Visiortyes. Vol. 17, pp. ,095 to 1103. Pergamon Press 1977. Printedm Great Britain.
VERSUS DARKNESS BRIGHTNESS ENHANCEMENT ENHANCEMENT AT A BORDER ARNULF
REMOLE
School of Optometry, University of Waterloo, Waterloo, Ontario, Canada (Received 18 November 1976; in revised form 3 March 1977) Abstract-The dark and bright portions of the enhancement effects perceived at an abrupt luminance discontinuity are measured with respect to their widths. The luminances of the fields on either side of the discontinuity constitute the independent variable. It is found that the width of the dark band is affected more strongly by luminance changes in the bright field than the bright band is affected by changes in the dark field. At moderate photopic luminance% the dark band is the broader. Towards lower luminances, both bands increase, the bright band more rapidly to approach and sometimes overshoot the dark band. It is suggested that the asymmetries result from a one-way effect between the two fields, from stray light effects, and possibly because the bands may represent separate systems for the perceptions of brightness and darkness.
Key Words-Border enhancement.
INTRODUCTION
The phenomenon of simultaneous border contrast enhancement has been known since antiquity and was described first by Leonardo da Vinci (Jung, 1973). However, the amount of quantitative information on the phenomenon is not large, particularly when compared with the great number of studies on the closely related Mach effects generated by a finite stimulus gradient. Most observations on border enhancement have been incidental to such Mach studies (Mathews, 1966; B&k&y, 1968; Shipley and Wier, 1972). Border enhancement effects resemble conventional Mach effects in exhibiting a band of increased brightness in the bright field and a band of increased darkness in the darker field. However, at an abrupt border, the bands appear to meet right at the boundary, thus heightening the contrast between the two fields. Watrasiewicz (1966) measured subjective luminance distributions near abrupt borders transmitted by optical instruments. Davidson and Whiteside, (1971) made similar measurements near a sharp border in the context of Mach theory. Mathews (1966) and Wildman (1974) tested threshold sensitivities near a border. Remole (1974, 1976) measured the effects of retinal blur, retinal illumination and retinal location on the width of the bright band. In many cases, such measurements parallel similar findings with classical Mach bands, whereas in other cases it is difficult to draw direct comparisons. The main concern of the present study is that the bright and the dark portions of the border contrast phenomenon are quite dissimilar, the dark band usually appearing the larger. In conventional Mach effects, such asymmetries are well explored (Shipley and Wier, 1972). Marimont (1963) noted that the bright Mach effects are several times as bright as the dark bands are dark. However, this apparent difference may be due to the logarithmic compression of the stimulus magnitudes in the sensory system (BCkesy. 1968). That is. a greater amount of light
would be required to match or measure a certam brightness increment at a high luminance level than to match a similar increment at a lower level. Measurements of the brightnesses of the two bands would, therefore, if given in linear units, result in a spurious asymmetry. Since this is common to all perceptual processes, it may not qualify as a true difference between the two bands. Differences in width between the two bands constitute perhaps a better criterion for establishing genuine asymmetries in the Mach bands, or in the border effects, than differences in subjective brightness. Variations in width do not follow variations in the brightness distributions in the bands (McCollough, 1955; Bliss and Macurdy, 1961). Fiorentini and Radici (1958) found, while testing the effect of luminance gradient on the bands, that the bright band was usually narrower than the dark band. Shipley and Wier (1972) found that the dark band became wider than the bright band at low photopic luminances. They suggest that this may be due to different processes for the two bands. It is of interest to determine if similar asymmetries exist also in the border contrast phenomenon. If this effect is to be better understood, we must establish how the bright and dark portions of the enhanced region compare, whether they are mirror images of each other or exhibit basic differences. In this study, the widths of the bands generated by an abrupt step between two fields differing in luminance were measured for various combinations of the magnitudes of the two luminances. In one experiment, the luminance of the dimmer field was varied while the luminance of the brighter field was constant. This would show the effect of variations in the low-luminance field on the dark band. Also, since there is interaction between the differently stimulated retinal areas, it would show the effect of such variations on the bright ,band. In a second experiment, the higher luminance field was varied while the lower luminance field was constant. This would
109h
AKNIILI
RfMoLI
show the effect of these variations on the bright band as well as effects induced in the opposite field, on the dark band. In both experiments, findings were obtained for several luminance levels of the constant
field. APPARATUS
AND METHOD
A plan of the apparatus is shown in Fig. I. Two diffusing plastics sheets. transilluminated by light projected from tungsten sources. were superimposed optically to form the stimulus field. The dimmer field was seen directly, whereas the brighter field was neen by reflection in a semisilvered mirror. The diffuser producing the brighter field was masked by a razor edge so as to produce a sharp transition with the darker field. To the observer. the fields appeared as two equa1 size semicircles, together subtending approximately 12” at the eye. To measure the spread of the enhancement effects from the border, a thin wire was suspended vertically, parallel to the border. It could be positioned anywhere in the field via a mechanical control operated by the subject. A small portion of the wire was illuminated by a narrow beam whose intensity could be varied to bring about a moderate contrast between the hairline and the field under observation. To avoid parallax, great care was taken to focus the razor’s edge in the plane of the hairline. Lenses were placed in front of the observer‘s eye so that the hairline and border were optimally focused on the retina. A 2-mm dia artificial pupil was positioned 8 mm in front of his cornea. The subject measured the spread of the enhancement by bringing the wire towards the border region from an initial position in the periphery of the field. All the time maintaining fixation of the border, he would bring the hairline in coincidence with the termination of the enhanced region in the field. The dark band was always measured before the bright band. Figure 1 shows the stimulus presented to the subject. fn the first experiment. the bright field was kept constant while the darker field was reduced in 0.20log unit steps from the luminance of the bright field through a range of 4 log units. This was repeated for each of five luminances of the bright field, 100, 10, 1.0, 0.1 and 0.01 cd/m2. Five series of data were obtained for all these luminance combinations. To plot the data. moving averages of results
Fig. 2. Stimulus field with illuminated portion of hairline visible in the dark field. L, luminance distribution; B, perceived brightness distribution. The arrows indicate the widths measured.
