Dark Earths and manioc cultivation in Central Amazonia: a window on pre-Columbian agricultural systems?

Many commentators highlight the fertility of Anthropogenic Dark Earths (ADE), emphasizing their potential for sustainable agriculture. Some scholars believe that terra mulata (the less fertile, more extensive form of ADE) was created by means of agricultural practices used by large settled populations of pre-Columbian farmers. But what was it that these Amerindian farmers were growing? Until recently, scholarly consensus held that manioc does not perform well on ADE. New research on the middle Madeira River is showing, however, that this consensus was premature. In this region, the most common crop in ADE fields is bitter manioc. Farmers there have various landraces of manioc that they believe yield particularly well on ADE, and logically plant more of these varieties on ADE. Aspects of the behaviour and perception of manioc cultivation among 52 farmers at the community of Barro Alto were measured quantitatively on four terra firme soil types (Terra Preta, Terra Mulata, Oxisols and Ultisols). These farmers plant different configurations of landraces in different soils, according to their perception of the suitability of particular landraces and their characteristics to certain soil types and successional processes. This, in turn, shapes selective pressures on these varieties, as new genetic material incorporated from volunteer seedlings is more likely to contain traits present in the most prevalent landrace(s) in each soil type. Owing to localized population pressure at Barro Alto, manioc is under more intensive cultivation systems, with shorter cropping periods (5-10 months) and shorter fallow periods (1-2 years). The outcome of these processes is different co-evolutionary dynamics on ADE as opposed to non-anthropogenic soils. Further anthropological study of manioc swiddening in one of the richest agricultural environments in Amazonia can fill a gap in the literature, thus opening an additional window on the pre-Columbian period.


INTRODUCTION
Many commentators highlight the fertility of Anthropogenic Dark Earths 1 (ADE), emphasizing their potential for sustainable agriculture across the humid tropics (e.g., Glaser, 2007).In Amazonia, a scientific consensus has emerged, with scholars from diverse fields of inquiry agreeing that the origin of some of the ADE of the region (those soils known as Terra Preta do Índio) is a legacy of the habitational activities of Amerindian peoples (accumulation of household and kitchen waste, especially animal bones, yard sweepings burnt with low intensity fire, used palm thatch etc.) in the late pre-Columbian period.This has been termed the 'midden model' of ADE formation (Kämpf et al., 2003).Terra preta is black or dark brown in colour, extremely fertile, with high levels of phosphorus, calcium and other nutrients, and large quantities of organic matter, especially charcoal, and pieces of ceramics (Lehmann et al., 2003;Glaser and Woods, 2004).The oldest sites date to around 5.000 years (Meggers and Miller, 2006), but ADE only becomes widespread in the archaeological record around 2.000 years ago (Neves, 2008).
While the midden model is now almost completely uncontroversial, there is much debate and disagreement surrounding a second kind of ADE.Terra mulata is thought to have been formed through intensive agricultural practices involving low intensity burning and mulching (Sombroek, 1966;Woods and McCann, 1999;McCann et al., 2001;Denevan, 2001Denevan, , 2004Denevan, , 2006;;Mora, 2006).While some researchers have provided evidence of present day analogues of ADE formation through agriculture, both in Amazonia (Hecht and Posey, 1989;Hecht, 2003), on the coast of French Guiana (Rosaine, 2008) and in West Africa (Leach and Fairhead, 1995;Fairhead and Scoones, 2005), the matter remains largely unresolved.This means that despite ADE being associated with agricultural potential today, evidence 2 of their use for agriculture in the pre-Columbian era is still lacking (but for a promising start see Arroyo-Kalin, 2008a).
If ADE were used for agricultural purposes, we must first ask the questions: What crops were Amerindian farmers growing?Which cultivation practices did they use?And where were they planting?Manioc 3 (Manihot esculenta) emerged as the principal crop cultivated by Amerindian farmers as they became increasingly dependent on food production systems about three thousand years ago 4 (Lathrap, 1970; Piperno and Pearsall, 1998; Oliver,  2001).Manioc was probably domesticated between 10.000 and 8.000 years ago, given the presence of archaeological remains on the Pacific coast of Peru before 6.000 years ago (Pearsall, 1992), and this clonally propagated tuberbearing herbaceous shrub was dispersed from its birthplace in Southwestern Amazonia throughout lowland South and Central America and the Caribbean by the time Columbus arrived.Accounts of first chroniclers of the Solimões River are full of references to an abundance of manioc and associated foodstuffs (see Sweet, 1974;Espinosa, 1948).Maize has been present in Amazonia for 6.000 years, but human bone isotopic evidence suggests that it only became a primary staple in the last few centuries before European contact (Arroyo-Kalin, 2008b).
Research on manioc agriculture in Amazonia has focussed principally on long fallow shifting cultivation in marginal environments (i.e., the old, heavily leached and infertile soils of the upland terra firme) (Carneiro, 1983;   1 We use "anthropogenic" rather than "Amazonian" to reflect the fact that these soils are a global phenomenon, rather than one that is restricted to the Amazon basin.See Graham (2006) for a discussion. 2 Studies are being planned on the phytoliths that exist buried in ADE sites.They can provide direct evidence as to what was grown on these soils, as well as to the plants used as mulch. 3While the term 'cassava' is often used in English, 'manioc' is a more suitable name, as it is closer both to the scientific term Manihot and the Brazilian term mandioca.Its roots lie in the Tupi word maniot.The term 'cassava' comes from casaba, an Arawak word that refers to manioc bread, rather than the plant itself (Gade, 2002). 4While maize arrived in Amazonia some 6.000 years ago, it did not become widely cultivated until the late-pre-Columbian chiefdoms of the major rivers (Roosevelt, 1980(Roosevelt, , 1999)).Chernela, 1987;Boster, 1984;Elias et al., 2000b;Wilson and Dufour, 2002;Emperaire and Peroni, 2007).There have been few studies of manioc swiddening in richer environments (i.e., floodplain soils and ADE) despite the fact that sixteenth -and seventeenth-century accounts affirm that maize and manioc were the primary floodplain crops of the Amerindian nations, such as the Omagua and Tapajó Indians (Meggers, 1996).The pre-Columbian territories of these groups contain numerous large ADE sites today.Laura German studied the cultivation of crop assemblages (of which manioc formed a part) on ADE, but she concludes that these soils are unsuitable for manioc (German, 2001(German, , 2003)).This lack of study in the richest agricultural environments makes ethnographic projection extremely problematic (Roosevelt, 1989) and may help explain why manioc is considered a terra firme crop that requires longfallow shifting cultivation.An analogy with crop landrace definitions helps understand this situation: Anton Zeven (1998, p. 127) examined the history of the definition of this term and concluded that "an autochthonous landrace is a variety with a high capacity to tolerate biotic and abiotic stress, resulting in a high yield stability and an intermediate yield level under a low input agricultural system".Note that, by definition, a landrace is confined to low input agriculture, which, for a modern agronomist or plant breeder, means a marginal environment, like the nutrient poor soils of the Amazonian terra firme.Ethnographic projection from this definition would confine agricultural systems to marginal environments instead of the best ones available at any given time in the past.Landraces are so defined today because of the expansion of industrial agriculture that dominates the best agricultural landscapes; traditional agriculture is marginal today for the same reasons.As Anna Roosevelt (1989) cautions, the pre-Columbian period was quite different from the post-conquest period about which we have much more information.This contribution is designed to start to fill a gap in the literature on traditional manioc agriculture in one of the richest agricultural environments in Amazonia, thus opening an additional window on the pre-Columbian period.

