unit, but that they severely underestimate the detectability of extended trajectories that span multiple units. For instance, a 200 ms trajectory is approximately three times more detectable than two isolated 100 ms trajectories presented together within an equivalent temporal interval. This result suggests a nonlinear interaction among local motion units. This interaction is not restricted to linear trajectories because circular trajectories with curvatures larger than 1 degree are almost as detectable as linear trajectories. Our data are consistent with a flexible network that feeds forward excitation among units tuned to similar directions of motion.
We address a surprising result in a previous study of speed discrimination with multiple moving gratings: discrimination thresholds decreased when the number of stimuli was increased, but remained unchanged when the area of a single stimulus was increased [Verghese & Stone (1995). Vision Research, 35, 2811-2823]. In this study, we manipulated the spatial- and phase relationship between multiple grating patches to determine their effect on speed discrimination thresholds. In a fusion experiment, we merged multiple stimulus patches, in stages, into a single patch. Thresholds increased as the patches were brought closer and their phase relationship was adjusted to be consistent with a single patch. Thresholds increased further still as these patches were fused into a single patch. In a fission experiment, we divided a single large patch into multiple patches by superimposing a cross with luminance equal to that of the background. Thresholds decreased as the large patch was divided into quadrants and decreased further as the quadrants were maximally separated. However, when the cross luminance was darker than the background, it was perceived as an occluder and thresholds, on average, were unchanged from that for the single large patch. A control experiment shows that the observed trend in discrimination thresholds is not due to the differences in perceived speed of the stimuli. These results suggest that the parsing of the visual image into entities affects the combination of speed information across space, and that each discrete entity effectively provides a single independent estimate of speed.
Little is known about how or where the visual system parses the visual scene into objects or surfaces. However, it is generally assumed that the segmentation and grouping of pieces of the image into discrete entities is due to 'later' processing stages, after the 'early' processing of the visual image by local mechanisms selective for attributes such as colour, orientation, depth, and motion. Speed perception is also thought to be mediated by early mechanisms tuned for speed. Here we show that manipulating the way in which an image is parsed changes the way in which local speed information is processed. Manipulations that cause multiple stimuli to appear as parts of a single patch degrade speed discrimination, whereas manipulations that perceptually divide a single large stimulus into parts improve discrimination. These results indicate that processes as early as speed perception may be constrained by the parsing of the visual image into discrete entities.
We used speed discrimination tasks to measure the ability of observers to combine speed information from multiple stimuli distributed across space. We compared speed discrimination thresholds in a classical discrimination paradigm to those in an uncertainty/search paradigm. Thresholds were measured using a temporal two-interval forced-choice design. In the discrimination paradigm, the n gratings in each interval all moved at the same speed and observers were asked to choose the interval with the faster gratings. Discrimination thresholds for this paradigm decreased as the number of gratings increased. This decrease was not due to increasing the effective stimulus area as a control experiment that increased the area of a single grating did not show a similar improvement in thresholds. Adding independent speed noise to each of the n gratings caused thresholds to decrease at a rate similar to the original no-noise case, consistent with observers combining an independent sample of speed from each grating in both the added- and no-noise cases. In the search paradigm, observers were asked to choose the interval in which one of the n gratings moved faster. Thresholds in this case increased with the number of gratings, behavior traditionally attributed to an input bottleneck. However, results from the discrimination paradigm showed that the increase was not due to observers' inability to process these gratings. We have also shown that the opposite trends of the data in the two paradigms can be predicted by a decision theory model that combines independent samples of speed information across space. This demonstrates that models typically used in classical detection and discrimination paradigms are also applicable to search paradigms. As our model does not distinguish between samples in space and time, it predicts that discrimination performance should be the same regardless of whether the gratings are presented in two spatial intervals or two temporal intervals. Our last experiment largely confirmed this prediction.
Observers were asked to locate a target in a brief, two-scale display. Accuracy of locating the target was measured as a function of the ratio between the two scales. At each scale ratio, the probability of locating the target as a function of the number of elements is well fit by the idea that the observer accurately monitors only a "critical" number of elements. The dependence of critical number on scale ratio is well accounted for by a model that assumes that the observer's decision is based on an evenly spaced array of samples. The sample spacing is under attentional control, but is always uniform.
We measured the probability of detecting the target in a visual search task, as a function of the following parameters: the discriminability of the target from the distractors, the duration of the display, and the number of elements in the display. We examined the relation between these parameters at criterion performance (80% correct) to determine if the parameters traded off according to the predictions of a limited capacity model. For the three dimensions that we studied, orientation, color, and spatial frequency, the observed relationship between the parameters deviates significantly from a limited capacity model. The data relating discriminability to display duration are better than predicted over the entire range of orientation and color differences that we examined, and are consistent with the prediction for only a limited range of spatial frequency differences--from 12 to 23%. The relation between discriminability and number varies considerably across the three dimensions and is better than the limited capacity prediction for two of the three dimensions that we studied. Orientation discrimination shows a strong number effect, color discrimination shows almost no effect, and spatial frequency discrimination shows an intermediate effect. The different trading relationships in each dimension are more consistent with early filtering in that dimension, than with a common limited capacity stage. Our results indicate that higher-level processes that group elements together also play a strong role. Our experiments provide little support for limited capacity mechanisms over the range of stimulus differences that we examined in three different dimensions.
BACKGROUND: A retrospective study of five patients with the clinical or magnetic resonance-based diagnosis of carotid dissection was done. Clinical data, imaging studies, treatment, and outcome were reviewed. The potential applicability of three-dimensional time-of-flight magnetic resonance angiography in these settings was evaluated. SUMMARY OF REPORT: This modality reliably showed vascular abnormalities and focal, segmental, or aneurysmal dilatation when correlated with conventional angiograms in three patients. Turbulence and magnetic susceptibility at the acute turn of the carotid in the petrous canal led to a false-positive diagnosis on magnetic resonance angiography in one patient (in whom subtle fibromuscular hyperplasia was found with conventional angiography but missed with magnetic resonance angiography). CONCLUSIONS: When combined with appropriate clinical signs, magnetic resonance imaging and magnetic resonance angiography can reliably establish the diagnosis of carotid dissection. Pitfalls of magnetic resonance angiography are discussed.
Is visual attention mediated by a general-purpose processor with a small data capacity? Such an attentive processor could perform a wide range of transformations upon a small amount of image data. We suggest that this limited capacity corresponds to a fixed amount of information, measured in bits. We measure how much information an observer's attention can handle by measuring how much we can restrict display information without impairing the observer's performance. The attentive visual tasks we study are the detection of a stationary dot in a field of moving dots, and the detection of a static square in a field of flashing squares. Performance of these tasks is perfect up to a critical number of elements (the span of attention) and then falls as the number of elements increases beyond this critical number. The display information required for unimpaired performance in each of these tasks is low; the results indicate that visual attention processes only 30 to 60 bits of display information.