Perimetry has long been a part of clinical vision examinations,1 and in the past decades, static automated perimetry, such as the Zeiss Humphrey Field Analyzer (HFA), has become a “gold standard.”2 – 8 Threshold visual fields have been extremely useful for diagnosis and prognosis in patients with diseases of the visual system.9 However, in patients with eccentric viewing and/or unsteady fixation, the uncertainty about the exact position of the retina during testing reduces the clinical utility of these perimetry tests.10 – 21 In an attempt to address these problems, Scanning Laser Ophthalmoscopes (SLO) have been modified to allow fundus viewing while conducting perimetric tests. These instruments have not been widely available and/or easily modified for clinical work.15,18,22 – 26 Newer, clinically friendly perimetry instruments, such as the Nidek Inc. MP1 (MP1) and Opko Instrumentation Spectral OCT/SLO (OSLO) microperimeter track fixation and correct target placement for eye movements during testing. These instruments allow for precise microperimetric assessment of field sensitivities (i.e., fundus-guided perimeter).7,26 – 28
There are many similarities between the new microperimeters and standard clinical perimeters (such as the HFA), including similar spatial patterns, similar threshold algorithms, and similar stimulus sizes and durations. However, there are important differences in stimulus configuration that may produce discordant results. For example, the HFA perimeter uses a projection system with a broad range of stimulus intensities, whereas the MP1 and OSLO microperimetry instruments use small solid-state monitors to present the targets over a limited range of intensities. Stimuli are presented on a background of 10 candela per meter square (cd/m2) in the HFA and OSLO, but on a background of 1.27 cd/m2 in the MP1.
Given the increasing availability of microperimetry instruments, we illustrate the consequences of the choices of stimuli and the influences of hardware choices made by each particular equipment on perimetry results. To do this, we compare the results obtained with two microperimetry instruments (MP1 and OSLO) to those obtained with an HFA. The purpose of this study was not to collect normal values on a large number of subjects or to demonstrate disease influences on thresholds; many studies of this type have been previously published.26,29,30 The data of this study were used to give a reader an understanding of the relationships among the measures obtained from each instrument, to allow direct comparison of results on a standard scale, and to discuss the psychophysical implications.
We recruited six normally sighted subjects to participate in this study. Inclusion criteria were 20/20 or better acuity, refractive errors not exceeding ±5.00 D sphere and −2.00 D cylinder, no history of ophthalmologic or neurologic disease, and undilated pupil diameters of at least 4 mm. Subjects gave informed consent to participate, and the research was approved by the NYEE IRB. The study followed the tenets of the Declaration of Helsinki. Five normally sighted subjects with a mean age of 25.8 years participated in the study. The right eye of each subject was tested.
Subjects' thresholds were measured on the HFA, Nidek MP1 (MP1), and OSLO (Fig. 1). Pupil were not dilated for any of the tests and pupil diameters were measured for each subjects during testing. The order of testing was counterbalanced across subjects. On all machines, visual field testing was done with a 10-2 spatial distribution of test points, using Goldmann Size III and Size I stimuli, with a presentation time of 200 ms, and using a 4-2 threshold algorithm. These two test sizes were chosen to allow examination of the sensitivities over the entire range of physical stimuli available in the microperimetry instruments. Two tests per subject at each size were performed on each machine. Subjects were given rest periods after each test. The order of testing was counterbalanced. Testing was done in two sessions of approximately 2 h each, conducted on 2 consecutive days.
The individual parameters were as follows:
Humphrey Field Analyzer
The background level of the HFA was 10 cd/m2 (31.5 apostilbs (asb)).31 The average pupil diameter for our subjects under a 10-cd/m2 background was 5 mm. HFA stimuli can be presented over a range intensities from a maximum of 10,000 asb (3183 cd/m2) to a 5 log-unit attenuated value of 0.01 asb (0.03 cd/m2).31 The standard viewing distance of 30 cm was used. Fig. 1A shows the output of a typical HFA field test on a normally sighted subject. This output plots thresholds in dB attenuation as a function of visual field location. There were 68 test points presented in the HFA perimetry.
The background of the MP1 was nominally 1.27 cd/m2 (4 asb).29 The average pupil diameter for our subjects under a 1.0-cd/m2 background was 6 mm, which equates to an adapting illuminance of 1.52 log td in the MP1. The maximum stimulus intensity was 127 cd/m2 (400 asb), attenuated over a 2-log-unit range.29 The MP1 has an automated tracking system that shifts the position of the stimulus to compensate for movement of the eye. An image of the retina was taken, and a landmark region of 128 × 128 pixels was selected for tracking. During testing, infrared images were aligned to the reference image and stimulus positions adjusted accordingly. The time for tracking was <2 frames (80 ms).29 Fig. 1B shows the threshold values for a MP1 test overlaid on an image of the fundus. There were 68 test points presented in the MP1 perimetry.