MANIOC AND DARK EARTHS IN THE PRE-COLUMBIAN PERIOD
The ADE sites that are found throughout the Amazon basin today are the legacy of the long-term sedentary activities of pre-Columbian Amerindian groups who cultivated principally manioc.That these two phenomena (manioc agriculture and ADE) co-existed we can be sure.The question is: what kind of relationship existed between them?The work of prominent scholars from disciplines such as paleoecology, geography and archaeology suggests links between sedentism, the intensification of manioc agriculture and the formation of ADE.Dolores Piperno and Deborah Pearsall (1998, p.125) propose that [...] this efficient, undemanding carbohydrate source [manioc] became an increasingly attractive crop as human populations grew and pressure on land increased (i.e., shortening fallow periods and decreasing time for recovery of soil fertility).Its increasing importance may correlate with evidence for the development and intensification of swiddening [...] These scholars link the increasing dependence of people on manioc to population pressure, shorter fallow periods and intensive swiddening -different from the longfallow shifting cultivation practised today.Other prominent students of agricultural origins in the Americas have argued that pre-Columbian agriculture, and therefore manioc cultivation, would have been qualitatively different from the long fallow shifting cultivation practised today owing to the lack of metal axes (Sauer, 1936(Sauer, , 1952(Sauer, , 1958;;Carneiro, 1974Carneiro, , 1979aCarneiro, , 1979b;;Denevan, 1992Denevan, , 2001)), and this is where the notion of terra mulata comes in (c.f.Denevan, 2004Denevan, , 2006)).Archaeologists have also suggested that there may be a correlation of soil improvement, or terra mulata, with intensive agriculture and sedentism (Eden et al., 1984;Andrade, 1986Andrade, , 1988Andrade, , 1990;;Mora et al., 1989Mora et al., , 1991;;Herrera et al., 1998;Herrera, 1992;Mora, 2001Mora, , 2003Mora, , 2006;;Schmidt and Heckenberger, 2008;Arroyo Kalin, 2008b).Arroyo-Kalin et al. (2008, p. 5) suggest that "the emergence of anthropogenic dark earths tracks the expansion and intensification of bitter manioc during the first millennium AD".
The degree to which intentionality was involved in the formation of ADE remains unresolved.Are all ADE purposeful indigenous creations, as suggested by Clark Erickson (2003, p. 486)?On the other hand, others believe that ADE is the unintentional by-product of settlement patterns (Kern, 1988(Kern, , 1996;;Smith, 1980;Kämpf et al., 2003).A recent study provides a more nuanced hypothesis of ADE formation, offering a theoretical middle ground between the extremes of intentionality and unintentionality, which may help bridge the gap between the midden and agricultural models.Based on data from some 4.000 soil samples taken throughout a Kuikuru community in the upper Xingu River basin, Morgan Schmidt shows how ADE formation is occurring today as part of a process whereby the human activities of manioc cultivation and fishing create fertile middens and patches of improved soil.Schmidt observed Amerindian women picking up earth from fertile middens formed in the areas of habitation and taking it to manioc fields in order to enrich the soil there (Schmidt and Heckenberger, 2007).The authors conclude that in the past these areas of enriched soil may have had the effect of reinforcing the behaviour of the people who made them, initiating a "positive feedback loop of soil improvement and crop production that led to agricultural intensification in the region" (Schmidt and Heckenberger, 2008).
These introductory paragraphs have shown how scholars from various disciplines have emphasized links between manioc swiddening, sedentism and the formation of ADE.Laura German (2003) suggests links between ADE formation and use and intensified cultivation of crop assemblages.But no study has yet explicitly linked manioc cultivation with ADE, probably owing to a lack of data.This paper sets out to provide the missing data, drawing on empirical data from fieldwork that used qualitative and then quantitative methods to investigate manioc farming on ADE.

