Ivi REPORT 1869. 



Standard readings have been reduced to 32° ought in this case to be ex- 

 amined. When a Standard reading is evidently wrong it ought to be noted 

 as such on the curve, and should not be made use of either in calculating the 

 residual correction or the monthly mean difFerence between the Standard 

 and Barograph readings. By applying both the above tests any error in the 

 calculation of the residual correction will be detected, and ought to be remedied 

 at once. Having by this means obtained an accurately calculated residual 

 correction, the accuracy with which this is applied to the various hours oiight 

 to be tested by the Kew assistant, who, obscuring fi'om his view the coliamn 

 which embodies the values after the residual correction has been applied, 

 should independently apply it on a separate piece of paper, thus producing 

 a new column of corrected pressure, which ought to be compared with the old 

 one ; any error discovered by this comparison should be corrected at once. 

 Before leaving this subject, it ought to be stated that the tabulating instru- 

 ment as well as the subsidiary ivory scale are so arranged as always to ensure 

 reading the proper point of the curve for every odd hour. 



Should anjr portion of the curve be too faint for measurement with the 

 ordinary tabulating instrument, but not too faint for measurement with the 

 ivory scale, it ought to be measured with this scale, applying to the mea- 

 surements so obtained their own appropi-iate residual correction. Such read- 

 ings ought to be specially noted in the tabulation forms. 



Should any part of tlie curve be deficient from ivant of light or any other 

 cause, it ought not to be inked in. If the deficiency be in the border of the 

 temperature curve, it will be possible to correct it, but if it be in the baro- 

 metric curve, this cannot be done. 



All curves in which the clock has stopped or the cJoclc-stop has been out of 

 action, should be personally inspected by the Director of the Central Obser- 

 vatory, in order that he may ascertain if the tabulations have been properly 

 made. 



Finally, it is right to state that the accitracy of the method of checking 

 the tabulated values now described, has been practically confirmed by the 

 month of October at Ivew being independently measured by two observers. 

 The results of the two sets, when compared together, are found to differ very 

 slightly from one another, the greatest difference being -008 in., which may 

 be supposed to denote a difference in each of -004 on either side of the truth. 

 This extreme difference only occurs three times in the course of the month, 

 that is to say, in 744 observations. 



The method of subsidiary tabulations now described is thus proved to be 

 effective in discovering the larger errors that the observer is liable to make 

 when he measures the curve. But to ensure an efficient standard of correct- 

 ness, it is not only necessary that the larger errors should be altogether 

 eliminated, but smaller ei-rors should be reduced to a minimum. Thus an 

 observer might be sufficiently cautious in reading his scale to make no large 

 error, yet sufficiently incautious to read erroneously when he came to the 

 third figure of decimals. For rough results such an observer might bo 

 reckoned a good one, but for the more delicate class of investigations his 

 figures would be of less value. 



The only way of perfectly eliminating this class of errors is for two inde- 

 pendent observers to make separate measurements, each with a tabulating 

 instrument, a course involving much additional labour and expense. But it 

 is obvious that the Standard Barometer affords a ready approximate means 

 of estimating the coircctuess of an observer's results. For should he be an 

 incautious observer, the mean difference between the simultaneous readings 



