﻿W. 
  P. 
  White 
  — 
  Silicate 
  Specific 
  Heats. 
  

  

  tures 
  ; 
  hence 
  greater 
  precision 
  seemed 
  desirable 
  than 
  has 
  

   generally 
  been 
  sought 
  in 
  work 
  of 
  this 
  kind. 
  It 
  therefore 
  

   seems 
  specially 
  necessary 
  to 
  inquire 
  what 
  precision 
  was 
  

   attained. 
  

  

  A 
  glance 
  at 
  almost 
  any 
  comparative 
  table 
  of 
  calori- 
  

   metric 
  values 
  will 
  suffice 
  to 
  show 
  that 
  calorimetry 
  is 
  one 
  

   of 
  those 
  subjects 
  in 
  which 
  the 
  systematic 
  errors 
  are 
  

   usually 
  much 
  larger 
  than 
  the 
  accidental, 
  and 
  deserve, 
  

   therefore, 
  the 
  more 
  careful 
  consideration. 
  The 
  amount 
  

   of 
  accidental 
  error 
  in 
  the 
  present 
  work 
  is 
  easily 
  seen 
  

  

  Table 
  I. 
  

  

  Accidental 
  Errors. 
  

  

  Showing 
  one 
  of 
  the 
  best 
  and 
  one 
  of 
  the 
  worst 
  cases 
  of 
  agreement. 
  

  

  Original 
  interval 
  specific 
  heats 
  from 
  0° 
  to 
  the 
  temperature 
  given. 
  

  

  300' 
  

  

  500 
  ' 
  

  

  700 
  ' 
  

  

  900 
  c 
  

  

  1100' 
  

  

  1300 
  c 
  

  

  1400 
  ( 
  

  

  Quartz 
  

  

  Pseudo-Wollastonite 
  

  

  •21689 
  

  

  •20452 
  

  

  93 
  

  

  39 
  

  

  •23804 
  

  

  •21697 
  

  

  03 
  

  

  707 
  

  

  •25433 
  

  

  •22577 
  

  

  33 
  

  

  93 
  

  

  •25956 
  

  

  •23244 
  

  

  50 
  

  

  49 
  

  

  •26410 
  

  

  •23788 
  

  

  05 
  

  

  87 
  

  

  

  92 
  

  

  

  75 
  

  

  

  77 
  

  

  

  •24270 
  

  

  

  62 
  

  

  

  61 
  

  

  

  46 
  

  

  

  •24470 
  

  

  

  506 
  

  

  

  488 
  

  

  

  473 
  

  

  from 
  Table 
  I. 
  The 
  maximum 
  disagreement 
  in 
  any 
  series 
  

   runs 
  from 
  1-5 
  per 
  mille 
  in 
  the 
  worst 
  case 
  (pseudo-wol- 
  

   lastonite) 
  to 
  0-23 
  per 
  mille 
  in 
  one 
  of 
  the 
  best 
  (quartz). 
  

   It 
  averages 
  about 
  0-5 
  per 
  mille, 
  as 
  appears 
  from 
  Table 
  

   IV. 
  (The 
  mean 
  error 
  would 
  be 
  from 
  a 
  half 
  to 
  a 
  fourth 
  

   as 
  great 
  as 
  the 
  maximum 
  disagreement 
  here 
  given.) 
  

  

  The 
  systematic 
  errors 
  of 
  course 
  can 
  not 
  be 
  completely 
  

   known, 
  but 
  considerable 
  light 
  on 
  their 
  amount 
  can 
  still 
  be 
  

   obtained. 
  In 
  such 
  an 
  attempt, 
  however, 
  the 
  familiar 
  and 
  

   crude 
  distinction 
  between 
  accidental 
  and 
  systematic 
  

  

  