Accident Analysis and Prevention 59 (2013) 466–468

Contents lists available at SciVerse ScienceDirect

Accident Analysis and Prevention

journal homepage: [Link]/locate/aap

Studying the role of vision in cycling: Critique on restricting research

to fixation behaviour
J.P. Schepers a,b,∗ , B.P.L.M. Den Brinker c , D. De Waard d , D.A.M. Twisk b ,
A.L. Schwab e , J.B.J. Smeets c
a
Ministry of Infrastructure and the Environment, Centre for Transport and Navigation, PO Box 5044, 2600 GA Delft, The Netherlands
b
SWOV Institute for Road Safety Research, The Netherlands
c
Vrije Universiteit Amsterdam, Faculty of Human Movement Sciences, Amsterdam, The Netherlands
d
University of Groningen, Department of Psychology, Groningen, The Netherlands
e
Delft University of Technology, Delft, The Netherlands

a r t i c l e i n f o a b s t r a c t

Article history: In a recent study published in Accident Analysis & Prevention, Vansteenkiste et al. (2013) – as one of the
Received 21 May 2013 first in this field – investigated the visual control of bicycle steering. They undertook the interesting task of
Received in revised form 18 June 2013 testing cyclists’ eye fixation behaviour against Donges’ two-level model of steering, i.e. the guidance level
Accepted 20 June 2013
to anticipate alternations in the course of the road and the stabilization level for lane keeping. Although
the laboratory experiment itself is well conducted, we believe that its results cannot be used to test the
Keywords:
two-level model of steering as developed for driving. The test track was only 15 m long, was completely
Bicycling
straight and was known in advance. Accordingly, it did not provide adequate conditions for testing the
Cycling safety
Single-bicycle crashes
guidance level. Furthermore, as the experimental lanes were much narrower than real-world cycling
Two-level model lanes, the stabilization level differed considerably from that in the real world. The study by Vansteenkiste
Fixation behaviour et al. (2013) may provide valuable insight into the role of vision in ‘precision steering’, but, as we discuss
in the paper, more elaborate research paradigms are needed to achieve more comprehensive knowledge
of the role of vision in real-world cycling and cycling safety.
© 2013 Elsevier Ltd. All rights reserved.

1. Introduction 1. the guidance level involving the perception of the instantaneous

and future course of the road and the response to it in an antici-
Bicycle use and cycling safety are important topics in transport patory open-loop control mode;
policy and research (European Commission, 2010). Single-bicycle 2. the stabilization level whereby any deviation between the vehi-
crashes, i.e. falls and obstacle collisions, are a growing concern cle’s actual path and its desired path is compensated for in a
for cycling safety in countries where cycling is common, such as closed-loop control mode.
Belgium and the Netherlands (see e.g. Schepers, 2012; Dhondt
et al., 2013; De Geus et al., 2012). It has been found that vision
plays an important role in single-bicycle crashes (Schepers and Den
Brinker, 2011) and knowledge about its role in cycling is essential Guidance involves anticipation on changes in the road’s course,
for understanding bicycle crashes and the development of pre- while stabilization refers to lane keeping, i.e. maintaining lateral
ventive measures. Vansteenkiste et al. (2013) recently published a position. Vansteenkiste et al. (2013) studied fixation behaviour of
study in Accident Analysis and Prevention in which they explored the cyclists who were required to ride in a short, straight, narrow lane.
interesting question of whether the visual control of bicycle steer- To date, theoretical models on cycle behaviour and research
ing could be explained by the two-level model of steering (TLMS) as paradigms applicable to the understanding of the cycling task are
developed for car driving (Donges, 1978). According to the model, scarce (Twisk et al., 2013; Schepers et al., 2013). Studies, such as
driver steering behaviour is conducted at two parallel levels: the study by Vansteenkiste et al. (2013) that aim to contribute to
the development of such a theoretical understanding are of great
value, as they advance the field by generating debate on methods,
∗ Corresponding author at: Ministry of Infrastructure and the Environment, Centre
validity and generalizability. In that context, this communication
for Transport and Navigation, PO Box 5044, 2600 GA Delft, The Netherlands.
needs to be understood as a critical appraisal of the suitability of the
Tel.: +31 887982457. experimental paradigm for the authors’ aim, namely, the study of
E-mail address: [Link]@[Link] (J.P. Schepers). visual control of steering in relation to safety. Note that, at times, the

0001-4575/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.
[Link]
 J.P. Schepers et al. / Accident Analysis and Prevention 59 (2013) 466–468 467

research undertaken by Vansteenkiste et al. (2013) will be referred

to as ‘the cycling study’.

