Brightness illusions evoke pupil constriction preceded by a primary visual cortex response in rats

Abstract

The mind affects the body via central nervous system (CNS) control of the autonomic nervous system (ANS). In humans, one striking illustration of the “mind–body” connection is that illusions, subjectively perceived as bright, drive pupil constriction. The CNS network driving this pupil response is unknown and requires an animal model for investigation. However, the pupil response to this illusion has long been thought to occur only in humans. Here, we report that the same brightness illusion that evokes pupil constriction in humans also does so in rats. We surveyed the role of most of rat cortex in this “mind–body” connection by recording cortex-wide EEG. These recordings revealed that, compared to a luminance-matched control stimulus, the illusion of brightness for a specific stimulus color and size, evoked a larger response in primary visual cortex (V1) and not in secondary visual, parietal, or frontal cortex. The response preceded pupil constriction suggesting a potential causal role of V1 on the pupil. Our results provide evidence that this “mind–body” connection is not confined to humans and that V1 may be part of a mammalian CNS network for bodily reactions to illusions.

Introduction

Mental processes mediated by the central nervous system (CNS) can affect the body via the autonomic nervous system (ANS). The “mind–body” connection is apparent from the effects of psychological stress on immune and gastrointestinal function (Glaser and Kiecolt-Glaser 2005; Mawdsley and Rampton 2005; Poller et al. 2022), and placebo effects on pain driven by a person’s beliefs (Geuter et al. 2016). One fascinating “mind–body” connection is that subjective illusions of brightness cause the eye’s pupil to constrict (Laeng and Endestad 2012) due to the CNS driving the parasympathetic arm of the ANS which constricts the pupil (McDougal and Gamlin 2017). The brain regions, cell types, and synaptic projections that mediate this “mind–body” interaction remain completely unexplored because there is no animal model permitting intracerebral (invasive) investigation.

In this study, we presented a brightness illusion (which humans subjectively perceive as bright (Laeng and Endestad 2012)) to head-fixed rats while performing pupillometry and simultaneous brain-wide 32-electrode EEG. We used the Asahi stimulus (Fig. 1), which was created by Prof. Akiyoshi Kitaoka (Department of Psychology, Ritsumeikan University, Osaka, Japan) based on luminance-gradient stimuli in earlier work (Zavagno 1997). We show that the Asahi stimulus drives a pupil constriction in rats. On the other hand, a luminance-matched control stimulus, which does not evoke illusory brightness in humans, did not cause a pupil constriction in rats. The Asahi stimulus also drove a larger cortex EEG event-related potential that was confined to primary visual cortex and preceded pupil constriction. Our results show that the rat is a viable animal model for studying how neural processing of illusions can affect autonomic control of the body.

Fig. 1.

Illustrations of the Asahi stimulus and the luminance-matched control stimulus. The Asahi stimulus is typically perceived by humans to have a brighter-than-white glare in the center. Rearranging the shape (control stimulus) abolishes the perception of brightness. The screen shots show that the actual luminance at the center of each stimulus is identical, despite appearing to be brighter in the center of the Asahi stimulus.

Materials and methods

Subjects

Male, Lister-Hooded rats (140–190 g) were obtained from Charles River. After a 7-day acclimation period, rats were implanted with a chamber and head-post and, in some cases, an EEG array. After implantation, rats were single housed. Experiments were carried out during the rats’ active phase (housing illumination from 7 P.M. to 7 A.M.). All procedures were carried out with the prior approval of local authorities and in compliance with the European Community Guidelines for the Care and Use of Laboratory Animals.

