The Scale of Tawantinsuyu, Part I: Sawsa
The Scale of Tawantinsuyu, Part I: Sawsa
1. Introduction
This is the first part of a hopefully 5-part piece on the prehistoric demography of Tawantinsuyu, the empire of the Inkas. As the largest and most complex* of the indigenous civilizations of the Americas, Tawantinsuyu occupies an important place in the ongoing debate on pre-Columbian demography, which I’ve touched on at least once previously. However, as you can probably tell from the ambitious number of parts I’ve divided this into, we’re dealing with a topic here that’s much broader than the previous piece on the population of Brazil’s Marajoara culture. Marajó Island, for all its fascinating historical and archaeological minutiae, was just a small corner of a very large Pre-Columbian Amazonian world. Tawantinsuyu under the Inka was an entire world unto itself.
*asterisk
Some perspective: at the eve of European contact, Inka control stretched from the Mira River in Colombia to the Maule River in Chile, a north to south distance of some 4,000 kilometers. The polities incorporated into Tawantinsuyu between its founding and the Spanish conquest numbered in the hundreds, from small-scale chiefdoms, like the Chachapoyas of the Andean cloud forests, to organized kingdoms, like the Aymara of the Altiplano, to at least one expansionist, bureaucratic state that had once functioned as a peer competitor to the Inka: the coastal empire of Chimor.
All this is to say that the human geography of Tawantinsuyu was bewilderingly diverse, encompassing a broad range of societies with varying levels of population density, political organization, and social & economic complexity. There were a couple of factors that unified the whole Andean region: subjection to the traditional labor tax, called (in Qusqu Qhichwa, the imperial lingua franca) the mit’a; maize and potato-based agriculture, supplemented with llama and alpaca pastoralism; and subsistence strategies wherein communities relied on the produce of many different ecological zones located at different altitudes, termed the “vertical archipelago” by John Murra (Murra 1968). But by and large, the pre-Columbian Andes were too diverse to be treated as a single unit in this kind of investigation. Demographic conclusions drawn from examination of, say, the Peruvian coast will be totally inapplicable to societies of the Peruvian sierra or the high Altiplano, and vice-a-versa. If we want to even begin to approach the question of how many inhabitants Tawantinsuyu was home to, we’re going to need to take a closer look at the individual components of the empire.
Here in Part I, we’re going to be starting off with exactly that sort of zoomed-in, ground-level approach. We’re going to be looking at a single wamani (province), and using documentary, archaeological, demographic, & agricultural evidence, we’re going to determine a range of probable population sizes for the province at the height of Inka power around 1520. Consider it a tasty little apéritif, before we gorge ourselves on the sumptuous main course to come.
Waiter, waiterǃ
2. The Mantaro Valley
Zoom in a little on that map of Tawantinsuyu. Ignore the fact that it gets the boundaries of the four quarters of the empire wrong, can’t seem to decide whether it’s writing in Qusqu Qhichwa or not, and features a number of locations that weren’t established until as late as 50 years after the Spanish conquest. Shoot for the bits between “Shawsha” (Hatun Sawsa, in the imperial lingua franca) and “Wankayu” (founded in the 1570s as a reducción after disease wiped out the majority of the area’s native population). Pause while I stop being a huge bitch about these entirely understandable mistakes. Ok, got it? Here, let me enhance that for you.
This is the Mantaro Valley, home to one of the Andes’ largest expanses of relatively flat, arable land. It will also serve as the setting for today’s little project. The wide, flat valley floor ranges from 3,350 meters above sea level in the north to about 3,100 in the south (Mayer 1980:14). Agriculture is possible on both the valley floor and along the mountainsides up to about 4,200 masl, although yields for all staple crops diminish rapidly with altitude.
I’ve decided to focus in on the Mantaro Valley for a couple of reasons:
It’s a fairly small region with clearly delineated boundaries
Early demographic data is available for the valley, including census records from the early Spanish period and probably also the Inka period
The area has been covered admirably by recent, in-depth archaeological surveys, primarily the Junín Archaeological Research Project and the Upper Mantaro Archaeological Research Project
There is some great modern work on land use patterns in the Valley, which will form the basis for a lot of the upcoming analysis of agricultural carrying capacity in Inka times
We are interested in the population of Tawantinsuyu as a whole, and we will absolutely be getting to that in later parts of this series. For now, consider the Mantaro Valley a case study: what we’re determining here is the kinds of densities that pre-Columbian Andean societies were able to achieve under ideal conditions. The numbers we arrive at won’t be extrapolated to the entire Inka domain, but they can absolutely be informative if we treat them as a single small piece of this very large puzzle.
First, a brief word on naming conventions. The region in question was inhabited in late prehistoric times by two Qichwa-speaking ethnic groups, the Shawsha of the northern Mantaro Valley and the Wanka of the south. When the Inka conquered the area in the 15th century, they organized these two ethnic groups into a single wamani (province), which they called (according to records from the 1580s published in the 19th century by Spanish historian Marcos Jiménez de Espada) Wanka. When Pizarro’s Spanish conquered the region in 1534, they kept both the province and its internal divisions intact. But, because nothing can ever be easy, they decided to switch up the names. The province remained divided into Hatun Xauxa, Hurin (“lower”) Guanca, and Hanan (“upper”) Guanca, but it was always listed in colonial records as Xauxa. That name has stuck, living on today as the modern Peruvian city and province of Jauja.