MANIOC AND DARK EARTHS TODAY
The interest in studying the relationship of manioc and ADE grew initially out of empirical rather than theoretical concerns.In the course of fieldwork seeking to explore the relationship between agriculture and ADE in general (Fraser et al., 2008), the middle Madeira River and its affluent the Manicoré River were selected field locations, owing to an abundance of ADE sites and the settled, agricultural communities that have inhabited the terra firme bluffs where these soils are situated for generations.The Madeira River is a nutrient rich white water environment, with abundant floodplains and evidence of river movement within the floodplain (oxbow lakes, paranás, bluffs with ADE now far from the river etc.) (Figure 1).The people who live in these villages are referred to (both by each other and in the literature) as caboclos.This is the Amazonian peasantry, formed through the encounter between Amerindian peoples and successive waves of migrants from Europe and Brazil's Northeast, although today they are culturally much closer to the latter5 .Caboclos are thought to have inherited some of the pre-Columbian indigenous knowledge and technologies, such as manioc farming and knowledge about plants (for a recent overview see Adams et al., 2006).
This paper focuses on the community of Barro Alto.It is an ideal location for a comparison of agriculture on ADE with that of other soils because it has the most ADE farmers of all of the ADE sites on the middle Madeira and many others who plant on non-anthropogenic soils: it is the largest community in the municipality of Manicoré, numbering some 600 individuals.In Brazilian Amazonia the most common crop for small-scale subsistence farmers on all soil types, including ADE, is bitter manioc.Why is this?Bitter manioc, which is processed into farinha, is both the fundamental carbohydrate energy source and often the most important market crop for caboclo populations throughout Brazilian Amazonia (Murrieta and Dufour, 2004;Piperata and Dufour, 2007;Piperata, 2007).The municipality of Manicoré is said to be the biggest producer of farinha in the whole of Amazonas state6 .
Bitter manioc yields well enough on poor soils (Oxisols and Ultisols) if adequately fallowed, but may also be cultivated more intensively and achieve higher yields in more fertile ecotones (ADE and the floodplain) fallow or mature forest, cultivated for up to three years, and then abandoned and allowed to re-grow.This is often presented as a successful adaptation that parallels natural successional processes (Hecht, 1982).
Manioc was domesticated from a wild subspecies adapted to the forest-savannah transition, where fire and drought are common, and is well suited to cultivation in a somewhat analogous system (Pujol et al., 2005a(Pujol et al., , 2005b)).A large and growing body of research concerning manioc has advanced our understanding of domestication in manioc cultivation, and the selective pressures and coevolutionary dynamics this entails.This work focussed our attention on the importance of sexual reproduction and the incorporation of seedling 'volunteers' in maintaining genetic diversity of manioc landraces.While the research being conducted is multi-disciplinary in orientation and spans diverse geographic locations, there are two important commonalities.All of their studies are undertaken in marginal environments of low soil fertility, and have focussed on long fallow shifting cultivation.This means that we run the risk of assuming that long fallow shifting cultivation is the only way manioc can be cultivated on the terra firme.In order to test whether or not manioc agriculture on ADE is different, we decided to compare it with standard long-fallow shifting cultivation practised by non-ADE farmers in the region.
Further investigation found that on the middle Madeira ADE farmers sometimes cultivate manioc under more intensive systems, with shorter cropping periods (5-10 months) and shorter fallow periods (1-3 years).This is most evident as a result of localized population pressure in the fields of the residents of communities with limited land and relatively large populations, such as Barro Alto and Água Azul.Additionally, bitter manioc is sometimes cultivated almost continuously along with other crops in ADE homegardens8 .Farmers have manioc landraces that perform particularly well in ADE, and logically plant more of these varieties on ADE.We hypothesized that this leads to different selective pressures on manioc landraces, as new genetic material incorporated from volunteer seedlings favours traits (such as faster maturation and early fruiting) already prevalent in varieties selected for cultivation on ADE by local farmers.This paper presents perceptual and behavioural data to demonstrate how divergences in the practices of manioc farming on these different soils leads to different co-evolutionary dynamics on ADE as opposed to non-anthropogenic soils.

