96 FRIDTJOF NANSEN. M.-N. KI. 
minations, be about 35'1 %/00 (computed by means of Knudsen’s tables, see 
below). This water is consequently different from the bottom-water of the 
Norwegian Sea, if these determinations be correct. For.this reason it 
was assumed by the writer that there could be no open communication 
between the deep basin of the North Polar Sea and that of the Nor- 
wegian Sea; otherwise the heavy bottom-water would flow into the latter. 
same time referring the reader to the original Memoir now in question (Oceanography of 
N. P. Basin, pp. 40—58, 237—242). 
The instrument used (N. & Z. No. 75,680) had a correction of + 0'07° C. in May 
1893, and of + 0.00” C. in March 1898. The diminution of the correction is evidently 
due to the secular contraction of the glass, and it may consequently be expected that 
the actual error of the instrument had been somewhere between these limits during the 
expedition if the thermometer had not been lying, when out of use, in a room with 
low temperature. By comparison with another thermometer (N. & Z. No. 75,684, the 
zero-corrections of which were determined in May 1893, in October 1900, as well as 
by numerious comparisons with other thermomejers during the expedition) it was found, 
however, that the correction was —0'06 ? C. on July 29, 1895, and this was accepted as 
the probable one for that year, although in reality it might possibly have been anywhere 
between this and + 0'00” C. But the inaccuracy thus arising cannot consequently 
amount to many hundreths of a degree and the correction will on the whole tend to 
give a too low temperature. On February 5, 1896, Dr. Blessing determined the zero- 
correction of this instrument to be —o0'4? C. but as he himself says (see Oceanography 
etc., p. 58, foot-note) the method used for the determination cannot have been trustworthy. 
The method employed is described in the Memoir (p. 238); as is there pointed out 
it must inevitably have given a minus-corrections at least 0'28° C. too big, and thus 
the zero-correction found by Blessing on February 5, 1896, cannot at any rate have 
been greater than —o'12? C. But as Blessing himself says the determination cannot 
have been trustworthy, because he had to "keep the thermometer in his hand” while 
it was being reversed, (and consequently had to lift it out of the melting hoar-frost 
into the warm air before reversing); the real error was therefore less, and therefore 
approaching the error actually assumed. It is a well-known fact that if the zero-point 
determination be not very carefully done, a much too great minus-correction is generally 
obtained. Pettersson’s assumption that the zero-correction of —0'4? C. found by Blessing 
may have been correct, is consequently very improbable. There is, however, another 
circumstance, which might tend to make the temperature-readings too low, in cases like 
that mentioned by Pettersson, vis. that the temperature of the air on the day in 
question was low, —40° C.; but this, the only possible cause of an appreciable in- 
accuracy, is not mentioned by Pettersson. If the thermometer, after coming up from 
the water, had had time to assume the airtemperature before the reading was taken, 
its indication would have to be corrected by about + 0'55° C. The thermometer was, 
however, as a rule kept below the water-surface, until convenient for taking the reading, 
it was then hauled up and read off as soon as possible; but as this had to be done 
with a lens, and during the dark winter (as on the above Occasion) by the light of a 
lantern, it might take some time; thus the broken off mercury might have been 
somewhat cooled in the air, and accordingly have given too low indications. In 
making the corrections, however, it was assumed that the thermometer had only been 
cooled down to the temperature of the upper water-strata, and in the case mentioned 
by Pettersson the reading was therefore reduced by —o'o4° C. and a temperature of 
—o'69° C. thus obtained for 1900 metres on Dec. 2, 1895. Consequently the real 
temperature cannot have been lower than this value, provided that the thermometer was 
correctly read oft, but there is a possibility or even a probability that it may have been 
somewhat higher. As already pointed out, the deep-sea temperatures taken with the re- 
