The influence of cycle lanes on road users’ perception of road space

ABSTRACT

Despite the many benefits of cycle lanes for active travel, their implementation remains a persistent problem. Road users who believe that building cycle lanes will take away their road space may object because they believe there is not enough space to do so. This study aims to address the visual perception of road space by exploring the relationship of the road users’ country of residence with how they perceive road space. Through an online survey distributed in three countries (n = 1591) with different levels of implemented cycle lanes this study demonstrates that the road space context significantly influences the visual perception of road space. Residents in the Netherlands, where cycle lanes are a common element of the road space, demonstrate 10% more recognition of having space to implement cycle lanes than the respondents in the UK and Australia, where cycle lanes are not as common. The implications of this research into the recognition by the public of having space (or not) to implement cycle lanes demonstrate the importance of context and provide evidence to policymakers to address a persistent problem.

KEYWORDS:

• Cycle lanes
• visual perception
• road space
• road user
• active travel

1. Introduction

The implementation of infrastructure for active travel has been a focus of urban research in recent years. Cycle lanes have provided a safe travel alternative during the COVID-19 outbreak (Teixeira et al., 2021) with pop-up cycle lanes implemented in many cities around the world (Buehler & Pucher, 2021). But despite the benefits of implementing cycle lanes (Nurse & Dunning, 2020), one of the major problems with their implementation is the pushback from road users claiming that road space is being taken. This argument has been difficult to overcome. One way to address this pushback has been to demonstrate the benefits of implementing cycle lanes for health (Aldred & Croft, 2019; Mulley et al., 2013; Saunders et al., 2013), economy (Blondiau et al., 2016; Rutter et al., 2013), and the environment (Kingham & Tranter, 2015; Mindell, 2015). However, although the benefits have been widely demonstrated, the pushback persists. This paper argues that visual perception plays a role in the perception that there is space to implement cycle lanes.

Visual perception is the mind’s ability to use vision to interpret and make sense of an object. This study tests Rock (1985) hypothesis that ‘past visual experience can affect perception in various ways’ and argues that ‘past visual experiences’ concerning road space could be considered in the context of a specific urban area.

The purpose of this study is to identify how respondents living in different contexts (defined by country of residence) perceive road space for cycle lanes. This study explores the relationship of country of residence and presence of cycle lanes with perception of road space using an online survey. The survey was distributed in three countries (the Netherlands, the United Kingdom and Australia) with different rates of implemented cycle lanes to explore the extent to which context (country of residence) influences the visual perception of road space. The survey showed images before cycle lanes were implemented and explored people’s perceptions regarding road space.

The paper is organised as follows, first, there is a literature review of studies looking at visual perception, road space, and relevant theories which explain how perception works, the role of road space (re)allocation in public policy and the implications for this study. This is followed by the methodology section and a detailed explanation of the steps taken in survey design and administration. The next section presents and discusses the results of the visual perception analysis. Finally, the limitations of the study and concluding remarks are presented.

2. Literature context

Visual perception is the most dominant way for people to derive information about the world (Swanston & Wade, 2013). Cognition sciences research shows that visual perception can be shaped by contextual factors (Niedenthal & Kitayama, 2013; Otten et al., 2017). In most of these studies, the contextual factors are social contextual features some of which might be feelings (Barrett & Bar, 2009), goals or expectations (Balcetis & Dunning, 2006), and even socio-economic status (Bruner & Goodman, 1947). This study is interested in exploring if physical contextual factors in urban elements, namely the road space infrastructure (which itself will be influenced by geometrical and functional criteria), can influence how road users perceive the available road space.

The first model considering perceptions and the diverse relationships that occur within the road space sphere was the ‘Ecology of the Street model’ (Appleyard et al., 1981). This model is relevant because it illustrates how perceptions may influence behaviours. Moreover, it points out that although the allocation of road space seems to be a technical issue treated with neutrality, the allocation of road space is also a social and political issue (Alcántara De Vasconcellos, 2004; Appleyard, 1980; Hickman & Banister, 2014).

The role of public policy is important to this study. Previous research highlighted the importance of perception–behaviour in the policymaking arena (Marsden & Anable, 2021). Moreover, Nicholson-Crotty and Meier (2005) show that perceptions can be translated into public policies since people tend to manifest their perceptions to influence public opinion (Edelman, 2013). Consequently, in situations where individuals have poor incentives to retain factual views (e.g. cycling) and strong directional motives to embrace beliefs that are congruent with a group identity (e.g. use of private car), damaging misperceptions emerge (Nyhan, 2020).

