There are many definitions of social isolation which draw on structural indicators (e.g. living alone), functional indicators (e.g. social support) or both. This makes comparing prevalence rates across studies difficult and provides little guidance for practitioners and service providers to identify and target socially isolated clients. The purpose of the present study was to compare, within one large population-based data-set of Canadians aged 45–85, single-item and composite indicators of social isolation, by total sample and by socio-demographics (age, sex) and health. Data were from the Canadian Longitudinal Study on Aging (CLSA) which assessed features of social network, social support and social participation. Two composite scales were created to compare prevalence rates based on structural only or both structural and functional indicators. Results indicated overall low prevalence rates of social isolation, regardless of the measure used. A composite scale using only structural features identified 5.8 per cent socially isolated adults aged 45–85. This compared with a structural and functional scale that identified 9.8 per cent socially isolated adults. The composite measures showed less variation across socio-demographics than single-item measures. Results shed light on different ways in which social isolation can be defined and how single-item and composite definitions impact our understanding of identifying socially isolated adults in a given population. Results add to discussion of measures that can be used by researchers, services providers and practitioners.