Surgical procedures

The surgical procedure was identical to prior work (Vasilev et al. 2022). Briefly, the rat was anesthetized using isoflurane and head-fixed using ear bars. We administered buprenorphine (0.06 mg/kg, s.c.), meloxicam (2.0 mg/kg, s.c.), enrofloxacin (10.0 mg/kg, s.c.), and lidocaine (0.5%, s.c. over skull) and waited 10–15 min (and for lack of response to paw pinch) before beginning surgical procedures. An EEG array (Neuronexus, CM32) was laid onto the cleaned and dried skull and fixed in place using dental cement (two-stage, powder/liquid Paladur). A custom-made skull implant was used for head-fixation (machine shop, Max Planck Institute for Biological Cybernetics). The implant was fixed onto the skull using UV-curing primer and dental cement (Tetric Evoflow, Dental Bauer). A craniotomy was made on the left occipital bone for a ground wire (99.9% pure silver). One end was flattened using an industrial press into an ~ 1–2 mm wide rectangle, which then was twisted into a roll to fit the craniotomy and inserted into the space between bone and dura. A rolled shape was used to increase the potential surface area in contact with CSF. The craniotomy was filled with viscous agar, which stabilized the wire and provided a conductive medium between the ground wire and the CSF. The other end of the ground wire was soldered to the ground wire of the electrode interface board of the EEG array. The wires and array were buried under dental cement (Paladur). The skin around the implant was glued to the implant using tissue glue (Histoacryl, B. Braun). Post surgical recovery lasted five days. During the first three days (surgery itself was counted as day 1), the rat was injected either every 12 h with buprenorphine or every 24 h with meloxicam (same dosages as pre-operative). During the first 5 days, enrofloxacin was injected every 24 h (same dosage as pre-operative). A rehydrating, nutritious, easily consumed, and palatable food was provided during recovery (DietGel Recovery, Clear H2O).

Handling and habituation

Rats were handled daily for at least 5 min per day from the day of arrival in housing until the day of surgery, which was 7 days. Animals were neither food nor water restricted. Habituation consisted of one session (~25 min) of head-fixation on a freely-rotating treadmill in front of a computer screen. Rats were free to run or remain immobile and a mixture of locomotor activity was observed.

Visual stimuli and stimulus presentation

Stimuli consisted of the Asahi stimulus, the control stimulus, and a gray screen used during the inter-stimulus interval. All stimuli were equiluminant (15–16 lux measured at the head-post). Stimuli were presented 50 cm from the rat’s head. The resolution was 1280 × 720 pixels. Stimuli were created in Adobe Illustrator with a canvas set to match the screen resolution. These files were exported to JPEG at 72 ppi. The stimuli were presented using Psychtoolbox implemented in MATLAB. The stimuli were presented at four visual angles (10°, 20°, 40°, and 60°) and in two colors (yellow, 580 nm, HEX: #ffff00 and green, 508 nm, HEX: #00ff28). As many stimuli were placed on the screen as possible, thus covering the entire visual field or most of it. Stimuli were presented for a duration of 4 s. The inter-stimulus intervals were drawn from a distribution with a flat hazard rate that spanned 4–8 s with a 0.5-s resolution. The flat hazard rate was used to reduce expectation of stimulus onset.

Pupillometry data acquisition and processing

Videos were recorded at 45 frames per second from the rat’s right eye with near-infrared illumination (Thor Labs LED, M850L3 and Thor Labs Collimation optics, COP4-B). Frames were acquired using a near-infrared camera (Allied Vision, G-046B) and variable zoom lens, fixed 3.3× zoom lens, and 0.25× zoom lens attachment (Polytec, 1–60,135, 1–62,831, 6044). Acquisition occurred over a GigE connection (MATLAB image processing toolbox). The camera provided a TTL pulse with each video frame. These TTL pulses were recorded directly into the neurophysiology system (Neuralynx).

We used an in-house custom algorithm and computer code to extract pupil size from the recorded video frames. The procedure is reviewed in detail in prior work (Vasilev et al. 2022). Briefly, images were Gaussian blurred, converted into a binary image, and then subjected to edge detection, closed contour detection, and fitting of ellipses to the closed contours. In cases where the algorithm was not able to find an ellipse of an area bigger than predefined minimal allowed area (or smaller than maximal, respectively), then the value of the pupil in this frame was left blank. This was also applied to the frames which captured the animal blinking. Blank frames were linearly interpolated. The pupil detection algorithm was implemented using the OpenCV package in Python 3.7.

