Click for next page ( 3


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 2 CONTRAST SENSITIVITY FUNCTION BACKGROUND The most widely used measure of visual resolution is visual acuity. It is used both for clinical diagnosis and evaluation and for legal screening and selection. (See the report of another working group, National Research Council 1980, for a discussion of methods and standards for the measurement of visual acuity.) Acuity is based on the size of the smallest detail in a visual target (optotype) that permits some criterion level of identification or detection performance (75 percent correct, for example). The smaller the size of this critical detail, the better the vision of the observer. The value of visual acuity measurements is well proven for correcting refractive errors. Under some conditions, however, individual variation in standard measurements of visual acuity often is not able to predict individual variation in performance on some visual tasks, such as target detection and identification (Ginsburg, 1983; Ginsburg et al., 1982, 1983). A considerable body of empirical knowledge has been gained about the stimulus factors, such as size, exposure duration, contrast, and adaptation level, that influence detection of simple disk-shaped targets (Graham and Margaria, 1935; Lukiesch and Moss, 1940; Blackwell, 1946). Although these data in some circumstances do quite well in predicting detection of more complex targets, they often are inadequate in predicting recognition and identification of these targets. In addition, individual differences in performance with simple targets are not easily related to any measured characteristics of vision, nor were they related via any satisfactory theoretical framework. In the past two decades, a new method of assessing vision has emerged that may provide a universal language. This method is the measurement of the contrast sensitivity function and, for some purposes (described below), it complements visual acuity. The first study that demonstrated the power of contrast sensitivity to supplement acuity measures employed low contrast Landolt C target (Hecht et al., 1949). Today, however, the contrast sensitivity function is typically measured using sinusoidal grating patterns as targets. This use of sine wave gratings was first introduced in vision by Schade (1956) and was subsequently used by early investigators to measure basic visual sensitivity (Westheimer, 1960; DePalma and Lowry, 1962; Campbell and Robson, 1968). use the print version of this publication as the authoritative version for attribution. Sinusoidal gratings vary in frequency, contrast, and phase. In Figure 1 the left side shows such a grating pattern, and the right side shows the sinusoidal variation in luminance across space. The number of light-dark cycles of the grating that subtend 1 deg visual angle is a measure of the spatial frequency of the grating, expressed in cycles per degree (cpd). The human visual system is able to detect spatial frequencies up to about 60 cpd. There is no lower limit, but gener ally measurements are not made below about 0.1 cpd, often because of practical limits of display size. Borrowing the term octave (a doubling of frequency) from audition, the range of spatial frequencies usually measured by the contrast sensitivity function is about 10 octaves. A low spatial frequency consists of broad black and white bands; a high spatial

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 3 frequency grating has thin black and white bands. Spatial frequency is therefore related to the size of conventional objects. When viewing distance and slant are held constant, higher spatia frequencies correspond to smaller objects. FIGURE 1 Photographs (left) and horizontal luminance profiles (right) of vertical sinusoidal grating patterns, use the print version of this publication as the authoritative version for attribution. electronically generated at three levels of contrast. The contrasts of the gratings were 0.85 (A), 0.50 (B), and 0.10 (C). SOURCE: Photographs taken from transparencies provided by Michael Miller, Department of Neurology, Mount Sinai School of Medicine, and Wolkstein et al., 1980. Reprinted with permission from I. Bodis-Wollner. Copyright 1980 by J. B. Lippincott Company. The contrast of a sinusoidal grating is based on the maximum luminance (Lmax) and the minimum luminance (Lmin) in the grating (see Appendix A). It is a dimensionless variable having values ranging from 0.0 (a uniform field) to 1.0, the maximum possible. The phase of a grating measures its position in space relative to some predetermined reference point.

