The catastrophic release of gas from Lake Nyos, Cameroon, in 1986 caused substantial but incomplete mixing of the stratified water column. The post-release evolution of water-column structure has been monitored through April 1992. Changes began immediately after the event as rainfall and inflow brought dilute fluid into the surface layer. Inflow and surface mixing have gradually deepened the chemocline. The Total Dissolved Solids (TDS) values in the upper 40 m of the water column have dropped from a few hundred mg/kg just after the release to <100 mg/kg. The chemocline is presently strongest at 50 m depth; 5 m below this, the TDS = 570 mg/kg. From 55 to 150 m depth is a gentle gradient in which TDS reaches 920 mg/kg. Little change in water-column chemistry has occurred in this depth interval since the release. Between 150 m depth and lake bottom at 210 m depth, a strong secondary chemocline has formed. Temperature, CO₂ concentration ([CO₂]), and TDS have all increased in the deepest layer in response to recharge by warm, mineralized water, reaching values of 25.0°C, 320 mmol/kg, and 1800 mg/kg, respectively, 1 m above lake bottom. Considering all these changes in part as a “recovery” process, it is possible to construct a model of the pre-release water column. The data indicate that the pre-release chemocline was at least 50 m deep. Above the chemocline was a dilute layer containing a seasonal thermocline; below the chemocline was probably a gradient zone(s) with correlated increases in TDS and [CO₂] and a secondary chemocline near lake bottom. Maximum values of TDS and [CO₂] calculated for pre-release bottom water are 2400 mg/kg and 430 mmol/kg, respectively, based on tritium data. From this pre-release structure, a model of the gas release is proposed that is consistent with available chemical and observational data. An important feature of the model is that disruption of the pre-existing stratification was much more extensive than previously proposed, and even the deepest water layers were involved in the event. This model is not intended to limit possible gas release mechanisms, and thus complete re-establishment of pre-1986 water-column conditions is not a prerequisite for a future release. Spontaneous instability could occur at lake bottom in <20 yr if dissolved gas pressures continue to increase in this zone by 0.5–1 bar/yr as they have for the last 6 yr.