Measurement of neighborhood processes and attributes rests on a decision about the proper unit on which to make those measures. It is common for researchers to aggregate survey responses to some level of geography that is a proxy for neighborhood and to treat this aggregation as the neighborhood unit. This paper examines the effects of various levels and methods of aggregation on the properties of neighborhood measures. The data come from Annie E. Casey Foundation’s Making Connections survey of over 7000 households in selected neighborhoods of ten cities. This survey, among other things, queried residents about the following neighborhood attributes using five multiitem scales: Social cohesion/trust, shared expectations for informal social control, neighborhood safety, disorder and incivility, and relations with police. Individual level scale reliability is calculated for each scale using Cronbach’s alpha to determine the internal consistency among the items. The majority of the scales prove to be reliable at the individual level (α >.7). Survey records were geocoded and the following levels of aggregation are compared: Entire Making Connections area, project defined sub-areas, census tracts, census block groups and the neighborhood named by residents. Variance components and reliability coefficients are calculated for five scales at each of these levels of aggregation. For most scales, smaller geographic units yield higher reliability coefficients. However, resident named neighborhoods also yield highly reliable aggregate measures. Finally, several strategies for constructing resident defined neighborhood units in surveys are illustrated, including analysis of the names residents give to their neighborhoods and resident drawn maps of their neighborhood boundaries.