To determine new referents, or cutoff levels for tear film hyperosmolarity in the diagnosis of keratoconjunctivitis sicca (KCS) and to assess their effectiveness in independent patient groups. A meta-analysis was performed on published data for tear osmolarity in samples of normal eyes and various subtypes of dry eye, and pooled estimates of the mean and standard deviations for normal and (all) dry eye subjects were determined. Diagnostic referents were derived from the intercept between the distributions of osmolarity in the two samples and from receiver operator characteristic (ROC) curves. This referent was tested for effectiveness of diagnosis in independent groups with normal and dry eyes. An osmolarity referent of 315.6 mOsmol/L was derived from the intercept of the distribution curves, and 316 mOsmol/L from the ROC curve. When applied to independent groups of normal and dry eye subjects a value of 316 mOsmol/L was found to yield sensitivity of 59%, specificity of 94%, and an overall predictive accuracy of 89% for the diagnosis of dry eye syndrome. Tear hyperosmolarity, defined by a referent of 316 mOsmol/L, was superior in overall accuracy to any other single test for dry eye diagnosis (Lactoplate, Schirmer test, and Rose Bengal staining), even when the other test measures were applied to a diagnosis within the sample groups from which they were derived. For overall accuracy in the diagnosis of dry eye, the osmolarity test was found to be comparable with the results of combined (in parallel or series) tests.