BACKGROUND AND OBJECTIVE: Social support interventions have a somewhat chequered history. Despite evidence that social connection is associated with good health, efforts to implement interventions designed to increase social support have produced mixed results. The aim of this paper is to reflect on the relationship between social connectedness and good health, by examining social support interventions with mothers of young children and analysing how support was conceptualised, enacted and valued, in order to advance what we know about providing support to improve health. CONTEXT AND APPROACH: First, we provide a brief recent history of social support interventions for mothers with young children and we critically examine what was intended by 'social support', who provided it and for which groups of mothers, how support was enacted and what was valued by women. Second, we examine the challenges and promise of lay social support approaches focused explicitly on companionship, and draw on experiences in two cluster randomised trials which aimed to improve the wellbeing of mothers. One trial involved a universal approach, providing befriending opportunities for all mothers in the first year after birth, and the other a targeted approach offering support from a 'mentor mother' to childbearing women experiencing intimate partner violence. RESULTS: Interventions providing social support to mothers have most often been directed to women seen as disadvantaged, or 'at risk'. They have also most often been enacted by health professionals and have included strong elements of health education and/or information, almost always with a focus on improving parenting skills for better child health outcomes. Fewer have involved non-professional 'supporters', and only some have aimed explicitly to provide companionship or a listening ear, despite these aspects being what mothers receiving support have said they valued most. Our trial experiences have demonstrated that non-professional support interventions raise myriad challenges. These include achieving adequate reach in a universal approach, identification of those in need of support in any targeted approach; how much training and support to offer befrienders/mentors without 'professionalising' the support provided; questions about the length of time support is offered, how 'closure' is managed and whether interventions impact on social connectedness into the future. In our two trials what women described as helpful was not feeling so alone, being understood, not being judged, and feeling an increased sense of their own worth. CONCLUSION AND IMPLICATIONS: Examination of how social support has been conceptualised and enacted in interventions to date can be instructive in refining our thinking about the directions to be taken in future research. Despite implementation challenges, further development and evaluation of non-professional models of providing support to improve health is warranted.