Opko Instrumentation Spectral OCT/SLO
The background level of the our OSLO was 10 cd/m2 (31.5 asb to 2.29 log td), with a maximum stimulus intensity of 139.8 cd/m2 (439 asb), attenuated over a 2-log-unit range. The OSLO instrument employs a scanning laser to image the fundus and tracks eye movements by aligning a subset of fundus landmarks with a reference image. The images of the fundus and organic light-emitting diode are coherent; thus, eliminating concerns of misalignment.32 Stimulus position was adjusted, and presentation made in <100 ms. Fig. 1C shows the threshold results plotted on an OSLO image of the fundus. There were 52 test points presented using the OSLP perimeter.
Both the HFA and MP1 present 68 stimuli at equivalent field locations. For analyses between these instruments, all of the data were included. The OSLO presents only a 52 point subset of the points displayed on the other instruments. For all analyses with OSOL data, only the spatially equivalent 52 data points from the HFA and MP1 were used.
MP1 vs. HFA
In Fig. 2, the data for MP1 are plotted against the HFA data. For each subject, the averages of the two runs on each instrument for each stimulus position are plotted. The MP1 data were transposed into field view, and spatially corresponding data points were compared for the two tests. All the MP1 data fell below the values for their corresponding points on the HFA.
For the Goldmann Size III target (Fig. 2A), HFA thresholds ranged from 25 to 37 dB (median = 33 dB) across the field, whereas MP1 thresholds ranged from 12 to 20 dB (median = 19 dB). The Spearman correlation between these two measures was 0.24. For the MP1 data, there was a floor effect (a luminance increment lower than that of the 20 dB nominal intensity could not be presented in this instrument), with a disproportionate number of points falling at an MP1 value of 20 dB (from 35 to 61% of all points across subjects). The values of corresponding HFA points ranged from 27 to 37 dB.
For the Goldmann Size I target (Fig. 2B), HFA thresholds ranged from 14 to 30 dB (median = 24 dB), whereas MP1 thresholds ranged from 2 to 16 dB (median = 8.5 dB). All the MP1 data fell below the values for their spatially corresponding points on the HFA. The Spearman correlation between these two measures was 0.56.
When averaged into eccentricity rings, median thresholds decreased as a function of eccentricity for both HFA and MP1 (Fig. 3). There was no statistical difference in slope between the HFA or MP1 dB and ring eccentricity for either target size.
OSLO vs. HFA
In Fig. 4, the data for OSLO are plotted against the HFA data. The OSLO data were transposed into field view, and spatially corresponding data points were compared for the two tests. For each subject, the averages of the two runs on each instrument for each stimulus position are plotted. Many data points overlap due to the 4 to 2 threshold algorithm used.
For the Goldmann Size III target (Fig. 4A), HFA thresholds ranged from 25 to 37 db (median = 33 dB) across the field, whereas OSLO thresholds ranged from 12 to 20 dB (median = 18 dB). All the OSLO data fell below the values for their spatially corresponding points on the HFA. The Spearman correlation between these two measures was 0.49. OSLO thresholds decreased as a function of eccentricity at a rate that was not statistically different than the HFA data (Fig. 3; t = 0.91, p = ns).
For the Goldmann Size I target (Fig. 4B), HFA thresholds ranged from 14 to 30 dB (median = 24 dB), whereas OSLO thresholds ranged from 6 to 18 dB (median = 14 dB). All but four of the OSLO data fell below the values for their spatially corresponding points on the HFA. The Spearman correlation between these two measures was 0.38. For the Goldmann Size I target, OSLO thresholds decreased as a function of eccentricity at a rate that was not statistically different than the HFA data (t = 0.57, p = ns).
MP1 vs. OSLO
In Fig. 5, the data for MP1 are plotted against the OSLO data. For each subject, the averages of the two runs on each instrument for each stimulus position are plotted. For the Goldmann Size III target, the MP1 thresholds ranged from 12 to 20 dB (median = 19 dB), and the OSLO thresholds also ranged from 12 to 20 dB (median = 18 dB) (Fig. 5A). The Spearman correlation between these two measures was 0.28. For the Goldmann Size I target (Fig. 5B), the MP1 thresholds ranged from 2 to 16 dB (median = 8.5 dB), and the OSLO thresholds ranged from 6 to 18 dB (median = 14 dB). The Spearman correlation between the two measures was 0.54.
