1
Vol.:(0123456789)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports
Melanopic stimulation does
not alter psychophysical threshold
sensitivity for luminance flicker
Joris Vincent1*, Edda B. Haggerty2, David H. Brainard3 & Geoffrey K. Aguirre2
In addition to the rod and cone photoreceptors the retina contains intrinsically photosensitive retinal
ganglion cells (ipRGCs). These cells express the photopigment melanopsin and are known to be
involved in reflexive visual functions such as pupil response and photo-entrainment of the circadian
rhythm. It is possible that the ipRGCs contribute to conscious visual perception, either by providing
an independent signal to the geniculo-striate pathway, or by interacting with and thus modifying
signals arising from “classical” retinal ganglion cells that combine and contrast cone input. Here,
we tested for the existence of an interaction by asking if a 350% change in melanopsin stimulation
alters psychophysical sensitivity for the detection of luminance flicker. In Experiment 1, we tested
for a change in the threshold for detecting luminance flicker in three participants after they adapted
to backgrounds with different degrees of tonic melanopsin stimulation. In Experiments 2 and 3, this
test was repeated, but now for luminance flicker presented on a transient pedestal of melanopsin
stimulation. Across the three experiments, no effect of melanopsin stimulation upon threshold flicker
sensitivity was found. Our results suggest that even large changes in melanopsin stimulation do not
affect near-threshold, cone-mediated visual perception.
Under daylight conditions, signals originating in the cone photoreceptors pass through several classes of retinal
ganglion cells to support visual perception. Also active in daylight are the melanopsin-containing, intrinsi-
cally photosensitive retinal ganglion cells (ipRGCs). The ipRGCs mediate numerous physiologic effects of light,
including variation in pupil size and photoentrainment of the circadian rhythm1,2. Beyond these “reflexive”
visual functions, several studies have examined if signals from the ipRGCs contribute to visual perception (for
a comprehensive review, please see3. When human observers are presented with stimuli that include an increase
in melanopsin stimulation, participants report that the spectral change appears as increase in “brightness”4–11.
Studies of this kind support the claim that ipRGC signals have a direct effect upon perception.
In addition to this direct effect, signals from the ipRGCs may interact with those carried by the cones, produc-
ing effects upon perception by an indirect mechanism (Fig.1a). Such interactions could occur within the ipRGCs
themselves, as these cells receive signals from the cones. In addition, in both the primate and rodent retina, a
subset of ipRGCs send recurrent axon collaterals to the inner plexiform layer12 where they are hypothesized
to influence the sensitivity of cone inputs to the “classical” RGCs (i.e., midget, parasol, small bistratified). This
provides another possible site for an indirect effect of melanopsin on signals originating in the cones. Electro-
retinogram (ERG) recordings in mice support this idea, as b-wave responses to cone-directed light flashes are
reduced in amplitude and latency with melanopsin-directed stimulation13,14. A third potential point of interaction
is present in the projections of the ipRGCs to the lateral geniculate nucleus (LGN), where they signal overall
retinal irradiance in both a tonic and phasic manner4,15–17. Therefore, signals from the melanopsin-containing
ipRGCs could interact with the classical RGC pathways downstream from the retina.
Here we examine if the threshold for detection of flickering light directed at the luminance channel is altered
by modulation of the melanopic excitation produced by the stimulus background. Over three experiments,
we measured the threshold at which human observers could detect a 5Hz modulation of cone contrast. In
each experiment, measurements were made under conditions of low- and high-melanopic excitation, with the
experiments differing in the timing of the melanopsin modulation, and in the means by which the stimuli were
produced. We choose 5Hz flicker as we expect that this stimulus will predominantly drive the conventional
RGCs (e.g., the parasol and midget classes), and less so the relatively sluggish ipRGCs. We find no evidence in
OPEN
1Computational Psychology, Institute of Computer Engineering and Microelectronics, Technische Universität
Berlin, Berlin, Germany. 2Department of Neurology, Perelman School of Medicine, University of Pennsylvania,
Philadelphia, PA, USA. 3Department of Psychology, School of Arts and Sciences, University of Pennsylvania,
Philadelphia, PA, USA. *email: joris.[email protected]
2
Vol:.(1234567890)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
our measurements that the change in melanopsin excitation alters the detection of luminance flicker beyond
what could be accounted for by imprecision in stimulus generation. These results constrain the ways in which
signals from classical and melanopsin-containing RGCs might be theorized to interact.
6 mm artificial
pupil
Pharmacologically
dilated pupil
3 primary
DLP
beamsplitter
OneLight
digital light
integrator Focusing
lens
Macular
blocker
Diffuser
Wavelength (nm) Wavelength (nm)
Low (dark gray) and high (red) melanopic SPDs LMS flicker SPDs, +-5% contrast
Spectral Radiance (W/sr/m2/nm)
0.00
0.01
0.02
0.03
400 500 600 700 800
0.00
0.01
0.02
0.03
400500 600700 800
Cones RGCs
1
2
3
LGN
Spectral Radiance (W/sr/m2/nm)
Figure1. Theoretical background and experimental design. (a) Retinal wiring schematic. There are several
hypothesized locations at which signals from the melanopsin-containing ipRGCs (blue) might interact with
signals from the cones, which are predominantly conveyed by the classical RGCs (gray). (1) Recurrent axon
collaterals from ipRGCs can provide inhibitory signaling on cone pathways; (2) ipRGCs can project to LGN cells
that also receive inputs from the classical RGCs; and (3) cone and melanopsin signals may interact within the
ipRGCs themselves. (b) Spectral sensitivities of the L, M and S cones and of the melanopsin-containing ipRGCs.
Each sensitivity is normalized to a peak of unity. (c) Left: spectral radiance of the low- (dark gray) and high-
(red) melanopic backgrounds. Right: example spectral modulations for LMS directed flicker. The dark gray line
is the spectral radiance of the low-melanopic background (same as in the panel on the right). The black and light
gray lines provide the spectral radiance of the low and high (± 5% contrast) arms of the LMS directed flicker
around this background. (d) Apparatus schematic. Light produced by the digital light integrator was passed
through a randomized optical fiber and lens assembly (not shown) and was imaged on a diffuser, to produce
a spatially uniform field. Light from a 3 primary DLP projector was optically admixed through a beam splitter
with the light from the digital light synthesizer (lenses and neutral density filters not shown). The participant
viewed the combined stimulus through a lens, such that the diffuser subtended 27.5° (diameter), with the central
5° (diameter) macular region occluded.
3
Vol.:(0123456789)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
Methods
Overview and participants. Psychophysical sensitivity to flicker directed at the L, M, and S cones was
measured across two conditions that differed in melanopic excitation. Measurements were made in three sepa-
rate experiments. The three experiments were conceptually similar, but differed in the details of the stimuli. A
total of six participants contributed data. All participants had normal color vision as assessed by the Ishihara test
for color deficiencies. In Experiment 1, the three participants were the three authors of the study (P1, P2, P3), all
male, ages 48, 58, and 26years old, respectively. Three participants naïve to the hypotheses of the study (P4, P5,
P6) were also recruited (2 male, ages 29, 25, 23, respectively) and participated in Experiment 2. Experiment 3
obtained data from participants P3, P4, and P5. This study was approved by the University of Pennsylvania Insti-
tutional Review Board, and studies were performed in accordance with the Declaration of Helsinki. Informed
consent was obtained from all participants. Due to an oversight, the three investigators who also served as par-
ticipants completed their consent documents after data collection. These investigators have also provided written
consent to the possibility that they may be identifiable within the data set.
