A Bluff Model of Riverine Settlement in Prehistoric Amazonia
Author(s): William M. Denevan
Source: Annals of the Associati...
A Bluff Model of Riverine Settlement in
Prehistoric Amazonia
William M. Denevan
Departmentof Geography, Universityof Wisco...
A BluffModel of Settlement 655
AmazonBasin e
0MB I A1%N
r 1.
ac e,a n t
Ilpa ~ ~ Tuua aau ota.
I'?&ivo :YuriZBFLIVA
656 Denevan
Table 1. Amazonian Population Density Estimates, 1492 (per kM2).
E. Total Total
and Population,Population,
A BluffModel of Settlement 657
Cross Section of the Upper Amazon Floodplain
Orchard Pi I y
Pineapple  F'i t
Manioc Umar11f...
658 Denevan
CentralUcayali Floodplain,Bluff, andUpland
s /? Gail /to-^XEFLOODPLAIN
t lC11#> gear =g, 
A BluffModel of Settlement 659
"on a hillside,""situatedon high land,""on a
slope of a hill,"and warriors"upon the high
660 Denevan
0 Village c M e
Is. Island tocations of Io er Settlements
.a- Bluff
I. l _~~~~~~~~~~~~~~~~~...
A BluffModel of Settlement 661
0 I0
FIgr4I coIntinued.
Interaction betwe...
662 Denevan
*Terra PretaSite
ocationsof ArchaeologicalSites Varzea (floodplain)
Figure5. Locationsof 7...
A Bluff Model of Settlement 663
<S~~t>^8H-3-:-80O -0 / 2 Ci L ~~~~~~OMB IA 7 _
--:t:-:0W-....... .......... Upper Amazo...
664 Denevan
an exaggeration of village populations and
more reliabilityof house numbers.(Avillageof
20 to 60 houses at 40 ...
A BluffModel of Settlement 665
Terrapretas are usually former settlement
sites whose darkcolor is mainlydue to residue
666 Denevan
example, the largeFincaRiverasite nearLeticia
(Colombia) (1.5 km long, 45-60 m wide,
16 ha) is up to 30 cm dee...
A BluffModel of Settlement 667
vial silt annuallyto maintainsoil fertilityat Site
3 according to Herreraet al. (1992:111)....
668 Denevan
(Sippelet al.1994:75). Hydrographrecords for
Manaus and Manacapurufrom 1903 to 1985
show 17 discharges near th...
A BluffModel of Settlement 669
poorer colluvial materialwashed down from
uplands.Floodplainterraces cover 0.7 percent
or a...
670 Denevan
cultivated, as it is today. In addition, brown
anthropicsoils (terramulatta)adjacentto black
soils probablywer...
A BluffModel of Settlement 671
(Myers 1990:33, 57). However, other Shipibo
have alltheirfields in the floodplain (Bergman
672 Denevan
patches of plots indifferentmicrohabitatsvary-
ing in lengthand durationof flooding and with
soil fertilityand...
A BluffModel of Settlement 673
t Missions 66? 6O+ 54? 48
A Mission/Fort
x Fort Locationsof ColonialMissionsOCity
0A L N I
674 Denevan
and do not allow for deep hunting and agro-
forestry sustaining areas on terra firme.
Using a riverine total o...
A BluffModel of Settlement 675
in combination with seasonal cultivation of
floodplain soils. Variabilityand flexibilitypre...
676 Denevan
7. A sixteenth-century Spanish league (legua)
measured about 3 2/3 mi or 5.9 km (Medina
1934:47); however, the...
A BluffModel of Settlement 677
In Indigenous Peoples and the Futureof Ama-
zonia:An EcologicalAnthropologyof an Endan-
678 Denevan
of Friar Gaspar de Carvajal and Other Docu-
ments, compiled by Jose Toribio Medina, ed.
H. C. Heaton, pp. 383-...
A bluff model of riverine settlement in prehistoric amazonia
A bluff model of riverine settlement in prehistoric amazonia
A bluff model of riverine settlement in prehistoric amazonia
Upcoming SlideShare
Loading in …5

A bluff model of riverine settlement in prehistoric amazonia


Published on

Published in: Technology
1 Like
  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

A bluff model of riverine settlement in prehistoric amazonia

  1. 1. A Bluff Model of Riverine Settlement in Prehistoric Amazonia Author(s): William M. Denevan Source: Annals of the Association of American Geographers, Vol. 86, No. 4 (Dec., 1996), pp. 654-681 Published by: Taylor & Francis, Ltd. on behalf of the Association of American Geographers Stable URL: http://www.jstor.org/stable/2564346 . Accessed: 31/05/2011 18:30 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=taylorfrancis. . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Taylor & Francis, Ltd. and Association of American Geographers are collaborating with JSTOR to digitize, preserve and extend access to Annals of the Association of American Geographers. http://www.jstor.org
  2. 2. A Bluff Model of Riverine Settlement in Prehistoric Amazonia William M. Denevan Departmentof Geography, Universityof Wisconsin-Madison The populationdensity and level of culturalcom- plexity achieved by protohistoricAmazonians is one of the most controversialtopics in American archaeology.(Meggers 1993-1995:91) n Amazonia (Figure1), it is well established that late prehistoric settlement and agri- culture were concentrated along the major rivers.Inthe Handbook of SouthAmericanIn- dians, Julian Steward (1948:886) stated that "the important ecological differences were those between water-front and hinterland peo- ples. . . . The differences were in resources, and these partly determined population den- sity and community size"; this is reflected in Steward's (1949:659) map of South American population densities at European contact. Betty Meggers' 1971 Amazonia book is organized on the basis of va'rzea (floodplain) and terra firme (upland or interfluve) distinctions. The same emphasis was made by Donald Lath- rap (1968a; 1968b; 1970), Robert Carneiro (1970; 1995), Anna Roosevelt (1980), myself (Denevan 1966b; 1976), and others. This riverine orientation is supported by the archaeological and ethnohistorical records, al- though both are meager; few river tribes sur- vived the colonial period. The riverine pattern seems explicable by the good soil and rich wildlife resources of the va'rzea compared to the terra firme, in addition to ease of move- ment. The general assumption has been that settlement and resource use were focused in the floodplain. Roosevelt (1980), for example, repeatedly refers to agriculture and dense populations "in"the floodplains.1 However, the floodplain is a high-risk habitat because of regular annual flooding plus periodic extreme flooding that overflows even the highest ter- rain, in addition to daily tidal flooding in the delta region. Flooding is what floodplains do, and "floodplain" is defined here accordingly.2 Meggers has retreated somewhat from her earlier assertion (1971:122) that "population densityand level of culturaldevelopment were considerablygreateron the v-rzea thanon the terrafirme at the time of Europeancontact." She now argues in an article on "Prehistoric PopulationDensity in the Amazon Basin"that overall densities were comparable for both varzea and terra firme, 0.3/km2 (Meggers 1992:203), and that "[r]iverine villages are [were] not significantlylargerand more perma- nent than hinterland ones" (Meggers 1993- 1995:97). Forterrafirme,her 0.3/km2 is based on an average of present indigenous densities which varyfrom 0.04 to 1.0/km2 with a mean of ca 0.3. No references are given. However, 0.3/km2 is close to the 0.23/km2 (0.6/mi2) of Stewardand Faron(1959:52-53)3 and my own 0.2/km2 (Denevan 1976:226)4 (Table1). Meggers (1992:203) applies the same den- sity of 0.3/km2 to the vairzea.She admitsthat "higherconcentrations could have existed on the vairzea[but that]these are offset by large uninhabitableregions."She believes thatflood- plain occupation and agriculturewould have been limitedby periodic highfloods that cov- ered entire sectors of floodplain:"Onthe var- zea, the highfertilityof soils is offset by unpre- dictable variationsin the timing and intensity of annualriverfluctuations,with attendantrisk of frequent crop loss"(Meggers 1994:416). As a result,"theterrafirme [was] an equallyif not a more reliablehabitatfor humans"(Meggers et al. 1988:291). Riverinesettlement, being at riskfromfloods, depended on supplementary food production from terra firme (Meggers 1993-1995:106), presumablyfrom long-fallow shifting cultivation supporting low densities, hence the application of the 0.3 terra firme density to the overall vairzea. Meggers (1984:642; 1971:133) believes that sites in the varzea were few, mainly "temporarycamps . . . during low water" or "fishing stations" (of which few have survived). My initialreactionto this argumentwas that Annalsof theAssociationof AmericanGeographers,86(4), 1996, pp. 654-681 ?1 996 by Associationof AmericanGeographers Publishedby BlackwellPublishers,238 MainStreet,Cambridge,MA02142, and 108 Cowley Road,Oxford,OX4 1JF,UK.
  3. 3. A BluffModel of Settlement 655 AmazonBasin e 0MB I A1%N Ruig r 1. ac e,a n t Ilpa ~ ~ Tuua aau ota. I'?&ivo :YuriZBFLIVA 0 mstgo4n TS Alm Figure1. AmaonBasn, locting rvers, plae, nribs mnind it is contrary to all my previous thinking about the cultural ecology of prehistoric floodplain settlement (Denevan 1966a; 1970; 1976; 1984; 1992a). I suggested overall floodplain densities ranging from 5 to 15 per km2 (Table 1). A re- consideration of riverine settlement and agri- culture, as presented here, now convinces me that Meggers is essentially correct about the demographic significance of high floods, but that she underestimates the contribution of bluff agriculture and both local and overall riverine population densities and numbers (Denevan n.d.). I propose a model that identifies valley-side bluffs as favored sites for relatively large and semipermanent settlements. Ibelieve that most prehistoric "riverine" settlement was not lo- cated in the floodplain but rather on those fringing bluff tops that were adjacent to active river channels. Subsistence was a multiple strategy that involved the seasonal utilization of floodplain playas5 and levee soils and wildlife in combination with more permanent bluff- edge gardens and agroforestry. However, bluff settlement was spatially sporadic rather than continuous, being dense mainly along sectors where a river channel impinged against a bluff. These sectors could continue for many kilo- meters, alternating with unpopulated bluffs iso- lated from active river channels. The importance of bluff location of prehis- toric sites in Amazonia has been noted pre- viously by archaeologists and others, but with- out elaboration. Here Iwill review the evidence and suggest an integrated bluff/v~rzea strategy of complementary resource use and reexamine spatial patterns of settlement and population density. The Riverine Bluffs Amazon River bluffs are near-vertical walls rising above and confining the entrenched val- ley, separating the recent alluvium of the var- zea from the impoverished soils of the Tertiary and Pleistocene uplands (Figure 2). Heights are variable depending on location and seasonal water level; 10-20 m during low water is com- mon, but some bluffs are much higher. The bluffs, as well as floodplain levees, are subject to collapse (bank caving) when the river chan- nel is immediately adjacent and migrating (Sternberg 1960:402-404), such collapse de- stroying past and present settlement sites on the bluff edges (Lathrap 1968a:69, 76). The
  4. 4. 656 Denevan Table 1. Amazonian Population Density Estimates, 1492 (per kM2). E. Total Total and Population,Population, Terra S.E. Wet Dry Lower Overall Greater Amazon Source Riverine Firme Coast Savanna Savanna Montane Densitya Amazoniab Basinc Steward .20-.60 .10-.40 .60 .15-.60 .10 .35 .220 2,153,000 1,302,000 (1949:659- 663) Steward and .39-.78 .23 .78 .43-.78 .12 .39 .278 2,720,000d 1,645,000 Faron (1959:52-53) Hemming .275e 2,686,000 1,627,000 (1978:492) Meggers .30 .30 .300 2,931,000 1,775,000g (1992:203) Denevan 5.30 .20 9.50 1.30-2.00 .50 1.20 .589 5,750,000 3,485,000 (1970:79-82) Denevan 14.60 .20 9.50 1.30-2.00 .50 1.20 .696 or 6,800,000 4,118,000 (1976:226- .522 or or 234) 5,100,000h 3,088,000 Denevan 13.90 .580 5,664,000 3,431,000 (1992a:xxv- xxix) Newson 2.95 .41-.51 1.61 .760-.830' 7,424,000 4,496,000 (1996:10, 11, (E. (E. (E. to to 15) Ecuador) Ecuador) Ecuador) 8,108,000J 4,910,00ok Denevan (n.d.) Bluff patches: .561 5,487,000' 3,319,000m 10 or more; remainder: .30 Notes: aTotalpopulationestimatefor GreaterAmazoniadivided by the totalapproximateareaof GreaterAmazonia,9,769,000 km2, unless otherwise noted. Greater Amazonia, as I have defined it, includes the tropical interior of South America east of the An- des and north of the Tropic of Capricorn, including the Orinoco Basin but not the Gran Chaco (Denevan 1976:230, 231). bTotals are either for Greater Amazonia or extrapolations to Greater Amazonia in the cases of Hemming, Meggers, and Newson. Totals are based on tribal, regional, or habitat counts and estimates or density estimates, including projections back- wards in time. COverall density multiplied by the area of the Amazon Basin (catchment area) of 5,916,000 km2 as given by Sternberg (1975:15). dFor the several categories of tropical forest, eastern Brazil, and northern Venezuela, which together are approximately equal to Greater Amazonia. eDerived from Hemming's total of 2,184,000 for tropical Brazil (total for Brazil minus the three southern states). fMeggers' average density for both va'rzea and terra firme. gMeggers' own total is 'depending on the boundaries employed ... 1,500,000 to 2,000,000 for Amazonia as a whole." hThe second estimate is a 25 percent reduction to allow for unoccupied buffer zones between antagonistic social groups. 'Newson's density for eastern Ecuador. Extrapolation of Newson's density for eastern Ecuador to Greater Amazonia. kNewson's own total for the Amazon Basin (size not given) is 4,860,000 to 5,460,000, based on her density for eastern Ecua- dor. IThis is an adjustment of my 1992 total, changing only the riverine density, area, and population. mNewson (1996:15) indicates that my (n.d.) new Amazon Basin total should be 3,640,000, but her method for deriving this figure is not indicated. likelihood of site destruction from lateralero- sion is considerable for the highlymeandering tributariesof the UpperAmazon (Figure3), but much less so for the more stable, more linear mainstreamAmazon (Figure4) and Orinoco. Many prehistoric sites have survived on the bluffs, as evidenced by dated anthropogenic soils, bones, and culturalmaterial.Inthe central Amazon there arehighlevee sitesthatareover 2,000 years old, suggesting long-termchannel stability (Sternberg 1975:18). The Taperinha shell midden site east of Santaremin Brazilis 7,000 to 8,000 years old (Roosevelt 1991). A bluffsite is only attractivefor settlement, however, ifthere is navigablewater inthe river channel immediatelybelow. Such a proximate
