Strengths and limitations
Our study has a number of strengths and limitations. Strengths include the separate analyses of the explanatory variables associated with walking and cycling, the large sample size of predominantly healthy working adults commuting from both urban and rural areas, and the use of reliable measures to assess perceptions of the environment and relevant psychological constructs, albeit applied in this study to a particular environmental context (the route to work) and a particular type of travel behaviour (car use) respectively. We also used time-based measures of travel to and from work which are more detailed than the more commonly used measures of 'usual' or 'main' mode. On the other hand, we cannot assume that walking or cycling were undertaken as sole modes of travel because the use of combinations of travel modes within one journey is relatively common in Cambridge . In this sample, 28% of participants reported using a combination of travel modes on their journey to or from work at least once in the last seven days, for example by using public transport in combination with walking, or by driving part of the way and cycling the remainder.
Data were collected over a six month period between May to November to reduce the confounding effect of seasonal variations in travel behaviour, and as a result the reported levels of walking and cycling to work in this study may be higher than average in this setting. Participants reported the characteristics and conditions on or along their route to work; however, data were not available on the perceived supportiveness of the neighbourhood environment, which may also be important. Rather than asking participants to complete a battery of psychological measures relating to each of several potential modes of transport, we chose to focus on psychological measures regarding car use because the longitudinal study aims to assess whether changes to the environment designed to promote a shift away from car use are associated with changes in travel behaviour. In particular, we intend to assess the mechanisms underlying such changes and whether these operate via changes in, for example, attitudes towards car use. However, one consequence of this longitudinal focus is that these measures may not be optimally matched to all the possible behavioural summary measures considered in baseline analysis. Another limitation is that our sample contains a higher proportion of participants educated to degree level and a smaller proportion of obese adults than the general population of Cambridgeshire , no doubt reflecting the focus of this particular study on the predominantly healthy working adult population.
Cambridge is a city known for its cycling culture and its relatively high prevalence of cycling . Nonetheless, comparison with local travel-to-work survey data suggests that cyclists may have been over-represented in our sample (47% versus 21% ). However, the local survey used a much cruder measure of travel behaviour and sampled participants from workplaces across the county, whereas our sample was drawn from workplaces in the city. Since higher levels of cycling tend to be observed in urban areas than rural areas and nearly half of our participants both lived and worked in Cambridge, these differences in sampling and data collection may account for the differences in the apparent prevalence of cycling, which would limit any concerns regarding potential selection bias. Whilst these relatively high levels of cycling observed allowed us to explore the explanatory variables associated with walking and cycling separately, the relative importance of the environmental and psychological factors may not be mirrored in other contexts where cycling is less popular or less embedded in social practices, and therefore the generalisability of these findings to these contexts may be limited. On the other hand, more than 85% of our participants also live in households with access to a car, and understanding the reasons why people choose, or do not choose, to walk or cycle despite having access to a car may be important for the development of strategies to promote active travel more generally.
Relative contribution of explanatory variables
Individual and household characteristics accounted for a larger proportion of the variation in cycling behaviour than the other putative explanatory variables. This is generally similar to findings reported elsewhere [13, 23, 40], however we also found that environmental perceptions were associated with walking behaviour. In this study, many of these individual characteristics, such as educational level and BMI, were associated with cycling but not with walking. Positive associations between educational status and cycling behaviour have been reported previously  and given that ours was a relatively healthy and well educated sample, it is possible that these associations may be stronger in the population at large. One possible mechanism for the observed association is that owing to the high cost of housing in Cambridge city and its immediate surroundings, those who are well educated and have higher levels of disposable income may be more likely to be able to afford to live closer to work and therefore more likely to have the option of cycling. Due to the relatively small numbers of participants who reported walking and who had lower levels of education in this sample, the analysis may have been underpowered to detect comparable associations for walking.
