Correct sector size for artifact ``splash zone''

Remember that we are considering a household compound to be composed of a small number of adobe structures together with associated unenclosed activity areas. In the present, these activity areas include places for storing agricultural by-products and firewood, as well as for tethering animals, threshing and winnowing cereals and for other domestic tasks. The human activities in these structures and activity areas generate debris, some of it durable, like ceramics. This debris is the source of the artifact scatters by which we identify archaeological sites.

However, the correlation between the extent of the artifact scatter and the extent of the habitation area (structures and unenclosed activity areas) is not direct. This is so for two reasons. First, deposition of refuse is not restricted to habitation areas, though it is certainly concentrated in habitation zones. Second, debris is subject to post-depositional processes which tend to disperse what may initially have been a relatively discrete concentration. These processes include trampling and kicking by livestock and humans (cf. [Nielsen 1991]) and, perhaps most important on the Taraco Peninsula, displacement due to the plowing of agricultural fields (cf. [Roper 1976]). The result is that an artifact scatter (archaeological sector) will include both a habitation area and its associated ``splash zone,'' a corona of displaced and peripheral artifacts deriving from that habitation area. This is illustrated in Figure 4.2. The first area directly represents human occupation; the second does not. To arrive at a population estimate we must therefore subtract the splash zone from the total area of the sherd scatter in order to derive the habitation area.

Figure 4.2: Habitation area and ``splash zone''

a) dispersed and b) nucleated residential patterns

The simplest case is that illustrated in Figure 4.2a. This is the isolated household compound, surrounded by cultivated fields. I personally observed a number of such cases in the field, and I took informal measurements on the habitation area and the area of the sherd scatter. I refer to the house compounds of the present-day Aymara inhabitants of the region, discussed in preceding paragraphs. My observations indicated that the average habitation area was no less than 30x30 meters (0.09 ha) in area. This seems to be broadly consistent with what we know of prehispanic household compounds as well. For example, the Pacajes-Inka compound partially excavated by Wise at Lukurmata consisted of at least three rooms and seems to have been enclosed by a stone wall ([Wise 1993]). It is therefore very similar to modern compounds. The compound measures 16 m on one side, though the other dimension could not be determines. Assuming it was square, the architectural compound alone covered 0.03 ha. Since this was an isolated household (``dispersed habitation'') man of the activity areas would have been located outside of the architectural compound itself, as is the case with modern houses.

At least some Tiwanaku urban household compounds - those excavated by Janusek at Tiwanaku - were considerably larger than this, suggesting that household activities were more contained within the architectural limits of the compound wall. Excavations in the AKE1-M and AKE-2 areas revealed very large residential compounds. These pertained to the Tiwanaku IV phase. They were delimited by large encircling compound walls, and contained a variety of domestic structures and outside activity areas. These are examples of household compounds in a nucleated habitation context. The dimensions of the compounds themselves are somewhat difficult to determine, since Janusek excavated only portions of them. However, the eastern wall of the AKE1-M compound seems to be at least 25 m in length, and the width of the compound could be no less than 8 m ([Janusek 1994], Figure 7.1). The absolute minimum size for the compound, then, would be 0.02 ha. However, its known dimensions are consistent with an area of 0.06 ha - 25 m square - or even more; roughly the same size in terms of total habitation area as the modern and late prehispanic household compounds discussed above.

A habitation area of 30x30 m is, in my small sample of modern household compounds, associated with an artifact scatter of approximately 50x50 m (0.25 ha). That is to say that the splash zone extended for approximately 10 meters on any side of the habitation zone. I consistently found this figure to be a good descriptor of the splash zones in modern houses, so I will use it here.

If we measure area in hectares and model the plan of the artifact scatter as a square, then:

H = 
$\frac{\left(\sqrt{A*10000}-20\right)^{2}}{10000}$




where A = scatter area


H = habitation area

Figure 4.2b shows the more complicated case of nucleated habitation in towns and villages. In this case we have densely-packed habitation areas (residential compounds) with overlapping splash zones. Almost all of the interior of the settlement is habitation area, with a rim of uninhabited splash zone around the edge. If we view a village in this way, then we can apply exactly the same correction equation to it that we applied to the isolated household compound. Therefore, there is no need to distinguish between different residential patterns in order to correct for the splash zone.

The actual correction method I used in my settlement database therefore consisted of three steps:

1. Derive the habitation area from the scatter area using the equation given above.
2. Make sure the corrected size of a sector is no less than the area of a single habitation unit (0.09 ha, as above).
3. Do not apply the correction to sites that are not artifact scatters (in this case, raised fields or terrace groups).

The correction method described here has the salutary effect of reducing the relative importance of small sites at a regional scale. When applied to a regional settlement dataset, the cumulative habitation area (``corrected area'') of small sites is reduced relative to that of larger sites. This corrects an error which has been common in Titicaca Basin settlement archaeology, and reverses some prior conclusions regarding the demographic effects of the Tiwanaku collapse. This particular case will be discussed at more length in Chapter 9.

Of course, matters are never this simple. Habitation density (persons per hectare) may vary in contexts of nucleated or dispersed habitation, so that the relation between population and habitation area is not direct. Other factors can affect the relation, as well. For example, habitation density will typically be higher in walled settlements than in unwalled ones ([Wenke 1975]), as people exchange living space for a reduction in collective labor investment. However, to derive a useful relation empirically would require very extensive horizontal excavations (or excellent surface visibility and architectural preservation; see [Montmollin 1987] for a Mesoamerican example). This has been accomplished in some contexts (see [Milner and Oliver 1999]: 90-93 for an example from Cahokia) with excellent results, but we are very far from this point in Titicaca Basin archaeology. Mississippian or Mesoamerican data clearly cannot be used directly. Such measures are specific to particular historical contexts. The method I use here for calculating habitation area is dramatically simplified, but is consistent with what little we know of contemporary and prehistoric habitation in the region and therefore serves as a useful starting point. I fully expect it to be replaced by a more reliable correction method when the necessary data become available.