Objectives We estimated associations between objectively determined neighbourhood 'walkability' attributes and accelerometer-derived sedentary time (ST) by sex, city or type of day. Design A cross-sectional study. Setting The URBAN (Understanding the Relationship between Activity and Neighbourhoods) study was conducted in 48 neighbourhoods across four cities in New Zealand (August 2008 to October 2010). Participants The response rate was 41% (2029 recruited participants/5007 eligible households approached). In total, 1762 participants (aged 41.4±12.1, mean±SD) met the data inclusion criteria and were included in analyses. Primary and secondary outcome measures The exposure variables were geographical information system (GIS) measures of neighbourhood walkability (ie, street connectivity, residential density, land-use mix, retail footprint area ratio) for street network buffers of 500 m and 1000 m around residential addresses. Participants wore an accelerometer for 7 days. The outcome measure was average daily minutes of ST. Results Data were available from 1762 participants (aged 41.4±12.1 years; 58% women). No significant main effects of GIS-based neighbourhood walkability measures were found with ST. Retail footprint area ratio was negatively associated with sedentary time in women, significant only for 500 m residential buffers. An increase of 1 decile in street connectivity was significantly associated with a decrease of over 5 min of ST per day in Christchurch residents for both residential buffers. Conclusion Neighbourhoods with proximal retail and higher street connectivity seem to be associated with less ST. These effects were sex and city specific.