pers. They were the so called "citrate of magnesia" bottles made by the Whitall, Tatum Co. The water was often stored in these

bottles for several days before analyzing it.

The temperature of the water was determined by a Negretti and Zambra deep-sea thermometer, which was attached to the cable carrying the water bottle and directly beneath it. This thermometer had previously been compared with a thermometer standardized by the Deutsche Physikalische Technische Reichsanstalt.

The volumes of sea-water taken for analysis and the volumes of silver nitrate solution required to react with them were measured in a Knudsen pipette and a Knudsen burette respectively; both were made by R. Goetze, Leipzig. The former is an ordinary pipette of about 25 c. c. capacity, provided with a three-way cock at the top. This arrangement allows the liquid to pass beyond the cock when the pipette is being filled; but upon turning the cock so that the body of the pipette is in connection with the air through its third opening, the pipette empties itself and the excess of liquid remains behind. Thus an exact filling is always attained. The Knudsen burette also has a three-way cock at the top, which is used in the same way. It is filled through a side tube entering at the bottom. The lower part is graduated in terms of the standard used in Knudsen's Tables. The volume between the smallest graduation marks is about .05 c. c. and the total capacity of the burette is about 42 c. c. The burette used in this investigation was carefully standardized and the graduations were found to be equal within the limit of accuracy of the readings.

5

A silver nitrate solution, containing about 42 grams of the salt per liter, was prepared and stored in a large bottle of brown glass. This bottle, which was placed on a shelf several feet above the table, was provided with a two-hole stopper, through one hole of which a glass tubule extended from the bottom of the bottle to the inlet tube at the bottom of the Knudsen burette. The other hole of the stopper was kept closed except during the filling of the bottle.

The solution was standardized as follows. A tube of standard sea-water, obtained from the “Conseil International" at Copenhagen, was opened and t Knudsen pipette was immediately filled from it. The water was run from the pipette into a beaker, allowing one minute for drainage, and three drops of a one percent sodium chromate solution were added as indicator. Silver nitrate solution was then

5 Knudsen, Mar Gad, and London,

Hydrographical Tables etc. Copenhagen, G. E. C. ams & Norgate. 1901. v + 63 pp.

run in from the burette, at first rapidly, but at the end drop by drop, until a faint reddish tinge in the precipitate was permanent for thirty seconds. This was taken as the end point. The difference between the volume of silver nitrate used, as expressed in burette divisions, and the figure accompanying the standard sample was the value "a" of the Knudsen tables. Obviously this method of standardization shows only the strength of the solution as compared with the standard upon which the Knudsen tables are based; but since the analyses also are expressed in terms of this standard, no further knowledge of the concentration is required.

A secondary standard sea-water was prepared by diluting ordinary sea-water till approximately the same volume of silver nitrate was required to react with it as was required to react with the original standard. The exact ratio of these standards was determined with great care, since the secondary standard was the one constantly used during the investigation. At the end of the work, however, the silver nitrate was again compared with the original standard and was found to be unchanged.

In making a series of analyses, the silver nitrate solution was each day titrated against the standard water, as just described. The various samples of water were then titrated in an exactly similar way, and finally the solution was again compared with the standard. The temperature of the room, was noted during the progress of the work, but in no case did it vary enough to require any correction of the results. All determinations were made in duplicate.

Accuracy.

Since the method of analysis consists of comparing the amount of silver nitrate solution necessary to react with a definite amount of sea-water with that necessary to react with the same amount of seawater of known composition, no standardization of the pipette used for measuring the water was necessary. The amount delivered by the pipette was constant, as the time allowed for drainage was always the same.

The determination of the capacity of the burette in absolute units was not required. Only the relation of the divisions to each other had to be known, and these were found to be equal within the limit of accuracy to which the volume could be read. These readings could be relied upon to one one-hundredth of a unit. of solution used in a determination was about twenty units, the

As the total volume

proportional error was thus one in two thousand. This measurement limited the reliability of the whole analysis, which was thus trustworthy to five hundredths of one percent.

Corrections for change of temperature are unnecessary when the standardization and analyses are carried out under conditions sufficiently similar. As the limit of accuracy of reading the burette was one in two thousand, this allowed a variation in temperature of 8° C., which was a greater change than ever took place.

That no other sources of incidental error existed was shown by the facts that duplicate analyses always agreed to one part in two thousand or better, and that comparisons between the silver nitrate solution and the standard water always showed the same ratio to exist.

The positions were usually determined by sighting conspicuous objects on shore.

The "depth below surface" and the "depth of bottom" were measured directly on the iron cable which carried the bottle. For the positions marked with an asterisk the depth of the bottom was not determined, but is that marked on the chart of the "Bermuda Islands❞— issued by the Hydrographic Office, Washington, D. C., and corrected