  5. 5. A BluffModel of Settlement 657 Cross Section of the Upper Amazon Floodplain Orchard Pi I y Pineapple F'i t Manioc Umar11f Village Village Manioc/ Manioc/ Manioc/ Orchard Maize Swamp Forest Mai/Maic - --I~i~li WalterLevel--------{ea / {;F) ,_Maize Orchard Maize 7----- -7------ PeANutsNMaizeEL Hi~h Like UPLAND Sanu d bits Lon levpli nearse har MAIN ~PIayIy~S FLOODPLAIN Figure2. Representativecross section of the UpperAmazon floodplainnearIquitosshowing the mainchannel, side channel, water levels, naturallevees, bluffs,villages,and a terrapretasite. Source:adapted from Coomes 1992:164. channel (or lake) must be navigable year- round, not just seasonally,and if not the main channel then itmustbe connected to the main channel. As a rivershiftsits course away from a bluff site, settlement is likely to be aban- doned. In Peru, a migratingriver channel is likely to returnto a bluff edge in about 500 years (Lathrap1968a:75), at which point a for- mer village site may be reoccupied. At San Franciscode Yarinacocha(Figure3), a bluffsite on the RloUcayaliin Peru,there are18 archae- ological levels (components) with an average durationof about 100 yearsand a maximumof 200 years (Lathrap1968a:67). Sections of bluff may extend for tens of kilometers without being fringed by a river channel, the channel or multiplechannels be- ing eitheratthe opposite bluffor in mid-flood- plain.This is illustratedin Figure4. Settlement avoids such areas.Present-daytowns, smallvil- lages, and individualdwellings are invariably located on bluffsoverlooking active channels, as were colonial missions (see below). The same was apparentlytrue of most prehistoric bluffsettlements (Figure5), and this has pro- found significancefor prehistoricdemography. Dense riverinepopulationsand associated ag- ricultureapparentlywere sporadic ratherthan continuous. In addition, bluff sites must have access to freshwater.Haulingwater in pots or gourds up a steep bluffis possible but difficult.Many bluff sites today are located where small upland streams meet and dissect the bluff. Ethnohistorical Evidence for Settlement Location and Size The remarkableSpanish voyages (Figure6) on the Amazon in the mid-sixteenth century give some indicationof the topographical lo- cation, size, and distributionof indigenous set- tlement. High ground is consistently referred to, but the terms used are not necessarily in- dicative of fringingbluffsin contrastto natural levees withinthe floodplain.Inthe description of the Orellanaexpedition down the Amazon in 1541-1542, "highbanks,""hills,"and "ele- vated spot overlooking the river"are likely bluffs,but "highland"is more ambiguous(Car- vajal 1934a:199, 201, 204, 209, 217, 277; 1934b:424, 425).6 Inthe landof ChiefArripuna (Tapaj6Indians), Carvajal(1934a:227) noted that for several leagues,7 "down close to the shore there were no settlements, for they all appeared to be inthe interiorof the land,"and that fortresses were "about two or three leagues back from the river."In his second version of the Orellana voyage, Carvajal (1934b:427, 430, 431, 440) mentioned villages
  6. 6. 658 Denevan CentralUcayali Floodplain,Bluff, andUpland s /? Gail /to-^XEFLOODPLAIN t t lC11#> gear =g, ~~~(Vdrzea) Figure 3. The Central R(o Ucayali Floodplain in Peru. Note the strongly meandering channel, numerous lakes, islands,and side channels.The floodplainis over 25 km wide here. The city of Pucallpais on the western bluff adjacentto the main channel. The Shipibovillageand archaeologicalsite of Yarinacochais on the bluffover- lookinga cut-offlakethatwas untilrecentlypartof the mainchannel.The lakestillprovidesaccess to the Ucayali via a smallchannel. The Shipibovillageof Panaillois within the floodplainon a high levee that is flooded in some years.The villagedoes not have access to terrafirmeresources,so highfloods area seriouscrisis.Source: adaptedfrom Bergman1980:13.
  7. 7. A BluffModel of Settlement 659 "on a hillside,""situatedon high land,""on a slope of a hill,"and warriors"upon the high bankof the river."Salinasde Loyola(1965:201- 202, 207) in 1557 observed thatthe RfoUcayali tribes "have their settlements on top of the ravines[barrancas]of the river." Villages also were observed within the floodplains, particularly on islands. Myers (1992b:132) mentions two reports from the 1560-1561 Pedro de Ursua (Orsu'a)/Lopede Aguirreexpedition of largeOmaguavillageson Garcia Island on the Amazon west of the mouth of the RloNapo. The issue here, how- ever, is not whether settlements were located in the floodplain, but ratherthe importanceof complementary settlement on the bluffs. There are several mentions of village sizes by members of the mid-sixteenth century Spanish expeditions. CaptainAltamirano,who was with Ursu'aand Aguirre,reportedOmagua villages (RfoNapo to RfoPutumayo)number- ing 2,000, 8,000, 6,000, and 2,000 Indians (Vazquez de Espinosa1948:383-384). Myers (1992b:132-1 33) listsseveral earlyaccounts of Omagua villages with 30 to 60 large houses each. Salinasde Loyolareported villages (Co- cama)on the Ucayalinumbering200, 300, and 400 houses, with about 20 per house; the largest could have contained 8,000 people (Myers 1974:140-141). Carvajal(1934b:426) described a Yurimaguavillage as having 500 houses. Evaluatingthese early reports, Myers (1992b:138-139) estimates that in 1541-1542 there were 30 Omagua villages, averaging20 largecommunal houses each, averaging40-50 occupants in each house,8 or 800-1,000 per village,for a total of 24,000 to 30,000 people, with a reduction to possibly 6,000 by 1651. Thus it is feasible that the largestOmagua vil- lages held a few thousand people, but not likely 6,000-8,000. Smallervillages were also mentioned. Ina recent examinationof the early voyages, Porro (1994:86) for the Yurimagua (Yoriman)between Tef6and Coariin 1647 es- timates that villages averaged 20-24 houses, witha reportedfour,five, or morefamilieseach (Acuha1942:79), for an average of about 550 people in each village.This is stilla significant village size, and undoubtedly there had been population reductions since the initialSpanish contacts a century before. Sixteenth-centuryaccounts also provide an indicationof the degree of continuityof settle- ment along the Amazon. Carvajal(1934a:202, 203, 212; 1934b:428) described linear9Oma- gua and Tapaj6villagesthat were 1, 2, 2.5, 4, and 5 leagues long. An Omagua village "stretchedfor five leagues without there inter- vening any space from house to house," and for 80 leagues "therewas not from village to village a crossbow shot," and at most under halfa league (Carvajal1934a:198). One village was broken up into sectors, each with its own "landingplace down on the river"(Carvajal 1934a:203).Altamiranodescribed a Yurimagua village3 leagues long with the houses touching one another,and anotheron a high bluff("bar- rancas muy altas")reached by a staircasewith over 100 steps cut into the bluff(Vazquez de Espinosa 1948:385). Francisco Vdzquez (1981:225), who also was with Ursuiaand Aguirre, described the same village as being only 2 leagues long. Acuha (1942:79) reported a Yurimaguavillagethatwas over 1 league long 80 years later. As to the spacing of villages, conflicting figures are given. Foran Omagua territoryof over 80 leagues, Carvajal(1934a:184, 188) in- dicated 26 overlords (subchiefs), probably heads of individuallargevillages,foran average spacing of 3 leagues. Participants in the Ursuta/Aguirreexpedition 20 years later re- ported 10 to 15 and 15 to 20 Omagua villages over 150 leagues, but the historianOrtiguera, who apparentlyinterviewedsurvivors,gave 25 to 30 villages over the same distance (Myers 1992b:133).Tento 30 villagesover 150 leagues would give an average spacing between vil- lages of from5 to 15 leagues. These potentially largeseparationsconformto the model of spo- radicbluffsettlement proposed here. Initialexpeditions on the Amazon above the Rio Negro reported largeuninhabitedsectors, apartfrom broad village spacing, possibly 25 percent of the mainstreambetween the Negro and the Putumayo and 50 percent along the lower Napo and the Ucayali (DeBoer 1981:376). Such percentages are probablytoo high overall; below the Rio Negro large gaps apparently were absent. The approximate locations of the empty sectors are mapped by DeBoer (1981:368), based on Carvajal (1934a:179, 200) in 1541-1 542, Salinasde Loy- ola (1965:207, 213) in 1557, and Acuria (1942:72) in 1639. The reported lengths of these empty lands varied from 10 to 200 leagues. In addition, Gonzalo de Zu'niga (1981:6), who was with Ursulaand Aguirrein
  8. 8. 660 Denevan Town/City----- 0 Village c M e Is. Island tocations of Io er Settlements .a- Bluff IS. I. l _~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 0 0 Figure4. Amazon floodplainbetween Manausand Monte Alegre, Brazil,showing channels, islands,bluffs, and recent settlementlocations.There are 153 villages,towns, and largecities shown. One hundredand six of these are on bluffedges; 35 are withinthe floodplain,most on islands,allbut two being smallvillages;and 12 are on terrafirmewithin10 km of the floodplain(additionaltowns furtherfromthe bluffsare not shown; most are alongroads).Of the 106 bluff-edgesettlements,most areadjacentto majorchannels, and of the remainder, mostareadjacentto largefloodplainlakes.Sources:adaptedfrommapsinRADAM1976 (Vol.10);1978 (Vol.18). 1560-1561, reported three unpopulated sec- tors along the upper Amazon measuring 300, 150, and 300 leagues separating occupied provinces of 150 and 200 leagues; the empty sectors only contained fishing camps of one or two houses. DeBoer (1981) and Myers (1976) believe that the longer empty sectors were buffer zones or "no-man's lands" between hostile groups. Their argument is convincing; how- ever, such empty lands, or some or parts of them, also could have been sectors where the river channels did not impinge on bluffs and thus were unattractive to settlement. Myers (1992a:88) points out that "[wihen Orellana and Orsua were out of touch with people, they were traveling through stretches where the course of the Amazon passed through the middle of the wirzea, remote from the high banks which were the site of human habita- tion."Also, the Spaniardsintentionallyavoided the settled sides of the river (Carvajal1934a: 227). Stocks (1983:265) suggests that the empty sectors reflect the uneven distribution of fluvial lakes (prime fishing sites), but this does not seem likely(Myers 1990:14). On the lower Napo, Carvajal(1934b:420) reportedvil- lages "completely abandoned . . . because the river overflowed its course and inundated everything."These villagesmay have been re- located to bluff sites during seasonal high water,which would explainwhy rivertravelers did not observe settlement when they were out of view of the bluffs.Another explanation for empty lands could have been regionalde- population from early European-introduced epidemics, but this probably would not ac- count for the spatialpatternsobserved.