for three consecutive log increments were used. In Fig. 3, error bars enclosing k 1.00 S.E. of the mean and representing moving average populations of 15 have been shown at every second datum point. In the second experiment, the luminance of the darker field was kept constant while. from this luminance, the bright-field luminance was increased in 0.20 log increments through 4 log units or less, the upper limit being lOOcd!m’. Five series of such data were compiled. The plotting method, shown in Fig. 4, is similar to that used in the first experiment. Three subjects, IR, WP and AU, participated in the project. Only the right eye was tested. The respective refractions in this eye were +0.75 D/ - 0.25 DC x 180”. + 1.25 D/ - 0.50 DC x 180’ and -4.25 D. Only subject AR had any previous experience in making measurements on the border enhancement response.
RESULTS
Fig. 1. Plan of apparatus. P,, P,, P,, projectors; ND,. ND,, neutral filter and wedge combinations; T, transformer: DS, , DS,, diffusing screens; L, lens for focusing E, razor edge, in plane of H, hairline, via SM, semisilvered mirror; LB, lens battery for focusing stimulus; A, artificial pupil; MG, microgauge.
Figure 3 shows the effect on the enhancement bands of luminance changes in the dark field. The luminance values heading the graphs refer to the constant bright field. With high luminances, there is a marked increase in the dark band through the first portion of the dark field luminance decrement range. After its initial increase, the function representing the dark band becomes rather irregular, showing troughs and peaks through the remainder of the range. An initial increase in the dark band does not always occur at low luminances, as, for example, for subject AR for the LOcd/m” level. An important observation is that, with the higher bright-field luminance& changes in the low-luminance field have little or no effect on the width of the bright band, this band remaining nearly constant through the entire decrement range. This observation does not hotd for the lowest iuminances of the brighter field, where some irregular changes in the bright band do occur with decreases in the dark-field luminance%
Brightness enhancement versus darkness enhancement 100 cd/m2
Y----e+
1cd/m2
10 cd/m2
1
1
2
0.01 cd/m:!
0.1 cd/m2
9---k++
3 4 Log luminance
w decrement
w
Fig. 3. Effect on border enhancement extent with constant bright-field luminance and decreasing darkfield luminance. Results are shown for five luminances of the bright field, 100, 10, 1.0, 0.1 and 0.01 cd/m2. IR, WP and AR refer to the subjects tested. 0, the dark portion; 0, the bright portion of the enhancement.
1234
1 Log
234 luminance
1
234
1
2
3
4
increment
Fig. 4. Effect on border enhancement extent with constant dark-field luminance and increasing brightfield luminance. Results are shown for four luminance levels of the dark field, 1, 0.1, 0.01 and 0.001 cd/m2. The subjects and symbols are the same as in Fig. 3.
109x
ARNI.L~RI:MOLI-
On the other hand, the bright-field luminances strongly affect the width of the bright band. In one subject, IR, the width of this band is a minimum for the l-cd/m’ level and increases both towards higher and lower luminances, whereas in the other two subjects. the minimum width occurs towards the highest luminances. The data agree qualitatively with previous findings (Remole, 1976) for the bright band, which showed a large increase towards low luminances for all subjects as well as a small increase also towards higher luminances for some subjects. In subject AR, the bright band increases towards low luminances to the extent that it far overshoots the dark band. whereas in the other subjects the bright band merely converges towards the dark band. Such large individual differences are typical of border enhancement responses at low luminances. They are possibly due to differences in fixation performance. For example, if the subject were to fixate slightly eccentrically at low luminances. the enhancement bands would be affected. In Fig. 4, the bright-field luminance was increased while the dark field was constant. The four headings give the luminances of the dark field. As one would expect from Fig. 3, the bright band increases with ascending luminances in the bright field in subject IR but decreases in AR. In WP. the width of this band decreases with increasing luminance for the two lower levels but shows little change for the two higher levels, the results again being in fair agreement with Fig. 3. The apparent discrepancy between subjects IR and AR with respect to the slope of the graphs is merely an illustration of the fact that some subjects show a bright band minimum at moderately high luminances, such as, for example, lOcd/m’. whereas others appear to have a minimum at much higher luminances. Again, this difference may be a reflection of the individual fixation performance. Although the two experiments in most instances agree reasonably well for identical luminance combinations, there are a few cases where the agreement is not good. For subject IR, at the dark-field level of 0.001 cd/m” in Fig. 4. the bright field increases 4 log units through 0.01, 0.1, 1.O and 10 cd/m2. We should expect a minimum for the bright band at l.Ocd/m’ for the bright field since this corresponds to the minimum for the bright band in Fig. 3. The bright band curve does dip down at the 3 log unit interval representing this luminance, but not nearly enough to agree well with Fig. 3. Perhaps in this case. the manner of fixating the border was affected somewhat by the sequence of presenting the data. If this iS the case. it is a factor that is extremely difficult to control. Contrary to the relationship between the dark field and the bright band, the dark band varies markedly with changes in the bright field. In most cases in Fig. 4, there is an initial increase in the dark band as the luminance of the bright field is increased. Again, especially towards low luminances. the individual differences become very marked. For example. for subject WP at the dark-field luminance of 0.001 cd/m’. the dark band decreases from a very large initial value. The dark band varies also with changes in the dark field. In general, this consists in an increase towards lower luminances. similar but less pronounced than