Although perceptions are important because they can influence policy decisions, perceptions are intuitive. Kahneman (2011) shows that in decision-making, two types of cognitive activities are used: System 1 (intuitive) and System 2 (analytical). System 1 thinking is frequently referred to as a reflex system that is ‘intuitive’ and ‘experiential’ or as ‘pattern recognition’, which initiates an automated way of thought (Tay et al., 2016). It is created without much conscious effort and channels the given data through a sub-conscious pattern recognition system based on similar past experiences (Hogarth, 2005). In contrast, System 2 is deliberate and analytical (Croskerry, 2009), and therefore not associated with perceptions.

From the literature considered here, and recognising that perceptions are intuitive (Kahneman, 2002), two points are evident and relevant to this study: 1) perceptions influence people’s behaviour, and 2) people tend to manifest their perceptions to influence public policies. The public policy in this study is the implementation of cycle lanes. The rationale for this study focuses on visual perception and road space in this way: if people perceive there is space for implementing cycle lanes, then their attitude (and behaviour) will be positive towards the implementation of cycle lanes. And vice versa, if a person perceives there is no space for cycle lanes, their attitude (and behaviour) will be negative towards the implementation of cycle lanes.

3. Methodology

Table 1 displays a variety of methods that can be used to investigate visual perception of urban infrastructure. The current study employs survey research to explore potential differences in visual perception among respondents living in different countries. The benefit of using this method lies in its capacity to acquire large samples, its cost-effectiveness, its potential to ensure consistency of data collection, and its ability to be remotely administered to participants in multiple countries.

Table 1. Research methods used to address visual perception or urban elements.

Method	Description	Advantages	Disadvantages	Examples
Observational	These involve researchers observing and recording participants’ responses to various road elements and urban infrastructure.	real-world setting, direct observation, large sample size, cost-effectiveness	lack of control, ethical concerns, observer bias, limited information.	(Tilt et al., 2007); Straatemeier et al. (2010); (Chamseddine & Ait Boubkr, 2020)
Experimental	These involve manipulating the road elements and urban infrastructure to investigate their impact on visual perception and measuring participants’ responses.	experimental control, establishment of causal relationships, precise measurement of effects.	artificial setting, demand characteristics, potential ethical concerns, limited generalizability.	(Fossum & Ryeng, 2022; Hull & O’holleran, 2014)
Survey research	These involve administering questionnaires or interviews to participants to obtain their perceptions and opinions about road elements and urban infrastructure.	ease of administration, cost-effectiveness, large sample size, access to participant perceptions and opinions.	potential for response bias, limited insight into underlying mechanisms, reliance on self-report measures, lack of experimental control.	(Perovic & Folic, 2012; Rosso et al., 2016)
Simulation	These involve using computer simulations or virtual reality technology to create realistic driving scenarios and investigate the impact of road elements and urban infrastructure on visual perception.	control over road elements and urban infrastructure, ability to create realistic scenarios, precise measurement of effects.	potential for artificiality, lack of ecological validity, potential for demand characteristics, expense of equipment.	(C. J. Wilson & Soranzo, 2015; Jonescu et al., 2020; Rybarczyk, 2014)
Eye-tracking studies	These involve using specialized equipment to track participants’ eye movements and measure their visual attention to road elements and urban infrastructure.	precise measurement of visual attention, direct measurement of eye movements, high temporal resolution, access to unconscious processes	potential for artificiality, expense of equipment, lack of ecological validity, limited insight into underlying mechanisms.	(Gholami et al., 2021; Lisińska-Kuśnierz & Krupa, 2020)
Neuroimaging	These involve using brain imaging techniques, such as fMRI or EEG, to investigate the neural processes underlying visual perception of road elements and urban infrastructure.	access to underlying neural processes, ability to investigate complex relationships, high spatial resolution.	expense of equipment, artificiality of tasks, limited ecological validity, potential for demand characteristics.	(Bolouki, 2022; Tang et al., 2017)

Australia, the Netherlands, and the United Kingdom, were chosen, based on their common attributes. The rationale for selecting these countries is threefold. Firstly, they all belong to the category of modernized developed nations (UN, 2021). Secondly, they have a shared Western-oriented culture (Henrich, 2020). Lastly, they exhibit a varied degree of incorporation of cycle lanes. The Netherlands, with its extensive network of connected cycle lanes, boasts the highest degree of implementation. On the other hand, the UK has launched several initiatives to introduce cycle lanes in recent times, and Australia, despite having greater investments in transport infrastructure, has fewer cycle lanes than both the Netherlands and the UK.

