Située à l'extrême Sud de Madagascar, la région Androy connaît plusieurs particularités par rapport au reste du pays dans le domaine scolaire : les indicateurs de scolarisation et d'alphabétisation y sont les plus faibles de l'Île et les filles y apparaissent singulièrement plus scolarisées que les garçons au niveau primaire. Cherchant à comprendre ces singularités, cette recherche a pour objet l'étude de la (non-)scolarisation dans la région, sur le temps long (depuis l'avènement de l'école jusqu'à aujourd'hui). Afin de saisir la dynamique du fait scolaire et le sens des pratiques de scolarisation dans l'Androy, il s'agit de mettre à jour l'évolution des logiques et des enjeux qui se jouent dans les inter-relations entre politiques éducatives, offre scolaire et demandes sociales et familiales de scolarisation. Plus spécifiquement, il s'agit, d'une part, de comprendre le "retard" scolaire et caractériser l'évolution de la scolarisation dans l'Androy et, d'autre part, de révéler la pluralité des pratiques sociales et familiales face à l'école et les logiques qui les sous-tendent. Au croisement de la sociologie, de la démographie, de l'histoire et de l'anthropologie, cette recherche mobilise plusieurs sources, quantitatives et qualitatives, recoupées et critiquées à la lueur de leurs conditions de production : discours et statistiques (administratifs et scolaires) tirés des archives de la période coloniale, statistiques et énoncés de politiques éducatives du ministère de l'éducation, recensement de la population de 1993, enquête démographique et de santé 2008-09, enquête socio-anthropologique menée dans le district de Tsihombe. Contrairement à l'argument souvent avancé d'un "refus" de l'école par la population de la région, longtemps considérée comme "archaïque", "arriérée" et refusant le "progrès", cette recherche amène à entrevoir la complexité des facteurs expliquant le "retard scolaire" de l'Androy. Ces facteurs se déclinent et s'articulent entre le faible intérêt des pouvoirs successifs (coloniaux et post-coloniaux) pour le développement - scolaire ou plus général - de cette région; le refus de la "soumission" au pouvoir central et la méfiance vis-à-vis de l'administration et de l’État (fanjakana) et de l'institution qui en a formé les agents; les enjeux locaux de pouvoir dans la quête du savoir et de la culture scolaires; enfin, la capacité de l'école elle-même à permettre l'acquisition réelle de savoirs scolaires variable dans le temps et dans l'espace. En considérant la pauvreté comme multidimensionnelle et relative, cette recherche permet également de relativiser l'importance du capital économique dans les pratiques de scolarisation des familles. L'adhésion à l'école n'apparaît pas dialectiquement liée à la seule dimension monétaire et économique du capital possédé par les familles, mais plus intimement au capital social - au sens de Bourdieu - et à la capacité du capital scolaire à produire du capital symbolique. Dans la continuité de ce raisonnement, si les filles sont davantage scolarisées que les garçons depuis le milieu des années 1980, elles l'ont été dans un premier temps "par défaut" avant qu'une véritable préférence pour la scolarisation des filles n'émerge plus récemment. Ce phénomène récent semble renvoyer à l'élargissement du champ des possibles pour les femmes au sein de la société en même temps qu'un focus sur la scolarisation des filles et l' "empowerment" des femmes devient plus prégnant au niveau des pouvoirs publics et plus encore des organisations internationales intervenant dans la région. Si la scolarisation s'est fortement accrue au cours de la dernière décennie, la question des inégalités face à l'école demeure : moins d'un enfant sur deux accède à l'école dans la région, et moins d'un garçon sur trois, l'écart entre filles et garçons tendant à se creuser au détriment de ces derniers, du moins d'un point de vue statistique au niveau régional. / Located at the extreme South of Madagascar, the Androy Region has several specific characteristics compared to the rest of the country when it comes to schooling: enrollment and literacy indicators there are the lowest of the island, and school attendance is particularly higher among girls than among boys at the primary level. Seeking to understand these particularities, this research aims to study, over the long term (from the emergence of schools to today), school attendance or lack thereof in the region. With the goal of understanding school dynamics and school enrollment practices in the Androy Region, the research attempts to reveal the evolution of logics and stakes that play out in the inter-relations between education policy, school supply, and social and family demands regarding school attendance. More specifically, the aim is to understand the academic "lag" and describe school enrollment trends in the Androy Region; and reveal the plurality of social and family practices regarding schooling and their underlying logics. At the crossroads of sociology, demography, history and anthropology, this research mobilizes several quantitative and qualitative sources that are compared and criticized in light of their production conditions: discourse and statistics (both administrative and school-related) drawn from colonial period archives, the statistics and educational policy announcements by the Ministry of Education, the 1993 population census, the 2008-09 demographic and health survey, and socio-anthropological study conducted in Tsihombe district. Contrary to the frequently touted argument that school is "rejected" by the population of the region, which has long been seen