that found with the bright band. However, many individual differences are superimposed on this trend. In order to average out some of these individual differences, the means for the three subjects have been plotted in Fig. 5. The upper portion of ths figure indicates more clearly the increase in the dark band with decreasing luminance of the dark field, the greater increase occurring with the inital log reductions. Also. this is generally evident from the lower portion. where four luminances of the dark field are represented. Another type of summary table, integrating data from the two experiments. is shown in Fig. 6. Plottings representing identical contrasts have been selected from Figs. 3 and 3 and have been replotted against their corresponding luminances. The two contrasts shown are those represented by the abscissa values of 0.4 and 2.0 log units. Identical contrasts for the two bands obviously do not imply identical stimulus situations. For a given luminance of the dark field, a given contrast is produced by a certain bright-field luminance. If the given luminance refers to the bright field, the same contrast is produced by a certain darkfield luminance. Not only is the mean luminance greater in the first situation. but the contrasts are also in opposite directions. Therefore, the functions for the two bands will not coincide. However. if the changes in bandwidth on both sides of the border were symmetrical. their corresponding functions would be similar in form. Except for subject WP. for a contrast of 2.0 log difference between the two fields, the pairs of curves in Fig. 6 are quite dissimilar. thus presenting another example of assymmetry. When the bright and dark bands are compared in ‘Figs. 5 and 6, certain general trends transcend the individual characteristics. Thus. a comparison between the two bands can bc summarized as follows: (I ) for the higher luminances. the width of the bright band is hardly affected by variatiohs in the dark-field luminance, whereas the dark band increases markedly with ascending bright-field luminances. However, towards low mesopic luminances, the two bands behave more equally in this respect. (2) For photopic and high mesopic luminances, the dark band is wider than the bright band. (3) The bright band increases markedly towards low luminances of the corresponding field. and. in some subjects. slightly towards high luminances. The dark band exhibits similar but less pronounced changes as the dark-field luminance is varied. (4) Functions describing the effect of luminance on the bandwidth for constant contrasts differ in form for the two bands. Extrapolation of these functions suggests that the bright band is the larger at some luminance levels. In one subject, AR, the bright band does. indeed. become wider than the dark band at low luminances DISCUSSION
Some individual differences have been pointed out in the description of the results, and it has been suggested that small differences in the way the border is fixated may be an important factor here. There may be other factors as well, such as the rate and extent of involuntary eye movements. A third factor is the focussing performance of the subject: small fluctuations in accommodation could easily affect the blur
1099
Brightness enhancement versus darkness enhancement 0.01 cd/m?
t
::~~~;~ .E E
-+-e?ww
I-_1 Log
z
luminance
2
3
4
decrement
aI
5
234 Log
luminance
1
increment
Fig. 5. Upper portion: mean results for the three subjects in Fig. 3. Lower portion: similar mean results for Fig. 4. The symbols are the same as in Figs. 3 and 4.
of the retinal image and hence the border enhancement extent (Remole, 1974). Thus, the border enhancement response is relatively labile, and that there are irregularities in the findings is therefore not surprising. However, in spite of these difficulties, the general trends described are quite obvious. The general trends defining the differences between the two bands need to be discussed in terms of explanatory hypotheses. First, the relatively pronounced effect of the bright field on the dark band is reminiscent of the one-way effect often observed with respect to the interaction between a bright and
a dark field. Although the theory of lateral neural inhibition (Ratliff, Hartline and Miller, 1963; Motokawa, 1970) assumes mutual interaction between visual units, the brighter field is perceptually the most powerful in transmitting effects across the border. For example, the perceived brightness of a field is affected by a brighter surround, but not by a darker surround (Marks, 1974). In our study, the logarithmic luminance scale may also have contributed to this effect. The relatively stronger effect of the bright field may result in part also from the distribution of intraocular
t WP
Log luminance,
ccl/m*
Fig. 6. Effect on border enhancement of the corresponding field luminance for two constant contrasts, represented by log differences, C, of 0.40 and 2.00. The data points have been selected from Figs. 3 and 4.