The pupil size was normalized to a pre-stimulus baseline (1 sec duration) by calculating a z-score. The z-score was calculated on each trial by subtracting the baseline mean from each pupil size data point and then dividing this array by the standard deviation of the baseline data points. The latency for pupil constriction was calculated as in prior work (Bergamin and Kardon 2003). We first smoothed the pupil size with an 11-point, second-order Savitzky–Golay filter (sgolayfilt in MATLAB). The signal was differentiated to obtain velocity and the velocity was lowpass filtered at 6 Hz with a second-order Butterworth filter (filtfilt in MATLAB). This signal was then differentiated to obtain acceleration and the latency to constrict was defined as the time point with the largest negative acceleration.

E‌EG signal acquisition and analysis

EEG signals were recorded using a flexible polyimide array with 32 platinum electrodes (Neuronexus, H32). Signals were recorded against animal ground, pre-amplified at the rat’s head (Neuralynx, HS-36), and then amplified and digitized at 32 kHz (Neuralynx, Digital Lynx SX). The analysis focused on bilateral electrodes placed over frontal cortex (locations relative to Bregma: 1.5 mm anterior, ±1.2 mm lateral; 3.6 mm anterior, ±1.2 mm lateral) and visual cortex (locations relative to Bregma: 5.0 mm posterior, ±1.5 mm lateral; 5.0 mm posterior, ±3.0 mm lateral; 5.0 mm posterior, ±4.4 mm lateral; 7.0 mm posterior, ±1.5 mm lateral; 7.0 mm posterior, ±3.0 mm lateral; 7.0 mm posterior, ±4.4 mm lateral). EEG signals were first low pass filtered at 5 Hz and then downsampled to 320 Hz. The entire signal was mean subtracted. EEG topographical plots were produced using the MATLAB command, scatteredInterpolant with natural neighbor interpolation. The contour plots (contourf function in MATLAB) used 50 levels.

Statistics

We used estimation statistics to report effect sizes and the confidence intervals for effect sizes (DABEST toolbox in MATLAB (Ho et al. 2019; Calin-Jageman and Cumming 2019a)). Bayesian statistics were used for assessing evidence (or lack thereof) for the null hypothesis and for the alternative hypothesis (Keysers et al. 2020). Bayesian statistics were calculated in JASP software. The hypothesis tests performed assessed whether the evidence favored H₁ (alternative hypothesis) over H₀ (null hypothesis), which is BF₁₀ in the nomenclature of Keysers et al. (2020).

Code accessibility

Data and code will be shared upon request.

Results

In sum, 14 male, lister-hooded rats were head-fixed on a non-motorized treadmill and passively exposed to visual stimuli. Rats were free to walk or sit immobile during the experiment; however, in the absence of rewards or a goal-directed task, they remained immobile and passively viewed the stimuli. Pupillometry was performed in a closed faraday cage in total darkness (except for the computer screen) such that environmental illumination was held constant throughout the experiment and across subjects. All sensory stimuli as well as the gray screen during the inter-trial interval, were equiluminant. We presented the Asahi stimulus (Fig. 1), which humans perceive to have a bright glare in the center (Laeng and Endestad 2012). We also presented a luminance-matched control stimulus (Fig. 1) that due to the rearrangement of the Asahi stimulus into a new structure, does not evoke a brightness percept in humans (Laeng and Endestad 2012). We stimulated most or all the entire visual field by tiling the computer screen with as many stimuli as possible (Supplementary Figs. 1–3). We tested the hypothesis that the Asahi stimulus would cause a pupil constriction relative to the luminance-matched control stimulus.

Given the lack of prior work in animals, as well as differences between the human and rodent visual system (e.g. spectral sensitivity and visual acuity), we screened a range of stimulus sizes (10°, 20°, 40°, and 60°) and two colors. Although yellow stimuli (~580 nm wavelength) were used in the human study (Laeng and Endestad 2012), the spectral sensitivity of the rat retina is diminished around 580 nm and lacking sensitivity over 590 nm, whereas it is very sensitive below 530 nm (Peirson et al. 2018). Therefore, we presented 580 nm (yellow) and 508 nm (green) stimuli.