All of the OSLO data from the current study (Size I and III combined) were regressed against the HFA data. The slope of this fit was 0.327 and the intercept was 5.0 db. HFA and OSLO data from 20 eyes with glaucoma were available from the report of Lima et al.33 The range of HFA data for these patients was from 30 to 6 dB and the range for the OSLO was 18 to 6 dB. This range of dB values was very similar to the range of values across target sizes collected in the present study. A regression fitted to the patient data of Lima et al. yielded a slope of 0.327 and an intercept of 4.5 db. There were no statistically significant difference between the slopes or intercepts for the control and patient data.34 This finding points to the generalizability of the findings of current study to studies reporting data from patients.
The MP1 and OSLO threshold values were consistently lower than the HFA values at corresponding retinal locations, and there was not a one-to-one correspondence among these data. This might reflect the differences in scales, with the microperimeters' attenuation values ranging from only 0 to 20 dB. But, it is more subtle than this. The dB values for each test represent attenuation from different maximum values. In the HFA, attenuation is calculated from a maximum stimulus intensity of 10,000 asb, whereas the microperimeters maximum values are approximately 2 log unit dimmer. In order to accurately compare these tests, all the data should be converted into equivalent threshold values. Because the HFA data are a “gold standard” for perimetry, we converted the MP1 and the OSLO data into equivalent HFA values. In psychophysics, thresholds are standardly expressed as increments, measured in delta intensity units above the background level. Table 1 lists the published HFA,31 MP1,29 and measured OSLO increment intensities at each instrument's dB level. The conversion to HFA equivalent values was accomplished using the formula:
where 3183 is the maximum intensity of the HFA perimeter in cd/m2; μ is the value of the microperimeter (MP1 or OSLO) increment in cd/m2; and the mulitplier 10 converts the value to dB.
Using this conversion, the smallest increment displayed in the MP1 (1.27 cd/m2) was equivalent to 34 HFA dB, and the brightest increment displayed by the MP1 was 14 HFA dB (127 cd/m2). The smallest increment displayed in the OSLO (1.56 cd/m2) was equivalent to 33.1 HFA dB, and the brightest increment displayed by the OSLO was 13.6 HFA dB (137 cd/m2) (Fig. 6).
MP1 Equivalent vs. HFA
In Fig. 7, the data for MP1 thresholds, converted to equivalent HFA dB based upon their increment intensity values, are plotted against their spatially corresponding HFA-measured values. For the Goldmann Size III target (Fig. 7A), MP1 thresholds ranged from 26 to 34 equivalent HFA dB (median = 33 dB). For the Goldmann Size I target (Fig. 7B), MP1 thresholds ranged from 14 to 28 equivalent HFA dB (median = 21 dB). For both Goldmann sizes, the data clustered around the line of unity.
OSLO Equivalent dB vs. HFA
In Fig. 8, the data for OSLO thresholds converted to equivalent HFA dB are plotted against their corresponding HFA-measured data. For the Goldmann Size III target (Fig. 8A), OSLO threshold ranged from 25 to 34 equivalent HFA dB (median = 30 dB). For the Goldmann Size I target (Fig. 8B), OSLO threshold ranged from 19 to 31 equivalent HFA dB (median = 27 dB). For the larger target size, the data clustered around the line of unity, whereas the OSLO data for the smaller target were generally above (lower thresholds) than the HFA data.
To calculate the agreement within these perimeters, the differences between repeated tests on the same instrument for the Size III target were plotted against the average values as suggested by Bland and Altman.35 The agreement between repeated tests on the same instrument was calculated as the standard deviations of the differences. For the Size III target, the agreement was 1.6 dB for repeated measurements on the HFA, 2.0 dB for the MP1, and 2.5 dB for the OSLO. For the Size I target, the agreement between two repeated measurement was 2.6 dB for both the HFA and MP1 and 2.3 dB the OSLO.
As a comparison, agreement in thresholds between instruments for the Size III target was similar in magnitude to that within an instrument. Agreement between the HFA and MP1 was 2.8 dB, between the HFA and OSLO it was 2.3 dB, and between the MP1 and OSLO it was 2.4 dB. For the Size I target, agreement within instruments was better than between instruments. Agreement between the HFA and MP1 was 3.3 dB, between the HFA and OSLO it was 2.8 dB, and between the HFA and OSLO it was 3.6 dB.