  9. 9. A BluffModel of Settlement 661 T~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 0 I0 FIgr4I coIntinued. Interaction between the floodplains and the bluff fringes and interior is indicated by early Spanish reports of roads extending inland from the rivers. Carvajal (1934a:200, 202, 203, 209, 210) mentioned "many roads," "like royal high- ways and wider," "leading into the interior," "roads that came down to the river" in the Omagua region.10 Sancho Pizarro, with Ursda, reported following an Omagua road for 30 leagues for 30 days into the interior (Vdzquez de Espinosa 1948:385-386). Nimuendajd (1952:11) observed prehistoric roads near the Rio Tapaj6s that were 1-1.5 m wide and 30 cm deep that "run almost in a straight line from one black earth [old village site] deposit to another." These roads may have been simply wide trails, but they are suggestive of resource use inland from the bluffs, as well as social interaction between varzea and terra firme. Evaluation of the Early Accounts Meggers (1992:203; 1993-1995:93-96,102- 104) considers information in Carvajal and the other early Amazon accounts to be unreliable, and she discounts the reports of large popula- tions and villages, whereas Myers (1992b), Roosevelt (1987:154), Porro (1994:86), and Whitehead (1994:35-36) disagree. Carvajal's (1934a; 1934b) 1542 accounts, it should be noted, were written within weeks or a few months after completion of the voyage, not years later when memory had faded. Hyper- bole and fantastic happenings are generally ob- vious. Certainly more credence can be given to direct observations by the Spaniards in con- trast to what the Indians reportedly told them. And more credence can be given to numbers of villages and houses, which are readily count- able, than to estimates of numbers of warriors attacking or people in a village or clustered on a river bank. Carvajal'srecord of the Orellana voyage con- sists of two versions (Medina 1934:7-11). The first written was published by the Chilean his- torian Jose Toribio Medina in Spain (Carvajal 1894) and later in English translation edited by H. C. Heaton (Carvajal 1934a). This exists as two transcripts of the lost original. The first
  10. 10. 662 Denevan *Terra PretaSite kTerraFirme(upland) ocationsof ArchaeologicalSites Varzea (floodplain) Figure5. Locationsof 75 archaeologicalvillagesites alongthe Amazon in the vicinityof Santaremand the Rio Iapajos.Most sitesareon bluffs,butseveralareon terrahirmesouthof Santarem.Source:adaptedfromMeggers 1971:132;1984:642,which is based on Nimuendajun d. (Mufoz version) is incomplete. The second is that compiled by Medina. There are only a few differences, but some are relevant to our dis- cussion. In the Muioz version, armies of 10,000 and 8,000 Indians are mentioned in the Province of Machiparo (Aisuari), whereas in the Medina version the information given is, first, "more than 400 Indians" and, second, "many Indians" and "beyond count" (Carvajal 1934a:196-1 97). The second account by Carvajal was pre- pared for the great historian of the Indies Gon- zalo FernAndez de Oviedo y Valdes, who was in Santo Domingo when Orellana's voyage ended. Oviedo requested a report from Car- vajal and then published it in his Historia de las Indias. Heaton published a translation (Carvajal 1934b) in the same volume with the Medina version. Heaton (1934:385-389) believed that the Medina version was probably written has- tily by Carvajal in late 1542 on the Island of Cubagua as a report to Orellana; that the ver- sion for Oviedo was written more carefully some months later; and, given that some pas- sages are almost identical whereas others are quite different, it is likely that Carvajal had the earlier account in front of him and corrected or amplified some passages but not others as he prepared the new version. Changes may also have been made by Orellana in the first version and by Oviedo in the second on the basis of interviews he had with other members of the expedition; also changes could have been made by transcribers. The point to be made is that the Oviedo version is probably more reliable than the Medina version. However, it is the Medina ver- sion that most modern scholars cite. There is more likely to be exaggeration of armies and village sizes in the Medina version than in the Oviedo version. There is no mention of 50,000 or 10,000 and 8,000 warriors as there is in the Medina (and Muhoz) versions (Carvajal 1934a:190, 196, 197), nor of villages 4 or 5 leagues in length. There is one mention of 5,000 warriors in the Oviedo version (Carvajal
  11. 11. A Bluff Model of Settlement 663 <S~~t>^8H-3-:-80O -0 / 2 Ci L ~~~~~~OMB IA 7 _ %, --:t:-:0W-....... .......... Upper Amazonl i.:iEiEE.i.tE;E-,.............. ' ..... Ca .....0-E.... EEC ADO R . > -g0-0k^<XX~uayaquil,<!II -i; ..< ......... /t <oSA 0 jan f ~~~~~~~~~~~~~~~~~amshiyacu e ... >....... Santa /..............1 <t P ER U > Rosa )/ SE:;:iE-::-;;EE:Egi0E;!;-0E:EChachapoyas ^ _ ' _p jurJv 9 :E:;::;;:-;;-E--;g:::-:-::::E~gg ( Po- / fArchaeological Sites ..:-g..... 0....;. Xgg;;0;;- 1.Araracuara :::-:;::E:i::E:!<;:;:;:::;S:;:::E;!Ej;;;::;:eE 5 S: < ' )2. FincaRivera ....i!0-:00: .. 0E:0 . i:;5:lti0:-0;.i . ........00000;00.;0 .d ~ u l.......... 3. Cushillococha... ...... 4.~e Cumancaya EE ED! y |...........;; .t Pucallpa 5. Sarayacu ................Wi:g |dE-!i00-gil:tifi~ii :id: ~Ei~~;06. Yarinacocha i~iSE:-12i-:,EE-~n~iE~v;iEEIE;Sur;E:l:;:E~f:!i:!!-E;:EE;!2E~i~ R e ,,Orellana Expedition, ....... :E....0 ) > t 1541-1542(upperroute)......... ..;S:;;;:;;-:EEi:i;Sf:;!;:;:;iXEE::E:;;;;;E:Eg;Sg;;;~:;;:0E j< VS 'U s aAg i r x e ii n - i;- 80tE:--E 0- -00 Xe<; < ~ ~ ~~~~~~~~~~~~0200 400 Figure 6. Upper Amazonia, locating places and archaeological sites mentioned and the upper routes of the Orellana (1541-1542) and Ursuia/Aguirre(1560-1561 ) expeditions. Sources for routes: Medinal1934:48 and Minta 1994:vi-vii. 1934b:440). Inthe latter,there are variousgen- eralreferences to numerousvillages,both large and small, and to large numbers of Indians. Thus the Oviedo version of Carvajalprovides a more moderate and believableview of Ama- zon demographyand settlementthandoes the often criticized Medina version. Villagelengths of from 1 to 5 km were re- ported by four differentpeople between 1542 and 1639; some are probably exaggerations, but villagesat least 1 league (5.9 km) long are possible. There are archaeologicalsites along bluffs that are indeed several kilometers in length (see below). Meggers' (1993-1995:98) argumentthat all large prehistoricsites repre- sent multiplereoccupations, and not single vil- lages at one point in time, is weakened by the severalsixteenth- and seventeenth-century re- ports of large,linear,occupied villages. As for the figures from the other mid- sixteenth century expeditions-Ursda/Aguirre and Salinas-there were village sizes reported as large as 6,000 to 10,000, but with houses numbering only 20 to 60. This again suggests
  12. 12. 664 Denevan an exaggeration of village populations and more reliabilityof house numbers.(Avillageof 20 to 60 houses at 40 to 50 people per house would total 2,400 to 3,000 Indians.)The esti- mates of villages with 200 to 500 houses for the Cocama would mean 4,000 to 10,000 each, using Myers'estimate of 20 persons per house; however, house sizes may have been smallerin these provinces. Ingeneral, there is a degree of consistency inthe variousaccounts of the three voyages. Archaeological Evidence The concentrationof prehistoricAmazonian riverine settlement sites on bluffs was ob- served by Nimuendajd(1952:11) in 1949 for the Santaremregion. Lathrap(1968a:77)forthe central Ucayali believed that while there was settlement in the floodplain, itwas "moretypi- cally on the bluffsadjoiningthe flood plain." Myers (1973:240-243) lists 40 archaeological sites for the central Amazon, Rio Napo, and central Rio Ucayali, of which 32 are on high banks, bluffs, or hills. Meggers' (1971:132; 1984:642) map of sites centered on the Rio Tapaj6sshows 75 by my count, based on a manuscript map by Nimuendajd (n.d.) (Fig- ure 5). Nimuendajd(1952:9-11) said that he had located 65 sites, less than halfof those in the region, almost all "foundon high ground, safe from inundation."Some are on interior terrafirmeeast of the RioTapaj6s.A 1924 map by Nimuendajd(1952:9ff)of the areajustwest of the Tapaj6s shows 28 terra preta (black earth) sites. N. Smith (1980:563) indicates 17 terrapretariverinesites between Santaremand Manacapuru.Meggers et al. (1988:289) map 23 sites along the Rio Xingd, and Meggers (1992:199) maps 37 prehistoricsites along the Rio Tocantins,most seeming to be on bluffs. Andrade (1986:22-23) maps 51 terra preta sites, includingthose of N. Smith. Archaeologicalsites have been discovered in the floodplains, usually on high levees, but they do not seem to be common. Most of the many that once existed have been destroyed by river erosion or buried under sediment. Myers (1973:241) reportsfloodplain lake sites at Cushillococha on the Peru/Brazil border. Lathrap(1968a:74)described them at Cuman- caya on the Ucayali. Sternberg (1960:417) found sites 1,000-2,000 yearsold on levees on Careiro Island near Manaus. Most floodplain villages, however, probably did not have sufficientdurationto create terrapreta of sig- nificant depth. Furthermore, most garbage probablywent into the river,ratherthan accu- mulating.Meggers (1984:642) points out that "Temporarycamps undoubtedlyexisted on the flood plain during low water to facilitateagri- culturalwork and other subsistence activities, but evidence has been obliteratedby shifts in the courses of the channels and deposition of sediment." Many villages consisted of houses on pilings that could survive periodic high floods; however, crops cannot, although vul- nerabilityvaries with crop and with duration and depth of flooding. TerraPreta Terrapreta, or terrapreta do indio (Indian black earth),is an anthropicsoil of prehistoric origin, black or dark brown, rich in organic material,and ladenwith culturaldebris(ceram- ics, bones, ash). Such soil has been reported along the Amazon, Orinoco, Negro, Guapor6, Tocantins, Tapaj6s, Xingc, Napo, Ucayali, Caqueta, Corentyne (Guyana), other rivers, Maraj6Island,and also in the Colombian Lla- nos (Nimuendajd1952; N. Smith 1980; Eden et al. 1984; Balee 1989:10-14; Herreraet al. 1992; Katzer 1944; Meggers 1993-1995:98; Andrade1986:22-23; and Woods 1995). Dates are as old as 100-450 B.C. (Eden et al. 1984:126). Most terrapreta sites are on bluff edges. Some are enormous, such as the one underlying much of the city of Santarem. Roosevelt (1989b:45) believes that some of these sites were large nucleated towns of chiefdom status,with a permanence of several hundred years. "Santaremhabitationsites ex- tend almost continuously along the river for hundreds of miles"(Roosevelt 1989a:82). An earlydescription of terrapreta soils was provided by the EnglishtravelerHerbertSmith (1879:238), who found AmericanConfederate families farmingthem near Santarem:"It[to- bacco] is cultivated on the rich black lands along the edge of these bluffs.... Allalongthis side of the Tapajds ... [which] must have been lined with these villages, for the black land is almostcontinuous, and at many points pottery and stone implements cover the ground like shells on a surf-washed beach."
