Many successful theories of liquids near the melting temperature assume that small length scale density fluctuations follow Gaussian statistics. This paper presents a numerical investigation of density fluctuations in the supercooled viscous regime using an enhanced sampling method. Five model systems are investigated: the single component Lennard-Jones liquid, the Kob-Andersen binary mixture, the Wahnström binary mixture, the Lewis-Wahnström model of ortho-terphenyl, and the TIP4P/Ice model of water. The results show that the Gaussian approximation persists to a good degree into the supercooled viscous regime; however, it is less accurate at low temperatures. The analysis suggests that non-Gaussian fluctuations are related to crystalline configurations. Implications for theories of the glass transition are discussed.