Bicycle Conspicuity Aids: Do They Work?

Marc Green

In the echo chambers that are bicycle enthusiast forums, I've read many rationalizations by bicyclists for not wearing high-visibility clothing. They frequently say that it is not safe because of the moth-effect even though it is likely very rare. Bicyclists, like pedestrians, drastically overestimate their visibility (Wood, Lacherez, Marszalek & King, 2009; Wood, Tyrrell, Marszalek, Lacherez & Carberry, 2013) and are frequently unaware of the putative benefits offered by conspicuity aids, especially biological motion (Fylan, King, Brough, Black, King, Bentley, & Wood, 2020). They often deny the usefulness of conspicuity aids and simply blame drivers for inattention which is circular reasoning and an attempt to deny responsibility and to blame someone else. They say why should they have to do anything because it is all the drivers' fault. Really? If you are killed in a crash does it matter whose fault it was? You should be more concerned with how to improve your odds while you are still alive.

Bicyclists, as well as other road users and humans in general, tend to rationalize self-serving explanations for event causation. One online survey (Lacherez, Wood, Marszalek, & King, 2013) found that only one percent of bicyclists who had been in a vehicle collision said that the cause was low visibility while 61 percent attributed the collision to driver inattention. Yet, only 34 percent of the bicyclists reported wearing high visibility clothing and many rode without lights. Another study (Setiawan, 2009) found that even when lights were legally mandated, only 28 percent of bicyclists were in compliance and 39 percent had no lights at all. In general, bicyclists only infrequently use conspicuity aids (e.g., Hagel, Lamy, Rizkallah, Belton, Jhangri, Cherry, & Rowe, 2007).

Any efficacy of conspicuity aids for bicyclists depends on several variables. First, conspicuity measures fall into two groups, those concerning rider apparel and those concerning the bicycle itself. Second, optimal conspicuity aids in daylight and at night theoretically differ. Fluorescent yellow clothing should be beneficial during the day but would not be much help at night because man-made sources contain little ultraviolet light. However, yellow is still a high reflectance color and certainly more conspicuous than dark clothing. At night, lights and retroreflection produce more contrast. Some studies do not clearly differentiate between day and night effectiveness. Lastly, different types of study disagree. Controlled experimental studies generally find large effects of various conspicuity aids. Conversely, two studies that mined archival data found no compelling safety improvement of conspicuity aids but a third found a large benefit. A third methodology, a quasi-randomized control study, fit one set of bicyclists with a conspicuity aid and compared them with another group without the aid. Bicyclists in both groups rode normally for a year. A survey then found that riding with the conspicuity aids was safer. Lastly, an examination of road users in general shows that conspicuity aid effectiveness is accepted for all other major road user classes. Are bicyclists somehow different?

Experimental Evidence. Controlled experimental research finds that the choice of clothing can greatly affect visibility/conspicuity. Although most vehicle-bicycle collisions occur during the day (e.g., Isaksson-Hellman, 2012), only a few studies have specifically examined daylight conspicuity. Two daylight studies found improved bicyclist conspicuity with a fluorescent-yellow vest but only under some circumstances. One simulator study (Rogé, Laurent, Ndiaye, Aillerie, & Vienne, 2019) found that detection distance was greater but only in conditions of "high visibility" but not in "low visibility." Visibility conditions were judged by the bicyclists, which seems rather backward since the drivers are the ones who had to do the seeing and since bicyclists are poor at judging their visibility (e.g., Wood, Lacherez, Marszalek, & King, 2009). Another study (Fekety, Edewaard, Szubski, Tyrrell, & Moore, 2017; also apparently reported as Edewaard, Fekety, Szubski, & Tyrrell, 2020) examined driver ability to detect fluorescent-yellow bicyclists in daylight. Vests alone afforded no advantage, but when paired with fluorescent-yellow pants, drivers detected bicyclists at much greater distances. The authors attributed the improvement to biological motion. There is evidence that biological motion pops out of the background.

