Several measures were used to assess the validity and reliability of the measurement model (see Table I). Convergent validity is the extent to which the individual items in a construct share variance between them, and is measured based on the variance extracted from each construct. The variance extracted for all constructs was close to or exceeded the recommended 50 per cent. Reliability was also considered when evaluating constructs. All constructs exhibited levels of composite trait reliability that were close to or exceeded 0.7. Discriminant validity examines whether the constructs are measuring different concepts, and is assessed by comparing the variance extracted to the squared inter-construct correlations. The variance extracted should be larger than the corresponding squared inter-construct correlations, and this condition was met in all cases, apart from one in which the squared correlation between ambience and design was somewhat higher than the average variances extracted (see Table 1). However, since the literature (e.g. Bonn et al., 2007) conceptualizes and treats ambience and design as two separate constructs, it was decided to keep them separate in the structural model.