This result supports the extractivists’ statement that they avoid establishing

crops or pastures in forests with BN trees or other valuable extractive resources. Second-cycle sites showed a higher average regeneration density, but it is usually after the third cultivation cycle that the BN tree density becomes substantial. The impressive BN regeneration density at some of the sites with long histories of agricultural use (we registered up to 104 trees ha−1) is perhaps better explained by a combination of factors. At the end of each SC cycle, the mature crop is an attractive source of GSK126 solubility dmso food to the agoutis (Balée, 1994). This phase of the crop cycle coincides with site abandonment for forest succession. The dense and entangled colonizing vegetation shelters the natural disperser activity of the agoutis (Silvius

and Fragoso, 2003) and is also a favorable microhabitat for seed and seedling establishment (Peña-Claros, 2001 and Uhl, 1987). The BN seedling has a large nutrient reserve and may survive for several years under low-light conditions (Zuidema et al., 1999) but it depends on large forest gaps to thrive (Myers et al., 2000). This light-gap condition also occurs in fallows, as measured by Cotta et al. (2008). However, it is the MAPK inhibitor frequency of SC disturbances in addition to the species’ resprouting capability that ultimately results in the higher BN densities of fallows relative to BN densities in the nearby undisturbed forest. Past agricultural Tolmetin use did not appear in the PCA because it was included as a grouping variable. However, this factor directly influenced the regeneration density observed (Fig. 2b). The higher light intensities offered by pastures may favor the growth of BN seedlings (Zuidema et al., 1999), but the frequency with which pastures are burned is incompatible with forest succession processes. Burning degrades the soil fertility and homogenizes the environment, eliminating seedling-establishment micro-sites and making seed dispersal from the surrounding forest improbable (Uhl, 1987 and Uhl et al., 1988). The frequency of burning cycles, the absence of

fallow intervals, and the presence of grazing animals tend to prevent vegetation regrowth. These properties of pastures probably discourage Dasyprocta dispersal activity because we rarely found gnawed-open fruits in the pastures, even though they were abundant in SC fallows and crops. This finding reinforces our assumption that the successful colonization of SC sites by BN trees depends as much on the disturbance events as on the consecutive fallow periods. The fact that pastures established in sites previously used for SC presented a regeneration density almost as high as those sites exclusively used for itinerant agriculture does not invalidate this conclusion. To show this argument correct, we must consider the characteristics of the regeneration that occurred in pastures established in areas previously used for plant crops.