Bicyclists have fewer collisions at night, but the risk is higher when their fewer numbers are taken into account (e.g., Consumer Product Safety Commission, 1995; Twisk, 2013). The hugely beneficial effect of biological motion is a theme that also dominates research on nighttime conspicuity clothing. Numerous studies reported large visibility/conspicuity gains by placing retroreflective material or lights on cyclists' joints (Blomberg, Hale, & Preusser, 1986; Wood, Tyrrell, Marszalek, Lacherez, Carberry, Chu, & King, 2010; Koo & Dunne 2012; Wood, Tyrrell, Marszalek, Lacherez, Carberry, & Chu, 2012; Koo & Huang, 2015; Stapleton & Koo, 2017; Hemeren, Johannesson, Lebram, & Eriksson 2017). The greatest benefit occurs when the lights are on ankles, thighs, and knees (Koo & Huang, 2013). This agrees with the daytime finding that pants were effective when vests alone were not.

As in daytime, the conspicuity improvement provided by a simple retroreflective vest was not as compelling. For example, one study (Wood, Tyrrell, Marszalek, Lacherez, Carberry, Chu, & King, 2010) tested conspicuity aids on a dark test track and found that biological motion increased response distance by a factor of six compared to black clothing, but a retroreflective vest improved response distance by only a factor of two. However, another study (Hemeren, Johannesson, Lebram, & Eriksson, 2017) found no visibility improvement for bicyclists wearing a vest. The difference is probably due to the different environments. Hemeren, Johannesson, Lebram, & Eriksson (2017) conducted their study on a route that included some well-lit areas where the advantage of retroreflective material over simple contrast detection declines.

Others have examined the effects of conspicuity aids attached to the bicycle. One experimental test (Wood, Tyrrell, Marszalek, Lacherez, Carberry, & Chu, 2012) found that lights on the handlebar did not help visibility. Naturalistic studies, on the other hand, found that running lights decreased accidents by 19 percent (Madsen, Andersen, & Lahrmann, 2013) and 47 percent (Lahrmann, Madsen, & Olesen, 2018). However, another experimental study (Fotios, Qasem, Cheal, & Uttley, 2017) with distracted viewers found that bicycle lights had no positive effect on well-lit roads. One common research theme is that lights and similar aids are more effective in rural than in urban environments. The reason is likely that bicycle lights become lost in the city clutter of bright points while high ambient illumination increases overall visibility.

Another study (Costa, Bonetti, Bellelli, Lantieri, Vignali, & Simone, 2017) found that putting retroreflective markings on the bicycle frame helped conspicuity, but the bicycle seemed stationary at a distance. This would be expected because the narrow width created little looming. The study also showed that putting the retroreflective material on the pedal cranks improved nighttime detection, another biological motion effect. Other experimental research (Edewaard, Fekety, Szubski, Tyrrell, & Rosopa, 2017) examined rear lights and concluded that the best conspicuity was created by lights on the heels, again creating biological motion. The study also found that if only a single taillight were used, flashing was better than steady. Taillights also improved detection distance in daylight (Edewaard, Szubski, Tyrrell, Duchowski, 2019; Edewaard, 2020).

Abdur, Aya, Teppei, & Hisashi (2021) tested the novel approach of wrapping reflective strips around the bicycle's tire tubes to produce alternating bands of different color. The moving bicycle tires then produced a strong sense of motion and were detected at relatively long distances. In some lighting conditions, alternating red-white bands were superior to black-white.

Bhagavathula, Gibbons, Williams, & Connell (2020) tested detection distance when drivers approached bicycles from the front, rear, or side in both daylight and nighttime conditions. The results are complex because relative aid effectiveness varied with approach angle and lighting conditions. A general overview of findings is that:

• Lights. Overall, headlamps and taillights were highly effective. Flashing lights produced among the longest detection distances in most conditions but steady lights were almost as good. However, there was an anomalous finding that the lower intensity headlamps were often better than higher intensity ones. Biomotion. Another surprise was the relatively limited effectiveness of biomotion. It produced only moderate detection distances (relative to other aids) from the forward viewing angle and short detection distances from the rear and from the side. The authors suggested that biological motion disappears in the urban environment's visual clutter, an effect already reported (Moberly & Langham, 2002). But why then would viewing angle matter?


