Detection of single, fluorescently labeled biomolecules is providing a powerful approach to measuring molecular transport, biomolecular interactions, and localization in biological systems. Because the biological molecules of interest rarely exhibit sufficient intrinsic fluorescence to allow observation of individual molecules, they are usually labeled with fluorescent dye molecules, fluorescent proteins, semiconductor nanocrystals or quantum dots, or fluorescently doped silica or polymer nanospheres to allow their detection. Differences in the photophysical and spectral properties of different labels allow one to identify individual molecules by distinguishing their corresponding labels. A simple approach to measuring fluorescence spectra of individual fluorescent labels can be implemented in a standard wide-field fluorescence microscope, where a grating or prism is incorporated into the path from the microscope to an imaging detector to disperse the emission spectrum. In this work, principal components and cluster analysis are applied to the identification of fluorescence spectra from single fluorescent labels, with statistical tests of the classification results. Spectra are determined from diffracted images of fluorescent nanospheres labels, where emission maxima are separated by less than 20 nm, and of single dye-molecule labels with 30 nm separation. Clusters of points in an eigenvector representation of the spectra correctly classify known labels (both nanospheres and single molecules) and unambiguously identify unknown labels in mixtures.