may be more likely that in child twin samples such as ours heritable parenting measures are more indicative of non-passive rGE than passive rGE. That said, in the current study it was not possible to definitively distinguish between types of rGE so we do not advocate one explanation over the other. Indeed, they are not mutually exclusive of one another so it is quite possible that they are each operating in tandem. In order to make distinctions between passive and non-passive forms of rGE it is necessary to combine parent-based twin datasets with child-based twin datasets. Such designs (e.g. the extended-children-of-twins-method; Narusyte et al. 2008; and the technique of comparing twin parent and twin children samples; Neiderhiser et al. 2004) demonstrate that both kinds of rGE may be important.
