''"jjnulry';  m^: }  Repoft  071  Atomic  Weights.  3 1 
Hydrogen. — A  very  thorough  investigation  by  Burt  and  Edgar ^ 
on  the  volumetric  composition  of  water  has  given  the  volume  ratio 
of  hydrogen  to  oxygen  as  2.00288  :  i.  From  this  value,  taking  the 
normal  liter  weights  of  oxygen  and  hydrogen  as  1.42900  and  0.089873 
Gm.,  respectively,  the  atomic  weight  of  hydrogen  becomes  1.00772,  or, 
rounded  off,  1.0077.  Guye,^  from  a  discussion  of  Burt  and  Edgar's 
data,  accepts  this  value  as  lying  between  the  two  extremes  of  1.00767 
and  1.00773.  If>  however,  instead  of  trusting  to  the  densities  of  the 
gases  and  their  physical  constants  exclusively,  we  take  into  account 
the  admirable  researches  of  Morley,  Noyes  and  others,  upon  the 
synthesis  and  analysis  of  water,  ^  the  most  probable  general  mean 
for  the  atomic  weight  of  hydrogen  becomes  1.0078,  which  differs 
from  the  volumetric  value  by  only  1/ 10000;  that  is,  the  two  dis- 
tinct lines  of  attack  upon  the  problem  agree  within  the  limits  of 
actual  uncertainty.  For  ordinary  purposes  the  approximate  value 
1.008  is  close  enough.  It  must  be  remembered  that  the  tables 
prepared  by  this  committee  are  for  the  use,  not  so  much  for  special- 
ists, as  for  working  chemists  in  general ;  and  too  much  refinement  will 
only  lead  to  confusion.  No  determinations  of  these  or  any  other 
constants  can  be  absolute  and  final.  All  are  subject  to  errors  which 
may  be  reduced  nearly,  but  not  quite,  to  insignificance,  but  never 
eliminated  entirely.  For  example,  in  the  determination  of  atomic 
weights  from  gaseous  densities  it  is  not  possible  to  guarantee  the 
absolute  purity  of  the  gases,  or  to  avoid  errors  in  weighing,  in  re- 
ductions to  a  vacuum,  or  in  the  values  given  to  the  physical  constants 
that  are  used  in  the  final  computations.  Some  of  these  errors  may 
be  so  small  as  to  be  negligible,  and  in  the  aggregate  they  may  tend 
either  to  reinforce  or  to  compensate  one  another,  but  their  extreme 
magnitude  can  be  estimated  with  some  approach  to  accuracy,  and 
expressed  by  means  of  the  usual  ±  sign.  At  present  an  accuracy 
to  within  1/ 10000  is  the  best  we  can  expect  to  obtain.^ 
Carbon. — ^Two  investigations  on  the  atomic  weight  of  carbon 
were  reported  from  the  Geneva  laboratory  in  19 18.    First,  Stahr- 
1  Phil.  Trans.,  216A:  393,  1916.  This  research  was  noted  in  the  previous 
report  for  191 7.    Its  review  by  Guye  renders  its  repetition  desirable  here. 
2  /.  chim.  phys.,  15:  208,  191 7. 
3  Computation  by  F.  W.  C. 
^  For  an  elaborate  discussion  of  sources  of  error  in  atomic  weight  determina- 
tions, see  Guye  and  his  colleagues  (M.  Germann,  Moles  and  Renard)  in  /.  chim. 
phys.,  14:  25,  195,  204,  1916;  15 :  60,  360,  405,  1917;  16:  46,  1918. 