So not only do we need to choose between two competing names, but also between four distinct orthographies that produce wildly differing results. True to form, I’ve decided to opt for the one solution that will please no one. In this piece I’ll be referring to the province as Sawsa, using the Cuzco Qhichwa spelling of the name that wasn’t applied to the place until the arrival of the Spanish. Why? None of your business !!!
The archaeological record of the Mantaro Valley stretches back into the dimmest days of prehistory, prior to the spread of agriculture across the Andes. For our purposes, though, beginning at the real beginning is hardly useful -- we don’t want to spill too much ink on the dynamics of the hunter-gatherer and early farming societies of Sawsa’s Preceramic and Initial periods. In this brief intro, we’ll be focusing on the period of prehistory that has the best coverage in the archaeological record -- and, incidentally, leads directly into the period we’re actually concerned about, the brief stretch of Inka domination of the valley in the 15th and 16th centuries.
So we’ll gloss over thousands of years of agricultural settlement in the Mantaro Valley, and even the initial establishment of the Wanka and Shawsha civilizations. We’ll open with the human landscape of the Mantaro in the Late Intermediate Period, the era of prehistory immediately preceding the establishment of the Inkas’ empire. The Wanka II period, as it’s termed in archaeological investigations. This time was largely defined by the concentration of the population into a number of large, well-fortified centers, located in easily-defensible mountaintop locations. It’s likely that this era saw the formation of nascent states on the local level, for purposes of coordinating defense and urban planning.
The conquest of Sawsa by the Inka and its incorporation into Tawantinsuyu fundamentally changed life for the area’s inhabitants. Most of the large, fortified hilltop centers that had characterized the Wanka II phase, like the massive Tunanmarca (7,955-13,259 inhabitants; D’Altroy 2001:68) and Umpamalca (3,889-6,482 inhabitants; id:71) were completely abandoned, and the rest were severely reduced in population (e.g. from 6,633-11,055 to 2,466-4,110 in Hatunmarca). New settlements established during Inka times, like Chucchus and Marca, were smaller in size, larger in area, and tended to be located along the valley floor or shorter, broader upland ridges.
The massive difference between pre-Inka and Inka-era settlement patterns is pretty striking when you start comparing site maps:
Why?
Well, larger settlements located in naturally defensible locations are correlated pretty strongly with high rates of intercommunity violence, which is well-attested for the Wanka II period. So it’s possible that the sort of Pax Incaica brought on by the region’s incorporation into Tawantinsuyu made this settlement pattern obsolete, prompting people to quite literally move on to greener pastures at the valley floor. This tracks pretty well with the situation we see in a settlement like Hatunmarca, where the population declined by about 60% but the site remained inhabited under Inka rule.
But, reminder: Hatunmarca is an aberration. Most of these hilltop sites were completely abandoned after the Inka conquest, with no sign of a gradual filtering down of people to the hinterlands. The archaeological disconnect is incredibly sharp; only in Hatunmarca do we see Inka-style ceramics and architecture intruding on a settlement built prior to the conquest. By the looks of things, this change was extremely sudden, and it wasn’t voluntary on the part of the Wanka and Shawsha people. Rather, their new Inka overlords appear to have forcibly resettled them from their hilltop strongholds to the valley floor below. We can see this manifest in the archaeological record by comparing the frequency of settlements of certain populations between the pre-Inka and post-Inka eras:
Again: why?
The first and most obvious goal of this resettlement program would have been to discourage rebellions, both the overt and the subtle kind. A population inhabiting a fortified, naturally defensible settlement has significantly more leverage over its rulers than an equally large population dispersed over a broad, flat plain, simply because that overlord will need to bring more force to bear against them in the event of a rebellion. By removing the population from their fortified settlements, the Inka significantly reduced the threat that a potential rebellion in Sawsa could pose to their rule.
The second goal appears to have been economic. By moving populations out of their prior concentrations on the fortified hilltops, the Inka were able to resettle huge numbers of people across the fertile, naturally irrigated farmland on the valley floor. This would have allowed more land to be brought under cultivation, and freed up labor to be invested in the task of increasing yields. Simply put, the Inka wanted to maximize the agricultural productivity of their new acquisition, and resettlement was just another tool to accomplish this end.
Here’s how things appear to have gone down: the Inka rolled into town in the latter part of the 15th century, probably during the reign of Pachakutiy Inka Yupanki or Tupaq Inka Yupanki. The people of Hatunmarca may have surrendered quickly, or else allied with the Inka in some way. Either way, they and only a couple of others were permitted to keep their fortified settlements. The rest of the region’s inhabitants were forcibly removed from their homes and made to settle on the valley floor. This is a playbook that the Inka would employ again and again in their conquest of new provinces. See Lorandi & Boixados (1988) for the extremely similar case of Argentina’s Valle Calchaquí.
So the period of Inka rule in Sawsa begins with a massive shift in the agricultural economy, with hilltop centers abandoned in favor of a more diffuse, lower-density distribution of settlements along the valley floor. This served Tawantinsuyu’s long-running twin goals of discouraging rebellion and intensifying agricultural production.
On the second point, resettlement appears to have been a fantastic success, as the archaeological and ethnographic records indicate that considerable food surpluses were generated in Sawsa during the period of Inka domination. D’Altroy (2014:411) reports that 2,753 qullqa (Inka modular storehouses) are known from the archaeological record of the upper Mantaro Valley, primarily found near the provincial center of Hatun Sawsa and its suburbs.
