Content uploaded by Richard Manton
Author content
All content in this area was uploaded by Richard Manton on Nov 13, 2017
Content may be subject to copyright.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
1
Using mental mapping to unpack perceived cycling risk
Richard Manton
a,b,
*, Henrike Rau
b,c
, Frances Fahy
b,d
, Jerome Sheahan
e
, Eoghan Clifford
a,b
a
Department of Civil Engineering, College of Engineering and Informatics, NUI Galway, Galway, Ireland
b
Ryan Institute for Environmental, Marine and Energy Research, NUI Galway, Galway, Ireland
a
Department of Geography, Ludwig-Maximilians-University Munich, Germany
d
School of Geography and Archaeology, College of Arts, Social Sciences and Celtic Studies, NUI Galway,
Galway, Ireland
e
School of Mathematics, Statistics and Applied Mathematics, College of Science, NUI Galway, Galway, Ireland
Abstract
Cycling is the most energy-efficient mode of transport and can bring extensive
environmental, social and economic benefits. Research has highlighted negative perceptions
of safety as a major barrier to the growth of cycling. Understanding these perceptions through
the application of novel place-sensitive methodological tools such as mental mapping could
inform measures to increase cyclist numbers and consequently improve cyclist safety. Key
steps to achieving this include a) the design of infrastructure to reduce actual risks and b)
targeted work on improving safety perceptions among current and future cyclists.
This study combines mental mapping, a stated-preference survey and a transport
infrastructure inventory to unpack perceptions of cycling risk and to reveal both overlaps and
discrepancies between perceived and actual characteristics of the physical environment.
Participants translate mentally mapped cycle routes onto hard-copy base-maps, colour-coding
road sections according to risk, while a transport infrastructure inventory captures the
objective cycling environment. These qualitative and quantitative data are matched using
Geographic Information Systems and exported to statistical analysis software to model the
individual and (infra)structural determinants of perceived cycling risk.
This method was applied to cycling conditions in Galway City (Ireland). Participants’
(n=104) mental maps delivered data-rich perceived safety observations (n=484) and initial
comparison with locations of cycling collisions suggests some alignment between perception
and reality, particularly relating to danger at roundabouts. Attributing individual and
(infra)structural characteristics to each observation, a Generalized Linear Mixed Model
statistical analysis identified segregated infrastructure, road width, the number of vehicles as
well as gender and cycling experience as significant, and interactions were found between
individual and infrastructural variables. The paper concludes that mental mapping is a highly
useful tool for assessing perceptions of cycling risk with a strong visual aspect and significant
potential for public participation. This distinguishes it from more traditional cycling safety
assessment tools that focus solely on the technical assessment of cycling infrastructure.
Further development of online mapping tools is recommended as part of bicycle suitability
measures to engage cyclists and the general public and to inform ‘soft’ and ‘hard’ cycling
policy responses.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
2
1. Introduction
Cycling safety is receiving increased attention as researchers, transport planners and cycling
advocates seek to increase uptake of the mode. A Stop Killing Cyclists protest (or ‘die in’) by
more than 1,000 cyclists in London in November 2013 dramatically highlighted the
continued risk of fatalities (The Guardian, 2013), and called for more suitable roads for
cycling. Cyclists are classed as ‘vulnerable road users’; in 2010, 1994 cyclists were killed on
the roads of 20 EU countries. Although cyclist fatalities in Europe have declined over the last
decade, cyclists remain among the most vulnerable road users. Furthermore, the decline in
cycling fatalities has not been as steep as for other road users, and cyclists now account for a
greater proportion of overall road fatalities at 7% (ERSO, 2012).
Perceived cycling safety acts as a major barrier to increasing cycling (Pucher & Dijkstra,
2000). According to Parkin et al. (2007a): “While actual, or objective risk, is relatively high
for cycling compared with other modes, the perceived risk, that is the risk that is assumed to
exist by existing and would-be mode users, is the important criterion in terms of behavioural
response”. This applies equally to people’s decision to cycle at all, their choice regarding
particular routes (e.g. avoiding roundabouts)as well as their actual behaviour (e.g. lane
position). Consideration of perceived safety is also central to successful cycling design
(Parkin & Koorey, 2012), yet there has been a lack of research into both the objective
characteristics of cycling environments as well as cyclists’ perceptions of these environments
(Ma et al., 2014).
Mental mapping is a research method that offers ample potential for recording and analysing
safety perceptions but which has not yet been fully utilised. This paper uses mental mapping
as part of a mixed-method approach to capture perceptions of cycling safety and their
relationship to the physical environment. By matching qualitative data on the perceived
quality of the cycling environment to quantitative and qualitative data on the physical
environment, the paper ‘unpacks’ major determinants of perceived cycling risk. This is tested
against a case study carried out in Galway, a university city in the West of Ireland. The
methodology and results of this paper will be relevant to engineers, planners, policymakers
and cycling advocates as part of an interdisciplinary response to improving actual and
perceived safety and increasing sustainability in transport.
2. Literature Review
2.1 Environmental Perceptions and Travel Behaviour
The relationship between environmental perceptions and spatial behaviour has interested
social scientists for decades. In the field of transport studies and traffic psychology, a body of
work contends that attitudes, perceptions, and preferences strongly influence individual’s
travel behaviour, including recent contributions from Spears et al.(2013) and Gehlert et
al.(2013). Indeed, several studies have indicated that attitudes towards public transport as
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
3
well as concerns about personal safety and traffic all play a significant role in the decision to
use public transport (Elias & Shiftan, 2012).
Transport researchers have applied attitude and behavioural theories from environmental and
cognitive psychology, such as Fishbein & Ajzen’s (1975) Theory of Reasoned Action (TRA)
and later Ajzen’s (1991) Theory of Planned Behavior (TPB), to explore the psychological
dimensions of travel behaviour and modal choice. The TRA and related models from the field
of cognitive psychology assume that individual variables such as attitudes and perceptions are
the dominant drivers of behaviour (this approach has been advocated for promoting bicycle
use by Bamberg (2012). A number of empirical studies support this contention (e.g.
Thogerson (2006)).
While often contested, the influence of perceptions cannot be ignored. Geographical and
sociological studies of crime in cities and perceptions of neighbourhood safety (Rengert &
Pelfrey, 1997; Austin et al., 2002) have shown that perception is often more important than
objective reality in shaping people’s use of the built environment, including transport
infrastructure and services. However, approaches derived from the TRA and similar theories
have increasingly been criticised for overstating the influence of perceptions and almost
completely neglecting of the role of structural and contextual factors in shaping individuals’
behaviour (Nye & Hargreaves, 2009; Davies et al., 2014). As a result the past decade has
seen the growth in perception behaviour models which incorporate contextual and situational
factors. For example, the premise of Spears et al.’s (2013) Perception-Intention-Adaptation
(PIA) model is that both cognitive processes and the physical environment have a direct
effect on travel behaviour. Similarly, Kazig and Popp (2010) have argued for a practice-
theoretical approach to how people orient themselves in urban spaces which combines
cognitive and affective aspects as well as elements of the (infra)structural context.
2.2 Cycling Risk
2.2.1 Cycling Safety and Perceptions
Safety is the primary factor in choosing whether to commute by bicycle (Noland, 1995;
Whannell et al., 2012). The major cause of cycling collisions is interaction with motorised
vehicles: 82% of cyclist fatalities and 87% of cycling injuries occur in collisions with
motorised vehicles (ERSO, 2012). Junctions pose a particular danger to cyclists: 35% of
cyclist fatalities take place at junctions, compared to 20% for pedestrians and 17% for car
users (ERSO, 2012). The main injuries to cyclists are to the legs, head and arms and the most
common types of injury are fractures (34%), bruising (31%) and open wounds (13%). Injured
cyclists spend, on average, an extra day in hospital than those injured in car collisions
(ERSO, 2012) and are classed as ‘vulnerable road users’. An uptake in cycling is seen as
particularly important from a road safety perspective as the ‘Safety in Numbers’ theory holds
that the likelihood of cycling collisions is inversely related to levels of cycling (Jacobsen,
2003).
