The Morelos censuses were undertaken for the Spanish Crown because of a dispute with Hernán Cortes' administrators over tax collections. Although the tribute question was the sole purpose of this enumeration, systematic data were also collected on the inhabitants of each household, including information on relationship to the conjugal family or household head. The actual date of these documents is unrecorded, but Carrasco places them between 1534 and 1544, possibly around 1537, when Viceroy Mendoza ordered the enumeration of Cortes' vassals. Carrasco concludes that these documents are of "gran antiguedad" because indigenous titles for governors are used throughout and only a small fraction of the population had been baptized. As Spanish colonialism engulfed the western hemisphere, native scribes adapted old glyphic traditions of portraying household tax obligations to the Roman alphabetic script introduced by Christian friars, producing an extraordinarily detailed account of the household dynamics of the ancient Nahuas.

Ethnohistorians have studied other sets of Nahua censuses of differing character and quality. The Santa María Asunción Codex, for example, resembles a population register, depicting births, deaths, marriages and migration over a half century or more, although the dating of events went unrecorded. Precise kin relationships cannot be determined beyond the nuclear family because the register is made up entirely of glyphs. These ethnohistorical studies offer valuable insights into the workings of indigenous households, Nahuatl linguistics, and the peculiarities of Nahua censuses. I use the most detailed census lists extant, those translated by Cline, because they provide systematic information on almost all individuals enumerated. To analyze these data, I reject the household approach of anthropologists and family historians of the Laslett school. Instead I, like most population historians, study the individual in the family context as a more revealing way to understand family dynamics both in the past and the present. Moreover, as we shall see below, Nahua households were highly dynamic, and their boundaries extremely fluid.

What we find in the Morelos censuses transcribed by Cline are native patterns of marriage and household residence. The "spiritual conquest" was just beginning in this region. Only one Christian marriage appears in these lists. Five men were reported in polygynous unions, involving sixteen women. Nine of these were listed as concubines, four of whom were baptized. Since less than 7% of the population was baptized, this means that a polygamist and his concubines were much more likely to obtain the sacrament of baptism than a monogamist or his spouse. This should not be surprising because missionaries first courted the native elite and readily overlooked their transgressions to win strategic converts. Later, church officials tenaciously fought to exterminate polygamy, but in the early moments of conversion, moderation was often the rule. In any case Christian penetration in this region was minimal. Christian names were recorded for only 164 individuals; another fifteen were noted as "not yet baptized."

It seems likely that for the conquest generation, with its elites not fully coopted, neither Spanish encomenderos nor Catholic friars concerned themselves with such intimate details of native life as age at marriage, coupling practices, or residence patterns. In any case, Carrasco, Díaz Cadena, Cline and other specialists agree that the Morelos censuses are authentically Nahua in form and content. Kin terms are quintessentially Nahuatl, following indigenous principles of social organization, which were almost wholly alien to the European or Spanish mind. Kin ties were expressed from the reference point of ego, here, either the head of the household or family. The form of expression was always possessive instead of absolutive; thus, in the above example, we read "Cuilol's uncle" instead of "Matlalihuitl, uncle". The possessive case appears for almost every individual mentioned in the document, aside from household heads. Likewise, while an occasional reference that a brother or sister was younger or older than the head would not strike a dissonant chord in the ears of native speakers of any European language, the systematic usage of such terms--instead of age--monotonously recorded in these lists, does sound odd to the European ear. Identification of individuals by sex was common to both Nahua and Hispanic linguistics, but only among the Nahua did the expression of sex for unmarried minors seem incongruous and therefore go unreported in the document.

Carrasco and Cline are convinced, and I concur, that the Morelos censuses faithfully reflect native society, virtually untouched by the reformist zeal of Europeans. The censuses for Huitzillan and Quauhchichinollan show that Christian marriage was almost unknown in this region. Among almost 700 couples, there was a single case of Christian marriage, noted as follows: "y otro hermano de Mexicatl que está aquí, llamado Nicolás, Ministro de Huehuetocan y su mujer, de nombre Magdalena Tlaco, casados por la iglesia ya hace un año. . .." Mexicatl's first younger sibling, unmarried, was 20 years old, so the married Christian was still in his teens when the census was taken a year later. His wife's age was unrecorded, but it seems likely that she was younger than the groom, probably four or five years younger. Her recently widowed aunt, also present in the household, was accompanied by two of her own children aged seven and eight years.

Early, universal marriage. If Christian marriage was almost wholly absent from this census, Nahua marriageways were ubiquitous, including concubinage although much less common than what scholars suppose for other times and places. What constituted marriage among the Nahuas? The Morelos censuses state nothing of courtship, parental arrangements, or the workings of professional matchmakers. Nor is there much evidence on the degree, or even expectation, of affection between spouses. What the listings show is the existence of nuptial unions, apparently sanctioned by kin, community, and certainly by the individuals involved, males and perhaps females, although when it came time to marry their youthfulness probably left them at the mercy of elders. Cline translates the Nahua text for marriage as "they took each other," "to acquire a woman," or simply "his woman." An anonymous Spanish translation renders "se juntaron" or "su mujer," but there are no ethnographies of how unions were formalized in rural Morelos.