• Retroreflective vests. Confirming the studies cited above, retroreflective vests were one of the least effective aids upon side and forward approaches during the day but also at night. On the other hand, vests were among the best aids when drivers approached from the rear during both day and night. Oddly, the combination of vests and biomotion also had shorter detection distances than vests alone.


• Reflectors. Normal bicycle reflectors produced moderate detection distances in all conditions.


• Spoke lights. Self-illuminating spoke lights were the best conspicuity aid for side approaching drivers in daylight but were only moderately effective at night when a steady headlamp was best.

The study's value is limited. The authors concluded that "active" aids were better than passive, although the data often show little difference. Several results were also anomalous. Nevertheless, the authors concluded that "Bicyclists should take every precaution to make themselves visible to traffic from all directions" and recommended the use of all the conspicuity aids in all conditions. Given this conclusion, the study appears pointless.

The study also embodies problems inherent in all the experimental research on bicycle conspicuity. At best, they tested only visibility. The "drivers" expected the bicycles so the task was "search conspicuity" as opposed to the more realistic "attention conspicuity" condition of "just noticing" and identifying (Cole & Hughes, 1984). In most experimental studies, a conspicuity aid is compared to a condition where the rider is only in dark clothing in very dim illumination. This magnifies visibility effects but does not necessarily represent real-world conditions where the bicyclist may be more visible for other reasons. It also minimizes the more important cognitive conspicuity issues, which are discussed below. In short, the relevance of the experimental research to real-world bicycle conspicuity is uncertain.

Data mining and survey evidence. Other studies have mined previously collected data on bicyclists and their conspicuity aids. Thornley, Woodward, Langley, Ameratunga, & Rodgers (2008) examined the effect of bicyclist clothing using the novel metric of missed days of work. They found that bicyclists who never wore fluorescent colors missed eight times as many days of work as those who always wore fluorescent colors. This is suggestive but does not address collision probability directly. Moreover, the populations of aid wearers and non-wearers may have differed in other ways.

A data mining and a survey study are often cited as showing that conspicuity aids are not beneficial. This is not quite what the papers show upon close reading. Tin Tin, Woodward, & Ameratunga (2014) classified bicyclists into four groups based on their assumed conspicuity derived from a composite score of many different factors. They also divided bicyclists into groups depending on how often they wore the conspicuity aids. They then correlated group score with hospital data. The results are highly confusing. Overall, there was no difference between the most and least likely to use conspicuity aids, but one of the middle use groups had fewer collisions. The results looked different when divided between bicycling in the city of Auckland vs. elsewhere. In the city, those who wore conspicuity aids the most had an odds ratio greater than one, i.e., more collisions. They also had more collisions than those who wore the fewest conspicuity aids. Outside Auckland, the situation reversed. Those who most often wore conspicuity aids had the fewest collisions. This suggests that conspicuity aids are more helpful on roads with less traffic. However, the study had 2,520 participants but only 162 of them had crashed so the sample size was small. The conspicuity score was based on a mishmash of factors including clothing color and retroreflectivity, helmet color and bicycle color. Helmet and bicycle color are not always regarded as conspicuity aids. Further, the composite score might average out the positive effect of any specific conspicuity device. In addition, half of the bicyclists wearing clothing that affords front conspicuity compromised rear conspicuity by improper adjustment or by wearing a backpack (Raftery & Grigo, 2012). Lastly, biological motion is perhaps the most critical sensory conspicuity factor, but it was not one of the conspicuity factors in the composite score.

