g. changes in parking provision) may be more effective in reducing car trips. Changes in only a few specific perceptions of the route environment were associated with changes in commuting behaviour. Together with our previous paper (Panter et al., 2013a), our complementary approaches to longitudinal analysis strengthen the evidence for causality (Bauman et al., 2002) and the case for the evaluation of interventions aiming to provide safe, convenient routes for walking and cycling and convenient

public transport. These findings are consistent with the conclusion of a recent systematic review that studies with designs capable of supporting more robust causal inference in this field (e.g. those attempting to assess temporal precedence) tend to find more null associations than cross-sectional studies (McCormack and Shiell, 2011). In keeping with previous research (Humpel et al., 2002 and Humpel et al., 2004), check details we found that those who reported unsupportive conditions for walking or cycling at t1 tended to report that conditions had improved at t2, whilst those who already perceived the environment to be supportive tended to report no change or small decreases. This may represent regression to the mean (Barnett et al., 2005). Further research using multiple measures over time may help to disentangle effects of regression to the mean

on exposure or outcome measurement in cohorts. Quasi-experimental studies that specify and test casual pathways leading to behaviour change would also provide more rigorous Erlotinib assessment of the effects of environmental change on walking and cycling (Bauman et al.,

2002). Researchers studying changes in travel behaviour have used a variety of metrics including changes in trip frequency (Hume et al., 2009) or in time spent walking or cycling (Humpel et al., 2004) or uptake of specific behaviours (Beenackers et al., 2012, Cleland et al., 2008 and Sugiyama et al., 2013), all of which relate to different research questions. Changes in reported time spent walking or cycling can be used to infer changes in time spent in moderate-to-vigorous intensity physical activity and consequent quantifiable health benefits, but such changes may largely reflect existing walkers or cyclists making more or longer trips (Ogilvie et al., 2004) or self-report measurement error new (Rissel et al., 2010). Measures of uptake of new behaviours, including switching between usual modes of travel, may therefore also be valuable, particularly for understanding the effectiveness of interventions in promoting activity among the less active. In summary, analysis of multiple outcome measures in combination may help to ensure that robust conclusions are drawn. Key strengths of this study include the large longitudinal sample of urban and rural working adults and the use of several complementary metrics of travel behaviour change.