While these censuses offer little formal description or definition of marriage, four facets clearly emerge from the contents of the lists. Coresidence was the primordial consideration, regardless of the ages of spouses. Absences, however infrequent, were clearly noted. Neither married women nor married men were recorded as residing alone without an explanation being noted in the document. In household Q10, we find the rare case of a married woman alone, an example which helps define the importance of marriage. The eleventh person in Q10 is listed as Tlaco, the older sister of the head, who, in Cline's translation, "just accompanies him [the head]. She is just an abandoned person; she was married someplace else." She is married, but alone, unaccompanied by children, and the last person reported in the household, listed where orphans, unrelated dependents and tribute helpers invariably appear. Normally older sisters regardless of marital condition are recorded much closer to the head, before younger sisters or more distant kin.

The second rule of marriage, but of almost equal importance, was the expectation of permanence, barring death and the occasional abandonment or separation (as noted for five women and two men). Third was the importance given to procreation, with the number of surviving, co-resident offspring noted for every nuptial union. Where there was no issue, almost without exception the number of years of the union was noted (e.g., "no tiene hijos y ya tienen diez años de estar juntos"), perhaps as an explanation for subfecundity or to suggest the earnest persistence of the union. Fourth, marriage was essential to attain, and retain, full adult status, "a step toward adult responsibility," according to Cline. No one could become head of a family unit without having been married. Indeed, even a young married male might displace an elder from headship, when the latter became widowed and failed to re-marry within a brief interval. Finally, although Nahua marriage was not necessarily monogamous, concubinage or polygyny was rare in these communities, limited to a handful of the highest political figures in each village.

The censuses show that marriage frequently occurred in what we would today call "childhood." Historians have long disagreed on marriage age among the Nahuas. For females, estimates range from close to fifteen years, to the late teens, and even to the mid-twenties. Gibson thought that Nahuas probably married at younger ages than Spaniards, but he declined to put a figure on it. Other historians descry the Black Legend in Native American marriage patterns, subscribing to the notion that with conquest greedy European ecclesiastics and encomenderos forced natives to abandon pre-Columbian sobriety for the debauchery that came with colonial rule. It is argued that with conquest marriage age was forced down to the early teens to increase Spanish income from tribute and religious fees, but there is little contemporary evidence for this view. I think the contrary is more likely. With Spanish colonization, marriage age probably rose rather than fell, if for no other reason than because, to the extent that natives accepted Catholic marriage, Christian clerics would have refused to wed girls who had not attained the age of reason (twelve years) and thus were considered incapable of voluntarily taking the nuptial vows.

In recent decades evidence for early marriage among the Nahuas has accumulated steadily, thanks to the work of ethnohistorians. When taken as a whole and analyzed from a demographic perspective, these data lead to the conclusion--unequivocal, it seems to me--that most girls probably married before their fifteenth birthday, indeed many before their twelfth.

In the well known Codex Mendoza native artists used pictographs to portray the life stages of males and females from birth to death. At age thirteen we see a girl at work on the metate, grinding corn for tortillas (boys carry loads and paddle canoes), and then at age 14, the girl is weaving with a backstrap loom (while the boy fishes).

For girls the next scene is marriage, although boys are shown receiving additional instruction in the arts of life. Accompanying the marriage scene are glyphs which show marriage occurring at age fifteen. A strict reading would be that both sexes are ready for marriage at that age, but for males union was probably delayed to acquire additional skills, as illustrated in the pictographs.

Contemporary Spanish texts on native marriage customs are often viewed skeptically by historians, but the rhetorical flourishes of a letter written by Viceroy Martin Enriquez in 1577 point in the right direction, at least for rural Morelos. The Viceroy informed his successor that before Christianity was introduced girls married almost at birth because none reached age twelve without marrying ("siendo costumbre en t[iem]po de su infidelidad casarse casi en naciendo porq[ue] no llegava muchacha a doze años q[ue] no se casase").

Inferring sex of unmarried individuals. Beyond the ethnographic evidence, demographic proof of early marriage among the Nahuas comes from a tally of the population by age, sex and marital status. Before this table can be constructed, however, the sex of each unmarried individual must be inferred, sometimes solely on the basis of the Nahuatl name. The ethnohistorian's inclination is to remain faithful to this text. Since the Nahuatl does not indicate sex of the unmarried, inferring sex would violate a canon of the discipline. For the demographic historian such information, even if it must be inferred, is essential to clarify the question of age at marriage. If unmarried females are not distinguished from males, the extreme precocity of marriage is not readily discernible.