scattering. Luminance thresholds near the border (Yonemura, 1962; Wildman, 1974) show that the scattered light in the retinal image of the low-luminance field increases towards the border. Such an increase is also in accordance with the distribution of light scattered from the cornea and eye lens (DeMott and Boynton. 1958: Trokel, 1962; Feuk and McQueen, 197I ). We expect_a corresponding reduction near the border of the high-illuminance portion of the retinal image. When the bright-field luminance increases, the amount of scattered light in the low-illuminance portion of the retinal image will increase and fill in the lower flexure adjacent to the distribution representing the border. This could result in a widening of the dark band. On the other hand, the corresponding changes in scattering effects on the bright side of the border, accompanying changes in the low-ium~nance field, are relatively minute and cannot be expected to affect the bright band. These two factors, the one-way effect and the scattering effect, may also partially explain why the dark band is generally the greater. However, the observation that the bands increase with decreasing field luminance, and that the bright band increases more rapidly than the dark band, cannot readily be explained in such terms. The marked widening of the bands with reduced luminance must be due to changes at the neural level. in terms of distances traveled by the inhibitory process. If there is an increase in convergence and coarseness of the neural mechanism towards lower luminances (Pirenne, 1962), the enhancement effects would be expected to spread farther as over a more thinly distributed mosaic. A parallel to these changes if found in the increased enhancement spread when the border stimulation is moved away from the fovea centralis (Remole, 1976). Here, the increase is clearly associated with the increasing neural convergence towards the retinal periphery. On the other hand, the relatively small increase in enhancement shown by some subjects for high luminances may originate from small changes in fixation or focussing performance. That the two bands increase at different rates with decreasing luminance is shown also in the constant contrast functions in Fig. 6. The simplest explanation for this difference is that the bands represent a dual system, one for brightness and one for darkness, now widely recognized in principle (Marks. 1974). Such a system has been identified with two types of receptive fields (Barlow, Fitzhugh and Kuffler, 1957; Jung, 1973; Gerrits and Vendrik, 1970; ~a~ussen and Glad. 1975).The different rates of change in the bright and dark bands may be ascribed to differences that are bound to exist between two such systems as they adjust to luminance changes. ~c~lzo~~~~g~~n~nrs-I wish to thank Dr. W. H. Pym and Dr. 1. 0. Revill for their assistance with the project. The work was supported by Grant No. A9951 from the National Research Council of Canada.
RI”.FERENCES
Barlow H. B.. Fitzhugh R. and KuH?er S. W. (1957) (~‘hangc of organization in the receptive fields of the cat’s retinn during dark adaptation. J. Pity.+/.. Lo&. 137. ?iX 3%. BCkCsy G. von (1968) Brightness distribution across the Mach bands measured with flicker photometry. and the linearity of sensory nervous interaction. .I, opt. SCU .Am
58. I-X. Bliss 3. C. and Macurdy W. B. (1961) Linear models for contrast phenomena. J. npr. SOL.&I. 53. 1373 1379. Davidson M. and Whiteside J. A. (1971) ?tuman hrightness perception near sharp contours. J. cr/~t. ,Soc, clnt. 61. SXt536. DeMott D. W. and Boynton R. M. (1958)Retinal dtstribution of entoptic stray light. J. opt. Sot. Aln. 48. 13 -22. Feuk T. and McQueen D. (1971) The angular dependence of light scattered from rabbit corneas. It7wsr. ~~~rj~~. to. 294-299. Fiorentini A. and Radici T. (1958) Brightness, width and position of Mach bands as a function of the rate of variation of the luminance gradient. .4tti Fond. Giorgio Ronchi 13, 145-155. Gerrits H. 3. M. and Vendrik A. J. H. (1970) Simuitaneous contrast, filling-in process and information processing in man’s visual system. Expt. bruin Res. 11, 41 l-430. Jung R. (I9731 Visual perception and neurophysiology. In Handbook of Sensory Physiology. Vol. VII/3 (Edited by Jung R.). Springer. Berlin. Maenussen S. and Glad A. (1975) Brightness and darkness enhancement during flicker: perceptual correlates of neuronal B- and D-systems in human vision. E.q~l. Bruin Rex 22, 399-413. Marimont R. B. (1963) Linearity and the Mach phenomenon. J. opt. Sot. .4m. 53, 40@401. Marks L. E. (1974) Srrxor,~ Processes. Academic Press, New York. Mathews M. L. (1966) Appearance of Mach bands for short durations and sharply focused contours. .I. opt. sot. Am. 56. 1401--1402. McCollou~ C. (1955) The variation in width and position of Mach bands as a function of luminance. J. e~p. Psychol, 49. 141-152. Motokawa K. (1970) Physiology c~f Colour urd Putturn Vision. Igaku Shoin, Tokyo. Pirenne M. H. (1967) Visual acuity. In The qe. Vol. 2 (Edited by Davson H.). Academic Press, New York. Ratliff F.. Hartline H. K. and Miller W. H. (1963) Spatial and temporal aspects of retinal inhibitory intcr~iction. J. opt. Sot. Al?!.53. 11(&120. Remole A. (1974) Relation between border enhancement extent and retinal image blur. Visiotl Res. 14. 9X9-995.
Remole A. (1976)Effectof retinal illuminance and eccentricitv on border enhancement extent. l’ision Rus. 16. 1353--l317. Shipley T. and Wier C. (1972) Asymmetries in the Mach band phenomena. Kybern~tik 10. 181-.1X9. Trokel S. (1962) The physical basis for transparency 01 the crystalline lens. iriu;st. Ophthd. 1. 493-501. Watrasiewicz B. M. (1966) Some factors affecting the appearance of Mach bands. J. opt. Sot. An]. 56. 53G-536. Wildman K. N. (1974) Visual sensitivity at an edge. vision Rrs. 14, 749-755. Yonemura G. T. (1962) Luminance thresonds as a function of angular distance from an inducing source. J. opf. Sot. .4m. 52. 1030-1034.