Varying background melanopic excitation. In each experiment, a flicker stimulus was presented
around one of two different background light spectra that varied in their melanopic excitation (low or high).
The pair of low- and high-melanopic spectra were designed through silent substitution18,19 to target melanopsin
while attempting to silence the L, M and S cones despite their overlapping spectral sensitivities (Fig.1b,c).
Our approach to generating photoreceptor-directed spectral modulations has been described previously18,20.
Briefly, cone fundamentals were based on the International Commission on Illumination (CIE) physiological
cone fundamentals21. The CIE standard only specifies fundamentals up to field sizes of 10°, so we obtained esti-
mates for our 27.5° field by extrapolation of the formula. A 6mm pupil was assumed. Spectra were tailored to
account for the age-predicted lens density for each participant. We elected not to adjust further the stimuli for
individual observers via a psychophysical nulling procedure, as we have previously found that the test–retest
reliability of nulling values produced by individual observers is not high22, and such adjustment reduces the
available gamut in our modulations. The melanopic backgrounds and pedestals were generated with a One-
Light digital light synthesizer (VISX Spectra), controlled by an Apple Macbook Pro, using custom MATLAB
(Mathworks) software. The change from the low- to high-melanopic background constituted a nominal 350%
increase melanopsin excitation. In Experiments 1 and 2, the OneLight was also used to produce cone-directed
flicker (Fig.1c—right). The luminance of the stimulus field for the background spectrum varied somewhat from
session to session but was on the order of 275cd/m2.
Stimulus delivery. The spectral output of the OneLight was imaged onto a custom diffuser to produce a
circular, uniform field of 27.5° diameter. Because spatial variation in macular pigment produces variation in
photoreceptor spectral sensitivity, the central 5° (diameter) of the field were obscured with black ink. We consid-
ered the possibility that scattered light from the stimulus field onto the obscured macular region could confound
the measurement. To minimize this, steady light provided by an HP Notebook Companion digital light projector
(DLP) was admixed with the optical stimulus from the OneLight (Fig.1d). The admixed light consisted of a cen-
tral spot overlapping with, and slightly extending past, the central 5°; and an annulus just inside of, and extend-
ing outward from, the outer edge of the 27.5° stimulus field. The pixels of the DLP aligning with the central
blocker and the outer stimulus were set to full-white (RGB input settings = [1 1 1]), and everywhere else to off ([0
0 0]). In the center of this white central region, the DLP output was set to form a small, red cross that participants
were instructed to use as a fixation target. The output intensity of the DLP was reduced by passing the projector
beam through 3.3-log units of neutral density filter, then imaged onto a custom diffuser screen to form an image,
which was then admixed with the optical signal from the OneLight using a beam splitter. The admixed light from
the “on” region of the projector image was ~ 11cd/m2. In informal observations we found that this was sufficient
to render the flicker stimulus undetectable in the central region, even at the highest contrast level used. In sum,
the participant viewed the full 27.5° annular stimulus region produced by the OneLight stimulus component,
with the central 5° and region exterior to the annulus masked by admixed white light. Supplementary FigureS1
illustrates the spatial elements of the stimulus field.
Spectroradiometric measurements of the stimulus field were taken with the admixed steady light. Measure-
ments integrated over a 1° portion of the stimulus, centered at an eccentricity approximately halfway between
the obscured macular region and the outer edge of the stimulus field. These measurements were made with the
projector light present and absent, and with the OneLight full-power output present or absent. The median
spectrum added by the projector light, across conditions, was taken to represent the scatter of the projector light
onto the stimulus field. During pilot testing, we found that the projector spectrum contribution to the stimulus
region was 2.63cd/m2. This scatter from the admixed steady light from the projector onto the stimulus field was
accounted for when generating stimuli of a specified contrast.
Prior to data collection, the right eye of the participant was pharmacologically dilated with a 1% tropicamide
ophthalmic solution following administration of 0.5% proparacaine as a local anesthetic. Subjects viewed the
stimulus field through a 6mm diameter artificial pupil.
Psychophysical sensitivity task. Psychophysical sensitivity to flicker was measured using a two-interval
forced-choice task. On each trial, one of two temporal intervals (the target interval) contained the target stimu-
lus, which was a sinusoidal flicker (5Hz) designed to stimulate the L, M and S cones. The other (reference)
interval contained no modulation. Each interval was 500ms in duration, and the start of each was indicated with
an auditory cue. A 500ms inter-stimulus interval identical to the reference interval was presented in between
4
Vol:.(1234567890)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
the stimulus intervals. Both the target and reference stimuli were refreshed at the same nominal frame rate, to
minimize the possibility that an artifact related to stimulus updating would provide a cue to the target interval.
After the intervals, the participant indicated by keypad button press their judgment as to which interval
contained the target. The response interval was untimed, and participants were allowed take a break or blink
before responding; the next trial started immediately after each response. Participants did not receive feedback.
The contrast of the flicker stimulus was varied from trial to trial through three interleaved adaptive staircases.
A correct response by the participant led to a decrease in nominal flicker contrast, while 3, 2, or 1 consecutive
incorrect responses (depending on the staircase for that trial) led to an increase. The participant was asked to hold
their eye in the eyepiece throughout a given experimental condition to maintain adaptation to the background.
During data collection, both the investigator (author JV) and the participant were masked to the order of
the adaptation conditions in that session, the flicker contrast on any trial, and which interval contained the tar-
get stimulus. Particularly for those participants who were also investigators, it is possible that familiarity with
the perceptual quality of the different conditions might have nonetheless allowed the participant to identify a
particular condition.
Threshold estimation. Pyschophysical sensitivity was assessed by measurement of the threshold contrast
of a detectable flicker stimulus. Data for each condition for each session separately were fit with a cumulative
Weibull psychometric function using the Palamedes 1.8.2 toolbox23 for MATLAB for maximum likelihood fit-
ting. These psychometric functions express the probability of detecting the flicker stimuli as a function of flicker
contrast (Experiments 1 and 2), or of flicker stimulus RGB value (Experiment 3). The Weibull functions were fit
to the data from all 120 trials in a block; guess rate was fixed at 50% for the two-interval forced-choice task, lapse
rate was free to vary between 0 and 5%, and the slope parameter was unrestricted.
Threshold flicker contrast was estimated, separately for each condition for each session, from these psycho-
metric functions, as the contrast corresponding to 70.71% correct detection. The 4 (or 6) repeated sessions (for
Experiments 1 and 2, and for Experiment 3, respectively), yielded 4 (or 6) estimates of flicker detection threshold
per participant for each melanopsin condition.
To quantify measurement uncertainty, a 95% confidence interval (CI) for this median was calculated across
four (or six) measurements, based on the interquartile range (IQR): CI = median ± (1.58 × IQR)/sqrt(n)24. This
calculation was performed in R25, using the built-in median and IQR functions. Further information about the
algorithm used to calculate the IQR can be found in the documentation for the quantile function25. R and the
tidyverse package26 were used to generate figures.
