Contrasting Blue-on-Yellow with White-on-White Visual Fields: Roles of Visual Adaptation for Healthy Peri- or Postmenopausal Women Younger Than 70 Years of Age

Abstract

purpose. To test the hypothesis that differences between short-wavelength automated perimetry (SWAP) and white-on-white visual field sensitivities are related to between-individual variation in the visual adaptation properties of SWS cone pathways.

methods. Twenty-six healthy amenorrheic (peri- or postmenopausal) women not using hormonal medication were tested. Subjects ranged in age from 48 to 68 years. They were tested by using foveal increment–threshold techniques and also with two types of 24-2 visual field tests: a full-threshold SWAP blue-on-yellow (B/Y) test and a white-on-white (W/W) test obtained using a Swedish Interactive Threshold Algorithm (SITA Standard). The age-corrected sensitivity differences between the two types of visual fields were compared against foveal measures of visual sensitivity and adaptation, which were obtained psychophysically using dim and bright yellow backgrounds. All measurements for each subject were made at a single testing session. The comparisons were made for the entire visual field and for separate portions of the visual field. The analyses also included pupil size data obtained during visual field testing.

results. The B/Y minus W/W (B/Y − W/W) mean deviation difference was described (R = 0.80) by a multilinear model with three significant factors: (1) an adaptation factor and (2) a baseline sensitivity factor, each derived from the foveal psychophysical data for short-wavelength test stimuli, and (3) a pupil size factor, as recorded for SWAP. The total deviation differences in the periphery of the visual field (∼22° from fixation) were described (R = 0.87) by a model with four significant factors, the fourth being an “eccentricity factor” describing the rate of change of the B/Y − W/W total deviation difference measured as a function of increasing retinal eccentricity ∼9°–17° from fixation. More than 40% of the variance in the B/Y − W/W mean deviation differences was accounted for either directly or indirectly (via effects of pupil size) by variations in adaptation to the yellow background used for SWAP.

conclusions. Much of the extra variability in SWAP sensitivities for a select group of healthy women can be accounted for by differences in the degree of desensitization induced by the yellow background used for SWAP. For clinical practice, pupil status (dilated or undilated) should be altered only with caution from one SWAP testing session to another.

Short-wavelength automated perimetry (SWAP) has gained fairly wide acceptance as a clinical test since its initial use more than 10 years ago. SWAP uses moderately bright yellow background fields to depress visual sensitivity that is mediated via long- and middle-wavelength-sensitive (LWS and MWS) cones. This selective sensitivity reduction allows visual sensitivity mediated via short-wavelength-sensitive (SWS) cones to be measured in isolation. ¹ Although SWAP can effectively isolate the SWS-cone–mediated response, the same yellow backgrounds that depress LWS- and MWS-cone–mediated responses can also desensitize the SWS-cone–mediated response to some extent. ¹ If the degree of desensitization for SWS-cone–mediated response varies between people, SWAP sensitivities may vary concurrently. In fact, the effects of visual adaptation on SWS-cone–mediated thresholds are known to vary between people, at least at the fovea. ^{2 3 4} For these reasons, it is important to evaluate the degree to which variability in visual adaptation properties may contribute to SWAP variability. 

Although SWAP and W/W visual field scores share many common determinants, there are some that are either unshared or are more important for one type of visual field test than for another. Differences between SWAP and W/W visual field scores will depend on such determinants. The most well known and well documented of these determinants is lens density, ^{5 6} which varies across the population, even for young people, ⁷ and increases with age. ⁸ Because between-person lens density differences are greatest at short wavelengths, ⁷ the importance of lens density for visual field scores is greater for SWAP than for W/W visual fields, even after standardized age corrections. ⁶ However, lens density is probably not the only major factor that contributes more to SWAP variability than to W/W variability. Differential visual adaptation effects may also be important. ⁹  

In particular, the stimulus conditions used for SWAP and W/W perimetry lead to different adaptation effects. Whereas the test/background combinations for W/W perimetry normally result in Weber’s Law threshold behavior (i.e., test threshold proportional to background light level), ¹⁰ the test/background combinations used for SWAP typically do not. ^{1 9} Instead, the rate of threshold elevation with increasing background illuminance is lower than described by Weber’s Law. This non-Weber behavior occurs because the specialized afferent visual pathways that transmit SWS cone signals are desensitized only indirectly, by MWS and LWS cone signals that oppose the SWS cone signal, ^{11 12} but SWS cones themselves are scarcely affected by the yellow SWAP background. ^{1 12} Because SWS-cone–mediated sensitivities depend predominantly on the responses of these distinctive pathways (at least for the incremental, long duration SWAP test stimuli ^{13 14} ), differences between SWAP and W/W visual field scores may depend on visual adaptation properties that are unique for SWS-cone–mediated vision. The present study focused on this hypothesis, which relies on the knowledge that visual adaptation properties for SWS-cone–mediated vision vary between people. ^{2 3 4}  

In addition, the present study sought to develop a model with a minimal number of factors, including visual adaptation, that can account for a large proportion of the difference between SWAP and W/W sensitivity levels across a normal visual field. 

