The Lighting Environment, Its Metrology, and Non-visual Responses

Light is essential for vision, but starting from the earliest weeks of life (1–5) it also drives important non-image-forming (NIF) effects that are powerful determinants of sleep (6), circadian rhythms (7), alertness (8, 9), mood (10) and hormone secretion (11). This paper is intended for lighting professionals, policy makers and researchers with a practical interest in lights' eye-mediated NIF effects, chronobiology and health. It explains and discusses a standardized light metrology (12) that is based on five retinal photoreceptor types, each of which has a distinct spectral sensitivity and may contribute to non-visual or NIF responses (13). Significantly, melanopsin is the functional photopigment for one of these five photoreceptor types.

Accumulating evidence (6, 14–21) suggests that the spectral sensitivity of melanopsin is the most successful and parsimonious model to predict responses to medium and long duration exposures to ambient light like circadian phase shifting, or modulations in pupil-size, alertness, and melatonin secretion. However, no single action spectrum or proxy will ever provide the complete picture (13, 22) for all the testable variations in intensity, timing, duration, and patterns of light exposure that can be created in laboratory settings (23, 24). Moreover, the effects of light in field settings are often confounded by various uncertainties which may be due to non-photic effects, interindividual variations in sensitivity to light (25), differences in the populations studied and the reduced environmental and behavioral control in real-life environments. Whilst acknowledging these limitations, some examples will be presented to suggest ways in which the melanopsin-based quantities from the standardized light metrology (12) can already be applied in practice.

The pineal hormone melatonin is an important, commonly used marker of circadian rhythms and the effects of light on its nocturnal secretion are well-established (11, 14, 15, 26, 27). In humans, melatonin facilitates sleep initiation and sleep consolidation (28), and is only secreted (resulting in detectable levels) during the period that we habitually sleep. Nocturnal light exposure acutely suppresses circulating melatonin levels (11), but being awake, or asleep, by itself has no direct effect on urinary melatonin (29). Under constant dim light conditions, melatonin levels start rising in the evening and peak at night about 2 h before the core body temperature reaches its nadir (denoted as CBTmin), with this nadir typically occurring a further 2 h before (habitual) wake-up time (30, 31).

The sleep-wake cycle closely follows the 24 h melatonin cycle: habitual bedtime is about 2 h after the melatonin onset (in dim light), while habitual wake-up typically occurs about 10 h after melatonin onset (in dim light), with melatonin onset being defined as the time point at which the salivary melatonin concentration increased to and stayed above either 4 pg/ml or 25% of its fitted amplitude (32, 33). Around the habitual wake-up time, melatonin concentrations are decreasing and drop to undetectable levels, even in dim light conditions. When living outdoors for a week in summer, camping under natural light and without any electric light exposure, average melatonin onset occurs near sunset, while average melatonin offset occurs before wake time, just after sunrise (34). An abrupt change of the sleep-wake cycle leaves the melatonin 24 h profile (virtually) unaffected (35), whilst a single laboratory light exposure with the appropriate timing and duration can shift the phase of the melatonin rhythm by up to 3 h (27, 36). However, negative feedback in the genetic clock mechanism, regulated by Sik1, limits the phase-shifting effects of light (37) and in jet-lagged humans and most other mammals behavioral phase shifts remain restricted to about 1 h per day (one time zone) (38).

The effects of light on the 24 h melatonin profile are shown schematically in Figures 1A–D. Morning light exposure advances the timing of melatonin secretion, facilitating earlier bedtimes and sleep onset, while evening light exposure postpones melatonin secretion, thus delaying the drive to go to bed (27). The circadian system considers light exposure that occurs before the CBTmin to be evening light, whereas light exposure that occurs during the hours after the CBTmin is considered to be morning light (27). Daytime light exposure can enhance nocturnal melatonin secretion (39), strengthen the body clock and reduce sensitivity to late evening/nighttime light exposures (40–45). Even a single 2.5 h bright light exposure in the early evening is sufficient to reduce the acute sleep-disruptive effects of late evening light exposure (46).