We presented the Asahi and control stimuli (4 s duration) 50 times each, in random order, and at unexpected times to avoid confounds of rhythmically entraining the pupil or brain activity. However, the surprising onsets of stimuli drive a large pupil dilation (Beatty 1982; Murphy et al. 2011; Breton-Provencher and Sur 2019) that can easily out-compete constriction driven by brightness perception. Any influence of illusory brightness on parasympathetic nervous system activity controlling the sphincter pupillae muscle (constriction) must compete against the antagonistic dilator pupillae muscle controlled by sympathetic nervous system activity. Thus, it is possible that strong surprise-evoked sympathetic activation will out-weight any influence that sensing the brightness illusion has on parasympathetic activation.

The Asahi illusion evokes a pupil constriction in rats

Despite the apparently large dilation evoked by both the Asahi and the control stimuli (Fig. 2A and B), we observed a strong and robust constriction only after the Asahi stimulus. The surprise-evoked dilation makes the constriction merely appear small due to the range of the y-axis in Fig. 2A and B; therefore, we also plot the pupil size in the 300–700 ms window after stimulus onset in the figure insets. We compared the average pupil size (500–700 ms after stimulus onset) between the Asahi and the control stimulus. The Asahi stimulus was associated with a lower pupil size relative to the control stimulus for six of the eight screened stimulus conditions (Table 1). We report effect sizes quantifying how much the baseline z-scored pupil size differed between Asahi and control stimulus in Fig. 2C. We judged which stimulus colors and sizes were optimal drivers of pupil constriction using the 95% confidence intervals around the effect size (shown in Fig. 2C, right panel). The confidence intervals estimate the probable effect sizes that would be observed in large population experiments (Calin-Jageman and Cumming 2019a, 2019b). For instance, the yellow 20° Asahi stimulus evoked a significantly smaller z-scored pupil size than the control stimulus (Table 1) with an effect size of −0.525 (Fig. 2C, yellow circle), yet the confidence intervals around this effect size tempered our beliefs of a strong difference between the Asahi and control stimuli. In this case, the estimated effect sizes ranged from the z-scored pupil size after the Asahi stimuli being as much as 1.230 lower, but potentially 0.187 higher than its control stimulus counterpart. As a contrasting case, consider instead the green 20° stimuli. The 95% confidence intervals estimated that the weakest likely difference in z-scored pupil size would still be 0.232 lower for the Asahi stimulus compared with the control stimulus.

Fig. 2.

The Asahi stimulus evokes a pupil constriction in rats. (A, B) The SEM of pupil size is plotted from 1 sec before stimulus onset until 1.75 s later. Pupil size was normalized relative to the pre-stimulus pupil size using a z-score. The gray line is the pupil size around the control stimulus and the colored line is relative to the Asahi stimulus. The insets show the pupil size in a 300–700 ms window after stimulus onset. The y- and x-axes are identical across the insets. The dotted line shows the baseline pupil size. Constriction is negative and dilation is positive. (A) plots the pupil response to yellow stimuli, whereas (B) plots the response to green stimuli. (C) The mean pupil size (z-scored to 4-s pre-stimulus baseline) from 500 to 700 ms after stimulus onset is plotted for each stimulus. Dots are individual rats. In the yellow 60°, 40°, and 20° conditions and the 10° green condition only 13 rats are plotted due to corruption of the file containing stimulus onset time markers. The right panel shows the effect size between the Asahi stimulus and the control stimulus.

Table 1.

Results of Bayesian one-tailed paired t-test of the alternative hypothesis that the average pupil size was lower after the Asahi stimulus compared to the control stimulus.