The manuscript examined the relationship among thresholds obtained on three visual field instruments for the same sample of control individuals. We attempted to cover much of the available ranges of physical stimulus intensities available in these instruments by changing stimulus target size, rather than by including patients with diseases that could affect sensitivities. We used a smaller target size to extend the range of stimulus intensities. Across stimulus size, our range of threshold values for the HFA was from 37 to 14 dB, for the MP1, thresholds ranged from 20 to 2 dB, and for the OSLO, thresholds ranged from 20 to 6 dB. These values covered most of the ranges of intensities for the two microperimeters and much of the HFA range.
Our goal was to discuss the relationships among the testing algorithms and stimulus parameters, rather than to examine the relationship between sensitivities and disease. We found that, other than the floor effect, there was a linear relationship between instruments from very small to very large luminance increments. The present data would be invalid only if a disease caused differences in thresholds to stimuli of the same physical intensities when presented in different instruments. We have presented evidence from one group of patients that the relationships between instruments holds for patients as well as for targets of different sizes.
We found that a comparison among the results obtained with the three instruments did not yield unitary correspondence, mainly due to the differences in stimulus parameters. However, when all the data were expressed as equivalent HFA dB, there was a one-to-one relationship among most of the measures. One exception to this trend was the Goldmann Size III data from the MP1. There was a floor effect observed in these data, with many of the MP1 data points falling at the lowest increment value (20 dB). This was not observed in the OSLO data. This was most likely caused by the dimmer background used in the MP1 relative to the other two instruments. A 20-dB stimulus in the MP1 was equivalent to a 1.27 cd/m2 increment above the background, whereas a 20-dB OSLO stimulus was equal to an increment of 1.56 dB. At first glance, it would appear that the MP1 should be more sensitive at lower threshold levels due to its smaller increment. However, when the background illuminance is taken into account, the Weber contrast of the MP1 20-dB stimulus was 1.0 (Delta I /Background I), whereas the Weber contrast of the OSLO 20-dB stimulus was 0.15. As a comparison, the calculated median central HFA threshold was 35 dB for a Goldmann Size III target, which was equivalent to a Weber increment of 0.1. This means that the first contrast step displayed in the MP1 was considerably larger than the minimum threshold increment needed for a normally sighted subject to see the target. In fact, a Weber contrast of 1.0 was equivalent to a threshold of approximately 24 HFA dB. This may explain the correspondence of MP1 20-dB thresholds with HFA values between 25 and 35 dB.
Background illuminance plays a large role in determining increment thresholds. Reports on threshold vs. intensity functions have been published for human psychophysics and electrophysiology.36 – 38 In these functions, there is a range of lower adapting illuminance levels over which thresholds are unchanged. Luminance gain (linearity of the Weber function) begins at around a background illuminance level >2.0 log td.37 The low background level used in the MP1 was approximately 0.5 log unit below the level required for linearity and well into the mesopic range assuming an undilated pupil diameter of 6 mm. When measuring thresholds at adapting illuminance levels in the mesopic range, thresholds can be mediated by mixed rod-cone system responses and by the interactions between photoreceptor types.39 – 43 As a consequence, thresholds measured under mesopic conditions will vary with stimulus spectral content, retinal location, spatial frequency, and temporal properties.43
Although most of the data aligned with the HFA results when converted to equivalent HFA dB, the data for the Goldmann Size I target in the OSLO fell consistently above the line of unity. This indicates that threshold increments were lower for this size target in the OSLO instrument than in the HFA. Ricco's Law states that threshold is determined by the product of stimulus area and luminance.44 Within a critical area, the product is constant. In our study, median thresholds increased between the Goldmann Size III and Goldmann Size I targets for the HFA from 33 dB (1.6 cd/m2) to 23 dB (12.64 cd/m2). The area of a Goldmann Size I stimulus is sixteen times smaller than the area of the Goldmann Size III stimulus. The median increment energy needed to reach thresholds for Goldmann Size I targets was 9.8 times higher than that required for the Goldmann Size III in the HFA, indicating incomplete spatial summation. This deviation from Ricco's Law is most likely due to the use of relatively large stimulus sizes and the fact that we averaged over a range of eccentric locations.45,46 For the MP1, thresholds increased from a median of 19 dB for the Goldmann Size III target to 8 dB for the Goldmann Size I target, or an increase in luminance of 10.1 times, again demonstrating less than complete summation. For the OSLO data, the change in median thresholds between the two stimulus sizes was only 3 dB (from 17 to 13 dB), or an increase in threshold energy of only 2.6 times. One explanation for the discrepant finding with the OSLO might be that the size of the Goldmann Size I stimulus in the OSLO was too large. After consulting with the manufacturer, it was discovered that the Goldmann Size I target presented by OLSO was 2.5 times larger in area than a standard Goldmann Size I target. This error has been corrected in newer versions of their software.