This study uses the same methodology as in Loyola et al. (2022) which considered the visual perception of speed limits and the implementation of cycle lanes and is summarized in the following sub-sections. The online survey was created in Qualtrics software and was distributed through Facebook. Figure 1 shows the flow diagram of the methodology, and the subsequent sub-sections elaborate on the reasoning that underpins the decision-making process in the survey design.

Figure 1. Flow diagram of the methodology.

3.1. Survey design

The online survey was designed to answer the research question of the study, namely, to explore how the visual perception of road space (dependent variable) changes according to the context of the respondents (independent/explanatory variable). The survey consisted of two sections with questions about visual perception and socio-demographic characteristics. Table 2 describes the survey sections.

Table 2. Survey design outline.

SECTION 1 Visual perception	1) Visual perception of road space:
	- Set a) ‘Without taking out one lane (either parking or traffic lane). Is there space for a cycle lane?’ (Images: 1 fixed + 3 random of a batch of 6)
	- Set b) ‘Taking out one lane (either parking or traffic lane). Is there space for a cycle lane?’ (Images: 1 fixed + 3 random of a batch of 5)
	2) Visual perception of road elements:
	- Two images (road elements: footpath, cycle lane, lane, car)
SECTION 2 Socio-demographic characteristics	Country of residence
	Age
	Gender
	Travel mode of commute (or most usual trip)
	Education
	Occupation
	Income
	Access to car
	Driver license

3.1.1. Visual perception section

This section of the survey presented street images and asked about the visual perception of road space for each image. All the images were taken from Google Maps (2021) and selection criteria for the images to be included were devised. The full set of images are presented in Appendix 1.

3.1.1.1. Selection criteria for the street images

1. Images from different countries. The reason for this was to avoid bias from the respondents who might be familiar with the street images.

2. Balanced street design/geometry. The selected images were a mix of urban and rural streets, some streets were wide while others were narrower. The purpose was to avoid fatigue from the respondents.

3. Representation of people and vegetation in the images. As a control measure, some images contained people and others didn’t. The purpose of this was to avoid the influence of the presence of people and vegetation on the respondents’ perception of road space.

4. Streets where a cycle lane is currently implemented. This study considered only images with access to a past (former) image of the street (i.e. without the cycle lanes). The image without the cycle lane was displayed, and the respondents were asked if there is space to implement a cycle lane. The purpose was to control for the correct responses. In other words, the analysis will demonstrate the difference between the respondents and the benchmark will be the correct answers (respondents who say ‘yes’ to the question if there is space to implement cycle lanes).

3.1.1.2. Visual perception of road space

Figure 3 is an example of how the questions were presented to the respondents. The chosen images were first displayed in two sets (sets a and b in Table 2). Each set displayed four images that were part of a batch of images. For the first set (set a), the survey displayed 1 fixed image+3 random images from a batch of six images. For the second set (set b), the survey displayed 1 fixed image+3 random images from a batch of five images. The purpose of using randomized images from a batch of images is to avoid respondent bias. In other words, to minimize the variability of the evaluation and avoid misleading findings (Antony, 2014). The reason for having one fixed image per set is to have an image that is presented to all the respondents (per set) that will be used for the regression analysis to determine the extent of visual perception in the recognition of road space to implement cycle lanes (Section 4).

The first set (set a) referred to respondents’ perception of visual perception without taking out a traffic lane. The survey questions were ‘Without taking out one lane (either parking or street/car lane). Is there space for a cycle lane?’ The second set (set b) considered the same questions but taking out a traffic lane. The questions for the second set were: ‘Taking out one lane (either parking or street/car lane). Is there space for a cycle lane?’. The reason for making these two sets was to test how respondents change their answers when they consider taking out a traffic lane for a cycle lane implementation.

The respondents’ options to answer the question regarding road space measurements and speed limits were different. The road space questions were single-answer questions (see Figure 2). The options were: a) Definitely NOT, b) Most likely NOT, c) Definitely YES, b) Most likely YES. The absence of a neutral option forced respondents to decide.

Figure 2. Example of the question presented in the online survey. (Image source: Google Maps).

The answer to the question if there is space (or not) was known in advance by the research team (selection criteria – d). All the images were older images from locations where currently there is a cycle lane implemented. Thus, the focus/analysis is on who correctly answered the question.