2. Evaluating the task

To examine whether the visual control of bicycle steering can be

explained by the TLMS, Vansteenkiste et al. (2013) measured fixa-
tion behaviour of cyclists riding 15-m lanes with widths of either
10, 25, or 40 cm indoors, in a gymnasium. Participants were asked
to cycle the lane at self-selected speeds (low, preferred, or high)
without crossing the white tape delineating the lane edges. In the
introduction, the authors argue: “If this model can be applied to
cycling as for driving, cyclists would mainly look at distant points
while they maintain centred in the lane by attending the proximal
pathway peripherally.” As the results of the cycling study showed
that “In contrast to the two-level model of steering during car driv-
ing, the near region was actively looked at.” the authors concluded
that the fixation behaviour of the participants could only partly be
explained by the two-level model.
There is an important difference between the design of the
experimental circuits used in car driving studies and those used
Fig. 1. 10, 25 and 40 cm width lanes delineated on a bicycle track by white tape to
in this cycling study. The former generally used simulated wind- represent the width of the experimental lanes.
ing road circuits (requiring guidance), with realistic lane widths
(requiring only modest stabilization) (Land and Horwood, 1995;
Donges, 1978; Brooks et al., 2005). For instance, Donges (1978) (2013) is based on the assumption that the TLMS makes predictions
studied participants’ path along a simulated 3,2 km winding two- on fixation strategy. This is not the case; the model makes predic-
lane road. He showed that drivers constantly steered to stay in tions only on which areas in the visual field are of importance (Land
their lane (stabilization), but also steered some time before entering and Horwood, 1995; Donges, 1978). The ambient-focal dichotomy
curves to anticipate (guidance). The cycling study used very short (Leibowitz and Post, 1982; Leibowitz and Owens, 1977) helps to
(15 m long), straight lanes that participants were able to observe explain that the frequency of fixations on a certain area do not nec-
before they began the task. As it was performed indoors on a smooth essarily correspond with the importance of visual input from this
surface without irregularities or unexpected events, this means that area. For some ‘focal’ processes (for instance object recognition),
the task did not incorporate the guidance level. the fovea should indeed be directed to the relevant information (i.e.
As regards the stabilization level, it should be noted that the focal vision). However, other processes such as motion perception,
TLMS was developed for four-wheeled vehicles and, given the are equally effective if the relevant area is in the retinal periphery
fundamental differences between them and single-track vehicles, (i.e. ambient vision), i.e. if the area is not fixated. This focal-ambient
application of the TLMS to the latter is problematic. Riding a single- dichotomy corresponds nicely with Donges’ guidance and stabiliza-
track vehicle requires steering input not only for lane-keeping but tion level (Schieber et al., 2008). The fact that car drivers rarely
also for balancing. In the absence of speed a two-wheeler falls either fixate the near road (Land and Lee, 1994) is thus unrelated with
to the left or right. Maintaining the balance of a moving bicycle the importance of this area (Salvucci and Gray, 2004; Land and
requires the rider to steer into the direction of the undesired fall Horwood, 1995). It follows then that the viewing patterns iden-
so that the wheel contact points are back under the centre of mass tified in the cycling study cannot be related to the validity of the
of the system (Kooijman et al., 2011). Critical to interpretation of TLMS.
the results of the cycling study is the lateral road space needed for
balancing a bicycle. From the bicycle model presented by Meijaard
4. Discussion
et al. (2007) and steering motions observed during normal cycling
(Van den Ouden, 2011), it has been estimated that the necessary lat-
What is the reason for the difference in gaze behaviour between
eral space at a forward speed of 18 km/h is around 40 cm. Cyclists at
the cycling study and the car driving studies? We think it may
normal speeds are able to reduce this space to around 20 cm with
be that, due to the experimental cycle lanes being very narrow,
more active steering control (Godthelp and Wouters, 1979). The
lane keeping became a precision task requiring focal vision. The
experimental lanes used in the cycling study were extremely nar-
approach and findings of the research by Vansteenkiste et al. (2013)
row (10, 25 or 40 cm; see Fig. 1 for an impression of these widths),
may be relevant to cycling safety in situations requiring precision
imposing severe constraints on the stabilization. Not surprisingly, it
steering, e.g. negotiating hazards such as potholes, rain puddles, or
was found that participants who cycled in a 10 cm wide lane (about
slippery drain covers on narrow bicycle tracks. Resulting fixations
the width of a normal edge line) were able to keep the bicycle within
in the near region could interfere with other tasks which primarily
this lane for only about 60% of the time. Given that the Flemish
rely on the focal visual channel (Horrey et al., 2006), e.g. scanning
design standard indicates a minimum width of 175 cm for bicycle
for traffic while crossing an intersection. This could be tested by
tracks (Flemish Government, 2012), the laboratory task does not
observing cyclists’ viewing behaviour near intersections, e.g. less
mirror real life lane-keeping.
fixations on traffic and more on road surface hazards (as an exam-
ple of studying intersection safety using viewing behaviour, see
3. Testing the model e.g. Summala et al., 1996). Alternatively, one could use a labora-
tory setting such as the cycling study, to determine the chance that
Even if an experimental task resembling every day cycling cir- important peripheral information goes unnoticed, e.g. by adding
cumstances were used in the cycling study, the interpretation stimuli next to a cycle lane to which cyclists need not respond
would still be questionable. The reasoning by Vansteenkiste et al. directly, but are asked for afterwards (see e.g. De Waard et al.,
468 J.P. Schepers et al. / Accident Analysis and Prevention 59 (2013) 466–468