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
4
Perception of cycling safety may be more important than objective reality in determining
uptake of cycling. These perceptions are influenced by attitudes, social norms and habits
(Heinen et al., 2010; Ma et al., 2014). Drivers’ attitudes to cyclists, for example, present a
significant barrier to cycling (Lawson et al., 2013; Wooliscroft & Ganglmair-Wooliscroft,
2015). Cyclists themselves consider many more factors than users of other modes
(Fernández-Heredia et al., 2014). Horton’s (2007) ‘fear of cycling’ goes beyond that of
collisions and traffic to include the fear of being on show, of harassment or violence, and of
seeming inept or unfit. Many of these fears are culturally embedded and socialised, e.g.
parents constrain the travel behaviour of their children based on risk perceptions (Timperio et
al., 2004; Carver et al, 2010). Collective perceptions of risk also manifest in social pressure to
wear disliked safety clothing, such as high-visibility vests and helmets (Aldred & Woodcock,
2015; Deegan, 2015); however, these do not improve perceptions of safety among cyclists
(Lawson et al., 2013).
To date, few studies of perceived cycling risk have included the characteristics of the cyclist
(e.g. age, gender and cycling frequency) (Lawson et al., 2013; Black & Street, 2014; Bill et
al., 2015), which is a gap that this paper seeks to address. The UK Department for Transport
considers the perception of cycling risk as a potential barrier to cycling and includes
perceived cycling safety in the British Social Attitudes survey (UK DfT, 2014). 61% of
people in the UK consider the roads to be too dangerous to cycle on and this varies with age
(47% of 18-24 y/o, 76% of 65+ y/o), gender (69% of women, 53% of men) and cycling
experience (48% of those who cycled in the last year, 67% of those who did not) (UK DfT,
2014). Several studies identified age and gender as factors which influence perceptions and
which also shape responses to segregated cycling infrastructure (Garard et al., 2008; Black &
Street, 2014; Ma et al., 2014; Dill et al., 2015). Cycling experience has also been shown to
influence risk perceptions and inexperienced cyclists are more likely to perceive road
conditions as hazardous (Bill et al., 2015). Sanders (2015) suggests that additional experience
and skills gained may make these cyclists more tolerant of risks, although even experienced
cyclists are concerned about a variety of possible causes of injury.
2.2.2 Infrastructure
Many authors, across various disciplines, have examined the connection between the built
environment and cycling behaviour. Key infrastructural and traffic factors that affect
perceived cycling risk include: motorised traffic volume and speed, presence of cycling
facilities, driving lane width, number of junctions and roundabouts, pavement surface, parked
cars and traffic mix (Lawson et al., 2013; Bill et al., 2015). Increased perception of cycling
crash risk can be found in areas of low density, non-mixed land uses as opposed to compact,
mixed-use neighbourhoods. This was even found when the latter areas experienced greater
actual crash risk (Cho et al., 2009). Bicycle-friendly neighbourhoods (connected streets, low-
traffic etc.) improve residents’ perceptions of the environment and these residents cycle more
often due to these positive perceptions (Ma et al., 2014).
Major streets with shared lanes are associated with greatest perceived risk while shared-use
paved paths are considered the safest form of infrastructure (Winters et al., 2012). Parkin et
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
5
al. (2007a) found that cycling facilities at roundabouts did not reduce the perceived hazard.
Cycling infrastructure on roads with heavy traffic marginally reduced perceived risk, while
completely off-road, traffic-free routes significantly reduced this perception. Cycle tracks are
perceived as the safest form of cycling infrastructure, preferred to raised cycle lanes, cycle
lanes, and on-road in traffic in Copenhagen (Jensen et al.,2007). Approximately 45% of
respondents felt ‘very safe’ cycling on cycle tracks, compared to 32% on cycle lanes and 11%
on road in traffic. These results confirm existing evidence of cyclists’ preferences for
segregated infrastructure, although there are limits to the additional travel time that cyclists
are willing to spend in order to use segregated infrastructure (Sener et al., 2009; Caulfield et
al., 2012).
2.2.3 Existing measures of cycling risk perception
The landscape of existing measures of cycling risk perception shows clear tendencies towards
infrastructural and technical considerations for practical application in traffic engineering and
urban design, e.g. cycling level of service (LoS), facility suitability, friendliness and
compatibility. The empirical backgrounds of these measures typically model infrastructural
and traffic factors associated with perceived risk(e.g. road width, traffic volume).Such
measures are useful as road sections can be rated and mapped to assist cyclists in route choice
and identify route sections in need of improvement. To clarify inconsistent terminology and
to classify measures spatially, Lowry et al. (2012) proposes three definitions:
• ‘bicycle suitability’ (perceived comfort and safety along a linear section of road)
• ‘bikeability’ (comfort, coherence, and convenience of a bicycle network)
• ‘bicycle friendliness’ (laws, policies, education, bikeability of a community)
Lowry et al. identified 13 measures of ‘bicycle suitability’ developed between 1987 and 2011
(e.g. Bicycle Compatibility Index (Harkey et al., 1998)), which vary according to
infrastructural characteristics considered, points system and weighting (see Parkin & Coward
(2009) for a review of cycle route assessments). Factors considered in these measures are:
road facility type; lane width, number and markings; cycle facility type and width; motorised
traffic volume and speed; cyclist traffic volume and speed; percentage of heavy vehicles;
presence of on-street car parking; number and type of junctions/driveways; pavement
condition and presence of a curb. The factors are weighted as adjustment factors and
combined to yield a score for bicycle suitability or perceived comfort or perceived safety.
The data collection methods for 13 perceived cycling safety studies have also been
summarised by Lawson et al. (2013) to include: video recordings, video simulations,
completion of a test course, interviews and questionnaires (see Doorley et al. (2015) for a
novel application of heart rate monitors in the assessment of perceived risk). However, only
two of the studies reviewed by Lawson et al. (2013) considered the characteristics of the
cyclists: Møller & Hels (2008) and Noland (1995). Møller & Hels investigated cyclists’
perception of risk at roundabouts, finding that safety perceptions are determined by a
combination of the characteristics of the individual cyclist (age and gender), the design of
infrastructure (e.g. cycle facility) and traffic volume.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
6
2.3 Mental Mapping: Visualising Cycling Risk Perceptions
To better understand road safety perceptions among cyclists requires a combination of
methods of data collection and analysis that can handle both quantity and quality.
Importantly, the successful application of videos, computer simulations, interactive maps and
other visual aids points towards the key role of visualisation in road safety research (cf.
Prendergast & Rybaczuk (2005) for a more general discussion of visualisation in spatial
planning). Mental mapping, a creative process that seeks to draw out and subsequently
visualise people’s experiences of their physical and social surroundings, deserves particular
attention in this context.
Mental maps are defined as “an amalgam of information and interpretation reflecting not only
what a person knows about places but also how he or she feels about them” (Johnston et al.,
1986). While all maps can serve as texts for exploring human perceptions of the landscape
(Soini, 2001), mental maps in particular have long been associated with cartography that
explores human perceptions of landscape. Lynch’s (1960) research on images in the city
represents an early landmark study in this field that reveals how different social groups view
and respond to the same environment in diverse ways. Mental maps have served to explore a
range of subjects including perceived desirability of neighbourhoods, orientation and way-
finding, perceptions of crime and migration propensities (Gould & White, 1993; Fahy& Ó
Cinnéide, 2009).
Growing interest across a range of disciplines in representations and the social construction
of places has coincided with an increased appreciation of mental mapping (Gregory, 2009).