3.1.1.3. Visual perception of road elements

In addition to the two sets of images addressing the visual perception of space to implement cycle lanes, this study explored whether perception differs between countries on other road elements (not only related to space to implement cycle lanes). The purpose of this question is to confirm if the visual perception of space differs for road elements other than the road space. To achieve this, two street images were displayed and asked the respondents to write the measures (in metres) for a series of road elements from the images. The road elements were a traffic lane, cycle lane, footpath, and car. The elements were marked on the images to avoid confusion or misunderstanding of the exact location/element the study is referring to. Moreover, adding this question to the survey provides a means of verification: if extreme values were found in one of the road elements, all of a respondent’s responses were dismissed since it meant the respondent was not being thoughtful.

3.1.2. Socio-demographic section

The purpose of the socio-demographic section of the survey was the identification of patterns in the data that could be relevant to the visual perception revealed in the previous section. This section included questions regarding country of residence, age, gender, travel mode (commute), education, occupation, income, access to a car, and driver’s license.

3.2. Survey distribution

To collect the data, this study used an online survey distributed using social networks. Facebook ‘Ads’ was used as a channel to distribute the survey to the targeted countries. The online survey was translated into Dutch for its distribution in the Netherlands. One of the benefits of using Facebook Ads is that it shows the number of respondents taking the online survey (clicking on the Ad). This is relevant because it allows the survey designer to increase (or decrease) the budget for the Ads according to the desired number of respondents per country. The time to complete the survey was anticipated to be around 8–10 minutes (the average actual duration of the online survey was 10.8 minutes). Before starting the survey, the respondents had to provide their consent (as required by institutional ethics approval) and were provided with information about the purpose of the anonymized survey.

4. Analysis

The present study utilises a median estimate and logistic regression models to analyse the recognition of road space and its relationship with the socio-demographic characteristics of respondents.

4.1. Sample description

The total number of respondents to the online survey was 1591. The number of people who took the survey was 3829 before cleaning the data and discarding unfinished surveys and outliers. Table 3 shows the sample description with the considered predictors (or explanatory variables). Most of the respondents were in the UK with 801 (50.3%), followed by Australia with 397 (25%), and 393 (24.7%) in the Netherlands. The gender distributions of the total sample were 59.3% female, and 38.7% male; the responses from non-binary, and respondents who ‘prefer not to say’ their gender were omitted because they were not a representative sample for each country and therefore would not offer reasonable estimates. Figure 3 shows the age distribution of the sample. Most of the respondents were adults around 50–70 years old, except in Australia where 33.5% of respondents were younger adults between 20–50 years old. The reason for the older age respondents’ rate might be that there are more adults than younger users willing to take a survey on Facebook. Also, respondents’ age is associated with their occupation. Most respondents were Retired (39.1%), Working full-time (27.2%), or Working part-time (14.7%), followed by Other occupations (9.8%), Self-employed (7.4%) and students (1.8%). The survey also explored the respondents’ access to a car. Here, the respondents from Australia led with 97% stating they have access to a car, compared to the UK (92.4%) and the Netherlands (87.3%). The online survey also included questions regarding education. Because different countries have different classifications of studies and degrees, education was grouped into three groups, people who studied up to high school, people above high school but below advanced degree, and people with an advanced degree (PhD or professional degree such as JD, MD). Income was also considered and converted to US Dollars to include the values (as continuous) in the regression models (see section 4.2.3). Finally, the parameter risk aversion was categorized into no risk, for respondents who did not play the lottery at the end of the survey; risk-avoiding, for respondents who choose option 1 out of 10 gift cards valued in AUD 10; risk-neutral, who choose 1 out of 4 gift cards value in AUD 25; and risk-seeking, who choose 1 gift card value in AUD 100. The rationale behind this classification was explained in the survey since the number of participants for each lottery will depend on the number of participants who choose that lottery (Benartzi & Thaler, 2011). For the data analysis, all responses were calculated as a percentage of responses from their country and grouped in the categories displayed in Table 3. Grouping the responses ensured a significant number of responses for each category (parameter) per country and thus reasonable stable estimates avoiding overfitting (Babyak, 2004; Peduzzi et al., 1995).

Figure 3. Age distribution by country.

Table 3. Sample description (source: Loyola et al. (2022)).