  13. 13. A BluffModel of Settlement 665 Terrapretas are usually former settlement sites whose darkcolor is mainlydue to residue fromfiresfor cooking and warmth;soil carbon is high. N. Smith (1980:561-562) found that phosphorus levels are high, the result of ash, fish and game and human bones, feces, urine, and shells. Bones also account for a high cal- cium content. The pH levels are also higher thanforadjacentsoils, and aluminumlevels are moderately low. The soil fertilityof terrapreta is significantlyhigherthan for most terrafirme soils.11Currently,both Indiansand non-Indians seek out terrapreta soil for theirfields.12 Terrapreta on bluffsoccurs on a variety of terrafirmesoils, includingoxisols, ultisols,eu- trophic soils (terra roxa), and spodosols (N. Smith1980:557).Most of these soilsareof very low naturalfertility,the terra roxa being the main exception. Soildarknessvariesconsider- ablyand there is no agreement on color criteria for a terrapreta. The AraracuaraProject re- searcherson the RioCaquetainthe Colombian Amazon found brown soils adjacentto or sur- rounding pockets of black soil. They believe that the brown soils are not settlement zones but ratherpermanentor semi-intensiveagricul- turalsectors that were maintainedby organic additives that produced the brownish color. The brown soils differfrom the black soils in color, less depth, less phosphorous, and hav- ing fewer cultural remains (Andrade 1986: 53-54; Mora et al. 1991:75-77). Sombroek (1966:175) pointsout the occurrence of brown soil (terra mulatta) in the Belterra area east of the RioTapaj6s.This is a soil lighterthan terra preta, without artifacts, occurring in bands around terra preta on terra firme. He believed that "thissoil has obtained its specific proper- ties from long-lasting cultivation."Undoubt- edly, different kinds of terrapreta originated and evolved through different pathways and on differentparentsoils (Woods 1995). Terrapreta bluff sites are linear, paralleling the rivers.The 17 river-edgesites examined by N. Smith (1980:563) range in size from about 1 to 90 ha and average 21.2 ha. A site on the Rio Xingu near Altamirais 1.8 km long and 500 m wide, covering 90 ha, and one at Mana- capuru on the Amazon is 4 km long and ex- tends 200 m inland,totaling 80 ha according to N. Smith(1980:560;2 km by 400 m accord- ing to Myers 1973:240). Roosevelt (1987:157) says that the terrapreta underlyingSantarem covers 5 km2 (500 ha).The site at Tef6is 6 km long (terra preta?) (Myers 1973:240). Although some terra preta sites have been excavated, apparentlynone have been mapped. Darkness and depth of terra preta are prob- ably indicative of length of occupation, whether continuous or periodic. The riverine sites examined by N. Smith (1980:563-564) had depths of up to 2 m with an average of 0.73 m. He suggests an accumulation rate of about 1 cm per 10 years of occupation; thus 2 m depth would mean 2,000 years of settle- ment, butthisseems unlikely.Ceramics,which are common in black earths, vary in style, in- dicating different culturalphases and hence, possibly, discontinuous settlement. Meggers (1992) demonstrates this on the basis of seria- tion sequence analysis of ceramics in sites along the Rio Xingc, Rio Tocantins,and other tributaries.She believes that sites with large surface areas "represent multiple reoccu- pations ratherthan largesingle villages"(Meg- gers et al. 1988:291) and that, "[h]ence, the surfaceextent of archaeologicalsites cannot be used to infer village size, as has often been assumed" (Meggers 1995:29). Roosevelt (1989b:45-46), on the other hand, believes that many largesites were of long duration,as do Mora et al. (1991:39, 61, 77) and Herrera et al. (1992:110)for Araracuara.Reoccupation was probably a factor, but it remains to be demonstrated how long and how large most specific occupations were, a criticalissue for estimatingsite populations. There have been a few terra preta sites re- ported withinfloodplainson high levees or on river terraces (Sternberg 1960:417, 419; Coomes 1995a). However, most are on terra firme bluffs. N. Smith (1980:562) found that even the 12 interfluvesites he examined were located either along a small river or within a few hundred meters of a perennialstream, in- dicatingthe importanceof navigation,potable water,and relativelyeasily cleared riparianfor- est. Bluffsites areoften justabove fallsor rapids or where bluffs jut into rivers or at tributary junctions (N. Smith 1980:562-563; Myers 1990:19), suggesting strategic considerations. Sombroek (1966:1 75) observed that terra preta sites in eastern Brazil"areespecially frequent at outer bends of the rivers,where no flood- plains occur between the water and the up- land,and where the waterwaycan be scanned freely." Certainly not all bluff sites are terrapreta. For
  14. 14. 666 Denevan example, the largeFincaRiverasite nearLeticia (Colombia) (1.5 km long, 45-60 m wide, 16 ha) is up to 30 cm deep, with no indication of blackearth,though there isa brown midden layer(Bolian1975:22, 27). Numerous small terrapreta sites have been reported in interfluveforests in Brazil,includ- ing a totalof about 50,000 ha between the Rio Tapajosand Rio Curua-Una(Katzer1944:35- 38). Such sites are usuallymuch smallerthan most bluffsites; N. Smith (1980:563) obtained an average of 1.4 ha for 12 interfluve sites. Some are as smallas 0.3-0.5 ha. Most are shal- lower than bluffsites, suggesting shorter peri- ods of occupation, but sufficient to create blackearth.They tend to be circularand prob- ably represent a few large communal houses or a circle of smallerhouses. However, there are also some very large (100 ha or more) interfluve terra preta sites, such as between the lower Tapajosand Arapiunsrivers, some being distant from any perennial source of water (McCann 1994); however, Nimuendaju (1952:11) reported prehistoricwells in this re- gion. Present-day terra firme Indian settlements seldom produce terrapreta, undoubtedly be- cause they are of short durationand are sup- ported by shifting cultivation.Permanent vil- lages would requiresome form of stable, sus- tainable agriculture. Exceptions might occur with some of the older mission villages, and these should be examined for black earthfor- mation. Blackearth is stillbeing formed today in the backyardsof Amazoniantowns in Brazil visited by N. Smith (1980:555-556). He notes that the rate of formation is probably slower now than in prehistorydue to pig and chicken scavaging and the currentpractice of building fires on above-ground platforms. The TerraPreta of Araracuara Colombian scholars have studied the terra preta bluff soils at Araracuaraon the Rio Caqueta in the Colombian Amazon (Figure6) (Andrade1986; Moraet al.1991; Herreraet al. 1992; and Cavelieret al. 1990; also Edenet al. 1984). Analyses of soils, pollen, phytoliths, plantremains,and ceramics, plus radiocarbon datingprovide systematicevidence for the na- ture of prehistoric bluff settlement and land use. Site 2 is on the 140 m highAraracuarasand- stone plateau overlooking the Rio Caqueta. The site was originallyoccupied about 2700 B.C.;anthropicsoil is lacking.The second pe- riod of occupation was continuous from A.D. 385 to A.D. 1175, nearly 800 years, covering an areaof 6 ha of brown anthropicsoil. Open zones of savannawere created. The dominant crop was maize, with some manioc. Palmsand other fruittrees included Iriartea,Oenocarpus, Jessenia, Mauritia,Astrocaryum, Bactris, At- talea,and Lepidocaryum,all importanthouse- garden and swidden-fallow trees today. This suggests that agroforestrysystems had been established. By A.D. 800, agriculturehad be- come intensified (and was nearlycontinuous) with additionsof chilipeppers, caimito(Poute- riacainito),and varietiesof manioc. The maxi- mum percentages of palms coincide with maximums of cultivated crops. Algae and silt occur in the soil, possibly from swamps in the floodplain, suggesting the transportof alluvial siltand organicmatterto the fields to improve fertilityand to reduce erosion. Most of the sur- rounding forest was maintained even with a growing population. By A.D. 1200, the settlement was aban- doned, and the savannasdisappeared(burning ceased?)and were replaced by forest.Cultivars decreased in variety, but some manioc re- mained, suggesting the continuation of small swidden plots.A varietyof fruittrees persisted, such as avocado, star apple (Chrysophyllum cainito), guava (Psidiumguajava),and peach palm (Bactrisgasipaes), suggesting swidden- fallow management. The settlement itself may have combined with an existingvillage(Site3) 3 kmaway,which expanded ataboutthe same time. Abandonment, however, was probably not for ecological reasons, as the originalsite had been productivefor hundreds of years. The anthropicsoils at Site 3 span from A.D. 0-1800. Initialfields were small with long fallows. After A.D. 800 the site experienced intensifiedagriculture(long cropping, shortfal- lowing), assisted by soil additivesincludingdo- mestic waste, dead leaves, wood, and weeds, plus silt and algae. These additives occur in greater quantities than at Site 2. The site is largerthan Site 2, with 14.5 ha of brown and blacksoil, extending forabout1 km.The black soil probablyoriginatedat house sites and the brown soil at intensive field and garden sites. Ittook an estimated 245 tons per ha of allu-
  15. 15. A BluffModel of Settlement 667 vial silt annuallyto maintainsoil fertilityat Site 3 according to Herreraet al. (1992:111).How- ever,the present average depth of the siltlayer is only 36 cm, much less than would be ex- pected. Also, this materialwould have had to have been carriedat least900 m fromthe river bank and up the bluff, a considerable labor requirement.Thus,a massive movement of silt can be questioned, and local organic inputs may have been more important. The Araracuaraproject seems to provide support for the bluff model presented here: lengthy permanentsettlement and largepopu- lations; permanent agriculture involving ar- tificialsoil fertilitymaintenance, supplemented by forms of agroforestry;and interactionwith an adjacentfloodplain. Thatthis land-use pat- tern was widespread in prehistoricAmazonia seems likely,but remainsto be demonstrated. Agricultural Evidence What was the primaryhabitatthat sustained prehistoricriverinefarmers?The archaeological and early ethnohistoricalevidence is meager and inconclusive. Laterethnohistoric and re- cent ethnographic evidence is probably not indicative of prehistoric conditions. Thus the model suggested here is largelyinferential,but nevertheless feasible and probable. VarzeaCultivation Floodplainagricultureis described by Meg- gers (1971:125-126), based on earlyaccounts, primarilyfor the Omagua Indians. By 1700 most indigenous riverine farming had disap- peared. One of the few survivingrivertribes is the Shipibo on the RioUcayaliin Peru,whose subsistence atPanailloisdescribed by Bergman (1980). The largefloodplains are highlyvariedenvi- ronments in terms of flooding regimes, soils, and microrelief(Denevan 1984). Forthe Ama- zon, varzea width ranges between about 10 and 50 km, generally being narrower below the RioNegro. Duringlow water largeamounts of terrainare exposed as playas, islands, and low levees forsufficientlengthsof timeto allow the cultivationof fast-growing crops such as maize, peanuts, beans, and watermelon. The highest naturallevees are above flood levels most years and can be plantedwith the longer maturingvarieties of manioc, other annuals, plantains,and nativefruittrees.13A gradientof plantingswas thus possible inrelationto water- level duration,as is illustratedfor the Shipibo by Bergman(1980:60). The marginsof va'rzea lakes were also preferred sites for settlement and cultivationbecause of rich,year-roundfish resources (Lathrapet al. 1985:42). Thefloodplainshave clearadvantagesforag- riculture.Soils, while variable in texture and nutrients,are mostly of relativelyhigh fertility, renewed by annualflood deposits, with ade- quate moisture availabilitywhich is capillary and not necessarilydependent on direct rain- fall, although sandy soils may dry out. Forest clearance, ordinarilydifficultin upland forests using stone axes, often is not a problem here. The playasand low levees lack largetrees and may be cleared of herbaceous vegetation by the annualfloods and deep silt deposits. The high levees may be forested with earlysucces- sionalgrowth which is relativelyeasilycleared, butthey mayalso have large,maturetrees. The need for weeding varies with site. Cropping can be annual,althoughplayasfrequentlyshift in location.Forthe RroPachiteain Peru,Campa Indianfields on levees are fallowed for only one to three years, and only because of weed invasion,not decliningfertility(Allenand Tizon 1973:145). However, at SanJorgenear Iquitos, swiddens on high levees are cropped for two to three years followed by about ten years of agroforest prior to reclearing (Hiraoka 1989:92-93). Highlevee soils only infrequently receive flood silt deposits and hence fertility may be less thanforthe low levees and playas, and this results in field fallowingaccording to Hiraoka(1985:8, 15). The disadvantage of floodplains resides in the irregularityand variabilityof flooding. On the tributariesdestructive floods can occur without warning, even during low-water sea- sons. The rise and fallof the main Amazon is more regular,but periodic extreme floods oc- cur, filling the entire floodplain, topping the naturallevees, and destroying most crops. On the RfoUcayali(Figure3), floods covering the highest levee for a month or so occur about every ten years, and slightlylower levees may be topped twice every ten years (Bergman 1980:53; Chibnik 1994:221). At Manacapuru, duringthe seven-year period 1979-1985, the entirefloodplainwas flooded only inJune1982
