We agree with Jos Barlow that arguments about the historical influence of pre-Columbian peoples in Amazonian forests must be based on robust evidence. That’s why we provided different types of evidence (existence of Amazonian Dark Earths (ADE) and anthropogenic forests, presence of charcoal and useful arboreal species) to support the statement that “past forest manipulation… extended over interfluvial forest areas”, which was the focus of Barlow's comment. Although we understand Barlow’s concerns, we do not agree that they weaken our evidence or compromise our conclusions, as discussed below point-by-point.

1. Replication and pseudo-replication –The sampling design used in the study seems to have not been understood by Barlow when he states: “Furthermore, the analysis treats samples from the same plots as independent replicates, when they are clearly pseudo-replicates (they are not spatially independent samples)”. Plots within each of the 6 sites were separated by at least 1 km (some are more than 2 km apart), so they clearly are not the “same plot”. Distance between plots was indeed planned to improve independence between them, and is larger than used between most plots in ecological studies.
It could still be argued that plots closer are expected to be more similar in terms of proportion of useful species than plots in different sites. We tested this autocorrelation with Moran’s I, and the residuals of the model testing the relation between relative abundance of useful species and distance to rivers were not significantly autocorrelated at any of the distance classes (p=0.64 for the first distance class that corresponds to plots within sites; see Figure below). Hence, the results from the analyses presented in the paper are not spurious due to autocorrelation.
Even with our small sample size, we found a relation between relative abundance of useful species and distance to rivers. Useful palm species were abundant not only in both plots of the M1 site, as highlighted by Barlow, but we also found an abundance of patauá in both M2 plots and babaçu in both M6 plots; all these sites are close to main or secondary rivers. Even excluding all palms from the analysis, the relationship between the abundance of dicotyledonous trees and the distance to major rivers was significant (p = 0.05). Therefore, Barlow's affirmation that "this species alone [Euterpe precatoria] could drive the negative exponential pattern" is not true. Palms are one of the mostly used and managed plant groups, and we expected that their removal from the analysis would weaken the relationship. The pattern, however, was still significant. We agree that distinct ecological conditions in one site could influence the abundance of palm species locally; however, we documented a suite of useful trees (Table S1) with different ecological “preferences” occurring together at the same sites. Hence, we argue that the higher abundance of useful palms and trees closer to rivers than far away cannot be explained solely by ecological variables. Furthermore, since açaí and patauá are especially useful and babaçu is frequently an indicator of anthropic influences, removing them from the analysis would be like removing the evidence of ADE near rivers.

2. Ecologically meaningful variables – We agree that a suite of ecological variables change with distance from rivers. That’s why we have an entire section in the Discussion called “Ecological factors and past human management influence useful tree and palm abundance and distribution.” Numerous ecological studies have already shown this relationship; however, very few mentioned the past human effect on the vegetation. Because the soil-water level (both below and above ground) is very important in the area, we selected this variable as the main environmental gradient to be controlled. Water fluctuation is indeed an important predictor of any of the vegetation aspects studied so far in this region (biomass, tree density, tree height, tree productivity) and therefore there was no reason to expect that floristics would not be one of them. Indeed it was, as shown in the paper, but the important message is that even controlling for this factor we still detect an effect of “distance to rivers” that is independent from the soil water level.
We are aware of the contents of the letter “Developing evidence-based arguments to assess the pristine nature of Amazonian forests” (Barlow et al. 2012. Biological Conservation 152: 293-294). In this letter they suggest that “it is necessary to focus on the kinds of evidence that would indicate widespread human influence” and “the spatial extent of ADEs is very limited”. We essentially followed one of Barlow’s suggestions by sampling areas without any a priori knowledge of ADEs. In fact, all of PPBio’s sites were positioned without prior knowledge of ADEs or other evidences of anthropogenic modifications in the landscape. We found that ADEs are limited in number, but are common in areas near secondary rivers. The simple fact that some ADE sites were found within the interfluve, very far from the sites previously known along the main rivers is already strong and undisputable evidence that people indeed used the interfluves. How strong their impact in these areas was may be discussed. Our evidence is that areas of significant manipulation are indeed concentrated near “major to intermediate sized (secondary) rivers”, but this is not the end of the story.
Our group collected nearly 3500 soil samples in more than 110 kilometers of trails distributed along 620 kilometers of this same interfluve in 11 PPBio study sites. None of our samples recorded ADEs, but some of them are rich in charcoal and have useful species in the same plots. ADE is only one extreme of anthropogenic landscape transformation, restricted to occupation sites and in our paper we show that more subtle anthropogenic transformations also occur in Amazonia, as they do elsewhere (see references in our paper). Barlow et al. mention 200 transects in 18 stream catchments that did not find ADE, but maybe they contained soil charcoal and concentrations of useful species; it would be interesting to know. Past human influence in the forest is not restricted to ADE and its evidence outside of ADE sites is more subtle, such as in useful species composition of forest communities. In the letter mentioned above, Barlow et al. argue that one species of monkey is more likely to effect forest composition than occasional humans, but Barlow et al. seem to forget the fact that during thousands of years humans modified forest landscapes also and that they would affect the useful-to-humans tree component of forest composition, rather than the components useful to Neotropical primates. This lack of recognition of the cumulative consequences of occasional human actions is why so many forests are commonly considered primary forests today. We recognize that our data is limited, which is the reason we need to do more studies to better understand how past human legacies in Amazonian forests are related with modern landscapes.