With its densely-settled population and high agricultural productivity, Sawsa predominantly served the Inka as a source of labor for their infrastructure projects and food for their armies. The second of these was particularly important; we have evidence that Sawsa was able to produce considerable surpluses of food. Spanish records report that the army of Challkuchima, an Inka general in the service of the Sapa Inka Atawallpa, camped at Hatun Sawsa for a long period in 1532-3 (D’Altroy 2014:335). This army is reported to have numbered 35,000 men.
But how agriculturally productive was the region? And how densely settled? This is going to be our main guiding question for the remainder of this post. We need some good-quality evidence, and it needs to come from multiple avenues and disciplines. More importantly, these disparate sources of evidence should all be telling us roughly the same story. The following sections will mostly be geared towards exploring our evidence for the population of Sawsa from four angles: documentary sources, archaeological evidence, agricultural carrying capacity, and the rate of depopulation in the region over the first 50 years of Spanish domination.
3. A word on the pre-Columbian demography debate
This is a topic that’s been touched on in our previous piece on the Marajoara culture of the Amazon Delta, so forgive me if I deal with it only briefly here. Here’s the CliffsNotes: essentially, there has been a debate raging in historical-demographic circles for the better part of the last century, centered on both the overall population of the pre-Columbian Americas and the levels of population density individual pre-Columbian societies were able to achieve. This debate has largely seen specialists divide themselves into two camps: one that favors a stupidly low population of the Americas (Kroeber 1939 reports no more than 9 million inhabitants), and one that favors a stupidly high population (Dobyns 1966 calculates 90-112.5 million inhabitants). More recently, compromise estimates ranging from 40 to 60 million inhabitants have been more in vogue (e.g. Denevan 1992’s 53 million), and the stupidly low side of the debate has fallen off considerably in the face of new evidence from archaeology and population genetics. So the primary debate has been thoroughly reframed; the main camps are now the stupidly high and compromise counts of pre-Columbian population.
In a specifically Andean context, different sources have calculated dramatically different total populations of Tawantinsuyu on the eve of European contact. Rowe (1947) reports 311,257 indigenous tributaries counted in the 1571 Spanish census; each of these is assumed to correspond to a single household, which is assumed to have averaged around 5 people -- 1.5 million residents, in total. By examination of reported rates of depopulation in a small sample of 5 Inka wamani, Rowe (who, despite his missteps, remains one of the undisputed greats of English-language Andeanist research) determines an average of 75% depopulation across the Andes in the 50 years after European contact. So, 6 million residents of the empire at its height.
The attentive reader should notice a couple of basic problems with this train of reasoning. First and foremost, an average household size of 5 is pretty low for a population composed primarily of subsistence farmers. Households in these kinds of societies actually tend to average between 6 and 8 individuals, depending on their specific social circumstances. And second, assuming strict continuity between the areas controlled by the Inka and those controlled by the Spanish is simply not correct. The Spanish conquest of Tawantinsuyu led to a significant contraction of the realm, as societies along the southern and eastern borders of the empire took advantage of the chaotic exchange of power to assert their own independence. Toledo’s 1571 census included only areas that had been the core of the Inka realm in the Peruvian coast, sierra, and Altiplano. The size of the area in question is dramatically different; so, too, would have been the population. For these reasons and others that we don’t really have time to get into, Rowe’s estimate must necessarily be a dramatic undercount.
On the complete other side of the spectrum, Dobyns (1966) uses a flat 90-95% rate of total depopulation from European contact to the final population nadir to establish a population estimate of 30 to 37.5 million. There is some good evidence for this methodology, but it isn’t primarily derived from examination of the Andean region specifically; rather, Dobyns considered it to be the expected rate of decline of all indigenous communities after contact with Europeans, ignoring the specific regional variations that are all-important in this part of the Americas. Smith (1970) reports that, for many parts of the Andes, a 90-95% rate of decline does not align with the documentary evidence from Spanish sources of the time. Many areas, especially around the coast, experienced higher rates of decline, while highland areas by and large saw significantly lower ones.
Ad-hoc compromise estimates tend to cleave closer to Rowe than Dobyns, hedging their bets with language like that used in Wikipedia’s entries on the Inka population: “may have ruled over more than 10 million residents” is a mainstay of pop history on Tawantinsuyu. Again, like with the case of writing on the Marajoara culture discussed in a previous post, this is a very interesting way of saying “it’s basically a mathematical impossibility that there were less than 10 million residents.”
Hence, this project. There is a real need for estimates of Tawantinsuyu’s total population derived from in-depth examination of specific, more constrained areas of study, with recognition of the stunning diversity of Andean geographies and the importance of incorporating evidence from multiple disciplines. Here, we’ll start with a simple case study. Only later will we begin to spill ink about the shape of the Inkas’ empire and overall estimates of its population and density. For now: Sawsa!
4. Sources
Where prior projects posted on this blog have suffered from a lack of good sources on the topic at hand, here we have a somewhat different problem. We have access to dozens of sources on Inka history written (in Spanish, though not exclusively by Spanish authors) throughout the late 16th and early 17th centuries. Hell, we have at least seven primary sources on the conquest of Tawantinsuyu by Pizarro, set down by members of the company themselves (Rowe 2006:2). Later authors, writing in the latter half of the 16th century and the early part of the 17th century, were able to draw on this considerable body of primary sources and their own experiences of a recently Inka-ruled Andes in their treatments of Inka history, culture, and lifestyles.
Unfortunately, I have limited time on my hands, and I really hate the prospect of slogging through thousands of pages of records written in old-timey Castilian. And besides, this isn’t strictly a project on Inka history. This is a project that zeroes in on a single aspect of that history: the demography of Tawantinsuyu, with particular emphasis on the size of the population under Inka rule at the empire’s height in the 1520s.