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 4 The minimum contrast at which a grating can be distinguished from a uniform field with some fixed level of accuracy is the contrast threshold. The reciprocal of threshold contrast is called contrast sensitivity. The contrast sensitivity function is obtained by measuring contrast thresholds over a range of spatial frequencies. A typical photopic contrast sensitivity function is shown in Figure 2. What is important about the contrast sensitivity function seen in Figure 2 is that there is a range of spatial frequencies around 2 to 5 cpd where sensitivity is maximum. Sensitivity falls off for lower spatial frequencies and rapidly falls off for higher spatial frequencies. Eventually a high spatial frequency is reached that requires a contrast of 1.0 to detect (the high frequency cutoff). Spatial frequencies higher than this cutoff frequency cannot be detected by an observer. Relationship Between Acuity and Contrast Sensitivity Visual acuity, because it is measured in terms of the smallest identifiable, high-contrast target, and because small sizes correspond to high spatial frequencies, measures visual sensitivity largely in the higher frequency regions of the contrast sensitivity function. In brief, visual acuity is measured in terms of the size of the critical detail (stroke width of the Snellen letter, for example), but this feature is use the print version of this publication as the authoritative version for attribution. FIGURE 2 Photopic contrast sensitivity function of the human visual system for sinusoidal gratings. Both coordinates are logarithmic. SOURCE: Campbell and Robson, 1968. Reprinted with permission from F. W. Campbell and J. G. Robson. Copyright 1968 by The Physiological Society.

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 5 not the only important one. Snellen acuity letters corresponding to acuity of 1.0 have a height of 5 min arc. The spatial frequencies necessary (but not sufficient) for correct identification after detection of these small letters fall in the approximate range from 18 to 30 cpd (Ginsburg, 1981a). This range of critical spatial frequencies necessary for identification of letters at a visual acuity of 1.0 is shown with the contrast sensitivity function in Figure 3. Does the measurement of sensitivity within this range of spatial frequencies (as with visual acuity) adequately describe the rest of the contrast sensitivity function? Extensive psychophysical data that deal with abnormal contrast sensitivity and individual differences in contrast sensitivity functions, independence of contrast thresholds at different spatial frequencies, and masking and adaptation experiments lead us to conclude that the answer is no. use the print version of this publication as the authoritative version for attribution. FIGURE 3 Photopic contrast sensitivity function showing the range of spatial frequencies necessary to achieve a visual acuity of 1.0 with Snellen letters. SOURCE: Ginsburg, 1981a. Reprinted with permission from A. P. Ginsburg. Copyright 1981 by Cambridge University Press.

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 6 Visual acuity measurements, which are related primarily to high spatial frequency sensitivity, cannot predict contrast sensitivity to low spatial frequencies because thresholds of spatial frequencies separated by more than about a factor of 2 (one octave) are statistically independent of each other (Blakemore and Campbell, 1969; Graham and Nachmias, 1971; Sekuler et al., 1984). This independence of widely separated spatial frequencies is consistent with a model of the visual system containing separate mechanisms, each of which is selectively sensitive to a limited range of spatial frequencies (Campbell and Robson, 1968; Blakemore and Campbell, 1969; Graham and Nachmias, 1971; Stromeyer and Julesz, 1972; Ginsburg, 1984a). The contrast sensitivity function has the potential of adding more information about the functioning of the visual system than that given by visual acuity, because it assesses sensitivity over a wide range of spatial frequencies, while visual acuity measures primarily sensitivity at the high spatial frequencies. A primary source of evidence showing that visual acuity measurements do not characterize the whole contrast sensitivity function comes from clinical studies of people having abnormal visual function. People with identical high spatial frequency sensitivity may have very different low spatial frequency sensitivity (Bodis-Wollner, 1972; Bodis-Wollner and Diamond, 1976; Regan et al., 1981). The clinical applications of contrast sensitivity function are summarized in Proenza et al. (1981). This dissociation between visual acuity and the contrast sensitivity function was first established in patients with cerebral lesions who, although they had visual acuity of 0.5 or better, complained of blurred vision. The contrast sensitivity functions of these patients are shown in Figure 4. A convenient way to illustrate the changes of sensitivity seen in Figure 4 is by plotting deviations from the “normal” contrast sensitivity, as shown in Figure 5. In the upper part of Figure 5 are two contrast sensitivity functions: one for normal observers and one from a patient complaining of reduced vision. The contrast sensitivity function of this patient shows that sensitivity is reduced at all spatial frequencies. The difference between the normal sensitivity curve and that of the patient is plotted in the lower part of Figure 5. This plot of the difference (on a logarithmic scale) between the normal sensitivity and that obtained for an individual is called a visuogram (see Lundh and Arlinger, 1984, for a discussion of three different ways to construct a visuogram). A visuogram is the visual equivalent of an audiogram, used by audiologists and otologists to illustrate deviations from normal auditory sensitivity at different frequencies. The audiogram has proved valuable in identifying different types of deafness and in making diagnoses about their causes (Davis and Silverman, 1960). Perhaps this same value will be realized in vision use the print version of this publication as the authoritative version for attribution.

