International Dark-Sky Association (IDA) — Information Sheet #136
Some Issues in Low Light Level Vision
Introduction
The sensitivity of the eye is different at different lighting levels. At higher levels, the cones in the eye dominate. At
low lighting levels, the rods dominate. Almost all texts on the eye, photometry, or lighting discuss these differences,
some in great detail. The effects are generally well known, but they are not often taken into account in considerations
of nighttime lighting. There is now, however, a growing interest in such issues. This information sheet discusses
some of the aspects. It is only a very brief introduction to the issues of vision at low light levels.
Scotopic vision is rod vision. The rods are very sensitive, allowing us to see at very low lighting levels (below
approximately 0.01 cd/m2). The rods are located away from the fovea (center area of the retina), with a maximum
density about 10-20° off axis, while the cones are located mainly in the fovea area (about 2° in size). A side view
(averted vision, or out of the corner of the eye) is the way to maximize the use of the rods at very low lighting levels.
Peripheral vision is mainly vision with the rods. The rods have no color vision, so there is no color perception at
such very low lighting levels, no matter what the color of the light source or of the object being viewed. As the
illumination level increases, we enter the mesopic range, where both rods and cones are in use, so vision is both
peripheral and central, and color begins to become apparent. At higher illumination levels (above about 3 cd/m2), we
are in the photopic range, where cones have taken over most of the visual task. At much higher levels, the eye cannot
tolerate the intensity, and one shades one’s eyes, turns away, or otherwise avoids the very high intensity.
The edges of the retina, where rods dominate, are particularly sensitive to motion. At the lowest lighting levels, one
sees motion but little or no detail. The eye is very sensitive, but visual acuity is missing. With more illumination,
say moonlight (a maximum of 0.1 lux), there is still no color but acuity is now fairly good. At about 1 lux,
color is becoming apparent as the cones are now beginning to be used. Fovea vision is taking over, and acuity is
good. A number of the references given at the end of this information sheet give considerable detail about these
transitions of the eye.
The whole issue of vision at low lighting levels is quite complex, but it may be possible to summarize much of
it in a few statements:
At very low lighting levels (scotopic vision), the rods rather than the cones are the way the eye sees.
! The rods are not concentrated in the center of the eye (the fovea) as the cones are, so they are not much utilized
when one is looking directly at an object. Averted vision is the best way to maximize the efficacy of the rods,
as any amateur or professional astronomer knows well. You can not look directly at an object and see it as
well as if you look a bit to one side. Note that we do not drive at night, or walk for that matter, by using
averted vision, except in an attempt to avoid the adverse effects of glare.
! The rods are more sensitive in the blue than are the cones. Much of what has appeared in the recent literature
is related to that fact. Authors discuss the two visual relative spectral sensitivity curves, represented by the
two CIE V( ) curves: V( ) for the photopic sensitivity function (cones) and V´( ) for the scotopic sensitivity
function (rods). These functions are only valid for the conditions under which they were determined. The
rods have no color sensitivity; you cannot see colors at scotopic light levels. This effect is discussed
in greater detail below
continued
© IDA, Inc. • 3225 N. First Avenue • Tucson, Arizona 85719-2103
520-293-3198 (voice) • 520-293-3192 (fax) • Web: www.darksky.org • E-mail: ida@darksky.org
1
06/96
!
Visual acuity is different for the rods and the cones, with the cones having considerably more visual acuity.
For sharp vision, we need to be in or near the photopic vision levels and looking right at the object (within the
2° cone of the fovea). With the rods, we see less detail, but we do see motion well.
!
It is important to remember that for all visual levels, everything possible should be done to eliminate glare.
Glare never helps visibility. Remember also never to overlight. Minimize the direct uplight and wasted light.
Such principles are as valid at the lowest lighting levels as at photopic levels. The eye can see very well at
quite low lighting levels if it is not compromised by glare or by nearby areas that are overlit.
Glare is a very important issue. In the presence of glare, the iris in the eye closes down. Vision at low lighting levels
is not as good as without the glare source. Another important issue is transient adaptation, which refers to the time it
takes the eye to adjust in going from bright levels of lighting to faint ones, and vice versa. While the eye can see over
a very wide range of lighting levels, it can’t accommodate them all at once: it needs time to adjust to changing levels.
