During a one-year period, the variability of throughfall and the standard errors of the means associated with different gauge arrangements were studied in a lower montane rain forest in Puerto Rico. The following gauge arrangements were used: (1) 60 fixed gauges, (2) 30 fixed gauges, and (3) 30 roving gauges. Stemflow was measured on 22 trees of four different species. An ANOVA indicated that mean relative throughfall measured by arrangements 1 (77%), 2 (74%), and 3 (73%) were not significantly different at the 0.05 level. However, the variability of the total throughfall estimate was about half as high for roving gauges (23%) as for fixed gauges (48-49%). The variability of stemflow ranged from 36% to 67% within tree species and was 144% for all sampled trees. Total stemflow was estimated at 4.1% of rainfall, of which palms contributed about 66%. Comparative analysis indicated that while fixed and roving gauge arrangements can give similar mean values, least 100 fixed gauges are required to have an error at the 95% confidence level comparable to that obtained by 30 roving gauges. © 2005 Elsevier B.V. All rights reserved.