Miller, Kendrick, Coupland, & Coffey (2017) also failed to find any benefit of conspicuity aids. They conducted a study restricted to nonfatal collisions and near misses (taking action that led to an injury) in bicyclists who commuted or who were on their way to shop and excluded leisure riders. In absolute terms, mishaps rose with exposure and were more common in riders of mountain bikes than in fold-up bikes. Using a composite conspicuity aid score, they compared self-report data for 76 "case" and 272 control member groups. Despite attempts to match cases and controls, the groups differed in important ways. The cases were younger, more likely male, have less than a year of bicycling experience, and less likely to have a driver's license. They are groups who would be expected to have more collisions. The authors attempted to perform a correction for the mismatches. After performing a correction for each individual factor, results showed that those riding with conspicuity aids were more likely to have a collision or near-miss, although there were some exceptions such as a protective effect when riding a racing or mountain bicycle vs. a fold-up bike.

Again the use of a composite conspicuity score and the absence of biological motion obscure the interpretation. The reliance on self-report data always allows the possibility of self-selection bias. Moreover, with only 76 cases, the correction likely left very few cases in each category, so the N was small. This is probably why the confidence intervals were also very large, so the noise was high. Correcting for only one factor at a time leaves open the possibility of interactions between factors. The study used only a restricted range of bicyclists, those most likely to be riding in the highest traffic environments. Miller, Kendrick, Coupland, & Coffey (2017) discussed the possibility that risk compensation may make their results difficult to interpret. Those wearing the aids may feel safer. Bicyclists who believe themselves to be more conspicuous take more risks or travel in more dangerous conditions. Bicyclists are also more likely to wear conspicuity aids in situations that they consider dangerous (Aldred, & Woodcock, 2015). The beneficial effects of conspicuity aids are then offset by the more dangerous environment. It is plausible that city riders wear more aids because they frequently ride in higher volume roads with narrower lanes.

Lastly, both Miller, Kendrick, Coupland, & Coffey (2017) and Tin Tin, Woodward, & Ameratunga (2014) used a "case control" method which is easy and cheap but which has many weaknesses. Most obviously, they know little about the details of the subjects and their exposure to the risk being measured. It is usually considered a weak form of evidence. It is certainly weaker than the "randomized control" method that assigns subjects to different groups and then measures their outcomes in the future. The only such study on bicycle conspicuity aids is discussed next.

Quasi-Randomized Control Trial. One study (Lahrmann, Madsen, Olesen, Madsen, & Hels, 2018; see also; Lahrmann, Madsen, & Olesen, 2018) attempted a randomized control trial (RCT), which is usually considered more rigorous than the case-control method. It is essentially an experiment performed in the real-world. The study tested the safety effects of a fluorescent yellow jacket that had small areas of retroreflective material during normal bicycling. A population of about 6800 bicyclists was divided in half, one group given the jacket and the other not. After a year of bicycling in normal conditions, the riders filled mishaps. The data represented a sizeable number of bicycles over an extended period and hence had good ecological validity.

Table 12.2 shows the "adjusted" results for multiparty crashes. ARR is "accident rate ratio", with values of one meaning no reduction and lower values meaning fewer crashes. A multiparty "accident" occurred when either there was an actual collision or the bicyclist was toppled or injured due to the action of a counterparty. Overall, the jacket group only had 62 percent as many crashes as the no-jacket group. The decrease in crashes with motorized counterparties (cars, trucks, buses) was even lower at 52 percent. Moreover, bicyclists in the jacket group often failed to wear their jackets. When only high jacket use bicyclists were examined, the collision rate fell further to 47 percent overall. Lastly, these reductions likely underestimate the benefits because the no-jacket group bicyclists wore their normal clothing, which may have employed some other conspicuity aids.

Table 12.2 The effect of yellow jacket conspicuity aid on bicycle collisions and near-misses. ARR under 1.0 signifies a reduction in collisions and near misses. After Lahrmann, Madsen, Olesen, Madsen, & Hels (2018).


This RCT study supported the controlled research in finding that wearing a conspicuity aid correlated with a marked reduction in collisions. There was no difference in the jacket's effectiveness across lighting conditions. In this case, the only biological motion at night would have been provided by a thin strip of retroreflective material on the sleeves. Unlike a vest, however, the moving jacket arms provided some biological motion. Lastly, the yellow jacket would create more contrast than darker clothing.