Figures 1E,F show the effect of morning and evening light on the sleep-wake cycle within a double-plotted actogram. When the light-dark cycle has a low amplitude, i.e., insufficient contrast between day and night, the circadian rhythm is free-running. A person that lives in constant dim light, has a sleep-wake cycle that shifts slowly to a later time every next day. This is depicted in Figure 1G, and is due to the fact that under dim light the circadian rhythm is free running at its endogenous period, which is on average about 24.2 h for humans (35, 47–50). The right combination of morning and evening light exposure entrains the circadian rhythm, so that it remains in sync with the 24 h light-dark cycle, see Figure 1H.

Evidence from the US suggests people in modern society may spend around 90% of their time indoors (51–53). The typical human indoor environment provides relatively little light during daytime, especially compared to the natural light outdoors, where illuminances may be 1, 2, or even 3 orders of magnitude higher. For instance, the European standard for lighting of work places (54) specifies minimum values for maintained horizontal illuminance in offices between 200 and 750 lx, depending on the specific task, whereas the horizontal illuminance outdoors can be as high as 150 klx (55). In the late-evening hours and at night, the widespread use of electrical light and luminous display devices results in extended exposures to light (56). Through their impact on circadian rhythms, these unnatural lighting conditions enhance eveningness (34). Moreover, modern lifestyles and (unnatural) light exposures are known to result in more “social jet-lag,” and this has negative consequences for sleep, performance, well-being and health (57, 58). Evolution shaped us to live in much brighter daytime conditions than present in our modern indoor life. For a healthy lit environment, people with a normal diurnal activity pattern (i.e., day-oriented, and usually in bed at night) need bright white light during the day, and especially in the morning, while they should reduce prolonged exposures in the late evening and avoid light as much as possible at night [see also CIE position statement (59)].

Although the introduction concentrates on chronobiology, it should be noted that chronobiological responses are just a subset of non-visual responses to light. The non-visual metrology tools described in this paper, and the information presented below, can also be applied to other retinal responses to ambient light.

Early this century a new class of retinal photoreceptor, the intrinsically-photosensitive retinal ganglion cell (ipRGC), was discovered (60). In addition to receiving extrinsic input signals from rods and cones, this class of photoreceptor expresses melanopsin which gives rise to the intrinsic light sensitivity after which it is named (13). Figure 2A shows the spectral sensitivities of the five classes of photoreceptors involved in non-visual photoreception (12), together with the well-known V(λ) function officially denoted as the spectral luminous efficiency function for photopic vision. In humans, melanopsin photoreception occurs efficiently across the short wavelength range of the visible spectrum between 420 and 560 nm, with a peak sensitivity in vivo at ~490 nm (13). Melanopsin-based signaling is more sluggish in onset and more sustained than rod or cone signaling (63–65). At least six subtypes of ipRGCs, M1–M6, have been identified in the mammalian retina (M1–M5 to date in humans) (66–69). Unlike rods and cones, ipRGCs have photosensitive dendrites that extend transversely across the retina. Figure 2B shows the relative densities of the rods, cones and ipRGCs as a function of retinal eccentricity. Melanopsin-based photoreception predicts both clock-mediated and acute non-visual responses under a range of everyday light exposures (21). The clock-mediated effects include regulation of the sleep-wake cycle and circadian phase shifting, whereas melatonin suppression, control of alertness and the steady state pupil diameter are examples of acute responses to light (17, 18, 20, 21).

During the first 5 years after birth, the crystalline lens in the human eye is still transmissive for short wavelength visible light and even for ultraviolet radiation (UVR) close to 320 nm (70). It becomes opaque to UVR at about an age of 5, and as age increases, the lens transmittance in the short wavelength range (i.e., violet and blue) of the visible spectrum decreases. Consequently, retinal photoreceptors receive less light input at older ages, particularly the short-wavelength sensitive photoreceptors (rods, S-cones and ipRGCs). Although adaptation mechanisms and neural plasticity may compensate for the age-induced decline in short-wavelength light that actually reaches the retina, the number of ipRGCs drops with age advancing beyond 50 (71). This loss of ipRGCs is accompanied by changes in cell morphology and an observable increase in randomness of the ipRGC distribution pattern.