Experiment 1. Experiment 1 measured the effect of adaptation to different levels of steady background
melanopic content on sensitivity to cone-directed flicker. Flicker sensitivity was tested on a low-melanopic back-
ground and a high-melanopic background in separate blocks of trials, with each block starting with 5min of
adaptation to the background (Fig.2a, upper panel). As part of Experiment 1, similar measurements were car-
ried out for low and high cone-directed backgrounds (with no change in melanopsin stimulation). This served
as a positive control, demonstrating that the experimental manipulation leads to detectable Weber-like changes
in sensitivity with background cone content (see Supplementary Information).
The target cone-directed flicker in this experiment was designed using silent substitution and delivered by
the OneLight. For each background, a spectrum was designed to deliver 5% contrast in the positive direction
(increase in excitation compared to background), and another to deliver 5% contrast in the negative direction
(decrease in excitation compared to background), directed at the LMS cone photoreceptors (while keeping
melanopsin excitation constant). The desired contrast of the cone-directed flicker on each trial was produced
by scaling the two modulation spectra following the staircase procedure, providing a nominal contrast range of
0 to 5%, in 0.1% contrast steps. Sinusoidal variation in the relative weight of these components then produced
flicker nominally directed at just the cone photoreceptors with the specified contrast. The background remained
constant across all trials in a block (Fig.2a, lower panels).
Each session consisted of four blocks, each testing one background condition: the low- and high-melanopic
backgrounds, as well as low and high cone-directed backgrounds that served to provide a positive control. The
four blocks (one for each background) were acquired in pseudo-random order in each session. The participant
adapted to each background for five minutes at the start of each block, and then completed 120 flicker-detection
trials. Each participant completed 4 full valid sessions.
Experiment 2. Experiment 2 measured the effect of a pulsed pedestal of melanopic content on sensitivity
to cone-directed flicker. Flicker sensitivity was tested on interleaved trials on either a steady low-melanopic
background, or on a pulsed high-melanopic pedestal (Fig.3a, upper panel). Each session started with 5min
of adaptation to the low-melanopic background, followed by 120 trials of each type (240 total) presented in
pseudo-random order.
Half of the trials contained a pulsed melanopsin pedestal on which both stimulus intervals were presented
(high-melanopic condition). The high-melanopic pedestal was windowed by a 500ms half-cosine ramp. After
250ms, the first interval was presented. After the second interval was presented, there was a 250ms delay and
then the melanopic content cosine-ramped back down to the low-melanopic background over the course of
500ms. On the other half of trials the stimulus intervals were presented on a steady background with low-mel-
anopic content (low-melanopic condition). On the trials without the melanopic pedestal, the steady background
was presented for the combined 750ms before the first interval, as well as for the combined 750ms after the
second interval (Fig.3a, lower panels).
5
Vol.:(0123456789)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
Time
1
4.5
Melanopic adaptation level
adaptation (5 min) trial 1
(2s)
... trial 120
(2s)
...
- - -
- - -
\\
adaptation (5 min) trial 1
(2s)
trial 120
(2s)
1 4.5 1 4.5 1 4.5
Melanopic adaptation level
LMS JND, normalized
Participant 1 Participant 2 Participant 3
Melanopic adaption level
LMS contrast
Time (s) LMS contrast
Time (s)
Melanopic adaption level
-0.3
-0.2
-0.1
0
0.1
0.2
0
0.5
1
1.5
2
2.5
3
1
1.5
2
2.5
3
3.5
4
4.5
-0.3
-0.2
-0.1
0
0.1
0.2
0
0.5
1
1.5
2
2.5
3
1
1.5
2
2.5
3
3.5
4
4.5
−1
0
1
JND difference
0
1
2
3
4
5
−1
0
1
JND difference
−1
0
1
JND difference
Figure2. Experiment 1 stimulus and results. (a) Each block contained only a single level of melanopic background
adapting stimulus (vertical axis gives normalized level: 1 or 4.5) which remained constant through the entirety of that
block. Blocks started with an adaptation period of 5min, during which the participant fixated on the background for
that block. Participants then completed 120 trials of a 2IFC task. Individual trials consisted of two 500ms intervals,
each followed by a 500ms ISI. During either the first or second interval, sinusoidal flicker (5Hz) directed at the L,
M, and S cones was presented around the melanopic background (low-melanopic background case shown on left,
high-melanopic background case shown on right); during the other interval no such flicker was presented. After both
intervals were presented, participants were asked to indicate which of the two intervals contained the flicker—this
response was untimed. Intervals were indicated with an auditory cue. Participants could take a variable-length break
in between blocks. Block-order was pseudorandom. (b) Experiment 1 results. The plots show cone-directed flicker
detection on low- and high-melanopic backgrounds, for the three participants in Experiment 1. Dark gray lines
express detection as Just-Noticeable-Difference (JND) from the background in four single sessions. All JNDs for a
participant were normalized to that participant’s median JND on the low-melanopic background. If detection of cone-
directed flicker was mediated by melanopic stimulation in a Weber’s law fashion, JNDs on the 4.5× higher melanopic
background would be 4.5× higher than on the low-melanopic background (dashed horizontal line). Insets: the four
red dots show the within-session differences between JNDs at low- and high-melanopic adaptation level. The thicker
horizontal line and gray whiskers show the median and confidence interval of the four within-session differences.
6
Vol:.(1234567890)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
Time
1
4.5
Melanopic level
adaptation (5 min) trial 1
(3.25s)
... ... ... ...
trial 240
(3.25s)
- -
-
-
- -
1 4.5 14.5 1 4.5
Melanopic level
LMS JND, normalized
Participant 4 Participant 5 Participant 6
Melanopic level
LMS contrast
Time (s) LMS contrast
Time (s)
Melanopic level
-0.3
-0.2
-0.1
0
0.1
0.2
0
0.5
1
1.5
2
2.5
3
1
1.5
2
2.5
3
3.5
4
4.5
-0.3
-0.2
-0.1
0
0.1
0.2
0
0.5
1
1.5
2
2.5
3
1
1.5
2
2.5
3
3.5
4
4.5
0
1
2
3
4
5
−1
0
1
JND difference
−1
0
1
JND difference
−1
0
1
JND difference
Figure3. Experiment 2 stimulus and results. (a) Temporal structure of trials in Experiments 2 and 3, where the target
flicker was presented with or without a melanopsin-directed pedestal. Sessions started with an adaptation period of 5min to
the low-melanopic background. Participants then completed 240 trials of a 2IFC task, half of which contained a pedestal of
melanopic stimulation. Individual trials consisted of two 500ms intervals, separated by a 500ms ISI. On trials containing the
pedestal, melanopic stimulation rose during a 250ms cosine window, and stayed high for the two intervals and ISI. During
either the first or second interval, sinusoidal flicker directed at the L, M, and S cones was presented around the melanopic
pedestal; on the other interval no such flicker was presented. After both intervals were presented, melanopic stimulation
ramped back down to the low-melanopic background during another 250ms cosine window, after which the participant
was asked to indicate which of the two intervals contained the flicker—this response was untimed. On trials without the
melanopic pedestal, melanopic stimulation stayed at the low-melanopic background level for the duration of the trial,
including the 500ms preceding the first interval. Intervals were indicated with an auditory cue. Participants could take a
variable-length break in between blocks. Trial order was pseudorandom. (b) Experiment 2 results. Same format as Fig.2b.
7
Vol.:(0123456789)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
The cone-directed flicker in this experiment was produced by the same method as in Experiment 1. Each
session consisted of a single block of interleaved trials. Each participant completed 4 full, valid sessions.