So, demography. The first truly reliable data we have out of Peru comes from the 1570s, when Francisco de Toledo instituted a census to track the rapidly-dwindling population of the Viceroyalty. This wasn’t quite analogous to a modern-day census, as the goal of de Toledo’s visitadores wasn’t to provide an exact count of the number of people in the province, but rather the number of taxpayers (tributarios). However, we can use it as an index for overall population nonetheless.
Under the Inka, a taxpayer was defined as a married, male head of household between the ages of 30 and 50. The households of local rulers (kurakakuna), including the leaders of villages as small as 100 households, were exempted from tribute and thus were not counted, as were yanakuna (slaves), the ethnically Inka ruling class, and households belonging to the estates of the Inka nobility.
After the Spanish conquest, Pizarro’s new Viceroyalty of Perú largely retained the basic organizational structure established by the Inka -- the wamani of Tawantinsuyu became corregimientos under the Viceroyalty, with their pre-Spanish subdivisions, termed repartimientos, awarded to individual conquistadors as their personal fiefs. This is, of course, a simplification of a complicated process, where names were switched around, boundaries were occasionally redrawn, and multiple wamani were often merged into a single corregimiento.
Fortunately, Sawsa appears to have gone into the era of Spanish domination with its boundaries and internal divisions roughly intact. For the corregimiento of Xauxa, the 1575 census reports:
Hatun Xauxaː 1,079 tributaries
Hurin Guancaː 3,374 tributaries
Hanan Guancaː 2,500 tributaries
various mitmaqkunaː 538 tributaries
Totalː 7,491 tributaries
(via Cook 1982)
We have evidence that, across a sample of several Inka wamani, a fairly consistent ⅓ of households were not subject to taxation, being headed up by ineligible men or widowed women (Pärssinen 1992:391). And the average family within the Inka Empire is believed to be comparable in size to other agrarian societies of the period, averaging somewhere between 6 and 8 individuals per household. Smith (1970) is broadly in agreement with these numbers, using data collected in the Lake Titicaca Basin province of Lupaqa/Chucuito to determine that each listed tributario probably corresponded to roughly 8-10 inhabitants.
I like these numbers, and I like even more that they’re generally in sync with one another. For the remainder of this project, I will be using the Smith (1970) ratio of 8-10 inhabitants for every 1 taxpaying tributary recorded. In 1575, that gives us a population within Sawsa of 55,624-69,530 individuals. And there it is! Our final number, the likely range of total population within this province in the Inka period. Please ignore the scroll bar on the right-hand side of your screen.
I joke, I joke. I kid, I kid! The problem with this number is that, while it may tell us a lot about the population of Sawsa under the Spanish in the 1570s, it’s virtually useless as an indication of the area’s population under Tawantinsuyu, because it can only tell us about a Sawsa that’s right in the midst of a historical period defined by the worst rates of epidemic mortality in all of human history. What follows here is adapted from Dobyns (1963)ː
Smallpox, in all of its various forms, has an attested case fatality rate of around 35% without modern medical treatment. In 1519, it struck the Americas for the first time, running roughshod over a completely unprepared continent. There’s some evidence indicating that the variant of the disease involved here was significantly deadlier than its more modern cousins, with as much as 50% mortality reported among the Cakchiquel people of Central America. This epidemic likely achieved a similar fatality rate when it reached the Andes in the 1520s; the Sapa Inca Wayna Qhapaq, his son and heir, his uncle, his brother, and his sister all died of the disease, laying the groundwork for the 1529-1532 civil war between his sons Waskar and Atawallpa.
Measles has a modern-day case fatality rate, in developing countries, of up to 15%. In 1531, an epidemic form of this disease struck the Americas, with high rates of mortality reported across Central America and the Caribbean. It’s likely that this disease continued on to the Inca Empire, causing further disruption in advance of the invasion of Francisco Pizarro in 1532.
In 1546, a disease known to the Aztecs as matlazahuatl struck the Andes. Although the Spanish referred to it as peste, meaning “plague,” the disease is speculated by some to have been typhus, some forms of which have a case fatality rate of up to 70%. 1558 saw another epidemic strike -- smallpox and measles again, reportedly, compounded with an influenza outbreak that originated in Europe.
Allowing for a generous 1% annual rate of increase in the interims between these diseases, and assuming that matlazahuatl refers to a type of typhus with a case fatality rate closer to 30%, the population of Tawantinsuyu may have fallen by as much as 55% by 1575, due solely to the transmission of these diseases. And that’s just the impact of epidemic diseases -- contact with Europeans also saw the introduction of a number of extremely nasty endemic ones to the Andes, chiefly malaria and tuberculosis. This is of course to say nothing about the impact of deaths from violence, economic disruption, and compounding systems collapse across the former Inka domain.
But indications are that disease was the major killer, and in fact these diseases were the primary motivation behind the Viceroy’s census in the 1570s -- the population of the Andes was declining so precipitously during this period that the process of extracting tribute from native communities had become an unwieldy mess. Indigenous communities were simply not able to provide the same tribute to the Spanish as they had to the Inka, due to the lack of available labor. The eventual fruit of this information-gathering was the reducciones, where native inhabitants of the depopulated countryside were concentrated and resettled in Spanish-style towns, to ease the administrative burden of tax collection.