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 7 once a large enough demographic data base for the contrast sensitivity function becomes available. FIGURE 4 The normal contrast sensitivity function, shown with heavy lines, based on data from 10 observers. The two other curves represent patients having the same visual acuity, but vastly different contrast sensitivity functions. SOURCE: Bodis-Wollner and Diamond, 1976. Reprinted with permission from I. Bodis-Wollner. Copyright 1976 by Oxford University Press. Pathophysiological studies of contrast sensitivity function reveal that loss of sensitivity can occur at all spatial frequencies or at only restricted bands of spatial frequencies (e.g., Bodis-Wollner and Diamond, 1976; Proenza et al., 1981). The major point to be taken from these data is that identical visual acuity may be found in use the print version of this publication as the authoritative version for attribution. people with different contrast sensitivity and that a single measure, such as visual acuity, cannot predict sensitivity at other sizes or spatial frequencies. This dissociation between visual acuity and contrast sensitivity has been found in patients; but what about people with “normal” vision? Figure 6 illustrates three different contrast sensitivity functions from three Air Force pilots having visual acuities of 1.33, 1.00, and 0.80. The visual acuities of the pilots are predicted from the contrast sensitivity in the high spatial frequency range. The higher the high frequency sensitivity, the higher the visual acuity. Note the wide

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 8 variations in low and high frequency sensitivity, and that a low sensitivity at high frequencies does not necessarily imply a low sensitivity at low spatial frequencies. FIGURE 5 Contrast sensitivity function (top) and visuogram (bottom). The upper plot shows two contrast sensitivity functions: the average contrast sensitivity functions of normal subjects with acuity of 1.0 (smooth curve), and a contrast sensitivity of a patient with visual defects (open circles). Contrast sensitivity (ordinate of upper plot) is specified on a logarithmic scale in decibels. The lower plot is a visuogram. It shows the contrast use the print version of this publication as the authoritative version for attribution. sensitivity deficit (downward arrow) at each of the tested spatial frequencies for the patient whose contrast sensitivity is given above. A loss of 6 db signifies a twofold reduction of contrast sensitivity; a loss of 20 db signifies a tenfold reduction. The spatial frequency scale (abscissa) of the visuogram is the same as that of the contrast sensitivity function above it. SOURCE: Wolkstein et al., 1980. Reprinted with permission from I. Bodis-Wollner. Copyright 1980 by J. B. Lippincott Company.

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 9 FIGURE 6 The contrast sensitivity function of three pilots having visual acuities of 1.33, 1.00, and 0.80. Note the variation of sensitivity below 7 cpd. SOURCE: Ginsburg, 1981b. Reprinted with permission from A. P. Ginsburg. Copyright 1981 by the Air Force Aerospace Medical Research Laboratory. Visual Acuity and Contrast Sensitivity Function in Normal Vision use the print version of this publication as the authoritative version for attribution. While visual acuity cannot predict the spatial contrast sensitivity function in people with abnormal vision, visual acuity also cannot predict contrast sensitivity in people with assumed normal vision. There are individual differences in contrast sensitivity as well as changes in the contrast sensitivity function with age (Ginsburg, 1981a, 1984b; Ginsburg et al., 1984; Owsley et al., 1983). These findings raise important questions: Does an observer with 1.33 acuity have a correspondingly higher contrast sensitivity function than one having acuity of 1.0? Or does the first have a slightly wider contrast sensitivity function? Is it better to have a higher peak contrast sensitivity function or a broader one for performing various visual tasks? There is some experimental evidence that under some conditions

