The growth of avalanches and the transition to a visible discharge has been studied in air, nitrogen, argon, carbon dioxide, hydrogen and oxygen by means of an expansion cloud chamber. The effect of expansion ratio, vapour filling and type of gas on the formation of droplets was investigated. Pulse lengths of 0·11, 0·08 and 0·05 μs and pressures of 400, 580 and 680 mmHg were used in air and agreement with previous mobility measurements found. As the voltage was increased the shape of the avalanche changed, its breadth increased and finally a filamentary visible discharge was produced. The position of the avalanche head at the time of transition has been correlated with the occurrence of a constriction in the visible discharge at the same point. A single avalanche has been found to be sufficient to form a streamer. Streamer thresholds have been measured as a function of pulse length, and vapour filling. The addition of methyl, ethyl or propyl alcohol to dry nitrogen reduced the streamer threshold by about 14%. Reduction of pulse length increased the streamer threshold. Nitrogen was very similar to air in regard to the effect of pulse length, vapour filling and voltage. The streamer constriction could be moved nearer the cathode with increasing voltage and its position ( x er. being distance from cathode) was given by α x er. = 14 to 20 depending on pressure and pulse length. The X/p value of streamer threshold was unaffected by a change of pd from 150 to 780 mmHg cm. Streamers were produced in argon but overvoltages of up to 200 % were required. Effect of vapour and pulse length was similar to air. Carbon dioxide was difficult to use owing to its dissociation by ultra-violet light, and agreement with mobility measurements in the dry gas is not very good. Streamers were obtained with overvoltages. Results for hydrogen were in reasonable agreement with previous mobility measurements. Velocity measurements in oxygen were less than the previous values, but this is most likely caused by electron attachment which reduced the apparent length of the avalanche. Streamers with bright knobs at the constriction were observed. The position of the avalanche head has been calculated based on the assumption that it coincided with the neck in the streamer, and some agreement with observation particularly at high X/p values was found. Thermal energy derived from avalanche width and velocity differs by as much as a factor of 2 and a possible explanation is put forward.