It has been suggested that a decline in melanopsin photoreception with age could play a significant, deteriorating role in sleep and neuro-cognitive effects of aging (71), including those related to dementia as well as general senescence. It is plausible that these effects may be partly mediated by the negative effects on sleep due to a compromised non-visual circadian regulation with increasing age (72–74). Partly corroborating this hypothesis, it has also been observed that more fragmented and less stable sleep-activity patterns are associated with a higher all-cause mortality (up to ~20%) in the middle-aged and the elderly, independently of age (75).

Traditional lighting practice primarily targets visual performance, comfort and other aspects of the visual domain, quantifying lighting designs and installations and light exposures using luminous flux (in lumens), illuminance (in lux) and other visually related quantities. These quantities describe the luminous sensation of a light source under photopic conditions [i.e., for luminances above 5 cd/m² (76)], where cones drive human visual responses. Scotopic vision occurs while the eye is adapted to very low luminances (below 0.001 cd/m²). Under scotopic conditions, visual responses are driven by rods. The conversion between luminance and illuminance depends on the apparent source size measured in steradians, so general scotopic and photopic thresholds cannot be expressed in lux.

Individually, photoreceptors follow the principle of univariance, meaning they cannot discriminate between a change in intensity and a change in wavelength (77). As such, the spectral sensitivities of the human luminous sensation for photopic and scotopic vision can be described by the spectral luminous efficiency functions V(λ) and V′(λ), respectively, see Figure 2A. The spectral power of light, for instance, can be photopically-weighted or scotopically-weighted by multiplying each wavelength by V(λ) or V′(λ), respectively. Photometric units (such as the lumen, lux or candela) are obtained after summing the result (which is now a photopically- or scotopically-weighted spectrum) over all wavelengths and multiplying the result by the corresponding efficacy constants (K_m and K′_m, respectively), as described below.

By definition, monochromatic radiation with a frequency of 540 × 10¹² Hz, (which corresponds to the wavelength 555 nm in standard air1) has a luminous efficacy of 683 lm/W (78). Since the V(λ) function reaches its peak value at 555 nm, this is where the maximum luminous efficiency for photopic vision (denoted by constant K_m) equals 683 lm/W. The maximum luminous efficiency for scotopic vision (denoted by constant K′_m) equals 1,700 lm/W, which follows from the relationship K_m·V(555 nm) = K′_m·V′(555 nm).

The ratio of the luminous output (of a source) as evaluated using the scotopic efficiency function to the luminous output evaluated using the photopic efficiency function is known as the S/P ratio_. The S/P ratio is a characteristic of the spectral distribution of the light, and by definition, equals 1 for monochromatic radiation with a frequency of 540 × 10¹² Hz, or a wavelength of 555 nm (in air). An S/P ratio above 1 denotes that a light source is more activating to rods per (photopic) lumen than 1 lumen of monochromatic light at 555 nm.

Mesopic vision occurs while the eye is adapted to light levels that are in between photopic and scotopic conditions. In this range, i.e., in the mesopic regime, the combined action of rods and cones defines the human visual response. However, ipRGCs are implicated in retinal adaptation (79) and may be involved in the regulation of mesopic and photopic visual sensitivity (80).

Do and Yau (81) provided an extensive review of ipRGCs and their functions, including their roles in visual responses. Already in 2002, Hankins and Lucas had demonstrated that adaptations of the human primary cone visual pathway according to time of day are driven by a non-rod, non-cone photopigment with a spectral sensitivity profile that matches the standard profile of an opsin:vitamin A-based pigment with a peak at ~483 nm (79). The resulting curve is now widely accepted as the prototype action spectrum of the photopigment melanopsin and describes the intrinsic light sensitivity in ipRGCs. Another demonstration that melanopsin can drive visual perception comes from a case study of a blind individual lacking functional rods and cones who could report whether a monochromatic light stimulus of 480 nm was on or off, but failed to do so for other wavelengths (82).