From a land use planning perspective, approaches incorporating mental mapping offer
significant advantages over survey methods or other scale-based measures because of their
place-specific attributes (Brown and Raymond, 2007). Indeed, Brown and Raymond (2007:
108) argue that “the mapping of landscape values and special places can provide an
operational bridge between place attachment and applied land use planning that seeks to
minimize potential land use conflict”.
Research into mental maps and travel behaviour is sparse and existing studies focus
predominantly on travel route choice. As noted by Mondshein et al. (2010:849):“the limited
research to date suggests that transport infrastructure and way-finding on overlapping,
distinct modal networks – sidewalks, bike lanes, transit routes, local streets and roads, and
freeway networks – affect the development of cognitive maps and, in turn, travel behaviour”.
The limited research on transport and mental mapping that exists suggests that mode of
transport influences level of detail and quality of maps, which has significant implications for
transport planning, accessibility, and wider public policy (Mondshein et al., 2010, 2013). For
cyclists, Snizek et al. (2013) used mental mapping to study route experience in a ‘high
cycling’ environment in Denmark, whereby an online questionnaire in Google Maps allowed
participants to award positive and negative experience points. Their approach points to a
wider field of online GIS-based platforms and sensors for crowd-sourcing perceptions of
cycling safety and identifying localised risks (cf. Loidl (2014), Nelson et al. (2015) and Zeile
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
7
et al. (2015)). However, Snizek et al. (2013) did not consider the individual characteristics of
the cyclists and the effect that these may have on route experience. The following section
details our own methodological approach which responds to both opportunities and gaps
identified in the literature review.
3. Methodology
This study combines mental mapping, a stated-preference survey and a transport
infrastructure inventory to unpack perceptions of cycling risk and to make visible both
overlaps and discrepancies between perceived and actual safety risks. The results of mental
mapping and the stated-preference survey captured perceptions of the cycling environment,
while a transport infrastructure inventory collected characteristics of the objective cycling
environment. The resulting qualitative and quantitative data were matched using Geographic
Information Systems and exported to statistical analysis software to construct a model of the
individual and structural determinants of perceived cycling risk. In this context this paper
makes a significant contribution to cycling safety research by exploring the perceptions of
cycling risk through the application of mental mapping as part of a larger mixed-method
study.
3.1 Study Area
Ireland has established a national cycling target of 10% modal share by 2020, yet safety
concerns remain a major impediment to increasing cycling uptake (DTTAS, 2009a; 2009b).
Between 2013 and 2014, there was a 27% increase in vulnerable road user deaths; there were
12 cyclists killed in 2014, compared to 5 in 2013. Cyclists represent 6% of all road fatalities
despite accounting for only 2% of road users (RSA, 2014). Issues surrounding cycling safety
are gaining attention in the Irish media as shown by one recent current affairs programme
entitled ‘The growing war between cyclists and motorists, what’s happening on our streets?’
(RTÉ, 2015). This discourse has centred on conflicts between the behaviour of cyclists
(breaking red lights, cycling on footpaths) and the behaviour of motorists (aggression, verbal
abuse, speeding, dangerous driving). Short & Caulfield (2014), for example, discuss the
safety challenge of increased cycling and the incorporation of safety in policy.
To achieve the national cycling target, small, compact urban areas with a young population
are deemed to harbour significant potential for modal shift away from the car and towards
active travel modes. The present study was conducted in Galway, a university city of 75,000
people on the west coast of Ireland. The study area is affected by a number of issues that
might impede uptake of cycling and a recent qualitative study that investigated modal shift
among the workforce of a large employer found perceived safety risks in the city to be an
important barrier to walking and cycling (Heisserer, 2013).Galway experiences mean annual
rainfall of 1193 mm and the mean annual temperature is 10°C (Met Éireann, 2015). The city
has a cycling modal share of 5%, while 57% residents travel to work by car, either as a driver
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
8
or passenger (CSO, 2012). Recent cycling-related developments include the installation of
raised cycle lanes, a series of greenways and a bike-share scheme.
3.2 Survey Sampling
In this study, people in Galway City who cycle to work, school or college make up the study
population. Convenience sampling was utilised by presenting the paper-based survey to
potential participants at large events in 2013; (random sampling techniques (e.g. simple
random, cluster or stratified sampling) could not be generated due to the lack of a sampling
frame; an intercept survey was also deemed unfeasible due to the time required to complete
the survey). The National University of Ireland, Galway campus was chosen for its central
location (1 km from Galway City centre) and relatively large cycling population (cycling
modal share 12%, campus population 17,000 students and 2,000 staff (Manton and Clifford,
2012)). As the sample was not randomly selected, it was not possible to make statistical
inferences about all cyclists or indeed the population of this study (Smith, 1983); however,
the use of non-random samples does not necessarily compromise the generality of the results,
allowing for interesting quantitative findings to be generated(cf. Chow, 2002).
3.3 Mental Mapping
While traditional mental mapping studies asked participants to draw a freehand sketch
(Lynch, 1960),this study utilised a base-map of Galway City roads and streets as an assist.
Participants were provided with one map each (which included a brief written introduction,
outlining the task) and coloured pens. They were asked to draw their regularly used (at least
weekly) cycling routes and to colour each route section according to their perception of the
safety of that section of their route: Green for safe, Amber for unsafe, and Red for very
dangerous. The use of this traffic-light sequence allowed for easy expression of risk,
compared to more complex rating scales. Participants found their origin and destination on
the base map and translated their mental map into coloured ratings of risk along the route.
The mapping task was undertaken independent of any interaction with the researcher and
there were no time restrictions placed on any of the participants. Participating in this mental
mapping exercise offered respondents a chance to reflect on their everyday cycling practices
and to offer some practical local improvements.
3.4 Stated-Preference Survey
Following the mental mapping exercise, participants completed a stated-preference survey of
28 questions that reflected the findings of the reviewed literature. Questions on participants’
general cycling experience and preferences (e.g. cycling frequency, trip purpose, self-
ascribed cycling skill, typical infrastructure used, preferred infrastructure) preceded questions
on cycling safety, including involvement in road collisions. The order of questions was
designed to invoke the memory of any previous cycling collision before the participant
answered specific questions on factors affecting cycling safety, including the volume of cars
passing, volume of trucks passing, roundabouts, adjacent car parking, speed limits, road lane
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
9
width, cycle lane width, and number of junctions. Due to the level of detail involved in these
questions, participants were challenged to carefully consider each factor before ranking them
in order of importance. Finally, participants were asked to provide demographic details
including: age, gender, years spent living in Galway, employment status, household
composition, and car availability.
3.5 Transport Infrastructure Inventory
Data on infrastructural and traffic-based factors affecting safety were collected using a
transport infrastructure inventory of Galway City. These included traffic volumes (cars and
the proportion of HGVs), on-street car parking, cycling facilities, road width, and junctions.
The roads in the study area were divided into sections of similar length (generally between
junctions and using named roads where possible) and data on each road section were
collected through desk studies and site visits. The volumes of light vehicles (predominantly
cars), heavy vehicles (predominantly trucks) were retrieved from Galway City Council
(2013), based on annual traffic counts conducted between 7am and 7pm on a standard day in
November. The locations of adjacent car parking were identified on site and by using Google
Streetview. The speed limit on all roads was 50 km/h, with the exception of the NUI Galway
campus, which has a speed limit of 20 km/h. The locations of segregated cycling
infrastructure were identified from Galway City Council (2012). The widths of road and
cycle lanes were measured on site. The number of junctions in each road section was counted
from mapping. A shapefile of the road network was imported to ArcGIS and the polylines
were split according to road section and inventory data were then added as attributes to each
road section. Limitations to the assessment of perceived safety include the under-reporting of
cycling collisions, the avoidance of particular routes and the variation in route types and
location (Parkin et al., 2007a).
