It was found that the amount of sulphate of alumina employed was more important than any other factor in determining the bacterial efficiency, and special experiments were made to establish the effect of more and of less coagulant than used in the ordinary work. These experiments were made upon the Warren filter during May, and with the Jewell filter during June. The monthly averages for these months are thus abnormal and are not to be considered. The remaining six months for each filter may be taken as normal and as representing approximately the work of these filters under ordinary careful working conditions. During the six months when the Warren filter was in normal order the raw water contained 11,531 bacteria and the effluent 201, the average bacterial efficiency being 98.26 per cent. The bacterial efficiency was very constant, ranging only, by months, from 97.48 to 98.96 per cent. During the same period a sand filter receiving the same water yielded an effluent having an average of 105 bacteria per cubic centimeter. The Jewell filter, for the six months in which it was in normal order, received raw water containing an average of 11,481 bacteria and yielded an effluent containing an average of 293, the bacterial efficiency being 97.45 per cent, and ranging, in different months, from 93.23 to 98.61 per cent. WASTING EFFLUENT AFTER WASHING FILTERS. After washing a mechanical filter the effluent for the first few minutes is often inferior in quality to that obtained at other times, and if samples are taken at these times and averaged with other samples taken during the run, an apparent efficiency may be obtained inferior to the true efficiency. To guard against this source of error, whenever samples have been taken at such times, the average work for the day has been taken, not as the numerical average of the results, but each sample has been given weight in proportion to the amount of time which it could be taken as representing; so that the results represent as nearly as possible the average number of bacteria in the effluent for the whole run. As a matter of fact, however, comparatively few samples were taken during these periods of reduced efficiency, and thus most of the results represent the normal efficiency exclusive of this period. A study has been made, however, of the results of examinations of samples taken directly after washing, somewhat in detail. The following is a tabular statement of the average results obtained from each filter by months, including only the results obtained on those days when samples were taken within twenty minutes after washing, the results of other days being excluded. AVERAGE NUMBER OF BACTERIA IN EFFLUENT.