Abstract The analyses of 18 volcanic gas samples collected over a two-hour period at 1075°C from Erta' Ale lava lake in December 1971 and of 18 samples taken over a half-hour period at 1125–1135°C in 1974 display moderately to intensely variable compositions. These variations result from imposed modifications caused by (1) atmospheric contamination and oxidation, (2) condensation and re-evaporation of water during collection, (3) analytical errors, and (4) chemical reactions between the erupted gases and a steel lead-in tube. Detailed examinations of the analyses indicate the erupted gases were at chemical equilibrium before collection. This condition was partially destroyed by the imposed modifications. High-temperature reaction equilibria were more completely preserved in the 1974 samples. Numerical procedures based on thermodynamic calculations have been used to restore each analysis to a composition representative of the erupted gases. These procedures have also been used to restore the anhydrous mean compositions reported for two series of collections taken at the lava lake in January 1973. The corrected analyses for 1971 and 1973 have similar compositions (69.6–71.3% H 2O, 1.6–2.4% H 2, 17.8–19.4% CO 2, 0.8–1.6% CO, 4.9–8.8% SO 2, 0.2–0.5% S 2, and 0.4–1.0% H 2S); those for 1974 were somewhat different (77.1% H 2O, 1.6% H 2, 11.7% CO 2, 0.5% CO, 7.4% SO 2, 0.3% S 2, 0.9% H 2S and 0.4% HCl). The O 2 and S 2 fugacities of all restored analyses are similar when compared at the same temperatures. O 2 fugacities are close to those of the quartz-magnetite-fayalite buffer. The restored analyses show no evidence of significant short-term (minutes, hours) variations in the compositions of the gases released from the lava lake, and evidence of long-term variations is limited. The restored analyses indicate the O 2 and S 2 potentials of the lava lake remained nearly constant from 1971 to 1974. However, there is a relative decrease in CO 2 between the 1973 and 1974 corrected compositions. This change is qualitatively compatible with magma outgassing controlled by solubility differences and could reflect a longer term trend. The chemistry of the Erta' Ale gases is consistent with models invoking the generation of basalts resembling mid-ocean ridge tholeiites along axial zones of crustal spreading in the Afar Triangle. The CO 2-rich vesicle gases of submarine basalts and the relatively H 2O-rich volcanic gases of Erta'Ale appear to be end-member compositions of an outgassing trend controlled largely by the effect of pressure on volatile solubilities in tholeiitic basalts from spreading zones. By analogy with submarine basalts H 2O and sulfur in the Erta' Ale gases are mainly contributed by magma outgassing down to shallow depths (50–250 m); CO 2 contributions probably come from larger volumes of magma down to depths of at least 1000 m.