So what do? There have been a couple of efforts to study population decline rates in the Andes, most notably in Dobyns (1966). However, these studies always seem to come up against the same problem: the apparent rates of population decline are incredibly schizophrenic in the Andes, with some areas ending up completely depopulated of their native inhabitants and other areas seeing at most fairly modest reductions. The likelihood that any pan-Andean population decline figure would be applicable to Sawsa specifically seems low, no matter what kind of methodology you rely on to determine it.
We’ll address other forms of evidence in Sections 5 through 7; right now, we’re focused on documentary sources, and so what we desperately need is more and earlier demographic data. Our gold standard, naturally, would be data collected during the Inka period, which we have abundant reason to believe existed. Documentary sources from the time paint the Inka as frankly obsessive recordkeepers, conducting regular census of both the size of their taxpaying population and the amount of tribute levied from them. This information was recorded on khipu, a system of dyed and knotted ropes used to record numeric and sometimes linguistic information.
Unfortunately, we don’t have a very large set of surviving khipu -- many were deliberately destroyed by the Spanish, and more simply degraded over time, leaving us with a modern-day corpus of about 850 (Urton 2015:151). Even more unfortunately, the systems used to encode information in the khipus are not currently understood, so while we may have some, we’re not yet up to the task of actually reading any of them.
But back in the early 1500s, when the Spanish first rolled into the Andes, khipu-reading was both a necessary and an extremely common skill among the people of indigenous communities, especially among the administrators of indigenous states. Early Spanish documentary reports are rife with population or army size numbers that are clearly informed by khipu readings collected from indigenous experts. Often, the origins of these pieces of data are explicitly remarked upon by these authors; in other cases, we can make assumptions on a more contextual basis.
There is, fortunately, one such count reported for the tributary population of Sawsa in “Inca times” -- i.e. the late 1510s or early 1520s. In his work Relaciones Geográficas de las Indias, the 19th-century Spanish historian Marcos Jiménez de la Espada reprints a 1581 description of the province of Xauxa, which identifies the tributary population “en tiempos pasados del Inca” as having totaled 27,000 -- 6,000 in Hatun Sawsa, 12,000 in Hurin Wanka, and 9,000 in Hanan Wanka. If we apply our standard ratio of 8-10 inhabitants to every 1 tributary, this yields a population of between 212,000 and 270,000 for the wamani as a whole.
This number is broadly in line with other early documentary sources on the populations of individual Inka provinces. Supposedly, wamani were intended to cleave close to an idealized count of 20,000 tributary households, divided into 2 moieties (basically smaller administrative units) of 10,000. However, empire-building is a messy business, and it’s clear that there was quite a bit of variation between provinces as to how closely they matched this number. Pärssinen (1992:299-303), analyzing early Spanish reports of tributary counts from individual provinces, finds an average of 17,189 tributary households per province, with individual wamani ranging between 5,000 at the lowest end and 50,000 at the highest. So our documentary sources on Sawsa would appear to put it around the middle/high range in terms of overall population, with most provinces exhibiting lower populations and a few coming in somewhat higher.
5. Archaeology
Does the archaeological record line up with our 212k-270k estimate?
I haven’t been able to get access to full writeups of the two major projects that have surveyed the area in depth, the Junín Archaeological Research Project and the Upper Mantaro Archaeological Research Project. But what I’ve been able to find has been pretty encouraging for this number.
Unfortunately, we have no direct, smoking-gun evidence for this part of the project. The archaeological record of the Inka period in the Mantaro Valley is severely disrupted, making a full-on survey of the number of households found at archaeological sites occupied between 1470 and 1520 impractical. And, to be fair, this is totally on the Inka. The aforementioned resettlement program shunted the people of the region down to the valley floor, where their diffuse, low-density settlements spent centuries being abandoned, reoccupied, dismantled, buried, dug up, and tilled over by their descendents.
But that isn’t to say we have no good archaeological evidence. The record of the period immediately preceding Inka rule in the Mantaro Valley, the Wanka II Phase, is shockingly well-preserved, thanks to the remote locations of the mountainside population centers and their rapid abandonment when the Inka gained control of the valley. As already reported in an earlier section of this post, these settlements were very large for their assumed level of sociopolitical complexity -- the fortified urban core of Tunanmarca alone would have been home to a maximum of around 13,000 residents, and there were a number of other large settlements dotting the area, as well as smaller villages and hamlets.
In total, D’Altroy (2002:67) reports a population of 60,862 for the Wanka II Phase of the Upper Mantaro and Yanamarca Valleys, assuming full and contemporaneous occupation of the 39 settlements uncovered in the author’s survey. This tracks almost perfectly with the documentary evidence: the area investigated (the northern extremis of the valley) corresponds closely with the Shawsha division of the wamani, for which the reported population in colonial documents from “Inca times” was 6,000 households (48,000-60,000 inhabitants).
The implication here is that the size of the population of the Mantaro Valley changed little during the period of Inka dominion; the main changes were to its concentration and distribution. Or, put more simply: the hypothetical yield increases owing to greater agricultural efficiency failed to trickle down to the subsistence farming population. Instead, larger surpluses only led to more appropriation of resources by the Inka state.
6. Agricultural capacity
The conjunction of documentary and archaeological evidence is compelling, but not ironclad. If we want to be more secure in this conclusion, we should interrogate it through a number of different lenses. The first and most obvious question is: would the Mantaro Valley have been able to feed 216k-270k residents in 1525?
Interesting question.
Agriculture in the Andes is old, originating (as far as we know) sometime in the 4000s or 3000s BCE along the Peruvian coast. First to be domesticated were maize, beans, and squash, followed some time later by highland plant staples, including cereals like amaranth & quinoa and tubers like potatoes, olluco, mashua, and oca. Cash crops included peppers, coca, and cotton.