Marriage is so fundamental to Nahua household structure that I deliberately pursued an aggressive strategy, inferring the sex of almost all unmarried individuals. Sexing was accomplished solely upon the basis of names in Cline's translation (I used whole names and suffixes), but without reference to partial totals accompanying the lists. I was genuinely surprised to find close agreement between my inferences and the internal partial totals recorded in the document (as in Table 3 below) because it did not occur to me to make this check until several weeks after the sexing of names was finished, the three-way table of marital status by age and sex prepared, and the argument about child marriage nearly complete. From the close correspondence between summary totals in the document and the inferences, I conclude that the exercise was a success.

Confidence was heightened with the even more recent chance encounter in the Museo Nacional de Antropología e Historia of a second transcription of a portion of this census accompanied by a Spanish translation. This anonymous work, neatly pencilled in two inexpensive notebooks (unfortunately on impermanent, highly acidic paper) faithfully transcribes the text for 50 households of Quauhchichinollan. Cline did not consult this work, because while correspondence between the two transcriptions for these households is close, there are differences which could only be explained by independent readings. In addition, one of the notebooks is a Spanish translation, which offers an often explicit gendering of names. For 446 individuals, there were twelve differences between the gender assigned in the Spanish notebook and my inferences. The net effect would be to reduce the number of females in my dataset by ten, but I am convinced that this adjustment is unwarranted (in any case, it would bolster the argument which follows about the scarcity of never-married females). Since gendering of Spanish names may be often determined by a single letter, a translator may easily render, as in this notebook, "Xocoyotl Maria" as "Benjamín María" instead of "Benjamina María," Teyacapan as "Primogénito" instead of "Primogénita," or "Unico" (Centehua) for "Unica", etc. The suffix "ehuatl" as in Tecayehuatl (audaz) is invariably a male name (but rendered as female in the Spanish notebook) while "cahua" as in Teiztlacahua is female (translated as male). I see no reason to change these or other disagreements found with this unsigned, undated, and unpublished anonymous translation. In any case, doing so inflates the surplus of unmarried males, thus reinforcing the theses which follow of extremely precocious child marriage and a socially constructed "marriage-squeeze".

As further verification of the gendering of names, and after all the above was completed, three additional tests were conducted. First, Díaz Cadena's translation of the Tepoztlan data was used to extract the sex of 57 unique Nahuatl names. For 636 individuals in my dataset with names occurring in the Díaz Cadena publication, there is only one disagreement of inferred sex: "Tecapanton" is rendered as male by Díaz Cadena, but I think female is more likely. A second test was constructed from Prof. Carrasco's published translation of the "Libros de Tributos" for seven households in Molotla, consisting of 72 individuals. Of 34 unique names occurring therein, one disagreement was found ("Ichpochton") and was reclassified in my dataset as female (affecting two individuals). As a final test, Prof. Carrasco kindly reviewed the 661 names in my database. His specific queries led to the correction of sex for two names (five individuals). Overall, the discovered error rate for my gendering of names is less than 0.3%, a much lower degree of error than historical demographers customarily work with.

In fact, the task of sexing is not as difficult as non-Nahuatlahtos might imagine. (Nor is my approach original; Prem's analysis of the lists for Tepetenchic and Molotla in ms. 551 is also based on the sexing of unmarried children.) First, the 661 Nahuatl names studied here were rigidly sex-typed, more so than any other social construction the historian is likely to encounter. Prof. Carrasco (private correspondence) raises the possibility that several kinds of names, such as those based in calendrics or birth order, may be shared by both sexes. In practice, in the lists studied here, such cross-naming occurred infrequently--only eight instances, once each for six distinct names: Teicuh ("The Second One," 183 females vs. one male), Necahual ("The Quiet One," 151 females), Xocoyotl ("Benjamina," 38 females), Quauhtli (a calendric meaning "Eagle," 18 males vs one female), Tecolotl (2 females), and Teyauh ("Our Cloud," 2 females). Only one name, "Ollacatl," ("Flexible Cane", five occurrences) was shared by both sexes in more than one instance, with three males and two females. When the violations of the rule are summed (8), compared with the same names which conform to rule (397) and the odds computed (8/397=2.0%), we must conclude that, as a practical matter, names were rigidly gender-typed among these rural folk. The figure shrinks to insignificance when names with no violations are included in the computation (8/2475=0.3%).

