OBJECTIVE: The objective of this study was to test the accuracy and cross-sensitivity of commercially available anesthetic gas monitors. METHODS: Using gas chromatography (GC) as a reference method, the accuracy, cross-sensitivity, and ability to recognize an erroneously selected agent were determined in the following 10 monitors for volatile anesthetics: Datex Capnomac Ultima-S, Datex Capnomac, Ohmeda 5330 agent monitor, Iris Dräger, Andros Dräger PM 8020 (all monochromatic, infrared analyzers), Nellcor N-2500E, Criticare POET II, Irina Dräger (all polychromatic, infrared analyzers), Siemens Servo Gas Monitor 120 (a piezoelectric analyzer), and Brüel & Kajer Type 1304 (a photoacoustic analyzer). Accuracy was determined at 0.5, 1, 2, and 4 times the minimal alveolar concentration (MAC) of either halothane or isoflurane in oxygen (O2). The cross-sensitivity tests were performed with 70 vol% nitrous oxide in O2, 5 vol% carbon dioxide in O2, 0.032 vol% alcohol in O2, and 70% water vapor in O2. The photoacoustic analyzer showed a higher accuracy for isoflurane than the polychromatic infrared monitors. The greatest inaccuracy with isoflurane was found in the Iris Dräger monitor, which had a maximal bias percentage by volume (vol%) of 0.09 at 0.5 MAC. (This bias was within the manufacturer's specified tolerance of +/- 0.1 vol% or 10% relative difference of reading, whichever is greater.) Irina Dräger was the most accurate analyzer with halothane (mean % bias [relative %] +/- SD, 0.9 +/- 2.0%). The greatest bias with halothane was found in the monochromatic infrared analyzers, with a maximal % bias at 0.5 MAC of 50.3% of the GC reading (12.4% with a new inner Nafion tube) found in the Datex Ultima monitor. The Siemens gas monitor showed a cross-sensitivity for water vapor (-0.248 vol%). The monochromatic infrared analyzers showed a small sensitivity to alcohol (additional deviation of 0.011 to 0.147 vol% at 2 MAC isoflurane) but no sensitivity to nitrous oxide. No cross-sensitivity was found in the polychromatic infrared and photoacoustic analyzers. An incorrect selection of anesthetic agent when using a monochromatic infrared analyzer can be fatal; for example, when using halothane and selecting isoflurane the values measured by the Datex Capnomac monitor were nearly 6 times: below the actual value (i.e., 1 vol% "isoflurane" on the display = 6 vol% halothane in reality). CONCLUSIONS: The photoacoustic measurement principle is more accurate than the other methods, although the polychromatic infrared analyzers are safer because they detect erroneously selected agents.