VERSUS DARKNESS BRIGHTNESS ENHANCEMENT ENHANCEMENT AT A BORDER ARNULF
REMOLE
School of Optometry, University of Waterloo, Waterloo, Ontario, Canada (Received 18 November 1976; in revised form 3 March 1977) Abstract-The dark and bright portions of the enhancement effects perceived at an abrupt luminance discontinuity are measured with respect to their widths. The luminances of the fields on either side of the discontinuity constitute the independent variable. It is found that the width of the dark band is affected more strongly by luminance changes in the bright field than the bright band is affected by changes in the dark field. At moderate photopic luminance% the dark band is the broader. Towards lower luminances, both bands increase, the bright band more rapidly to approach and sometimes overshoot the dark band. It is suggested that the asymmetries result from a one-way effect between the two fields, from stray light effects, and possibly because the bands may represent separate systems for the perceptions of brightness and darkness.
Key Words-Border enhancement.
INTRODUCTION
The phenomenon of simultaneous border contrast enhancement has been known since antiquity and was described first by Leonardo da Vinci (Jung, 1973). However, the amount of quantitative information on the phenomenon is not large, particularly when compared with the great number of studies on the closely related Mach effects generated by a finite stimulus gradient. Most observations on border enhancement have been incidental to such Mach studies (Mathews, 1966; B&k&y, 1968; Shipley and Wier, 1972). Border enhancement effects resemble conventional Mach effects in exhibiting a band of increased brightness in the bright field and a band of increased darkness in the darker field. However, at an abrupt border, the bands appear to meet right at the boundary, thus heightening the contrast between the two fields. Watrasiewicz (1966) measured subjective luminance distributions near abrupt borders transmitted by optical instruments. Davidson and Whiteside, (1971) made similar measurements near a sharp border in the context of Mach theory. Mathews (1966) and Wildman (1974) tested threshold sensitivities near a border. Remole (1974, 1976) measured the effects of retinal blur, retinal illumination and retinal location on the width of the bright band. In many cases, such measurements parallel similar findings with classical Mach bands, whereas in other cases it is difficult to draw direct comparisons. The main concern of the present study is that the bright and the dark portions of the border contrast phenomenon are quite dissimilar, the dark band usually appearing the larger. In conventional Mach effects, such asymmetries are well explored (Shipley and Wier, 1972). Marimont (1963) noted that the bright Mach effects are several times as bright as the dark bands are dark. However, this apparent difference may be due to the logarithmic compression of the stimulus magnitudes in the sensory system (BCkesy. 1968). That is. a greater amount of light
would be required to match or measure a certam brightness increment at a high luminance level than to match a similar increment at a lower level. Measurements of the brightnesses of the two bands would, therefore, if given in linear units, result in a spurious asymmetry. Since this is common to all perceptual processes, it may not qualify as a true difference between the two bands. Differences in width between the two bands constitute perhaps a better criterion for establishing genuine asymmetries in the Mach bands, or in the border effects, than differences in subjective brightness. Variations in width do not follow variations in the brightness distributions in the bands (McCollough, 1955; Bliss and Macurdy, 1961). Fiorentini and Radici (1958) found, while testing the effect of luminance gradient on the bands, that the bright band was usually narrower than the dark band. Shipley and Wier (1972) found that the dark band became wider than the bright band at low photopic luminances. They suggest that this may be due to different processes for the two bands. It is of interest to determine if similar asymmetries exist also in the border contrast phenomenon. If this effect is to be better understood, we must establish how the bright and dark portions of the enhanced region compare, whether they are mirror images of each other or exhibit basic differences. In this study, the widths of the bands generated by an abrupt step between two fields differing in luminance were measured for various combinations of the magnitudes of the two luminances. In one experiment, the luminance of the dimmer field was varied while the luminance of the brighter field was constant. This would show the effect of variations in the low-luminance field on the dark band. Also, since there is interaction between the differently stimulated retinal areas, it would show the effect of such variations on the bright ,band. In a second experiment, the higher luminance field was varied while the lower luminance field was constant. This would
109h
AKNIILI
RfMoLI
show the effect of these variations on the bright band as well as effects induced in the opposite field, on the dark band. In both experiments, findings were obtained for several luminance levels of the constant
field. APPARATUS
AND METHOD
A plan of the apparatus is shown in Fig. I. Two diffusing plastics sheets. transilluminated by light projected from tungsten sources. were superimposed optically to form the stimulus field. The dimmer field was seen directly, whereas the brighter field was neen by reflection in a semisilvered mirror. The diffuser producing the brighter field was masked by a razor edge so as to produce a sharp transition with the darker field. To the observer. the fields appeared as two equa1 size semicircles, together subtending approximately 12” at the eye. To measure the spread of the enhancement effects from the border, a thin wire was suspended vertically, parallel to the border. It could be positioned anywhere in the field via a mechanical control operated by the subject. A small portion of the wire was illuminated by a narrow beam whose intensity could be varied to bring about a moderate contrast between the hairline and the field under observation. To avoid parallax, great care was taken to focus the razor’s edge in the plane of the hairline. Lenses were placed in front of the observer‘s eye so that the hairline and border were optimally focused on the retina. A 2-mm dia artificial pupil was positioned 8 mm in front of his cornea. The subject measured the spread of the enhancement by bringing the wire towards the border region from an initial position in the periphery of the field. All the time maintaining fixation of the border, he would bring the hairline in coincidence with the termination of the enhanced region in the field. The dark band was always measured before the bright band. Figure 1 shows the stimulus presented to the subject. fn the first experiment. the bright field was kept constant while the darker field was reduced in 0.20log unit steps from the luminance of the bright field through a range of 4 log units. This was repeated for each of five luminances of the bright field, 100, 10, 1.0, 0.1 and 0.01 cd/m2. Five series of data were obtained for all these luminance combinations. To plot the data. moving averages of results
Fig. 2. Stimulus field with illuminated portion of hairline visible in the dark field. L, luminance distribution; B, perceived brightness distribution. The arrows indicate the widths measured.
