To increase the likelihood of response, many survey organizations attempt to provide sample members with a mode they are thought to prefer. Mode assignment is typically based on conventional wisdom or results from mode choice studies that presented only limited options. In this paper we draw heavily on research and theory from the mode effects and the survey participation literatures to develop a framework for understanding what characteristics should predict mode preferences. We then test these characteristics using data from two different surveys. We find that measures of familiarity with and access to a mode are the strongest predictors of mode preference and measures of safety concerns, physical abilities, and normative concerns are unexpectedly weak predictors. Our findings suggest that variables that may exist on sample frames can be used to inform the assignment of “preferred” modes to sample members.