A study of the uncertainties in infrared (IR) and microwave (MW) sea surface temperature (SST) retrievals by satellite is described here. The availability of both infrared (VIRS) and microwave (TMI) sensors on the Tropical Rainfall Measurement Mission (TRMM) provided an opportunity to perform this detailed SST intercomparison from two distinct but colocated sensors. Infrared retrievals provide good spatial resolution, but they are limited to clear-sky conditions. Microwave observations provide supplemental SST information in cloudy areas, but they are subject to other limitations such as poor spatial resolution and possible wind biases. One year of colocated VIRS and TMI SSTs are analyzed here, together with other colocated ancillary data sets including surface winds and water vapor, and potential sources of error are presented. The analysis shows some residual IR SST's dependence on atmospheric water vapor, despite corrective terms in the algorithm, with biases of a few tenths of a degree. The inclusion of a water vapor term in the IR algorithm reduces these biases. The results also show a higher uncertainty in MW SST retrievals for winds greater than 12 m/s, and a warm bias of MW SSTs of the order of 1¿C within 50--100 km from land. Other potential sources of uncertainties discussed in the paper are undetected clouds, choice of IR SST algorithm, diurnal warming of the skin layer, and satellite maneuvers. The results of this intercomparison are valuable for the development of a future generation of a blended IR/MW SST data set, which combines the strengths of each observational method.