There has been a large amount of good work recently on environmentally mediated effects, known in the literature as ‘genetic nurture’, ‘indirect genetic effects’ or ‘social genetic effects’. I prefer to call them heritable environmental effects or environmentally mediated genetic effects, however, to emphasize the role of the ‘environment’ in their mediation.
Recent work in both model organisms (Ashbrook & Hager 2017; Baud et. al 2017; Baud et. al 2019; Sartori & Mantovani 2013; Wu et. al 2019) and humans (Bates et. al 2019; Kong et. al 2018; Willoughby et. al 2019) have suggested that behavioral traits are substantially influenced by these ‘social genetic effects’, and researchers have started to incorporate them into evolutionary and quantitative genetic models (Bijma 2014; Fisher & McAdam 2019; Hunt et. al 2019; Marjanovic et. al 2018; Saltz 2019). In humans, educational attainment traits have been shown to have social genetic effects that range from 30% of the direct effect (Kong et. al 2018) to 40% (Willoughby et. al 2019). Some have posited that the observed deflations from between-families to within-families (Lee et. al 2018; Selzam et. al 2019) may be explained by assortative mating or unobserved genes (Hart et. al 2019; Selzam et. al 2019), but such an explanation would be unable to account for the decreased predictive power of PGS in adoptive children (Cheeseman et. al 2019) or the fact that between vs within family disparities disappear when controlling for parental socioeconomic status and other parental variables (Wertz et. al 2018; Wertz et. al 2019; Willoughby et. al 2019) or that it is attenuated following controls for prenatal environment (Armstrong-Carter et. al 2019) or that adoptive parental PGS predicts adoptive child outcomes (Domingue & Fletcher 2019). Finally, despite the claims that this effect may be unique to educational attainment, it has also been observed for IQ PGS  (Trejo & Domingue 2019) and health traits (Mandemakers & Otten 2019; Sotoudeh et. al 2019).
 Trejo & Domingue (2019) does demonstrate that within-family regressions will underestimate direct genetic effects, but the effect is quantitatively small and likely has an upper bound ~11%.