Four approaches were used to assess the effect of inundation on methane emissions from floodplain wetlands in Victoria, Australia: (i) field observations following natural rainfall events; (ii) experimental manipulation of water levels in small floodplain depressions; (iii) experimental manipulation of water levels in replicated mesocosms; and (iv) in vitro incubation of floodplain sediment under laboratory conditions. Raftery's Swamp, a large (150 ha) wetland on the floodplain of the Goulburn River, became inundated in June 1993 following autumn-winter rainfall. Methane emissions peaked (1.7+/-0.05 mmol m(-2) h(-1)) some six months later, and the methane content of sediment gas bubbles reached 59% v/v, even though the positive sediment redox potentials (176 to 243 mV) indicated that sediments were only moderately reducing. Three small (< 1 ha) depressions on the floodplains of the River Murray and Kiewa River were inundated either naturally (by rain and/or overflow from nearby rivers) or artificially by flooding at specific times of year; emissions from these sites were usually negligible after flooding in autumn or winter. In contrast, the onset of methane emission was very rapid (within 3 to 6 days) after the depressions had been flooded in summer, and the methane content of sediment gas bubbles could then reach 36% v/v. At their peak, emissions from the ephemeral wetlands were similar to peak emissions from permanent wetlands in south-eastern Australia. Emissions from replicated wetland mesocosms (4.5 m diameter, 0.9 m deep) were always very small (< 0.2 mmol m(-2) h(-1)), regardless of time of flooding, water depths, or season. In vitro incubation of wetland sediment under anaerobic conditions indicated a progressive decrease in benthic methanogenesis with sediment desiccation and exposure to air. Ephemerally inundated floodplain wetlands may be sites of significant methane emission, especially over the summer months. Moreover, the survival and rapid reactivation of methanogenic archaea after prolonged drying of wetland sediments suggests that methanogenesis is possible even from re-wetted floodplain environments that had earlier experienced an extended dry phase.