This paper covers the first comprehensive study of the distribution of heat absorption in a natural-gas-fired, water-cooled, steam-boiler furnace. The use of a fuel free of incombustible residue revealed factors in furnace performance because ash was not present to mask the influence of such factors. The tube-surface temperatures on the furnace walls were essentially constant at any given rating and excess air, and were completely reproducible. After a lapse of more than a week a check test was made, and it was found that the indicated tube-surface temperatures were within a degree or two of their previous values. The obvious conclusion is that the shape of the “flame” envelope (invisible with natural gas) was independent of load, at least with fixed burner position and at fixed mixing-vane angles. There is no reason to suppose that the flame envelope and gas path in a furnace fired with pulverized coal would not also be the same at the same rating and excess air and with fixed positions of burners and mixing devices. This factor has not been apparent on previous tests because of the masking effect of variable ash deposits on the furnace tube surfaces. The test boiler, in normal service, was used at high load factor and high capacity. The mixing vanes and flows in the burners had been adjusted properly for this load condition after extended experience. Readjustment for each test load would have been time-consuming so vane adjustment was not attempted. When isotherms were drawn on plots of the furnace walls it was strikingly apparent that the pattern was similar at all loads with eight burners and with four burners. The gradient, or space between isotherms, varied with rating but the pattern was constant. This constancy of pattern indicated that as long as burners and mixing devices were not readjusted, the hot gases would impinge on the walls at the same spots at all ratings and, therefore, that the path of gases through a furnace is determined by the burner position and the geometry of the furnace and not significantly by rating or excess air. This factor was not apparent from the results of the previous pulverized-coal-fired tests because of the masking action of ash deposits on the surface of the furnace tubes. The usual flame, as observed in coal and oil-fired furnaces, was completely invisible in this natural-gas-fired furnace. Computations of radiant transfer indicate that about 80 per cent of the heat absorbed by the furnace was transferred by radiation. Estimates for luminous flames indicated that from 95 to 100 per cent of the heat absorbed in the furnace is transferred by radiation. In the first attempts to measure the furnace gas temperature at the burner level several impressive occurrences indicated that we were dealing with much higher temperatures than were expected. The porcelain shields, used to reduce radiation loss, shattered upon reaching the hot gases. The platinum and even the platinum-rhodium wires of the thermocouple melted. The analysis of the data indicated, by extrapolation, that temperatures of 3400 F were common within 10 ft of the burner. Such temperatures indicated effective mixing of fuel and air and very rapid combustion. Conservative estimates of heat release indicated values of 160,000 Btu per cu ft of space utilized for the completion of combustion. When the velocities of gas and air were lowered, for testing at lower loads, temperature measurements indicated that the mixing action was not as effective because the burner vanes were not readjusted to compensate for the reduced velocities. At partial loads studies of the gas temperature indicated that the impingement of the hot gases on the rear or target wall produced sufficient mixing to complete combustion at the burner level. No unburned products were found in the gas samples taken above the burner level at any load. It can be concluded, therefore, that with sufficient hot-air pressure and with a well-designed mixing burner, natural gas can be burned rapidly and completely at high heat-release rates. The temperatures of the gases leaving the furnace at constant steamload decreased with increase in excess air but the heat content of the mixture increased. The patterns of the isotherms of gas temperature at the furnace outlet vary markedly with the load. The addition of a widely spaced screen of tubes in front of the rear wall reduced the gas temperatures leaving the furnace from 50 to 100 F but also changed the pattern of isotherms so that the pattern was essentially the same at all loads. Probably the furnace-wall screen not only absorbed additional heat but also aided in the mixing action of the gases impinging on the wall and thus made the isotherms more uniform.