58  Mr.  L.  E.  Ingersoll  on  the  Faraday  and 
4*125  cms.  long  and  at  a  temperature  of  29°  0.  The  field- 
strength  was  approximately  6000  c.G.s.  units,  and  was  given 
by  a  magnet-current  of  18*7  amp.  The  columns  marked 
2dl±  and  "2dl2  give  the  means  of  the  several  galvanometer- 
deflexions  as  explained  above,  the  second  column  being  made 
with  the  galvanometer-connexions  reversed.  All  deflexions 
are  stated  in  terms  of  the  highest  sensibility,  and  have 
galvanometer-proportionality  corrections  applied  in  all  cases* 
but  because  of  a  certain  amount  of  unsteadiness  which  is 
unavoidable — due  mainly  to  small  variations  of  the  source — 
readings  are  given  only  to  the  nearest  millimetre,  and  in 
some  cases — where  a  large  reading  is  obtained  from  one 
which  is  actually  very  much  smaller,  by  reducing  to  a  con- 
stant sensibility  as  a  basis — only  to  the  nearest  centimetre. 
The  considerable  differences  which  are  seen  to  occur,  in  the 
values  of  I  in  separate  determinations  on  the  same  wave- 
length, are  due  to  different  slit-widths,  or  to  different 
brilliancies  of  ihe  glower  on  the  two  occasions. 
The  last  column  has  been  obtained  from  the  previous  one 
by  the  application  of  the  correction  factors  previously  de- 
duced, necessitated  by  the  existence  of  partially  polarized 
light  and  by  impurity  of  the  spectrum.  These  results  have 
been  plotted  in  tig.  2,  the  circles  representing  the  points  of 
series  a  ;  the  crosses,  those  of  b  ;  and  the  triangles,  those  of  c. 
The  small  circles  connected  by  dotted  lines  represent  the 
determinations  of  Moreau,  plotted,  for  comparison's  sake,  to 
the  same  scale,  so  that  the  rotation  at  \='ld  fju  shall  be  the 
same  in  each  case. 
From  the  nature  of  the  problem  it  is  impossible  to  do  more 
than  estimate  the  probable  error  of  the  results,  but,  as  pie- 
viously  mentioned,  it  is  believed  that  the  best  idea  of  it  may 
be  gained  by  noting  the  variation  of  the  points  from  the 
mean  curve.  For,  since  the  points  were  all  determined 
independently  and  with  many  different  adjustments  and 
changes  of  slit-width,  &c,  it  is  difficult  to  see  how  any 
systematic  error  could  remain  undetected,  save  perhaps  one 
which  applied  equally  to  all  wave-lengths,  and  this  would 
only  alter  the  scale  of  the  diagram.  The  probable  error  of 
some  points  is  much  less  than  of  others.  Thus,  in  making 
measurements  on  the  shorter  wave-lengths  of  series  c,  a 
water-cell  was  interposed  in  the  path  of  the  beam,  to  absorb 
the  slight  amount  of  stray  infra-red  radiation,  which  was 
thought  to  have  caused  too  low  values  in  the  corresponding 
points  of  series  a,  so  these  points  in  series  c  are  according^ 
given  greater  weight  in  drawing  the  curve.  Other  things 
being  equal,  those  points,  for  the  determination  of  which  the 