The main factors limiting agriculture in the Andes have always been precipitation and altitude. At the very highest altitudes, above 4,000 meters above sea level, agriculture is fully impossible; the main productive use of land is for grazing alpacas and llamas, which are raised for their meat and wool. Down to about 3,500 meters, only varieties of potato and other tubers can survive. Below that is the maize-growing zone, which extends all the way down to sea level. Maize is highly sensitive to temperature fluctuations, which can affect its growing season profoundly. Close to sea level, maize cultivars tend to go from planting to floration in 80-90 days; above 2,000 meters, this period increases linearly to a maximum of roughly 150-160 days around the upper growing limit (Earls 1998:15).
While maize’s shorter low-altitude growing period generally allows for multiple harvests per year at locations along the Pacific coast, this region happens to have some of the lowest levels of precipitation in the world, as it’s sandwiched between the rain shadows of the Andes and the Pacific trade winds. Here, agriculture is only possible along a couple of dozen rivers, which carry glacial meltwater from the mountains down to the Pacific. Despite these generally unforgiving conditions in both the highlands and the lowlands, thousands of years of continuous innovation have resulted in agricultural practices that, by the time of European contact, allowed Andean peoples to achieve an impressive degree of food security and generate large food surpluses.
In this section, I’m going to approach the question of the Mantaro Valley’s agricultural carrying capacity from two different angles. For the first part, I’m going to review the information presented in a wonderful little 1980 work on land use in the Mantaro Valley, which is not nearly as useless as it might seem. As it turns out, the practices and limitations of subsistence farming in the Andes appear to have changed shockingly little between the 1520s and the 1970s (besides the introduction of some European cereal crops, obviously -- which we’ll do our best to control for here). For the second part, we’ll review what our primary sources from the 1500s have to say about the lifestyles of subsistence farmers under Tawantinsuyu. Hopefully, I’ll be able to satisfactorily demonstrate that these sources are telling us roughly the same thing about how many people could be supported by a given area of land in Andean prehistory.
Part the first. There are around 65,000 hectares of arable land in the Mantaro Valley (Mayer 1980:15), and the range of agricultural uses this land can be put to is determined by altitude. Land between 3,950 and 4,250 masl makes up the “high” zone, where nighttime frosts are frequent and the main crop grown in the modern day is bitter potatoes (id:35-40). If not artificially fertilized, this land requires lengthy fallow periods to remain productive, with plots generally in use for 1-2 years over a 4-10 year cycle. This modern-day land use pattern appears to be largely unchanged since prehistory; in Qusqu in the 1500s, lower-altitude potato lands would be sown as rarely as one year out of every five, while their highest counterparts would be sown one year in ten (D’Altroy 2014:311).
Land between 3,500 and 4,000 masl makes up the “intermediate” zone, where potatoes, other tubers, legumes, and quinoa are grown in the modern day -- alongside European grains (Mayer 1980:40). Fallow periods in this band vary with altitude; most farmers in the 1970s observed one of two 7-year cycles, either 3 years cropping/4 fallow or 4 cropping/3 fallow. Some low-lying areas in this zone have a 4-year cycle, 2 and 2. A 1:1 ratio of cropping time to fallow time, basically.
Land between 3,000 and 3,500 masl makes up the “low” zone, which is defined by a lack of nighttime frosts (except in July and August) and the resulting ability of farmers in this area to grow maize (Mayer 1980:41-42). This band is divided in the modern day into a number of subzones, primarily distinguished by whether or not they’re irrigated. Fields in the low zone are only very rarely fallowed, and some areas are able to produce two maize harvests per year. When the study was conducted in the 1970s, parts of this zone were covered with commercial farming enterprises, which produce primarily for the market and benefit from economies of scale. However, since we’re only using this information to calculate land use, and not yields, this shouldn’t affect our analysis of agricultural carrying capacity at all.
So in the ‘70s, there were roughly ~19,700 hectares of land under cultivation in the high and intermediate zones, and another ~24,600 hectares under cultivation in the low zone. You’ll note that this isn’t quite the 65,000 hectares determined by the study, but that’s probably due to the fact that about half of the land in the high & intermediate zones would have been sitting fallow during the study period. Since the determining factor here is altitude, I think it’s unlikely that the modern land use pattern represents a radical departure from how things were in prehistory. With that in mind, I’ll be using it as a proxy here.
We have our land use patterns! Determining our other independent variable is going to be a bit more difficult -- time to talk about yields.
In 1980, the per-hectare annual yield for potatoes was 2.0 tonnes in the low zone and 2.5 in the intermediate and high zones (Mayer 1980:49). For maize, both zones averaged 1.0 tonnes/ha. Obviously, taking these numbers and extrapolating them back to the Inka period isn’t the ideal way forward for this project. Modern yields and prehistoric ones are separated by over 400 years of innovation and selective breeding. Worse, there’s very little good-quality research on maize yields in late prehistory. This runs us out of the realm of hard numbers and into airy speculation. Not ideal at all!
But for those interested, airy speculation is one of my main specialties. For potatoes, it’s reasonable to assume that yields under Tawantinsuyu were actually higher than modern ones. In the Mantaro Valley specifically, wheat and barley have displaced potatoes from the most productive parts of the intermediate zone, owing to their status as cash crops. Potato yields across much of the rest of the Andes range from 1.6 to 6 tonnes per hectare (Morris 1999:289), and some heirloom tubers can achieve yields as high as 18 tonnes per hectare (Reategui et al 2019:268).