To develop a model with these properties, we modified a three-factor multilinear model that we had developed to account for SWAP sensitivities alone. ¹⁵ This model accounted for much of the variance in the peripheral SWAP visual field sensitivities of women who used the selective estrogen receptor modulator tamoxifen as adjuvant therapy for early-stage breast cancer. ¹⁵ In the modified model, one factor (an “adaptation factor”) represents the degree of SWS-cone–mediated sensitivity loss induced by a yellow background, a second factor (a “baseline sensitivity factor”) represents the level of SWS-cone–mediated sensitivity measured under a minimal degree of yellow background adaptation, and a third factor (an “eccentricity factor”) concerns the rate of sensitivity reduction that occurs with increasing retinal eccentricity, which is relatively steep for SWAP. ^{6 16} Lens density effects would be incorporated mainly into the baseline sensitivity factor. We added a fourth factor, pupil size, to the model because variations in pupil size are expected to compound or interact with strictly neural variations in visual adaptation. That is, under non-Weber conditions such as those used for SWAP, the increase of sensitivity resulting from the extra test light passing through an enlarged pupil will not be entirely offset by a complementary reduction in sensitivity resulting from the extra background light passing through the same pupil, even though the test and background retinal illuminance levels change simultaneously by the same degree. Pupil size is recorded routinely by the latest generation of automated perimeters. 

Because commercial visual field tests are conducted at preset background luminance levels, we measured the baseline and adaptation factors using a custom-built device (a Maxwellian view) that allowed background light levels to be readily manipulated. This same device also allowed us to project the test and background lights onto the retina along a pencil of light smaller than the natural pupil, which eliminated any effects of pupil size variation on retinal illuminance. For the foveal conditions used for testing (see the Methods section), the adaptation factor has a standard deviation of approximately 0.3 log units (3 dB), as calculated for subjects meeting the same eligibility criteria as for this study. ² This degree of variability approaches that obtained at the same test wavelength (440 nm) for lens density measurements made in a more diverse population with a wider age range. ⁵ However, the adaptation factor appears to be independent of age. ^{2 17} In contrast, the baseline sensitivity factor is related inversely to lens density ² and to age. ¹⁷  

The subject population for this study comprised healthy amenorrheic (i.e., peri- or postmenopausal) women not using any hormonally acting medication, such as hormone replacement drugs. These women also served as control subjects for a broader investigation concerning the effects of adjuvant endocrine breast cancer medications on visual function ² and intraocular anatomy. ¹⁸ However, these subjects’ data are important in their own right, as these women were old enough to be at some risk for age-related eye disease, yet they were young enough so that excellent ocular health and visual function were not atypical. Because this subject group was relatively homogenous, extraneous sources of variability may have been reduced; and because these subjects were amenorrheic, any effects of cyclic hormonal change on SWS-cone–mediated sensitivities ^{19 20} were minimized. The Discussion section addresses the generality of the results. 

Because short-wavelength sensitivities—SWS-cone–mediated sensitivities, in particular—decrease with age, ^{21 22 23} clinical evaluations typically emphasize the need to employ age corrections that reduce visual field variability that is not due to disease. In fact, such age corrections are incorporated into many of the statistics that are output by SWAP and by W/W testing instruments. ²⁴ In particular, age corrections are incorporated into the “total deviation” statistic, which quantifies a person’s departure from an age-corrected population average at each separate locus in the visual field. Age corrections are similarly incorporated into the “mean deviation,” which is a corresponding summary statistic for the entire visual field. Over the age range for this study, the age correction for the SWAP mean deviation is approximately 2 dB per decade, approximately three times what it is for a W/W field. However, not all of this change is due to an age-related increase in lens density. Neural response capabilities can change with age, ^{25 26} and pupil size tends to change with age, becoming smaller for any given ambient illumination level. ²⁷ For all these reasons, the analyses involving visual field data for this study were conducted by using the age-corrected visual field statistics. However, these age corrections were no doubt derived from a mixed-gender population that included women who were premenopausal or using hormone replacement, and the age corrections for SWAP and for W/W visual fields were not obtained from a single group of people. Thus, we repeated the analyses for some of the main results in our study using visual field data that were not age-adjusted. 

Methods

Subjects

Subjects were eligible for this study if they were healthy women 40 to 69 years old, not using hormonal medication, and amenorrheic for at least 6 months. All subjects met the following eligibility criteria for excellent ocular health: (1) 20/20 or better visual acuity in one eye and 20/25 or better visual acuity in the other, (2) no evidence or suspicion of eye disease on undilated direct ophthalmoscopic examination and on evaluation of stereoscopic color fundus and optic nerve head (ONH) photographs obtained with pupil dilation at the end of the testing session, (3) no history of eye disease or ocular hypertension, (4) no diabetes, (5) intraocular pressure (IOP) ≤22 mm Hg on Goldmann applanation and no between-eye IOP difference more than 2 mm Hg, (6) no myopia more than 5 D, (7) no use of any medication known to affect vision, (8) no history of ocular surgery, and (9) normal color vision (i.e., no more than a single minor transposition error on the D-15 test performed under standard illuminant C, ²⁸ which is the light provided by Macbeth Easel Lamp illumination [no longer manufactured]). 

Twenty-seven subjects met these eligibility criteria. However, one subject was excluded because of abnormal visual fields, as her pattern standard deviations were high, 7.33 dB for a W/W Swedish Interactive Threshold Algorithm (SITA) Standard 24-2 field and 6.18 dB for a full-threshold SWAP 24-2 field. The pattern standard deviations for the other 26 subjects averaged 1.58 ± 0.36 dB (SD) for the W/W field and 2.96 ± 0.79 dB for the SWAP field. There were no other outliers. The mean deviations for these 26 subjects averaged −1.44 ± 1.53 dB for the W/W field and −2.31 ± 2.70) dB for the SWAP field. The slightly negative mean deviations and lower SWAP mean deviation may have reflected the fact that many or all of these subjects were unpracticed at visual field testing, ⁶ as all subjects had medical histories that were negative for eye disease or suspicion of eye disease. 