METHODOLOGY
Manioc was found to be the only crop in the overwhelming majority of ADE fields and the dominant crop 9 in all other fields on the terra firme.Only a handful of farmers sometimes planted crops other than bitter manioc in ADE and interviews revealed that this is because other crops are of marginal importance for livelihoods 10 .Hence, smallscale swidden agriculture on the terra firme of the middle Madeira is almost exclusively bitter manioc cultivation.This was the case not only in Barro Alto, but at the other central research location at Barreira do Capanã and Boa Vista, the Água Azul Coast and Vista Alegre.Quantitative methodology was developed to allow comparison of manioc cultivation across four soil types.
The first parameter is the Total Landrace Frequency Index (TLF).Farmers were asked how many feixes (bundles) of maniva (stem cutting used for propagation) of each landrace they planted in each field and what kind of soil the field contained.From this the % of total space occupied by each landrace was calculated.The mean of all of the fields in which the landrace was present was then divided by the number of fields in the particular soil type to give TLF for each landrace in each soil type.The TLF was calculated with data from 19 fields in terra preta, 19 in terra mulata, 21 in Ultisols and 33 in Oxisols gathered during various fieldtrips in 2007.The second parameter is the Fraca/Forte or Strength Index.Farmers were asked to place landraces they know well on an ordinal scale, from the most fraca (weak) to the most forte (strong).The Strength Index was then calculated from the average of 42 individuals' perceptions.Further data were collected to quantify whether primary forest was being cleared, the ages at which secondary vegetation was being cleared for planting, and the extent to which volunteer seedlings are being incorporated into landraces.

CASE STUDY: BARRO ALTO
Forty years ago the community was comprised of only eight families, numbering some sixty-two people.In 2007 there were nearly 600 people living in the community.This is an interesting anomaly in the general trend of ruralurban migration, which has seen an emptying out of rural Amazonia over the last forty or so years (Becker, 1995).The village is situated on a high terra firme bluff at a bend in the Manicoré River on an ADE site, with approximately 20 ha of terra preta and 30 ha of terra mulata.After being abandoned by its former Amerindian occupants, the site was re-occupied by migrants from the Northeast during the rubber boom; these new inhabitants engaged in smallscale rubber production and manioc farming.The land tenure of the community is comprised of four adjacent historic landholdings: Barro Alto, Raimiro, Parintintin and Liberdade.We will concentrate on Barro Alto and Liberdade.Land-tenure is linked to kinship and family histories within the community.Once a field is cleared from mature forest it comes under the ownership of the family who cultivated it.In this way families have come to own various capoeiras11 , and these are passed from father to son upon the death of the former.Manioc is cultivated on the ADE of all of these landholdings, apart from Parintintin where the ADE is occupied by the residential area, homegardens and agroforests.
The most intensive manioc planting takes place on the ADE belonging to the Barro Alto and Raimiro landholdings.According to community elders, this intensive cycle first began about thirty years ago, when the population began to grow.At Liberdade, manioc is planted on ADE, but longer fallows are used, owing to the fact that only a few people are allowed to plant there.The Oxisols and Ultisols that extend into the centro (meaning inland) are where a greater proportion of the population practice long fallow shifting cultivation.Around 100 ha of new fields are planted every year, cleared from young and old capoeira, and old growth forest (see Figure 2)

Local soil classifications
Residents of the community and the wider region classify soils according to their colour and texture, the stage of successional vegetation that covers them, and their known history of use.Soils are divided into two encompassing types, barro (clay) and areia (sand).As general types, the former coincide roughly with Latossolos (Oxisols) and the latter with Argissolos (Ultisols) of the Brazilian Soil Classification System (Embrapa, 2006) 12 .Clayey soils are easily recognised, and called barro amarelo, barro vermelho, or barro branco/ tabatinga depending on yellow, red or grey colouring.The residents also recognize that soils may be a mixture of these types, such as areia misturado com barro.Terra Preta is also unmistakable owing to its very dark colouring, ceramics, different successional processes and the distinct suite of volunteers associated with it.Texturally they are designated as either solto (loose) or duro (hard), or somewhere in between.Some volunteers, such as the caiaué oil palm (Elaeis oleifera) or a weed locally known as maria-preta or rabo-de-gato (Acalypha brasiliensis, Euphorbiaceae) are so ubiquitous on terra preta that they could be designated signature species rather than merely indicators.Certainly, local people strongly associate them with terra preta.
Terra mulata and sandy Ultisols are more ambiguous, however.In terms of soil colour and texture, they are easily confused, not only by over enthusiastic researchers, but also by locals.This ambiguity is reflected in the diversity of terms that were recorded as referring to them: terra preta, terra preta misturada com areia, areia misturada com barro, areia preta and areião.Ultisols are sometimes confused with ADE because of their dark colouring.Terra mulata are less often confused with Ultisols because of their obvious higher fertility.They yield maize and also exhibit some of the same kinds of weeds as terra preta.Thus it is through differences in fertility and successional processes that terra mulata can be differentiated from Ultisols.

Landraces, soils and successional processes: forte and fraca
In the vast majority of interviews, local people employed the terms fraca (weak) and forte (strong) to describe both the characteristics of different manioc landraces and soil fertility, the latter expressed as the stage of regrowth that the successional vegetation has reached.This was also the case in the other communities in the study area.Accordingly, mandioca fraca is seen as best suited to capoeira fraca or terra fraca and mandioca forte to capoeira forte or terra forte.The origins of the use of these terms are difficult to pin down, but the categories strong and weak are present in traditional agriculture in other areas of the world (e.g., Harlan, 1992, p. 148): Landrace populations are often highly variable in appearance, but they are each identifiable and usually have local names.A landrace has particular properties or characteristics.Some are considered early maturing and some late.Each has a reputation for adaptation to particular soil types according to the traditional peasant soil classification: e.g., heavy or light, warm or cold, dry or wet, strong or weak.(emphasis added).
Taking these local notions of the relative strength or weakness of natural agents as an entry point, this section describes people's understandings of landraces, soils and successional processes, and then looks at how these elements are interrelated.