The study has some limitations. The subject selection was not truly random. It suffers from loose controls because the riders did not always follow the experimental protocol and because the bicyclists might have employed additional conspicuity devices. The data are self-reports rather than objectively recorded events. As explained here, these issues are common and explain why science has so much trouble trying to answer apparently straightforward real-world questions are so hard to answer.)

Other road users. Another way of looking at bicyclist conspicuity aids is to compare them to aids used by other road users. Fluorescent colors reduce motorcyclist collisions (Wells, Mullin, Norton, Langley, Connor, Jackson, & Lay-Yee, 2004). The retroreflective tape required on trucks greatly reduces collisions (e.g., Morgan, 1961). Daytime running lights reduce automobile collisions (e.g., Elvik, Christiansen, & Olsen, 2003). Normal pedestrians seldom dress in conspicuity aids, but workers, especially near roadways, typically must wear the fluorescent yellow-green and fluorescent orange colors specified in ANSI 107. These are required by law. Similarly, fluorescent yellow-green signs are commonly used to attract driver attention about potentially hazardous conditions such as school zones.

This brings up some questions. "Why would conspicuity aids improve safety for pedestrians and motorcyclists and not for bicyclists?" Is there some fundamental difference between bicyclists and other road users?" "Are people wasting their money and effort requiring workers to wear conspicuity clothing and putting up fluorescent yellow-green signs to attract driver attention?

In reviewing the evidence, however, the data suggest that conspicuity aids are beneficial but not completely effective. This is why I would have told the bicyclist to get a fluorescent yellow vest but not to bet his life on it. Even the most valid evidence, the yellow-jacket study, found only halving of risk, which is substantial but not complete. Biological motion is likely helpful, but no existing data allow for a realistic assessment of benefit magnitude. The major problem is that there is another critical factor when it comes to evaluating sensory conspicuity aids in the wild--cognitive conspicuity.

There are two types of conspicuity at length and noted that they operate in somewhat different circumstances. Sensory conspicuity is more important when the viewer searches for a known target, as in the experimental research, and less effective for unexpected objects (but see Schieber, Willan, & Schlorholtz, 2006). Cognitive conspicuity is relatively more important in perceiving unexpected objects, which represents many real-world scenarios. Some (e.g., Rogé, Ndiaye, Aillerie, Aillerie, Navarro, & Vienne, 2017) have suggested that lack of cognitive conspicuity is the main bicycle safety issue, an effect already noted in motorcycle collisions (e.g., Wulf, Hancock, & Rahimi, 1989). Cognitive conspicuity may dominate driver perception, which explains why data mining studies find little effect of sensory conspicuity aids.

Bicycle cognitive conspicuity is largely a function of driver expectation, which is subject to two main factors. Expectation is partially due to prevalence (signal probability). Objects with low prevalence are often missed (e.g., Rich, Kunar, Van Wert, Hidalgo-Sotelo, Horowitz, & Wolfe, 2008). As the frequency of bicycles increases, they are more likely to be noticed (e.g., Thompson, Wijnands, Mavoa, Scully, & Stevenson, 2018). However, bicyclist behavior is the most important factor in preventing collisions. When bicyclists violate driver expectation, they are less likely to be noticed because they are not cognitively conspicuous. Even when the presence of a bicycle is expected, viewer response is slower and more errorful when the location is unexpected (Theeuwes, 1991).

The conspicuity aid question is best summarized by my own experiences. When I, and all the other people that I have ever worked with, go on or even near a roadway to inspect a collision site, we wear fluorescent yellow-green vests with retroreflective material. I also wear a flashing LED armband at night and some others wear additional aids such as retroreflective hats. We know that we are often violating driver expectation so the aids don't stop us from constantly checking for approaching vehicles, and if, possible having a spotter to warn us. We hope for the best but prepare for the worst.

---

A division of 2057949 Ontario, Inc.
Copyright © 2024 Marc Green, Ph. D.
Home Page
Contact Us