Stimulus condition	Bayes factor	Evidence favoring or against alternative hypothesis
Green, 10°	BF10 = 28.515	Strong evidence favoring
Green, 20°	BF10 = 102.058	Strong evidence favoring
Green, 40°	BF10 = 0.280	Moderate evidence against
Green, 60°	BF10 = 4.196	Moderate evidence favoring
Yellow, 10°	BF10 = 200.140	Strong evidence favoring
Yellow, 20°	BF10 = 18.734	Strong evidence favoring
Yellow, 40°	BF10 = 13.001	Strong evidence favoring
Yellow, 60°	BF10 = 0.298	Moderate evidence against

In summary, our results demonstrate that, in six of the eight stimulus conditions tested, the Asahi stimulus evoked a significantly smaller pupil size to some degree (circles in Fig. 2C are below 0). Examination of which confidence intervals are below zero indicates that the green 20° and 10° Asahi stimuli are the most optimal drivers of pupil constriction within the stimulus conditions tested in this study. Therefore, the effect of illusory brightness on pupil size may be restricted to specific color- and size-dependent conditions.

We next quantified the magnitude of constriction to better understand how large of an effect the two optimal Asahi stimuli had on pupil size relative to the pre-stimulus baseline. Constriction magnitude was defined as the mean pupil size from when the pupil began constriction (specific to each rat) until 700 ms after stimulus onset. We found that the optimal stimuli for evoking pupil constriction (green 20° and green 10° Asahi stimuli) evoked baseline z-scored constrictions of −0.528 (a 299-fold reduction) and − 0.303 (a 154-fold reduction). In this same time window of constriction to the Asahi stimulus, the control stimulus instead evoked a purely dilatory response. Pupil size increased by z-scores of +0.304 (an 820-fold increase) and + 0.267 (a 193-fold increase) for the green 20° and 10° stimuli, respectively. The pupil dilations are large but not outside the natural physiological range of the pupil, which increases to z-scores of nearly 6.0 after the stimulus (Fig. 2A, B). Thus, constriction was specific to the brightness illusion, whereas after the control stimulus there was no constriction and, in fact, the opposite occurred. Importantly, the constriction evoked by the Asahi stimulus was large and robust across subjects, despite the antagonistic competition from surprise-evoked sympathetic activation of the dilator pupillae muscle. We focus the remaining analysis on these two most effective stimuli.

It is unlikely that fixation on local changes in contrast could explain the observed pupil constriction. Stimuli tiled nearly all, or all the visual field. Fixation on local contrast changes cannot explain the pupil constriction given that rats do not have foveal vision (Euler and Wässle 1995) and the visual acuity of rats is poorer than the sharp changes in local contrast of the 10° and 20° stimuli (Artal et al. 1998; Prusky et al. 2002). Moreover, the same black-to-white contrast changes occur in the yellow and green stimuli, yet only the green Asahi stimuli were associated with a pupil constriction. Nevertheless, we compared eye position between the Asahi and control stimuli during the first 600 ms after stimulus onset found that they did not differ. Bayesian paired t-tests comparing the location of the center of the pupil between the green Asahi stimulus and green luminance-matched control stimulus supported the null hypothesis that location did not differ (x-position for 20° and 10° stimuli were BF = 0.34 and 0.34, respectively; y-position for 20° and 10° stimuli were BF = 0.29 and 0.34, respectively). Moreover, saccades were exceptionally rare (81% of trials had not a single eye movement, SEM, n = 14 rats). Given the low proportion of trials with eye movements (19%) and the visual acuity and lack of fovea in rats, it is unlikely that local contrast changes or eye movements explain the observed constriction.

We also assessed whether ongoing running could affect our findings given that pupil size changes during running in rodents (Erisken et al. 2014). We observed the rats to remain inactive and passively view the stimuli (Supplementary Fig. 4). We compared the treadmill velocity during these experiments with the velocities recorded during other experiments in which rats were trained to respond to a Go stimulus by running on the treadmill. In the goal-directed behavioral experiments, the velocities were pooled from 14 rats, 306 sessions, and 74,671 hit trials. In comparison to active behavior, the stimuli in this experiment were passively viewed without locomotion.