3.6 Data analysis
This final stage of the empirical part of the study constructed a model of perceived cycling
risk by matching the perceived environment (mental map) to characteristics of the physical
environment (inventory data). Mental maps were uploaded to ArcGIS by attributing the
colour-coded ratings of each participant (along with demographic information) to road
sections (cf. Boschmann & Cubben (2014) for sketch maps and qualitative GIS, and Snizek et
al. (2013) for map matching). This yielded a dataset in which each row represents one
observation (the rating given by one participant to one road section); this dataset was then
imported into the statistical software package SPSS (version 21) for analysis. The perceived
risk rating is the response of interest and is a qualitative variable with values Green, Amber,
Red in order of increasing perceived risk. Factors (qualitative/categorical input variables) and
covariates (quantitative input variables) include the physical characteristics of the road
section and the demographics of the individual participant. A statistical model was then
developed to identify the significant factors and covariates in perceived cycling risk.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
10
A number of features associated with the study design posed challenges for the model.
Firstly, the response data are qualitative and ordinal. Secondly, as each participant rated
several roads, observations for any given participant may be correlated. Thirdly, interactions
between several of the variables can (as in any study) also arise. Of particular interest here
are the interactions between individual-level and infrastructural variables. The presence of a
significant interaction would imply that the effect of one independent variable (e.g. an
infrastructural characteristic) on perceived risk, which is a dependent variable, differs
according to a second independent variable (e.g. a characteristic of the cyclist). Also some
variables can seriously mask the effect of others (e.g. when present, multicollinearity may
have such a masking or other adverse effect) and it was considered appropriate to exclude
certain variables (e.g. fitness)from the analysis. Bearing in mind the design and goals of the
study, it was decided to employ logistic regression and to adjust the technique for the above
mentioned possibility of correlations between participants’ ratings and allow interactions
between input variables. A Generalised Linear Mixed model was applied to investigate multi-
category responses that could accommodate the within-subject correlation through random
effects (McCullogh et al., 2008). Interaction terms were introduced for all two-way
interactions and then excluded on the basis of lack of significance at the 5% level.
Red (dangerous)was chosen, arbitrarily, as the reference category for the response variable,
Rating. Following SPSS’s mixed model analysis for multinomial regression, the
(multinomial) logistic model employed models:
as a linear function of variables representing the factors and of the covariates, along with a
random error term. The coefficient, β
i
, of a covariate, X
i
, (such as age and road width)
represents the change in the above log-odds for a unit increase in that variable; while for a
binary input variable (such as gender or segregation) the coefficient of that variable
represents the expected change in the log-odds between the reference category of that
variable to the other category. For the only input variable which has three categories, cycling
experience, there were two parameters involved to represent changes from the reference to
each of the other two categories (i.e. from inexperienced to competent and highly skilled).
For most input variables, of interest is whether a change in levels of this variable increases
the log-odds (rather than changes the log-odds); that is, tests for which the
alternative/research hypothesis is one-sided, e.g. are women are more likely than men to
perceive cycling risk (as suggested by the literature) rather than simply whether there is any
difference between men and women in perceiving cycling risk. For other input variables
(such as age), a two-sided hypothesis test is applied (the p-value for a one-sided hypothesis
test is half that of a two-sided test). In practice, it may be easier for interpretation purposes to
exponentiate the log-odds ratios, so that then the linear function described above is replaced
by an exponentiated version and one can carefully interpret the corresponding coefficients as
pertaining to changes in odds rather than changes in log-odds.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
11
4. Results and Discussion
4.1 Sample Characteristics
The number of survey participants was 104 and the total number of observations (i.e.
perceived risk ratings) was 484,an average of 4.65 observations per participant. The average
distance (subsequently included in the analysis) rated per participant was 1.95 km.
Participants’ ages ranged from 17 to 58 years (mean = 30.8 years; standard deviation= 10.7
years). The majority of participants were male, 60.6%, and this reflects the national cycling
gender gap – in Ireland 73% cyclists are male (CSO, 2012).The sample included 36% people
at work, 36% undergraduate students, 21% postgraduate students, and 6% other employment
statuses. More than half of the participants cycle everyday (51%), a further 29% cycle several
times per week and the remaining 20%cycle less often. 29% of cyclists in the study classified
themselves as highly skilled, 64% as competent and 7% as inexperienced. 14% of the sample
classified themselves as very fit, 51% as fit, 29% as of average fitness and 6% as unfit. The
majority of participants (61%) had not been involved in a collision as a cyclist. The most
common cycling purpose was commuting, followed by leisure, and health/fitness.
4.2 Perceived Environment
A total of 38 road sections in Galway City received a rating. Only road sections with a
minimum number of ten ratings were included (as road sections will be compared with
respect to a set of variables rather than compared to each other on the basis of rating, this
sample size was considered satisfactory), leaving 27 road sections in the final analysis. The
average length of these road sections was 419 metres and the total length of road network
included in the analysis was 11 km. The River Corrib divides Galway City approximately in
half, east and west. As the NUI Galway campus and the majority of residences are located
west of the river, road sections at that side of the city received the majority of ratings. The
most frequently rated roads were in the immediate vicinity of the university. Figure 1 shows a
sample mental mapping response across a route from Salthill, a seaside suburb, to the
university at the banks of the river. The start (residential roads) and end (canal towpath and
university roads) are rated as Green (safe), while one road section is coloured Amber and
another Red.
Of the 484 road section ratings, almost half (48.6%) were Green, 29% were Amber and 22%
were Red. This suggests that the majority of roads are perceived to be unsafe or very
dangerous. Furthermore, route choice, whereby cyclists avoid dangerous roads, could mask
the true extent of this perceived risk (Snizek et al., 2013). Of interest here is the relative
influence of individual and infrastructural factors in determining this ordinal rating. For
illustrative purposes in Figure 2, the three response colours have been weighted with values
1, 5 and 10 in order of increasing perceived risk. Averaging these values and forming three
equally-sized categories allows a rough comparison of perceived risk across the road
network.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
12
Figure 1 – Sample mental mapping response (Male, 31 years old)
Figure 2 – Galway City road network, indicative perceived safety ratings and locations of
cycling collisions
Also shown in Figure 2 are the locations of the 32reported collisions involving cyclists in
Galway City in 2005, 2006, 2007, 2008 and 2010 (RSA, 2014).There were no cyclist
fatalities in Galway in this period though it is believed that cycling collisions are subject to
major under-reporting (Short & Caulfield, 2014). In the absence of more reliable measures
River Corrib
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
13
(e.g. collision intensity), this source of cycling collisions was judged to be an acceptable but
basic representation of actual cycling risk. Of the 32 collisions, 23occurred on road sections
included in this study. Four collisions align with the safe category, 15 with the unsafe
category and four with the very dangerous category. It is interesting that all of the collisions
on road sections perceived as very dangerous actually took place at roundabouts, though it
should be noted that the weighting system yielded just three very dangerous road sections
other than roundabouts. Roundabouts were rated as very dangerous by all participants and
require further research for cycling safety. Within the limitations of the arbitrary weighting of
response colours and the under-reporting of cycling collisions, this suggests that some
perceptions of risk align with location of actual collisions. This is envisaged as part of a
complex connection between perception and reality, whereby actual risks play some role in
influencing cyclists’ risk perceptions, although a linear relationship is not necessarily
implied.
4.3 Physical Environment
The transport infrastructure inventory compiled the engineering and traffic characteristics of
the 27 road sections covered by mental mapping. Traffic volumes ranged between0 (canal
towpath) and 14,791vehicles per day, the proportion of HGVs between 0–4%, road lane
width between2–4 m. There were two types of segregated cycling infrastructure: raised cycle
lanes and the canal towpath (Figure 3). On-street car parking is available in some areas and
the number of junctions ranged from two to nine. Images of typical types of road and cycling
infrastructure in Galway City are shown in Figure 3.
