This paper examines when conceptual misalignments in dialog lead to consequential miscommunication. Two studies explore misunderstanding in survey interviews of the sort conducted by governments and social scientists, where mismeasurement can have real social costs. In 131 interviews about tobacco use, misalignment between respondents' and researchers' conceptions of ordinary expressions like “smoking” and “every day” was quantified by probing respondents' interpretations of survey terms and re-administering the survey questionnaire with standard definitions after the interview. Respondents' interpretations were surprisingly variable, and in many cases they did not match the conceptions that researchers intended them to use. More often than one might expect, this conceptual variability was consequential, leading to answers (and, in principle, to estimates of the prevalence of smoking and related attributes in the population) that would have been different had conceptualizations been aligned; for example, fully 12% of respondents gave a different answer about having smoked 100 cigarettes in their entire life when later given a standard definition. In other cases misaligned interpretations did not lead to miscommunication, in that the differences would not have led to different survey responses. Although clarification of survey terms during the interview sometimes improved conceptual alignment, this was not guaranteed; in this corpus some needed attempts at clarification were never made, some attempts did not succeed, and some seemed to make understanding worse. The findings suggest that conceptual misalignments may be more frequent in ordinary conversation than interlocutors know, and that attempts to detect and clarify them may not always work. They also suggest that at least some unresolved misunderstandings do not matter in the sense that they do not change the outcome of the communication-in this case, the survey estimates.