Survey respondents may start the survey but fail to complete it, resulting in breakoff. This behavior occurs in interviewer-administered surveys but it occurs at high rates in web surveys, necessitating further attention. Breakoff is one of multiple response behaviors, each of which can be affected by unique and common causes. We present a framework within which to study different response behaviors, unit nonresponse, breakoff, item nonresponse, and measurement properties, and the factors affecting them in web surveys. Theories within this framework for explaining breakoff are proposed and tested. Features within the survey that are only seen after starting are predictive of breakoff rate, distinguishing this behavior from unit nonresponse. Education, a proxy for respondent cognitive sophistication, was significantly related to breakoff rates. Furthermore, respondents who broke off did not seem inattentive, supporting further efforts in their retention. Question and questionnaire design characteristics were predictive of survey breakoff, and placed in the context of the survey response process model and respondent burden. Implications for survey design and needs for future work are discussed.