as "archaïc", "backwards" and refusing "progress", this research suggests the complexity of factors explaining the "school lag" in the Androy Region. These factors include the low interest of the successive (colonial and post-colonial) powers in the development of the region, whether in regard to schools or in general; the refusal to "submit" to the central authorities and the mistrust towards the administration and the State (fanjakana) and the institution that trained its agents; local power stakes in the quest for knowledge and school culture, which vary over time and depend on individuals' and groups' positions within the social arena; and finally, the ability of school itself to enable real acquisition of academic knowledge, which varies over time and by location. Taking a multidimensional and relative vision of poverty, this research also places the importance of economic capital in families' school enrollment practices into perspective. School enrollments does not appear to be dialectically linked to the monetary and financial dimension of families' assets, but rather more intimately linked to their social capital -in the Bourdieusian sense- and the ability of educational capital to produce symbolic capital. Continuing on from this line of reasoning, while school attendance has been higher among girls than among boys since the mid-1980s, this was the case initially "by default" until a true preference for girls' enrollment emerged more recently. This recent phenomenon seems to relate to the broadening of possibilities for women in society while a focus on school attendance among girls and women's empowerment is taking stronger hold among the public authorities and even more within the international organizations actions in the region. Even though shool attendance had grown during these last fifteen years, the issue of inequalities in schooling remains : less than one out of two children accesses school in the area, and only one out of three boys, the gender gap widening and leaving boys behind, at least in statistics point of view at the regional level.

The development of human societies rests on functioning ecosystems. This thesis builds on integrated theories of linked social-ecological systems and complex adaptive systems to increase the understanding of how to strengthen the capacity of ecosystems to generate services that sustain human well-being. In this work, I analyze such capacity in human-dominated production ecosystems in Tanzania and Madagascar, and how this capacity is related to local management practices. Resilience of social-ecological systems refers to the capacity to buffer change, to re-organize following disruption, and for adaptation and learning. In Papers I and II, qualitative interview methods are used for mapping and analyses of management practices in the agroecosystem of the Mbulu highlands, Northern Tanzania. Practices such as soil and water conservation, maintenance of habitats for pollinators and predators of pests, intercropping, and landscape diversification, act to buffer food production in a variable environment and sustain underlying ecological processes. The practices are embedded in a decentralized but nested system of institutions, such as communal land rights and social networks, that can buffer for localized disturbances such as temporary droughts. Paper II compares these findings with practices in a farming system in Sweden, and suggests that similar mechanisms for dealing with uncertainty and change can exist in spite of different biophysical conditions. In Papers III and IV, interviews are combined with GIS tools and vegetation sampling to study characteristics and dynamics of the dry forests of Androy, southern Madagascar. Paper III reports on a previously underestimated capacity of the dry forest of southern Madagascar to regenerate, showing areas of regeneration roughly equal areas of degenerated forest (18 700 ha). The pattern of forest regeneration, degradation, and stable cover during the period 1986-2000 was related to the enforcement of customary property rights (Paper III). Paper IV reports on a network of locally protected forest patches in Androy that is embedded in a landscape managed for agricultural or livestock production and contributes to the generation of ecosystem services and ecosystem resilience at a landscape scale. Forest protection is secured by local taboos that provide a well-functioning and legitimate sanctioning system related to religious beliefs. In Paper V, two spatial modeling tools are used to assess the generation of two services, crop pollination and seed dispersal, by the protected forest patches in southern Androy. The functioning of these services is dependent on the spatial configuration of protected patches in the fragmented landscape and can be highly vulnerable to even small changes in landscape forest cover. In conclusion, many of the identified practices are found to make ecological sense in the context of complex systems and contribute to the resilience of social-ecological systems. The thesis illustrates that the capacity of human-dominated production ecosystems to sustain a flow of desired ecosystem services is strongly associated with local management practices and the governance system that they are embedded in, and that, contrary to what is often assumed, local management can and does add resilience for desired ecosystem services. These findings have substantial policy implications, as insufficient recognition of the dynamics of social-ecological interactions is likely to lead to failure of schemes for human development and biodiversity conservation.