2010). This could yield some very valuable data with regard to Fabriek, E., De Waard, D., Schepers, J.P., 2012. Improving the visibility of bicy-
overload in the focal visual channel and capability of the ambi- cle infrastructure. International Journal of Human Factors and Ergonomics 1,
98–115.
ent visual channel to alert the focal channel to the occurrence and Flemish Government, 2012. Vademecum Fietsvoorzieningen (Standard for Bicylce
location of important events. Facilities). Flemish Government, Environment, Nature and Energy Department,
Identifying which areas in the visual field are critical to safe Brussels.
Godthelp, H., Wouters, P.I.J., 1979. Course Holding by Cyclits and Moped Riders.
cycling (e.g. Fabriek et al., 2012) and how visible or conspicuous Institute for Road Safety Research, Leidschendam.
information needs to be to avoid single-bicycle crashes is important Horrey, W.J., Wickens, C.D., Consalus, K.P., 2006. Modeling drivers’ visual attention
as well (e.g. Schepers and Den Brinker, 2011). Given the findings allocation while interacting with in-vehicle technologies. Journal of Experimen-
tal Psychology: Applied 12 (2), 67–78.
emanating from research into visual control and car driving, it
Kooijman, J.D., Meijaard, J.P., Papadopoulos, J.M., Ruina, A., Schwab, A.L., 2011. A bicy-
would appear that experimentally manipulating the dimensions of cle can be self-stable without gyroscopic or caster effects. Science 332 (6027),
information helps to study which parts of the visual field are impor- 339–342.
Land, M., Horwood, J., 1995. Which parts of the road guide steering? Nature 377
tant, e.g. occluding part of the visual field (near, far, peripheral,
(6547), 339–340.
central, etc.) (Schieber et al., 2008; De Waard et al., 2004). Land, M.F., Lee, D.N., 1994. Where do we look when we steer. Nature 369, 742–744.
To conclude, the research paradigm used in the cycling study Leibowitz, H.W., Owens, D.A., 1977. Nighttime driving accidents and selective visual
has shed light on the role of vision in ‘precision’ steering, but more degradation. Science 197 (4302), 422–423.
Leibowitz, H.W., Post, R.B., 1982. The two modes of processing concept and some
elaborate research paradigms are needed to test the complexities implications. In: Beck, J. (Ed.), Organization and Representation in Perception.
of the role of vision and the applicability of Donges’ TLMS to the Lawrence Erlbaum, Hillsdale, pp. 343–363.
cycling task. Meijaard, J.P., Papadopoulos, J.M., Ruina, A., Schwab, A.L., 2007. Linearized dynam-
ics equations for the balance and steer of a bicycle: a benchmark and review.
Proceedings of the Royal Society A 463 (2084), 1955–1982.
Salvucci, D.D., Gray, R., 2004. A two-point visual control model of steering. Percep-
