Overpopulation and over-exploitation of resources continues to strain the process of development for many countries in the tropics. In Latin America deforestation and the subsequent marginalization of these lands has put pressure on the agricultural research community to develop appropriate land-use systems for these areas. Agroforestry is one class of such systems that are presently receiving much attention. The tropical mixed garden, in particular, is one such system that has attracted attention from researchers in various countries. The present study has focused upon the traditional mixed garden system, as found in Costa Rica, with the objective of determining its potential for increased contribution to small farming systems. Development of the data base for this assessment included a survey of 225 farms distributed throughout Costa Rica, year-long case studies of six farms, divided between two contrasting ecological zones, and a simple simulation model of a mixed garden agroforestry system. The mixed garden is clearly an important component of small farming systems in Costa Rica. Though half of the gardens studied were only between 0.01 to 0.20 hectares in size, half were greater, and a few encompassed a hectare or more of land. As a percent of total farm size, mixed gardens were most important in the Tropical Dry Forest and Tropical Moist Forest life zones. Mixed gardens are more common in economically depressed areas and less so in areas where farmers are well off. The ranking of various factors representing ecological complexity of mixed gardens is what one would expect if difference in garden complexity were determined solely by between-zone differences in the environment, thus supporting hypothesis 1. On the other hand, multivariate analysis of species presence/absence data for mixed gardens suggest that the hypothesis (Hypothesis 2) that Holdridge's system of ecological classification is an adequate means of stratifying the variation in species composition in mixed gardens is false. The findings also support the hypothesis (Hypothesis 3) that the mixed garden has a higher energy benefit-cost ratio than commercial cropping systems. The commercial cropping systems on the farms studied consumed between 9 to 10,000 times the amount of cultural energy as did the mixed gardens. Mixed gardens on small farms have the potential to contribute much more to the cash economy of the farm household than they generally do at present. The observations reported here concerning labour patterns and management practices, together with the economic analysis, support the hypothesis (Hypothesis 4) that the output of the mixed garden can be improved. The economic and labour use analysis presented here also supports the hypothesis (Hypothesis 5) that "the mixed garden exists as a supplementary enterprise whose primary function is to absorb excess farm labour." With due regard for the limitations of a simulation of the type used in this thesis, I find support for the contention that the traditional mixed garden in Costa Rica can be developed into an ecologically conservative yet commercially viable cropping system. In particular, the incorporation of high-value timber species shows the potential to significantly improve the long-term economy of the farm. Integrating animal production, as Wagner (1957) had advocated earlier, also can enhance garden productivity. / Graduate and Postdoctoral Studies / Graduate