3. >50 m secondary rivers – Pre-Columbian peoples would have certainly navigated almost everywhere, but the probability that pre-Columbian peoples also navigated and stayed for a longer time in large rivers is higher than along small rivers. Furthermore, small rivers in the Purus-Madeira interfluve are intermittent; in the dry season running water disappears and during the wet season the areas are flooded. For these reasons, as well as the fact that TM images don’t allow enough precision for narrower rivers, we chose only secondary rivers > 50m in width. Anyway, the fact that we analyzed only rivers larger than 50 m in width doesn’t affect our conclusions. Including smaller rivers could have added more precision to the small distances, but would not change the pattern already observed, as is clear from the graphs presented.

4. Brazil nut – The “castanhais” vary considerably in number of individuals from one location to another, with densities as low as 1 tree in 6 ha and as high as 20 trees in 1 ha having been reported (Prance and Mori. 1990. Advances in Economic Botany, 8, NYBG, New York). The threshold of >1 individual per hectare is similar to the density of individuals found in “castanhais” close to the study area reported by Peres et al. (2003. Science 302: 2112–2114). The density of “castanhais” from RADAMBRASIL (1978) showed in the Figure 1 vary between 2 to 8 individuals/hectare. Given that most “castanhais” are probably of anthropogenic origin (Shepard and Ramirez. 2011. Economic Botany 65: 44-65), we use the occurrence of this species as an indication of human influence in the forests.

5. Rubber – We agree that more evidence that rubber tappers planted or increased the abundance of rubber trees in the Amazon forest is needed. Our work may be the first that reported the higher abundance of rubber trees related to the interference of rubber tappers in the forest away from the margin of main rivers. “Seringais” with high abundance of trees are relatively common along the high floodplains of the Solimões, Amazonas and Madeira Rivers in many areas (Hund and Ohly, 2002, in Junk et al. 1997. The Central Amazon Floodplain. Springer), and were probably planted during the rubber boom, as they have similar diameters to those we observed.

6. All species –The information about all the species in these plots will be published in a forthcoming article (Souza et al. in prep.) and will be available in the PPBio web site.

7. Current management – While the effect of logging and hunting activities by recent arrivers may influence useful species abundance and diversity, modern people have different needs and knowledge than pre-Columbian peoples, so would be unlikely to affect the entire suite of species. Furthermore, the environmental conditions of the interfluve (poor soils and flooding) often constrain tree growth rates in our study sites. Trees with 10 cm of DBH were on average 42 years old, ranging from 22 to 116 years (Bruno Barçante Ladvocat Cintra, INPA, personal communication). So trees with small DBH can be very old and most of the trees we sampled are older than the most recent human occupation.

We think that our major contribution is the whole set of landscape transformations that we report away from the edges of the main rivers. Furthermore, this is the first time that the abundance of useful species in forest landscapes has been interpreted historically and we hope that it will inspire others to search for and report similar types of transformations.