The visual cortex responds to the Asahi stimulus in rats

We next assessed whether the Asahi stimulus differentially engaged any cortical regions in comparison to the luminance-matched control stimulus. In 10 of the 14 rats, we recorded EEG from anterior frontal cortex to visual cortex using a 32-electrode array implanted directly onto the skull and aligned to bregma. We obtained event-related potentials (ERPs) for each electrode in a window starting 750 ms before stimulus onset and lasting until 750 ms after stimulus onset. A topographical plot revealed that the response was confined to posterior electrodes laying over visual cortex. The Asahi and the control stimuli evoked a visual cortex response for both yellow and green stimuli of all sizes (Supplementary Fig. 5A, B). A Bayesian one-sided t-test of the hypothesis that the ERP peak was larger for the “brighter” stimulus (i.e. the Asahi stimulus) in comparison to the control stimulus was supported in the case of the 10° green Asahi stimulus (BF₁₀ = 5.67, Fig. 3A, B). Notably, the 10° green Asahi stimulus also evoked a pupil constriction. Intriguingly, the 20° green Asahi stimulus was also associated with a large pupil constriction but not a larger ERP. One possible explanation for this is that neuronal response to the 10° stimulus is stronger and registered more efficiently at the EEG electrode.

Fig. 3.

An Asahi stimulus that evoked pupil constriction also evoked a larger event-related potential in visual cortex. (A) The SEM of the visual cortex EEG is plotted from 750 ms before and until 750 ms after stimulus onset. The gray line is the ERP around the control stimulus and the colored line is relative to the green 10° Asahi stimulus. (B) The maximal potential in the ERP is plotted for each stimulus. Dots are individual rats. The inset shows the effect size between the Asahi stimulus and the control stimulus in mV. The asterisk indicates support for the alternative hypothesis (BF > 3). (C) The average ERP peak magnitude across 10 rats is shown on a scale of 0 (blue) to 0.06 mV (red). Electrode locations are shown relative to bregma at the origin.

A topographical plot of the ERP peak magnitude for the 10° green Asahi stimulus shows that it was confined, not to all visual regions, but specifically to primary visual cortex (V1) for the Asahi and the control stimuli (Fig. 3C). These electrodes were located at 5.0 and 7.0 mm posterior to bregma with three electrodes (per hemisphere) at each posterior location situated laterally from bregma at 1.5 (overlaying V2MM), 3.0 (overlaying V2ML at the anterior electrode and V1M at the posterior electrode), and 4.4 mm (overlaying V1 at the anterior electrode and V1B at the posterior electrode). The ERP was largest over V1.

Finally, we sought to determine whether the V1 field potential response preceded pupil constriction, which could suggest a potential causal role of V1 on the pupil. The SEM of the ERP peak latency after stimulus onset was 244.1 ± 8.0 ms, whereas the SEM of the constriction latency was 341.0 ± 24.1 ms for the green 10° Asahi stimulus. The V1 response preceded the pupil response by ~ 100 ms.

Discussion

The effect of brightness illusions on autonomic control of the pupil are a powerful tool for studying how parts of the CNS involved in higher mental functions can affect the body (Laeng and Endestad 2012). The neuronal correlates of this “mind–body” connection are unknown, and their discovery requires an animal model that permits probing cell types and specific neuronal projections using single cell recordings and optogenetic tagging of neurons by their projection target.

Here, using a novel combination of head-fixation, pupillometry and brain-wide, global EEG recordings in rats, we show that the pupil constricts after the same brightness illusion that causes pupil constriction in humans. Thus, our results establish the rat as an animal model for studying how sensing a brightness illusion drives a physiological reaction in the body. We note two differences between the dynamics of the pupil response in our rat model and the prior work in humans (Laeng and Endestad 2012). First, in the human study, a dilation began prior to stimulus onset, which we did not observe in rats. We suspect the dilation in human work may have occurred due to the use of a fixed 0.5 s inter-stimulus interval, which would produce an expectation of a rhythmically appearing stimulus that may have led to preparatory changes in arousal and visual system activity. Indeed, sensory signal oscillations can entrain brain activity and affect expectancy (Lakatos et al. 2008; Cravo et al. 2013). In contrast, we presented stimuli at random intervals from a distribution with a flat hazard rate to diminish expectation. Second, in the human study, the Asahi stimulus evoked a long lasting constriction for ~ 400 ms, whereas we observed only a brief constriction interrupted by a large dilation. In our experiments, the stimulus presentations were unexpected and surprising, which drives stimulus-evoked dilation (Beatty 1982; Murphy et al. 2011; Preuschoff et al. 2011; Liao et al. 2016; Breton-Provencher and Sur 2019). These dilations may have prevented the constriction from continuing for a longer duration in our experiments. In contrast, the study in human subjects used rhythmically presented stimuli. Prior work has shown that such periodic stimuli either do not evoke pupil dilations or evoke smaller pupil dilations that diminish over time-on-task (Beatty 1982; Liao et al. 2016). Therefore, in the absence of a surprise-evoked pupil dilation, the constriction evoked by the Asahi stimulus would be allowed to continue uninterrupted in human subjects.

