Flicker in traffic

– phantom array and phantom wiggle –

In this post, I:

  • briefly summarize what saccadic eye movements do and how they interfere with otherwise imperceptible flicker
  • describe a newly named phenomenon of mislocalization due to flicker: The Phantom Wiggle – Best Illusion of the Year 2021 finalist
  • give some outlook on LEDs, flicker, and the future (they don’t need to intersect… but so far, it seems like they will)

Flickering light sources in traffic turn the light source on and off at a frequency which is high enough so that most people, in most situations, do not perceive the flicker directly. These lights are primarily LED daytime running lights and tail lights, where dimming is achieved through pulse-width modulation.

Our vision is affected by this unnatural blinking whether we notice it or not. Certain situations can reveal these effects even for those who are otherwise not sensitive to flicker. In this post, I will briefly discuss two of these situations: the phantom array and the phantom wiggle.


The phantom array

When can we most easily see flicker? This depends on the size of the light source, the frequency of the flicker, and the angular velocity of displacement (how fast the image moves on the retina). When the source is small and the velocity of movement is high relative to the frequency of flicker, a series of light images appear (beads, or ghosting, or phantom array).

We sometimes see this artifact when a fast object, like a car, moves across the visual scene. Most commonly, however, it happens when we make fast (saccadic) eye movements – and in traffic we make particularly fast saccades, very frequently.

Motion blur with continuous and flickering light sources in night scene.
Phantom array by the daytime running lights of a Volvo truck.

Saccadic eye movements

Saccades are quick eye movements to connect fixations. Foveal vision is much more detailed than peripheral vision, so the eyes need to fixate the most informative parts of a visual scene one after the other. The longer the saccade, the faster it gets – but all saccades are so fast that comprehensive vision during the movement is impossible. Yet, our perception of the world is continuous…

So basically, the visual system is “filling in” the missing visual information for the duration of the saccade, with predictions based on the length and target of the planned saccade and peripheral visual information from before the saccade. It also takes into account the impoverished visual information during the saccade, which is smeared due to the great velocity.

motion blur

During the saccade a flickering light source deprives the system of the reliable visual cues of image blur and continuous change in retinal image. Instead, the flashes of light suddenly appear as salient objects* in unpredicted locations, leading to a confusion of space as well as time.

(*: relative to the direct (normal, intended, continuous) percept of a subliminally flickering source at e.g. 50% duty cycle, the light image in a phantom array is twice as bright! And at 25% duty cycle it is four times as bright, etc.)

Animation of a continuous and a flickering light source as seen during a saccadic eye movement.
Slow-motion animation of two, approximately isoluminant, light sources as seen during an 80 ms saccadic eye movement segment (recorded in an unrelated experiment). Flicker at 200 Hz with 50% duty cycle. While we assume the two sources to be equally bright, during eye movements, the flickering source appears as a "phantom array" of higher luminance and smaller surface.

The visual image just before the onset of a saccade is particularly informative for the predictions to complete the percept during the eye movement. The blurred image during the saccade is normally suppressed from perception (while it still plays a role in where the saccade will land). Therefore, different parts of the phantom array created by a flickering light are mislocalized differently, saccades become imprecise and perceptual efficiency decreases.


The phantom wiggle


In the past ten years, many have also reported another kind of percept as a result of flicker in traffic. I have made a video demonstration of it (selected to be a finalist in the Best Illusion of the Year 2021 contest) and gave it the name “phantom wiggle”:

A requirement for this phenomenon seems to be a relatively low frequency and/or duty cycle of flicker, as it is most commonly experienced with earlier models of LED daytime running lights and tail lights (e.g. Mercedes M-Class W166 / E-Class W212 / C-Class W204, Volkswagen Passat B6/B7 / Golf 5, etc.), aftermarket products, or bicycle lights. The phantom wiggle is frequently seen when the image is shaky (e.g. due to potholes) and especially when viewed through a rear-view mirror. It may also happen though, when the light and the viewer are stationary: I see some flickering lights as if they were floating, kind of a subdued version of the phantom wiggle. This, I assume, is due to small eye movements (more on this further below).

