This study was conducted to expand the knowledge of the ACRES regarding methodological issues as well as to compare the commuting route environment profiles of inner urban and suburban areas and to interpret the consequences in terms of bikeability, as defined below.
The main results show: (1) a considerable criterion-related validity of the ACRES; (2) general concordance between advertisement- and street-recruited bicycle commuters regarding both mean ratings of different items, distribution of the values and in the sizes and directions of differences in ratings between route environments; and (3) clear differences in commuting route environment profiles between the inner urban and suburban areas, demonstrating a higher level of bikeability of route environments in the suburban areas. The evidence for these interpretations of the results will be discussed below.
The first aim of this study dealt with a criterion-related validity assessment of the ACRES. In evaluating this issue, we have combined findings from all three Results sections. The first way of illuminating this issue was based on whether or not differences between the inner urban and the suburban areas were reflected by differences in perceptions of the two environments. And this was the case. First, directions of the ratings of the four ACRES items, exhaust fumes, noise, congestion: all types of vehicles and greenery, by both advertisement-recruited (Both I&S) and street-recruited men and women bicycle-commuting in inner urban and suburban areas, corresponded with the directions of the existing objective measurements. Interestingly, the same findings were noted when the ratings for the groups of advertisement-recruited men and women who only cycled in either the inner urban (Only I) or the suburban (Only S) area were compared. Second, the sizes of the differences in ratings of the route environments by Both I&S, men and women, corresponded reasonably well with the ratings of the experts. These results are in agreement with our previous findings , thus further supporting and strengthening the criterion-related validity of the ACRES. Our previous study was based on a smaller sample of street-recruited participants. In this study, we used both a larger sample of participants recruited through morning newspapers, and an enlarged sample of street-recruited participants. The similar results therefore also strengthen the external validity.
A third dimension of criterion-related validity relates to the question of whether ratings of an area would be affected by whether the participants had also experienced and rated another area. It was possible to evaluate this by comparing the ratings of commuters in Only I or Only S and Both I&S. Since the inner urban route environment setting is rather discrete and homogeneous in nature, compared to the suburban setting, which has a gradient all the way from close to the inner urban qualities in some suburbs near the inner urban area to rural qualities, we thought that the best comparison in this respect would be to compare the ratings of the inner urban area between Both I&S and Only I. The deviances noted between the two female groups are remarkably small (mean absolute difference: 0.45 ± 0.35, n = 18). The same is true of the male groups (mean absolute difference: 0.66 ± 0.41, n = 18; cf. Table 4 and Figures 7 and 8). Significant differences exist in some cases, but the magnitudes are small. We therefore interpret this result as a sign of robustness of the ACRES. In our mind, these results further strengthen the criterion-related validity of the ACRES. Thus, summing up these, we regard the criterion-related validity of the ACRES to be considerable.
Interestingly, however, significant differences in most of the items were seen in ratings of the suburban route environments between women Both I&S and Only S. Note that the differences that generate statistical significance are small (mean absolute difference: 1.14 ± 0.44, n = 14), considering that the response scales have 15 points, and that this reflects that the sample size was fairly large. Still, this finding is worth considering, particularly since the corresponding deviance is not noted in the males. At present, we can only speculate about possible explanations. In our minds, a plausible explanation is that Only S women's suburban areas are located more distant from the inner urban areas and therefore might constitute a slightly different suburban area as compared to Both I&S women's suburban settings. Differences in the spatial distribution of workplaces for men and women , as well as differences in the distances covered by them in their bicycle-commuting , support such an interpretation. The findings indicate that there is much to be learned about the actual route taken, in relation to route environments, and potential differences between men and women in this respect.
Options for future studies on the validity and other psychometric properties of the ACRES with finer distinction include the possibility to break down the commute route into distinct segments and have the participants to rate each of them. Indeed, further studies in this respect are welcomed.
Possibilities of direct comparisons with previous research regarding validity, are limited by the lack of research in this area and because we study a specific behaviour in a specific environment. Previous studies have mainly focused on physical activity and the neighbourhood. Some validity results have been reported on the Neighborhood Environment Walkability Scale (NEWS) [20–26]. In three studies, one carried out in the USA  and two in Australia [23, 25], researchers used the NEWS or a modified Australian version (NEWS-AU) in a high- versus low-walkability comparison design. This design is similar to ours in comparing perceptions of two different environments. Our findings are consistent with those of these three studies, supporting the view that differences in environment characteristics can be assessed by self-reports.
