The clinical effectiveness of published algorithms in correcting serum total calcium (Car) for the effects of protein, albumin, and pH was tested. Corrected calcium (Cac) values obtained by 13 of these methods were compared with values of measured free calcium (CaF) in 55 samples from normal controls and 404 samples from patients with various disorders of calcium metabolism. Three criteria were used to compare either Cac or Car with measured CaF: 1) the correlation coefficient, 2) the average absolute deviation from measured CaF of the values of CaF predicted by the linear regression of CaF on each Cac, and 3) the number of samples in which Cac or Car gave a different impression of normality than measured CaF. Application of the 13 published algorithms produced varied results, but none produced substantially better agreement between Cac and CaF than was found between Car and CaF. The application of additional algorithms derived by multiple linear regression using our data base gave slightly better results than any of the published algorithms, but many values of Cac remained which were disparate from the measured value of Cap. Correction of measured total calcium by using other concurrently obtained chemistry values does not seem to adequately predict calcium status as measured by free calcium.