Recent studies suggest the possibility of further melanopic influences on visual responses. Human brightness perception can be greater when the light stimulus has a larger melanopic content while being isoluminant for rods and cones (83), and further experiments have quantified the effect of melanopsin on brightness perception in more detail (84, 85). Melanopsin effects can increase brightness perception by up to 10%, especially for brightness discrimination tasks that involve little or no differences in luminance and hue (86). Finally, it is worth noting that melanopsin photoreception can also improve the detectability of coarse patterns (80). Together these results indicate that melanopsin is not only implicated in non-visual responses and visual adaptation, but may also contribute meaningfully to further visual responses like brightness perception and pattern recognition. However, proper demonstration of melanopic influences to vision is methodologically complex and still faces many challenges (87). At present, the relevance of melanopsin-based photoreception for brightness perception beyond laboratory settings is not yet settled and merits further investigation.

As detailed above, the melanopsin-based photoreception of ipRGCs constitutes an important driver of non-visual responses. In their work, many lighting designers already draw on a wide understanding of the visual, architectural and psychological aspects of light and lighting. Awareness amongst lighting professionals is increasing that next to cone-dominated metrics such as correlated color temperature (CCT), illuminance and luminance, there is a need to consider melanopsin-based photoreception in specifications, codes, recommendations and research. All these metrics are useful tools for quantifying or comparing individual aspects within a lighting scheme, but they cannot replace an experienced designer's overall appreciation of the interplay between the diverse effects of light. In addition, NIF photoreception relates to the light arriving at the eyes from all directions. This requires recommendations framed in terms of light arriving at eye level—e.g., measured normal to the visual axis in the vertical plane—rather than with reference to the light falling on the horizontal plane, walls or object surfaces.

No single action spectrum or proxy can describe all eye-mediated non-visual responses to light (13, 22). All five known receptor types can contribute to these responses, and the relative contribution of each individual photoreceptor type can vary depending on the specific response and upon light exposure properties such as intensity, spectrum, duration, timing (external and internal/circadian), prior light history and sleep deprivation state of the individual. Based on the Lucas et al. review paper (13), the International Commission on Illumination (CIE)—the worldwide body responsible for developing international standards and reports on light and lighting—has published CIE S 026:2018 “CIE System for Metrology of Optical Radiation for ipRGC-Influenced Responses to Light” (12). This new International Standard defines spectral sensitivity functions, quantities and metrics to describe optical radiation for its ability to stimulate each of the five retinal photoreceptor classes that, via ipRGCs, can contribute to the non-visual effects and functions of light in humans.

The Lucas et al. (13) authors used an opsin template and a lens transmittance function to establish five action spectra that describe the spectral sensitivity of all five known retinal photoreceptors that can contribute to non-visual responses. The CIE standard (12) adopts the same melanopsin action spectrum as the Lucas et al. (13) authors. However, for consistency with existing standards and psychophysical data, CIE S 026 adopts the 10-degree cone fundamentals (88) and the spectral luminous efficiency function for scotopic vision, V'(λ), to describe the cone and rod action spectra, respectively.

Figure 2A shows the five spectral weighting functions or action spectra, s_α(λ), for the five retinal photoreceptor classes: S cone, M cone, L cone, rhodopsin and melanopsin-encoded photoreception of ipRGCs as defined in CIE S 026. For each of these five (α-opic) photoreceptors, an α-opic irradiance can be calculated from the spectral irradiance, E_e,λ, of a (test) light source, see Table 1. The α-opic irradiance of a test light divided by its illuminance, E_v, defines its α-opic efficacy of luminous radiation (α-opic ELR). The ratio of this α-opic ELR to the α-opic ELR of standard daylight (D65) defines the α-opic daylight (D65) efficacy ratio (α-opic DER) of the test light.