  16. 16. 668 Denevan (Sippelet al.1994:75). Hydrographrecords for Manaus and Manacapurufrom 1903 to 1985 show 17 discharges near the 1982 level or greater,foran averageof aboutone yearinfive (Richey et al. 1989:246). Forthe centralAma- zon, Meggers (1971:12) points out that "[a]lthoughthe 1953 crest reached only ten feet above average, it had a disastrousimpact on crops and cattle."Thiswas the highest river stage duringthe period 1903-1953 (see Stern- berg 1975:22). Thusfloodplainsare a relatively rich but very high-riskhabitat. As for crops, the accounts in the sixteenth and seventeenth centuries mention the im- portance of both maize and manioc in the floodplains. Forthe Omagua, additionalcrops included sweet potatoes, peanuts, beans, to- bacco, achiote (Bixaorellana),cotton, gourds, peppers, pineapples, cacao, and avocado and other fruits (Meggers 1971:125). Carvajal (1934a:192, 200, 211) reported Orellana'smen obtainingvery largequantitiesof food in some villages.However, generallyit is not clearas to whether the primaryproduction was coming from playasand islands,levees, or terrafirme, or from all, which is likely,and which crops were grown where. There is historicaland ar- chaeological evidence that both maize (mud flats)and manioc (levees) were primaryflood- plain crops (Fritz 1922:50; Acuha 1942:35; Heriarte1952:17;also see Roosevelt1980:112- 159). Food Storage Food was stored in order to feed people duringthe high-waterand flood periods. The storage of manioc in pits in the floodplain, keeping for up to two years without rotting,is mentioned by Fritz (1922:50) and Acuha (1942:35-36). The Tapaj6stored maize in bas- kets buried in ash for protection from weevils (Carvajal 1934b:432). Maize was normally stored in raftersand in raised cribs (Carvajal 1934b:398). Bittermanioc was made into cas- sava flour,or mixed maize-manioc flour (Car- vajal 1934b:398, 425), and also stored. Such farinhatoday keeps for long periods. In addi- tion, fish were smoked and turtleswere kept live in pens (Acuha 1942:39). Orellana'smen obtained 1,000 turtlesina singlevillageaccord- ing to Carvajal(1934a:193), and Ursuia'smen reported a village with 4,000 turtles (in cor- rates)caught inthe dryseason foreatinginthe wet season (Vdzquez de Espinosa1948:385). The capacity for food storage alleviated the problem of flooding, but may not have been sufficientto counter high floods of long dura- tion or floods occurring several years in se- quence. The loss of crops to flooding was mentioned by Fritz(1922:50):"whenthe River is in high-flood, they are left without a chagra [field]and not seldom without anythingto live upon."Likewise,Acuha (1942:35) saidthatthe Indians"areexposed to a great reduction and loss [of crops], because of the powerful floods." In addition, fish availabilityis greatly reduced in the river channels during high water. Storage of food was mainly seasonal, and at best provided for a year or so of non- production. More than food was involved, however. Seed andtubercuttingsmustbe pre- served in adequate quantityfor future plant- ings. Seed can be stored, if not eaten in emer- gencies, but Iknow of no long-termstorageof manioc cuttings. If manioc plants were de- stroyed by flooding, new cuttingswould have to have been obtained from terrafirmefields. Thus non-floodplain (terrafirme)sources of food would have been essentialforthe support of largenumbers of people over the long run. Today,of course, floodplainfarmersandfisher- men have access to market sources of food; however, inthe Iquitosregiontradingforstaple food between floodplainand uplandfarmersis stilltheir "safetynet,"not the market(Coomes 1995b). River TerraceCultivation The premise of the BluffModel isthatvairzea agriculture,while very productive,is at riskbe- cause of periodic highfloods thatfillthe flood- plains.This is not entirely so, however. There are large riverterraces within the valley-sur- faces of Pleistocene and Holocene age thatare higher than the highest levees and are not flooded. Ages rangebetween about 5,000 and 100,000 years. Dumont et al. (1990:128, 131) map terraces in the R(o Ucayali valley with dates from 8,520 B.P.to 32,750 B.P.Thus river terrace soils are relatively young and less weathered than terra firme soils and have a relativelyhigh nutrientcontent (RAsAnenet al. 1993:211). Other terrace soils, however, are
  17. 17. A BluffModel of Settlement 669 poorer colluvial materialwashed down from uplands.Floodplainterraces cover 0.7 percent or about 4,800 km2of the PeruvianAmazon, based on measurementson Landsatphotogra- phy (RasAnenet al. 1993:211). In the eastern Amazon, river-terraceformations have been explained by sea-level oscillations, whereas tectonic activityinthe sub-Andeanfaultsystem is believed responsible in the upper Amazon (Saloand Kalliola1991:248-249). Riverterraces, in partat least, have a consid- erable potentialfor settlement and cultivation because they do not flood, but only ifthey are readilyaccessible. On the largeriversthey do not seem to be utilized much today, possibly because they are distantfrom the main chan- nels, not easilyreached by canoe, and because they are fragmented. The same difficulties probablyprevailedin prehistory.Riverterraces have seldom been examined by archaeolo- gists, so their past agriculturalimportance re- mains unknown. Moraet al. (1991:6) map ter- races along the RioCaqueta in the Colombian Amazon and show several prehistoricsites on them. Where present and accessible, they may have provideda nonflooded alternativeto bluff occupation. Bluff Cultivation The bluffsare partof the well-drained, low- fertility terra firme where prehistoric agriculture has been portrayed as long-fallow shifting cultivationcomparable to that of survivingIn- dians today (Meggers 1971:42, 99; Roosevelt 1980:87). However, there are contemporary Amazonianexamples of more permanentpro- duction, based on soil protection and mainte- nance, such as the short-fallowswiddens and house gardensof the Kayap6(Hechtand Posey 1989), Waika (Harris 1971), Siona (Siona- Secoya) (Vickers 1983:37-38), and Amuesha (Salick 1989:201-205). Agroforestry systems based on perennials have been examined for the Bora (Denevan and Padoch 1988) and Runa(Irvine1989). An important consideration for prehistoric terra firme agriculture in Amazonia is the use of stone axes to clearforest (Denevan 1992b). Giventhe inefficiencyof stone axes, compared to metal axes, enormous amounts of time and energy were required to clear mature forest, particularlyhardwoods and trees of large di- ameter.Experimentsindicatethatsteel axes are 7 to 60 times more efficient than are stone axes. Consequently, such clearing was prob- ably minimized or avoided. Once a clearing was established, itwas likelymaintainedsemi- permanentlyin production, possibly gradually enlarged, with only short periods of fallow- ing.14Smallplots could be established at tree- fall openings.15Otherwise, fields would have been concentrated along the edges of rivers and streamswhere growth was largelysecon- dary,dominated by smalltrees and softwoods thatcould be comparativelyeasilyclearedwith stone axes. Fieldsizes became largerwith the use of metal axes (Denevan 1992b:157, 162). A second consideration in reconstructing prehistoricterrafirme agricultureis that agro- forestrysystems were probablyintegratedwith permanent gardens and swiddens. Such agro- forestry would have included swidden-fallow management (enriched fallows, successional management)and fruitorchards(Denevan and Padoch 1988) and formsof forest manipulation (Posey 1985:144-152). The use and cultivation of fruit trees is mentioned frequently in the earlyaccounts, attestingto their importance.16 Fruitswere cultivated, managed in disturbed sites, and wild. Cultivatedfields, in particular, suggest permanentfields and settlement, since perennialsare less common in short-cropping, long-fallow systems where people move fre- quently.The highconcentrationof fruittrees at the Araracuarabluffsite (see above) is indica- tive of agroforestrysystems, such as orchards, house gardens, and managed fallows (Mora et al. 1991:43). The location of fruittrees is generallynot clear,but they were probablyon both high levees and bluffs. Laureanode la Cruz (1885:188)in1651 reportedthatOmagua Indians"went into the interiorforests [monte adentro] to search for the fruitsof palms and othertrees,"possiblytrees growinginswidden fallows, indicatingforest utilizationinlandfrom the bluffs. Permanent and semipermanent production systems were probably characteristicof the bluffzones where there is archaeologicaland ethnohistoric evidence of relatively dense population concentrations. Terrapreta soil is indirectevidence of intensive terrafirmeculti- vation, in that permanent or semipermanent settlement creating terrapreta is usuallyasso- ciated with permanent or semipermanent fields. The black soil itself was undoubtedly
  18. 18. 670 Denevan cultivated, as it is today. In addition, brown anthropicsoils (terramulatta)adjacentto black soils probablywere either agriculturalsoils en- riched over time by organic additives (Som- broek 1966:175; Andrade1986:54; Moraet al. 1991:77) or possibly they were zones of spo- radicsettlement. Whether agriculturalpressure on bluff-zone vegetation was sufficient to create savannas and soil degradation is uncertain. Carvajal (1934a:227)mentioned Tapaj6"fortressesscat- tered along the tops of hillsand for the most partstripped bare,"but this was "twoor three leagues backfromthe river."These could have been interiornaturalsavannas,as occur today in the Santarem region. However, Carvajal (1934b:435) at apparentlythe same site stated that [a]illalong that side of the river . .. there were not only savannas, but also uplandsand slopes and hills cleared of trees." Thus inten- sive agriculturemay have created anthropic savannas. This is also suggested for the Araracuara site in Colombia (Mora et al. 1991:39). BluffCultivationin Contemporary Peru Analogs of prehistoric bluff agricultureand agroforestryexist today but have been little studied. One example is that of riberefo set- tlers along the Amazon in the Iquitosregion (Figure6). The villageof Tamshiyacuis on the east bluffof the floodplain,which is 15-25 km wide, adjacent to the main channel (Hiraoka 1986). Thisis Tertiaryage upland(altura),40 m above mean riverlevel.A smallsector of levees and sandbarsis planted in rice, maize, beans, and vegetables at low water. Cultivationex- tends inlandfrom the bluffedge for between 1 and 5 km, the greatest distance being from the villageitself.The immediatebluffzone (1.0 to 1.5 km wide) is used for short-fallowswid- dens. A middle zone (1.0 to 2.5 km wide) is used for agroforestry.The most interiorzone (1.0 to 2.0 km wide) is used for scattered long- fallowshiftingcultivation,beyond which is ma- ture forest (Hiraoka1986:358). However, for villages along the nearby Rio Tahuayo just south, Coomes (1992:182-188) did not find this zonation, althoughthe same types of fields occur. He describes a patchwork of these types with the mix related to the age of the settlement, the older villages having more managedfallowand morefieldsthataredistant from the village (4 km or more). Inthe short-fallowfields at Tamshiyacu,one or two crops of manioc are obtained over two to three years, with yields of 9.0-9.5 metric tons/ha/yr. No fertilizers are needed, even though the soils are very acidic and low in nutrients.Other crops do poorly,however. For manioc, several weedings are necessary, and this is the main problem ratherthan declining yields. The low-protein manioc staple is sup- plemented with fishingand huntingand flood- plain seed crops. The agroforestryzone con- sists of a young managed fallow or transitional stage (two to six years), with pineapple domi- nating,plusfruittrees such as Pourouma,Inga, starapple, cashew, and peach palm.The older agroforest (over six years) is dominated by groves of umarf(Poraqueibasericea)which last 20 to 30 years. Other older economic trees include Mauritiaflexuosa, Astrocaryumcham- bira, Brazil nuts, peach palm, and avocado. These agroforestrysystems are derived from Indian antecedents. The zone most distant from the bluffat Tamshiyacuconsists of long- fallowshiftingcultivationwithinprimaryforest. A year of maize, rice, and beans isfollowed by one to two years of manioc and plantains,fol- lowed by mostly naturalgrowth for 20 years or more. Most familiesalso have a home gar- den (1,000-2,000 mi2). At Santa Rosa, another riberefo village on the lower Rfo Ucayali, farmers also cultivate both varzea and terrafirme habitats(Padoch and de Jong 1992). The village was originally on a naturallevee, but was moved recently to a bluffsite because of high floods. The flood- plain playas and levees continue to be culti- vated, includingthe old village-sitelevee, but most farmersalso have swidden and agrofores- tryplots inthe bluffzone. Avarietyof cropping combinations representing 12 different pro- duction systems are employed, but the most common component is swidden-fallow agro- forestry.On the RfoTahuayomost floodplain farmers have terra firme plots for security (flood insurance), whereas many terra firme farmers do not also have floodplain plots (Coomes 1992:171). Thisis indicativeof which habitatis most risky.On the other hand,flood- plain farmers move to higher ground after flood "wipe outs," but stillprefer to locate as close to the riveras possible. The Shipibo at San Fernandode Yarinacochaand at Sarayacu plantcrops on both floodplainand terrafirme