Then too, female names were drawn from a tiny pool. The top ten names occurring fifteen or more times accounted for 83% of the enumerated female population; in addition to the above, Teyacapan ("First Born," n=313), Tlaco ("Middle One," 182), Xoco ("Last Born," 53), Centehua ("My woman/wife," 42), Tlacoehua ("Second Born," 21), and Tepin ("Older Sister," 15) appeared with monotonous regularity. In total, 1,196 females shared 87 indigenous names. Male names had more imaginative meanings and were drawn from a vastly larger pool, totalling 574 names. While there were eight occurring fifteen or more times--in addition to Quauhtli, Yaotl ("Rival or Enemy," n=74), Matlalihuitl ("Rich Feather," 63), Nochhuetl ("Ideal Bean," 52), Coatl ("Serpent," 48), Tototl ("Bird," 19), Tochtli ("Rabbit," 17), and Çolin ("Quail," 16)--they constituted less than one-fifth of the names of all males.

Finally, since the sex of the married and widowed appears in the document (n=1,610), their names may be used to infer the sex of the never-married (572 instances). Christian names were also helpful in a few cases (n=107). That left the sex of 177 individuals to be inferred on the basis of suffixes and other linguistic elements (29 others could not be determined for lack of a name and nine were gendered by other information). This exhausting series of tests (and their prolonged description here) is necessary because conventional demographic techniques yield their most insightful findings when gender is taken into account. Moreover, the results are sufficiently surprising that the reader's first reaction may be to discount the arguments which follow as simply due to faulty methodology, error or ignorance.

Child Marriage. The Morelos censuses offer the most conclusive evidence available on marriage age among ordinary Nahuas, and a demographic analysis of these data points to younger ages than previously suspected. It seems that many females were married by age twelve, indeed, some as early as age eight, with the average probably falling between twelve and fourteen years. Note, however, that these ages were of no special significance to the Nahuas whose favored ages were the digits "10" (n=140) and "15" (n=104), but not "11", "12", "13" or "14" (n=12!) or, for that matter, even "16" through "19" (n=2!). Population historians rarely encounter such massive age-heaping, yet even these data can be made to yield their demographic secrets.

Table 1. Population profiles from two ancient Morelos censuses
Age, sex, and marital status
(probable date: 1537)

Source: computed from translated census lists published in Cline, 1993b; inferences of sex, where necessary, are mine.

The proof of average age comes from a detailed examination of the libro de tributos data on age, sex, and marital status (Table 1). To obtain a numeric answer to the question of average marriage age, three rather straight-forward assumptions are necessary because the lists do not report ages for adults or children once they have married. First, assume that adulthood began at age ten--that is, that all married females were at least ten years old--and second, that age declarations were roughly correct. With these assumptions we can then disregard unmarried children less than ten years of age (250 females and 282 males) and construct table 2. (While neither assumption is completely true, both are acceptable simplifications. Surmising the fraction married among girls aged 5-9 years is a tricky matter. If it could be ascertained, it would strengthen the argument which follows.) For females, the currently married account for 74.8% of "adults" ten years of age and over, and the widowed for an additional 16.3%, totalling 91.1% for what demographers call the "ever-married." There were only 83 unmarried girls aged ten years or older, and they comprised 8.9% of the "adult" female population. If adulthood is placed at fifteen years (thereby disregarding never-married girls aged fourteen years and under), the percent of ever-married females must rise, topping 95% (40 unmarried and 845 ever-married). This figure is twenty-five percentage points greater than that for medieval England and forty points above Hajnal's ceiling for Western European marriage patterns, including those of Spain and Portugal. These Nahua women married at much younger ages than females anywhere in Western Europe from medieval times on.

Table 2. Sex ratios by sex, marital status and age

Source: computed from translated census lists published in Cline, 1993b.

Note: "Married" includes one abandoned woman and five women and two men who were noted as "no longer married." The document rarely reports ages for the ever-married.

While Nahua ages should not be taken literally, neither should one think that "10" (note the large number of ten year olds in Table 1) was only a symbol for coming of age or for puberty. Cline observes "a lack of concern with precision" in Nahua ages. Yet, the Nahua knew how to count and reckon ages. In the census, we find a certain Quauhtemoc characterized as "a little child, not yet big. . .now eight years old". These figures are more reliable as categories than quantities. Nonetheless, if we are to understand marriage among the Nahua, determining average age is fundamental. Combing both qualitative and quantitative data shows that many females began the transition to marriage as early as age ten and almost all had completed it by age fifteen. Adulthood meant coresident marriage for Nahua females, however young the chronological age might be.

The document itself confirms these arguments. Folio 36 of ms. 549 reports population totals by marital status for the community of Quauhchichinollan. This contemporary summary drafted in Nahuatl records 287 currently married women (135 in primary families and 152 "not yet in their own separate residences"), 70 widowed, and 24 unmarried (Table 3). Converting to percentages, we find 94% of "adult" women of Quauhchichinollan, as defined by the scribe, characterized as either married or widowed (including the few who were separated or abandoned). The corresponding figure for Huitzillan district is 96%. For males, the figures are 78.2% and 77.5%, respectively.

