Twenty-five coating grade roofing asphalts, derived from eleven crude oils and five major producing areas throughout the world, were exposed in replicate to the 51-9C cycle in one twin-arc and two single-arc weatherometers. These asphalts ranged in durability from 22 to 89 cycles at the 50 per cent failure level. The durability of an asphalt measured in the twin-arc weather-ometer averaged 6 per cent less than in the single-arc weatherometers, but the standard deviation was higher and the results were less reliable. Hence, the following results are based on single-arc exposures. The two single-arc machines were statistically equivalent.

Little difference was noted between exposing panels of unfilled asphalts at the same time in equivalent machines and exposing the panels at different times in the same machine. The 10 per cent and 50 per cent failure levels have about the same reproducibility.

The estimates of the standard deviations and confidence limits vary with the number of panels of an asphalt exposed and the number of times or machines in which it is exposed. In general, the machine variations contribute the major portion of the error; hence, it is recommended that single panels be exposed in a number of weatherometers or in one weatherometer a number of times.

The exposure of a single panel on only one occasion is not recommended. Under this condition, at the 95 per cent confidence level the confidence interval of the true durability is 28 per cent at the 10 per cent failure level and 33 per cent at the 50 per cent failure level. As the number of sets of single panels is increased, these limits converge rapidly. Thus, if three sets of single panels are exposed, the 95 per cent confidence range is reduced to 16 per cent at the 10 per cent failure level and 18 per cent at the 50 per cent failure level. When comparing two or more asphalts directly, it is recommended that single panels of each be exposed side by side in at least three machines or in the same machine on at least three occasions.