for three consecutive log increments were used. In Fig. 3, error bars enclosing k 1.00 S.E. of the mean and representing moving average populations of 15 have been shown at every second datum point. In the second experiment, the luminance of the darker field was kept constant while. from this luminance, the bright-field luminance was increased in 0.20 log increments through 4 log units or less, the upper limit being lOOcd!m’. Five series of such data were compiled. The plotting method, shown in Fig. 4, is similar to that used in the first experiment. Three subjects, IR, WP and AU, participated in the project. Only the right eye was tested. The respective refractions in this eye were +0.75 D/ - 0.25 DC x 180”. + 1.25 D/ - 0.50 DC x 180’ and -4.25 D. Only subject AR had any previous experience in making measurements on the border enhancement response.
RESULTS
Fig. 1. Plan of apparatus. P,, P,, P,, projectors; ND,. ND,, neutral filter and wedge combinations; T, transformer: DS, , DS,, diffusing screens; L, lens for focusing E, razor edge, in plane of H, hairline, via SM, semisilvered mirror; LB, lens battery for focusing stimulus; A, artificial pupil; MG, microgauge.
Figure 3 shows the effect on the enhancement bands of luminance changes in the dark field. The luminance values heading the graphs refer to the constant bright field. With high luminances, there is a marked increase in the dark band through the first portion of the dark field luminance decrement range. After its initial increase, the function representing the dark band becomes rather irregular, showing troughs and peaks through the remainder of the range. An initial increase in the dark band does not always occur at low luminances, as, for example, for subject AR for the LOcd/m” level. An important observation is that, with the higher bright-field luminance& changes in the low-luminance field have little or no effect on the width of the bright band, this band remaining nearly constant through the entire decrement range. This observation does not hotd for the lowest iuminances of the brighter field, where some irregular changes in the bright band do occur with decreases in the dark-field luminance%
Brightness enhancement versus darkness enhancement 100 cd/m2
Y----e+
1cd/m2
10 cd/m2
1
1
2
0.01 cd/m:!
0.1 cd/m2
9---k++
3 4 Log luminance
w decrement
w
Fig. 3. Effect on border enhancement extent with constant bright-field luminance and decreasing darkfield luminance. Results are shown for five luminances of the bright field, 100, 10, 1.0, 0.1 and 0.01 cd/m2. IR, WP and AR refer to the subjects tested. 0, the dark portion; 0, the bright portion of the enhancement.
1234
1 Log
234 luminance
1
234
1
2
3
4
increment
Fig. 4. Effect on border enhancement extent with constant dark-field luminance and increasing brightfield luminance. Results are shown for four luminance levels of the dark field, 1, 0.1, 0.01 and 0.001 cd/m2. The subjects and symbols are the same as in Fig. 3.
109x
ARNI.L~RI:MOLI-
On the other hand, the bright-field luminances strongly affect the width of the bright band. In one subject, IR, the width of this band is a minimum for the l-cd/m’ level and increases both towards higher and lower luminances, whereas in the other two subjects. the minimum width occurs towards the highest luminances. The data agree qualitatively with previous findings (Remole, 1976) for the bright band, which showed a large increase towards low luminances for all subjects as well as a small increase also towards higher luminances for some subjects. In subject AR, the bright band increases towards low luminances to the extent that it far overshoots the dark band. whereas in the other subjects the bright band merely converges towards the dark band. Such large individual differences are typical of border enhancement responses at low luminances. They are possibly due to differences in fixation performance. For example, if the subject were to fixate slightly eccentrically at low luminances. the enhancement bands would be affected. In Fig. 4, the bright-field luminance was increased while the dark field was constant. The four headings give the luminances of the dark field. As one would expect from Fig. 3, the bright band increases with ascending luminances in the bright field in subject IR but decreases in AR. In WP. the width of this band decreases with increasing luminance for the two lower levels but shows little change for the two higher levels, the results again being in fair agreement with Fig. 3. The apparent discrepancy between subjects IR and AR with respect to the slope of the graphs is merely an illustration of the fact that some subjects show a bright band minimum at moderately high luminances, such as, for example, lOcd/m’. whereas others appear to have a minimum at much higher luminances. Again, this difference may be a reflection of the individual fixation performance. Although the two experiments in most instances agree reasonably well for identical luminance combinations, there are a few cases where the agreement is not good. For subject IR, at the dark-field level of 0.001 cd/m” in Fig. 4. the bright field increases 4 log units through 0.01, 0.1, 1.O and 10 cd/m2. We should expect a minimum for the bright band at l.Ocd/m’ for the bright field since this corresponds to the minimum for the bright band in Fig. 3. The bright band curve does dip down at the 3 log unit interval representing this luminance, but not nearly enough to agree well with Fig. 3. Perhaps in this case. the manner of fixating the border was affected somewhat by the sequence of presenting the data. If this iS the case. it is a factor that is extremely difficult to control. Contrary to the relationship between the dark field and the bright band, the dark band varies markedly with changes in the bright field. In most cases in Fig. 4, there is an initial increase in the dark band as the luminance of the bright field is increased. Again, especially towards low luminances. the individual differences become very marked. For example. for subject WP at the dark-field luminance of 0.001 cd/m’. the dark band decreases from a very large initial value. The dark band varies also with changes in the dark field. In general, this consists in an increase towards lower luminances. similar but less pronounced than
