Total plasma carbon dioxide (TCO2) concentrations were measured in Standardbred horses to determine criteria to discriminate between normal horses and horses with excessive TCO2 concentrations on raceday. TCO2 concentrations from stabled horses were distributed normally with a mean of 30.2 mmol/L and a standard deviation of 1.2 (n = 192) while pre-race TCO2 concentrations were not normally distributed. The results indicate that about 50 horses per million are likely to have TCO2 concentrations greater than or equal to 35 mmol/L and that it is extremely unlikely that a normal horse would have a resting TCO2 concentration above 36 mmol/L. These values were associated with sensitivities of 67% and 59%, respectively, and with a specificity of 100%. TCO2 concentrations were relatively stable in blood samples stored at 4 degrees C for 4 days, whereas the TCO2 in specimens stored at room temperature (25 degrees C) and at ambient temperature (16-28 degrees C) declined progressively over 5 days. The accuracy and precision of the Beckman EL-ISE Auto Analyser were acceptable and within the manufacturers specified range. Paired specimens analysed using a Beckman EL-ISE Auto Analyser and a Kodak Dry Chemistry Analyser were not significantly different. However, the measurements made using the Kodak Dry Chemistry Analyser averaged 0.5 mmol/L higher than those analysed on the Beckman EL-ISE. The significance of these sources of variation in TCO2 concentration in relation to the testing of horses for 'milkshake' administration are discussed.