References tion 33 (10), 1233–1248.
Schepers, J.P., 2012. Does more cycling also reduce the risk of single-bicycle crashes?
Brooks, J.O., Tyrrell, R.A., Frank, T.A., 2005. The effects of severe visual challenges on Injury Prevention 18 (4), 240–245.
steering performance in visually healthy young drivers. Optometry and Vision Schepers, J.P., Den Brinker, B.P.L.M., 2011. What do cyclists need to see to avoid
Science 82 (8), 689–697. single-bicycle crashes? Ergonomics 54 (4), 315–327.
De Geus, B., Vandenbulcke, G., Int Panis, L., Thomas, I., Degraeuwe, B., Cumps, E., Schepers, J.P., Hagenzieker, M.P., Methorst, R., Van Wee, G.P., Wegman, F., 2013. A
Aertsens, J., Torfs, R., Meeusen, R., 2012. A prospective cohort study on minor conceptual framework for road safety and mobility applied to cycling safety.
accidents involving commuter cyclists in Belgium. Accident Analysis and Pre- Accident Analysis and Prevention, [Link]
vention 45, 683–693. In press.
De Waard, D., Schepers, J.P., Ormel, W., Brookhuis, K.A., 2010. Mobile phone use Schieber, F., Schlorholtz, B., McCall, R., 2008. Visual requirements of vehicular guid-
while cycling: incidence and effects on behaviour and safety. Ergonomics 53, ance. In: Castro, C. (Ed.), Human factors of visual and cognitive performance in
30–42. driving. CRC Press, Boca Raton.
De Waard, D., Steyvers, F.J.J.M., Brookhuis, K.A., 2004. How much visual road Summala, H., Pasanen, E., Räsänen, M., Sievänen, J., 1996. Bicycle accidents and
information is needed to drive safely and comfortably? Safety Science 42 (7), drivers’ visual search at left and right turns. Accident Analysis and Prevention
639–655. 28 (2), 147–153.
Dhondt, S., Macharis, C., Terryn, N., Van Malderen, F., Putman, K., 2013. Health burden Twisk, D.A.M., Vlakveld, W.P., Dijkstra, A., Reurings, M., Wijnen, W., 2013. From
of road traffic accidents, an analysis of clinical data on disability and mortality Bicycle Crashes to Measures; Brief Overview of what we know and do not know
exposure rates in Flanders and Brussels. Accident Analysis and Prevention 50, (yet). Institute for Road Safety Research, Leidschedam.
659–666. Van den Ouden, J.H., 2011. Inventory of Bicycle Motion for the Design of a Bicycle
Donges, E., 1978. A two-level model of driver steering behavior. Human Factors 20 Simulator. Delft University of Technology, Delft, MSc thesis.
(6), 691–707. Vansteenkiste, P., Cardon, G., D’Hondt, E., Philippaerts, R., Lenoir, M., 2013. The visual
European Commission, 2010. Towards a European road safety area: policy orienta- control of bicycle steering: the effects of speed and path width. Accident Analysis
tions on road safety 2011-2020 (COM(2010) 389). Brussels. and Prevention 51, 222–227.