22  jEypedmental  Xegielation
four  opportunities  of  experiment  in  the  same  subject ­
  during  the  longest  political  career.  If  we
divide  up  the  country,  and  try  one  experiment  on
one  town  or  county,  and  another  on  another,
there  is  a  possibility  of  making  an  almost  unlimited
number  of  valid  trials  within  ten  or  twenty  years.
But,  apart  from  this  consideration,  a  general  legislative ­
  change  is  not  a  true  experiment  at  all,
because  it  affords  no  clear  means  of  distinguishing
its  effects  from  the  general  resultant  of  social  and
industrial  progress.  Statistical  facts  are  usually
numerical  or  quantitative  in  character,  so  that,  if
many  causal  agencies  are  in  operation  at  the  same
time,  their  effects  are  simply  added  together
algebraically,  and  are  inextricably  merged  into  a
general  total.  Thus,  the  total  numbers  receiving
poor-law  relief,  or  the  numbers  apprehended  in
the  kingdom  for  drunkenness,  are  numerical
results  affected  by  the  oscillations  of  trade,  by  the
character  of  the  seasons,  the  value  of  gold,  etc.,
etc.,  as  well  as  by  the  Acts  of  the  legislature.  To
make  a  valid  experiment  we  must  have  a  certain
thing  subject  to  certain  constant  conditions,  and
we  must  introduce  a  single  definite  change  of
condition,  which  will  then  be  probably  the  cause
of  whatever  phenomenon  follows.  It  is  possible,
indeed,  to  experiment  upon  an  object  of  varying
conditions,  provided  we  can  find  two  objects  which
vary  similarly  ;  we  then  operate  upon  the  one,
and  observe  how  it  subsequently  differs  from  the