Thirteen years ago, this journal published an opinion paper entitled, “Parental supervision: a popular myth”, which argued that the concept of supervision was poorly defined in terms of injury risk and that published calls for greater parental supervision to reduce child injury risk had no empirical basis.1 As both Morrongiello’s case–control study in this issue,2 and Petrass’s recent systematic review3 demonstrate, empirical data are beginning to accumulate which show that caregiver supervision does in fact reduce child injury risk and which help us to develop a better understanding of the issues around defining and measuring caregiver supervision.

DEFINITIONS AND MEASUREMENT OF SUPERVISION

Most contemporary definitions of supervision focus on three attributes: attention to the child’s behaviour, proximity to the child, and continuity of supervision.3^–6 Attention refers to both visual and auditory attention to the child’s activity and behaviour and can range from full undistracted attention to completely absent attention. Proximity refers to the caregiver’s physical proximity to the child, which ranges from physical contact (necessary, for example, with infants in swimming pools) to very long distances away (fairly common in playground and sometimes pedestrian settings). Continuity of supervision can range from constant visual and auditory attention, without interruption, to intermittent visual and auditory checking on the child’s behaviour (eg, when a parent is cooking upstairs but checks on a toddler in the basement every 5 minutes or so).

A significant challenge in using the taxonomy of attention, proximity and continuity to describe caregiver supervision of children for research purposes is identifying appropriate strategies to assess each attribute. Many experts feel that naturalistic observation is the richest source of ecologically valid information because it offers real-world assessment of how caregivers supervise children in real environments. Logistically, however, naturalistic observations can be difficult, especially if the goal is to measure real-world …