One current limitation of our animal model is that the optimal conditions for evoking pupil constriction may not be fully characterized and will require psychophysical studies. We screened a variety of stimulus sizes in two colors and found that, in most cases, the brightness illusion evoked pupil constriction. Although the spectral sensitivity of the rat retina is diminished around 580 nm (yellow) light, there is still some sensitivity which would permit yellow Asahi stimuli to have an effect. However, the constriction was particularly strong for green (530 nm) stimuli of small size (10° or 20°). Lower wavelengths and smaller size stimuli were the most optimal stimuli for driving a pupil response to illusory brightness in rats. Constructing Asahi stimuli from even lower wavelengths, such as 480 nm (blue), may prove more optimal in rats. In humans, brightness illusions evoked by blue stimuli are subjectively perceived as brighter and evoke a stronger pupil constriction than higher wavelengths (including yellow, red, and magenta stimuli) (Suzuki et al. 2019).

Although our model does not include a behavioral report of perceived brightness by the rats, our paradigm allows connecting changes in the body with a high-level mental process (Dum et al. 2019). In this case, the mental process is the processing of the sensory information present in the Asahi stimulus. The processing of sensory information has been termed “sensing” and is distinct from “perception”, which can be thought of as imbuing an interpretation on sensory information, or awareness of sensory information (Charbonneau et al. 2022). Our findings link the “mental process” of sensing illusory brightness to a change in the body state.

In support of this mental process interceding in the very fast reflexive and automatic ANS control of the iris sphincter muscle (Clarke and Ikeda 1985; Young and Lund 1994), we found that the pupil constriction to illusory brightness was delayed in comparison to the pupillary light reflex (PLR). In humans, the PLR requires approximately 250 ms after stimulus onset for high intensity light and as much as 400 ms for very low intensity light (Ellis 1981; Bergamin and Kardon 2003; Fotiou et al. 2007). A direct comparison of these latencies in rats is not possible because prior work defining PLR latency in rats used the time constant of an exponential fit (Liu et al. 2017), whereas human studies and our study defined latency as the time of maximal negative acceleration. The brightness illusion (e.g. the 10° green Asahi stimulus) required 341 ms to evoke constriction similar to very low intensity light in humans. The delayed constriction may be due to the time (additional forebrain and hindbrain synapses) required for a “mental process” to physiologically intervene in the brainstem-ANS neuronal activity controlling the PLR.

V1 may be a key node in a shared mammalian neural network for bodily reactions to illusions

We used an array of 32 electrodes to cover cortex bilaterally from frontal back to visual and identify which cortical regions respond and which cortical regions do not. The brightness illusion evoked a larger ERP than the control stimulus in only one of the two conditions that optimally drove pupil constriction (i.e. the green 10° stimulus). Surprisingly, the green 20° Asahi stimulus drove a similar pupil constriction without evoking a larger ERP relative to the control stimulus. It is possible that neurons are better tuned to the 10° stimulus and, therefore, a stronger ERP is measured at the EEG electrode. However, the ability to speculate why the larger ERP is observed after only one of the two stimuli that evoked pupil constriction is limited by the spatial resolution of EEG signals. Psychophysical studies can be used to identify the optimal color and size of stimuli for driving the pupil response in rats, and then use those optimized stimuli to drive a more effective V1 neural response. Our animal model creates the possibility to perform invasive recordings targeting single neurons that are tuned to the color and size of these optimal stimuli. These experiments will establish a tighter link between V1 neuronal activity and the pupil response to illusory brightness.

