LED Lighting, Screens and Health

A review of the scientific literature

"Light works as if it’s a drug, except it’s not a drug at all."

-George C. Brainard, PhD, director of the Light Research Program at  Thomas Jefferson University in The New York Times (2011).  While Dr. Brainard was referring specifically to light's effect on melatonin release and circadian rhythms, there is significant evidence that light flicker also has biological effects and alters brain chemistry, as described further below. There is not scientific support for an assumption that light cannot adversely affect human health.

Is flickering light unhealthy?

Biological effects of light flicker in the scientific literature

Research on the biological effects of light flicker is very limited, largely restricted to visible flicker <100 Hz (<100 flashes per second), which is slower than most of the flicker produced by artificial lighting, and insufficient to predict what kind of lights or screens might cause adverse health effects. As LED lights became the predominant kind of light available to the public, The Institute of Electrical and Electronics Engineers (IEEE) issued a major report on the known biological effects of light flicker, stated an urgent need for further research, and recommended limits on visible and invisible flicker in LED lighting (IEEE std 1789, 2015). Despite the call for further research on potential health effects of LEDs, very little such research has been done and lighting manufacturers have supported limits on LED flicker that are less stringent than those proposed in the IEEE report (see Background: LED Lights for descriptions of how light sources flicker, metrics for measuring flicker, and proposed limits on flicker in lighting). Although some government agencies have begun to recognize that there is a need for research on the health effects of flicker, including in a 2019 ANSES report for the French government, very little research has been published.

Although flicker is not directly assessed, emerging research on the effects of screen use on the eyes and rates of headache and light sensitivity suggest that a majority of the population (at least 66-74%) is being harmed by screen use (see More screen sensitivity literature).

"There's no such thing as flicker fusion." 

-Arnold Wilkins (2014), Professor Emeritus, University of Essex, and author of hundreds of scientific articles related to lighting, flicker, pattern, visual discomfort, and health.

There is a historic emphasis on distinguishing between visible and "invisible" flicker in lighting research. These are traditionally defined as in IEEE std 1789, 2015:

Visible flicker: Variation in light intensity that occurs at a frequency that is sensed by the brain and consciously perceived - flicker that a person is conscious of seeing. Flicker is generally easily visible when slower than about 60 Hz (60 flashes per second), but can also be obviously visible as flashing light up to about 100 Hz, depending on the person and on environmental conditions (reviewed in IEEE std 1789, 2015). 

"Invisible" flicker: The variation in light intensity is fast enough that the light appears to have a constant intensity.  Historically, flicker with a frequency higher than about 90 Hz has been called "invisible"and this kind of flicker is not usually consciously perceived as flashing light [reviewed in IEEE std 1789, 2015]. "Invisible" flicker is still flicker.

Critical flicker fusion (CFF) frequency  is a very long-standing hypothesis that there is a flicker frequency above which an optical illusion occurs in which flickering light appears to people to be continuous, meaning that the flicker is "invisible" (see Landis, 1953, for an annotated bibliography of over a thousand early studies of the perception of flicker, with the first recorded attempts to measure CFF, also studied as "persistence of vision," published in 1740 by Segner and in 1765 by d'Arcy, who estimated that a flaming coal swung in a circle appeared to form a continuous circle of light at 0.133 seconds/revolution, or 7.5 Hz). The idea behind this hypothesis is the assumption that, at a high enough frequency, the human nervous system fuses discrete images created by flicker. Thus, flicker fusion makes the flicker "invisible" by creating an optical illusion that fools people into thinking that the light is continuous. Historically, the estimate of the CFF frequency for humans is somewhere in the range of 60 Hz to 100 Hz (reviewed in Brundrett, 1974) and there continues to be an emphasis on this CFF estimate in current lighting research. 

Wilkins was referring to CFF when arguing that "there's no such thing as flicker fusion," (Wilkins,  2014) and therefore not a scientific basis for dividing light flicker between "visible" and "invisible" at a frequency somewhere between about 60 Hz and 100 Hz.

There is now evidence that much of the flicker in the >90 Hz, historically "invisible," range is actually visible too. At least 800 Hz flicker can be visible as flashing light (Davis et al., 2015). Flicker up to at least 11,000 Hz can be visible as a phantom array (Brown et al., 2019). The stroboscopic effect can be visible if objects are moving under flickering light up to at least 10,000 Hz (Bullough et al., 2012). Together, these data refute the assumption that people can't see flicker above 90 Hz. They also suggest that if there is a point at which flicker becomes "invisible" to humans, it is at some unknown point in excess of 11,000 Hz (discussed further below).

CFF frequency was significantly underestimated in early studies due to limitations of the design of those studies.  The historic estimates of CFF frequency were largely based on laboratory studies of flicker perception in which the observer's gaze was fixed on a point and the field of view was uniform. Therefore, CFF frequency estimates depend upon the following assumptions that are rarely present in real-world scenarios:

Professor Wilkins argues that data from more realistic scenarios do not support either the hypothesis that there is flicker fusion by the nervous system or the hypothesis that flicker faster than about 100 Hz is invisible. He cites the existence of the phantom array effect, the observation of multiple discrete images of a flickering light during a gaze shift with flicker of at least 2,000 Hz (Wilkins, 2014), as evidence both that the flicker is not fused by the brain and that the flicker is visible. Later evidence further refutes the hypothesis of flicker fusion and further extends the range of visible flicker, suggesting that if there is a point at which flicker becomes "invisible" to humans, it is at some point in excess of 11,000 Hz (Davis et al., 2015; Brown et al., 2019; see discussion in Flicker ≥100 Hz).

