A LAST WORD. 347 



thermometer, at least once a day ; the direction and force of the 

 wind three times a day — first, middle, and latter part — at the hours 

 eight P.M., four A.M., and noon ; the variation of the compass 

 occasionally ; and the set of the current whenever encountered. 

 These observations, to be worth having, must be accurately made ; 

 and as every thermometer and every barometer has its sources of 

 error, consequently, every ship-master who undertakes hereafter to 

 co-operate with us, and keep an abstract log, should have his ba- 

 rometer and thermometer accurately compared with standard in- 

 struments, the errors of which have been accurately determined. 



999. These errors the master should enter in the log ; the in- 

 struments should be numbered, and he should so keep the log as 

 to show what instrument is in use. For instance, a master goes 

 to sea with thermometers Nos. 4719, 1, 12, etc., their errors hav- 

 ing been ascertained and entered on the blank page for the pur- 

 pose in the abstract log. He first uses jSTo. 12. Let it be so stated 

 in the column of Remarks, when the first observation is recorded, 

 thus : Thermometer No. 12. During the voyage, No. 12 gets bro- 

 ken, or for some reason is laid aside, and another, say 4719, is 

 brought into use. So state when the first observation with it is 

 recorded, and quote in the column of Eemarks the errors both of 

 Nos. 12 and 4719. Now, with such a statement of errors given in 

 the log for each of the instruments, according to the number, the 

 observations may be properly corrected when they come up here 

 for discussion. 



1000. It is rare to find a barometer or a thermometer that has 

 no error, as it is to find a chronometer without error. A good 

 thermometer, the error of which the maker should guarantee not to 

 exceed in any part of the scale one degree, will cost, in the United 

 States, not less than $2, perhaps $2 50. 



1001. The errors of thermometers sometimes are owino- to in- 

 equalities in the bore of the tube, sometimes to errors of di\^sion 

 on the scale, etc. Therefore, in comparing thermometers with a 

 standard, they should be compared, at least, for every degree be- 

 tween melting ice and blood heat. 