Landraces
Mandioca fraca matures quickly, but is liable to rot or not soften when soaked if left in the ground for more than a year.The Tartaruga landrace is planted preferentially on ADE (Table 1).It seems likely that this and other fraca landraces originated in the floodplain.Tartaruga is planted in the floodplain in the communities of Verdum, Amparo and La Delicia, about four hours downstream from Manicoré, where it is considered a floodplain landrace.The characteristics exhibited by fraca landraces strongly suggest a floodplain origin (fast maturing, fast seeding but not long lasting in the ground), as they are traits that serve well in the floodplain.It is possible that having come from the floodplain these landraces are more capable of taking advantage of the greater nutrient levels in ADE.Mandioca forte on the other hand, matures slowly, but is more durable in the earth, lasting as long as three years or more.This is the type of manioc most often referred to in studies of long-fallow shifting cultivation systems.
Farmers also characterise manioc in terms of starch content, and this is expressed in terms of water content.The more starch, the less water and the more farinha.
Farmers complain that some (generally fraca) landraces are muito aguada (very watery) and quebra na farinhada (quebrar here means to diminish in volume from massa to farinha).Forte landraces are said to contain the least water, and therefore they diminish the least during processing.

Soil fertility and succession
Local people deduce soil fertility from the stage that successional processes have reached at any given point in time.People say that "Terra fraca é terra com capoeira baixa, terra forte é terra com capoeira alta" (Weak soil is soil with low second growth, while strong soil is soil with tall second growth).Soil is said to be cansado (tired) from having been cultivated too much, and the resultant capoeira fina (thin secondary growth) provides an indicator of this.Farmers obviously do not understand soil fertility in terms of nutrients as soil scientists do, but in terms of their location in anthropogenic successional processes and their known history of use.Nonetheless, their understanding can also be explained by soil science, since the growth of capoeira is fuelled by nutrient pumping from deep soil horizons and by nutrient inputs from rainfall (Nye and Greenland, 1960).
This perception of soil fertility as being determined by age of capoeira and known history of use might have Table 1.Perceptions and behaviours of bitter manioc farmers in the communities of Barro Alto and Liberdade, Manicoré, Amazonas, Brazil.The Strength Index is the farmers' perception of the relative strength of each landrace, where the weakest landrace is coded as one and the strongest depends upon the number the farmer comments on (the value in parentheses is the number of people who named the landrace).The Total Landrace Frequency Index is the proportion of each landrace planted in fields on the different soil types.something to do with several generations of manioc farming under long fallow shifting cultivation.Soil fertility is expressed in a schema that makes sense for nonanthropogenic soils under long fallow shifting cultivation, but falters when it is used to explain the fertility of ADE.ADE is called terra fraca when it has been intensively cultivated (but still yields manioc, maize and watermelon) and an old capoeira on Oxisols is terra forte (even though it will not yield maize or watermelon).

Fraca and forte: preferences for landraces in different scenarios
Different manioc landraces are said to be better suited to different stages in the successional process.Many farmers claim that landraces belonging to the fraca category are better suited to fields cleared from capoeira fina; conversely mandioca forte is better suited to fields cleared from old capoeira or primary forest."Arroz (a forte landrace) é bom para mata" (Arroz is good for the forest [soils]) they say, or "Arroz só dá bem na terra forte" (Arroz only does well in strong soil).People perceive certain landraces as more or less suitable to different soil types (Table 1), and to different stages in the successional process.The categories fraca and forte apply to both manioc and soil fertility, and are combined when people state "mandioca fraca na capoeira fraca, mandioca forte na capoeira forte" (weak manioc in weak second growth, strong manioc in strong second growth).