One must also be conscious of the level of the ambient lighting. All of these items, and others, must be taken into
account in investigating vision at low lighting levels. No one item operates in isolation from the others.
To summarize, in very faint light the eye is using the rods more than the cones, the rods have poor sensitivity
right at the center of the eye’s viewing angle (at the fovea), they are more sensitive in the blue than the yellow-as
opposed to the cones, and they offer considerably less visual acuity than do the cones. Vision is different at low
lighting levels than at high levels. Avoid glare, and do not overlight. For optimal visibility at night, it is necessary
to avoid glare and overlighting.
The Color Shift in Going from Photopic to Scotopic Vision
We will summarize this issue by quoting from a recent paper by Werner Adrian, who has done extensive research
on low light level vision. Adrian is a professor at the University of Waterloo, in Waterloo, Ontario, Canada, and
is an active member of all the IESNA and CIE technical committees that deal with these issues. He is also one
of the renowned experts on glare, active in all the IESNA and CIE committees on that issue as well. He is also
a longtime member of the IDA.
With decreasing luminance levels, the spectral sensitivity of the eye changes and becomes more blue-sensitive. This
goes in concert with the fading of colors until, at low levels, the eye perceives only brightness. It can be expected
that the basic visual functions such as contrast sensitivity, visual acuity, and performance, as well as the pupil size,
are dependent on the perceived brightness and not on the photopic luminance, which in the range of mesopic and
scotopic vision is an inappropriate measure as it is based on V( ) only. A photometric quantity that accounts for the
changing spectral sensitivity of the eye is called the Equivalent Luminance, Leq. An algorithm has been developed
to allow for its direct calculation. The results show that the brightness expressed in terms of Leq indeed controls
the visual functions. Consequently, the equivalent luminance for light of different spectral power distributions can
be calculated for levels of mesopic vision to show the effect on visual performance. As the spectral sensitivity of
the eye is shifting to the blue with lower light levels, light with a blue rich power distribution appears brighter and
allows higher visual performance to be achieved.
After considerable research and discussion, he concludes:
It has been shown that the basic visual functions such as contrast sensitivity, visual acuity (resolution of
detail), and visual performance which appears to be a composite of visual acuity and time required for the task are
determined by the perceived brightness in which these functions are measured. In photopic levels, the brightness is
expressed by luminance. In mesopic and scotopic levels, the equivalent luminance reflects the perceived brightness
and appears to be the physiologically reasonable and appropriate measure. The results suggest that the equivalent
luminance is suitable to predict the achievable level of visual performance in various spectral light distributions.
continued
2
The table given below summarizes the results as a figure of merit for each distribution given as a function of the
lighting level. Metal halide lamps achieve the highest figure of merit, but this is offset by the efficacy of the lamps,
so that the lowest figure of merit lamp, the low pressure sodium lamp, appears to be even with the metal halide
at a lighting level of 0.1 cd/m2.
Table I. Relative power in watts based on 100 lm at 10 cd/m2 to achieve equal visual performance (lower values
in the table mean better efficacy):
Source
10-4
MH
Hg
HPS
LPS
1.38
2.79
3.12
2.66
Luminance level, cd/m2
10-3
10-2
10-1
Scotopic
1.40
1.40
1.44
2.81
2.73
2.50
2.78
2.03
1.48
2.78
2.42
1.39
1
3
Photopic
1.52
2.06
0.96
0.75
1.49
2.16
1.13
1.01
10
1.6
2.1
0.9
0.6
Lower numbers reflect better efficiency. So at any lighting levels above 0.1 cd/m2, the LPS lamp offers the best
energy efficient performance, while at lower levels, the metal halide lamp dominates.
John Bullough, of the Lighting Research Center, Rensselaer Polytechnic Institute, Troy, NY, also addressed the issue
at the USNC/CIE meeting in Cleveland, Ohio, on 1 Nov 1997. He noted the requirements for mesopic photometry:
One must represent the combined activity of rods and cones and adhere to combined additive and proportionality
laws (0.5 lumen + 0.5 lumen = 1.0 lumen). While acknowledging that much more research is needed, he presented a
table giving the efficacies for different light sources at different lighting levels. It is summarized below. The relative
figures of merit differ somewhat from those in Adrian’s study, but the conclusions are about the same.