For this same reason -- displacement by wheat and barley in the low zone -- average maize yields are probably about the same as they were at European contact.
Assuming that all of our low land was planted with maize, and we got an average of 1 harvest per year, these 24,600 hectares would produce 24,600 tonnes of the grain, or about 80 billion food calories. On an average intake of 2,000 kcal/day, that’s enough to support 110,194 individuals.
Assuming that an average of ½ of the higher-altitude potato land (~19,700 ha) was planted annually, and that per-hectare yields averaged a middle of the road 3.8 tonnes, we’d get an annual crop of 74,860 tonnes of potatoes. Under our typical dietary requirements laid out before, that’d be enough to support 106,650 individuals.
Obviously this is all a radical simplification -- Andean peoples didn’t grow only 2 crops, but hundreds of cultivars of dozens of different crops, and they planted not in two rigidly segregated climatic zones to maximize yields, but in dozens of carefully selected microclimates to minimize risk. That being said! Our total ad-hoc agricultural carrying capacity for the Mantaro Valley is 216,844 individuals, or about 3.34 individuals per cultivated hectare. This is probably something of an underestimate, as it fails to take into account other sources of food calories: llama and alpaca-based pastoralism in the high puna was certainly practiced in Sawsa, and some maize-growing land was and still is suitable for 2 annual harvests, rather than 1.
Our other source of data on the agricultural carrying capacity of areas within Tawantinsuyu is ethnographic reports of land use patterns under Inka rule. Garcilaso de la Vega wrote that, upon being married, each newlywed couple received from their ayllu (the Andean local community, on which land use was generally based) a tupu of land, which was not a rigidly defined unit like a hectare but rather seems to have been a general term for “however much land will support two people” (D’Altroy 2014:311). Naturally, this varied with altitude and location. De la Vega’s measurement puts it at around 1 hectare, but we don’t know where that measurement was taken, or what kind of land it was taken on (most likely the higher-altitude areas around Qusqu, the Inka and later Spanish colonial capital).
Further tupu were distributed to households upon the birth of new children: 1 tupu for a son and ½ of a tupu for a daughter. Again, it seems like each tupu was intended to support roughly 2 people with maybe a little bit of surplus left over, and some tupu were about 1 hectare in size.
In our own survey area, the Mantaro Valley, a potato-growing tupu -- the minimum amount of land required to feed 2 individuals -- would have been about .74 hectares. A maize-growing tupu could be significantly smaller, around .4 hectares. If our average household contains 6 individuals -- let’s say a married couple, 2 sons, and 2 daughters -- then we would expect each household to farm an average of 4 tupu. At that kind of farm size, your average subsistence farming household would be able to both meet their own caloric needs and produce a pretty substantial surplus, as they’re on average producing enough food to feed 8 people, but only consuming enough for 6.
Assuming our simplified land use breakdown evens out to be roughly accurate, and our napkin-math calculations of yield changes from the 1500s to the 1970s holds, the Mantaro Valley would have contained around 53,243 tupu of potato-growing land and 61,500 tupu of maize-growing land, for a total of 114,743 tupu.
If we take the 27,000 tributary households implied by the ethnographic data at face value, then our number of tupu farmed by each household is about 4.25 -- almost right on the money. However, as loath as I am to ruin such a perfect number, we should not do this. Because the total number of households implied by a tributary count of 27,000 is 40,500 -- remember, about ⅓ of households were not counted as tributaries. If we take these untaxed households into account, the average number of tupu farmed by a household is only 2.83.
This is much lower than the somewhat idealized number implied by Garcilaso de Vega’s work, but it’s not really very surprising in the least. I personally find it to be much more realistic. What we’re talking about here is a region where the vast majority of the population are subsistence farmers -- the operative word there being subsistence. Societies living in this kind of economy tend to expand until they hit up against the carrying capacity of the land they farm, and then a little further. In an era before widespread, effective birth control, the main way population sizes remain stable long-term is through a constant cycle of expansion and contraction on the level of individual households, mediated by the ever-present specters of disease and malnutrition.
While 2.83 tupu for your average 6-person household implies widespread caloric deficit (with each household only producing enough food calories to support 5.66 people), it should be noted that our numbers generally underestimate the potential agricultural productivity of Sawsa, ignoring things like pastoralism and the possibility in certain places of 2-harvest years. Most likely, your average farming household in the Mantaro Valley functioned much like a subsistence farming household anywhere else, able to generate enough food to meet their basic dietary needs, along with a small surplus.
The average farm size this gives us, roughly 1.6 hectares, also makes sense when put in context with the dynamics of Eastern Hemisphere subsistence agriculture. In wheat-farming regions of the Old World, the minimum area needed to support your average subsistence-farming family tended to be between 1.5 and 2.0 hectares across most of premodern history. The average small Roman farm ranged between 1.2 and 2.0 ha, while in northern China prior to the Han dynasty it was around 1.9. In rice-farming regions, higher per-hectare yields allowed for household farms to be much smaller, generally averaging under 1 hectare in area (see this article by Bret Devereaux, and its attendant series). Maize and potatoes, the staple crops of the Andean region, tended to produce per-area yields that were lower than those of rice, but higher than those of wheat and barley. With this in mind, we would expect average farm sizes to be somewhere between those of rice-farming and wheat-farming regions. And that’s exactly what these numbers are telling us.