Characteristics	Categories	Country (per cent per country)			Total (n)	% of total
		Australia	The Netherlands	The UK
Gender	Female	56	56	63	943	59
	Male	42	41	36	615	39
	Non-binary	2	1	1	16	1
	Prefer not to say	1	2	1	17	1
Age	(continuous)				1591	100
Travel mode	Private vehicle	79	58	75	1138	72
	Bicycle	6	26	3	147	9
	Walk	5	5	11	130	8
	Bus	2	1	5	50	3
	Train	4	2	2	40	3
	Other	3	4	3	50	3
	Bus + train or Train + bus	3	2	1	26	2
	Ferry (+ bus, train, or other)	0	0	0	2	0
	Micromobility (e-scooter, e-bike, etc.)	0	1	0	5	0
	Taxi (Uber, OLA, Didi, etc)	0	0	0	3	0
Education	Degree	66	55	59	950	60
	High school	26	36	31	498	31
	Advanced degree	8	9	10	143	9
Occupation	Retired	41	21	47	622	39
	Work full-time	29	31	25	433	27
	Work part-time	16	15	14	234	15
	Other	4	21	7	156	10
	Self-employed	7	10	7	117	7
	Student	4	2	1	29	2
Income	(continuous)				1591	100
Access to car	Yes	97	87	92	1468	92
	No	3	13		123	8
Driver license	Yes	98	91	94	1496	94
	No	2	9	6	95	6
Lottery risk aversion	No risk	51	69	72	1051	66
	Risk-seeking	15	18	16	255	16
	Risk-neutral	28	8	7	196	12
	Risk-avoiding	7	5	5	89	6

4.2. Visual perception of road space

To explore the visual perception of road space, the average responses and a binary logistic regression were analysed. This study analysed the combined sample (three countries) and per country. To build the model the study first conducted univariate logistic models to determine the statistical significance (or confounding) of all predictors with a p-value<0.25. Secondly, a binomial logistic regression with the selected variables from the univariate logistic regression (p-value<0.25) was conducted. Figure 4 shows the approach to determine the extent of visual perception in the recognition of road space to implement cycle lanes.

Figure 4. Diagram flow for the analysis of the visual perception of road space (sub-sections in brackets).

The analysis dichotomized (grouped) responses from the single-choice question options into affirmative and negative answers for one of the images presented to all the respondents on the online survey. If respondents selected ‘Definitely NOT’ or ‘Most likely NOT’ those responses were grouped into negative (or NO), and if respondents selected ‘Definitely YES’ or ‘Most likely YES’ into affirmative (or YES). The dependent variable which measures the recognition of having road space is the grouped response (YES). Since the linear probability model is heteroskedastic and may predict probability values beyond the (0,1 or NO, YES) range, the logistic regression model is used to estimate the factors/predictors which influence the recognition of having road space.

4.2.1. Average responses

Residents in the Netherlands show more recognition (i.e. were more likely to answer ‘yes’) of having road space to implement cycle lanes. The results in Figure 5 show this recognition between countries for the images from set a (without taking out a traffic lane) and the images from set b (taking out a traffic lane). The respondents in the Netherlands (in both sets) show a higher recognition than the respondents in Australia and the UK. For set a, the respondents in the Netherlands show on average 59.4% of recognition having space to implement cycle lanes, while respondents in Australia and the United Kingdom averaged 44.5% and 45% respectively. The country difference is significant because respondents in the Netherlands show around 14% more recognition than respondents in the UK and Australia. For set b, respondents in the Netherlands answer affirmatively 88.5% while Australia and the UK averaged 77.5% and 79.1% respectively. The Netherlands’ responses were higher at 11.1% and 9.4% than respondents in Australia and the UK respectively. The difference between set a and set b (the high values in set b) is explained by the premise of the question in set b. In other words, if a traffic lane is taken out to implement a cycle lane it follows that more respondents will think there will be space. Therefore, the subsequent analysis refers to the set a (without taking out a traffic lane to implement a cycle lane).

Figure 5. Average recognition of having space for cycle lanes.

4.2.2. Combined responses

4.2.2.1. Univariate logistic regression

Table 4 shows the results from the univariate logistic regression models for each parameter. The dependent variable for these models was the recognition of having space to implement cycle lanes (affirmative/YES). The parameters that show a p-value<0.25 (in bold) were country of residence, age, gender, travel mode, education, occupation, and income. The parameter with higher significance was country of residence (<0.001), followed by age (0.005) and education (0.007). Interestingly, travel mode was also significant, and looking at the results closely it made sense that the estimate was positive only for the respondents who cycle to work (although the results per individual category were not significant).

Table 4. Univariate logistic regression of visual perception of road space.

