The history of criminal statistics bears testimony to a search for a measure of “criminality” present among a population, a search that led increasingly to a concern about the “dark figure” of crime—that is, about occurrences that by some criteria are called crime yet that are not registered in the statistics of whatever agency was the source of the data being used. Contending arguments arose about the dark figure between the “realists” who emphasized the virtues of completeness with which data represent the “real crime” that takes place and the “institutionalists” who emphasize that crime can have valid meaning only in terms of organized, legitimate social responses to it. This paper examines these arguments in the context of police and survey statistics as measures of crime in a population. It concludes that in exploring the dark figure of crime, the primary question is not how much of it becomes revealed but rather what will be the selective properties of any particular innovation for its illumination. Any set of crime statistics, including those of survey research, involve some evaluative, institutional processing of people’s reports. Concepts, definitions, quantitative models, and theories must be adjusted to the fact that the data are not some objectively observable universe of “criminal acts,” but rather those events defined, captured, and processed as such by some institutional mechanism.