7. Depopulation
Here’s the final test. Our population numbers derived from Spanish sources, the archaeological record, and the theoretical agricultural carrying capacity of the Mantaro Valley seem to line up pretty well. We have a population range of 212k-270k inhabitants at the peak of Inka power around 1525. And we have data from the colonial Spanish census suggesting that, by 1575, this had declined to around 55k-70k.
How realistic is a loss of 72.3% of the population over this 50-year period?
Well, as stated before, my (admittedly lowballed) maximum rate of population loss from epidemic diseases during this period is around 55%, which gets us most of the way there. The remaining percentage points could easily be attributable to mortality from endemic diseases, violence, starvation, etc. However, if you want my honest opinion, this isn’t very strong evidence. Made-up numbers will only get you so far in an exercise like this; we need something more concrete.
I think we have a more interesting route available to us. Sawsa isn’t the only wamani with population figures reported for both “Inca times” and the Toledo census in the 1570s. In fact, by my count there are 21 wamani for which we have entries in both of these columns. Among them, Sawsa appears to group very strongly with its neighbors in terms of the rate of population decline attested in 1575:
Neighboring Yauyos (13 on the map) was reportedly home to 2,939 taxpaying households in 1575, down from 10,000 in Inka times -- a 70% decline, nearly identical. Willka (Vilcas) went from 30,000 to 5,672 -- an 81.1% decline.
I’ll speak more to this in a later part of this project, but by and large, the data we have suggests that Sawsa and other mid-altitude wamani of the Peruvian sierra tend to cluster together, cleaving very close to an average population decline of 77.14%. This contrasts sharply with the wamani of the Peruvian coast and the eastern side of the Andes, which saw declines of up to 95% during this period, and those of the high sierra and the Altiplano, which averaged around 46.66%.
But that’s a story for another time. For now, suffice to say that the population decline implied for Sawsa by our data is entirely typical of neighboring wamani at similar elevations.
8. Conclusions
Spanish records claim that the wamani of Sawsa was home to a population of around 27,000 tributaries around the year 1525. On the basis of more detailed census evidence gathered from other Peruvian wamani, this is assumed to correspond to a population totaling 212,000-270,000 individuals.
The archaeology of the area is in agreement with these numbers. Prior to the Inka conquest, the Upper Mantaro and Yanamarca Valleys were likely home to a population of around 60,000, based on the number of households found at surveyed sites (D’Altroy 2002:67). Assuming that this area corresponds roughly with the territory of the Shawsha people, the Inka-era population attested in our documentary sources would have been roughly similar, with 6,000 claimed tributaries giving us a population of 48,000-60,000.
A population of this magnitude would not have put significant strain on the agricultural carrying capacity of the land they lived on. On the basis of documentary evidence and comparison with modern-day yields of heritage potato and maize varieties, we’ve determined that the 65,000 hectares of arable land in the Mantaro Valley would have been able to support a population of at least 216,000 people under Tawantinsuyu.
And finally, Sawsa’s apparent 72.3% population decline between the period of Inka rule and the first Spanish census in 1575 matches up closely with the population decline seen in similar provinces where data is available, most obviously the neughboring wamani of Yauyos and Willka.
With 212k-270k people distributed over the roughly 600 km² of the valley floor, the Sawsa wamani would have had a population density of 353-450 inhabitants per square kilometer. Initially, this might seem a bit ridiculous to an attentive student of premodern demography. Compared to other premodern societies, 353-450 is extremely dense, outstripping the overall population density of any single country in Europe during the same period. The closest any European country comes, according to this infamous post by Lyman Stone, is Belgium, with an average of 50 inhabitants per square kilometer.
Here I think we’re running up against an issue of commensurability. Population density estimates derived from the demography of an entire country are necessarily going to be lower than those arrived at by examination of a single small area with abundant arable land, because an entire country is going to include areas of extremely low density that are nonetheless necessary for overall production efficiency: forests, mountains, pastures, lakes, etc. In this case, I don’t think the evidence is implying a number that’s at all out of sync with reality. Obviously, when we start talking about the population density of the whole of Tawantinsuyu in later parts, this number will go down precipitously as less densely-populated areas are included in our investigation.
But one thing is certain: those lowball estimates ranging from 3 to 6 million inhabitants we talked about earlier are going to be in serious trouble if a single densely-populated wamani was home to a population on the order of 212-270k inhabitants at the height of Inka power. In Part 2 of this piece, we’ll be putting Sawsa in context by talking about the shape of Tawantinsuyu, in which the Mantaro Valley was only one of over a hundred wamani. And here’s an added spoiler: though Sawsa was a core part of the Inka empire, it was neither the largest nor the most densely-populated of their many possessions. In fact, in most respects it appears to have been a thoroughly typical province, defined more by its compactness and density than the size of its population.
The main course, beloved reader, is yet to come. Hopefully, no one is stuffed just yet.
References
Cook, David Noble (1982). Population data for Indian Peruː Sixteenth and Seventeenth centuries.
D’Altroy et al (2001). Empire and domestic economy.
D’Altroy, Terence (2014). The Incas.
Dobyns (1963). An outline of Andean epidemic history to 1720.
Earls (1998). The character of Inca and Andean agriculture.
Jiménez de la Espada, Marcos (1897). Relaciones geográficas de las Indias.
Lorandi & Boixados (1988). Etnohistoria de los Valles Calchaquíes en los siglos XVI y XVII.
Mayer, Enrique (1980). Land use in the Andes.
Murra, John (1968). An Aymara kingdom in 1567.
Pärssinen (1992). Tawantinsuyuː the Inca state and its political organization.
Smith, C. T. (1970). Depopulation of the central Andes in the 16th Century.