Objectives. A primary justification for collecting U.S. racial statistics is the need to monitor racial discrimination. This article aims to show how analyses of Hispanics—who may officially be of any race—tend to miss discrimination based on racial appearance by relying on data that instead capture racial self-identification, a different aspect of race that often does not correspond.

Methods. The study analyzes 60 qualitative interviews with Dominican and Puerto Rican migrants in the New York metropolitan area. It employs multiple measures to represent theoretically distinct aspects of the lived experience of race.

Results. Respondents interpret the Census race question in different ways corresponding to different aspects of race, which often do not match one another. Although respondents experience discrimination on the basis of phenotype, their racial self-identification is a poor proxy for measuring their racial appearance.

Conclusions. Scholars need to develop a language of race that communicates the multiplicity of social processes involved. Social surveys must provide measures of these multiple components, including interviewer observations of racial appearance, to monitor discrimination on the basis of phenotype within Hispanic groups.