  19. 19. A BluffModel of Settlement 671 (Myers 1990:33, 57). However, other Shipibo have alltheirfields in the floodplain (Bergman 1980:90; Myers 1990:33). Elsewhere,Braziliancaboclos atCoarion the middle Amazon have some of their fields on the vairzea,but most of theirfields are on the upland as protection against loss of crops to high floods (Parkeret al. 1983:181-183). Tu- cuna Indiansnear Leticia,Colombia, live in the floodplain, but may have fields on the terra firme as "an insurance"garden when major floods are expected, and people on the terra firme have fields both there and in the flood- plain(Bolian1975:20). Projectingto Prehistory The long-fallow zone at Tamshiyacuprob- ably did not exist when stone axes were used. The short-fallow zone adjacent to the bluff edges probably consisted of semipermanent fields, orchards, house gardens, and young agroforest. This zone of intensive cultivation initiallywas probablynot very wide, given the difficulty of clearing primary forest, except alongtributarystreams.However, as the popu- lation grew, this strip could have been ex- panded as mature trees at the edge were graduallyfelled. The swidden-agroforest zone may have been as much a manipulatedforest as a managed agroforest,with, however, small gardens located at tree falls,their density de- creasingwith distancefrom riversand streams. Where shiftingcultivation(long or shortfallow) was practiced,the fallowvegetation was prob- ably managedto maintaina high proportionof usefulplants,as is done today by the Boraand other Amazon people (Denevan and Padoch 1988). These fallows would have been re- cleared while young, and tree felling was still feasible with stone axes. Discussion Bluff!VarzeaComplementarity I have examined historical, archaeological, and agriculturalevidence to supportthe thesis that prehistoric riverine settlement in Ama- zonia was primarilylocated on fringingbluffs ratherthan in the floodplains, based on a dual strategy or complementarity of va'rzea and terrafirmeresource use. Blufflocationsare less productive than the floodplain but are more reliable,and they arealso more defensible. The critical archaeological evidence is terra preta, anthropicblacksoil created by prehistoricset- tlement mainlyon the bluffs,and often of con- siderable extent and depth. The evidence for intensive bluffagricultureis mainlyinferential. Large,permanent villages require productive, stable agriculture.There is brown anthropic soil probably resultingfrom intensive cultiva- tion around some village sites. Integrated bluff/floodplainproduction systems today are indicative of what was possible. Bluff/flood- plain interaction is indicated by sixteenth- century locationalinformation,and by reports of roads leadingfrom bluffedges into the inte- rior,of landingplaces "down"on the river,of fruittrees being harvested in the interior,and of fish being traded inland. Also, there are manyexamples today of integratedbluff/flood- plainproduction systems. We do not know the nature of prehistoric socioeconomic complementarity between bluff and floodplain, nor do we know much more about nonmarket complementarity to- day. People may have had fields in both habi- tats, including satellite villages in one or the other. There may well have been a form of transhumance or shifting of residence from temporaryhuts on the playasto the bluffsdur- ing normal high water. People living in more permanent villageson high naturallevees may have had recourse to living temporarilywith kin on the bluffsduring exceptional floods.17 Thereprobablyalso was a seasonal shiftinem- phasisfrom vwrzea fishingduringlow water to bluff-zone hunting during high water, when fish availabilitydeclines. Therewas probablyan exchange of upland manioc for floodplainfish and of upland fruit, fish poison, drugs, and medicinals for floodplain products. This ex- change could have been by trade,via kinship, by control of multipleresource habitats,or all of these. In addition, there must have been exchange up and down river and from bluff edge to the interior,often involving consider- able distance and time. Inany event, it is not likely- that a floodplain community could manage over long periods without some means of access to upland resources. Hence, most floodplain villages were likely located where they had access by waterto the base of bluffs,a clusteringor patch patternparalleling that of bluffvillages. Cultivation in the vwrzea consisted of
  20. 20. 672 Denevan patches of plots indifferentmicrohabitatsvary- ing in lengthand durationof flooding and with soil fertilityand texture, with a zonation by elevation along the sides of levees and islands. Blufffields were a mosaic of permanent plots amidstmanagedforest, combined witha zona- tion based on distance from riverbank.Villag- ers relied on both bluff and va'rzea for both crops and wildlife.Subsistence strategyvaried depending on ecology, season, demography, and distance. Bluffcultivation and forest re- sources provided a safety valve when va'rzea fields were destroyed by high floods. These patterns can be observed along the Amazon today. V?rzea and bluff-edgevegetation was in largepartmanageablewith stone axes. Primary forests beyond the bluffedges were gradually modified and managed for several kilometers. Trekkingto the interiorfor plant and animal resource exploitationand for trade could have contributedto thisdiversifiedsubsistence strat- egy. Such an integrated subsistence system could conceivably have supported permanent settlements of several thousand people or more. Decline of BluffSefflement and Cultivation Bythe mid-seventeenth century there were reportsof villageson islandsand playas.Acuha (1942:35) in 1641 indicated that in Omagua territorythe river was full of islands, all of which were inhabited or cultivated, but that the Indians"areexposed to great loss, on ac- count of the powerful floods." Laureanode la Cruz (1885:86-88) in 1647 reported a village on the Islandof Piramotabelow the mouth of the Napo. Fritz (1922:50) in 1689-1691 said that Omagua houses and fields "aregenerally situated on islands, beaches or banks of the River;all low lying lands liableto be flooded," ratherthan on high land, even though their crops mightbe destroyed; this was because of custom or fear of forest Indians. Heriarte (1952:16), for the Tapajoregion in 1692, be- lieved that the river banks (bluffs?)were not inhabited because the Indianshad fled to es- cape the Portuguese. Myers (1992b:134) and Bolian(1975:15-16) suggest thatby the seven- teenth century much of Omagua bluff settle- ment had shifted to islandsfor defensive rea- sons. Thus bluff control by va'rzea Indians seems to have broken down early in some areas, probably because bluff settlement was unsafe given increased Europeanriveractivity. In 1616, a Jesuit mission was founded at Belem near the mouth of the Amazon, and others were soon established along the river. Hemming mapped 87 missions in 1759 along the centraland lower Amazon (Figure7). Pres- ton James in his text LatinAmerica noted the bluff location of these missions: "Upstream, wherever the river in its meandering course swept againstthe base of the valley bluffand so provided high ground next to the navigable river channel, mission stations were built" (James 1969:840). Other missions, such as Sarayacu(Franciscan)on the Ucayali,were lo- cated within but at the edge of the floodplain and crops were planted both in the floodplain and on the adjacent terra firme (Myers 1990:57). While many of the missions were abandoned as their Indian populations de- clined, others became riverports and rubber collection centers-the antecedents of present rivertowns and cities. By1850 there were very few riverineIndians survivingalongthe Amazon and itsmainlower tributariesas a resultof disease, slaving,fleeing to the interiorforests, and detribalization.De- population and tribalextinction were propor- tionally much greater than in the interfluves and along lesser tributaries.The rubberboom brought new people into riverine Amazonia, and new bluff settlements were established, but forest extractionwas more importantthan agriculture.With the decline of rubber and other extractiveproducts, subsistence agricul- turewas slowly reestablishedby the remaining people-detribalized Indiansfrom the interior, and settlers,allnow referredto as riberehosin the Upper Amazon and as ribeirinhosor cabo- clos in Brazil(Parker1985). They utilize both varzeas and bluffs, and unless markets are close by (a day or so by water), they are still dependent on bluff!va'rzeacomplementarity. Most of the larger,permanentsettlementscon- tinue to be on bluffs(Figure4). "Thereare no cities on the Amazon floodplain proper" (Gouldinget al. 1996:14). Demographic Implications If the model of bluff settlement presented here is basicallycorrect, what are the conse- quences for estimatinglate prehistoricriverine
  21. 21. A BluffModel of Settlement 673 t Missions 66? 6O+ 54? 48 A Mission/Fort x Fort Locationsof ColonialMissionsOCity 0A L N I AmapcA - N ; | Uaup 4< -s OCEANti:E:: Macapa t ~~~~~ ~ ~~~~~~Monte t Alegrem if iMrafro is. or JaurvapObidos ab on Blufs A 9io Manaus t at Sntar6m O~vena Tefet %?O Kilometers9~~O~~~O ~~~~0t 0~~ 50 100 66o 610o 5140 8 Figure 7. Central and lower Amazon showing locations of 75 missions and mission/forts in 1759. All but four or five appear to be on bluffs. Source: adapted from Hemming 1978:xxii-xxiii. populations in Amazonia? In three earlierat- tempts to estimate Amazon populations near the time of initialEuropeancontact, Iobtained samples of known densities for smallareas of floodplain (5.3, 14.6, and 13.9 per kiM2)and projected these to all floodplains to get esti- mated total populations (Denevan 1970:67; 1976:218; 1992a:xxvi). The same was done for other majorhabitattypes to obtaintotal popu- lationsfor GreaterAmazonia (Table1). Forthe floodplains, I now reject this entire methodol- ogy, given the argument here that settlement was not evenly dispersed in the floodplains, but ratherwas mainlyconcentrated in clusters on bluffsegments that impinged againstactive channels of the Amazon and its majortribu- taries. We do not know how many such vil- lages there were or how largethey were, ex- cept for a few terra preta sites that have been measured. Hence, it is impossible to estimate prehistoricriverinepopulationdensities ortotal populations. As an exercise, however, I used available population densities and sustaining areas for contemporary Indian and peasant villages to obtain a potential aboriginal population (Denevan n.d.). Sustainingareas (agriculture, plant foraging, hunting, fishing) average a ra- dius of 10 km for huntingand 5 km for fishing (usuallymuch less for agriculture).Fora bluff village, the average sustainingarea would be 196 km2.Forthe Amazon bluffvillageof Tam- shiyacu in Peruwith a populationof 2,040 (Hi- raoka 1986:357), the population density for 196 km2would be 10.4 per km2.Assumingthat an estimated20 percent of the bluffterrainwas on navigablechannels (based on an examina- tion of fluvialmaps), a 15 km wide sustaining area extending 10 km inlandfrom bluffedges and 5 km into the vwrzeas, each side of the river,and a density of 10 persons per kM2, the total riverinepopulationfor 20 percent of the Amazon Riverand 21 majortributaries(esti- mated sectors with floodplains) in Greater Amazonia would be 1,464,000. The restof the riverine zone was not devoid of people, al- though settlement was mostly sparse and un- stable. Average density is unknowable, but if Meggers' overall Amazon density of only 0.3 per km2 is used for the remaining80 percent of the riverinesustainingarea, this would add 176,000 people fora riverinetotalof 1,640,000. Table1 provides previous estimates of habi- tat densities, overalldensities, and total popu- lationsfor GreaterAmazonia and forthe Ama- zon Basinby myselfand others. Previousriver- ine density estimates range from 0.3 to 14.6 persons per kiM2,but are for floodplains only