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 10 peak contrast sensitivity may be more important than visual acuity for predicting detection and identification of objects (Ginsburg, 1981b). Contrast sensitivity functions are measured with sinusoidal grating patterns. Although contrast sensitivity functions could be measured with a wide assortment of targets differing systematically in size and contrast, the human visual system seems to be especially sensitive to sinusoidal targets (Guth and McNelis, 1969; Watson et al., 1983; Ginsburg, 1984b). These sinusiodal gratings have important mathematical properties (see Appendix A and Appendix C) that allow the application of linear systems analysis to the human visual system. This approach allows both the visual stimulus and the visual system to be described with the same language: that of sinusoidal spatial frequencies. The visual system is not completely linear, however, and there may be other types of visual targets (low contrast letter optotypes, for example) that may be particularly effective in detecting certain types of visual defects (Regan and Neima, 1983, 1984). The working group does not wish to preclude the use of other types of visual targets if their utility can be demonstrated. Contrast Sensitivity Function and Visual Performance Because the image of any object can be described as a set of spatial frequencies at various orientations, amplitudes, and phases (see Appendix C), there is the potential that an observers contrast sensitivity function can be used to predict visual performance with more complex visual material. The working group evaluated the evidence that the contrast sensitivity function can predict visual performance with complex stimuli and found that considerably more research in this area is needed. If the potential of these early experiments is verified by further studies, we believe that we will have a powerful way of studying individual differences in vision and accounting for some of the variability of individual performance on a wide variety of tasks that are primarily dependent on vision. There is some experimental evidence suggesting that the contrast sensitivity function can predict certain types of visual performance better than other measures can. In one series of experiments, subjects were asked to judge the visual similarity between all possible pairs of random complex grating patterns. These studies found that the similarity judgments were accurately predicted by using the spatial frequency content of the gratings in conjunction with the human contrast sensitivity function (Harvey and Gervais, 1978, 1981; Gervais, 1978). In another study, subjects were required to identify letters of the alphabet 6 min arc high presented for 30 msec (Gervais, 1978; Gervais et al., 1984). The researchers tried several models of visual processing to predict the pattern of identification confusions found among all 26 letters of the alphabet. The model with the greatest predictive power was one based on the amplitude and phase information in the spatial frequency content of each letter filtered by the human contrast sensitivity function. Gervais et al. (1984) also provide a critical review of other studies that have failed to find an advantage in spatial frequency models. use the print version of this publication as the authoritative version for attribution.

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 11 The contrast sensitivity function of infants has been used to successfully predict the amount of time that infants spend looking at different types of visual stimuli. Prior to using contrast sensitivity measures, the best predictor of looking time was thought to be the amount of contour in the stimulus, the contour density (Karmel, 1969). One study demonstrated that when stimuli were equated for contour density, infants still preferred some stimuli over others. These preferences were predicted by the spatial frequency characteristics of the stimuli (Banks and Stephens, 1981). Another study showed that when the spatial frequency content of the stimulus patterns were combined with the infant's contrast sensitivity function, both the infant's looking preferences and looking times were better predicted than by the contour density measure (Gayl et al., 1983). Finally, individual differences in contrast sensitivity functions may be the basis of individual differences in performance on complex tasks. Figure 7 shows the contrast sensitivity functions of three observers having visual acuities of 1.00, 0.66, and 0.40 (Ginsburg, 1981b). The two subjects with the acuity below 1.00 required optical correction but were tested without their glasses. These observers had their contrast thresholds measured for the detection and identification of both letters of the alphabet and airplanes of different angular size. The individual differences seen in the detection and identification of letters and planes were predictable from the relevant spatial frequencies of those targets required for detection and identification and the individual contrast sensitivity functions. Note in Figure 7 that the observer having the highest contrast sensitivity in the middle spatial frequencies also was best at the target detection and identification tasks, even though his visual acuity was not the best of the three. A second study of 11 Air Force pilots (Ginsburg et al., 1982) indicated that contrast sensitivity, not visual acuity, predicted simulated air-to-ground target detection. Visual acuity and contrast sensitivity functions were measured under high and low photopic levels of luminance. The correlations between the acuity measures and detection range was not statistically significant. There was a significant correlation (0.83, p < 0.01) between detection range and the peak sensitivity of the contrast sensitivity functions measured at low photopic levels. In a third study (Ginsburg et al., 1983), 84 Air Force pilots took part in field trials that required them to detect an approaching airplane from the ground; 10 sets of field trials were run under widely differing visibility conditions, with about 8 pilots participating in each set. An interesting pattern of correlations emerged between the distance of target detection and the contrast sensitivity at six different spatial frequencies. These correlations tended to be significant for frequencies greater than 8 cycles per degree under high visibility conditions. In contrast, the correlations tended to be significant at 8 cycles per degree and below under low visibility conditions. These results suggest that contrast sensitivity measurements taken over a range of spatial frequencies has potential for predicting detection performance under a variety of visibility conditions. Further studies are needed to determine the full potential for predicting this important kind of performance. use the print version of this publication as the authoritative version for attribution.