Parameter [p-value]		Estimate	SE	OR	OR 95% CI
Intercept		0.01	0.07	1.01	0.88	1.16
Country [0.001]	UK			1.00
	Australia	0.08	0.12	1.08	0.85	1.38
	Netherlands	0.56	0.13	1.75	1.37	2.25
Intercept		0.88	0.26	2.41	1.45	4.01
Age [0.005]	(continuous)	−0.01	0.00	0.99	0.98	1.00
(Intercept)		0.30	0.08	1.35	1.15	1.58
Gender [0.048]	Male			1.00
	Female	−0.21	0.10	0.81	0.66	1.00
Intercept		0.52	0.22	1.69	1.09	2.61
Travel mode [0.019]	Other			1.00
	Private car	−0.39	0.23	0.68	0.43	1.07
	Bus	−0.85	0.36	0.43	0.21	0.87
	Bicycle	0.05	0.28	1.05	0.61	1.83
	Train	−0.42	0.39	0.65	0.31	1.40
	Walk	−0.62	0.28	0.54	0.31	0.94
Intercept		0.24	0.17	1.27	0.91	1.77
education [0.007]	Advanced degree			1.00
	High school	−0.31	0.19	0.73	0.50	1.06
	Degree	0.04	0.18	1.04	0.73	1.48
Intercept		0.38	0.10	1.46	1.21	1.77
Occupation [0.114]	Work full-time			1.00
	Work part-time	−0.16	0.16	0.86	0.62	1.18
	Student	−0.17	0.39	0.84	0.40	1.80
	Self-employed	−0.22	0.21	0.80	0.53	1.21
	Retired	−0.34	0.13	0.71	0.56	0.91
	Other	−0.40	0.19	0.67	0.46	0.96
Intercept		−0.03	0.10	0.97	0.81	1.17
Income [0.011]	(continuous)	0.00	0.00	1.00	1.00	1.00
Intercept		0.17	0.05	1.19	1.07	1.32
Access to car [0.398]	Yes			1.00
	No	−0.16	0.19	0.85	0.59	1.23
Intercept		0.17	0.05	1.18	1.07	1.31
Driver license [0.774]	Yes			1.00
	No	−0.06	0.21	0.94	0.62	1.43
Intercept		0.16	0.06	1.18	1.04	1.33
Risk aversion [0.970]	No risk			1.00
	Risk seeking	0.02	0.14	1.02	0.77	1.34
	Risk neutral	0.02	0.16	1.02	0.75	1.39
	Risk avoiding	−0.09	0.22	0.91	0.59	1.40

4.2.2.2. Binomial logistic regression

Table 5 shows the results of the binary logistic regression after taking out the predictors that were not significant. The dependent variable was the recognition of having space to implement cycle lanes (affirmative/YES). The predictors that were not significant (p-value>0.05) despite having a p-value<0.25 in the univariate models (Table 4) were travel mode, occupation, and income. The statistically significant parameters were country of residence, gender, education, and age. The Hosmer-Lemeshow goodness of fit was not significant (0.862) indicating the model is correctly specified. Additionally, the −2log Likelihood = 2106.761 and the Nagelkerke R squared = 0.035. The independent variables country of residence: the Netherlands (B = 0.51, SE = 0.15), age (B = −0.01, SE = 0), gender: male (B = 0.24, SE = 0.11), and education: ‘degree’ (B = 0.39, SE = 0.11) were found to be significant (p < 0.05). The estimated odds ratio (OR) for the country of residence ‘the Netherlands’ is 1.66 (95% CI = 1.24–2.23) which means that residents in the Netherlands are 1.66 times more likely to believe there is space to implement cycle lanes than residents in Australia. The estimated OR for age is 1.0 suggesting no difference or little difference in age (although significant). The estimated OR for the country of gender ‘male’ is 1.27 (95% CI = 1.03–1.57) indicating that males are 1.27 times more likely to believe there is space to implement cycle lanes than females. The estimated OR for education: degree is 1.48 (95% CI = 1.18–1.85) suggesting that people who have a degree are 1.48 times more likely to believe there is space to implement cycle lanes than people with only high school education.

Table 5. Results: binary logistic regression.

Parameter [p-value]		B	S.E.	Sig.	OR	OR 95% C.I.
Country [< 0.001]	Australia				1.00
	Netherlands	0.51	0.15	0.001	1.66	1.24	2.23
	UK	0.00	0.13	0.991	1.00	0.78	1.28
Age [0.018]	(continuous)	−0.01	0.00	0.018	0.99	0.98	1.00
Gender [0.025]	Female				1.00
	Male	0.24	0.11	0.025	1.27	1.03	1.57
Education [0.003]	High School				1.00
	Degree	0.39	0.11	0.001	1.48	1.18	1.85
	Advanced degree	0.32	0.20	0.106	1.37	0.93	2.02
	Constant	0.33	0.30	0.268	1.39

4.2.3. Variables within a country

The result of the analysis within each country shows that for every country the potential predictors are different. For Australia the predictors showing significance (p-value<0.25) were age, gender, education, and income; while for the respondents in the Netherlands they were gender, travel mode, education, access to a car, driver’s license, and risk aversion; and for the respondents in the UK, only age was significant. Nonetheless after conducting a binary logistic regression for each country with the above-mentioned predictors results show no significance (p > 0.05) in any of the three models.

