There is growing recognition of the role of the physical environment in influencing physical activity and consequently public health. Planners are interested in the potential for environmental interventions to have population-wide effects on walking levels. However, although an association between the physical environment and physical activity has been identified, evidence on causality between specific characteristics of the environ- ment and physical activity has proved much harder to establish ( Handy, 2005). Developing audit tools to measure objectively the street-scale and fine-grain attributes of the physical environment in urban or suburban contexts relevant to physical activity is a challenge in a number of dimensions. The characteristics of the environment relevant to walking will vary according to climate, landscape, built form and cultural traditions, and audit tools should be sensitive to such differences. Audit tools have been developed in Australia (e.g. Pikora et al, 2002), North America (e.g. Brownson et al, 2004) and Europe (e.g. de Bourdeaudhuij, 2003), but problems remain and it is unclear how transferable different audit tools are from one urban context to another. The difficulties in developing a tool to audit environments such as residential streets and neighbourhoods in relation to their support for walking – their ‘walkability’ – occur at a number of levels. Firstly, it is difficult to define an appropriate neighbourhood or range of environments that are relevant to each individual. Where they walk, whether for transport or for leisure, will depend on age, employment, family circumstances, income and lifestyle. Secondly, determining what to measure objectively is not simple. While certain aspects of the environment are known to be likely to influence walking levels and may be straightforward to measure and record, other aspects, such as aesthetics or safety (Croucher et al, 2007) are less readily or meaningfully measured in objective ways. Audits of green space and natural environment factors are particularly difficult. Thirdly, related to this, the reliability of different items in an audit often varies considerably. Fourthly, finding sufficient variability in the physical environment to allow for adequate reliability testing of each element can be difficult. Fifthly, it is not straightforward to establish how the audit data for each participant’s neighbourhood area or walking route(s) is to be meaningfully summarized prior to comparing it with walking levels. Finally, such audit tools may involve collection and analysis of very large and unwieldy data sets. This paper will explore these issues, drawing on the authors’ experience of developing and testing the Scottish Walkability Assessment Tool (SWAT) (Millington et al., 2009) in Glasgow, Scotland.