sample size would reduce the number of singleton species, but the problem only got worse rather than better.
Pires and coworkers did not set themselves the ambitious goal of attempting to estimate the number of tree species in all of the Amazon but just in the particular “association” they sampled. Nevertheless, the quote above indicates that Pires et al. (1953) were aware of the difficulty of answering the “how many species” question without having a theoretical hypothesis concerning the distribution of relative species abundance. Two primary competing statistical hypotheses were available, then as now: Fisher’s logseries (Fisher et al., 1943) and Preston’s lognormal (Preston, 1948). The logseries predicts that the most frequent abundance class will be the rarest—singletons, which is what Pires and coworkers observed. Of the 179 species they found, 45 species (25%) occurred just once. Despite this observation, Pires et al. (1953) argued that the Preston lognormal was the “most reasonable” hypothesis, although they did not fit or mention Fisher’s logseries, of which Preston’s paper was a critique. When one does this exercise, Fisher’s logseries actually fits their data quite well (Fig. 6.1). But these data were from small plots in forest that was relatively species-poor by Amazonian standards. The question therefore arises: Which of these two distributions is a better fit to the distribution of relative tree species abundance in tropical tree communities in general and, more specifically, to relative tree species abundances in the entirety of the Amazon Basin?