Survey breakoff is less likely in online panels than in one-time web surveys – due, at least in part, to early panel attrition by less-committed panel respondents (see e.g., McCutcheon, Rao, and Kaminska, 2014) . Breakoff in online panels, however, remains problematic, and may serve as a prelude to panel attrition. Interest in breakoff involving internet survey respondents has been accelerated by the relatively recent availability of paradata collection methods for web surveys (Peytchev 2009). In addition to respondent and survey design characteristics, it is now relatively easy to obtain data such as the amount of time taken per survey item (response latency), number of response changes, time of day that survey breakoff occurs, and other factors that may contributors to survey breakoff. Internet panels offer the additional advantage of accumulated information regarding respondent characteristics. This study examines data from 18 waves of the Internet component of the Gallup Panel, a multi-mode, probability panel of American households. In addition to demographic characteristics and survey design factors (e.g., question complexity, topic, number of questions, grid patterns, survey length), the analysis will include self-reports on Internet usage and computer sophistication, paradata and, especially, device type to explore factors related to survey breakoff. Preliminary analysis indicates that while long-term panel members are less likely to breakoff, past behavior and respondent-reported computer usage are important predictors – those who have broken off in the past appear, and those who report less computer usage, appear to be more likely to breakoff again. The study will consider the potential use of predictive models for survey breakoff in developing possible adaptive design (Groves and Heeringa 2006) interventions for Internet surveys that may prove useful in delaying survey breakoff.