Figure 3 – Clockwise from top left: new raised cycle lane on main road, canal towpath,
typical roundabout, and a road without cycle facilities (Google, 2015)
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
14
4.4 Stated Preferences
Participants were asked to rank nine physical factors according to their impact on cycling
safety. Based on the number of 1
st
, 2
nd
and 3
rd
rankings, three of the major safety concerns
were found to be traffic-related: the number of trucks passing, speed of traffic and number of
cars passing. Infrastructure proved to be less of a concern than traffic; and cyclists consider
the presence of a roundabout, the width of the road lane and the presence of an adjacent car
parking lane to be the most concerning characteristics of infrastructure. Other factors
expressed in qualitative responses included road condition and driver behaviour.
Following the ranking of safety concerns, participants were then asked whether they felt two
types of traffic (trucks and cars) and two elements of infrastructure (roundabout and car
parking) compromised their safety while cycling, gauged on a 5-pointLikert scale.59.2%
agreed that the number of trucks passing compromised their cycling safety, while 54.5%
agreed that the number of cars passing was a major issue. 42.6% are concerned by the
presence of a roundabout, but adjacent car parking, which can result in ‘dooring’, concerned
just 14.9% of participants. The maximum speed limit of a road that most participants (57%)
would feel comfortable sharing with motorised traffic is less than50 km/h, 26% said 50-60
km/h and 17% said 60-80 km/h.
Participants were asked to rank their frequency of use and preferred type of cycling
infrastructure or on-road cycling positions. Figure 4 shows the results of the participants’
actual riding locations and shows that reasonable numbers always cycle on-road, mostly in
the secondary riding position (closer to the kerb, rather than ‘taking the lane’). Some
participants stated that they always cycle on the footpath, potentially indicating significant
fear of interaction with traffic. Figure 4 also shows the participants’ preferred cycling
locations with raised cycle lanes (footpath level), road-level cycle lanes and greenways
receiving the highest rankings. The disparity between this clear preference for segregated
cycling infrastructure and actual levels of on-road cycling suggests a deficit of dedicated
cycling infrastructure, a finding in line with Caulfield et al. (2012).
Figure 4 – Actual and preferred cycling infrastructure
23.2 23.0
15.4 14.0
9.2 9.4 5.8
5.7 1.6
21.3
47.5
4.9 2.5
16.4
0
5
10
15
20
25
30
35
40
45
50
On-road
(secondary
pos.)
On-road
(primary
pos.)
Road-level
cycle lane
Raised
cycle lanes
On the
footpath
Shared
bus-cycle
lane
Off-road
greenway
% Participants
Always use
Prefer to use
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
15
Finally, the impact of participants’ route choice must be considered. Cyclists may avoid roads
that they identify as dangerous, e.g. those with heavy traffic. This would lead to a disparity
between stated preference results and mental mapping results, as cyclists may not use the
roads they perceive to be most dangerous. However, this was not determined to be significant
factor in this survey as the mental mapping results show that the vast majority of participants
chose the most direct route between origin and destination, most likely due to the lack of
route choice in Galway City which does not have a grid pattern. Many cyclists will also
temper safety concerns with time and distance delays caused by alternative routing.
4.5 Modelling Perception of Cycling Risk
A Generalised Linear Mixed Model was built in SPSS, where the Subject was the participant
(using a unique participant number to identify repeated measurements) and the Target was
the perceived risk rating. The Measurements were the 484 observations, including associated
demographic and infrastructural data. The goal was to assess the extent to which the ordinal
variable Rating relates to nine main qualitative and qualitative effects (Table 1). The
qualitative variables are: gender, cycling experience [inexperienced/competent/highly
skilled], segregation [of cycling facility; yes/no], parking [adjacent car parking; yes/no]. The
quantitative variables are: age, LV [per 1000 light vehicles per day], %HV [percentage of
heavy goods vehicles], width [of road lane in metres], and number of junctions (Table 1).
Table 1 – Variable information
Variable Category n Percent
Minimum
Maximum
Qualitative
Rating
Green 235
48.6
Amber 141
29.1
Red 108
22.3
Gender Female 189
39.0
Male 295
61.0
Cycling experience
Highly Skilled 160
33.1
Competent 298
61.6
Inexperienced 26 5.4
Segregation Not Segregated
324
66.9
Segregated 160
33.1
Parking No Parking 230
47.5
Parking 254
52.5
Quantitative
Age (years) 484
17 58
LV (1000 veh) 484
0 15
%HGV 484
0 3.9
Width (m) 484
2 4
Junctions (no.) 484
2 9
Figure 5a displays the percentage of participants for each category of gender. These results
suggest that female participants perceived more roads as very dangerous and fewer roads as
safe. Figure 5b illustrates the corresponding summary for segregation, which appears to have
a strong effect: dedicated cycling facilities received a larger proportion of safe ratings than
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
16
road sections that involve cycling in motorised traffic. Chi-squared tests showed that there is
a significant relationship between gender and rating (X
2
= 6.632, p-value = 0.036) and
between segregation and rating (X
2
= 48.033, p-value = 0.000) (of course, these tests have not
removed the effect of other variables). Both of these observations were also suggested by the
literature and the potential interaction of individual and infrastructural variables is also of
interest. For example, female participants rated a greater proportion of segregated
infrastructure than their male counterparts – potentially as they are more likely choose a route
on segregated infrastructure – as did older people and inexperienced cyclists.
Figure 5 – Rating plotted against Gender (left) and Segregation (right)
To account for interactions between pairs of variables, all two-way interaction terms were
initially included in the analysis and then systematically dropped according to their effect on
the significance of main effects. Some variables have the potential to mask the effect of
others and it was deemed necessary to exclude these. Fitness, for example, was dropped at an
early stage of the analysis as it was found to be highly correlated with, and masking the effect
of, Cycling Experience; this was also the case with Years Living in Galway and Age.
Random Effects were included to account for within-subject correlations. The fitted
Generalized Linear Mixed Model components are shown in Table 2. In this table, each
coefficient,
, estimates the change in the log-odds of Green or Amber relative to Red for a
unit increase in a quantitative variable (units are denoted in parenthesis for quantitative
variables) and as the change in the log-odds between the reference and the other category (or
other categories) for qualitative variables. The exponentiated log-odds ratio, Exp(
), then
represents changes in odds; the 95% confidence interval for the true underlying odds, Exp(
),
is also shown in Table 2. Significance is implied by the magnitude of the p-value, displayed
in Table 2 for two-sided hypothesis tests and is halved for cases where the alternative
hypothesis is one-sided.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
17
Table 2 – Individual and infrastructural effects on perceived cycling risk
Ref=Red
Exp(
) 95% CI for Exp(
) p-value
Lower Upper
Individual characteristics
Age (years) 0.022 1.024 0.984 1.066 0.240
Gender [ref=Male]
Female 1.526* 4.601 1.336 15.847 0.008
Cycling Experience [ref=Inexperienced]
Highly Skilled -1.563* 0.210 0.045 0.982 0.024
Competent -1.694* 0.184 0.043 0.787 0.012
Infrastructural characteristics
LV (1000 vehicles) 0.176** 1.192 1.076 1.321 0.001
HV (percent) 0.304 1.355 0.903 2.035 0.142
Width (m) -0.977* 0.377 0.153 0.929 0.034
Junctions (number) 0.006 1.006 0.873 1.159 0.932
Parking -0.521 0.594 0.266 1.325 0.203
Segregation -2.993** 0.050 0.009 0.269 0.001
Interactions
Age*[Segregation] 0.070* 1.072 1.029 1.118 0.001
%HV*[Gender = Female] -0.500* 0.607 0.379 0.971 0.037
*Significant at the 5% level; **Significant at the 1% level
Individual characteristics
The coefficient for gender in the fitted model in Table 2 is
= 1.526 and the corresponding
exponentiated value is
= 4.6. This means that the estimated log odds of choosing Red
would increase by 1.526 for a female relative to a male (or equivalently, the estimated odds
of belonging to Red relative to the reference value Green or Amber is for a female 4.6 times
larger than its value for a male), when the other input variables are held constant. In other
words, female respondents are significantly more likely to rate a road section as dangerous
than are their male counterparts.
