We report exploratory experiments simulating gas-driven eruptions using the CO2-H2O system at room temperature as an analog of natural eruptive systems. The experimental apparatus consists of a test cell and a large tank. Initially, up to 1.0 wt % of CO2 is dissolved in liquid water under a pressure of up to 735 kPa in the test cell. The experiment is initiated by suddenly reducing the pressure of the test cell to a typical tank pressure of 10 kPa. The following are the main results: (1) The style of the process depends on the decompression ratio. There is a threshold decompression ratio above which rapid eruption occurs. (2) During rapid eruption, there is always fragmentation at the liquid-vapor interface. Fragmentation may also occur in the flow interior. (3) Initially, the top of the erupting column ascends at a constant acceleration (instead of constant velocity). (4) Average bubble radius grows as t2/3. (5) When viscosity is 20 times that of pure water or greater, a static foam may be stable after expansion to 97% vesicularity. The experiments provide several insights into natural gas-driven eruptions, including (1) the interplay between bubble growth and ascent of the erupting column must be considered for realistic modeling of bubble growth during gas-driven eruptions, (2) buoyant rise of the bubbly magma is not necessary during an explosive volcanic eruption, and (3) CO2-driven limnic eruptions can be explosive. The violence increases with the initial CO2 content dissolved in water.¿ 1997 American Geophysical Union