36 Zeitschrift für Ethnologie 115 (1990) effect of social stratification is sufficient to explain the data (Table 5). Furthermore, this effect is more than two times as high as on external war (Figure 5). POPULATION DENSITY FOOD STRESS TRIBOTE +74 | STRATI- FICATION Fig. 5. Logit-model for collection of tribute Acquisition of land This dependent variable is perhaps the most important for a test of the land-scarcity- hypothesis. If a society in fact fights wars because of land scarcity, one would expect that these wars are aimed at the appropriation of land to get rid of the shortage. With p — .098 for model 2, we reject the hypothesis of no interaction between the three independent variables and the acquistion of land as a motif of warfare. As can be seen from Table 5, none of the models containing only main effects (models 3 to 9) fit the data. Thus, at least one interaction effect between the independent variables will be necessary to explain the data. The most parsimonious models with interaction effects are models 10—12. Of these, only model 12, which contains the interaction between hunger and social stratification (HS*), yields an acceptable fit. It may be concluded, that those interactions involving population density — (PH*) and (PS*) — are not neces- sary to explain the independent variable. Neither does the addition of the main effect of population density (model 15) significantly reduce the L? of model 12 (A L? = 4.22—3.76 = .46, with df = 4—3 = 1). Thus, model 12 is the final model (Figure 6).