THE effect of branching on item nonresponse is an important research issue, in that it deals with an aspect in the design of mail questionnaires which is, in some situations, discretionary on the part of the researcher. There are many situations in which a researcher may choose between asking a two-step question with a branch and asking a single question with an expanded list of responses. For example, an ''own" versus "rent" question might be followed by a question for "owners" only on type of housing. Renters would be instructed to branch around the second question. Alternatively, these two questions could be combined into a single more complex question with responses of: own-single-family house; own-other type of dwelling; and rent. While this eliminates the branch, it correspondingy increases the complexity of the question, thereby creating the potential for ambiguity. This apparently leaves the researcher with a choice between the two approaches. However, the literature on item nonresponse does not indicate whether branching actually reduces item response rates, and, if it does, under what conditions. The purpose of this study is to investigate the effect of branching on item nonresponse.