Net all-wave radiation was observed at 22 surface sites during the 1987 observation year of the First ISLSCP Field Experiment (FIFE). Eight groups of investigators employed seven different designs of net radiometer by five different manufacturers. To establish true differences in received net radiation among sites requires knowledge of differences due to instrument performance. After careful Sun/shade calibration, side-by-side comparison revealed daytime differences as large as 5 to 7% for instruments of the same manufacture and 10 to 15% between manufacturer. The largest differences are between instruments with so-called ''thin windows'' and ''thick windows'' and between instruments of ''double-dome'' and ''single-dome'' design. Comparisons with four-component reference net radiation observations reveal that the double-dome and thick window instruments have substantially lower sensitivity to longwave (thermal) net radiation than to shortwave (solar) net radiation. The magnitude of the sensitivity difference is greater when the sky is clear than when cloudy. Observations with thin window instruments agreed more closely with the reference component net radiation. Field observations made with double-dome radiometers can be corrected when net shortwave radiation is separately measured. Such a correction is shown to reduce the systematic root-mean-square differences among instruments to between one half and one quarter of those shown by uncorrected measurements. When net shortwave radiation is not available, correction according to a regression comparison against a ''standard'' net radiometer may be used. This reduces the systematic root-mean-square differences by up to one half of their uncorrected values. From this analysis it is estimated that the regression corrected daytime net radiation observations reported to the FIFE Information System include systematic-root-mean-square instrumental differences of 15 to 35 W m-2. ¿ American Geophysical Union 1992