Satellite remote sensing of smoke aerosol-cloud interaction during the recent Smoke, Clouds, and Radiation--Brazil (SCAR--B) experiment is analyzed to explore the factors that determine the magnitude of the cloud response to smoke aerosol. Analysis of 2 years worth of data revealed that the response is greatest in the north of Brazil where aerosol optical depth is smallest, and tends to decrease as one moves southward, and as aerosol optical depth increases. Saturation in this response occurs at an aerosol optical depth of 0.8 in 1987 and 0.4 in 1995. To explore the reasons for this, a framework is developed in which the satellite-measured response can be compared to simple analytical models of this response and to numerical models of smoke aerosol-cloud interaction. Three types of response are identified: (1) cloud droplet concentrations increase with increasing aerosol loading, followed by saturation in the response at high concentrations; (2) as in type 1, followed by increasing droplet concentrations with further increases in aerosol loading. This increase in droplet concentration is due to the suppression of supersaturation by abundant large particles, which prevents the activation of smaller particles. This enables renewed activation of larger particles when smoke loadings exceed some threshold; (3) as in type 1, followed by a decrease in droplet number concentrations with increasing aerosol loading as intense competition for vapor evaporates the smaller droplets. The latter implies an unexpected increase in drop size with increasing smoke loading. The conditions under which each of these responses are expected to occur are discussed. It is shown that although to first-order smoke optical depth is a good proxy for aerosol indirect forcing, under some conditions the size distribution and hygroscopicity can be important factors. We find no evidence that indirect forcing depends on precipitable water vapor. ¿ 2001 American Geophysical Union