Meteorological clouds often exist in the liquid phase at temperatures below 0°C. Traditionally, satellite-derived information on cloud phase comes from narrow bands in the shortwave and thermal infrared, with sensitivity biased strongly toward cloud top. In situ observations suggest an abundance of clouds having supercooled liquid water at their tops but a predominantly ice phase residing below. Satellites may report these clouds simply as supercooled liquid, with no further information regarding the presence of a subcloud top ice phase. Here we describe a physical basis for the detection of liquid-top mixed-phase clouds from passive satellite radiometer observations. The algorithm makes use of reflected sunlight in narrow bands at 1.6 and 2.25 μm to optically probe below liquid-topped clouds and determine phase. Detection is predicated on differential absorption properties between liquid and ice particles, accounting for varying Sun/sensor geometry and cloud optical properties. When tested on numerical weather prediction model simulated cloud fields, the algorithm provided threat scores in the 0.6–0.8 range and false alarm rates in the 0.1–0.2 range. A case study based on surface and satellite observations of liquid-top mixed-phase clouds in northern Alaska was also examined. Preliminary results indicate promising potential for distinction between supercooled liquid-top phase clouds with and without an underlying mixed-phase component.