Some apology seems needed in presenting a new research on the atomic weight of an element already measured with a precision which the highest living critic has emphasised as “the magnificent accuracy of Stas’ determination." Moreover, the present experiments cannot claim an accuracy to be compared with any individual series of Stas’ ratios. But, on the other hand, Stas’ atomic weight of chlorine is derived indirectly from oxygen by a series of operations which include the determination of (1) the oxygen in potassium chlorate, (2) the silver equivalent to the molecule of potassium chloride, and (3) the composition of silver chloride. Stas himself has assigned different values to these ratios at different times; e. g ., in 1860 he found that 100 parts of silver were equal to 69·103 of potassium chloride, in 1882 he found 100 of silver equal to 69·119, and in his latest work to 69·123 of potassium chloride. Therefore, although Stas’ value 35·457 (O = 16) is in satisfactory agreement with Clarke’s value 35·447 re-calculated from all the best determinations, it is possible that some constant error may occur in some part of the long chain connecting the value of hydrogen with that of chlorine, an error which would be repeated from link to link., and would become evident only when the two ends of the chain were connected up. A direct comparison between hydrogen and chlorine might not only serve to detect any systematic error in this chain of ratios, but such a comparison, inasmuch as it does not involve the probable error of other ratios, would be cœteris paribus more exact. Again, the closing of the chain between hydrogen and chlorine with reasonable accuracy would permit the accidental errors to be distributed and prevent their accumulation at the unconnected end. The accumulated “probable error” in Clarke’s recalculated value for chlorine is ±·0048; the “probable error” of our nine experiments is ±·0019.