Table 3. Population totals by sex and marital status as reported for Quauhchichinollan community

"(Here are the people of an) altepetl named the Quauhchichinollan people; all of them total 135 houses [i.e., 135 primary families with a like number of married women].

"Here are the married men who are still just together with other people, not yet in their own separate residences: a total of 152 [and 152 married women].

"Here are the unmarried young men: 80.

"Here are the unmarried young women: 24.

"Here are the widows: 70.

"Here are the children: 226."

Source: Cline, 1993b, p. 219.

At what age did a rural Nahua female become an adult? The indigenous administrator seems to have considered adulthood for females as beginning around age ten. In the Quauhchichinollan portion of the census I tallied almost 300 married women, but only 18 unmarried females above the age of ten. To arrive at a figure of 24, the total in the document itself, exactly half of the ten year olds would have to be counted as "unmarried young women." Errors in inferring sex could not explain away this finding (although there was one unmarried twenty year old whose sex I could not determine). The same problem exists with males. I inferred that there were 77 unmarried males eleven years of age and older. My total falls three short of that found in the original document and would have to be made up from among the ten year old boys (or by reclassifying three girls). Only one conclusion is plausible: the internal evidence, as well as inferences from the raw data, point to substantial pre-teen marriage for both sexes, as early as ten years for girls and eleven for boys.

Cline describes two cases of extremely young married girls whose ages were reported--an eight-year old who had been married four years and a nine-year old married last year, both recorded as infertile--and suggests that these precocious unions may be due to scribal error or that age may have been noted to highlight their unusual youthfulness. She cautiously concludes that "marriage may have occurred early" among the Nahuas. It seems to me that a systematic examination of the data on age, sex and marital status proves such caution unwarranted. Childhood marriage was the rule, not the exception, among the Nahuas of rural Morelos.

Carrasco's study of Tepotzlan provides unexpected support for this conclusion. The demographic "decadence" which Carrasco spies in the Tepotzlan figures may instead point to astonishingly precocious marriage. His ratio of total inhabitants to married couples reveals only 3.3 inhabitants per couple in Quauhchichinollan and 3.7 for 9 barrios of Tepoztlán. For the same places four centuries later, ratios range from 5.5 to 6.0 (data compiled from the 1930 census). Subtracting couples from these averages, he discerned severe demographic disturbances in ancient Mexico, with only 1.3-1.7 sons, daughters, widows, orphans and others combined per married couple (compared with 3.5-4.0 in modern times). The conclusion that the low ratios for the sixteenth century are signs of "a population in decadence" seems inescapable, but a second interpretation may be more plausible. Consider that ratios, by definition, consist of two elements: numerator and denominator (here, total unmarried population and total number of couples, respectively). Carrasco's ratio may be pointing to a relative abundance in the denominator of married couples, including a sizeable number of married children less than fifteen years old, instead of an absolute scarcity of young children. Likewise, in modern times a high ratio may signal a relative dearth of couples, caused in part by delayed marriage (and in 1930, by the lingering devastation of civil war), rather than simply an abundance of children. Indeed, as we have seen, the ancient listings record very few unmarried girls aged ten years or more (Table 1), while in modern Tepotzlan, the great majority of girls aged 14-19 years (70.0%) were not, and had never been, in any form of union--civil, religious or consensual (fortunately all forms of union and disunion were reported in the 1930 census). If girls in 1930 followed the same pattern of nuptial precociousness as four centuries earlier (as discussed below: 50% married aged 10-14, 95% at ages 15-19, and 98%+ at ages 20 and above), instead of 3.8 unmarrieds per couple there would have been only 2.0, which is very close to the 1.6 figure computed for 1537. Thus, more than four-fifths of the apparent demographic decadence of the first decades of Spanish colonization must be attributed, not to an absence of children, but instead, to a proliferation of married children ([3.8-2.0]/[3.8-1.6]=1.8/2.2=82%)! The scarcity of children was not absolute (demographic), but relative--that is, a social construction. Children became adults upon marriage, and most children above the age of 10 years were married (or widowed, separated or abandoned).

Demographic models provide additional support for the early marriage thesis, although the lack of ages for married and widowed women prevents any straightforward analysis. For readers attentive to quantitative argument, I constructed a hypothetical age distribution from a model life table, assuming life expectancy at birth of 17.5 years and crude rates of 65 births and 60 deaths per thousand population (yielding an average annual growth rate of 0.5%; using "South" models throughout). Under these conditions 71% of the female population would be ten years of age or more. Since the census lists show 91% of females aged ten and over as ever-married, this translates into 65% of the entire female population, the same figure as that obtaining in the sixteenth century censuses.

