Life history theory predicts that increasing investments into reproduction compromises survival and future reproduction. However, demonstrating such costs is confounded by positive correlations between life history traits. For example, individuals in good condition may be good at both surviving and reproducing. We studied such processes in a viviparous snow skink lizard (Niveoscincus microlepidotus) from high elevation sites in Tasmania, Australia. Our results show a stark difference in costs of reproduction between unmanipulated females from the natural population versus experimentally manipulated females (using follicle stimulating hormones). In the unmanipulated females, females with relatively larger reproductive investments survived better than females with smaller reproductive investments. In the experimental group, however, females forced to 'over-invest' into a larger clutch survived less well than controls. Thus, our study confirms the potential dangers of non-experimental estimation of costs of reproduction.