For the 26 subjects whose data are reported in the Results section, the average age was 57.2 ± 5.0 years (range, 48–68). For reference, the median age of onset of menopause in this society is 51 years. ²⁹ All subjects were unpaid volunteers who, in most cases, responded to electronic or print ads stating the requirement of good reading vision, with corrective lenses if necessary. Several subjects were referred by previous participants. 

Procedures

Subject Screening and Training.

The screening procedures were conducted as just described and as described in greater detail elsewhere. ^{15 18} The Pelli-Robson spatial contrast-sensitivity test was administered before all visual field and psychophysical testing. This test is useful for teaching subjects to respond to threshold-level stimuli that may not be easily visible. It was not used for screening purposes. 

Written informed consent was obtained from all subjects after explanation of the nature and possible consequences of the study. All portions of the study adhered to the tenets of the Declaration of Helsinki and were approved by the Oregon Health and Science University Institutional Review Board. 

Visual Field Testing.

Visual fields were measured for one eye only, designated as the test eye for all postscreening visual function testing. The test eye was selected using the following criteria, applied in order as necessary: (1) the eye with the better best-corrected acuity, even if by just one letter, (2) the eye with the lesser degree of spherical equivalent refractive error, and (3) subject preference. 

All visual field testing was conducted with a single perimeter (Humphrey Field Analyzer II model 750i; Carl Zeiss Meditec, Dublin, CA) that is serviced yearly and has always met specifications. We administered the SITA Standard W/W visual field test first and the full-threshold SWAP test second, both according to conventional procedures (e.g., with optical correction, with size III spots [0.43°] for W/W fields and size V spots [1.72°] for SWAP, and with stimulus durations of 200 ms ²⁴ ). Each visual field test was administered with a 24-2 test pattern. Subjects rested between the two visual field tests for approximately 5 minutes. For reference, the background luminance level for SWAP is 100 cd/m² and the background luminance level for W/W perimetry is 31.5 apostilbs (10 cd/m²). ²⁴ The yellow background used for SWAP contains all visible wavelengths greater than approximately 530 nm, and the short-wavelength test stimulus is composed of light having a peak transmission at 440 nm, with a half-height bandwidth of 27 nm. ⁶ The SWAP stimulus combinations ensure approximately 13 dB of SWS-cone–mediated isolation at the fovea and approximately 9 dB at 20° eccentricity. ⁶  

Threshold data were recorded by using the StatPac (Carl Zeiss Meditec) output for the two types of visual fields, and pupil size was recorded automatically, typically after approximately 3 minutes of viewing the respective backgrounds for the two types of visual fields. Pupils were not pharmacologically altered for the visual field tests or for any visual function testing. 

Maxwellian View Psychophysical Testing.

After subjects rested, Maxwellian view testing was initiated for the same eye in which visual field data were collected. The device and its calibration procedures have been described previously. ³⁰ The device corrected internally for each subject’s spherical equivalent refractive error. It had an exit pupil of 1.2 mm, smaller than any subject’s natural pupil. In contrast to the visual field stimuli, all the stimuli for Maxwellian view testing were created using narrow-band interference filters, with half-height bandwidths of approximately 10 nm. In the present study, an eye-piece/chin-rest assembly facilitated proper head position. 

All background stimuli were 11° diameter disks. Subjects were instructed to fixate the center of these disks, aided by diagonal cross hairs with a 4° central gap. Test stimuli were 3° diameter disks centered inside the background cross hairs and square-wave modulated at 1.5 Hz with a 50% duty cycle. The test stimuli were presented continuously. 

Thresholds were obtained by using a method of ascending limits, in which the subject signaled when the test stimulus first became visible. All threshold data for this article were obtained for steady state adaptation conditions. The incremental step size was 0.06 log unit, and threshold was computed as the mean of four responses for each test stimulus–background combination. 

Steady state thresholds were measured for each of three 580-nm background fields in the following order: 2.0, 1.6, and 3.6 log td. SWS-cone–mediated thresholds were measured at 440 nm, and verification of SWS-cone isolation at 440 nm was made by comparing 440-nm thresholds with 490-nm thresholds. ³ The 440-nm thresholds of every subject were verified to represent SWS-cone–mediated response, as 440-nm sensitivities exceeded 490-mm sensitivities for each background. (In fact, the 1.6-log td background illuminance was chosen on the basis of being as dim as possible while still allowing routine proof of SWS cone isolation for these foveal stimuli.) Thresholds not mediated by SWS cones were measured for 580-nm test stimuli at all background levels. Thresholds for 510-nm test stimuli were obtained for the 3.6-log td background, to verify isolation of SWS-cone–mediated response at 490 nm, which, in turn, ensures an additional 1.3 log units of isolation at 440 nm for this background. ³ Threshold measurements were made after subjects viewed each background for at least 4 minutes. There was one exception; for the 1.6-log td background, an initial set of 440-nm thresholds was obtained beginning at 3 minutes. Data for this initial set of 440-nm thresholds are reported once, where their use is identified in the Results section. 

Data Analyses and Definitions

Visual Field Data.

Mean deviations and total deviations were obtained from the StatPac output. These data represent the sensitivity departure from an average age-corrected norm for the entire field (the mean deviation) or for each point in the visual field (the total deviation). Visual field data are expressed in decibel units (1 dB = 0.1 log unit). 