Consistent with a systematic review of the environmental determinants of physical activity - which found that three quarters of all associations tested returned evidence consistent with the null hypothesis - in this analysis, few of the environmental perceptions remained significant in final models. Of these, perceptions that it was pleasant to walk and that convenient public transport was available on the route to work were associated with walking, and the perceived convenience of cycle routes was associated with cycling. It is not possible to be sure from these results whether convenient routes facilitate cycling, or if cyclists are simply more aware than non-cyclists of the presence of convenient routes. In the context of this study, the provision of facilities which improve the convenience, quality and pleasantness of walking and cycling, such as traffic free routes, may be an important component of a broader intervention to promote active commuting, particularly if improvements are focused on routes which are frequently used, connect home and work locations within an acceptable cycling distance, and intersect with public transport stops.
We also found that participants who reported little traffic were less likely to spend time walking to work. This may reflect greater awareness and reporting of traffic in walkers compared with non-walkers, as suggested by Giles-Corti and Donovan  and Titze et al.  who also reported similarly unintuitive findings. Alternatively, it may represent a genuine association whereby walking to work is more prevalent in built up areas which have higher traffic levels. Further analyses in this study will draw on secondary data sources and GIS to explore the associations between travel behaviour and more objective assessments of the environment on and along the route to work.
In contrast to much of the literature which reports positive associations between psychological measures and walking or cycling behaviours [14, 40], we found that few of the psychological measures related to car use were negatively associated with walking and cycling. Favourable attitudes towards car use were positively associated with walking. The reason for this finding is not apparent, although it is possible that these measures captured the potential enjoyment of using the car, rather than the practical feasibility of doing so. For example, it may be possible for a participant to respond positively to an item assessing attitude towards car use, but also to respond negatively to an item assessing perceived behavioural control if it is not practical for that person to use a car. Alternatively, it may be that participants who reported walking and reported positive attitudes towards car use did so because they travel to work using a combination of driving and walking. Again, the reasons are not immediately apparent but it may be that enjoyment of cycling and car use co-exist. This hypothesis appears feasible given the local context of the study, in which the locations of park-and-ride sites facilitate the use of combinations of travel modes and in which cycling is socially patterned in such a way that relatively affluent individuals can afford to own a car but also to live close enough to work to cycle. A mixed method exploration of how and why people commute by car in this sample will be the subject of a further paper.
Previous research has also highlighted the distance to work as an important explanatory variable for both walking and cycling behaviours. Interestingly, in this study the effects of distance on walking in particular were much stronger in the subset without access to a car. It may be that those participants who have no access to a car walk all the way to work, whereas those who do have access to a car may walk as part of a longer journey, for example by using off-site car parks or park-and-ride sites which encourage walking for short distances. This may explain why distance between home and work is more strongly associated with travel behaviour in the subgroup without access to a car than in those with access to a car. However, given the relatively small percentage of participants who did not have access to a car, the analysis may have been underpowered to detect an association in this group, regardless of the explanatory variable under consideration. This opportunity to combine travel modes is relatively common in older cities in the UK, such as Cambridge, Oxford and York, where the geography and historical development of the area means that the availability of parking in city centres is often limited. Analyses exploring the characteristics of those who walk or cycle in combination with other travel modes as part of a longer journey will be subject of a further paper.
In summary, we found that individual and household characteristics, as well as perceptions of the route environment and psychological measures related to car use, were associated with walking and cycling to and from work; however the contribution of these explanatory variables to behaviour was relatively small and the individual and household characteristics explained more of the total variance in cycling than the other variables. This raises two important points. First, it identifies a need for greater consideration of the range of factors which influence behaviour and the interactions between them. It is likely that behavioural choices, particularly about travel to and from work, are made in the context of wider consideration of the needs and requirements of other people , especially other members of the household but also those from work or other social settings. Researchers should therefore consider moving beyond asking individuals to reflect on their own views and perceptions towards attempts to ascertain the social and physical contexts of travel decision-making, for example by using a combination of quantitative and qualitative methods to gain a greater understanding of the role of household, neighbourhood and workplace social contexts in shaping behaviour. Second, although perceptions of the route and psychological measures related to car use made a relatively modest contribution to the models, these factors do appear to have a role in explaining patterns of walking and cycling to work, especially in those with access to a car. Understanding the reasons for these associations in particular subgroups and how changes in perceptions of the environment or psychological orientation towards car use could be brought about is important, but it is likely that both psychological and environmental influences on travel behaviour will need to be tackled in order to bring about sustained behaviour change in the population as a whole .