Experiment 3. Experiment 3 also measured the effect of a pulsed pedestal of melanopic stimulation on
sensitivity to cone-directed flicker, but in Experiment 3 the target flicker was produced by the DLP projector
optically admixed with the stimulus background/pedestal component produced by the OneLight. The target was
thus a light-flux modulation rather than strictly cone-directed modulation. Flicker sensitivity was measured on
either a steady low-melanopic background, or on a pulsed high-melanopic pedestal, on separate but interleaved
trials (as in Experiment 2). Each block started with 5min of adaptation to the low-melanopic background, fol-
lowed by 120 trials of each type (240 total) presented in random order.
Melanopsin stimulation was produced in the same manner as in Experiments 1 and 2. The cone-directed
light flicker was provided by the HP Notebook Companion digital light projector (DLP), under control of
the same computer as the OneLight. The DLP was set at its half-on level (input settings = [0.5 0.5 0.5]) in the
annular stimulus region, providing a DLP background component around which the flicker modulated. The
5Hz sinusoidal flickering stimulus around this background was produced by time-varying the settings of the
DLP both in the positive direction (increase in excitation compared to background) and the negative direction
(decrease in excitation compared to background). The maximum excursion, i.e., the change in input settings,
was adjusted between trials.
The primary interest in Experiment 3 was to confirm the results of Experiment 2, but with a different stimulus
generation method. In this stimulus generation method, the same flicker stimuli from the projector were admixed
across the introduction of the melanopic pedestal. This did not require precise control over the projector, as the
goal was simply to make the test stimulus independent of the background/pedestal. In particular, no lineariza-
tion of the DLP stimulus was performed. Rather, the RGB input to the projector was varied at a nominal 8-bit
quantization, with the R, G and B inputs always set to the same value as each other. Additional quantization may
have been performed internally to the DLP. Flicker stimuli were drawn at a 60Hz refresh rate.
As described above, the output intensity of the DLP was reduced by passing it through 3.3-log of neutral
density filter, then imaged onto a custom diffuser which acted as a projection screen, and finally admixed to
the optical signal from the OneLight by passing it through the beam splitter. The participant saw the full 27.5°
annulus stimulus region as a mix of the OneLight stimulus component, and the DLP stimulus component, with
the central 5° and outer region masked by a dim white light from the DLP.
The dynamics of the melanopsin stimulation and the trial design were the same as in Experiment 2 (see
Fig.3a). Each session consisted of a single block of interleaved trials (120 trials with and without pedestal, 240
total). Each participant completed 6 validated sessions.
Results
In each of three experiments, sensitivity to cone-directed flicker was measured in low and high melanopsin
stimulation conditions. In Experiment 1, melanopsin stimulation was varied by adaptation to either a low- or
high-melanopic background, and the detection task was completed for each condition separately. In Experiments
2 and 3, melanopsin stimulation was varied on individual, intermixed trials by presenting the cone-directed
flicker around a low-melanopic background, or upon a high-melanopic pedestal. In Experiments 1 and 2, a single
device (the OneLight) was used to both modulate melanopic excitation and to produce cone-directed flicker.
In Experiment 3, the target flicker was produced by a conventional, 3 primary projector, and optically admixed
with the background/pedestal produced by the OneLight.
In Experiment 1 and 2, threshold was measured as the estimated contrast for 70.71% correct detection of the
flicker stimulus, separately for each melanopsin condition of each session. Table1 presents the LMS threshold
contrasts in each melanopsin condition, expressed as the median across sessions per participant. Supplementary
FigureS2 provides the fits of the psychometric functions to the detection accuracy data for each participant,
session, and experiment. Note that the staircase method employed is designed to place trials near threshold, and
thus provide data that constrain the threshold estimate. This method does not, however, optimize the estimate
of the slope of the psychometric functions.
Table 1. Participant flicker LMS contrasts at detection threshold. Median (over sessions) threshold contrast
for LMS flicker in low, and high melanopsin conditions.
Low melanopsin condition (median contrast) (%) High melanopsin condition (median contrast) (%)
Experiment 1
Participant 1 1.35 1.21
Participant 2 1.03 1.02
Participant 3 1.35 1.39
Experiment 2
Participant 4 2.04 1.62
Participant 5 2.14 1.53
Participant 6 1.53 1.12
8
Vol:.(1234567890)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
Additionally, sensitivity can be expressed as the Just-Noticeable-Difference (JNDs) between flicker stimulus
and background. Multiplying contrast values by the LMS content of the background spectrum gives the increase
and decrease in LMS stimulation produced by the flicker stimulus. Thus, multiplying threshold contrast values
by background LMS content expresses detection threshold as the difference in LMS stimulation between back-
ground and stimulus that is just noticeable.
In visualizations of results (Figs.2b, 3b), JNDs are normalized such that the median JND in the low mel-
anopsin stimulation condition has a value of unity. This allows for the interpretation of the JNDs in the high
melanopsin condition as a scalar increase of JNDs in the low melanopsin condition. The change in melanopic
contrast was 350%; thus melanopsin stimulation was increased by a factor of 4.5× between the low- and high-
melanopic backgrounds.
In an initial, positive control experiment, we measured the effect of a 4.5× increase in the luminance of
the background (L + M + S) upon the detection of superimposed luminance flicker. In agreement with Weber’s
Law, this change in the background produced a roughly 4.5× increase in JNDs for the three participants (Sup-
plementary FigureS3). In the experiments below, we ask what effect a 4.5× change in melanopsin contrast has
upon JNDs for detection of luminance flicker. We might consider the Weber’s Law effect as a (perhaps unlikely)
upper bound on how much a change in melanopsin excitation might indirectly influence luminance perception.
The effect of melanopsin on detection of cone-directed signals was investigated within participants, by vary-
ing the melanopsin stimulation within each session. Thus, to determine the presence or absence of an effect
of the melanopsin stimulation, the difference between JNDs in the high and low melanopsin condition was
calculated for each session (Figs.2b, 3b, insets). If melanopsin has no effect on detection, the JNDs would not
differ systematically between melanopic conditions, and these within-session differences would cluster around
0. We provide the median (and 95% confidence interval) across the four measurements in each participant as
the central tendency of the effect.
Experiment 1. In Experiment 1, the threshold for two-interval discrimination of cone-directed flicker was
measured on a low- or a high-melanopic background (Fig.2a). participants adapted to the low- or high-melano-
pic background for five minutes, and the background was held constant during collection of 120 trials.
Detection thresholds for cone-directed flicker on the constant high-melanopic background were not consist-
ently different from detection thresholds on the constant low-melanopic background (Fig.2b). For participants
1–3, median JNDs in the high melanopsin stimulation condition were, respectively, 80%, 100%, and 100% percent
of the median JND in the low melanopsin stimulation conditions. For each participant, the difference in JNDs
between melanopsin conditions is small compared to the variation in these differences across sessions (Fig.2b,
insets). The 95% CI about the median difference included zero for two of the three participants: [− 0.39, − 0.03],
[− 0.37, 0.23], [− 0.66, 0.17]. Therefore, no consistent difference in JNDs of cone-directed flicker between the low
and high melanopsin conditions was found; Participant 1 appears to show a slight decrease in sensitivity, while
the other two participants show no systematic difference in median threshold.