Why does the phantom wiggle happen?

We can only speculate; as opposed to the phantom array, this illusion has not been experimentally studied yet. The explanation may be similar and along the lines of the above described interference between eye movements and flicker.

Interestingly though, there are some clear differences between this phenomenon and the phantom array:

  • The image displacement is small (therefore, there are also no perceivable “beads,” or repetitions of the light image).
  • The effect appears during fixation and smooth pursuit (and not during saccadic eye movements or steady fixation with rapidly moving stimulus).
  • In case of the phantom wiggle, mislocalization is directly perceptible – in fact, it is the most central element of the phenomenon. With phantom arrays, however, mislocalization is not necessarily perceivable for the viewer (the “beads” appear as a steady snapshot of individual lights).
  • The effect seems to be enhanced when a continuous light source is adjacent to the flickering one. This is not the case with the phantom array.

The fact that the phantom wiggle appears during fixation means that large saccades are not part of the mechanism behind it. However, similarly to large saccades, space perception is altered before microsaccades too: foveal stimuli are perceived as more eccentric and vice versa. This might play a role in the generation of the phantom wiggle, but that remains an open question until tested.

Also noteworthy is the observation that the phantom wiggle is more prominent when external random shakiness is added to the stimulus as opposed to all retinal image displacement being produced by eye movements. This may have to do with either that the dynamics of the movement are different in these two cases (e.g. the amplitude being larger in the former) or maybe that the corollary discharge is used by the visual system to avoid mislocalization. The lessons learned from the phantom array (see above) suggest otherwise, but I just wanted to mention this as a possibility to explore further. (So basically the question is: how much are saccades and microsaccades similar regarding temporal light artifacts?)

Many people are bothered by perceptual effects of flicker while some never notice them. It is therefore another interesting question whether some people are more susceptible to the phantom wiggle and how this may contribute to discomfort and fatigue in traffic.

Traffic safety

The phantom array and phantom wiggle are great demonstrations of confusion of the visual system by flicker, as they can be easily experienced by most people. While the real-world safety impact is less straightforward to study and remains mostly unknown, the effect of such mislocalization on the road is clearly not positive.

Before LEDs, light sources used in automotive applications did not flicker. As LEDs are still on the rise (apparently until they replace all other types of lamps on the road), we can expect to see more flicker as well. This is technically not necessary, but pulse-width modulation is still the most common method of dimming due to practical reasons.

On the plus side though, here are three newer trends:

  1. The applied frequency of pulse-width modulated LEDs is generally rising. This can decrease the problem to some extent, particularly by making the directly visible artifacts (like those discussed in this post) become imperceptible. Subliminal aspects will remain though, even at much higher frequencies. Phantom arrays with very small light sources at high angular velocity still take place at up to 11kHz – so increasing the frequency seems to be more like a mitigation effort than a real solution.
  2. Flicker is also a problem for computer vision systems and while the negative effects on human perception have not been able to grab the attention of either industry or regulators, the great momentum behind automated driving is now calling for a solution. (Of course it may just be that software will become better able to deal with flicker than humans…)
  3. Light guide technology has developed a lot since the first introduction of LEDs on cars. This makes it much more viable to implement the dimming of LEDs without the introduction of flicker by integrating the output of several lamps.


There is a long history of studying the negative perceptual and physiological effects of flicker – for a thorough perspective, see e.g. the work of Professor Arnold Wilkins.

The conclusion is always the same: flicker is bad for us. Depending on the application, it leads to decreased reading speed, increased fatigue, headaches, mislocalization or other perceptual anomalies/illusions, or subliminal effects resulting in lower perceptual efficiency.

Regardless of the details and explanations, the effects of flicker are never positive. Yet, flicker has been ubiquitous all throughout the decades of fluorescent lamps and now increasingly again, with LEDs. The reasons for this are purely technical and can be eliminated with LEDs – so we just need to use them in a smarter way.

Image credit:

Photoholgic on Unsplash