The model can then be used to produce results for five-year age groups, as in Table 4, along with the likely fractions ever-married at specific ages required to reach the 91.1% married figure for the female population aged ten years and over. This exercise translates totals computed from the lists into conventional age groups and proportions married. Specifically, to reach the 91.1 figure for ever-marrieds would require that 50% of females aged 10-14 be married or widowed, rising to 95% for those aged 15-19, 98% at ages 20-24, and almost all women older than 24. This simulation is remarkably invariant to a wide range of growth rates (+/-1% per annum) and mortality levels (e0 15-25 years). The fraction of female children who were married between age ten and fourteen could not have been less than one-third and probably exceeded one-half, rising above nine-tenths for the age group 15-19. The table also shows that married girls younger than 20 years of age would account for almost one-fourth of the ever-married female population (15/65=23%).

Table 4. Likely percent married by age: Nahua females

Source: Column 1: computations from the inverse projection program Populate (McCaa and Pérez-Brignoli, 1989), using the age mortality pattern for region South. Column 2: Hypothetical figures yielding overall total of 65.0% married. Column 3: Column 1 x Column 2.

Note: In this hypothetical population with high mortality (life expectancy at birth of 17.5 years), moderate natural increase (0.5% per annum) and extremely precocious marriage (90% of population aged ten years or more married), ever-married females (including widowed and abandoned) would have constituted 65.0% of the entire female population.

These proportions married can be used to estimate average age at marriage, following Hajnal's singulate method (SMAM). If we accept the proportions married at specific ages as in Table 4, the mean age of marriage (SMAM) for females is 12.7 years. Four centuries later, the figure stood at 22.2 years (SMAM computed from the 1930 census including all forms of union and disunion), an increase of 9.5 years, one of the largest increases ever discovered by demographic historians. Comparing various marriage and mortality scenarios persuades me that the mean age could not have been much above 14 and was probably substantially less, perhaps below 13 years. The model is compelling evidence that widespread child marriage was a fact among the Nahua decades before Christian missionaries could root out polygamy, much less instill Catholic sacraments of baptism or marriage, and before Spaniards could change the rules of native tribute.

For the narrative historian, skeptical of quantitative methods, qualitative evidence from the texts themselves may be more persuasive. The language of the Nahua censuses points to the expectation of early marriage. For the first 89 households enumerated in Huitzillan community, there were only twelve unmarried females more than ten years of age. "Not yet married" was noted for seven of these. Of eleven girls explicitly recorded as ten year olds, one was noted as "not yet married". Thus, marriage was expected to take place soon for most of the few remaining unmarried girls above age ten and for at least one of the handful of unmarried girls as young as ten.

Child brides (and grooms) should have high rates of infertility due to sexual immaturity, and this is exactly what we find recorded in the census. Fertility data in the censuses point to the ubiquity of marriage before the age of biological maturity. Many recently married couples were reported as infertile. "They have not yet had children" was a common refrain. 94 couples "married last year" were recorded as not having children. We do not know how many others were married last year but who nonetheless had already born children because for fertile couples length of union was rarely indicated. Moreover childlessness was a social concept rather than a biological or demographic one, referring to the absence of children in the household. Thus, numbers on childlessness should perhaps be halved to allow for infant or child mortality. As the length of unions increased, the number of childless couples decreased, but remained substantial at all durations: 40 childless couples after two years of marriage, 24 at three, and still fifteen after five years of marriage. If a maximum crude marriage rate of 20 per thousand population per year is assumed, 50 marriages would have occurred annually. Even if as much as half the childlessness was due to mortality, this would mean that still some 15% of couples remained infertile after five years of union (7 observed of 50 supposed). It seems likely that the high rate of infertility in the first years following marriage was due to the fact that most brides were not yet pubescent. Fecundity was a concern of the Nahuas, and the documents distinguish between two types of childlessness: infertility and impotence. Three married males were reported as impotent, but length of their unions went unrecorded, perhaps because there was no expectation of any amelioration with time.

Revealing evidence of child marriage can be found in the sex ratios of children and young adults (Table 2). The overall sex ratio of the population was nearly balanced at 106 males per hundred females and, for children under ten years of age, slightly unbalanced at 113. For the never-married, the ratio jumps to 166, and for the never-married aged ten years and over soars to 328. At age fifteen and above there were more than four unmarried males available for each never-married female. The number of excess unmarried males at age ten alone is almost exactly identical to all never-married females aged eleven or more. These imbalances cannot be restored by the small number of women in polygynous unions (16 distributed among 4 males). Instead, it is the large number of widows when compared with widowers (151/14, or more than ten per widower) that offsets the lop-sided unmarried adult sex ratios. Adding the widowed to the never-married aged 15+ clinches the argument. The shortage becomes a surplus, with 106 marriageable females per 100 unmarried males. Marriage rules in rural Morelos rushed girls into unions, but by doing so created an artificial marriage squeeze which constrained the nuptial prospects of males, prolonged their presence in the household of birth, and extended their contribution to the paternal household economy.

