Learning in a multiple-cue judgment task: Evidence for shifts from rule based processing to similarity based processing

Sammanfattning: Cue abstraction (additively combining abstracted values) and exemplar memory (comparing with stored memory via similarity) are important processes in multiple  cue  judgments,  but  previous  studies  lack  insight  into  how  people  use  these  processes  while  learning  to  make  judgments.  The  present  study  investigates  the  learning  process  in  multiple  cue  judgment  tasks,  comparing  a  linear  structure  with  a  non--‐linear  and  modeling  participant  responses  with  formal  models.  Concurrent  verbal  reporting  (think  aloud)  was  used.  The  hypotheses  were  a)  that  initial  learning  would  follow  a  “rule  bias”  via  additive  integration,  b)  the  representation  of  the  task  would  shift  to  an  exemplar  memory  based  one  with  learning  in  the  non--‐linear  structure  and  c)  the  think  aloud  protocol  would  reflect  this  hypothesized  shift.  The  multiplicative  environment  enables  better  learning  of  the  material,  and  is  best  described  by  an  exemplar  memory  model  while  the  linear  group  performs  worse  and  is  equally  well  described  by  both  models.  Model  fit  in  the  non--‐linear  group  changes  from  equal  to  favoring  exemplar  memory  with  training.  Hypothesis  a  was  not  supported  in  the  results,  both  b  and  c  were  supported.  Furthermore  the  results  have  implications  on  the  question  of  Rule  Bias,  and  also  corroborates  previous  studies.