The second aim of this study dealt with the fact that active commuters, and particularly bicyclists, normally represent a small group within the general population in larger cities. Therefore, it is, at present, difficult to use population-based random samples. In our previous study, we recruited active commuting participants in the street, reaching for a sufficiently representative group of active commuters. This is, however, a difficult and impractical method when the aim is to reach a large sample of active commuters and to achieve a wide geographical coverage, particularly in larger cities. Therefore, this time we recruited participants by advertising in morning newspapers as an alternative method. Consequently, we were concerned about the representativity of the advertisement-recruited participants and therefore we compared the ratings of the inner urban and suburban route environments between the advertisement- and the street-recruited participants. First, in general, the sex-neutral means, as well as the sex-neutral mean values for the standard deviations, corresponded reasonably well between the advertisement- and the street-recruited participants for both the inner urban and the suburban route environments. Second, overall, the distributions of the response for both the advertisement- and street-recruited men and women ranged from nearly minimum to maximum. This can furthermore be interpreted as support for the use of the 15-point scales, a scale which enables other statistical analyses, for example correlation analyses, and finer distinctions between environments, in contrast to the majority of questionnaires in this area, which have Likert-type scales with fewer response alternatives [e.g. ]. Third, correspondence in both the significance and the direction of the differences was noted between the advertisement- and the street-recruited participants in the majority of the items, and for both men and women. Fourth, the sizes of differences between ratings of the inner urban and suburban route environments, expressed as sex-neutral means, generally corresponded reasonably well between the advertisement- and street-recruited participants. Furthermore, although not tested for significance, in general, no major differences appear to exist between the descriptive characteristics of the advertisement- and the street-recruited participants. Thus, overall, there was a good correspondence between the advertisement- and the street-recruited participants. This strengthens the argument for the use of the advertisement-recruited participants in this study as well as recruitment by advertisement as a feasible method. This result is encouraging, since, as stated previously, it is difficult to use population-based random samples, at present, when studying active commuters because they normally only represent a small proportion of the population in larger cities.
Commuting route environment profiles and bikeability
The third aim of this study involved commuting route environment profiles and the bikeability of different settings. Ratings of suburban route environments were compared to ratings of inner urban route environments for Both I&S as well as between Only I and Only S. In both of these comparisons, we noted significantly higher ratings by both men and women for the suburban environments in traffic safety and the extent to which the route environment stimulated bicycle-commuting. We regard these perceptions as major outcome perceptions with respect to bikeability of route environments (see below). Both of these components of bikeability are most likely composite outcomes of input from several of the environmental predictor items studied with the ACRES. And we view them at this point as potential mediators between the environmental items and the bicycling behaviour.
Given that the route environmental profiles are distinctly different (see Figures 7 and 8), we suggest that the higher ratings of bikeability of route environments in suburban areas are to be explained by factors that differ between these settings. The higher ratings of bicycle paths and greenery in the suburban as compared to the inner urban route environments, and the lower ratings of exhaust fumes, noise, flow of motor vehicles, congestion: all types of vehicles, congestion: bicyclists, conflicts, course of the route and red lights are therefore regarded at this point as explanatory candidates for the differences in the outcome perceptions of traffic safety and the extent to which a route environment stimulates or hinders bicycle-commuting. Interestingly, a recent study on motivators and deterrents of bicycling supports many of our findings regarding possible explanatory factors .
All our participants are bicyclists and therefore have a 'bicycle use behaviour'. Yet, the environmental conditions vary substantially when they bicycle-commute in the inner urban as compared to the suburban areas. Thus, cycling takes place despite the fact that a route environment is perceived as being, for example, more or less safe. From an analytical perspective, it therefore appears to be useful to differentiate between bicycle usage and the extent to which the route environments are bicycling-friendly, or even hostile.
In the predominant research and scientific dialogue dealing with walkability in more recent years, we most frequently note a paucity of such distinctions [cf. [28, 29]]. This could introduce a risk of a rather crude understanding of the linkage between environment and walking behaviour. The research on bikeability is at a much earlier state of development than that of walkability. In the methodological context of this study we therefore consider it to be useful to elaborate on the term bikeability and its definitions.
We suggest that the term bikeability should be used for factors associated with bicycling and the route environment, route distance and aspects of the interaction between the bicyclist and the bicycle which affect the conditions of a specific trip. In our minds, the term bikeability should preferably relate, in a wider perspective, to how these factors and aspects can interact with the perception and behaviour of bicycling for at least three different purposes: (1) transport; (2) recreation; and (3) exercise, as well as competition.
The purpose of bicycling may affect our understanding of bikeability. Given that bicycling with a transport purpose by definition always involves a destination, distinct distance demands are imposed. However, distance-decay relations, i.e. people's willingness to travel different distances to reach destinations [cf. ], may vary for different types of destination. Furthermore, the desired qualities of the route environments during cycling with the purpose of recreation and exercise might be higher than for the purposes of transport.
Our focus here will be limited to bikeability in relation to active commuting to one's place of work or study. We suggest that the following different environmental aspects should be included as components of possible importance for the perception of bicycling friendliness in relation to active commuting: (1) the means of transport - the bicycle; (2) the level of safety; (3) whether the route environment stimulates or hinders active commuting; and (4) the route distance and topography.
The means of transport - the bicycle - relates to various aspects of the fact that bicycling represents an interaction between a human being and technology, in which the bicycle stands for a technological environment. The effort needed for transport per distance and elevation, the possible speeds and in what kinds of environment it can be used are examples of issues that can affect bikeability. The level of safety relates to traffic safety and other forms of risks, such as crime. Whether the route environment stimulates or hinders active commuting will most likely relate to a complex of environmental variables (see above). The route distance and topography relate to issues of time allocation needed and acceptable levels of physical effort.
