﻿16 
  W. 
  P. 
  White 
  — 
  Silicate 
  Specific 
  Heats. 
  

  

  we 
  get 
  (2) 
  plus 
  9/^ 
  4 
  E. 
  Since 
  the 
  4th 
  difference 
  is 
  384/* 
  4 
  E 
  

   the 
  discrepancy 
  97i 
  l 
  E 
  can 
  usually 
  be 
  ignored. 
  Then 
  (3) 
  

   is 
  the 
  error 
  of 
  taking 
  (1) 
  for 
  the 
  true 
  heat 
  at 
  0, 
  and 
  may 
  

   be 
  subtracted 
  to 
  get 
  a 
  better 
  result. 
  The 
  effect 
  of 
  acci- 
  

   dental 
  error 
  can 
  be 
  estimated 
  as 
  in 
  the 
  former 
  case, 
  and 
  

   proves 
  to 
  be 
  about 
  the 
  same 
  for 
  the 
  same 
  effective 
  values 
  

   of 
  6 
  and 
  P. 
  

  

  The 
  applicability 
  of 
  both 
  methods 
  of 
  deriving 
  the 
  true 
  

   heat 
  is 
  shown 
  in 
  Table 
  VII, 
  where 
  the 
  even-hundred 
  

   values 
  were 
  obtained 
  by 
  the 
  second 
  method, 
  that 
  of 
  equa- 
  

   tions 
  1, 
  2, 
  3, 
  and 
  the 
  odd-hundred 
  by 
  the 
  first 
  method. 
  

   There 
  is 
  no 
  systematic 
  difference 
  between 
  the 
  two. 
  The 
  

   even-hundred 
  set, 
  having 
  the 
  interval 
  P 
  effectively 
  half 
  

   as 
  great, 
  show 
  greater 
  irregularities 
  resulting 
  from 
  the 
  

   original 
  accidental 
  errors. 
  

  

  The 
  original 
  results 
  for 
  cristobalite 
  were 
  smoothed 
  

   before 
  the 
  derivation 
  of 
  the 
  true 
  heat, 
  because 
  they 
  were 
  

   shown 
  by 
  the 
  run 
  of 
  the 
  differences 
  to 
  lie 
  on 
  a 
  curve 
  less 
  

   smooth 
  than 
  those 
  of 
  the 
  other 
  substances. 
  The 
  agree- 
  

   ment 
  of 
  the 
  original 
  duplicates 
  was 
  excellent, 
  however. 
  

   In 
  the 
  smoothing 
  no 
  change 
  was 
  greater 
  than 
  1 
  per 
  mille, 
  

   the 
  tolerated 
  error; 
  the 
  maximum 
  resulting 
  change 
  in 
  a 
  

   true 
  heat, 
  at 
  1200°, 
  was 
  14 
  per 
  cent. 
  

  

  Eesults. 
  

  

  The 
  Interval 
  Specific 
  Seat 
  Results. 
  

  

  Table 
  IV 
  gives 
  all 
  the 
  results 
  for 
  which 
  the 
  regular 
  pro- 
  

   cedure 
  was 
  used 
  without 
  accident. 
  It 
  also 
  contains 
  pre- 
  

   liminary 
  determinations, 
  most 
  of 
  which 
  were 
  not 
  repeated, 
  

   and 
  some 
  determinations 
  in 
  which 
  an 
  accident 
  occurred, 
  

   but 
  which 
  still 
  seemed 
  to 
  be 
  concordant. 
  Determinations 
  

   affected 
  by 
  accident 
  are 
  omitted 
  if 
  also 
  non-concordant, 
  

   but 
  none 
  are 
  omitted 
  unless 
  there 
  was 
  evident 
  cause 
  of 
  

   possible 
  error. 
  The 
  few 
  which 
  are 
  non-concordant 
  with- 
  

   out 
  special 
  evident 
  reason 
  are 
  in 
  parenthesis. 
  Deter- 
  

   minations 
  were 
  taken 
  as 
  concordant 
  which 
  did 
  not 
  differ 
  

   by 
  more 
  than 
  0-001 
  from 
  their 
  duplicates. 
  The 
  most 
  

   usual 
  mishap 
  was 
  a 
  failure 
  of 
  the 
  charge 
  to 
  drop 
  true 
  

   into 
  the 
  calorimeter. 
  In 
  most 
  of 
  these 
  results 
  one 
  unit 
  

   in 
  the 
  last 
  (fourth) 
  figure 
  is 
  evidently 
  of 
  some 
  sig- 
  

   nificance, 
  two 
  units 
  of 
  some 
  importance 
  in 
  making 
  com- 
  

   parisons. 
  It 
  is 
  clear 
  that 
  in 
  performing 
  arithmetical 
  

   operations 
  on 
  these 
  values 
  a 
  significant 
  distortion 
  might 
  

  

  