in the experimental receiver did not vary by a fraction of a tenth of a degree Centigrade during the whole period of the experiments. I mention this, not only to show that the change of apparent temperature indicated by the maximum index of the thermometer under compression was produced by action in the receiver, and was in no way due to changes of temperature in the atmosphere, but also to call attention to the very favourable temperature conditions offered by a ship in tropical regions for carrying out physical experiments. The choice of such a laboratory would usually be dismissed without examination, while a little experience would reverse the decision.

While in the Antarctic regions, the impossibility of determining the temperature of the water with the thermometers available gave me cause for much thought, and it was not until we had left the icy regions that a method of adapting them to the occasion occurred to me. As similar conditions might be found to obtain in the waters near the Falkland islands, which we should pass through on the way home, I caused some special Six's unprotected thermometers to be sent out to me. My intention was, if the temperature gradient was found to be inverted, to open the extremity of one of these thermometers, and send it down open. It would then be subject to the variations of temperature and pressure obtaining at the different depths. The pressure increases regularly with the depth, and produces a sliding scale for the temperature. The effect of pressure alone would be determined afterwards in an hydraulic receiver. The opportunity for using these instruments did not occur, but the mercury piezometer, which was a development of them, was successfully used by Professor Mohn in his researches in the Norwegian sea. It is usual now to use capsizing thermometers, and to overturn them by a messenger sent down the line when the thermometer has arrived at the required depth. Both of these devices are very old, and were used by Aimé in his remarkable researches in the Mediterranean between 1843 and 1846. For the investigation of the really deep water of the ocean, we require thermometers with such a scale that tenths of a degree Fahrenheit or the corresponding portion of a Centigrade degree can be determined with certainty. In the course of my researches in the Buccaneer, I passed over the point where three ridges meet, almost on the equator to the north of the island of Ascension. These ridges delimit three basins of the Atlantic, which are distinguished by the temperature of their bottom waters. The temperature at the bottom of the basin lying to the south and west of this point is 35.5°, in that to the north and west it is 360°, and to the eastward in the Gulf of Guinea it is 36-4° Fahr. It is obvious that, in order to be certain of differences of this character, the thermometers must have a wide and open scale.

We have already a large number of determinations of the distribution of temperature with depth made at different localities in the

ocean, but we have very few determinations of the distribution in the same locality at different seasons and in different years. In the Buccaneer, I made a point of repeating the serial temperature observations at all of the Challenger stations in the neighbourhood of the Guinea coast, and very considerable differences were found, especially in the surface layer of 100 fathoms in thickness. In the Gulf of Guinea I also carried out systematically the determinations of the temperature gradient in the layer of water extending from the bottom to 250 fathoms above it (Scottish Geographical Magazine, 1888, p. 13). This is a branch of inquiry which has received very little attention, but it deserves to be assiduously cultivated. We have many determinations of distribution of deep-sea temperatures, but we may say no discussion of them from a calorimetric point of view. In the Proceedings of the Royal Society of Edinburgh, 1885-1886, p. 423, I discussed a series of observations made at different dates in Loch Lomond from this point of view, giving the heat exchanges which take place in the course of the year. The beat-unit used in this discussion was the fathom-degree (Fahr.)-that is, a depth of 1 fathom heated one degree (Fahr.). If the fathom has a sectional area such that the volume of water weighs 1 pound, then the fathom-degree is the same as the ordinary British heat unit, the water-pound-degree. It is rather remarkable that, if the metre be used for measuring depth, and the Centigrade degree for measuring temperature, the resulting heat-unit for depth is the same as when the fathom and the Fahrenheit degree are used, because the fathom is 18 times the length of the metre, and the Centigrade degree is 1.8 times the Fahrenheit degree.

II.

The waters of different localities of the ocean are distinguished by the amount and nature of the saline matter dissolved in them. It has heen found that the nature of the dissolved contents can, for almost all purposes, be held to be constant, and that, therefore, a water is generally characterized by the amount of its dissolved contents, by its salinity. This salinity, within the limits met with in the ocean, varies directly with the density. The density can be determined with great accuracy even at sea by means of a suitable hydrometer. It has been found, also, that the preponderance of chlorides over other salts in sea-water is such that the salinity of a sea-water varies sensibly as the amount of chlorine which it contains. I myself always use the hydrometer, with which I can make sure of the density to one or two units in the fifth decimal place, as against distilled water of the same temperature determined at the same time and with the same instrument. The chlorine method is quite unsuitable for use at sea; first, because the quantity of chlorine is so large that the amount of water convenient for analysis is very small, and it cannot be weighed at sea. Then at sea nothing is free from chlorine-the air and everything is impregnated with chlorides; so

that, as a means of specifying and distinguishing oceanic waters, I consider the chemical method absolutely untrustworthy, except when made with all refinements in a laboratory on land. There is, of course, no comparison in the amount of time required compared with the hydro-meter method.

Many writers, in passing judgment on the hydrometer as an instrnment for the determination of the density of liquids, have only in their minds the hydrometer whose indications are determined by comparison with another or standard instrument; or by immersion in solutions, the densities of which have been otherwise ascertained. These instruments have no greater value than that of more or less carefully constructed copies of a standard, the method and the principle of the construction of which is not always given. Rightly, therefore, they prefer the density as determined by weighing a vessel filled with the liquid and comparing it with the weight of distilled water of the same temperature filling the same vessel.