For analysis purposes, the visual field data were subdivided in two ways. First, the average total deviations were derived for each of four rings, defined on the basis of their distance from fixation. Ring 1 consisted of the four visual field loci with x and y coordinates at 3° (distance from fixation = 4.2°); ring 2 consisted of the 12 loci with coordinates at 9° and 3° or at 9° and 9° (distance from fixation = 9.5° or 13.8°); ring 3 consisted of the 12 loci with coordinates at 15° and 3°, or at 15° and 9° (distance from fixation = 15.3° or 17.5°); and ring 4 consisted of the 20 loci with coordinates at 15° and 15°, at 21° and 3°, or at 21° and 9° (distance from fixation = 21.2°, 21.2°, or 22.8°). The positions in the visual field that corresponded to the blind spot and to its mirror image across the vertical meridian were omitted from the calculations, as were the two points at approximately 27° eccentricity. This procedure for subdividing the visual field on the basis of eccentricity is identical with the procedure we used previously (but had not described entirely correctly) for analyzing the SWAP data alone. ¹⁵ The outermost ring is considered to represent the peripheral portion of the visual field. 

Second, the average total deviation was computed separately for each of the four visual field quadrants: superior nasal, superior temporal, inferior nasal, and inferior temporal. We used the same set of data points used for computing the average total deviations for the rings. 

The “eccentricity factor” is defined as the average reduction in the total deviation difference for SWAP versus W/W visual fields from ring 2 to ring 3. A positive value represents a steeper falloff with increasing eccentricity for SWAP fields than for W/W fields. 

Psychophysical Data.

The “adaptation factor,” representing the degree of SWS-cone–mediated sensitivity loss induced by a yellow background, is defined as the difference in log thresholds at 440 nm for a 3.6- versus 1.6-log td 580-nm background. A positive value represents a threshold increase (i.e., a sensitivity decrease) from the dimmer to the brighter background. The “baseline sensitivity,” representing the level of SWS-cone–mediated sensitivity measured under minimal adaptation, is defined as log sensitivity at 440 nm for the 1.6-log td 580-nm background. 

Because the 490-nm sensitivities for the 3.6-log td background exceeded the corresponding 510-mm sensitivities for every subject, we concluded that the SWS-cone–mediated response was isolated at 490 nm for the 3.6-log td background for every subject. Because the 490-nm sensitivities were mediated via SWS cones, we were able to estimate lens density. ^{2 22} Estimating lens density (actually a constant plus a value proportional to lens density) is possible because the log sensitivity difference from 440 to 490 nm depends on the sensitivity profile of only a single class of photoreceptors (the SWS cones) and because macular pigment optical density is very nearly the same at 440 and 490 nm. ³¹  

Statistical Analyses.

All P-values for the results are two-sided, and all statistical relations were assessed parametrically (e.g., the correlation coefficients all represent Pearson correlations, and comparisons of central tendency are made using t-tests. Multilinear regression techniques were used for examining associations within the data. We conducted all statistical analyses with commercial software (SYSTAT ver. 10.2; SYSTAT, Richmond, CA). 

Nomenclature.

B/Y refers to results obtained using SWAP (i.e., blue-on-yellow) visual fields, W/W refers to results obtained using white-on-white visual fields, and the term B/Y − W/W refers to B/Y data minus W/W data. Unless stated otherwise, the pupil sizes for all data analyses corresponded to the pupil size measurements made for SWAP rather than for the W/W visual field. 

Results

Consistent with previous results, ^{2 17} the adaptation factor did not correlate with age (r = −0.01) for the subjects in this study. As predicted, the adaptation factor correlated with the difference between the B/Y and W/W mean deviations (r = −0.59, P = 0.002). In fact, the relation of the adaptation factor to the B/Y mean deviation itself was significant (r = −0.47, P = 0.015). The relation of the adaptation factor to the B/Y − W/W mean deviation difference is graphed in Figure 1 , and the relations of the adaptation factor to the average B/Y − W/W total deviation differences for each of the four visual field eccentricity rings are graphed in Figure 2 . Table 1lists the correlation coefficients and probabilities for all five graphs, and it includes the corresponding data for the four visual field quadrants. None of the corresponding regression line slopes (not shown) differed significantly from one another. 

The adaptation factor correlated significantly with the B/Y − W/W total deviation difference for every part of the visual field (Table 1) , even though the B/Y − W/W total deviation differences depended on retinal eccentricity (Fig. 2) . The strongest effect of retinal eccentricity seemed to involve the variability of the B/Y − W/W total deviation differences, which was greatest in the periphery, with the variance of the B/Y − W/W differences in the outermost ring being greater than the corresponding variance in the innermost ring (ratio of variances = 2.77; P < 0.02 based on an F test with ν₁ = ν₂ = 25). In addition, the B/Y total deviations tended to be slightly lower than the W/W total deviations near the center of the visual field (mean difference = 1.4 dB for ring 1; P < 0.001) but not in the periphery (mean difference = 0.2 dB for ring 4; P = 0.722). The change in the B/Y − W/W differences from ring 1 to ring 4 was nominally significant (paired t-test, P = 0.026). 

Developing a Four-Factor Model to Account for B/Y − W/W Differences in the Visual Field Periphery

The adaptation factor is not the only factor that distinguishes B/Y from W/W visual field sensitivities. For example, the reduction of sensitivity with increasing retinal eccentricity is known to be greater for B/Y visual fields than for W/W visual fields, ^{6 16} and so individual differences in eccentricity-dependent sensitivity changes are expected to contribute to between-subject B/Y − W/W sensitivity differences in the periphery. These effects can be captured by the eccentricity factor, which is derived from the data for the middle 2 rings and which is the only factor derived from the visual field sensitivities themselves. 