Since daylight is a naturally occurring stimulus under which we evolved, it is an interesting and relevant point of reference to evaluate and express the properties of human light conditions within the built environment. The CIE standard illuminant D65 is adopted as the reference illuminant in CIE S 026 (2018) to express each of the five α-opic irradiances as a photometric equivalent quantity. These quantities are the five α-opic equivalent daylight (D65) illuminances (α-opic EDIs). Each α-opic EDI is expressed in lx and corresponds to the illuminance of D65 radiation that is required to provide an equal α-opic irradiance as the test light, for a given α-opic photoreceptor. The term “test light” used here refers to the light being considered, to differentiate it from the reference illuminant. 0002

The photometric equivalent concept adopted in S 026 is not restricted to illuminance (unit lx), and luminance (unit cd/m²). It can also be applied to other quantities such as light exposure (unit lx·h), luminous energy (unit lm·s), and luminous intensity (unit cd) 3.

Returning to CIE S 026, when describing the spectral properties of a test light, the ratio of the α-opic EDI of a test light to its illuminance defines the α-opic DER of the test light, see Table 1. In other words, the melanopic DER represents the ratio of the melanopic flux (“M”) per photopic luminous flux (“P”) of a test light, and this dimensionless quantity can usefully be thought of as the new “M/P ratio.” By definition, this ratio is normalized to 1 for the reference illuminant D65. The S/P ratio is an established lighting metric. It equals 1 for monochromatic radiation of 555 nm, as the S/P ratio effectively uses radiation of 555 nm as its normalizing reference illuminant. In case the melanopic EDI is 30 lx, the test light has the same activating effect on ipRGCs as 30 lx of radiation conforming to the spectrum of D65 daylight. In the same way, a scotopic illuminance of 30 lx indicates that the test light has the same effect on rods as 30 lx of radiation at 555 nm.

There are three different mainstream metrological approaches for quantifying visible optical radiation:

In the SI system, radiometry is described as “the field of metrology related to the physical measurement of the properties of electromagnetic radiation, including visible light.” Radiometric quantities can be unweighted, but photobiological quantities are typically weighted according to a suitable action spectrum that describes the relative efficiency of radiation as a function of wavelength in producing an effect.

Energy-based radiometry is often used by physicists, whereas photobiologists and photochemists often use the photon system, and the light and lighting professions have a strong preference for photometry. Photometry uses special SI units like cd, lm and lx. Radiometry and photometry and their units are closely related through the current definition of the SI base constant K_cd (K_cd ≈ K_m, see earlier) and the corresponding SI base unit for the photometric quantity luminous intensity, namely the candela. Of the seven SI base units (and their defining constants) the candela and its defining constant K_cd are unique in relating to human vision, rather than a fundamental physical phenomenon. The photon system is very similar to the radiometric system with energy units replaced by number of photons (requiring an adjustment4 to spectral weighting functions and quantities), and is often expressed after taking logs, due to the very large numbers involved.

Figure 3 illustrates the deep connections between these three metrological approaches. The set of quantities (illuminance, luminous flux, luminance, etc.) in the photometric system has the analogs photopically-weighted (irradiance, radiant flux, radiance) in the radiometric system and the analogs photopically-weighted photon (irradiance, flux, radiance) in the photon system. These analogs have units (lx, lm, cd/m²), (W/m², W, W/sr/m²), and (m⁻²·s⁻¹, s⁻¹, sr⁻¹·m⁻²·s⁻¹), respectively. For melanopic quantities—with exactly the same units—the respective quantities are [melanopic EDI, melanopic equivalent daylight (D65) luminous flux, melanopic equivalent daylight (D65) luminance], melanopic (irradiance, radiant flux, radiance) and melanopic photon (irradiance, flux, radiance). Equally, for the other four α-opic quantities, the same relationships hold. Under CIE S 026 definitions, melanopic equivalent daylight (D65) luminance can be abbreviated to melanopic EDL.

