The eigenvector of the weights was:

Distance to Site:	0.2245
Favourable Slope:	0.2245
Distance to River:	0.1283
Distance to Large Lake:	0.1283
Distance to Stream:	0.1283
Favourable Aspect:	0.0745
Favourable Geology:	0.0449
Favourable Hydrology:	0.0459

Consistency Ration = 0.01
Consistency is acceptable.

I determined that the factors most important to site location would be distance to a known site and favourable slope - distance to a known site for the reasons of spatial autocorrelation, and favourable slope as people would be more likely to set up camps on areas with a gentle slope than otherwise. Distance to water (be it river, lake or stream) was selected as a close second in importance for the many factors that proximity to water is thought to subsume (ie. plant and animal resources, potablity, transport). I did not differentiate between river, lake and stream in terms of resource importance in this model. In terms of importance, I vacillated between slope and distance to the water factors, but finally decided favourable slope would capture more site types. The danger here is that some sites (for example portage sites or rock art sites) that would be located near water, but probably on unfavourable slopes, would not be modeled for. Aspect was given somewhat less importance as it was a factor I had generalized into "south facing". This may apply as a very general rule to site location, but I suspect that slope and distance to water weigh heavier in any assessment of site location. Surface geology and hydrology were given the least weight. This was in part because the results of the histogram were unexpected and I was not certain I was interpreting them correctly.

In the case of hydrology, the hydrologic type on which sites were most often found was marine aquitards. From what I understand from the supporting data in the Urban Geology of the National Capital Area these are areas that tend to allow water to penetrate very slowly into the soil thus creating the potential for significant surface runoff. Such a poorly drained area does not seem optimum for site location. I would also have expected more sites near areas of bogs and swamps as these environments often provide medicinal plants and good hunting opportunities.

In the case of surface geology, I was surprised by the relative number of sites that were located on bedrock. While some of this could be due to my inexact site location, at least two of the sites on this surface were ones with location's I had most confidence in (as the location was identified to street intersection). Without the results of this histogram, I would have probably put more emphasis on glaciofluvial deposts as areas that might contain older sites. The relative number of sites found on erosional and fluvial terraces is not surprizing however. Terraces above rivers are often good site locations and it is possible that many sites have been found eroding out of the erosional terraces.

The small sample size (only 24 sites) may account for some of this uncertainty - the very real possibility that these sites are not representative of the real site universe cannot be taken out of account. Many of the sites did contain pottery, so the marine clay found in both the hydrological and geological areas could have been a factor in site location.

Once the weights had been calculated, the module MCE was run on the pairwise comparison, with allwater (reclassed to reverse the 0 and 1 values to create the layer allwaterconstraint) serving as the constraint factor.