3 o8 RESPIRATION 



the presence of a fire or of central heating by warm air, the 

 amount of wind, etc. But even if there is little or no cause of ex- 

 change of air before the gas escape begins, the escape itself will 

 furnish a cause, since the gas is much lighter than air, so that air 

 to which gas has been added will tend to pass out by the roof. 

 Hence even under conditions least favorable to ventilation, the 

 gas can never accumulate to more than a very limited concentra- 

 tion in the air of a room. 



Another complication in connection with gas escapes is that the 

 gas may or may not mix evenly with the air of a room. Gas escap- 

 ing from a burner passes straight upwards to the roof and there 

 spreads. I found that unless the temperature of the windows and 

 walls was below the air temperature of the room the gas never 

 came down again to any very great extent. With a very rapid 

 escape of gas, as when a burner was completely removed or a 

 pipe cut, this was very marked. It was impossible to obtain a 

 poisonous atmosphere at the ordinary breathing level, but there 

 was a heavy concentration of gas near the roof. The danger of 

 poisoning was to persons in the floor above, and not to those in the 

 room where the escape was occurring. Near the floor level, how- 

 ever, a curious phenomenon was observed. The gas actually 

 present in the air was found to be nearly pure hydrogen. This 

 showed that it was only by diffusion, and not by convection cur- 

 rents, that gas had penetrated downwards. Hydrogen, being much 

 more diffusable than any of the other constituents of lighting gas, 

 had diffused downwards much more rapidly ; and in general it was 

 found that the hydrogen in lighting gas separates off by diffusion 

 very readily, leaving a mixture containing more of CO and the 

 other heavier constituents of the gas. At night, when the windows 

 were cold, and the tendency to convection currents down them 

 was consequently strong, mixture of the gas by convection was 

 much more apt to occur, especially if the escape was at a moderate 

 rate. There was consequently more danger at night to persons 

 sleeping in the room. 



When the percentage of gas was determined at intervals in the 

 air of a room with gas continuously escaping from a burner and 

 mixing by convection currents down the windows, I found that, 

 if the conditions of wind, etc., remained constant, the percentage 

 became constant after a certain time which depended on the size 

 of the room among other conditions, and might vary from about 

 one to three hours according to the size of the room, rate of gas 

 escape, amount of wind, etc. The maximum percentage obtained 