that found with the bright band. However, many individual differences are superimposed on this trend. In order to average out some of these individual differences, the means for the three subjects have been plotted in Fig. 5. The upper portion of ths figure indicates more clearly the increase in the dark band with decreasing luminance of the dark field, the greater increase occurring with the inital log reductions. Also. this is generally evident from the lower portion. where four luminances of the dark field are represented. Another type of summary table, integrating data from the two experiments. is shown in Fig. 6. Plottings representing identical contrasts have been selected from Figs. 3 and 3 and have been replotted against their corresponding luminances. The two contrasts shown are those represented by the abscissa values of 0.4 and 2.0 log units. Identical contrasts for the two bands obviously do not imply identical stimulus situations. For a given luminance of the dark field, a given contrast is produced by a certain bright-field luminance. If the given luminance refers to the bright field, the same contrast is produced by a certain darkfield luminance. Not only is the mean luminance greater in the first situation. but the contrasts are also in opposite directions. Therefore, the functions for the two bands will not coincide. However. if the changes in bandwidth on both sides of the border were symmetrical. their corresponding functions would be similar in form. Except for subject WP. for a contrast of 2.0 log difference between the two fields, the pairs of curves in Fig. 6 are quite dissimilar. thus presenting another example of assymmetry. When the bright and dark bands are compared in ‘Figs. 5 and 6, certain general trends transcend the individual characteristics. Thus. a comparison between the two bands can bc summarized as follows: (I ) for the higher luminances. the width of the bright band is hardly affected by variatiohs in the dark-field luminance, whereas the dark band increases markedly with ascending bright-field luminances. However, towards low mesopic luminances, the two bands behave more equally in this respect. (2) For photopic and high mesopic luminances, the dark band is wider than the bright band. (3) The bright band increases markedly towards low luminances of the corresponding field. and. in some subjects. slightly towards high luminances. The dark band exhibits similar but less pronounced changes as the dark-field luminance is varied. (4) Functions describing the effect of luminance on the bandwidth for constant contrasts differ in form for the two bands. Extrapolation of these functions suggests that the bright band is the larger at some luminance levels. In one subject, AR, the bright band does. indeed. become wider than the dark band at low luminances DISCUSSION
Some individual differences have been pointed out in the description of the results, and it has been suggested that small differences in the way the border is fixated may be an important factor here. There may be other factors as well, such as the rate and extent of involuntary eye movements. A third factor is the focussing performance of the subject: small fluctuations in accommodation could easily affect the blur
1099
Brightness enhancement versus darkness enhancement 0.01 cd/m?
t
::~~~;~ .E E
-+-e?ww
I-_1 Log
z
luminance
2
3
4
decrement
aI
5
234 Log
luminance
1
increment
Fig. 5. Upper portion: mean results for the three subjects in Fig. 3. Lower portion: similar mean results for Fig. 4. The symbols are the same as in Figs. 3 and 4.
of the retinal image and hence the border enhancement extent (Remole, 1974). Thus, the border enhancement response is relatively labile, and that there are irregularities in the findings is therefore not surprising. However, in spite of these difficulties, the general trends described are quite obvious. The general trends defining the differences between the two bands need to be discussed in terms of explanatory hypotheses. First, the relatively pronounced effect of the bright field on the dark band is reminiscent of the one-way effect often observed with respect to the interaction between a bright and
a dark field. Although the theory of lateral neural inhibition (Ratliff, Hartline and Miller, 1963; Motokawa, 1970) assumes mutual interaction between visual units, the brighter field is perceptually the most powerful in transmitting effects across the border. For example, the perceived brightness of a field is affected by a brighter surround, but not by a darker surround (Marks, 1974). In our study, the logarithmic luminance scale may also have contributed to this effect. The relatively stronger effect of the bright field may result in part also from the distribution of intraocular
t WP
Log luminance,
ccl/m*
Fig. 6. Effect on border enhancement of the corresponding field luminance for two constant contrasts, represented by log differences, C, of 0.40 and 2.00. The data points have been selected from Figs. 3 and 4.