Widows could have provided an escape valve for unmarried males (and remarriage, succor for females), but this option was little used. Widowhood killed the marital prospects of women. Widowed males remarried quickly, unlike widowed females, who remarried much more slowly, if at all. Of 14 widowers enumerated, more than half had been widowed for less than one year while among 122 widows with length of widowhood indicated, 70 had been without a spouse for three or more years. The median interval of widowhood for males was "200 days" versus four years for females. The over supply of widows in the Nahua marriage market matched the over-supply of young, unmarried males, but social constraints discouraged alliances between these groups--perhaps because husbands were expected to be older than their wives.

Nahua males married at older ages than females. A substantial age gap between spouses must have characterized many unions, although the size of the gap is difficult to estimate because of the lack of data on ages of married individuals. Mean age at first marriage for males was certainly less than 20 years (exactly 19.4, if we compute SMAM using the method in Table 4). The median age of the unmarried above eight years old is strikingly low, 15.2 years for males and 10.7 for females, or an age-gap of 4.5 years. "Not yet married" or "not yet taken a wife" was noted for 29 males in 89 households in Huitzillan. Of those aged 10-19 25% were deemed to be dawdling, rising to 60% for unmarried males aged 20 years or more. Brothers-in-law were frequently characterized in this way. Of 18 in these households, twelve were already married. Of the remaining six, four were listed as "not-yet-married", aged 10, 15, 20, and 30 years versus two "not-married" aged 10 and 12.

To summarize, the thesis of extremely precocious marriage is supported by seven tests using two distinct types of evidence (numbers and narratives) and methods (quantitative and linguistic). Three tests are based on inferences about sex--sex ratios by marital condition, proportions remaining single in the pre-teen years, and demographic models of the likely age structure of the population--but four rest wholly on direct verification--contemporary summary totals of the number of single compared to married females, Carrasco's ratio of contemporary totals (unmarried individuals to married couples), linguistic evidence regarding the expectation of early marriage, and high rates of infertility among recently married couples. All the evidence points toward extremely youthful marriage, in round figures, average ages of 12-14 years for females and of 17-19 for males.

Neither mortality nor migration can explain away these findings. While there is a great imbalance in the sexes among the single, the over-all sex ratio for the population aged ten years or more is reasonably balanced (965/928, or 104). If the scarcity of unmarried females is attributed to out-migration (of females, but not males), then the large surplus of widows would have to be explained by return migration of women (but not men) who, in the interim, had married outside the village. This artful hypothesis is contradicted by marital status totals from villages and barrios analyzed by Carrasco and Hinz and his collaborators. The demographics of all the Morelos villages are consistent with the early marriage hypothesis: a shortage of unmarried females, a surplus of widows, and a balanced adult sex ratio. Likewise, if the data are to be explained away by mortality, this would require much greater mortality for adolescent girls than boys, but much lower mortality for adult women than men--a scenario difficult to find anywhere in the historical record. The simplest, reasonable explanation is extremely early age at marriage for girls and little remarriage for widows.

Demographic hell. Historians of the family recognize three restraints on the formation of complex households: late marriage, neolocal rules of residence, and high mortality. As we have seen, for the Nahua, marriage and residence of newlyweds were no restraints at all. On the contrary, such youthful marriages required co-residency with other adults, at least until the age of biological maturity was reached. Mortality, on the other hand, severely restricted multigenerational families among the Nahuas. There is a single four generation household in these villages, notwithstanding the extremely early ages at marriage. Consider household H38 discussed above. With only nine individuals in the household, there were four conjugal family units of which two were already broken by the deaths of husbands, the head's father and his uncle's spouse's brother-in-law.

By shifting the perspective from households to individuals, to children and their rates of orphanhood, we can directly weigh the effects of mortality and, thereby, exclude noise or nuisance factors. While the census specifically lists only three children as complete orphans, that is, with no other kin present, other orphans can be inferred by searching for parents within the immediate household. It is clear from the enumerators' de jure method of enumeration that unmarried children were listed in the households of their parents when these were alive. For children under five years of age, 2.5% were complete orphans, lacking both father and mother (7/280). For 5-9 year olds the figure rises to 6.9% (18/261). This means that one of every fourteen children aged 5-9 were orphaned of both father and mother. Table 5 shows that for children under 5 years of age, excluding the few abandoned and illegitimate, the fathers of almost 10% were already dead, rising above 15% for those aged 5-9. Maternal orphans were common, but occurred less frequently, only 4% and 9%, respectively, but these are minima because, as we have seen, widowers rapidly remarried and the censuses rarely distinguish step- from biological mothers.

Table 5. Orphanhood at various ages was worse in Morelos
than for the poorest conditions contemplated in Model Life Tables

Source: Observed orphanhood computed from census lists in Cline, 1993b; model from Coale and Demeny, 1983.