The hydrometer which I constructed for the Challenger expedition, and used during the whole of it, is not an hydrometer in the above sense: it does not give comparative results; it gives absolute ones. By its means, the weights of equal volumes of the solution and of distilled water of the same temperature are determined directly. It is neither more nor less than a pyknometer, where the volume of liquid excluded up to a certain mark is weighed instead of that included up to a similar mark. In the pyknometer, the internal surface per unit of length of the stem can be made smaller than the external surface per unit of length of the stem of the hydrometer. On the other hand, the volume of the hydrometer can safely be made many times larger than that of the pyknometer, the dimensions of which must always be kept small on account of the difficulty of ascertaining its true temperature, which is always a matter of guesswork, because it is not measured directly. The temperature of another mass of liquid is measured, and the two are assumed to be identical. With the hydrometer, the liquid being in large quantity and outside of the instrument, its temperature can be immediately ascertained with every required accuracy.

Again, for every determination with the ordinary pyknometer, the weight of the liquid contained in it has to be determined by a separate operation of weighing. With the hydrometer the weight of the liquid displaced, being always equal to its own, is determined once for all by repeated series of weighings, where every refinement is used to secure the true weight of the instrument. This weight can then be increased at will by placing suitable small weights on the upper extremity of the stem. Their weight is also most carefully determined once for all, so that at any moment the total weight of the displacing instrument is accurately known. The stem of the instrument is divided over a length of 0.1 metre into millimetres, and its diameter is chosen so that 100

millimetres of it will displace 0.9 to 1 cubic centimetre; the total volume of the instrument is intended to be about 180 c.c., and the glassblower who supplies them generally fulfils this specification very closely. He loads the instrument so that it floats at 0 millimetre in distilled water of 30° C. The small weights used are in the form of spirals of aluminium wire for fractions of a gramme, and of brass or silver wire for greater weights. The system is such that any weight up to 10 grammes, increasing by steps of 0.05 gramme, can be added. It is thus possible, by making the first reading when the instrument is loaded so as just to be immersed to the lowest division (0 mm.) of the stem, to make a series of twenty-one independent determinations of the weight of twenty-one different volumes of the same liquid in a very few minutes. If the liquid is replaced in the cylinder by distilled water of the same temperature, twenty-one determinations of the weight of the same twenty-one volumes of distilled water of the same temperature can be made in as short a time, and we have as the result twenty-one perfectly independent determinations of the specific gravity of the liquid, that is, of the ratio of its density to that of distilled water of the same temperature; and the accuracy of each determination depends on nothing but the accuracy with which the original weighings have been carried out that is to say, it depends on the operation, which is capable of being performed with the greatest precision in the laboratory. In actual practice I use steps of 0.1 gramme, and I aim at having at least nine separate observations both in the liquid and in distilled water. It never happens that the successive readings in distilled water are identical with those in the liquid, but by repeated immersions in distilled water the stem is accurately calibrated, so that a correction can with safety be made for the difference of one or two millimetres between them. We then have a series of scale readings, and opposite each a pair of weights giving the weights of these identical volumes of the liquid and of distilled water, and the ratio of each pair gives the specific gravity of the liquid at the common temperature. There is no difficulty, when working in the circumstances which are alone suitable for determinations of the kind, in securing identity of temperatures within 0.1° C.

For all ordinary purposes it is not necessary to make a determination in distilled water along with each sample of sea-water or other liquid under examination. When a sufficient number of separate observations have been made in distilled water at different temperatures, we may either take the series made at the temperature nearest to that of the liquid, and compare the two after making the necessary small corrections, or we may construct a table by interpolation, giving the weights required to immerse the hydrometer up to, say, every tenth division of the stem in distilled water at different temperatures. From such a table we should be able at once to find the weights required to depress

the hydrometer to the same scale divisions as had been observed in the. liquid, and from them obtain the specific gravities. The table may, however, take another form. The weight of distilled water displaced at every observation is known by the weight of the hydrometer and added weights. If we know the volume of a cubic centimetre of distilled. water at all the temperatures covered by the experiments, we have directly the volume of the immersed portion of the hydrometer, and as such observations are made at different temperatures, we obtain the volumes of the hydrometer at different temperatures, and its rate of expansion. In constructing a table of the volumes of the hydrometer, it should always be stated what factors have been used, so that the absolute values depending on weighing alone can be recovered. For all important or normal determinations, the parallel series of observations in distilled water of the same temperature should not be omitted.

Assuming the correctness of our knowledge of the density of distilled water of different temperatures, and deducing the volume of the hydrometer from observations with it in distilled water of known temperature, we obtain directly the volume of a unit weight of the liquid, or its density; and for many purposes this is convenient.

The following table gives the results of determinations of specific gravity of samples of Mediterranean water collected in 1893 by H.S.H. the Prince of Monaco, which happened to be made at identical temperatures with different hydrometers. The specific gravities given are each the means of from nine to eleven separate observations on seawater and distilled water at the same time and at the same temperature. The greatest difference between any pair of values is 3.3 in the fifth decimal place, and the individuals of each pair depend on perfectly distinct sets of weighings, and are therefore quite independent.

It may safely be asserted that, working in this way, the specific gravity of a sea-water or similar solution can be determined with a probable error of not more than 1 in the fifth decimal place. In a water whose specific gravity is 103000, 1 in the fifth decimal place represents of the whole solid contents; so that, by the careful use