No other cortical regions responded to the brightness illusion as though it were physically brighter than the control stimulus. Visual cortex was likely to respond because illusions require the processing of the gestalt of a visual scene (Purves et al. 2004) and such high-level visual processing relies on visual cortex in non-human primates and in mice (Rossi et al. 1996; Roe et al. 2005; Pak et al. 2019; Saeedi et al. 2022). Surprisingly, however, other cortical regions did not respond more strongly to the brightness illusion. For instance, although frontal regions have been implicated in the awareness and interpretation of stimuli (Panagiotaropoulos et al. 2012) and can modulate both parasympathetic and sympathetic control of the pupil via monosynaptic input to brainstem noradrenergic neurons (Luppi et al. 1995; Joshi et al. 2016; Liu et al. 2017; Breton-Provencher and Sur 2019; Totah et al. 2021), frontal cortex was not particularly responsive to the brightness illusion. However, it is also possible that the activity of small populations of neurons in frontal cortex (or other cortical regions) is involved in the processing of the brightness illusion, but that their activity is not apparent in the EEG signal. An additional possibility is that this mind–body connection is mediated entirely within a brainstem-V1-brainstem loop.

Our data are consistent with the idea that V1 may be a potential cause of pupil constriction, given that the V1 response preceded pupil constriction. The latency of the maximal ERP after the 10° green Asahi stimulus was 244 ms, whereas the pupil did not constrict until 341 ms after this stimulus. V1 may control the ANS via either an unknown direct projection or via a poly-synaptic set of subcortical synapses. These forebrain circuits must eventually modulate one or two brainstem areas involved in this basic reflex (Clarke and Ikeda 1985; Young and Lund 1994). The midbrain olivary pretectal nucleus (OPN) is one target for forebrain neurons. The other is the post-synaptic target of the OPN: the preganglionic neurons of the Edinger-Westphal nucleus that project to the ciliary ganglion of the parasympathetic nervous system. The forebrain cannot directly influence the ciliary ganglion (located in the posterior orbital socket) or the iris sphincter muscle. A prudent interpretation of our data is that the role of V1 neuronal activity in pupillary constriction to illusory brightness is inconclusive given that the larger ERP only occurred for one of two optimal stimuli evoking pupil constriction. Resolving this question will require testing the role of V1 using optogenetic inhibition, as well as recording activity in projection-target defined V1 neurons (using opto-tagging) and in potentially involved sub-cortical structures. These experiments are impossible in humans but are now permissible using this rat model.

A new animal model of importance for studying “mind–body” interaction

Our finding establishes the first animal model for studying how the CNS response involved in sensing a brightness illusion drives a pupil response. By demonstrating that the same brightness illusion that drives pupil constriction in humans also does so in rats, we show that this type of “mind–body” connection is present at an earlier stage of evolution than previously thought. Illusory brightness-evoked pupillary constriction in rats and humans makes it possible that an elemental nervous system architecture, which supports perceptual influences on the pupil, is shared by early mammals and those that evolved more complex perceptual and cognitive processes (e.g. apes). Subjective perception and its effect on body physiology in primates may have grown from a seed that evolved in early mammals or developed independently in a case of convergent evolution. Our rat model can be used to uncover the physiological basis of this “mind–body” interaction in rodents and potentially primates.

Author contributions

Conceptualization—NT; Data acquisition and curation—DV, IR; Formal analysis—DV, NT; Methodology—DV, NT; Project administration—NT; Supervision—NT; Visualization—NT; Writing—NT.

CRediT authors statement

Dmitrii Vasilev (Data curation, Formal analysis, Investigation, Methodology), Isabel Raposo (Data curation, Investigation), Nelson Totah (Conceptualization, Formal analysis, Funding acquisition, Methodology, Project administration, Supervision, Visualization, Writing—original draft).

Funding

This work was funded by the Max Planck Society and the University of Helsinki (Helsinki Institute of Life Science).

Conflict of interest statement: None declared.