The extreme poor in America face significant problems in their daily lives. We explore the role of social capital in a homeless community in eastern Connecticut. While researchers have studied social capital in homeless populations using qualitative methods, there is very little quantitative evidence of if or how well the homeless trust other homeless people or are trusted by the nonhomeless in their communities. We explore the questions of altruism and trust within and between homeless people using behavioral games commonly used in economic research. We find that individuals currently experiencing homelessness are more trusting in general, compared to individuals not currently experiencing homelessness. We also find that those who are not currently homeless, but have been in the past, are more trusting of the homeless population than the non-homeless. We argue that these results have important policy implications and that economists and other social scientists should work more with the extreme poor to understand their constraints and needs.