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please use the print version of this publication as the authoritative version for attribution. CONTRAST SENSITIVITY FUNCTION Aerospace Medical Research Laboratory. subjects having 0.4 and 0.6 Snellen acuity without their glasses and one subject having uncorrected acuity of 1.0. FIGURE 7 Contrast sensitivity to sine wave gratings, detecting and identifying letter and aircraft silhouettes for two SOURCE: Ginsburg, 1981b. Reprinted with permission from A.P. Ginsburg. Copyright 1981 by the Air Force 12

OCR for page 2
About this PDF file: This new digital representation of the original work has been recomposed from XML files created from the original paper book, not from the original typesetting files. Page breaks are true to the original; line lengths, word breaks, heading styles, and other typesetting-specific formatting, however, cannot be retained, and some typographic errors may have been accidentally inserted. Please CONTRAST SENSITIVITY FUNCTION 13 Suprathreshold Contrast Functions The human visual system responds to suprathreshold grating patterns quite differently than it does to threshold patterns. Although the amount of contrast required for detection of gratings is heavily dependent on the spatial frequency of the grating, this dependency breaks down in the perception of apparent contrast of suprathreshold gratings (Georgeson and Sullivan, 1975; Cannon, 1979; Ginsburg et al., 1980). This effect, called contrast constancy, is found when subjects adjust the physical contrast of different frequency gratings in order to achieve the perception of equal apparent contrast. Particularly at high levels of physical contrast, the contrast required for constant apparent contrast is virtually independent of spatial frequency. There also can be a dissociation between detection sensitivity and spatial frequency discrimination (Began et al., 1982). Following adaptation to a suprathreshold sine wave grating, although detection threshold is greatly elevated at the adaptation spatial frequency, discrimination threshold is unaffected. Discrimination threshold is maximally affected at a frequency about twice that of the adapting frequency (Regan and Beverley, 1983). Studies using visual evoked potentials have also found a dissociation between the response to low contrast gratings and the response to high contrast gratings (Bodis-Wollner et al., 1979). We cite this evidence to point out that extrapolation from threshold measurements of contrast sensitivity to predictions of responses to suprathreshold patterns is theoretically and practically a complicated issue. This complication does not deter the working group from looking at the empirical evidence relating to the usefulness of the contrast sensitivity function in predicting visual processing of complex targets. Sensitivity to Phase The spatial phases of the sinusoidal components of complex patterns are as important as their amplitudes. There is growing evidence that the contrast sensitivity function does not capture all important individual differences, that phase sensitivity must be considered as well. The abnormal vision of some amblyopic patients is not well explained by changes in the contrast sensitivity function (Hess, 1984). These patients report perceptual distortions that seem like phase distortion effects when viewing sinusoidal gratings. The ability to discriminate between gratings containing two harmonically related spatial frequencies that differ only in their phase relationships is markedly reduced in the amblyopic eye. The ability of the human observer to recognize or identify an object is remarkably robust to distortions of the spatial frequency amplitude spectrum of the object but not to a distortion of its phase. A small amount of phase distortion renders the picture unrecognizable. Sensitivity to phase distortions both in photographs of objects and in random checkerboard textures is relatively independent of spatial frequency (Caelli and Bevan, 1982). use the print version of this publication as the authoritative version for attribution.