Two additional factors are hypothesized to be important. The first concerns the baseline level of SWS-cone–mediated sensitivity, which ideally would be the level of SWS-cone–mediated sensitivity in the absence of any background-induced desensitization. We approximated this factor using the sensitivity level measured for a relatively dim yellow background, which was 1.6 log td in this study. The average 440-nm sensitivity difference for 1.6 versus 2.0 log td 580-nm backgrounds was 0.01 log unit, supporting the use of this approximation. Whereas this operationally defined baseline sensitivity factor probably does not represent the SWS-cone–mediated sensitivity level that would exist for every subject in the absence of background-induced desensitization, this factor probably does adequately represent the lowest sensitivity level attainable for the same SWS-cone mechanism (Stiles’ π₁) that is responsible for detecting test stimulus increments on the 3.6-log td background. ¹² (In some subjects, SWS-cone–mediated thresholds at very low background levels might be detected via Stiles’ π₂ mechanism. ^{4 12} ) In either case, the baseline level of SWS-cone–mediated sensitivity, as defined operationally, is expected to be more important than the baseline sensitivity level for W/W stimulus conditions (which we did not measure), since Weber’s Law negates the effects of factors that alter the response to white background stimuli and to white test stimuli by the same degree. 

The second additional factor (i.e., the fourth factor in the model) is pupil size. As stated in the introduction, an increase in pupil size is expected to result in a net sensitivity gain for SWAP compared with W/W visual fields, since the visual sensitivities for W/W fields more closely obey ideal Weber’s Law behavior. For the subjects in this study, SWAP pupil diameters ranged from 3.4 to 6.5 mm, with a mean of 4.36 ± 0.81 (SD) mm. 

Because retinal illumination is proportional to pupil area (A), and because visual field sensitivities and light adaptation effects are measured in logarithmic units, we initially planned to use log(A) units when incorporating pupil size into our four-factor model. Because log(A) equals a constant + 2log(D), where D signifies pupil diameter, multilinear models could equivalently use log(D) units. However, for the range of pupil sizes recorded, the correlation between D and log(D) was very high (r = 0.995) and the residuals for only one subject (with the largest pupil) exceeded 0.1 mm when D was regarded as a function of log(D). Thus, to make the pupil size term for any multilinear model intuitive, we computed the coefficients for such models by using D as the index of pupil size. Individual differences in retinal directional sensitivities ³² and SWAP Weber’s Law departures, although unknown, would outweigh any effects of the D:log(D) substitution. This substitution raised the multiple correlation coefficients (R) by 0.01 unit for most of the multilinear models reported in the remainder of the Results section. 

The resultant four-factor model (using D as the pupil size factor) led to high multiple correlation coefficients, both for the overall B/Y − W/W mean deviation difference (R = 0.88) and for the B/Y − W/W total deviation difference at the outer ring representing the visual field periphery (R = 0.87). Because peripheral total deviations, unlike mean deviations, are computed independently of the data from the midperiphery, we assessed the significance of the model for the outer ring data only. Table 2gives the significance levels and model coefficients. All four factors were statistically significant, with P < 0.01 for each factor. Figure 3shows the fit of the model to the data. 

Had we used raw (i.e., non–age-corrected) visual field scores, the fit of the model to the B/Y − W/W peripheral differences would have improved marginally. The value of R would have increased from 0.87 to 0.89, and the P-values for each of the four component factors would have strengthened. (Conversely, had we used raw scores rather than age-corrected scores for Table 1 , the individual correlations with the adaptation factor would have weakened slightly, presumably because adaptation is independent of age.) Adding the lens density estimate as a fifth factor would have improved neither the age-corrected nor the raw-data model. For the age-corrected model, the lens density P = 0.61, whereas for the raw-data model, P = 0.40. 

The correlations among the adaptation, baseline sensitivity, eccentricity, and pupil-size factors were low, with only one of the pair-wise correlation coefficients exceeding 0.2 in magnitude. The exception was the adaptation/baseline–sensitivity pair, for which the r = 0.24. Although this correlation was not significant (P = 0.223), it nevertheless may have been influenced by the presence of measurement noise shared by the baseline sensitivity and adaptation factors, because each depended on a common set of thresholds measured for the 1.6-log td background. Had we instead used an alternative baseline sensitivity factor derived from the threshold measurements taken shortly after 3 minutes of adaptation to the 1.6-log td background, the correlation of the adaptation factor with the baseline sensitivity factor would have dropped from r = 0.24 to r = 0.02, but the model would have been unaffected, indicating that adaptation and baseline sensitivity are functionally independent. This independence provides further evidence that a 1.6-log td background can be used for generating an appropriate baseline sensitivity. 

Using a Three-Factor Model to Account for the B/Y − W/W Mean Deviation Difference

A reduced three-factor model employing no visual field sensitivity data accounted for more than half the variance in the B/Y − W/W mean deviation difference. The three-factor model with the eccentricity factor omitted yielded a multiple correlation coefficient of R = 0.80, with the three remaining factors—the adaptation factor, the baseline sensitivity factor, and the pupil size factor—all being statistically significant. Figure 4shows the fit of the model to the data, and Table 3gives significance levels and model coefficients. For the B/Y − W/W sensitivity difference in the periphery, the comparable three-factor model was less effective (R = 0.72), which is expected since sensitivity in the periphery depends greatly on the steepness of the hill of vision. The remaining three factors were statistically significant. 