Experiment 2. In Experiment 2, the threshold for two-interval discrimination of cone-directed flicker was
measured on a low-melanopic background, or a pulsed high-melanopic pedestal (Fig.3a). Participants adapted
to the low-melanopic background for five minutes, and on individual trials were presented either with only the
low-melanopic background or the high-melanopic pulsed pedestal on top of that background.
Figure3b presents the thresholds for cone-directed flicker on a pulsed high-melanopic pedestal as com-
pared to the low-melanopic background. For participants 4–6, median JND in the high melanopsin stimulation
condition were, respectively, 80%, 69%, and 74% percent of the median JNDs in the low melanopsin stimula-
tion conditions For all three participants, there was a small but consistent tendency for JNDs to be reduced on
the high-melanopic pedestal as compared to low-melanopic background. The within-participant differences in
JNDs between melanopsin conditions are shown in the insets (Fig.3b). For each participant, for each session,
the JND on the low melanopic background was lower than the JND on the pulsed high-melanopic pedestal.
The 95% CI about the median difference excluded zero for all three participants: [− 0.48, − 0.12], [− 0.61, − 0.18],
[− 0.39, − 0.14].
Experiment 2 seems to show a small, consistent decrease in JNDs of cone-directed flicker between the high
and low melanopsin conditions. We considered, however, the possibility that this difference between conditions
was due to small confounding stimulus artefacts, rather than a true effect of melanopsin stimulation. We discuss
this possibility next.
Potentially confounding stimulus artefacts. The results of Experiments 1 and 2 show either no effect,
or a small effect, of melanopsin background upon cone flicker detection. Two alternative hypotheses suggest that
either there is no true effect of melanopsin stimulation (as evidenced in Experiment 1), and the apparent effect in
Experiment 2 is due to confounds; or, alternatively, Experiment 2 shows a true effect that is masked by confounds
in Experiment 1. The measurement could be confounded in such ways if stimuli depart from their nominal spec-
tral content. There could be imprecision in the LMS content of low- and high-melanopic backgrounds or in the
magnitude of LMS flicker contrast itself, either of which could cause the LMS flicker to have contrast different
from that intended on the two backgrounds. Further, stimulus imprecision could result in the stimulus being
detected by a mechanism other than the one targeted (e.g., the highly sensitive L–M cone opponent mechanism).
Spectroradiometric measurements of the stimuli were obtained during each session of data collection, and these
measurements were used to assess each of these potential confounds.
9
Vol.:(0123456789)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
Imprecision in background spectra. In Experiments 1 and 2, the low-melanopic and high-melanopic
spectra were designed through silent substitution that targeted melanopsin while attempting to silence the L,
M and S cones. Imprecision in device control affects the efficacy of this silencing: residual contrast (“splatter”)
on the L, M, and S cones could have been present across the nominally cone-silent background exchange. Since
the pair of spectra was generated separately for each session of data collection, the imprecisions varied session-
to-session. Radiometric measurements of the background spectra were taken at the end of each session of data
collection to assess the magnitude of the background splatter.
Figure4a, left, shows the measured LMS content of the low-melanopic and high-melanopic backgrounds
in Experiment 1. LMS content here is the sum of L, M and S-cone excitations. Excitations were calculated by
multiplying measured spectral power distributions by cone fundamentals (normalized such that they peak at 1).
For all three participants, the LMS content does not vary systematically between the melanopic backgrounds.
To understand the effect of any difference in cone stimulation between backgrounds on the psychophysical
sensitivity, normalized JNDs were predicted for each session. This prediction was made as follows: First, a dif-
ference in melanopsin stimulation was assumed to have no effect on sensitivity. The nominal LMS contrast at
threshold, as measured on the low-melanopic background, was taken as the predicted threshold contrast for both
backgrounds, i.e., the threshold flicker LMS contrast was assumed constant across melanopic conditions. Second,
this predicted contrast was multiplied with the measured LMS content of each background. This expresses the
predicted threshold as the predicted JND on each background. The predicted JNDs were normalized in the same
way as the experimental results, such that the median (per participant) JND in the low melanopsin stimulation
condition has a value of 1. These normalized predicted JNDs express thresholds as the increase in LMS content
that would correspond to the same predicted LMS contrast on each background, given the measurements of that
background, under a Weber’s law assumption.
Figure4a, right, shows the median (per participant) of these normalized, predicted JNDs. They do not change
appreciably across the two melanopic backgrounds, confirming that the measured LMS contents of the low- and
high-melanopic backgrounds do not differ enough to lead to substantially different JNDs. Thus, imprecisions
in the nominally constant LMS content of the backgrounds would not produce an effect that could be masking
a true effect of melanopsin stimulation on psychophysical sensitivity to cone-directed flicker and leading to the
observed null-result.
Figure5a shows the same analysis, with the same conclusion of no artifact, for Experiment 2.
Imprecision in flicker. The flicker stimuli in Experiments 1 and 2 were also designed through silent substi-
tution to produce equivalent modulation of the L, M and S cones on both melanopic backgrounds. Radiometric
measurements of the flicker spectra were also taken, at the end of each session. A preliminary nominal threshold
was estimated from the staircase reversals—not from a psychometric function fit—and this preliminary nominal
threshold was used to generate stimulus spectra corresponding to the nominal stimulus at detection threshold.
From these spectra, LMS flicker contrast was calculated as the equally weighted sum of L, M, and S-cone con-
trasts between these spectra and the corresponding measured background spectra.
Figure4b, left, shows the measured flicker LMS contrasts, in the low- and high-melanopic conditions. Here
the measured contrast is expressed as a proportion of the nominal contrast. In the ideal case of no splatter, this
proportion would equal 1 (measured flicker LMS contrast equals nominal flicker LMS contrast); higher values
indicate the flicker stimulus had more cone-directed content than nominal, lower values indicate less than nomi-
nal content. On average, the flicker was slightly stronger than nominal in the low-melanopic condition, but closer
to nominal in the high-melanopic condition, although this is not consistent or robust across participants and
sessions. This suggests that the same nominal flicker threshold corresponds to a weaker flicker stimulus in the
high-melanopic condition than in the low-melanopic condition, thus potentially overestimating the sensitivity
in the high-melanopic condition.
To understand the effect of this increase on the psychophysical sensitivity, JNDs were predicted for each ses-
sion. This prediction was made as follows. First, a difference in melanopsin stimulation was assumed to have no
effect on sensitivity. The nominal LMS contrast at threshold, as measured on the low-melanopic background,
was taken as the predicted threshold contrast for both backgrounds, i.e., the threshold flicker LMS contrast
was assumed constant across melanopic conditions. Second, this predicted contrast was multiplied with the
ratio between measured and nominal LMS contrast of each flicker stimulus. This, then, predicts the LMS con-
trast (including splatter) of the flicker stimulus, at the predicted threshold. Lastly, this predicted contrast was
multiplied with the measured LMS content of each background. This converts the predicted threshold into the
predicted JND on each background.
Figure4b, right, shows these predicted JNDs, averaged per participant as the median over sessions for Experi-
ment 1. These differ minimally across the two melanopic backgrounds, indicating that imprecisions in the
nominally identical flicker stimuli did not produce an effect that could be masking a true effect of melanopsin
stimulation on psychophysical sensitivity to cone-directed flicker and thus producing an artifactual null-result.
