The paper considers problems with data collection methods for social network analysis specifically so called name generators that are widely used in surveys and interviews. The issue is exampled by a research of homeless people in Pilsen, Czech Republic. Beginning with the general discussion of relational data and basic network terms, the article then covers methods of data collection with focus on name generators. A description of non-representative research design on social support networks of homeless people follows. The selected results show that personal networks are small, with high density and rather based on kin or friendship ties. Further, major parts of the ties are outside the homelessness. However, the ecological validity of the results is questioned. On the basis of the triangulation with a data from the qualitative interviews the paper concludes that the personal networks are rather “cognitive networks” than the actual social relations.