The findings of our work emphasize the importance of comparing data using equivalent scales. There was good correspondence among these results when compared on equivalent increment threshold units. However, discrepancies in our findings allowed us to evaluate consequences of design choices made by the manufacturers. The lesson to be learned from this work is that users should make no assumptions about what the equipment is doing and should always evaluate the psychophysical consequences of the stimuli that are used by a particular instrument.
111 East 59th Street
New York, NY 10022
This work was supported in part by a grant from the U.S. Department of Veterans Affairs and the Bendheim Family Retina Center of the New York Eye and Ear Infirmary.
1. Von Graefe A. Uber die untersuchung des gesichtsfeldes bei amblyopischen affectionen. Von Grafes Arch Ophthalmol 1856;2:258–98.
2. Fankhauser F, Koch P, Roulier A. On automation of perimetry. Albrecht Von Graefes Arch Klin Exp Ophthalmol 1972;184:126–50.
3. Heijl A. Studies on computerized perimetry. Acta Ophthalmol Suppl 1977;132:1–42.
4. Heijl A, Krakau CE. An automatic perimeter for glaucoma visual field screening and control. Construction and clinical cases. Albrecht Von Graefes Arch Klin Exp Ophthalmol 1975;197:13–23.
5. Krakau CE. Aspects on the design of an automatic perimeter. Acta Ophthalmol (Copenh) 1978;56:389–405.
6. Lynn JR. Examination of the visual field in glaucoma. Invest Ophthalmol 1969;8:76–84.
7. Midena E, Vujosevic S, Cavarzeran F. Normal values for fundus perimetry with the microperimeter MP1. Ophthalmology 2010;117:1571–6.
8. Wild JM. Techniques and developments in automated perimetry: a review. Ophthal Physiol Opt 1988;8:295–308.
9. Choplin NT, Russel PE. Visual Field Testing With the Humphrey Field Analyzer, 2nd ed. Thorofare, NJ: SLACK Inc.; 1999.
10. Choplin NT. Scanning laser imaging devices: what do we measure? Clin Experiment Ophthalmol 2007;35:401–2.
11. Crossland MD, Dunbar HM, Rubin GS. Fixation stability measurement using the MP1 microperimeter. Retina 2009;29:651–6.
12. McMahon TT, Hansen M, Viana M. Fixation characteristics in macular disease. Relationship between saccadic frequency, sequencing, and reading rate. Invest Ophthalmol Vis Sci 1991;32:567–74.
13. Sunness JS, Applegate CA. Long-term follow-up of fixation patterns in eyes with central scotomas from geographic atrophy that is associated with age-related macular degeneration. Am J Ophthalmol 2005;140:1085–93.
14. Tarita-Nistor L, Gonzalez EG, Markowitz SN, Steinbach MJ. Fixation characteristics of patients with macular degeneration recorded with the mp-1 microperimeter. Retina 2008;28:125–33.
15. Timberlake GT, Mainster MA, Webb RH, Hughes GW, Trempe CL. Retinal localization of scotomata by scanning laser ophthalmoscopy. Invest Ophthalmol Vis Sci 1982;22:91–7.
16. Cacho I, Dickinson CM, Reeves BC, Harper RA. Visual acuity and fixation characteristics in age-related macular degeneration. Optom Vis Sci 2007;84:487–95.
17. Mori F, Ishiko S, Kitaya N, Takamiya A, Sato E, Hikichi T, Yoshida A. Scotoma and fixation patterns using scanning laser ophthalmoscope microperimetry in patients with macular dystrophy. Am J Ophthalmol 2001;132:897–902.
18. Schuchard RA. Preferred retinal loci and macular scotoma characteristics in patients with age-related macular degeneration. Can J Ophthalmol 2005;40:303–12.
19. Schuchard RA, Naseer S, de Castro K. Characteristics of AMD patients with low vision receiving visual rehabilitation. J Rehabil Res Dev 1999;36:294–302.
20. Sunness JS, Applegate CA, Haselwood D, Rubin GS. Fixation patterns and reading rates in eyes with central scotomas from advanced atrophic age-related macular degeneration and Stargardt disease. Ophthalmology 1996;103:1458–66.