If anything, the splatter analysis indicates that a true lack of effect might be revealed as a small decrease in JND
with increasing background melanopic content, consistent with the weak trend seen in the data for Experiment
1. Figure5b shows the same analysis, with similar conclusion, for Experiment 2.
Imprecise production of the LMS flicker stimuli could also have introduced L–M chromatic contrast, to which
the visual system is highly sensitive27. Given enough L–M splatter, detection of the flicker stimulus could have
been mediated primarily by this chromatic contrast, rather than by the experimentally manipulated LMS contrast.
From the same radiometric measurements of the flicker stimuli, the L–M contrast splatter was computed. To
compute the L–M contrast for other nominal values of L + M + S stimulus, we express the L–M splatter relative
10
Vol:.(1234567890)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
Melanopic adaption level
LMS responses
LMS content of stimulus background
Melanopic adaptation level
LMS JND, normalized
JNDs predicted from background LMS content
Melanopic adaption level
LMS flicker contrast, normalized
Realized vs. nominal LMS flicker contrast
Melanopic adaptation level
JNDs predicted from flicker LMS imprecision
Melanopic adaption level
(L−M)/(L+M+S) absolute contrast
Flicker L-M splatter
Melanopic adaptation level
LMS JND, normalized
JNDs predicted from flicker L-M splatter
14.5
14.5
LMS JND, normalized
0.0
0.1
0.2
0.3
0.4
0.5
14.5
0.0
0.5
1.0
1.5
2.0
14.5
0.0
0.2
0.4
0.6
14.5
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
14.5
11
Vol.:(0123456789)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
to the measured L + M + S contrast, as the contrast ratio (L–M)/(L + M + S). We scale this ratio by the nominal
value of L + M + S to obtain an estimate of L–M splatter as nominal target contrast is varied.
There was a trend for the (L–M)/(L + M + S) ratio of the flicker contrasts to be lower for the low-melanopic
condition than for the high-melanopic condition, for all participants in both Experiment 1 (Fig.4c, left) and
Experiment 2 (Fig.5c, left). To understand the effect of this chromatic contrast on the psychophysical sensitivity,
JNDs were predicted for each session, under the null-hypothesis that melanopic content does not affect sensitivity
and that detection of the flicker stimulus was mediated exclusively by the L–M chromatic contrast.
Specifically, for each session the absolute value of measured L–M contrast at threshold on the low-melanopic
background was taken as the predicted threshold contrast for both conditions, i.e., the L–M contrast was assumed
constant at threshold across melanopic conditions. By multiplying this predicted contrast with the ratio between
L–M and LMS contrast, the nominal LMS contrast of a flicker stimulus with that predicted L–M contrast on
the high-melanopic background was calculated. If the L–M splatter mediated detection and there were no effect
of background melanopsin content on thresholds, a decrease in threshold (increase in sensitivity) would be
observed on the high-melanopic background. Figure5c, right shows the same analysis, with the same conclu-
sion, for Experiment 2.
The results from Experiment 2 potentially reflect this kind of stimulus artifact. Overall, participants showed
slightly increased sensitivity for the flickering stimulus on the high melanopsin background. While this could
in principle be an effect of melanopsin excitation, such an effect in the data is consistent with the direction of
effects produced by the small amount of splatter in the stimuli, so that the small changes in threshold sensitivity
might be artifactual. To address this possibility, a third experiment was conducted, in which the generation of
the stimulus flicker was independent of the manipulation of the stimulus background.
Experiment 3. Experiment 3 was designed to address the possibility that imprecision in the flicker stimulus
could confound the perceptual measurement on different melanopic backgrounds. For this experiment, the low-
melanopic background and high-melanopic pedestal were produced by silent-substitution using the OneLight
as in Experiment 2. The flicker stimuli, however, were generated using the DLP projector, a device separate from
that used to produce the background and pedestal, and optically admixed. Thus, the spectral content of the
flicker stimuli was held constant across melanopic conditions (session-to-session variation may have occurred,
but this variation would be independent of, and not correlated with, the melanopic condition).
In Experiment 3 the threshold for two-interval discrimination of cone-directed flicker was measured on a
low-melanopic background and on a pulsed high-melanopic pedestal (as in Experiment 2). Participants adapted
to the low-melanopic background for five minutes, and on individual trials were presented either with only the
low-melanopic background or the high-melanopic pulsed pedestal on top of that background.
Detection thresholds for cone-directed flicker admixed on the pulsed high-melanopic pedestal, were not
consistently different from detection thresholds on the low-melanopic background For participants P1, P4 and
P5, median JNDs in the high melanopsin stimulation condition were, respectively, 106%, 100% and 89% of the
JNDs in the low melanopsin stimulation conditions (Fig.6). For each participant, the difference in JNDs between
melanopsin conditions is on average (median across sessions) small compared to the variation in these differ-
ences across sessions (Fig.6, insets), which for each Participant go in both the positive and negative direction.
The 95% CI about the median difference includes zero, for all three participants: [− 0.06, 0.001], [− 0.25, 0.63],
[− 0.26, 0.05]. Therefore, no consistent difference in JNDs of cone-directed flicker between the low and high
melanopsin conditions was found.
Discussion
We find that varying melanopsin stimulation does not measurably alter threshold sensitivity for cone-directed
luminance flicker. Interaction effects are absent both during adaptation to a continuous, high melanopsin back-
ground (Experiment 1), and against brief, high melanopsin spectral pulses (Experiments 2 and 3). This null
result was obtained in the face of slight imperfections in our control of the chromatic content of our stimuli. We
Figure4. Stimulus imprecision analysis for Experiment 1. (a) Left panel. LMS content of the low-melanopic
and high-melanopic backgrounds in Experiment 1, in cone coordinates. Stimulus spectra were multiplied by
cone fundamentals (normalized to peak at 1) to obtain coordinates for each cone class; these were then averaged
to get a combined LMS cone coordinate. Right panel. Predicted JNDs if background LMS content were driving
LMS detection threshold, under a Weber’s Law assumption for detection (JND proportional to background LMS
content). Same format as Fig.2b, but with median (across sessions) JNDs and ± standard error of the median
for each of three participants shown in a single plot. (b) Measured LMS content of the threshold flicker stimuli
in Experiment 1. Left panel. Measured LMS content of the flicker stimuli corresponding to detection threshold
in the low-melanopic and high-melanopic conditions, expressed as proportion of the nominal LMS contrast
of those flicker spectra (i.e., value of 1 indicates LMS contrast was as nominal). Each line plots median ± 1
standard error of the median for one participant, with the median taken across sessions. Right panel. JNDs for
Experiment 1, based on measured rather than nominal flicker spectra. (c) L–M chromatic contrast of the LMS
flicker stimulus spectra in Experiment 1. Left panel. Measured (L–M)/(LMS) contrast of the flicker stimuli in the
low-melanopic and high-melanopic conditions. Each line plots median ± 1 standard error of the median for one
participant, with the median taken across sessions. Right panel. Predicted JNDs, assuming that detection of the
flicker stimulus was mediated not by LMS contrast, but rather by the absolute magnitude of the L–M chromatic
contrast. Under this assumption, greater L–M contrast in the flicker stimulus predicts lower JNDs, under the
null-hypothesis that cone-mediated flicker detection does not change with background melanopsin stimulation.
