Assume that ancient household units are roughly comparable to modern ones

This is far from a safe assumption. First of all, as the archaeological literature on households clearly demonstrates, household configurations can change drastically over time (cf. [Ashmore and Wilk 1988,Netting 1989,Wilk and Netting 1984]). Secondly, there is concrete evidence of a significant shift in residential patterns at the end of the Tiwanaku period (ca. 1100 A.D.). Before this time, at least in the southern Titicaca Basin, people lived almost exclusively in sedentary villages. I have referred to this pattern as ``nucleated habitation'' ([Bandy 1999d]:24). After the collapse of Tiwanaku, however, people began to live in small isolated farmsteads. That is, people inhabited small clusters of one or a few households, separated from other such clusters by substantial tracts of uninhabited land, a pattern I have called ``dispersed habitation'' ([Bandy 1999d]:24). This is how the majority of the regional population still lives today. Apparently, therefore, residential patterns changed significantly over time.

However, there is some evidence that while residential patterns shifted from nucleated to dispersed, the configuration of the actual household remained consistent. At present, local residents live as do most Aymara communities in the Bolivian highlands. A single family group - a married couple together with unmarried children and frequently with elderly relatives - will inhabit a small compound of structures (perhaps 3-5 rooms altogether), often enclosed by an adobe wall, and separated from other such compounds by cultivated fields (cf. [Arnold 1988], Chapter 9). Figure 4.1 is a schematic representation of a typical residential compound. Each of these compounds is adjacent to outdoor activity areas - much of the daily work of the household takes place out of doors - where activities such as threshing, chuño processing, weaving and so on take place. There are no well-defined middens, and this entire inhabited area contains a sufficient density of cultural material to qualify as a ``site'' according to my working definition.

Figure 4.1: Schematic representation of a contemporary Aymara residential compound.

To date, very few post-Tiwanaku, prehispanic household units have been excavated in the southern Titicaca Basin. Those that have been excavated seem to indicate a residential pattern broadly similar to that of the present day. The Late Intermediate Period structures excavated at North Point in Lukurmata are tentatively interpreted by the excavator as forming a ``compound containing a number of houses and enclosed by a large wall'' ([Wise 1993]: 111). A Late Horizon residential compound was excavated at the same site, comprising at least two fieldstone structures and a fieldstone corral, all enclosed by a compound wall ([Wise 1993]:109). The limited information at our disposal therefore seems to indicate that the post-Tiwanaku household form has remained relatively stable, and has been broadly similar to that observed among the modern inhabitants of the area.

Somewhat more archaeological evidence exists for the Tiwanaku period although the excavated sample is still very small. Marc Bermann ([Bermann 1990,Bermann 1993,Bermann 1994,Bermann and Graffam 1989]) excavated a long sequence of domestic architecture from the site of Lukurmata. Though his sequence begins in the early Late Formative and continues through to the LIP, the bulk of his data pertains to the Tiwanaku III-V periods (LF2 and Tiwanaku periods, in my chronology). He excavated several examples of what he considers to be household compounds dating to these period. He describes the typical Tiwanaku household unit at the site as ``a house compound with three structures, two primarily used for a wide range of domestic tasks or as a dwelling, and one for specific and different purposes [storage, sleeping, etc] ... The household unit also included a small enclosed patio used for most activities'' ([Bermann 1993]: 128). Janusek has excavated similar compounds at Tiwanaku and the Misitón area of Lukurmata, these with clear enclosing walls ([Janusek 1994]). The excavations of the Proyecto Wila Jawira in the La Kkaraña area of Tiwanaku have suggested household compounds in the early Late Formative, as well (see [Escalante Moscoso 1994]). The domestic residential compounds built and inhabited by the modern Aymara can be traced with some certainty at least to the Late Formative period.

Sufficient evidence therefore exists for me to propose that the observable shift in residential patterns at the time of the Tiwanaku collapse reflects a change in the external rather than the internal configuration of household compounds. That is to say that the Tiwanaku and Formative Period villages were composed of agglomerations of many household residential compounds, and that the dispersal of population during the Early Pacajes period is best understood as a dispersal of household compounds. In the Tiwanaku period and earlier the compounds were grouped into large clusters which we refer to as ``villages,'' while more recently they have been scattered individually among agricultural fields.

I will proceed, therefore, on the assumption that during all periods of the area's prehistory people lived in household compounds as described above, and that towns and villages essentially comprised dense concentrations of such units.