were fixed each at 0.165 m high which would have caused large entrance and exit flow resistances to mask the variation of wall friction resistance with cavity width. It was observed that the predicted flow rates for Type 3 Chimney were higher than those for Type 2 Chimney. For example, at a cavity width of 0.6 m, the flow rate through Type 2 Chimney was 0.162 m is whereas the flow rate through Type 3 Chimney s 0.190 m3/s. The difference in the flow rate resulted from the difference in the net height of the storage wall. For Type 2 Chimney, part of the storage as the cavity width) became the inlet opening and so less heat was available to induce buoyancy. Besides, this part of storage wall adjacent to the chimney bottom had a larger stack height and so would be more effective in inducing buoyancy than the rest of the storage wall for the same wall surface area. This was confirmed by simulation of the same type of chimney (Type 3) but with no heat gain for part of the storage all opposite to the inlet opening (i.e. Type 3 modified in Fig. 11), which generally produced a similar flow rate to that for Type 2 Chimney with the same cavity width. However the optimum cavity width decreased to 0.55 m although the variation in the flow rate for cavity width between 0.5 and 0.6 m was less than 1%. It is also seen that the effect on the flow rate of cavity width larger than 0.6 m was more significant for Type 3 Chimney (both original and modified) than Type 2. That is to say, if air flows into the chimney through the storage wall, the flow rate would not decrease much when the cavity width is larger than the optimum alue. In contrast, if incoming air flows through glazing, the decrease in the flow rate would be significant when the cavity is wider than the optimum