Divergent agricultural practices
Manioc is cultivated under more intensive, shorter cropping cycles on the ADE at Barro Alto and Liberdade than when cultivated on non-ADE soils13 (Figure 2).In both Barro Alto and Liberdade, most manioc fields on ADE are cut from capoeira fraca (between one and five years of age).In both Barro Alto and Liberdade, several fields were cut from old growth forest on non-ADE (Primary forest first cut) and in Barro Alto numerous fields were cut from young capoeira soon after having been cut from old growth forest (Primary forest second cut).The most commonly planted landrace at both Barro Alto and Liberdade on ADE is Tartaruga (Table 1).Farmers explain that this is because Tartaruga yields well on ADE, and is suited to terra fraca, which is essentially equivalent to short-fallow swiddening on ADE.Tartaruga is also the most fraca landrace, with most people beginning to harvest only six months after planting.The market is almost certainly also a factor in the preference for fast maturing landraces, as the main source of income from agriculture comes with the sale of farinha.The reader will note that Tartaruga is also popular on Ultisols, and this is because most of the fields included in the study are close to the village, have been intensively farmed and are therefore designated as terra fraca and planted with more mandioca fraca accordingly.
Roxinha is preferred on terra forte, whether it is ADE or non-ADE.People prefer to plant more Roxinha when they have terra forte available as it yields the yellowest farinha.In this case the market provides a strong selective pressure.The factor determining its distribution then emerges not from soil classification per se, but rather the perceived strength of the soil, which is established from the stage of successional growth.The case of Liberdade shows this, as more Roxinha is planted on terra mulata than Tartaruga (Table 1).This is because fewer people are allowed to plant at Liberdade, lessening population pressure and allowing longer fallow periods even on TM, which makes the TM forte.
A similar thing occurs at Barro Alto; Roxinha is equally popular on terra mulata and Oxisols, because the terra mulata is located further behind the village, cultivated less intensively and therefore subject to longer fallow periods.These factors lead farmers to categorise the capoeiras on the terra mulata as forte, and they therefore are perceived to be suitable for planting of the more forte landrace Roxinha.Arroz (the most forte landrace) is most evident in fields cleared from mature forest (the most forte kind of land).
A comparison of average yields of ADE vs non-ADE shows that there is little difference in terms of sacks of farinha produced between a capoeira fraca on ADE and a capoeira forte on non-anthropogenic soil.Most farmers stated that half a hectare of manioc yields between 40 and 50 sacks in both cases.The important difference is that ADE fields are harvested after only six months, while fields on non-anthropogenic soils are only harvested after ten months.This means that production overall is higher on ADE, because two cropping cycles can fit into one cropping cycle on nonanthropogenic soils.How long ADE soils can resist this kind of intensive agriculture is still an open question, though observations suggest that they can be cultivated for longer periods of time than non-anthropogenic soils 14 .
Another point of divergence is the different months of the year in which fields can be planted.Fields cleared from older capoeira or mature forest on Oxisols and Ultisols can only be planted a few months into the dry season (august/september at the earliest).This is because cleared vegetation needs a month or two of sun to dry enough to burn.Because manioc is being planted in capoeira fina 14 German (2001) noted that the Açutuba site on the lower Negro River has withstood decades of intensive agriculture, although the fertility of the ADE is starting to degrade.on ADE, this does not need to be the case (as also noted by German, 2001).Some farmers take advantage of the ability to plant earlier -February to June -to be able to harvest when there is little farinha from non-anthropogenic soils and prices are higher.This, combined with more replanting (up to four times without fallow in some cases), also contributes to the greater intensiveness of manioc agriculture on these ADE.

Intensive agriculture and feedback loops
It appears then that Tartaruga is planted most on the ADE at Barro Alto as it is best suited to cultivation under an intensive cropping system.Its characteristics as a fraca landrace (fast maturing, but not durable) fit into a situation where fields are cleared from capoeira fraca.How to interpret the greater intensiveness of manioc farming on ADE?There is obviously a paradox between the perception that one should not plant repeatedly in what is termed capoeira fina and terra fraca, and the behaviour of continuing to plant on the ADE despite its designation as terra fraca.In Barro Alto and other communities, farmers almost never plant in the capoeira fina on non-ADE soils.
The capoeira is fina on ADE at Barro Alto and the adjacent landholding called Raimiro not because the soils are exhausted, but because the capoeira is not given time to develop; the median fallow is in the less-than-5-years class (Figure 2).Judged by yields and nutrient information, the soils are more fertile than any non-anthropogenic soil in the region.The agricultural intensification that has occurred over the last thirty years has not yet exhausted the soils, which still yield manioc well (40-50 sacks per 0.5 ha), maize and exhibit elevated pH, P and Ca compared to the Ultisols and Oxisols surrounding the village (Table 2).
Factors contributing to this intensification are complex.Demographic pressure certainly plays an important role in shortening fallow periods and selecting for fast maturing landraces.But this could not happen without greater soil fertility: Farmers would not plant in ADE if yields were not sufficient to justify the time and effort expended in clearing, planting, weeding and harvesting manioc15 .At Barro Alto, the ADE fields and young capoeiras are located much closer to the community, their proximity facilitating clearing, planting, weeding and harvesting.However, it takes three times as long to clear a four-year-old capoeira (which is all understory) than a fifteen to thirty year old capoeira (which has larger trees and much less understory).And ADE fields require much more weeding (up to five times) than non-ADE ones (once or twice between planting and harvest).The intensive cultivation of ADE cannot be explained by proximity alone, therefore, as more labour is required in their cultivation.
One way of explaining the farmers' behaviour is in terms of a short term positive feedback loop leading to a long term negative one.In this case it is the capacity of these soils to produce when cleared from capoeira fina that reinforces the behaviour of the farmer and makes him more likely to clear a field from capoeira fina in the terra preta again.As this cycle continues, agricultural intensification occurs, as observed by Morgan Schmidt and Michael Heckenberger (2008).
Feedback loops can be either positive or negative.Over the longer term, however, the action of the farmers will lead to soil degradation, as fallow periods are insufficient for soils to recuperate, and farmers do not use mulching or infield burning to restore fertility as their indigenous predecessors are thought to have done (Denevan, 2001).In a negative feedback loop, one change leads to a lessening of that change (i.e., eventual exhaustion of soils will mean that agriculture will become less intensive as it stops paying off vis-a-vis labour inputs).At Barro Alto, we already observe the effects of combined positive and negative feedback loops, with farmers complaining ADE doesn't produce as much as it used to.At the moment the positive feedback loop dominates at Barro Alto, as intensive swiddening on ADE still pays off against labour requirements.This behaviour will eventually lead to yields that are too low to pay off labour requirements.This is the negative feedback loop.In the end, if not checked, negative feedback loops will put an end to ADE cultivation.This occurred at Vista Alegre after two decades of yearly watermelon cultivation and was observed by Laura German (2001) along the lower Negro River.