  22. 22. 674 Denevan and do not allow for deep hunting and agro- forestry sustaining areas on terra firme. Using a riverine total of 1,640,000, adjusting floodplain and riverine sizes, and retaining pre- vious habitat estimates yields a new estimate for Greater Amazonia population of 5,487,000, a reduction of 177,000 from my 1992 estimate (Table 1), a relatively minor amount. Although the assumptions are tenuous and the densities of 10 and 0.3 persons per km2 are feasible but conservative, working numbers at best, the new perspective based on the bluff model raises new questions on Amazonian demogra- phy and settlement. As I have indicated, I now have serious res- ervations about all earlier estimates based on habitat densities. Thus Iam reluctant to suggest a new total population. However, based on 20 years of considering the question of Amazon Indian numbers as of 1492, new historical and archaeological research, and Newson's (1996; see Table 1) recent estimates resulting from a careful examination of ethnohistorical informa- tion from the Ecuadorian Amazon, I believe that reasonable ranges of estimates would fall between 5.5 and 7.5 million for Greater Ama- zon and between 3.5 and 5.0 million for the Amazon Basin, numbers well above those of Steward, Hemming, and Meggers (Table 1). Conclusions The concept that prehistoric riverine food production involved the integration of bluff and floodplain cultivation has been presented previously by several scholars, mainly inferred from the same forms of evidence examined here. One of the first was the archaeologist Peter Paul Hilbert (1957:2-3). According to Lathrap (1970:44): The tropicalforestfarmerlivingon the bluffof old alluviumadjacentto the active flood plain could simultaneouslyfarmthe limitedbut excellent re- cent alluvialsoils in an intensive and continuous mannerandthe poor butessentiallyunlimitedsoils of the old alluvialdeposits using slash-and-burn agriculture. Meggers (1991:199) comments that most known archaeological sites are "riverine in lo- cation, but the sustaining area is primarily terra firme" (also see Meggers 1984:632; 1993- 1995:106). Myers (1990:30) points out for the Shipibo that "avillage accessible to both kinds of terrainwould be most advantageouslysitu- ated. . . . Furthermore, a village with gardens on interfluvialland would be assured of pro- duce throughoutthe flood season." And Eden (1990:79) states that: "Adjacenttracts of the terra firme were also used for cultivation, pro- viding some insuranceagainstthe flood riskof the virzea." Carneiro (1995:57-58) empha- sizes that "Vdrzeacould never have been the only land reserved for growing crops. Some reliance must also have been placed on terra firme.... In years of excessive flooding, then, terra firme would have served as a kind of 'crop insurance' . .. unquestionably this [high bluff]must have been a choice site for an In- dian village." Finally, geographer N. Smith (1995:228) statesthat"[m]anyvillageswere es- tablished along the upland bluff overlooking floodplains so that the inhabitantscould take advantage of animaland plant resources from both vdrzea and upland environments" (also see Goulding et al. 1996:24). What we do not know is the degree to which individualhouse- holds utilized both habitats with multiple dwellings,or reliedupon specializationand ex- change. Thusprehistoricbluff!wvrzea complementar- ityforriverineAmazoniawas suggested earlier, but the implicationswere not adequately ac- knowledged in archaeological research on settlement patterns, demography, and subsis- tence. Ratherthan rationalizingthatmost river- ine sites were floodplain and thus destroyed, archaeologists should focus on bluff sites, which were primaryand which have survived to a considerable extent. The evidence is good that large settlements existed on the bluffs, contraryto Meggers' (1992:203) assertionthat "[T]he conclusion that early eyewitness ac- counts exaggerate the indigenous population density seems inescapable."18 However, the distributionof largebluffvillageswas sporadic ratherthan continuous, which has not been pointed out previously.They were separated by smaller,unstable settlements in the flood- plains and by sparse settlement in the bluff zones isolated from river channels. Accord- ingly,population densities alternatedbetween relatively dense (10 per km2 or more) and sparse (0.3 per km2 or less) as compared to Meggers'overallriverinedensity of 0.3. Perma- nent bluffsettlementwas made possible by the integrationof house gardens, intensive fields, fruitorchards, and managed secondary forest
  23. 23. A BluffModel of Settlement 675 in combination with seasonal cultivation of floodplain soils. Variabilityand flexibilitypre- vailed.There isevidence to supportthispatchy bluff/lvrzea complementarity model, but fur- ther research is needed. The model also may be relevantto other prehistoricbluff/floodplain contexts, such as the MississippiRiverwhere there are numerous settlement sites on both levees within floodplains and on adjacent up- lands.19 Amazonianist ethnohistorianAntonio Porro (1994:86, 91) has "no doubt about the demo- graphic density of the floodplain as a whole and the great size of many villages";and he believes the early "sources suggest that the [fluvial]provinces were territoriallydefined and socially stratified,"with "centralized political power" (i.e., chiefdoms). The bluffmodel pre- sented here should help substantiatethis river- ine argumentmade by Porro,as well as earlier by Lathrap,Carneiro, Myers, Roosevelt, and others. The bluff model, in part at least, ex- plainsconcentrated riverinesettlement; other- wise populationslikelywould have been thinly dispersed over thousands of kilometers of floodplain. Finally,the bluffmodel underminesthe con- ventionaldichotomy between virzea and terra firme settlement and agriculturebased on soil constraintsexpressed by manyAmazon schol- ars.Ifbluffsoils could be made to help support relativelylarge numbers of people, so could other terra firme soils. Interfluve terra preta sites confirm this. Thatsuch sites are sporadic was the result of other factors, including the use of inefficientstone axes, the limitedavail- abilityof animalprotein, lack of demographic pressure, and villagefissioning. The bluffmodel, if valid, has significantim- plicationsfor contemporaryAmazonian devel- opment. Uniformsettlement and cultivationof the majorriverinezones is unlikely,given that the floodplainsare subjectto periodicdestruc- tive floods (unless farmers are protected by creditor insurance),given the limitedextent of unflooded bluffs that are adjacent to active year-roundnavigableriverchannels, and given distances to markets.20The patch patternthat existed prehistoricallyexists today, and is likely to persistinthe future.Second, terrafirmesoils can be cultivated permanently by traditional methods. Tobe sure, high laborcosts, polycul- turalcropping, a diversity of food production systems, and forest management and extrac- tion, while appropriate for subsistence, may not be feasible for a strongly market-oriented economy. On the other hand, a complemen- tary bluff/floodplain mosaic system of land use can maintain some form of forest, ensure reli- able security, and support a modest population density, such as apparently existed in the in- digenous past. Acknowledgments I have relied heavilyon the researchand ideas of archaeologistsDonaldLathrap,ThomasMyers, Betty Meggers, and Anna Roosevelt. Draftsprofitedfrom critical readings by Oliver Coomes, Kent Mathewson, and ClarkErickson.Researchand writ- ingwere funded, inpart,by the NationalEndowment for the Humanitiesand the Wisconsin Alumni Re- searchFoundation.Thefigureswere preparedby the University of Wisconsin CartographicLaboratory. This paper was presented as my retirementlecture at the Universityof Wisconsin-Madison,December 3, 1993, and as the 13th Carl 0. Sauer Memorial Lecturedelivered at the Universityof Californiaat Berkeley,October 11, 1994. Ithank the geography departmentsat Madison,my academic home for 30 years, and at Berkeley,my almamater,for theirrec- ognitionand their hospitalityon these occasions. Notes 1. The "evidence from archaeologyand ethnohis- tory made it clear that dense populationsand advanced cultureshaddeveloped inAmazonian floodplain habitats in late-prehistoric times" (Roosevelt 1980:252). Later,however, Roosevelt (1987:154-155) statedthatlate prehistoricchief- doms were "settledalongthe banks,levees, and deltasof the majorfloodplains,"which is stillam- biguous as to precise location. She does not define "floodplain." 2. "Atmaximumriverstage, the entirefloodplainis covered by a continuous sheet of water"(Sippel et al. 1994:73). However, there are some high levees, ancientfloodplainsurfaces,and riverter- races that never seem to flood and can thus be considered to be within the valley but not part of the active floodplain. 3. Thisisa revisionof Steward's(1949:659)regional density estimatesof between 0.10 and 0.40/km2 for the nonriverineAmazon forests. 4. Population densities listed by Beckerman (1987:86) for 13 contemporary Amazon terra firmetribes average 0.39 per kmi2,with a range from 0.01 to 1.00. 5. InPeru,a distinctionis made between emergent sandbars(playas)and mudflats(barreales).Here, playa(beach) refersto both. 6. For example, "un pueblo que estaba en una loma," and "sobre un alto una hermosa poblaci6n"(Carvajal1894:54).
  24. 24. 676 Denevan 7. A sixteenth-century Spanish league (legua) measured about 3 2/3 mi or 5.9 km (Medina 1934:47); however, the length of a league varied somewhat in time and place, often 5 km being given. The accuracy of measurement during early Amazon Rivertravel is unknown; some dis- tances reported are clearly excessive. 8. Communal houses containing over 100 people have been observed in the twentieth century (Myers 1992b:138). 9. One seventeenth-century account reported Omagua villages perpendicular to the rivers, pos- sibly to focus on a prime river landing (Cruz 1885; in Myers 1992b:133). However, I know of no reported archaeological sites so oriented. 10. "Hab(a muchos caminos que entraban la tierra adentro muy reales" (Carvajal 1894:42, also see pp. 45, 46, 53, 67). "Highways" were not likely, but neither is it likely that these were simple trails, which the Spaniards were quite familiar with. 11. For Araracuara, Eden et al. (1984:134) found the terra preta soils to be higher in phosphorus and organic carbon than adjacent soils, but similar in acidity, calcium, and exchangeable aluminum. 12. Farmers today even transport black earths con- siderable distances by truck to fertilize their fields and gardens (Smith 1980:562). 13. Agroforestry systems with annuals and fruittrees on high ground within floodplains have been described in Peru by Padoch and de Jong (1987:190-192), Coomes (1995a), and Hiraoka (1989:81-84). On floodplain forest management for varied products on islands in the lower Ama- zon, see Anderson et al. 1995. 14. Lathrap was convinced that "shifting slash-and- burn agriculture was a secondary, derived, and late phenomenon within the Amazon Basin"and that both floodplain and upland farmers had in- tensive garden systems involving "a bewildering arsenal of crops and cycles of rotation extending up to 25 or 50 years" (Lathrapet al. 1985:54-55). 15. Tree-fall gaps comprise about 1 percent of the forest area of Amazonia at any time, averaging several hundred square meters each (Ruoko- lainen and Tuomisto 1993:141-142). 16. Acufia 1942:37, Carvajal 1934a:210, 217; 1934b:415, 426; Cruz 1885:188; Heriarte 1952:17; Ortiguera 1934:317; Salinas 1965:199, 200, 202, 205, 206, 207; Vdzquez 1981:211; Vdzquez de Espinoza 1948:384; Zuniga 1981:15. In the Iquitos region today, there are at least 193 species of useful fruits, of which 74 are cultivated (domesticates and semi-domesticates); see Vasquez and Gentry (1989:352-356). 17. Myers (1990:34) suggests that multiple habitat access and use may have been associated with economic security based on extended families rather than on the family or village. 18. This is a reversal of Meggers' earlier position that: "Inspite of their deficiencies, however, the early chronicles make it clear that population density and level of cultural development were consid- erably greater on the vdrzea than on the terra firme at the time of European contact" (Meggers 1971:1 22). 19. For the American Bottom region of the Missis- sippi, Woods and Holley (1991:50) indicate that "[I]tis very likely that [prehistoric] upland settle- ments were exploiting both upland and bottom- land environments." 20. For a discussion of the general problems of ag- riculturaldevelopment of the Amazon floodplain in Peru, see Chibnik (1994:221-223). Goulding et al. (1996:165, 166) indicate that the govern- ments of Amazonian countries are beginning to shift development efforts from the uplands to the floodplains, where serious environmental prob- lems are now occurring as a result. References Acuha, Crist6bal de. 1942 [1641]. Nuevo descu- brimiento del gran r(o de las Amazonas. Buenos Aires: Emec6 Editores. Allen, William L., and Holshouser de Tiz6n, Judy. 1973. Land Use Patterns among the Campa of the Alto Pachitea, Peru. In Variation in Anthro- pology, ed. Donald W. Lathrapand Jody Douglas, pp. 137-153. Urbana: IllinoisArchaeological Sur- vey. Anderson, Anthony B., Magee, Pennie, G6ly, Anne, and Jardim, Mario Augusto Goncalves. 1995. Forest Management Patterns in the Floodplain of the Amazon Estuary.Conservation Biology 9:47- 61. Andrade, Angela. 1986. Investigacibnarqueol6gica de los antrosoles de Araracuara. Bogota: Banco de la Republica. Bal6e, William. 1989. The Culture of Amazonian Forests. Advances in Economic Botany 7:1-21. Beckerman, Stephen. 1987. Swidden in Amazonia and the Amazon Rim. In Comparative Farming Systems, ed. B. L.Turner IIand Steven B. Brush, pp. 55-94. New York:Guilford Press. Bergman, Roland W. 1980. Amazon Economics: The Simplicity of Shipibo Indian Wealth. Dell- plain LatinAmerican Studies, No. 6. Ann Arbor, Michigan: University Microfilms International. Bolian, Charles E. 1975. Archaeological Excavations in the Trapecio of Amazonas: The Polychrome Tradition. Unpublished Ph.D. dissertation, De- partment of Anthropology, University of Illinois, Champaign-Urbana. Carneiro, Robert L. 1970. The Transitionfrom Hunt- ing to Horticulture in the Amazon Basin. Pro- ceedings, Eighth International Congress of An- thropological and Ethnological Sciences, Tokyo and Kyoto, 1968 3:244-248. Tokyo. 1995. The History of Ecological Interpreta- tions of Amazonia: Does Roosevelt Have ItRight?
