The IL482 CO-Oximeter uses four wavelengths of visible light to analyze blood samples for the relative percentages of oxy-, carboxy-, reduced, and met-hemoglobin. In the analysis, the absorption at each of the wavelengths is multiplied by a matrix of four coefficients to derive the quantities of the four hemoglobin types. The normal settings of the CO-Oximeter coefficients are those for adult human hemoglobin. However, animal blood can be measured provided that the appropriate matrix of coefficients is available. Instrumentation Laboratory has provided sets of coefficients for several animal species. The company has also published a protocol for determining coefficients for other animal species. This protocol was examined using sheep hemoglobin-A blood and found to be inaccurate. The IL482 protocol is unsatisfactory because, if the initial error is large, successive iterations to determine the coefficient matrix through revision of the estimates of residual hemoglobin types do not converge. With sheep type-A hemoglobin, the use of human coefficients for the initial estimate gave a value of 6%, whereas, by chromatography, the carboxyhemoglobin (COHb) was 0.12% (i.e., a better initial estimate would be zero). When this was done, the final COHb estimate on "as-drawn" blood was within 1% of the COHb measured independently by gas chromatography. Revision of the protocol gave a markedly better accuracy, within 2% for COHb over the whole range when tested against mixtures of CO and O2 tonometered blood.(ABSTRACT TRUNCATED AT 250 WORDS)