Pastoralism is a typical economic activity in the Italian Alps. Every summer thousands of sheep are taken to the uplands to exploit the high grazing areas. In these pastures there are many seasonal pastoral sites, called malghe. They are often composed of three functional parts: the shepherd’s hut (casera), where he lives and prepares cheeses; the sheep-pen or enclosure (stala or mandra), where the animals stay overnight and where they are milked twice a day; the storage area (casél), in which the shepherd puts the cheese to mature. These malghe are not randomly distributed in the uplands. The different activities that take place in the malghe, in fact, may be constrained by different spatial variables; and these constraints may have determined the choice of a specific settlement pattern. As we don’t have much information about the movement and seasonal settlement strategies of prehistoric pastoral groups in the Alps, an analysis of the locational criteria of the current shepherds and their malghe could be very useful in understanding more about the ancient shepherds.

This paper is about the application of logistic regression analysis to the malghe of an Italian sample area. This sample area is Val di Fiemme, in the Trentino regione, in the Italian Eastern Alps. It is a 450 km² basin with 82 malghe, only 26 of which are currently used. After carrying out the spatial analysis, the author took part in the seasonal activities of the shepherds in the uplands of Val di Fiemme, to better understand the constraints that influence their settlement strategies. This approach has enriched the computer analyses with ethnographic insight, and should be of great help for the purposes of predictive modeling and the archeology of pastoralism.

The first results of the regression seemed to match the ethnographic information: the location of the Val di Fiemme malghe was influenced by some specific and predictable environmental variables. But the regression exhibited unpredicted residuals, and these were not normally distributed around the regression line. The Moran’s I test for autocorrelation clarified that there was a significant spatial autocorrelation of the residuals. This meant that there was another, unknown, variable that influenced the pattern of our malghe. The analysis of the residuals has then shown that there was a positive autocorrelation between the malghe of the val Cadino area. Further ethnographic and ethnoarchaeological research has helped us to identify the missing variable: the closest malghe tend to be a fixed distance apart, because each malga needs a minimum grazing area to feed its animals. This outcome makes us rethink the relationship between shepherds and upland grazing areas. The environment remains the dominant variable. But the exploitation of pastures (and the consequent distribution of sites) is not as “free” as postulated by many archaeologists and ethnologists of pastoralism, but is, rather, constrained by economic factors.