A brief outline of the basic concepts of cloud filtering and atmospheric attenuation corrections used in the Multi-channel Sea Surface Temperature (MCSST) method is given. The operational MCSST procedures and products are described in detail. The comparative performance of AVHRR-based MCSST's is discussed via the use of the results of the JPL Satellite-Derived Sea Surface Temperature workshops. For the four data periods there is surprisingly good correspondence in the sign and location of the major monthly mean SST anomaly features derived from MCSST's and those from a screened set of ship-based SST's. With the partial exception of the one data period severely affected in some areas by volcanic aerosol from El Chichon eruptions, global statistical measures of the MCST anomalies relative to the ship data are as follows: biases, 0.3--0.4 ¿C (MCSST lower than ship); standard deviations, 0.5--0.6 ¿C; and cross-correlations, +0.3 to +0.7. A refined technique in use with NOAA 9 data in 1985 has yielded consistent biases and rms differences near -0.1 ¿C and 0.5 ¿C, respectively.