TESTING  TO  SPECIFICATION

III

stance,  spelling  should  count  twenty  or  thirty  per  cent
°f  the  total  result,  whether  arithmetic  should  be  valued  at
forty  or  at  twenty-five  per  cent.  The  scientific  method  of
determining  these  proportions  is  by  actually  giving  all  the
tests  involved  to  a  group  engaged  in  the  same  kind  of  work.
The  relative  importance  of  the  various  tests  can  then  be
computed  on  the  basis  of  the  correlations  which  are  obtained. ­
  The  mathematics  of  this  process  is  outlined  in  the
Appendix.
_  Hiring  by  specification,  as  here  described,  is  not  entirely ­
  free  from  arbitrary  and  unscientific  elements,  and
yet  it  can  be  seen  that  it  is  a  long  step  in  the  direction  of  a
genuinely  scientific  procedure.  When  a  quantity  of  steel
ls  r equired  for  manufacturing  purposes,  the  first  question  is
jTe  exact  function  which  that  steel  is  to  fulfill.  When  this
ls  known,  the  proper  specifications  for  steel  which  will  serve
d* a t  purpose  are  made  and  the  order  placed  accordingly.
■*-he  laboratory  then  makes  the  appropriate  tests  to  discover ­
  whether  or  not  the  steel  which  is  to  be  used  has  the
characteristics  specified.  Much  the  same  process  must  be
a pplied  to  human  beings  if  they  are  to  be  fitted  for  their
Wor k  with  the  same  consideration  and  care  with  which  we
Usually  select  steel.