◂
12
Vol:.(1234567890)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
LMS content of stimulus background
LMS responses
Melanopic level
LMS JND, normalized
JNDs predicted from background LMS conten
t
LMS flicker contrast, normalized
Realized vs. nominal LMS flicker contrast JNDs predicted from flicker LMS imprecision
(L−M)/(L+M+S) absolute contrast
Flicker L-M splatter
LMS JND, normalized
JNDs predicted from flicker L−M splatter
14.5
14.5
14.5
LMS JND, normalized
0.0
0.1
0.2
0.3
0.4
0.5
14.5
0.0
0.5
1.0
1.5
2.0
14.5
0.0
0.2
0.4
0.6
14.5
Melanopic level
Melanopic levelMelanopic level
Melanopic level
Melanopic level
0
1
2
3
4
5
0
1
2
3
4
5
0
1
2
3
4
5
Figure5. Stimulus imprecision analysis for Experiment 2. Same format as Fig.4.
13
Vol.:(0123456789)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
consider it notable that failure to measure and account for these small artifacts might have otherwise led to the
erroneous conclusion of an effect of melanopsin stimulation upon cone sensitivity.
Melanopsin resides within the ipRGCs, and these cells are known to project to the geniculo-striate visual
pathway15. There is evidence that melanopsin stimulation contributes to perceived brightness, in addition to the
perception of brightness provided by the cone-mediated luminance mechanism4,6,10,11,20. This brightness percept
follows the properties predicted from ipRGC physiology, including stronger percepts at higher background light
levels and lower temporal frequencies3. A related finding is that cone and melanopsin signals combine in a sub-
additive manner in producing ratings of visual discomfort28. These photoreceptor signals must be combined
somewhere prior to sites mediating verbal report, and in principle this combination could begin within the
ipRGCs themselves, as several classes of these cells receive extrinsic cone inputs.
We sought to test in the current study the specific proposition that melanopic signals from the ipRGCs inter-
act in human perception with cone-driven signals carried by the classic RGCs. Specifically, we began with the
assumption that cone-directed luminance flicker is processed predominantly by the classic RGCs. If manipula-
tion of the melanopic background (which can only be transduced by the ipRGCs) alters perceptual sensitivity
for luminance flicker, this would provide support for the idea that ipRGC and classic RGC signals interact in
perception (although with some caveats discussed below). Across our three experiments, we did not find reliable
evidence of a change in threshold cone sensitivity induced by changing melanopsin activation. While it is of
course possible that a small interaction effect exists that escaped the sensitivity of our measurements, our find-
ings nonetheless place tight bounds on how much interaction there could be between melanopsin stimulation
and luminance flicker detection at threshold.
An inferential nuance of our study is that a positive finding in our measurements would not have provided
definitive evidence regarding the mechanism of the interaction, and more specifically would not have shown
unambiguously an interaction of ipRGC signals with those of the classic RGCs. First, it is possible that the 5Hz
luminance modulation could be transduced by the ipRGCs themselves. Second, because Experiment 3 made
use of a light-flux flicker modulation (as opposed to pure luminance) it is possible that the melanopic compo-
nent of the light-flux flicker could have directly modulated ipRGC activity. Both of these considerations admit
the possibility that the background and flicker components of the stimuli could have interacted entirely within
the ipRGCs. In practice, we think that both of these theoretical mechanisms are unlikely to contribute much
to perception on prior grounds, and more generally these considerations are mooted by our negative results.
Prior studies with animals and humans have suggested that ipRGC and classic RGC signals can interact.
In rodents, a ten-fold increase in melanopic background leads to a measurable decline in the b-wave (post-
photoreceptor) ERG response to a brief flash of light13; a similar effect of melanopsin excitation on global mean
activity in the LGN was observed. This effect is proposed to arise from recurrent axon collaterals from the ipRGCs
that modulate cone inputs to the bipolar cells12. There are salient differences between our perceptual study in
humans and this electrophysiologic measurement in rodents that may explain the different findings of an effect
of melanopsin background. While our study provided substantial (350%) contrast on melanopsin between the
low and high background conditions, our stimulus manipulation is nonetheless unable to achieve the higher
contrast possible in transgenic mice. It is possible that an effect of melanopsin background upon cone sensi-
tivity arises only for changes in background beyond what we were able to achieve. Additionally, we examined
perception of a subtle modulation of cone signals at perceptual threshold, as opposed to the supra-threshold,
high-intensity flash of light used in the ERG measurement. It is possible that the effect of melanopsin activity
upon cone signals becomes evident only in the supra-threshold case. We also cannot discount the possibility
that interactions between melanopsin and cone signals are fundamentally different in humans than they are in
the rodent model system.
1 4.5 1 4.5 1 4.5
Melanopic level
LMS JND, normalized
Participant 3 Participant 4 Participant 5
0
1
2
3
4
5
−1
0
1
JND difference
−1
0
1
JND difference
JND difference
−1
0
1
Figure6. Experiment 3 results. Same format as Fig.2b.
14
Vol:.(1234567890)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
The prior study most similar to our own is that of Zele and colleagues7. They measured human psychophysi-
cal sensitivity to one second, cone-directed pulses, either in the presence or absence of a melanopsin pedestal.
In different measurements, contrast on melanopsin was varied between 9 and 22%. Despite this relatively small
degree of melanopsin contrast, Zele and colleagues reported that the high melanopic pedestal induced a 40–60%
increase in sensitivity to cone-directed stimuli. It is difficult to reconcile these findings with the negative results
of the current study. While there were differences in the cone-directed stimuli (steady pulses vs. 5Hz flicker)
and background light level (637cd/m2 in Zele and colleagues vs. ~ 275cd/m2 in the current study), these do not
strike us as fundamental differences for a posited interaction of ipRGC and classic RGC signals. In particular,
these small stimulus differences seem unlikely to explain why our study found no effect of a large manipulation
of melanopsin contrast upon cone-mediated perception, while Zele and colleagues found a large effect using a
small change in melanopsin contrast.
We were aware in the current study of the particular importance of measuring and accounting for even small
imprecision in the generation of stimuli. The current experiments (similar to the studies7,13) make use of the tech-
nique of silent substitution19 to selectively target photoreceptor classes. Apart from adjusting for age-predicted
lens density, we did not use psychophysical measurements to adjust our stimuli for each observer. Individual
biologic variation in (for example) macular pigment density and opsin spectral sensitivity is a further source of
contrast “splatter” beyond the device imprecision that we examined here. We believe it unlikely that systematic
differences between our participant population and our stimulus model masked a subtle, adaptive effect of mel-
anopsin upon luminance sensitivity, but we cannot completely exclude this possibility.
As we have described in this report, inevitable imperfections in device control produce deviations of the
stimuli from their nominal properties. Most critically, we found that our nominal luminance isolating flicker
stimulus in Experiments 1 and 2 developed ~ 0.5% of L–M chromatic contrast “splatter” when it was generated
on the high melanopsin background spectrum. While a seemingly small degree of stimulus error, we calculated
that this nonetheless could lead subjects to exhibit increased sensitivity for detecting the flicker on the high mel-
anopsin background. This observation motivated Experiment 3, in which the design of the apparatus precluded
an influence of the melanopsin background upon the flicker stimulus. This third experiment allowed us to reject
the presence of an interaction in perception of pulses of melanopsin stimulation and threshold sensitivity for
cone-directed flicker.