1
Turning to cycling experience, being a highly skilled or
competent cyclist decreased the odds of perceiving risk by a factor of 0.18 (p-value = 0.024)
and 0.21 (p-value = 0.012), respectively, compared to inexperienced cyclists. Significant
interactions were found between age and segregation and between gender and %HV. These
interactions confirm the hypothesis that the effect of some infrastructural variables differs
with individual characteristics, but complicate the interpretation of the main effects. These
results regarding gender and cycling experience confirm the findings of several other studies
(Lawson et al., 2013; Black & Street, 2014; Ma et al., 2014; Bill et al., 2015; Dill et al.,
2015). Future transport policymakers and planners should thus consider the roles of gender
and the lack of cycling experience in the promotion of cycling.
1
When is the corresponding true log odds, consider testing the null hypothesis
versus the (one-
sided) alternative hypothesis
, or equivalently testing the alternatives
versus
, the p-value associated with the estimate is 0.008.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
18
Infrastructural characteristics
Of the six infrastructural variables, the number of cars (LV), width of the road lane, and
cycling segregation were significant. The odds of rating a road section as dangerous
decreased with width by a factor of 0.38 (p-value=0.01) for each additional metre. The
number of cars passing increased the odds of perceptions risk by a factor of 1.2 (p-value
<0.005) for each 1000 vehicles. Segregation had a particularly strong effect (Exp(
)= 19.9,
p-value <0.005): the presence of a segregated cycling facility significantly increased
perceptions of safety. These findings confirm existing research on cyclists’ preferences for
segregated infrastructure (Caulfield et al., 2012; Lawson et al., 2013) as well as policy and
advocacy for reduced motorised traffic volumes and increased overtaking distances.
However, it is important to note that additional road lane width is unlikely to yield benefits
for cycling safety as motorists typically adapt their behaviour to these conditions by
increasing speed (cf. Lewis-Evans & Charlton (2006)).
Choice of model
The Generalized Linear Mixed Model (GLMM) correctly predicted 92% of Green (safe)
responses and the overall percentage correctly predicted was 67%. Two other models were
developed, namely multinomial logistic and ordinal logistic. Both of these models gave the
same results in terms of significance of the various factors and covariates but differed from
the mixed model multinomial logistic analysis in that segregation and the interaction between
%HV and gender each lost its significance. It is interesting to note that the mixed model
employed, a multinomial logistic, has allowed for possible correlation between observations
on the same person, whereas the (non-mixed) multinomial and ordinal logistic models
assume independence of all response observations. Future research could explore which
model is more appropriate for the analysis of data from this study design.
While the analyses illustrated in this study demonstrates the potential major factors in
determining perceived cycling safety, the fact that the data were not strictly generated by a
probabilistic sampling design method, and the fact that variations of models that were fitted
(e.g. different ways of modelling within-cyclist correlation) gave slightly different results for
the significance or non-significance of certain variables, it is suggested that the results above
may best be viewed as exploratory and as suggestions of approaches to be pursued on new
data by future researchers rather than as ‘definitive’ statistical inferential conclusions.
Overall, it is envisaged that the innovative methodology developed in this paper has opened
up a fruitful avenue for further mixed-method cycling safety research.
5. Conclusions
Perceived cycling risk has the potential to overshadow objective cycling risk as the major
barrier to increasing uptake of cycling. Perceptions of cycling have received substantial
academic attention over recent years; however, this work has focused on infrastructural
determinants of perceived risk and rarely considers the characteristics of the cyclist. This
study draws on attitude and behaviour theory to argue that cycling perceptions exist within a
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
19
broader model of attitudes, social norms and habits (Heinen et al., 2010) that need to be
understood and that new quantitative and qualitative methods are required to explore
perceptions of risk. The paper presents mental mapping, a stated-preference survey and a
transport infrastructure inventory to unpack perceptions of cycling risk and to make visible
both overlaps and discrepancies between perceived and actual characteristics of the physical
environment. While the more ‘traditional’ self-reported survey uncovered significant data
related to perceptions of cycling risk, we argue that the data derived from the mental mapping
approach has the potential to provide a more specific, placed-based assessment of these risks.
Upon critical reflection, the resulting maps display a snapshot of the geographical distribution
of selected elements but exclude cyclist’s in-depth cycling knowledge and experiences.
Further work is needed to include these qualitative aspects in analyses and debates regarding
perceived and actual cycling safety.
Participants’ mental maps (n=104) delivered rich perceived safety data (n=484) and initial
comparison with locations of cycling collisions showed alignment between perception and
actual conditions, particularly relating to danger at roundabouts. Attributing individual and
infrastructural characteristics to each observation, a Generalized Linear Mixed Model
subsequently identified segregated infrastructure, road width and traffic volume as well as
gender and cycling experience as significant. These results confirm previous research on
participants’ stated preferences and suggest interactions between the characteristics of the
cyclist and infrastructural conditions in the perception of cycling risk. Future data collection
could consider randomly-selected samples and more controlled physical environments to
better understand these interactions.
While the size and nature of the sample does not allow for inferences about the wider
population of cyclists, the findings nevertheless confirm observations made in cycling safety
documents and contributions to cycling policy by cycling campaigners and lobby groups in
low-cycling countries such as Ireland and the UK. Regarding cycling in traffic, these include
calls for reductions in traffic speeds and volumes, as well as for changes to legislation, such
as an increase in overtaking clearance distance to 1.5 m.This study also contributes to the
integration-segregation debate by demonstrating the importance of segregation for reduction
in perceived risk (cf. Parkin et al.,2007a; 2007b). Gaps between participants’ stated
preferences and actual cycling behaviour suggest a segregated cycling infrastructural deficit
in the city under study, whereby most would prefer to cycle in cycle lanes, yet in practice
cycle on road in traffic. Cyclists are a heterogeneous group however and characteristics such
as gender and cycling experience influence risk perceptions and infrastructure preferences.
Segregated infrastructure may well bring safety benefits for large sections of the population,
but space restrictions, indirect routes and junction requirements mean that sharing the road
with motorised traffic remains cyclists’ primary means of negotiating urban areas. A
combination of carefully-designed dedicated-space for cycling and making roads safer for
cycling, for example by reducing traffic speeds and volumes, is recommended for improving
safety perceptions among current and future cyclists.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
20
Moving beyond a focus on infrastructural provision, the findings presented in this paper have
significant implications for future cycling policy. As previous research reveals,
misconceptions among different groups of road users continue to negatively affect the safety
of vulnerable groups and remain a source of tension. The Irish government's target for 10%
cycling modal share by 2020 requires a serious commitment to changing current attitudes and
improving interactions between motorised vehicles and cyclists. National policy initiatives
could be designed to both dispel prevailing perceptions of risks and raise awareness of the
vulnerability of non-motorised road users. Furthermore, interventions could be targeted at
those user groups, for example women, which are particularly sensitive to perceptions of
cycling risk (cf. Garard et al. (2012)) as part of broader policy of dismantling the ‘fear of
cycling’.