CFF was a popular research focus in many different scientific fields in the first half of the 20th century with the advent of electric lights and motion pictures. CFF was measured in many species and in humans under many different conditions (see Landis, 1953, for an annotated bibliography). It was widely reported in the early flicker literature that individual sensitivity to flicker varied widely, so the historic estimate of CFF varied greatly among individuals (reviewed in Brundrett, 1974). 

The historic focus on CFF and historic methods for its measurement, the flicker frequency necessary to create an optical illusion, may have created bias against doing research focused on potentially harmful health effects of flicker at frequencies beyond these estimates.

There is not a scientific basis for assuming that light flicker must be visible in order to produce health effects, either in terms of obviously visible flicker or in terms of more subtle flicker for which the optical illusion associated with CFF might usually conceal the visibility of the flicker from one's conscious awareness. However, this assumption has dominated lighting research in recent decades. Brundrett (1974) argues that the flicker of workplace fluorescent light causes high rates of workplace headache and eyestrain in his study (discussed further in Flicker below 100 Hz) because there is a correlation between headache or eyestrain and the ability to see lamp flicker and because there is a correlation between having headache in the workplace and having greater flicker sensitivity as assessed by EEG (see discussion  in Flicker below 100 Hz). This argument and the concept of CFF frequency were conflated by the lighting industry and those studying lighting into the assumption that flicker can be harmful to people's health if people can see it, with the further assumption that flicker isn't harmful to people's health if people can't see it. The fact that normal control individuals tend to be annoyed by visibly flickering light further reinforced the tendency to focus only on noticeably visible flicker when considering health effects of artificial light sources. While there is a lot of data to support the idea that visible flicker can be harmful to human health, there is no data to support the hypothesis that flicker isn't harmful if people can't knowingly see it. Continued emphasis on measuring CFF frequency and suggested limits on flicker based on human awareness of flicker visibility have further promoted the baseless assumption that flicker isn't harmful to people's health if they don't knowingly see it. In 2016, the Commission Internationale de l'Eclairage (CIE) even redefined "flicker" for the lighting industry to only refer to noticeably visible flicker (CIE TN 006: 2016). This decision has the potential to create significant confusion or obfuscation when marketing lighting products or otherwise communicating with the public.

On this website, "flicker" refers to any repetitive change in light intensity, regardless of its visibility. This definition is consistent with the common usage of the word among the general public and also avoids introducing subjective judgment in deciding when the term should be applied. The discussions of flicker research in the next website pages are roughly divided between research on flicker <100 Hz and research on faster (≥100 Hz) flicker, the latter of which covers the flicker frequency range of modern artificial light sources. The former flicker range is more likely to be obviously visible, but the latter range includes visible flicker too.

ANSES. OPINION of the French Agency for Food, Environmental and Occupational Health & Safety on the “effects on human health and the environment (fauna and flora) of systems using light-emitting diodes (LEDs).” (2019). https://www.anses.fr/en/system/files/AP2014SA0253EN.pdf

Davis, J., Hsieh, YH. & Lee, HC. Humans perceive flicker artifacts at 500 Hz. Sci Rep 5, 7861 (2015). https://doi.org/10.1038/srep07861

Brown, E., Foulsham, T., Lee, C-s., Wilkins A.. Visibility of temporal light artefact from flicker at 11kHz. Lighting Research and Technology, 52, 371-376 (2019).  https://www1.essex.ac.uk/psychology/overlays/2019-249.pdf

Brundrett, G. W. Human sensitivity to flicker. Lighting Research & Technology. 6, 127-143 (1974). https://journals.sagepub.com/doi/abs/10.1177/096032717400600302?journalCode=lrtb

Bullough et al. Detection and acceptability of stroboscopic effects from flicker. Lighting Research & Technology 44 (2012) 477-483; first published online 2011. http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.876.126&rep=rep1&type=pdf

International Commission on Illumination. Technical note: Visual aspects of time-modulated lighting systems - definitions and measurement models. CIE TN 006:2016. CIE, 2016. http://files.cie.co.at/883_CIE_TN_006-2016.pdf

The Institute of Electrical and Electronics Engineers, Inc. IEEE Std 1789™-2015: IEEE Recommended Practices for Modulating Current in High-Brightness LEDs for Mitigating Health Risks to Viewers. 2015. http://www.bio-licht.org/02_resources/info_ieee_2015_standards-1789.pdf

Landis, C. An annotated bibliography of flicker fusion phenomena covering the period 1740 - 1952. The Armed Forces -National Research Council, Vision Committee Secretariat., June 1953, 130 pages. https://books.google.com/books?id=mY8rAAAAYAAJ

Wilkins, A. CFF is a seductive but misleading concept. Lighting Research and Technology, 46, 368 (2014). http://www1.essex.ac.uk/psychology/overlays/2014-222.pdf