Because the adaptation factor and the baseline sensitivity factor together completely determine foveal SWS-cone–mediated sensitivity for the bright (3.6 log td) yellow background, two-factor models with only pupil size and the bright-background foveal sensitivity as factors may approach the effectiveness of the corresponding three-factor models. In fact, for the B/Y − W/W mean deviation difference, the two-factor R equaled 0.78 (a decrease of 0.02 from the three-factor R), whereas it equaled 0.71 (a decrease of 0.01) for the B/Y − W/W peripheral-sensitivity difference. Moreover, the pupil size coefficients changed very little from the three-factor to the two-factor models. However, these two-factor models are incomplete, in that they cannot distinguish baseline sensitivity from adaptation. 

Adding age as a factor did not improve either of the models represented in Figure 3 or 4 . Similarly, adding lens density as a separate factor scarcely affected either model. 

Relating Adaptation and Pupil Size to the B/Y − W/W Mean Deviation Difference

Because the adaptation factor was defined using Maxwellian view data only, it was computed independently of any variation in pupil size. However, because pupil size affects background light levels for visual field testing and since the adaptation properties pertaining to SWAP visual fields differ from the adaptation properties pertaining to W/W visual fields, ^{1 9 10} we would expect the overall effect of adaptation on B/Y − W/W mean deviation differences to depend on afferent pathway (e.g., retinal) visual processes combined with pupillary processes. We found that the two-factor model with the adaptation factor and pupil size factor as independent variables accounted for 46% of the variance of the B/Y − W/W mean deviation difference (R = 0.68), with both factors significant. Table 4gives the statistics for this reduced (and incomplete) model. For the B/Y mean deviation data alone, the two-factor model would have accounted for 29% of the variance, whereas for the W/W mean deviation data alone, the corresponding two-factor model (using W/W pupil size and an adaptation factor derived from the 580-nm test stimuli) would have accounted for only 3% of the variance. 

Finally, we considered how pupil size itself is related to the B/Y − W/W mean deviation difference, independent of any model. By itself, pupil diameter was correlated with the B/Y − W/W mean deviation difference (r = 0.43, P = 0.028), indicating that effects of between-subject differences in pupil size remain even after visual field norms have been corrected implicitly for the tendency of pupil size to decrease with age. ²⁷ For subjects with pupil diameters larger than the median (4.35 mm), the B/Y − W/W mean deviation differences averaged 2.4 dB higher than they did for subjects with pupil diameters smaller than the median (nominal P = 0.003). Figure 5left shows the relation between pupil size and mean deviation for SWAP fields and their associated pupil diameters, and Figure 5right shows the corresponding relation for W/W fields and their associated pupil diameters. Whereas a modest pupil size effect may exist for W/W fields, there seems to be a more substantial pupil size effect for SWAP fields, an impression borne out by the analyses involving the B/Y − W/W difference data. 

We conclude by noting that since the pupil diameters for SWAP ranged from 3.4 to 6.5 mm, the retinal illuminances of the SWAP backgrounds ranged from 2.96 to 3.52 log td. The SWAP retinal illuminance for a 2-mm diameter pupil would be only 2.50 log td. 

Discussion

A drawback of SWAP is its relatively high degree of between-person and within-person variability compared with W/W perimetry. ^{6 33} By testing a relatively homogenous population and by analyzing differences between B/Y and W/W visual fields, we sought to identify some of the factors other than lens density that might contribute to SWAP variability. Using Maxwellian view psychophysics to test at the fovea, we were able to measure effects of visual adaptation, independent of any pupil size changes. We then compared these foveal adaptation data with visual field data obtained at the same testing session. Based on these comparisons and also by incorporating the effects of pupil size, we showed that in the population tested, approximately 40% to 45% of the variance in the difference between SWAP and W/W mean deviations was related directly or indirectly to variations in the ability of the visual system to adapt to the yellow backgrounds used for SWAP, even among untrained subjects. The proportion of variance attributable to adaptation effects may be even higher if the change of visual sensitivity with increasing eccentricity depended partly on eccentricity-dependent changes in visual adaptation. 

Generality of Foveally Based Models for Describing the Visual Field Data of Healthy Eyes

On a priori grounds, SWS-cone–mediated sensitivity at any individual retinal locus must depend on the ability of an adapting stimulus to change sensitivity from the baseline level existing in the absence of any adaptation-induced sensitivity change. (For SWAP, the adaptation stimulus is a 100 cd/m² yellow background.) However, measuring adapted and nonadapted SWS-cone–mediated sensitivity levels across the entire visual field—essentially obtaining a point-by-point adaptation factor and baseline sensitivity factor—would be difficult, if not impossible, considering the added need to verify isolation of SWS-cone–mediated response at baseline levels. One practical way to address this problem is to model the visual field data by using adaptation and sensitivity measurements from a limited number of loci. In our study, the models relied on adaptation and baseline sensitivity measurements made at just one locus, the fovea. The fovea defines the center of the visual field, it is the locus most amenable to psychophysical testing, and it is the locus with the least risk of rod intrusion into sensitivity measurements made at dim background levels. By assessing the degree to which foveally based models are able to account for extrafoveal sensitivity data, we can increase our understanding of the processes that determine sensitivities across the visual field. Although foveally based models were productive for our unique subject population, it is important to identify the limitations and to consider the generality of these models. 