4.3. Other road elements

Table 6 shows no significant variability in the responses to estimated measures (in metres) of road elements among respondents from the three countries. These results are relevant because they confirm that the visual perception of space for cycle lanes is not the same as the visual perception in general. The results in Table 6 show that there are no significant differences when testing the visual perception of other road elements. Therefore, these results confirm only that the visual perception of road space for implementing cycle lanes shows a significant difference between countries in this study. Although most of the reported measurements are slightly higher for the respondents in the UK, this is not the case in all the road space elements. Also, it should be noted that not all selected countries used the metric system of measurement. The UK has largely adopted the metric system but there are some exceptions, for instance, road speed limit signs don’t generally use the metric system (RAC, 2021). The survey specified that respondents answer these questions using metres, but for someone who is not used to the metric system this may be difficult. The images presented to the respondents are in Appendix 2.

Table 6. Average measures (metres) of road elements from the presented images.

Road element	The United Kingdom		Australia		The Netherlands
	Average measure (metres)	Standard deviation	Average measure (metres)	Standard deviation	Average measure (metres)	Standard deviation
Lane (photo 1)	2.14	0.60	2.00	0.60	2.08	0.81
Car (photo 1)	1.72	0.56	1.56	0.40	1.74	0.68
Cycle lane (photo 1)	1.08	0.41	0.99	0.38	0.98	0.39
Footpath (photo 1)	1.37	0.62	1.30	0.55	1.15	0.66
Lane (photo 2)	1.87	0.35	1.80	0.45	1.78	0.45
Car (photo 2)	1.63	0.56	1.49	0.41	1.69	0.66
Footpath (photo 2)	1.34	0.46	1.28	0.49	1.32	0.56

5. Limitations

Although the analysis has shown significant statistical differences it is important to acknowledge that there are some limitations of this study. An important limitation is the interpretation of context and its relationship with geometrical constraints. This study acknowledges that geometrical constraints can affect the visual perceptions of users and we have tried to offset this limitation presenting a randomized sample of images to all respondents (Antony, 2014). It is also important to recognise the implication of culture. The results from this study show that a different context is related to respondents recognizing that there is space to implement cycle lanes, but this study cannot refute that the recognition of having space to implement cycle lanes might be because the respondents in the Netherlands are from a culture more pro-cycling (which could include the implementation of cycle lanes in more favourable conditions) than respondents in Australia or the UK. Therefore, investigating the narratives of how cycle lanes were (and are) implemented in these three countries is an interesting area for future research that could shed some light on this matter.

Methodological limitations are also to be noted. Since the online survey was anonymous, it is impossible to determine the contextual conditions of the respondent when doing the survey. It is well known that self-selection bias is a problem in online surveys given that in any online community there are going to be respondents who are more eager to answer online surveys than others (Wright, 2005). To address the ‘less-than-thoughtful’ survey takers, all the responses were deleted if any answer was not coherent (see section 3.1.1.3). Another limitation might be that 60% of the respondents were aged between 50–70 years old. Following the institutional ethics protocol, respondents younger than 18 years old were not considered. The study counterbalances the respondent’s age with a representative sample size for the three countries (1591 valid respondents); and when analysing the data, models were run using age as a numerical variable and as a categorical variable with different age bands; in all cases, age was not a significant predictor for any of the models.

Another point to be highlighted is that different types of cycle lanes were not the focus of this study because people’s perceptions of road space are intuitive (is there space … or not?) and do not follow a deliberate analysis (e.g. thinking about the different geometries of different cycle lane types before answering a question about space). Moreover, previous research demonstrates that perceptions are intuitive and do not occur in deliberate operations of reasoning (Chudnoff, 2013; Kahneman, 2002; Tversky & Kahneman, 1983). It is recognised that human perception is largely influenced by a wide range of factors, many of which would be related to the characteristics of the person and other factors such as personal experience. This paper explores the difference in perception between residents of different countries; further research could be undertaken to explore the factors that are related to individual differences in perception.