The mixed method used in this study is a reflection of the interdisciplinary nature of the
project team, drawn from civil engineering, sociology, geography and statistics. There is
clearly potential to further develop the mapping and matching method as well as other mixed-
method approaches in transport studies in the future. Indeed, there is a dearth of research
exploring how transport brings individuals into cognitive and physical contact with their built
environments (Mondshein et al., 2013), and this study has shown that mental mapping has
latent potential as a research tool in this respect. Building on the success of this method,
further research is recommended on bicycle suitability measures and online mapping tools.
Engaging cyclists and the general public through GPS-based mobile applications and the
crowd-sourcing of data, including elements of mental mapping, can further unpack
perceptions of cycling risk and feed into ‘soft’ and ‘hard’ cycling policy responses.
Acknowledgements
This research was funded by NUI Galway through the College of Engineering & Informatics
Postgraduate Fellowship Scheme and by NUI Galway Students’ Union through the Explore
Innovation Initiative.
References
Ajzen, I. (1991).The Theory of Planned Behavior. Organizational Performance and Human
Decision Processes, 50, pp.79–211.
Aldred, R. and Woodcock, J. (2015). Reframing safety: an analysis of perceptions of cycle
safety clothing. Transport Policy, 42, pp.103-122.
Austin, M.A., Furr, A.L. and Spine, M. (2002). The effects of neighbourhood conditions on
perceptions of safety. Journal of Criminal Justice, 30, pp.417-427.
Bamberg, S. (2012). Understanding and promoting bicycle use – insights from psychological
research. Cycling and Sustainability, pp.219-246.
Bill, E., Rowe, D. and Ferguson, N. (2015). Does experience affect perceived risk of cycling
hazards? Scottish Transport Applications and Research (STAR) Conference. Glasgow,
UK. 20
th
May.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
21
Black, P. and Street, E. (2014). The power of perceptions: exploring the role of urban design
in cycling behaviours and healthy ageing. Transportation Research Procedia, 4, pp.68-
79.
Boschmann, E.E. and Cubben, E. (2014). Sketch maps and qualitative GIS: using
cartographies of individual spatial narratives in geographic research. The Professional
Geographer, 66(2), pp.236-248.
Brown, G. and Raymond, C. (2007). The relationship between place attachment and
landscape values: towards mapping place attachment. Applied Geography, 27, pp. 89–
111.
Carver, A., Timperio, A. and Hesketh, K. and Crawford, D. (2010).Are children and
adolescents less active if parents restrict their physical activity and active transport due
to perceived risk? Social Science & Medicine, 70, pp.1799-1805.
Caulfield, B., Brick, E., & McCarthy, O. T. (2012). Determining bicycle infrastructure
preferences – A case study of Dublin. Transportation Research Part D: Transport and
Environment, 17(5), 413-417.
Cho, G., Rodriguez, D.A. and Khattak, A.J. (2009).The role of the built environment in
explaining relationships between perceived and actual pedestrian and bicyclist safety.
Accident Analysis and Prevention, 41, pp.692-702.
Chow, S.L. (2002). Issues in Statistic Inference. History and Philosophy of Psychology
Bulletin, 14(1), pp.30-41.
Central Statistics Office (CSO) (2012).Results of Census 2011 – Profile 10 Door to door –
Commuting in Ireland. Cork, Ireland.
Davies, A, Fahy, F. and Rau, H. (2014).Challenging Consumption. In Davies, A.R., Fahy,
F. and Rau, H. (Eds.), Challenging Consumption: pathways to a more sustainable
future. London: Routledge, pp.3-19.
Deegan, B. (2015). Mapping everyday cycling in London. In: P. Cox (Ed.), Cycling Cultures,
Chester: University of Chester, pp.106-129.
Department of Transport, Tourism and Sport (DTTAS) (2009a). Smarter Travel. Dublin,
Ireland.
Department of Transport, Tourism and Sport (DTTAS) (2009b).National Cycle Policy
Framework. Dublin, Ireland.
Dill, J., Goddard, T., Monsere, C.M. and McNeil, N. (2015). Can protected bike lanes help
close the gender gap in cycling? Lessons from five cities. Annual Conference of the
Transportation Research Board (TRB). Washington DC, USA. 11-15
th
January.
Doorley, R., Pakrashi, V., Byrne, E., Comerford, S., Ghosh, B. and Groeger, J.A.
(2015).Analysis of heart rate variability amongst cyclists under perceived variations of
risk exposure. Transport Research Part F, 28, pp.40-54.
Elias, W. and Shiftan, Y. (2012).The influence of individual’s risk perception and attitudes on
travel behaviour. Transportation Research Part A, 46, pp.1241-1251.
European Road Safety Observatory (ERSO) (2012).Traffic safety basic facts 2012 – cyclists.
European Commission, Brussels, Belgium.
Fahy, F. and Ó Cinnéide, M. (2009).Re-Mapping the urban landscape: community mapping –
an attractive prospect for sustainability?Area,41, pp.167-175.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
22
Fernández-Heredia, Á., Jara-Díaz, S. and Monzón, A. (2014). Modelling bicycle use
intention: the role of perceptions. Transportation, published online.
Fishbein, M. and Ajzen, I. (1975). Belief, Attitude, Intention, and Behavior: An Introduction
to Theory and Research. Addison-Wesley, Reading, UK.
Galway City Council (2013).Galway traffic counts 2012: Manual classified junction counts.
Conducted by Abacus Transportation Surveys for Galway City Council.
Galway City Council (2012).Walking and cycling strategy. Galway, Ireland.
Garrard, J., Rose, G. and Lo, S. K. (2008). Promoting transportation cycling for women: the
role of bicycle infrastructure. Preventive Medicine, 46(1), pp.55-59.
Garrard, J., Handy, S, and Dill, J. (2012). Women and Cycling. In: J. Pucher and R. Buehler,
(Eds.) City Cycling, Cambridge, MA: MIT Press, pp. 211-234
Gehlert, T., Dziekan, K., and Garling, T.(2013).Psychology of sustainable travel behaviour.
Transportation Research Part A, 48, pp.19-24.
Google (2015).Google Maps. Available at: maps.google.com [Accessed 1
st
July 2015].
Gould, P. and White, R. (1993).Mental maps (3
rd
ed.). Boston; London: Allen & Unwin.
Gregory, D. (2009). The Dictionary of Human Geography Wiley-Blackwell, Oxford (5
th
Edition).
Harkey, D.L., Reinfurt, D.W. and Knuiman, M. (1998). Development of the Bicycle
Compatibility Index. Transportation Research Record, 1636, pp.13-20.
Heinen, E., Van Wee, B.andMaat, K. (2010). Commuting by bicycle: an overview of the
literature. Transport Reviews, 30(1), pp.59-96.
Heisserer, B. (2013). Curbing the Consumption of Distance? A practice-theoretical
investigation of an employer-based mobility management initiative to promote more
sustainable commuting. NUI, Galway: Unpublished PhD thesis.
Horton, D. (2007). Fear of cycling. Cycling and Society, pp.133-152.
Hunt, J.D. and Abraham, J.E. (2007). Influences on Bicycle Use. Transportation, 34, pp.453-
470.
Jacobsen, P.L. (2003). Safety in numbers: more walkers and bicyclists, safer walking and
bicycling. Injury Prevention, 9(3), pp.205-209.
Jensen, S.U., Rosenkilde, C. and Jensen, N. (2007). Road safety and perceived risk of cycle
facilities in Copenhagen. Presentation to AGM of European Cyclists Federation.
Johnston, R .J., Gregory, D., Pratt, G. and Watts, M. (eds.) (1986). The Dictionary of Human
Geography. Oxford (Blackwell) 2
nd
edition.
Kazig, R. and Popp, M. (2010).Unterwegs in fremden Umgebungen: Ein praxeologischer
Zugang zum „wayfinding“ von Fußgängern. Raumforschung und Raumordnung, 69(1),
3-15.
