The fit test panels currently used for respirator research, design, and certification are 25-subject panels developed by Los Alamos National Laboratory (LANL) and are based on data from the 1967 and 1968 anthropometric surveys of U.S. Air Force personnel. Military data do not represent the great diversity in face size and shape seen in civilian populations. In addition, the demographics of the U.S. population have changed over the last 30 years. Thus, it is necessary to assess and refine the LANL fit test panels. This paper presents the development of new respirator fit test panels representative of current U.S. civilian workers based on an anthropometric survey of 3,997 respirator users conducted in 2003. One panel was developed using face length and face width (bivariate approach) and weighting subjects to match the age and race distribution of the U.S. population as determined from the 2000 census. Another panel was developed using the first two principal components obtained from a set of 10 facial dimensions (age and race adjusted). These 10 dimensions are associated with respirator fit and leakage and can predict the remaining face dimensions well. Respirators designed to fit these panels are expected to accommodate more than 95% of the current U.S. civilian workers. Both panels are more representative of the U.S. population than the existing LANL panel and may be appropriate for testing both half-masks and full-face piece respirators. Respirator manufacturers, standards development organizations, and government respirator certification bodies need to select the appropriate fit test panel for their particular needs. The bivariate panel is simpler to use than the principal component analysis (PCA) panel and is most similar to the LANL panel currently used. The inclusion of the eight additional facial measurements allows the PCA panel to provide better criteria for excluding extreme face sizes from being used. Because the boundaries of the two new panels are significantly different from the LANL panel, it may be necessary to develop new respirator sizing systems. A new five-category sizing system is proposed.