There are two conclusions to be drawn from this trial. In the first place, this problem is universal and no country taking part maintains an acceptable standard of laboratory precision. Secondly, inspection of Fig. 1 shows that the ratio 37B/37A is significantly more constant for most constituents than are the corresponding absolute values. This implies that laboratories can compare two solutions more accurately than they can determine the absolute value of a constituent. It follows that estimations should be made whenever possible by comparing an unknown with a standard solution, a measure that would immediately eliminate one source of error.