  25. 25. A BluffModel of Settlement 677 In Indigenous Peoples and the Futureof Ama- zonia:An EcologicalAnthropologyof an Endan- gered World, ed. Leslie E. Sponsel, pp. 45-70. Tucson: University of Arizona Press. Carvajal, Gaspar de. 1894 [1542]. Descubrimiento del R(o de Orellana. In Descubrimiento del Rio de las Amazonas, compiled by Jos6 Toribio Med- ina, pp. 1-83. Sevilla: E. Rasco. * 1934a [1542]. Carvajal'sAccount. In The Dis- covery of theAmazon Accordingto theAccount of FriarGaspar de Carvajaland Other Docu- ments, compiled by Jose Toribio Medina, ed. H. C. Heaton, pp. 167-242. Special Publication No. 17. New York:American Geographical Soci- ety. . 1934b [1542]. The Version of Carvajal's Ac- count in Oviedo's "Historia."In The Discovery of the Amazon According to the Account of Friar Gasparde Carvajaland OtherDocuments, com- piled by Jose Toribio Medina, ed. H. C. Heaton, pp. 405-448. Special Publication No. 17. New York:American Geographical Society. Cavelier de Ferrero, Ines, Mora Camargo, Santiago, and Herrera de Turbay, Luisa Fernanda. 1990. Estabilidad y dindmica agricola: Las transforma- ciones de una sociedad agr(cola. In Ingenierias prehispinicas, ed. S. Mora Camargo, pp. 73- 109. Bogota: Fondo FEN. Chibnik, Michael. 1994. Risky Rivers: The Econom- ics and Politicsof FloodplainFarmingin Ama- zonia. Tucson: University of Arizona Press. Coomes, Oliver T. 1992. Making a Living in the Amazon Rain Forest: Peasants, Land, and Econ- omy in the Tahuayo River Basin of Northeastern Peru. Unpublished Ph.D. dissertation, Depart- ment of Geography, University of Wisconsin- Madison. .1 995a. Indigenous Agroforestry and Relic Riverine Features in the Amazon River Lowlands: The Case of the Yarinales of Esperanza, NE Peru. Paper presented at the Annual Meetings of the Association of American Geographers, Chicago. . 1995b. Personal communication. Montreal. December 18. Cruz, Laureano de la. 1885 [1653]. Nuevo descu- brimiento del Maraion, aio 1651. In Varones ilustresde laOrdenSeraficaen el Ecuadordesde la fundacidnde Quito hasta nuestros dias, ed. Francisco Maria Compte, 1:148-205. Quito: Im- prenta del Clero. DeBoer, Warren R. 1981. Buffer Zones in the Cul- tural Ecology of Aboriginal Amazonia: An Ethno- historical Approach. American Antiquity 46:364- 377. Denevan, William M. 1966a. The Aboriginal Cultural Geography of the Llanosde Mojos of Bolivia. Ibero Americana 48. Berkeley: University of Cali- fornia Press. . 1966b. A Cultural-Ecological View of the For- mer Aboriginal Settlement in the Amazon Basin. Professional Geographer 18:346-351. . 1970. The Aboriginal Population of Western Amazonia in Relation to Habitatand Subsistence. Revista Geogrdfica (Rio de Janeiro) 72:61-86. . 1976. The Aboriginal Population of Ama- zonia. In The Native Population of the Americas in 1492, ed. W. M. Denevan, pp. 205-234. Madison: University of Wisconsin Press. . 1984. Ecological Heterogeneity and Horizon- tal Zonation of Agriculture in the Amazon Flood- plain. In Frontier Expansion in Amazonia, ed. Marianne Schmink and Charles H. Wood, pp. 311-336. Gainesville: University of Florida Press. . 1992a. Native American Populations in 1492: Recent Research and a Revised Hemispheric Es- timate. In The Native Population of the Americas in 1492, 2nd ed., ed. W. M. Denevan, pp. xvii- xxxviii. Madison: University of Wisconsin Press. . 1992b. Stone vs Metal Axes: The Ambiguity of Shifting Cultivation in Prehistoric Amazonia. Journal of the Steward Anthropological Society 20:153-1 65. . n.d. The Native Population of Amazonia in 1492 Reconsidered. Manuscript. Denevan, William M., and Padoch, Christine, eds. 1988. Swidden-Fallow Agroforestry in the Peru- vian Amazon. Advances in Economic Botany, Vol. 5. Dumont, Jean Francois, Lamotte, Sandrine, and Kahn, Francis. 1990. Wetland and Upland Forest Eco- systems in Peruvian Amazonia: Plant Species Diversity in the Light of Some Geological and Botanical Evidence. Forest Ecology and Manage- ment 33-34:125-1 39. Eden, Michael J. 1990. Ecology and Land Manage- ment in Amazonia. London: Belhaven Press. Eden, Michael J., Bray, Warwick, Herrera, Leonor, and McEwan, Colin. 1984. TerraPreta Soils and their Archaeological Context in the Caquetd Ba- sin of Southeast Colombia. American Antiquity 49:125-1 40. Fritz, Samuel. 1922 [1723]. Journal of the Travelsand Labours of Father Samuel Fritz in the River of the Amazons between 1686 and 1723, ed. George Edmundson. Hakluyt Society, Second Series, No. 51. London: Cambridge University Press. Goulding, Michael, Smith, Nigel J. H., and Mahar, Dennis J. 1996. Floods of Fortune: Ecology and Economy along the Amazon. New York:Colum- bia University Press. Harris, David R. 1971. The Ecology of Swidden Cul- tivation in the Upper Orinoco Rain Forest, Vene- zuela. Geographical Review 61:475-495. Heaton, H. C. 1934. Editor's Foreword. In The Dis- covery of the Amazon According to the Account
  26. 26. 678 Denevan of Friar Gaspar de Carvajal and Other Docu- ments, compiled by Jose Toribio Medina, ed. H. C. Heaton, pp. 383-389. Special Publication No. 17. New York:American Geographical Soci- ety. Hecht, Susanna B., and Posey, Darrell A. 1989. Pre- liminary Results on Soil Management Techniques of the Kayap6 Indians. Advances in Economic Botany 7:174-188. Hemming, John. 1978. Red Gold: The Conquest of the Brazilian Indians. Cambridge, Massachusetts: Harvard University Press. Heriarte, Mauricio de. 1952 [1692]. The Province of the Tapaj6s. Trans. and ed. John H. Rowe. Kroe- ber Anthropological Society Papers 6:16-18. Herrera, Luisa Fernanda, Cavelier, Ines, Rodr(guez, Camilio, and Mora, Santiago. 1992. The Techni- cal Transformation of an Agricultural System in the Colombian Amazon. World Archaeology 24:98-113. Hilbert, Peter Paul. 1957. Contribuicao a arqueologia do Amapd. Boletim do Museu Paraense Emilio Goeldi: Antropologia 1:1-41. Hiraoka, Mario. 1985. Floodplain Farming in the Pe- ruvian Amazon. Geographical Review of Japan 58 (Ser. B):1-23. . 1986. Zonation of Mestizo Riverine Farming Systems in Northeast Peru. National Geographic Research 2:354-371. . 1989. Agricultural Systems on the Floodplains of the Peruvian Amazon. In FragileLands of Latin America: Strategies for Sustainable Develop- ment, ed. John 0. Browder, pp. 75-101. Boul- der, Colorado: Westview Press. Irvine, Dominique. 1989. Succession Management and Resource Distribution in an Amazonian Rain Forest. Advances in Economic Botany 7:223-237. James, Preston E. 1969. Latin America. New York: Odyssey Press. Katzer, F. 1994. A terra preta. Boletim da Seccdo de Fomento Agricola no Estado do Pari 3(2):35-38. Lathrap, Donald W. 1968a. Aboriginal Occupation and Changes in River Channel on the Central Ucayali, Peru. American Antiquity 33:62-79. . 1968b. The Hunting Economies of the Tropi- cal Forest Zone of South America: An Attempt at Historical Perspective. In Man the Hunter, ed. Richard B. Lee and Irven DeVore, pp. 23-29. Chicago: Aldine. . 1970. The Upper Amazon. New York: Praeger. Lathrap, Donald W., Gebhart-Sayer, Angelika, and Mester, Ann M. 1985. The Roots of the Shipibo Art Style: Three Waves on Imirfacocha or there were "Incas" before the Incas. Journal of Latin American Lore 11:31-119. McCann, Joseph M. 1994. Terra Preta do Indio in Lower Amazonia: Fertile Ground for Caboclo Agriculture and Theoretical Debates in Cultural Ecology. Paper presented at the International Meeting of the Conference of LatinAmericanist Geographers, Ciudad Judrez, Mexico. Medina, Jose Toribio. 1934 [1894]. Introduction by Medina. In The Discovery of the Amazon Ac- cording to the Account of FriarGasparde Car- vajaland Other Documents, compiled by Jose Toribio Medina, ed. H. C. Heaton, pp. 7-163. Special Publication No. 17. New York:American Geographical Society. Meggers, Betty J. 1971. Amazonia:Manand Culture in a CounterfeitParadise.Chicago: Aldine. . 1984. The Indigenous Peoples of Amazonia, their Cultures, Land Use Patterns and Effects on the Landscape and Biota. In The Amazon: Lim- nology and Landscape Ecology of a Mighty TropicalRiver and its Basin, ed. Harold Sioli, pp. 627-648. Dordrecht: W. Junk. . 1991. Cultural Evolution in Amazonia. In Profilesin CulturalEvolution,ed. A. Terry Rambo and Kathleen Gillogly, pp. 191-216. Anthropo- logical Papers No. 85, Museum of Anthropology. Ann Arbor: University of Michigan. . 1992. Prehistoric Population Density in the Amazon Basin. In Disease and Demography in the Americas,ed. John W. Verano and Douglas H. Ubelaker, pp. 197-205. Washington, D.C.: Smithsonian Institution Press. . 1993-1995. Amazonia on the Eve of Euro- pean Contact: Ethnohistorical, Ecological, and Anthropological Perspectives. Revista de Ar- queologiaAmericana8:91-115. . 1994. Prehistoric Cultural Development in Amazonia: An Archeological Perspective. Re- search and Exploration10:398-421. . 1995. Judging the Future by the Past: The Im- pact of Environmental Instability on Prehistoric Amazonian Populations. In Indigenous Peoples and the Futureof Amazonia:An EcologicalAn- thropology of an EndangeredWorld,ed. Leslie E.Sponsel, pp. 15-43. Tucson: University of Ari- zona Press. Meggers, Betty J., Dias, 0. F, Miller, E.T., and Perota, C. 1988. Implications of Archeological Distribu- tions in Amazonia. In Proceedings of a Work- shop on Neotropical DistributionPatterns,ed. W. R. Heyer and P. E. Vanzolini, pp. 275-294. Rio de Janeiro: Academia Brasileirade Ciencias. Minta, Stephen. 1994. Aguirre:The Re-Creationof a Sixteenth-CenturyJourneyacrossSouthAmer- ica. New York: Henry Holt. Mora C., Santiago, Herrera, Luisa Fernanda, Cavelier F., Ines, and Rodrfguez, Camilio. 1991. Culti- vars,AnthropicSoilsand Stability:A Preliminary Reportof ArchaeologicalResearchinAraracuara, Colombian Amazonia. University of Pittsburgh Latin American Archaeology Reports, No. 2. Pittsburgh. Myers, Thomas P. 1973. Toward the Reconstruction