Because we did not obtain between-session repeated measures for either the visual field data or the psychophysical data, we are unable to construct a general model that distinguishes between-person from within-person variability and that specifies the sources of these distinctions. Although our subjects were amenorrheic (and probably almost all were postmenopausal), which would reduce or eliminate any menstrual cycle effects on within-person variability, ^{19 20} we cannot rule out the possibility that individual differences in hormonal exposure influenced the data. ^{34 35 36 37} To reduce the ability of individual differences in macular pigment densities ^{23 38} to alter the baseline sensitivity factor, it would be useful to obtain baseline sensitivity data outside the fovea. Only by measuring full threshold-versus-illuminance curves at two or more extrafoveal test loci would it be possible to quantify the extent to which eccentricity-dependent reductions of sensitivity (and hence the eccentricity factor) depend on reductions of baseline sensitivity, changes of adaptation, or both. Presently, there is only limited information concerning these types of comparisons. ³⁹  

The extent to which the results in this study would apply to other healthy populations is unknown. It is possible, for example, that adaptation properties would depend on retinal locus in a way that would invalidate or compromise the use of foveally based models for any of the following populations: men, young women (for whom there is cyclic hormonal change), postmenopausal women using hormone replacement, or the elderly (for whom the hill of vision steepens ⁶ ). The visual adaptation properties of young women need to be compared for foveal versus peripheral vision as a function of phase of menstrual cycle. ^{19 20} This comparison can be achieved in Maxwellian view. 

Implications for Clinical Visual Field Testing

As predicted, the difference between SWAP and W/W sensitivities correlated with pupil size: the larger the pupil for SWAP testing, the greater the SWAP sensitivities relative to the W/W sensitivities. The effect of pupil size probably contributes more to between-subject SWAP differences than to natural within-subject SWAP changes, given that the standard deviation of SWAP pupil size for this study was approximately three times larger than an average between-session standard deviation reported in individual subjects. ⁴⁰ Nevertheless, within-session variation in pupil size would be capable of raising the short-term fluctuation. The effect on short-term fluctuation probably would be greatest for elderly people, since they tend to have the smallest pupils and densest lenses and consequently the lowest retinal light absorptions. In addition, age-related degenerative changes may reduce the ability of cone photoreceptors to absorb light, ^{41 42} Low retinal light absorption would be expected to result in the greatest departures from Weber’s Law, ^{1 12} which, in turn, would further amplify effects of variations or oscillations in pupil size. 

For clinical purposes, it would be useful to assess the ability of pupil dilation to affect SWAP visual field scores, since an increase of pupil diameter from 4 to 8 mm can be calculated to lead to a substantial SWAP sensitivity increase, up to approximately 6 dB in the most extreme cases. Moreover, the sensitivity increase may not be the same for each portion of the visual field. For this study, the B/Y − W/W mean deviation difference increased by approximately 1 dB for every 1-mm increase in pupil diameter from approximately 3.5 to 6.5 mm (Table 3) . However, there are three reasons for expecting a slower rate of change at the largest pupil areas: (1) cone photoreceptors typically are least sensitive to light entering near the dilated pupil margin, ³² (2) a given increase in pupil diameter causes the pupil area to be enlarged disproportionately less for a large pupil than for a small pupil, and (3) departures from Weber’s Law for SWS-cone–mediated vision become progressively less as retinal illuminance increases. ^{1 12} The small magnitude of the adaptation factor for several subjects (Fig. 1)suggests that the yellow background used for SWAP sometimes raises individual subjects’ SWS-cone–mediated thresholds above their baseline levels little or not at all. Such individuals would be among those in whom the effects of pupil size on SWAP sensitivities are the greatest. 

Relations to Previous Studies and Suggestions for Future Studies

Of all the physiologic and/or anatomic factors that affect SWAP sensitivities, pupil size is the only one that can be manipulated using reversible, interventional means. That is, pupils can be pharmacologically dilated or constricted and then later return completely to normal, thereby allowing effects of pupil size on SWAP sensitivities to be assessed for causality. Furthermore, these effects can be assessed point by point across the visual field. These considerations suggest several experiments involving SWAP visual field sensitivities, which are expected to be immune to any modest refractive errors that are pharmacologically induced. ⁴³ We would predict pupil constriction to affect SWAP scores more than would pupil dilation, because of directional asymmetries inherent in the three relations enumerated in the previous paragraph. 

Felius and Swanson ⁹ showed how within-person and between-person threshold variability can be relatively great under yellow-background stimulus conditions similar to those used for SWAP. As we did, they assessed effects of visual adaptation by using foveal Maxwellian view psychophysics, which eliminates pupil size as a source of variability, but they did not compare their visual adaptation data with visual field data obtained for the same subjects. Nevertheless, they showed how the non-Weber adaptation conditions used for SWAP result in relatively high-threshold variability, and their analyses are pertinent for our results that compare SWAP visual fields (with non-Weber adaptation properties) against W/W visual fields (with Weber adaptation properties). 

The most unexpected result of our study concerned the tendency of the SWAP total deviations to be slightly less (∼1.4 dB) than the W/W total deviations in the innermost portion of the visual field. Because this tendency vanished toward the periphery, the slope of the hill of vision in either the SWAP or W/W fields, or both, probably differed slightly from the corresponding slopes for those subjects on whose data the visual field norms were based. There is not enough information to assign a cause, but the direction of the effect is opposite to that expected for a hypothetical practice effect. It is possible that the average slope of the hill of vision for SWS-cone–mediated response for postmenopausal women not using hormone replacement differs slightly from the average slope for other populations. This possibility is based on data showing that SWS-cone–mediated visual response at the fovea can be related to hormonal status ^{2 19} or to gender. ^{3 44} Alternatively, the tendency of the SWAP total deviations to be slightly lower than the W/W total deviations for the innermost portion of the visual field may have represented a byproduct resulting from the use of separate populations to establish SWAP norms and W/W SITA norms. Distinguishing between the two alternatives may be accomplished by testing healthy men prospectively. 