scattering. Luminance thresholds near the border (Yonemura, 1962; Wildman, 1974) show that the scattered light in the retinal image of the low-luminance field increases towards the border. Such an increase is also in accordance with the distribution of light scattered from the cornea and eye lens (DeMott and Boynton. 1958: Trokel, 1962; Feuk and McQueen, 197I ). We expect_a corresponding reduction near the border of the high-illuminance portion of the retinal image. When the bright-field luminance increases, the amount of scattered light in the low-illuminance portion of the retinal image will increase and fill in the lower flexure adjacent to the distribution representing the border. This could result in a widening of the dark band. On the other hand, the corresponding changes in scattering effects on the bright side of the border, accompanying changes in the low-ium~nance field, are relatively minute and cannot be expected to affect the bright band. These two factors, the one-way effect and the scattering effect, may also partially explain why the dark band is generally the greater. However, the observation that the bands increase with decreasing field luminance, and that the bright band increases more rapidly than the dark band, cannot readily be explained in such terms. The marked widening of the bands with reduced luminance must be due to changes at the neural level. in terms of distances traveled by the inhibitory process. If there is an increase in convergence and coarseness of the neural mechanism towards lower luminances (Pirenne, 1962), the enhancement effects would be expected to spread farther as over a more thinly distributed mosaic. A parallel to these changes if found in the increased enhancement spread when the border stimulation is moved away from the fovea centralis (Remole, 1976). Here, the increase is clearly associated with the increasing neural convergence towards the retinal periphery. On the other hand, the relatively small increase in enhancement shown by some subjects for high luminances may originate from small changes in fixation or focussing performance. That the two bands increase at different rates with decreasing luminance is shown also in the constant contrast functions in Fig. 6. The simplest explanation for this difference is that the bands represent a dual system, one for brightness and one for darkness, now widely recognized in principle (Marks. 1974). Such a system has been identified with two types of receptive fields (Barlow, Fitzhugh and Kuffler, 1957; Jung, 1973; Gerrits and Vendrik, 1970; ~a~ussen and Glad. 1975).The different rates of change in the bright and dark bands may be ascribed to differences that are bound to exist between two such systems as they adjust to luminance changes. ~c~lzo~~~~g~~n~nrs-I wish to thank Dr. W. H. Pym and Dr. 1. 0. Revill for their assistance with the project. The work was supported by Grant No. A9951 from the National Research Council of Canada.
RI”.FERENCES
Barlow H. B.. Fitzhugh R. and KuH?er S. W. (1957) (~‘hangc of organization in the receptive fields of the cat’s retinn during dark adaptation. J. Pity.+/.. Lo&. 137. ?iX 3%. BCkCsy G. von (1968) Brightness distribution across the Mach bands measured with flicker photometry. and the linearity of sensory nervous interaction. .I, opt. SCU .Am
58. I-X. Bliss 3. C. and Macurdy W. B. (1961) Linear models for contrast phenomena. J. npr. SOL.&I. 53. 1373 1379. Davidson M. and Whiteside J. A. (1971) ?tuman hrightness perception near sharp contours. J. cr/~t. ,Soc, clnt. 61. SXt536. DeMott D. W. and Boynton R. M. (1958)Retinal dtstribution of entoptic stray light. J. opt. Sot. Aln. 48. 13 -22. Feuk T. and McQueen D. (1971) The angular dependence of light scattered from rabbit corneas. It7wsr. ~~~rj~~. to. 294-299. Fiorentini A. and Radici T. (1958) Brightness, width and position of Mach bands as a function of the rate of variation of the luminance gradient. .4tti Fond. Giorgio Ronchi 13, 145-155. Gerrits H. 3. M. and Vendrik A. J. H. (1970) Simuitaneous contrast, filling-in process and information processing in man’s visual system. Expt. bruin Res. 11, 41 l-430. Jung R. (I9731 Visual perception and neurophysiology. In Handbook of Sensory Physiology. Vol. VII/3 (Edited by Jung R.). Springer. Berlin. Maenussen S. and Glad A. (1975) Brightness and darkness enhancement during flicker: perceptual correlates of neuronal B- and D-systems in human vision. E.q~l. Bruin Rex 22, 399-413. Marimont R. B. (1963) Linearity and the Mach phenomenon. J. opt. Sot. .4m. 53, 40@401. Marks L. E. (1974) Srrxor,~ Processes. Academic Press, New York. Mathews M. L. (1966) Appearance of Mach bands for short durations and sharply focused contours. .I. opt. sot. Am. 56. 1401--1402. McCollou~ C. (1955) The variation in width and position of Mach bands as a function of luminance. J. e~p. Psychol, 49. 141-152. Motokawa K. (1970) Physiology c~f Colour urd Putturn Vision. Igaku Shoin, Tokyo. Pirenne M. H. (1967) Visual acuity. In The qe. Vol. 2 (Edited by Davson H.). Academic Press, New York. Ratliff F.. Hartline H. K. and Miller W. H. (1963) Spatial and temporal aspects of retinal inhibitory intcr~iction. J. opt. Sot. Al?!.53. 11(&120. Remole A. (1974) Relation between border enhancement extent and retinal image blur. Visiotl Res. 14. 9X9-995.
Remole A. (1976)Effectof retinal illuminance and eccentricitv on border enhancement extent. l’ision Rus. 16. 1353--l317. Shipley T. and Wier C. (1972) Asymmetries in the Mach band phenomena. Kybern~tik 10. 181-.1X9. Trokel S. (1962) The physical basis for transparency 01 the crystalline lens. iriu;st. Ophthd. 1. 493-501. Watrasiewicz B. M. (1966) Some factors affecting the appearance of Mach bands. J. opt. Sot. An]. 56. 53G-536. Wildman K. N. (1974) Visual sensitivity at an edge. vision Rrs. 14, 749-755. Yonemura G. T. (1962) Luminance thresonds as a function of angular distance from an inducing source. J. opf. Sot. .4m. 52. 1030-1034.