21. Whittaker SG, Cummings RW, Swieson LR. Saccade control without a fovea. Vision Res 1991;31:2209–18.
22. Fujii GY, De Juan E Jr., Humayun MS, Sunness JS, Chang TS, Rossi JV. Characteristics of visual loss by scanning laser ophthalmoscope microperimetry in eyes with subfoveal choroidal neovascularization secondary to age-related macular degeneration. Am J Ophthalmol 2003;136:1067–78.
23. Mainster MA, Timberlake GT, Webb RH, Hughes GW. Scanning laser ophthalmoscopy: clinical applications. Ophthalmology 1982;89:852–7.
24. Rohrschneider K, Bultmann S, Gluck R, Kruse FE, Fendrich T, Volcker HE. Scanning laser ophthalmoscope fundus perimetry before and after laser photocoagulation for clinically significant diabetic macular edema. Am J Ophthalmol 2000;129:27–32.
25. Sunness JS, Bressler NM, Maguire MG. Scanning laser ophthalmoscopic analysis of the pattern of visual loss in age-related geographic atrophy of the macula. Am J Ophthalmol 1995;119:143–51.
26. Van de Velde FJ. Microperimetry. Bull Soc Belge Ophtalmol 2006:37–49.
27. Podoleanu AG. Combining SLO and OCT technology. Bull Soc Belge Ophtalmol 2006:133–51.
28. Rohrschneider K, Bultmann S, Springer C. Use of fundus perimetry (microperimetry) to quantify macular sensitivity. Prog Retin Eye Res 2008;27:536–48.
29. Midena E. Microperimetry and the Fundus: An Introduction to Microperimetry. Thorofare, NJ: SLACK Inc.; 2007.
30. Rohrschneider K, Springer C, Bultmann S, Volcker HE. Microperimetry—comparison between the micro perimeter 1 and scanning laser ophthalmoscope—fundus perimetry. Am J Ophthalmol 2005;139:125–34.
31. Heijl A, Patella VM. Essential Perimetry: The Field Analyzer Primer. Dublin, CA: Carl Zeiss Meditec Inc.; 2002.
32. Woods RL, Vera-Díaz FA, Lichtenstein L, Peli E. Spatial alignment of microperimeters. Invest Ophthalmol Vis Sci 2007;48:E-Abstract 144.
33. Lima VC, Prata TS, De Moraes CG, Kim J, Seiple W, Rosen RB, Liebmann JM, Ritch R. A comparison between microperimetry and standard achromatic perimetry of the central visual field in eyes with glaucomatous paracentral visual-field defects. Br J Ophthalmol 2010;94:64–7.
34. Zar JH. Biostatistical Analysis. Engelwood Cliffs, NJ: Prentice-Hall; 1974.
35. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307–10.
36. Greenstein VC, Hood DC. Test of the decreased responsiveness hypothesis in retinitis pigmentosa. Am J Optom Physiol Opt 1986;63:22–7.
37. Hood DC. Lower-level visual processing and models of light adaptation. Annu Rev Psychol 1998;49:503–35.
38. Seiple WH, Holopigian K, Greenstein VC, Hood DC. Sites of cone system sensitivity loss in retinitis pigmentosa. Invest Ophthalmol Vis Sci 1993;34:2638–45.
39. Andrews DP, Hammond P. Mesopic increment threshold spectral sensitivity of single optic tract fibres in the cat: cone-rod interaction. J Physiol 1970;209:65–81.
40. Frumkes TE, Sekuler MD, Reiss EH. Rod-cone interaction in human scotopic vision. Science 1972;175:913–14.
41. Latch M, Lennie P. Rod-cone interaction in light adaptation. J Physiol 1977;269:517–34.
42. Makous W, Teller D, Boothe R. Binocular interaction in the dark. Vision Res 1976;16:473–6.
43. Stockman A, Sharpe LT. Into the twilight zone: the complexities of mesopic vision and luminous efficiency. Ophthalmic Physiol Opt 2006;26:225–39.
44. Davson H. Davson's Physiology of the Eye, 5th ed. London: Macmillan Academic and Professional Ltd.; 1990.
45. Matin L. Ricco's law: response as a power function of stimulus luminance and distance from target center. Vision Res 1975;15:1381–4.
46. Matin L, Kornheiser A. Linked changes in spatial integration, size discrimination, and increment threshold with change in background diameter. Vision Res 1976;16:847–60.