A final limitation is that the study considers the respondents of the country as the whole group and does not differentiate between cities (who might have different degrees of cycle lanes implemented), respondents in an urban and rural environment, or any other than the variables mentioned in the survey. Nevertheless, this might be an interesting future research topic.

6. Conclusions

The (re)allocation of road space to implement cycle lanes for active travel is a contested issue that has been generally addressed in terms of justice and fairness. One of the major obstacles to implementing cycle lanes is the pushback from road users who perceive that the road space is not wide enough to implement cycle lanes. The results from this study suggest that the way people (in the selected countries) perceive road space is contingent on the context of where they live and the built environment that they find around them.

This research analysed the results of an online survey (n = 1591) distributed in three countries with different levels of cycle lanes implemented. The average recognition of having space to implement cycle lanes was compared and logistic regression models were employed to find patterns explaining the difference in recognition of having space between respondents in different countries. The results illustrate that respondents living in the Netherlands (a country where cycle lanes are a common element of the road space) show 10% more recognition of having space to implement cycle lanes than the respondents in the UK and Australia (countries where cycle lanes are not as common). Compared to respondents in the UK, respondents in the Netherlands are 1.88 times more likely to perceive there is space to implement cycle lanes in streets without taking out a traffic lane; and 1.74 times more likely to perceive there is space to implement cycle lanes by taking out a traffic lane. Respondents in Australia did not differ significantly from respondents in the UK. In addition, this study tested the visual perception of different road elements between the selected countries. The results here did not show significant variability and this is relevant because it shows that the visual perception of space for cycle lanes is not the same as the visual perception of other road elements. Moreover, testing if respondents over- or underestimate the width of road elements is a potential area of future research.

The implications of these results are meaningful from a range of perspectives. From a psychological perspective, the results confirm and extend the findings from previous research stating that different forms of the built environment can cause different psychological effects (Bell et al., 2001; Kang et al., 2020; Kaptsevich, 2021). This is related to a social perspective, explaining the resistance from groups of people where cycle lanes are not a regular element of the road space, suggesting that people who oppose cycle lanes might perceive that there is no space (when there is). It may also be people choose to report strategically that there is no space, despite perceiving space, because of a previous disposition against cycle lanes.

From a policy perspective, these findings provide evidence for policymakers to deal with hostility to cycle lanes (A. Wilson & Mitra, 2020; Aldred et al., 2019; Aldred, 2012). For example, sharing these results with policymakers might encourage them to support the implementation of sustainable transport infrastructure under the assumption that antagonism to implement cycle lanes may be the result of a lack of familiarity with cycle lanes.

The findings from this study have important policy implications for policymakers in the Netherlands, Australia, and the UK. Policymakers in the Netherlands can use the results to reinforce their existing initiatives towards prioritizing sustainable transport infrastructure and promoting active travel. The findings can guide policymakers in Australia and the UK to develop a policy framework that prioritizes sustainable transport infrastructure, which can be used to guide decision-making and on-going maintenance of policies related to sustainable transport infrastructure. Policymakers can use the findings of the study to address resistance and misperceptions towards cycle lanes and promote active travel. Sharing the results with policymakers can increase their support for the implementation of sustainable transport infrastructure and the prioritization of the needs of active travellers. This approach can be instrumental in enhancing the public’s willingness to accept sustainable transport infrastructure, and addressing misperceptions related to cycle lanes, thus promoting the adoption of active travel.

Another way to understand the implications of this study is given by what is known as the implications of misperceptions. Because previous research shows that misperceptions might become public beliefs that are polarizing (Flynn et al., 2017; Gerber & Green, 1999; Koehler, 1993). In this case, this belief could be represented by a strong position against cycle lanes even though there is space to implement them without taking space from car lanes. Although previous research shows that for some people this belief will be maintained even if incentives are provided (Peterson & Iyengar, 2021). Any evidence-based response to the problem of misperceptions must begin with an effort to address the problem itself (Nyhan, 2020; Porter & Wood, 2019).

Acknowledgments

The authors acknowledge the technical assistance of Kathrin Schemann of the Sydney Informatics Hub, a Core Research Facility of the University of Sydney. We wish to thank the University of Sydney Business School via the “Engaged Research Accelerator” and the Institute of Transport and Logistics Studies (ITLS) for the financial and academic support in the development of this study.

Disclosure statement

No potential conflict of interest was reported by the authors.

APPENDICES Appendix 1

Figure A1. Images of the streets presented to the respondents.

Appendix 2

Figure A2. Images 1 and 2 (descending order) presented to respondents to estimate other road elements (section 4.3).