Table II. Reported efficacies for photopic, scotopic, and three levels of mesopic luminances.
Efficacies relative to HPS are given in ( ):
Source
Photopic
Mesopic
efficacy
2
HPS
MH
Incand
Clear Hg
LPS
95 (1.00)
78 (0.82)
15 (0.15)
52 (0.55)
180 (1.89)
0.3 cd/m
86 (1.00)
99 (1.15)
17 (0.20)
58 (0.57)
136 (1.58)
(lm/watt)
2
0.1 cd/m
74 (1.00)
115 (1.54)
19 (0.25)
62 (0.84)
89 (1.20)
Scotopic
2
0.03 cd/m
66 (1.00)
125 (1.89)
20 (0.30)
65 (0.98)
58 (0.99)
61 (1.00)
130 (2.12)
20 (0.33)
67 (1.09)
41 (0.66)
Here higher numbers reflect better efficiency, while the numbers in parentheses compare the relative efficiencies
with HPS taken as 1.00. If all the assumptions that went into the calculations are valid, one sees that LPS is the
most efficacious source down to lighting levels a bit above 0.1 cd/m2, and below lighting levels of that amount,
metal halide is the most efficacious source. At levels of 0.03 cd/m2, HPS, LPS, and Hg are about equal and metal
halide is twice as efficacious.
We note that the luminances recommended by the IESNA for various types of street lighting range from 0.3 to 1.0
cd/m2 (or about 3 to 10 lux). So a value of 0.1 cd/m2 is actually lower than the IESNA recommended levels
for street lighting. It appears that at the lowest of these street lighting levels (0.3 cd/m2), LPS is still the most
efficient lamp, just as it is at higher levels. Note, however, that these IESNA recommended lighting levels are not
always followed by communities, due to cost or other reasons. In Adrian’s table, at the level of 0.1 cd/m2, MH, HPS,
continued
3
and LPS are about equally efficient. Naturally, in “real life” the lighting situation is much more complicated (for
example, the effects of glare must be fully incorporated), and it is clear that much more research into these interesting
issues is both needed and most welcome.
Conclusion
While much more research is needed on understanding vision at scotopic and mesopic lighting levels,
it is clear that such vision is different than at photopic lighting levels. The rods and the cones in
the eye behave very differently. These differences should be taken into account for any nighttime
lighting designs. We must also carefully address all adverse environmental issues in such designs.
Quality outdoor nighttime lighting is important, and it should be well understood and well designed.
References
Some references are given below, in addition to the two papers cited above, which are listed first:
Adrian, W., 1997. The Influence of the Spectral Power Distribution for Equal Visual Performance in Roadway
Lighting Levels. Paper given at the International Lighting Conference in Durban, South Africa, September 1997,
whose theme was Lighting in Developing Countries.
Bullough, J., 1997. Mesopic Photometry: Issues and Implications. Paper given at the United States National
Committee of the CIE meeting in Cleveland, Ohio, November 1997.
Crawford, D., 1994. Terminology and Units in Lighting and Astronomy. IDA Information Sheet No. 99.
Erhardt, L., 1977. Radiation, Light, and Illumination.
He, Y., M.S. Rea, A. Bierman, and J. Bullough, 1997. Evaluating light source efficacy under mesopic conditions
using reaction time. Journal of the Illuminating Engineering Society, 26 (1), 125.
IESNA Lighting Handbook, Reference and Application, 8th Edition, 1993.
Kinney, JAS, 1958. Comparison of scotopic, mesopic, and photopic spectral sensitivity curves.
Opt Soc Am, 48, 185-190.
Lewis, A.L., 1995. Visual performance as a function of spectral power distribution of light sources at luminances
used for general outdoor lighting. Report to the LRI, New York.
Loe, D.L., and I.M. Waters, 1973. Visual performance in illumination of differing spectral quality. Environmental
Research Group, University College London.
McGowan, T. and M.S. Rea, 1995. Visibility and spectral composition: Another look into the mesopic. 70 Years
of CIE Photometry, CIE, Vienna.
Rea, M.S., Y. He, and A. Bierman, 1997. Toward a system of mesopic photometry based upon M-channel
response. Visual Scales: Photometric and Colorimetric Aspects, National Physical Laboratory/CIE-UK, Teddington.
Adrian’s paper lists many other references.
4