Subsampling has been widely applied in the laboratory to process freshwater macroinvertebrate samples. Currently, many governmental agencies and research groups apply the fixed-count approach, targeting a number of individuals per sample, and at the same time keeping track of the number of quadrats (fraction of the sample) processed. However, fixed-area methods are still in use. The objective of this paper was to evaluate the reliability of macroinvertebrate taxonomic richness estimates developed from processing a standard number of subsampling quadrats (i.e., fixed-area approaches). We used a dataset from 18 tropical stream sites experiencing three different levels of human disturbance (most-, intermediate-, and least-disturbed). With 12 quadrats processed (half the sample), the collection curves started to stabilize, and for more than half of the sites studied, it was possible to sample at least 80 % of the total taxonomic richness of the sample. However, we observed that the minimum number of quadrats to achieve 80 % of taxonomic richness was strongly negatively correlated with the number of individuals collected in each site: the fewer the individuals in a sample, the greater the processed proportion of that sample needed to represent it properly. Thus our results indicate that for any given areal subsampling effort (any fixed fraction of the sample), samples with different numbers of individuals will be represented differently in terms of the proportion of the total number of taxa of the whole samples, those with greater numbers being overestimated and those with fewer numbers being underestimated. Therefore, we do not recommend the use of fixed-area subsampling methods alone if the main purpose is to measure and analyze taxonomic richness; instead, we encourage researchers to use fixed-count approaches for this purpose.