We consider it possible that the results reported by Zele and colleagues7 contain an influence of stimulus
imprecision of this kind, although we note that the authors did consider and reject this possibility. A similar
note of caution should likely attend claims that melanopsin stimulation imparts or alters chromatic percepts29,30;
note as well that prior work4,18 by Brown etal. and Spitschan etal. cautions against concluding that melanopsin
stimulation induces a chromatic percept.
In summary, we did not find in the current study evidence that melanopsin-driven signals arising from the
ipRGCs influence human perception of cone signals conveyed by the classical RGCs. It remains possible that such
interactions exist for stimulus regimes different from those we examined. In particular, prior work in rodents
leads us to expect that, if such interactions are to be found, they are present for supra-threshold cone stimulation
against very high melanopic backgrounds.
Data availability
These experiments were the subject of pre-registration documents (E1: https:// osf. io/ jd4qh/; E2: https:// osf. io/
qgs49/; E3: h t t ps:// osf . io/ p45aj/). The datasets and code used for data analysis is publicly available (h t t ps:// gi thub.
com/ Brain ardLab/ MeLMS ens).
Received: 8 July 2021; Accepted: 24 September 2021
References
1. Berson, D. M. Phototransduction by retinal ganglion cells that set the circadian clock. Science 295, 1070–1073 (2002).
2. Lucas, R. J., Douglas, R. H. & Foster, R. G. Characterization of an ocular photopigment capable of driving pupillary constriction
in mice. Nat. Neurosci. 4, 621–626 (2001).
3. Lucas, R. J., Allen, A. E., Milosavljevic, N., Storchi, R. & Woelders, T. Can we see with melanopsin?. Annu. Rev. Vis. Sci. 6, 453–468
(2020).
4. Brown, T. M. et al. Melanopsin-based brightness discrimination in mice and humans. Curr. Biol. 22, 8 (2012).
5. Spitschan, M. Melanopsin contributions to non-visual and visual function. Curr. Opin. Behav. Sci. 30, 67–72 (2019).
6. Zele, A. J., Adhikari, P., Feigl, B. & Cao, D. Cone and melanopsin contributions to human brightness estimation. J. Opt. Soc. Am.
A 35, B19 (2018).
7. Zele, A. J., Adhikari, P., Cao, D. & Feigl, B. Melanopsin driven enhancement of cone-mediated visual processing. Vis. Res. 160,
72–81 (2019).
8. Allen, A. E., Storchi, R., Martial, F. P., Bedford, R. A. & Lucas, R. J. Melanopsin contributions to the representation of images in
the early visual system. Curr. Biol. 27, 1623-1632.e4 (2017).
9. Allen, A. E., Martial, F. P. & Lucas, R. J. Form vision from melanopsin in humans. Nat. Commun. 10, 2274 (2019).
10. Yamakawa, M., Tsujimura, S. I. & Okajima, K. A quantitative analysis of the contribution of melanopsin to brightness perception.
Sci. Rep. 9(1), 1–8 (2019).
11. DeLawyer, T., Tsujimura, S. I. & Shinomori, K. Relative contributions of melanopsin to brightness discrimination when hue and
luminance also vary. JOSA A 37(4), A81–A88 (2020).
12. Joo, H. R., Peterson, B. B., Dacey, D. M., Hattar, S. & Chen, S.-K. Recurrent axon collaterals of intrinsically photosensitive retinal
ganglion cells. Vis. Neurosci. 30, 175–182 (2013).
13. Allen, A. E. et al. Melanopsin-driven light adaptation in mouse vision. Curr. Biol. 24, 2481–2490 (2014).
14. Hankins, M. W. & Lucas, R. J. The primary visual pathway in humans is regulated according to long-term light exposure through
the action of a nonclassical photopigment. Curr. Biol. 12, 191–198 (2002).
15
Vol.:(0123456789)
Scientific Reports | (2021) 11:20167 | https://doi.org/10.1038/s41598-021-99684-0
www.nature.com/scientificreports/
15. Dacey, D. M. et al. Melanopsin-expressing ganglion cells in primate retina signal colour and irradiance and project to the LGN.
Nature 433, 749–754 (2005).
16. Brown, T. M. et al. Melanopsin contributions to irradiance coding in the thalamo-cortical visual system. PLoS Biol. 8, e1000558
(2010).
17. Schmidt, T. M. et al. A role for melanopsin in alpha retinal ganglion cells and contrast detection. Neuron 82, 781–788 (2014).
18. Spitschan, M., Aguirre, G. K. & Brainard, D. H. Selective stimulation of penumbral cones reveals perception in the shadow of
retinal blood vessels. PLoS ONE 10, 1–22 (2015).
19. Estévez, O. & Spekreijse, H. The, “silent substitution” method in visual research. Vis. Res. 22, 681–691 (1982).
20. Spitschan, M. et al. The human visual cortex response to melanopsin-directed stimulation is accompanied by a distinct perceptual
experience. Proc. Natl. Acad. Sci. U. S. A. 114, 12291–12296 (2017).
21. CIE. Fundamental Chromaticity Diagram with Physiological Axes—Parts 1 and 2. Technical Report. 170-1 (Central Bureau of the
Commission Internationale de l’ Éclairage, 2007).
22. Spitschan, M., Datta, R., Stern, A. M., Brainard, D. H. & Aguirre, G. K. Human visual cortex responses to rapid cone and melan-
opsin-directed flicker. J. Neurosci. 36, 1471–1482 (2016).
23. Prins, N. & Kingdom, F. A. A. Applying the model-comparison approach to test specific research hypotheses in psychophysical
research using the Palamedes toolbox. Front. Psychol. 9, 1250 (2018).
24. Krzywinski, M. & Altman, N. Visualizing samples with box plots. Nat. Methods 11, 119–120 (2014).
25. R Core Team. R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2021).
26. Wickham, H. et al. Welcome to the Tidyverse. JOSS 4, 1686 (2019).
27. Chaparro, A., Stromeyer, C. F., Huang, E. P., Kronauer, R. E. & Eskew, R. T. Jr. Colour is what they eye sees best. Nature 361, 348–350
(1993).
28. McAdams, H. et al. Selective amplification of ipRGC signals accounts for interictal photophobia in migraine. Proc. Natl. Acad. Sci.
U. S. A. 117, 17320–17329 (2020).
29. Cao, D., Chang, A. & Gai, S. Evidence for an impact of melanopsin activation on unique white perception. JOSA A 35(4), B287–
B291 (2018).
30. Zele, A. J., Feigl, B., Adhikari, P., Maynard, M. L. & Cao, D. Melanopsin photoreception contributes to human visual detection,
temporal and colour processing. Sci. Rep. 8, 3842 (2018).
Acknowledgements
This work was supported by grants from the National Eye Institute (R01EY024681 to GKA and DHB; Core Grant
for Vision Research P30 EY001583) and the Department of Defense (W81XWH-151-0447 to GKA).
Author contributions
J.V., D.B. and G.K.A. designed the experiments, and participated in some of the experiments; corresponding
author J.V. implemented the experimental design in software, collected data, analyzed data, and drafted the
methods and results; Joint supervisors D.B. and G.K.A. drafted discussion and introduction, and edited methods,
results; E.B.H. designed figures and provided copy-editing. All authors reviewed the manuscript.
Funding
Open Access funding enabled and organized by Projekt DEAL.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary Information The online version contains supplementary material available at https:// doi. org/
10. 1038/ s41598- 021- 99684-0.
Correspondence and requests for materials should be addressed to J.V.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
© The Author(s) 2021
