The accuracy of temperature measurements from drifters is first examined for 16 drifters (manufactured either by Metocean Data Systems or by Pacific Gyre) deployed with two temperature sensors in the tropical or North Atlantic Ocean. One of these sensors is the SST thermistor commonly used on Surface Velocity Program (SVP) drifters since the late 1980s; whereas the other sensor is a platinum temperature probe associated with a Seabird conductivity cell. The authors find (for 19 separate deployments) an average positive offset of the SST thermistor measurements in 17 out of 19 cases, exceeding 0.1 degrees C in five instances. Among the five drifters that were at sea for a year or more, two present a large trend in this offset (0.10 degrees and -0.10 degrees C yr(-1)); and in two other cases, there is a clear annual cycle of the offset, suggesting a dependency on temperature. Offsets in 9 out of 12 drifters with sea time longer than 4 months present a negative trend, but the average trend is not significantly different from zero. The study also examined 29 drifters from four manufacturers equipped only with the usual SST thermistor, but for which either a precise initial temperature measurement was available or a float was attached to provide accurate temperature measurements (for a duration on the order of a month). These comparisons often identify SST biases at or soon after deployment. This initial bias is null (or slightly negative) for the set of Clearwater Instrumentation's drifters, it is very small for two out of three sets of Technocean drifters, and positive for the third one, as well as for the set of Pacific Gyre drifters (on the order of 0.05 degrees C).