Selective pressures on manioc landraces
Manioc is a clonally propagated crop.Once harvested, the stems of mature plants are stored in the shade, and when planting is to take place, these stems are chopped into 10-15 cm sticks (called manivas), and planted in a shallow hole dug with a hoe.This clonal shoot develops from the same genetic material as the plant that was harvested.Hence, a landrace is comprised of one or more 'exceptional individuals' (Zohary, 2004).In a study in French Guiana, Marianne Elias and colleagues demonstrated that 72% of landraces were not monoclonal.Rather, landraces were made up of more than one clone, which when compared had very similar phenotypes but different genotypes.The clonal propagation of manioc serves to reproduce and conserve the desired characteristics of these few 'exceptional individuals', most of which are quite heterozygous and may exhibit greater vigour and resistance to pests and diseases because of this.The problem with clonal reproduction is that only a fraction of the original population is reproduced, and this leads to a genetic 'bottleneck' resulting in a steady loss of diversity.However, manioc has retained its ability to reproduce sexually, and it is through incorporation of new clones, which appear as volunteers, that genetic diversity is maintained (see Elias et al. 2001aElias et al. , 2001b;;Martins, 2001).
Once a plant matures, it flowers, is pollinated (the main agents of pollination are bees; Silva et al., 2001), its fruits grow, mature and burst, dispersing the seed.The germination ecology of manioc is thought to be pre-adapted to slash and burn systems, owing to its having evolved in transitional forest-savanna ecosystems predominant in Amazonia in the pre-Holocene period (Pujol et al., 2002).Seeds are capable of surviving 50 years or more in capoeiras and shifting cultivation creates a mosaic of seed banks lying dormant in successional vegetation (Pujol et al., 2007).In the process of a field being cut and burnt anew, seeds germinate (stimulated * Please note that a full set of soil samples from Barro Alto is currently being analysed and will be presented in a future publication.These samples are included only to demonstrate the greater fertility of ADE even under intensive cultivation.The Oxisols of Central Amazonia normally have a P content of 1-5 and Ultisols normally have a P content of 1-7.Calcium content ranges from 0,01 to 0,5 in both soil types (Wenceslau Teixeira, Embrapa Amazônia Ocidental, personal communication, february 14, 2008).The nutrient values presented in the table correspond to the expected ranges for Terra Preta and Terra Mulata (see Teixeira and Martins, 2003;Madari et al., 2003).
by the high temperatures) and appear as seedling 'volunteers' in the fields about a week or so later, often before the clones have been planted.These seedlings grow from new genetic material, often different from the clones planted into the new field.Many farmers allow these seedlings to grow and later, intentionally or unintentionally, they are incorporated into landraces to which they bear a phenotypic resemblance.This process serves to bring new genetic material into landraces and maintain their vigour over time.The outcome of the practice of selective incorporation of volunteer seedlings is that landraces come to be formed of various clones that are morphologically similar but with distinct genotypes.Seedlings recruited from seed banks and incorporated into landraces transform pure clonality into a mixed clonal/sexual reproductive system.Seed banks and how farmers use them play a crucial role in maintaining genetic diversity in landrace populations (Pujol et al., 2007).A corollary of this is that the more widespread, popular and well known a landrace becomes, the greater the likelihood that seedlings will be incorporated, leading to increased genetic diversity.

Local practices that shape selective pressures
Farmers were found to intentionally incorporate, unintentionally incorporate, or remove seedling volunteers from fields (Figure 3).These three categories of behaviour were constructed on the basis of insights from anthropological fieldwork in the middle Madeira River region.While there is no space here for a full analysis of local understandings of sexual reproduction, suffice it to say that they vary between individuals, families and communities, and appear to be based on personal experiences rather than a common understanding.All farmers recognised the morphological differences, stating that the seedlings grow straight up without growing branches16 , and that the seedling grows only one tuber that goes straight down, and is difficult to pull up.At one extreme are the people who intentionally take cuttings from those volunteers they consider to be most healthy or attractive17 , and plant them separately to see 'what kind of manioc it is', 'if it yields really big tubers' or 'if it has a really yellow tuber'.People in this intentional category differ in how they name the resulting variety of manioc.A greater proportion of people believe that the seedlings always yield clones of an existing variety (this was also the finding of Marianne Elias and colleagues (2001a), who state that 67% of seedlings were found to be close enough to known landraces to be assigned to them).If it does not closely resemble a type present in their community, they will try to find the name with relatives or friends from other communities.Other people recognise that clones taken from the seedling are always different from existing qualities of manioc.They often seem loath to name them, calling them a 'kind from seed', maniva achada, or maniva nativa.Naming may also occur when the new clone is passed from one farmer to another.Often the original name may be forgotten, and the name of the person who gave the type of manioc is used (for example Glai and Capão).
The second category is that of unintentional incorporation.During the course of fieldwork it became apparent that many people, while not intentionally selecting and planting from seedlings, did not remove (some or all of) them during weeding.Rather these people simply harvest them together with the rest.In most instances they do not separate the maniva of the seedlings from the rest of the clones.This leads to the unintentional incorporation of clones from the seedlings.The perplexing (for the farmer) result of this is that unknown types of manioc appear in the middle of what they thought were blocks of one kind of manioc only.
The third category is formed from those who purposefully remove the seedlings during weeding.When asked why they do this, they often state that the seedlings 'não presta' (they are no good) because they only yield a single tuber, and are hard to remove from the ground.
Another interesting tendency was that some farmers plant landraces either fully mixed, partially mixed or separately.At Barro Alto and Liberdade 18 farmers mix the manioc in their fields and ten maintain them separate.Some farmers cited agronomic reasons for this (landraces are said to affect one another's growth when interplanted).Some said that when they plant in a capoeira fina, Roxinha will yield better when interplanted with Tartaruga than if planted alone18 .Others said they did it in order not to lose the landraces they harvested first19 .The unintentional genetic outcome of this is that planting different landraces mixed together greatly facilitates inter-landrace pollination, which should enhance heterozygosity and may lead to healthier seedlings.

