We excluded throughfall from humid tropical forests in Puerto Rico for a period of three months to determine how drought affects greenhouse gas emissions from tropical forest soils. We established five 1.24 m2 throughfall exclusion and five control plots of equal size in three sites located on ridges, slopes, and an upland valley dominated by palms (total of 30 plots). We measured weekly changes in carbon dioxide (CO2) and bi-weekly changes in nitrous oxide (N2O) and methane (CH4) in response to manipulation. We additionally measured the effects of throughfall exclusion on soil temperature and moisture, nutrient availability, and pH. Rainout shelters significantly reduced throughfall by 22 to 32 % and decreased soil moisture by 16 to 36% (top 10 cm). Rates of CO2 emissions decreased significantly in the ridge and slope sites (30%, 28%, respectively), but not the palm during the experimental drought. In contrast, the palm site became a significantly stronger sink for CH4 in response to drying (480% decline relative to controls), while CH4 fluxes in the ridge and slope sites did not respond to drought. Both the palm and ridge site became a sink for N2O in response to drought and the slope site followed a similar trend. Soil pH and available P decreased significantly in response to soil drying; however, available N was not affected. Variability in the response of greenhouse gas emissions to drought among the three sites highlights the complexity of biogeochemical cycling in tropical forested ecosystems, as well as the need for research that incorporates the high degree of spatial heterogeneity in experimental designs. Our results show that humid tropical forests are sensitive to climate change and that short-term declines in rainfall could result in a negative feedback to climate change via lowered greenhouse gas emissions and increased greenhouse gas consumption by soils.