These components can be viewed as a chain with four different links. Weakness in one link may be enough to break the chain. However, the characteristics of the different components might also very well interact. For example, a perceived high level of traffic unsafeness may be acceptable if the route distance is sufficiently short. Furthermore, a certain degree of hilliness is not problematic if one has a bicycle with several gears, and so forth. Given this background, we think that it is important to try to study all these aspects of bikeability and the relations between them. We also think that it is important to be specific when using the term bikeability by clearly indicating which aspect of the phenomenon has been studied.
The present results point to a higher level of stimulation of bicycle-commuting and traffic safety in the suburban route environments than in the inner urban ones. We therefore assert that the bikeability of the suburban route environments in these two perspectives is higher than in the inner urban area.
The importance of traffic safety in relation to bikeability is reflected by the fact that the perception of traffic unsafeness is commonly reported as the major hindrance to bicycling [cf. ]. It has recently been stated that the likelihood of bicycle use is higher when residential neighbourhood environments have higher residential density, greater mixed land use and higher connectivity of streets . This has been interpreted to be due to, not least, proximity to destination points, which means shorter distances in these areas . If this interpretation is correct, it can, in conjunction with our findings, point in the opposite direction as well: bicycle usage in settings with higher residential density and greater mixed land use may exist under conditions of clearly suboptimal bikeability from a population perspective. Recent findings of higher demands on route environments for transport among occasional and potentially new bicyclists  point to the importance of both safe and stimulating route environment qualities from a public health perspective.
Thus, combining proximity to destinations with a high degree of personal safety and stimulating route environments for active transport therefore appears to be an important goal for future urban and regional planning with the aim of creating more widely attractive environmental settings for active commuting. From this perspective, it is important to further our understanding of the bikeability of route environments and the factors that contribute to the perception of traffic safety and whether a route environment stimulates cycling or not. In other words; what constitutes cycling-friendly route environments?
Several possible limitations of the present study should be mentioned. First, the collected data relied on self-reports. More objective measurements may provide additional information about the route environments. On the other hand, studies have shown poor agreement between objective and perceived measures of environments [33–35]. Yet, perceptions of the environments are likely to influence people's physical activity behaviours [cf. ]. For example, if people think that the traffic environment is unsafe, although it is in fact safe, their perceptions could result in a non-active commuting behaviour. A combination of objective and perceived measures of the environment may be important to further knowledge about the possible associations between environments and physical activity behaviours. Second, the generalizability of this study is limited. The work with the ACRES is in a relatively early stage at present and we have only assessed it based on one city and two different samples of people. Studies are also desirable regarding active transports with other purposes, different route environments and different samples. Our participants were active commuters and therefore probably very familiar with their route environments. Hence, their perceptions might differ from non-active commuters . Non-active commuters' perceptions of the route environment are important to study to further a more comprehensive knowledge of the route environment in relation to active commuting [cf. ]. A slightly modified version of the ACRES could be used for such a purpose. Third, data on the different participant groups was collected during different months as well as during different years: in May, 2005, for the advertisement-recruited participants, in November and December, 2005, for the street-recruited participants and in September and October, 2009, for the expert panel. The compared ratings could therefore be based on somewhat different environments due to, for example, seasonal or built environment changes. On the other hand, only small, if any, changes have occurred in Greater Stockholm during this period, and these changes probably do not have an effect on the general picture of the route environment [cf. ]. The street-recruited participants appear to be characterized by all-year round active commuting, whereas the advertisement-recruited participants appear to include summer season active commuting as well. Active commuting is, however, normally a repetitive behaviour along a specific route. As mentioned, this probably makes the active commuters very familiar with their individual route environment and their perceptions of the route environment can therefore be considered relevant, irrespective of rather moderate variations in yearly trip frequency. Fourth, no adjustments for potential confounders were made. The descriptive characteristics of the commuting participants, based on self-reports, yielded a very homogeneous picture, with few distinctions of characteristics allowing for adjustments of potential confounders. The commuting participants were, however, separated into men and women so as to allow useful separate analyses.
Despite the above-mentioned possible limitations, this study has several strengths. One is that we studied active bicycle commuters and their route environments: a specific physical activity behaviour and the specific environment within which the behaviour occurred [cf. ]. In accord with this, an important strength of the ACRES is that it deals with the whole commuting route environment and not just the local neighbourhood, as many other questionnaires on physical activity and the environment do [e.g. ]. Another strength is that the advertisement-recruited participants provided a large sample size. This enabled, among other things, different groupings, for example, separation of those who bicycle-commuted in both inner urban and suburban areas and those who bicycle-commuted in only an inner urban or a suburban area, for comparisons. Indeed, this design also disclosed a robustness of the ACRES and further strengthened its criterion-related validity. Other important strengths are the criterion-related study design, using an expert panel as a way of handling the problem of non-existing objective data for comparison, and the assessments of representativity, using different samples recruited by different sampling methods. Furthermore, the different commuting route environment profiles of inner urban and suburban areas provide a basis for furthering the understanding of bikeability in relation to route environments. To our knowledge, this is a new approach.