To calculate α-opic quantities in the radiometric, photon and photometric systems, and convert from one system to another, CIE has published an interactive Excel^TM spreadsheet, the “CIE S 026 Toolbox” (90). Access to the toolbox is free on the CIE website [doi: 10.25039/S026.2018.TB↗], and also an introductory video and a user guide are provided. The toolbox features include weighting functions, spectral weighting charts and a concise glossary.

Toolbox users can enter a spectral measurement and calculate all the quantities that are the geometric analogs of irradiance and radiance, including the illuminance and α-opic EDIs for this spectrum (Figure 4A). Alternatively, even without spectral data, users can familiarize themselves with the links between the three systems using one of the five built-in spectral distributions selected from the CIE standard illuminants (A, D65, E, FL11, LED-B3; Figure 4B).

The CIE has proposed “integrative lighting” to be the official term for lighting that is specifically intended to integrate visual and non-visual effects, producing physiological and psychological effects on humans that are reflected in scientific evidence (59, 92). In the context of this promising new approach, we reconsider the light that people are exposed to in their daily lives. To investigate and characterize potential light exposures in relation to non-visual responses, a number of measurements of familiar sources of light were made, where possible re-using information from previous investigations.

The α-opic toolbox was used to evaluate the absolute and relative melanopic content of these sources in more detail. Taken together, subject to the potential limitations of the melanopic model for predicting NIF responses to light (see Introduction), the information provides useful context and further evidence for advice relating to light and health.

Daily variations in the light environment are important for sleep, well-being and long-term health. The knowledge base concerning the contributions and interactions of retinal photoreceptors in driving non-visual effects is becoming more mature. Although the science is by no means complete, measures of the environment expressed in terms of melanopic EDI are now thought to have ecological validity. New recommendations for future building and lighting standards are therefore expected to incorporate both minimum thresholds for daytime melanopic EDI and maximum thresholds for evening melanopic EDI. These recommendations should be carefully integrated with the visual components within existing lighting codes. One way of limiting evening melanopic EDI would be by recommending dimmer lighting, and this is more effective when simultaneously lowering melanopic DERs (i.e., reducing M/P ratios). Another recommendation could be to strive for near darkness wherever people are expected to sleep at night. The CIE S 026 Toolbox has been introduced, partly to support this expected shift in lighting practice, and partly to enable researchers to expand the evidence base for future lighting standards, guidance and health advice.

Figure 5D shows that the melanopic DER for daylight on a clear day is significantly greater than the melanopic DER within a recent sample of white LED lighting with a range of CCTs. This supports the viewpoint that the LEDs sampled are relatively inefficient at producing melanopic light for a given combination of CCT and luminous flux, in agreement with others (97, 98). New lighting products, including those with tunable M/P ratios, may help to address this. Higher M/P ratios, similar to daylight, might be considered a beneficial characteristic for the daytime indoor environment. Daylight entry within the built environment is a good way to achieve this.

If the aim is to minimize melanopic light exposures, the lighting used at night for navigation and perceptions of safety should be restricted to lower M/P ratios. Increased daytime light exposures can reduce the adverse effects of evening light (39–46), and daytime light exposure may be as important as avoiding bright light before bedtime. During the day, indoor electric lighting could reproduce the melanopic light exposures (and other facets) of the outdoor environment, although this entails greatly increased indoor illuminances. Nevertheless, daylight is an excellent, natural, energy-efficient source of melanopic-rich light, and public health policies should encourage a daytime (natural) light-seeking lifestyle, especially during the first morning hours after bed and starting from the very first days after birth.

The raw data supporting the conclusions of this article are subject to UK Crown copyright and will usually be made available by the authors, without undue reservation.

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication. Both authors have contributed extensively to the gray literature on this topic in unpaid voluntary roles, including (12, 59, 89, 90).

LS's full time position at Eindhoven University of Technology is partially funded by Signify. The remaining author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.