Lawson, A.R., Pakrashi, V., Ghosh, B. and Szeto, W.Y. (2013). Perception of safety of
cyclists in Dublin City. Accident Analysis and Prevention, 50, pp.499-511.
Lewis-Evans, B., & Charlton, S. G. (2006). Explicit and implicit processes in behavioural
adaptation to road width. Accident Analysis & Prevention, 38(3), pp.610-617.
Liang, K-Y.and Zeger, S.L. (1986). Longitudinal data analysis using generalized linear
models. Biometrika, 73(1), pp.13-22.
Loidl, M. (2014). How GIS can help to promote safe cycling. Proceedings of the
28
th
EnviroInfo Conference. Olderburg, Germany. 10-12
th
September.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
23
Lowry, M., Callister, D., Gresham, M. and Moore, B. (2012).Assessment of Communitywide
Bikeability with Bicycle Level of Service. Transportation Research Record, 2314,
pp.41-48.
Lydon, M. (2003).Community mapping: the recovery (and discovery) of our common
ground. Geomatica, 57, pp.131-44.
Lynch, K. (1960). The Image of the City. Cambridge MA: MIT Press.
Ma, L., Dill, J. and Mohr, C. (2014). The objective versus the perceived environment: what
matters for bicycling? Transportation, 41(6), pp.1135-1152.
Manton, R. and Clifford, E. (2012).A study of travel patterns to NUI Galway: lessons for
Smarter Travel in Universities. Irish Transport Research Network (ITRN) Conference.
University of Ulster, Jordanstown, UK. 29-30
th
August.
McCulloch, C. E.andNeuhaus, J. M. (2003). Generalized linear mixed models. John Wiley &
Sons, Ltd.
Met Éireann (2015).Monthly data for Athenry weather station. Irish Meteorological Service
Online. Available at: http://www.met.ie/climate/monthly-data.asp?Num=1875 (accessed
2
nd
July 2015).
Møller, M. and Hels, T. (2008).Cyclists’ perception of risk in roundabouts. Accident Analysis
and Prevention, 40, pp.1055–1062.
Mondshein, A., Blumenburg, E. and Taylor, B. (2013).Going mental: everyday travel and the
cognitive map. Access, 43, pp.2-7.
Mondshein, A., Blumenburg, E. and Taylor, B. (2010).Accessibility and cognition: the effect
of transport mode on spatial knowledge. Urban Studies, 47(4), pp.845-866.
Nelson, T.A., Denouden, T., Jestico, B., Laberee, K. and Winters, M. (2015). BikeMaps.org:
a global tool for collision and near miss mapping. Frontiers in Public Health, 3(53),
pp.1-8.
Noland, R.B. (1995). Perceived risk and modal choice: risk compensation in transportation
systems. Accident Analysis and Prevention, 27(4), pp.503-521.
Nye, M. and Hargreaves, T. (2009).Exploring the social dynamics of proenvironmental
behaviour change. Journal of Industrial Ecology, 14,pp.137-149.
Parkin, J., Wardman, M. and Page, M. (2007a).Models of perceived cycling risk and route
acceptability. Accident Analysis and Prevention, 39, pp.364-371.
Parkin, J., Ryley, T. and Jones, T. (2007b). Barriers to cycling: an exploration of quantitative
analyses. Cycling and Society, pp.67-82.
Parkin, J. and Coward, A. (2009).Comparison of methods of assessing cycle routes.
Proceedings of the ICE – Municipal Engineer, 162, pp.7-14.
Parkin, J. and Koorey, G. (2012) Network planning and infrastructure design. In: Parkin, J.,
ed. (2012) Cycling and Sustainability. Bingley: Emerald Books, pp. 131-160.
Prendergast, P. and Rybaczuk, K. (2005) Using visualisation techniques in planning to
improve collaborative governance in Ireland. Paper presented at CORP 2005, Vienna.
Pucher, J. and Dijkstra, L. (2000). Making walking and cycling safer: lessons from
Europe. Transportation Quarterly, 54(3), pp.25-50.
Raidió Teilifís Éireann (RTÉ) (2015). ‘The growing war between cyclists and motorists,
what’s happening on our streets?’.Broadcast26
th
May.
Published in Accident Analysis & Prevention 88 (2016) 138-149
doi:10.1016/j.aap.2015.12.017
24
Rengert G.F. and Pelfrey, W.V. (1997). Cognitive Mapping of the City Center: Comparative
Perceptions of Dangerous Places, In: David Weisburd and Tom McEwen, eds. (1997)
Crime Mapping and Crime Prevention, pp.193-217, Willow Tree Press, New York.
Road Safety Authority (RSA) (2014).Provisional Review of Road Crashes 2014.Ballina, Co.
Mayo, Ireland.
Sanders, R. (2015).Perceived traffic risk for cyclists: the impact of near miss and collision
experiences. Accident Analysis & Prevention, 75, pp. 26-34.
Sener, I. N., Eluru, N. and Bhat, C.R. (2009).An analysis of bicycle route choice preferences
in Texas, US. Transportation, 36(5), pp.511-539.
Short, J. and Caulfield, B. (2014).The safety challenge of increased cycling.Transport Policy,
33, pp.154-165.
Smith, T.M.F. (1983). ‘On the validity of inferences from non-random samples’.J. R. Statist.
Soc. A, 146(4), pp.394-403.
Snizek, B., Nielsen, T.A.S. and Skov-Petersen, H. (2013).Mapping bicyclists’ experiences in
Copenhagen. Journal of Transport Geography, 30, pp.227-233.
Soini, K. (2001). Exploring human dimensions of multifunctional landscapes through
mapping and map making. Landscape and Urban Planning,57, pp.225-239.
Spears, S., Houston, D. and Boarnet, M.G. (2013).Illuminating the unseen in transit use: a
framework for examining the effect of attitudes and perceptions on travel behaviour.
Transportation Research Part A, 58, pp.40-53.
Thogerson, J. (2006). Understanding repetitive travel mode choices in a stable context: a
panel study approach. Transportation Research Part A, 40, pp.621-638.
Timperio, A, Crawford, D., Telford, A. and Salmon, J. (2004).Perceptions about the local
neighbourhood and walking and cycling among children. Preventive Medicine, 38,
pp.39-47.
The Guardian (2013).Over 1,000 cyclists stage die-in protest outside Transport for London
HQ. Available at: http://www.theguardian.com/environment/bike-
blog/2013/dec/01/stop-killing-cyclists-die-in-tfl-protest [Accessed 1st July 2015].
UK Department for Transport (UK DfT) (2014).British Social Attitudes Survey 2013: Public
attitudes towards transport. London, UK.
Whannell, P., Whannell, R. and White R. (2012). Tertiary student attitudes to bicycle
commuting in a regional Australia university. International Journal of Sustainability in
Higher Education,13, pp.34-45.
Winters, M., Babul, S., Becker, H.J.E.H., Brubacher, J.R., Chipman, M., Cripton, P.,
Cusimano, M.D., Friedman, S.M., Harris, M.A., Hunte, G., Monro, M., Reynolds,
C.C.O., Shen, H. and Teschke, K. (2012). Safe cycling: how do risk perceptions compare
with observed risk?. Can J Public Health, 103(9), S42-S47.
Wooliscroft, B. and Ganglmair-Wooliscroft, A. (2014).Improving conditions for potential
New Zealand cyclists: an application of conjoint analysis. Transportation Research Part
A: Policy and Practice, 69, pp.11-19.
Zeile, P., Resch, B., Dörrzapf, L., Exner, J. P., Sagl, G., Summa, A. and Sudmanns, M.
(2015).Urban Emotions –tools of integrating people’s perception into urban planning.
Proceedings REAL CORP 2015 Tagungsband. Ghent, Belgium. 5-7
th
May.