Agricultural systems and co-evolutionary processes
The preceding section demonstrates that farmer selection of landraces is contingent on their perceptions of different types of soil (TP/TM/Ox/Ul) and vegetation, especially successional (young, middling, old capoeira and primary forest).Tartaruga is the most commonly planted landrace on ADE, especially when it is being intensively cultivated.Arroz is most common in fields cleared from older capoeira and primary forest on non-ADE soils.Roxinha comes somewhere in between fraca and forte, and is only planted on ADE when the soil has been fallowed for more than 10-15 years.
The selective pressures brought to bear through farmer decisions are shaped by other factors besides perceptions of soil type and age of fallow.These include the market (quick harvesting, yellowest farinha), social relations and kinship.The social shaping of selective pressures is evident in the fact that close kin are the most important sources of labour and manioc cuttings.Manioc farming is therefore ecologically, economically and culturally shaped.As a result of these selective pressures, different configurations of landraces are planted on different soil and fallow types.This means that certain selective pressures are brought to bear on manioc landraces.The seedlings resulting from pollination will be more likely to favour certain traits which were selected for by farmer agency (such as fast yielding, yields well in a certain soil, or produces very yellow farinha) and others that result from natural selection (e.g., fast seeding).
Many of today's most prominent scholars see the relationships that people have with cultivated plants as a form of co-evolved mutualism.Co-evolution is not only genetic, but an outcome of an interplay between genes, the organism, its environment and culture (Rindos, 1984;Piperno and Pearsall, 1998;Zeder, 2006).
The different selective pressures described above are a part of divergent co-evolutionary dynamics in two agricultural systems.Human action shapes the genetic material of plants, but in engaging with these plants and their particular characteristics, humans are themselves subtly moulded by the plants (Rindos, 1984;Zeder, 2006).For example, planting more fast-maturing but lessdurable Tartaruga in ADE means that it will be harvested earlier than if slower maturing landraces were planted.This alters labour patterns and also the precise moment when replanting occurs.Tartaruga was observed to seed earlier, something that people on the middle Madeira understand as a sign of maturation.
Farmers choose different soils and landraces in order to maximise yield vis-a-vis a plethora of social and natural externalities, including labour requirements, the market and seasonality.The different kinds of agricultural systems that emerge from this process are not the intentional, but are rather the unintentional outcomes of the interplay between human (understandings, preferences related to agriculture, selection for perceived landrace characteristics) and natural agencies (greater soil fertility, more weeds).

FINAL COMMENTS
This case shows that manioc is the most important crop for both ADE and non-ADE farmers in the communities studied along the middle Madeira River.It follows that ADE is often used for manioc cultivation, even if it would be logical to plant more nutrient demanding crops in them.Other crops are simply of little importance.This is probably also the case for other caboclo residents of rural Amazonia away from urban centres, roads and frequent river transport.This is less true for new colonists on the agricultural frontier, however.Papaya farmers in Iranduba, so often shown as an example of 'ADE agriculture', are not representative of the reality of most rural Amazonians (see Fraser et al., 2008).Iranduba, Rio Preto da Eva and other municipalities close to Manaus should be considered peri-urban, not rural.Most traditional farmers in Amazonia do not reside in locations so privileged for agricultural production, let alone do they have access to roads and trucks.Cultivation of other crops is often simply considered too risky (in terms of pests, inadequate rainfall) and logistically unfeasible (getting crops to a buyer without spoiling) to be worth it.
We suggest here that the intensification of manioc agriculture at Barro Alto demonstrates that it is possible to reduce fallow periods without a reduction of yield on ADE, at least in the short term.The case of Barro Alto shows how the presence of pre-Columbian humanenriched soils, along with population pressure and market demand for farinha, has led to a positive feedback loop of agricultural intensification.It shows how somewhat different agricultural systems have emerged on different soils, and how the distinctive configurations of landraces associated with each soil leads to divergent selective pressures on manioc landraces.
The study of manioc and ADE is important not only for our understanding of contemporary agriculture, but also shows that this would have been possible in the pre-Columbian period.The current archaeological consensus is that manioc was one of the most important crops in the pre-Columbian era (at least until the chiefdoms of the final pre-contact phase, who may have had maize as their staple), and that it fuelled the population expansion that is responsible for the thousands of ADE sites which exist in central Amazonia today (see Lathrap, 1970;Oliver, 2001;Arroyo-Kalin, 2008b).Anthropological and archaeological theory suggests that manioc farming in the past was more intensive than today, and this may have resulted in the formation of terra mulata.The case presented here can in no way be read as a reconstruction of Amerindian agriculture, but it demonstrates, for the first time, that intensive manioc agriculture is possible on ADE.This case opens a window onto the past, raising a hypothesis that can be tested by on-going archaeological work in Central Amazonia.
Figure 1.Location of communities participating in the research along the middle Madeira River and the Manicoré River, Amazonas, Brazil.

Figure 2 .
Figure 2. Frequency distributions of the age of capoeiras cut for manioc fields on all four soil types.