Note: The hypothetical case assumes mean age of childbearing of 27 for mothers and 30 years for fathers. Model proportions orphaned were computed from figures for Region South Level 1 (life expectancy at birth of 20 years) using conventional methods for estimating single year rates. Observed proportions of paternal orphans are based on 275 children aged 0-4 and 257 aged 5-9; for maternal orphans, totals are 280 and 261, respectively. Nine abandoned or illegitimate children are excluded in computing orphanhood. In the observed data, maternal orphanhood is understated because step-motherhood was rarely noted and widowers (but not widows) remarried without delay.

If prehispanic Mesoamerica is sometimes likened to paradise, the orphanhood evidence in these lists points to a demographic hell--life expectancies below the worst conditions found in standard model life tables (Table 5). This finding is in agreement with the fact that one-fifth of ever-married women were widows at the time of the census. Encountering such high levels of mortality should not be surprising since two virgin soil epidemics probably passed through this region a decade or so before the census was taken (war mortality, on the other hand, was probably insignificant). The smallpox epidemic of 1520-21 is widely recognized as one of the three most devastating epidemics to strike sixteenth century Mexico. Measles erupted in 1531, and smallpox returned in 1538, but we do not know whether the villages studied here succumbed to any of these tragedies.

Even in pre-Columbian times, Mesoamerica was no demographic paradise. Both the bioarchaeological and pictographic evidence point to severe nutritional stress and extremely high mortality before Spanish conquest and colonization began. There is just one study with mortality estimates from the late postclassic through colonial times for any region in the Central Mexican Basin. The author, Michel Hayward, uses skeletal and archival evidence to estimate remaining life expectancy from age fifteen. For the postclassic period Hayward places the figure at 34 years (e15=19), rising in the last half of the seventeenth century to 44 years (e15=29). Her highest estimate is still some 5 years worse than the deadliest conditions in standard model life tables (e15=34, South level 1 as above).

For the first century of Spanish conquest and colonization, there is insufficient evidence to estimate life expectancies of any kind. Yet, the Morelos evidence on orphanhood reveal the workings of a high pressure demographic system, where chronically high levels of mortality and morbidity were overcome by what I call the Amerindian mode of reproduction: early, universal marriage with few restraints on fertility.

Cemithualtin (complex households). Rather than parentage or lineage, the key to social linkages among the Nahua was coresidence, home, or "cemithualtin," those belonging to a yard or patio. Kin ties structured Nahua cemithualtin, and perhaps much of Nahua society--high mortality and child marriages notwithstanding. According to the Morelos censuses studied here, fewer than 60 individuals lived in households in which they had no kin tie with the head. Of these few, half lived in some conjugal group so that 98.99% of all individuals resided with maternal, paternal or spousal kin. True outsiders were limited to three orphans, 20 servants (tribute helpers) and an unmarried slave girl, who was recently acquired to help prepare tortillas for the Quauhchichinollan village leader's mother. Concubines cannot be called outsiders because they unmistakably contributed to household production as well as reproduction. Even migrants came (and went) in family groups, 28 of them, some with as many as thirteen individuals, although as a whole migrant families tended to be smaller than the overall average household size of eight. A typical migrant household is that of H#134:

"Family" or "relative" never occurs in the census document, yet kin ties are ubiquitous. Lockhart's thesis that among the Nahuas "inclusiveness is emphasized over precise descent" is amply confirmed in these lists. Lockhart argues that there was no commonly used Nahuatl term for family or household. Instead there was a series of "words that emphasize the setting in which a joint life takes place, not the origin of the relationships between those living together." A household consisted of "those in one house," "those in one patio," or "people who live together in one house". The household compound was often made up of several structures facing each other on two or more sides of a patio and was much less integrated than buildings in Western Europe or Spain.

If servants and other outsiders amounted to scarcely 1% of the population, relatives of the household head--but not part of the head's conjugal family--accounted for almost half of the population (47.7%). When no other kin tie could be remembered, "distant cousin" resolved the quandary, as in "Here are two distant younger siblings (cousins). . .."

Nahua rules of household formation were simple to an extreme, allowing a profusion of coresidence possibilities. The most inflexible rule, observed in 311 of 315 households, was that only married males could be heads. Of the four exceptions, three were cases of recently widowed mothers with offspring of marriageable ages, but not yet married or recently widowed themselves. In household H#49, the head's husband "died last year." Her daughter's husband died "three years ago." This female head was accompanied by a six year old grandchild and two unmarried offspring, a son aged twenty and a fifteen year old daughter. When either of these married, headship would likely devolve to the son or son-in-law. In household H#115 the female head also lost her husband the previous year. Her household included three unmarried sons, aged twenty, fifteen and three years, a widowed sister and two unmarried nieces, aged fifteen and ten. The last household in the document, H#139, is an anomaly. The head was also a woman, but she was married and her husband was present, but not listed as head. Cline's translation of the case reads: