How does great power security assistance affect civil-military relations in developing states? Great powers use security assistance in the form of arms, equipment, and training not only to build capacity but also to impart values and norms in developing militaries. The United States and other liberal powers, for example, explicitly try to promote civilian control of the military and respect for human rights. Yet security assistance frequently seems to produce norm-violating militaries instead. Policymakers tend to chalk failures up to insufficient emphasis on socialization, while scholars favor rationalist arguments that stress interest misalignment between providers and recipients. By contrast, I argue that norm violations tend to occur not because assistance fails to impart norms, but because it does not impart them quickly enough relative to increases in military capacity and because—in the case of liberal providers—it imparts conflicting norms. Moreover, counter-messages from competing providers dilute the efficacy of socialization attempts. In this dissertation, I argue that we must disaggregate how security assistance changes military beliefs as well as military behavior. Accordingly, the first part of the dissertation examines the conditions under which security assistance leads to shifts in military beliefs. I argue that security assistance can socialize recipient militaries to adhere to norms such as respect for human rights and civilian control of the military, but such norm-abiding behavior is likely to emerge only under certain conditions. First, because it is hard to change beliefs about standards of appropriate behavior in the security domain, socialization requires extensive military training and interaction over time. Moreover, socialization will only occur when there are no competing norms being promoted by other providers. Even if socialization occurs, however, there is no automatic guarantee that behavior will change because behavior is the output of multiple forces including norms, interests, and capabilities. In the second part of the dissertation, I argue that a powerful feature of foreign military training is its ability to alter all three elements of decision-making. However, foreign military training can strengthen military capacity faster than it socializes norms of restraint. When organizational interests are threatened, militaries with enhanced capabilities from security assistance may be more likely to intervene politically or abuse human rights. Second, liberal assistance imparts norms with potentially contradictory implications for behavior. Conflict between liberal norms can arise when political leaders, who militaries are supposed to obey, order the military to harm the population that they are supposed to protect. The contradiction can lead to perverse behavioral outcomes by reducing support for both of the conflicting norms. The dissertation uses micro-level, sub-national, and cross-national data to test the arguments both between and within countries. My empirical focus is on Africa, where many states receive assistance from multiple providers. To evaluate the effects of socialization on belief change, I conduct an original survey of the Liberian military, which the United States rebuilt after Liberia’s civil war ended in 2003. The survey includes an experiment in which soldiers hear a scenario about civilians ordering the military to repress protests, engendering conflict between the two liberal norms. I find that higher levels of training strongly increase support for liberal norms. The experimental evidence suggests, however, that exposure to norm conflict leads to reduced support for both norms and the effects are strongest among soldiers with more US training. To examine the effects of counter-messages from competing providers, I conduct a case study of the Tanzanian military, which Canada and China concomitantly tried to train during the 1960s. Canada attempted to build a liberal military in Tanzania, while China sought to shape a socialist military (China prevailed). The case study draws on hundreds of archival documents from the Canadian military training mission to process trace Canada’s influence and socialization attempts. Finally, to test the link between security assistance and military behavior, I build a new dataset of military involvement in politics and human rights abuses across Africa from 1999 to 2010. Quantitative analyses demonstrate that US foreign military training corresponds to less military interference in politics and repression. These effects are strongest at higher levels of training and training has stronger effects on military behavior than other forms of security assistance. But there is a catch: rapid increases in training appear to drive worse outcomes. By showing the ways, some of them unexpected, that security assistance can change military beliefs and behavior, this study illuminates both the promise and pitfalls of security assistance as a tool of statecraft.