Because all the subjects in this study appeared to be free of eye disease and because the degree of background-induced desensitization measured at the fovea was correlated with SWAP sensitivities across the entire visual field, any ability of visual adaptation to affect the SWAP visual fields of the people in this study appeared to be largely global. However, diseases such as glaucoma may affect SWAP by causing visual adaptation changes that are localized, regardless of the subject population. Eisner et al. ³ showed that the adaptation properties of SWS cone pathways, as assessed at the fovea, can be affected early in a glaucomatous disease process. Eisner et al. ⁴⁵ and McKendrick et al. ⁴⁶ each have provided evidence of other types of glaucoma-related visual adaptation changes. Eisner et al. ⁴⁵ found flicker response abnormalities at the fovea for a class of subjects having minimal field loss overall, while McKendrick et al. ⁴⁶ found contrast response abnormalities in relatively normal areas of the peripheral visual field. Thus, there is reason to assess visual adaptation at extrafoveal sites in the visual field, especially at early stages of the glaucomatous disease process, when functional alterations may precede ganglion cell death. ^{3 45 46 47 48 49} Part of the challenge for future studies of SWS-cone–mediated response is to assess visual adaptation at visual field loci that are tailored for individual patients, while controlling for pupil size. 

 Supported by National Eye Institute Grant EY014594 and by an unrestricted grant from Research to Prevent Blindness.

 Submitted for publication December 17, 2005; revised March 27 and August 1, 2006; accepted October 10, 2006.

 Disclosure: A. Eisne r, None; M.D. Toomey, None; L.J. Incognito, None; J.P. O’Malley, None; J. R. Samples, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

The difference between SWAP (B/Y) and W/W mean deviations plotted versus the adaptation factor for SWS-cone–mediated response. The adaptation factor is defined as the threshold increase caused by raising the 580-nm steady state background illuminance from 1.6 to 3.6 log td for a 440-nm test.

Figure 2.

 

The difference between SWAP (B/Y) and W/W) total deviations averaged separately in each of four rings of progressively greater retinal eccentricities plotted versus the adaptation factor for the SWS-cone–mediated response. Ring 1 corresponds to the parafovea; rings 2 and 3 to the inner and outer portions, respectively, of the midperiphery; and ring 4 to the periphery. See the Visual Field Data portion of the Data Analyses and Definitions section of the Methods for the specifications of the rings. Other details are as in Figure 1 .

Table 1.

 

Relation of the B/Y − W/W Sensitivity Differences to the Adaptation Factor for Various Portions of the Visual Field

Area of the Field	Pearson r	P
Mean deviation	−0.59	=0.002
Ring 1 (parafovea)	−0.65	< 0.001
Ring 2 (inner midperiphery)	−0.64	< 0.001
Ring 3 (outer midperiphery)	−0.56	=0.003
Ring 4 (periphery)	−0.51	=0.007
Superior nasal quadrant	−0.57	=0.002
Superior temporal quadrant	−0.59	=0.001
Inferior nasal quadrant	−0.52	=0.006
Inferior temporal quadrant	−0.40	=0.046

  n = 26.

Table 2.

 

Statistics for the Four-Factor Model of the B/Y − W/W Average Total Deviation Difference in the Periphery

Factors	Model Coefficient	Standardized Coefficient	P
Constant	−29.48	—	0.0050
Adaptation (log units)	−6.15	−0.45	0.0008
Baseline sensitivity [log(deg2/μW)]	4.53	−0.32	0.0087
Eccentricity (dB)	−1.10	−0.51	0.0002
Pupil size (mm)	1.29	0.34	0.0054

  n = 26, R = 0.87, R ² = 0.76, adjusted R ² = 0.71, P < 0.0001.

Figure 3.

 

The difference between the average SWAP (B/Y) and average W/W total deviations in the periphery (ring 4) plotted versus the values output by the four-factor multilinear model. The four factors are the adaptation factor, baseline sensitivity factor, eccentricity factor, and pupil-size factor. See Table 2for the model coefficients.

Figure 4.

 

The difference between SWAP (B/Y) and W/W mean deviations plotted versus the values output by the three-factor multilinear model. The three factors are the adaptation factor, baseline sensitivity factor, and pupil-size factor. See Table 3for the model coefficients.

Table 3.

 

Statistics for the Three-Factor Model of the B/Y − W/W Mean Deviation Difference

Factors	Model Coefficient	Standardized Coefficient	P
Constant	−28.00	—	0.0011
Adaptation (log units)	−6.33	−0.64	0.0001
Baseline sensitivity [log(deg2/μW)]	4.41	0.43	0.0035
Pupil size (mm)	0.99	0.36	0.0116

  n = 26, R = 0.80, R ² = 0.64, adjusted R ² = 0.59, P < 0.0001.

Table 4.

 

Statistics for a Two-Factor (Incomplete) Model of the B/Y − W/W Mean Deviation Difference

Factors	Model Coefficient	Standardized Coefficient	P
Constant	−2.19	—	0.3269
Adaptation (log units)	−5.31	−0.53	0.0023
Pupil size (mm)	0.95	0.34	0.0370

  n = 26, R = 0.68, R ² = 0.46, adjusted R ² = 0.41, P = 0.0008.

Figure 5.

 

Graphs showing the relations between pupil diameter and mean deviations for SWAP (B/Y) fields and their associated pupil diameters (left), and for W/W fields and their associated pupil diameters (right).

The authors thank Chris Johnson and Shaban Demirel for critiquing the manuscript.