UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Foundations of cultural learning Zhao, Wanying 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2016_february_zhao_wanying.pdf [ 2.27MB ]
JSON: 24-1.0221438.json
JSON-LD: 24-1.0221438-ld.json
RDF/XML (Pretty): 24-1.0221438-rdf.xml
RDF/JSON: 24-1.0221438-rdf.json
Turtle: 24-1.0221438-turtle.txt
N-Triples: 24-1.0221438-rdf-ntriples.txt
Original Record: 24-1.0221438-source.json
Full Text

Full Text

	  	  FOUNDATIONS	  OF	  CULTURAL	  LEARNING	  by	  Wanying	  Zhao	  B.Com.	  (Hons),	  University	  of	  Toronto,	  2007	  M.A.,	  The	  University	  of	  British	  Columbia,	  2010	  	  A	  DISSERTATION	  SUBMITTED	  IN	  PARTIAL	  FULFILMENT	  OF	  	  THE	  REQUIREMENTS	  FOR	  THE	  DEGREE	  OF	  	  	  DOCTOR	  OF	  PHILOSOPHY	  in	  THE	  FACULTY	  OF	  GRADUATE	  AND	  POSTDOCTORAL	  STUDIES	  (Psychology)	  	  THE	  UNIVERSITY	  OF	  BRITISH	  COLUMBIA	  (Vancouver)	  	  	  December	  2015	  	   	  ©	  Wanying	  Zhao,	  2015	  	   ii	  Abstract	  To	  acquire	  their	  local	  culture,	  infants	  must	  identify	  good	  cultural	  models	  to	  learn	  from.	  Doing	  so	  successfully	  requires	  learners	  to	  evaluate	  others’	  qualities	  as	  potential	  knowledge	  sources.	  The	  following	  body	  of	  research	  examines	  how	  the	  youngest	  humans	  identify	  good	  sources	  of	  conventional	  behaviours—a	  domain	  of	  cultural	  knowledge	  that	  lacks	  inherent	  properties	  for	  evaluation.	  	  Chapter	  2	  examines	  infants’	  preferences	  for	  individuals	  who	  performed	  a	  consensus	  action	  vs.	  an	  oft-­‐repeated	  action.	  Results	  revealed	  that	  preverbal	  infants	  are	  capable	  of	  making	  complex,	  context-­‐dependent	  evaluations,	  favouring	  conformists	  when	  the	  targets’	  prior	  knowledge	  cannot	  be	  assumed,	  and	  preferring	  mavericks	  when	  it	  can.	  Chapter	  3	  extends	  these	  results	  by	  showing	  that	  preschool	  aged	  children	  use	  some	  of	  these	  same	  cues	  to	  identify	  who	  may	  be	  good	  to	  learn	  from.	  	  Chapter	  4	  investigates	  infants’	  use	  of	  observed	  emotional	  communications	  to	  choose	  between	  social	  and	  asocial	  targets.	  Cultural	  and	  domain	  differences	  were	  found	  for	  12	  month	  old	  infants:	  target	  preferences	  were	  influenced	  by	  emotional	  reactions	  directed	  at	  social	  targets,	  but	  not	  by	  emotional	  reactions	  directed	  at	  asocial	  targets.	  A	  differential	  response	  to	  positive	  and	  negative	  emotional	  reaction	  only	  reliably	  affected	  European	  Canadian	  infants’	  choices,	  but	  not	  East	  Asian	  infants,	  nor	  European	  and	  East	  Asian	  mixed	  infants.	  Chapter	  5	  investigated	  how	  parents	  convey	  evaluative	  messages	  about	  objects	  during	  interactions	  with	  infants,	  and	  explored	  cultural	  differences	  in	  these	  pedagogical	  interactions.	  Results	  hint	  at	  cultural	  differences	  in	  the	  amount	  of	  valence	  congruent	  	   iii	  utterances	  caregivers	  make,	  resulting	  in	  differential	  experience	  with	  emotional	  communications	  as	  a	  means	  of	  learning	  about	  the	  world.	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	   iv	  Preface	  	  The	  work	  and	  ideas	  presented	  in	  this	  thesis	  are	  those	  of	  the	  author,	  developed	  in	  collaboration	  with	  research	  supervisor	  Dr.	  J.	  Kiley	  Hamlin.	  The	  projects	  and	  associated	  methods	  were	  approved	  by	  the	  University	  of	  British	  Columbia’s	  Behavioural	  Research	  Ethics	  Board	  [certificate	  #H10-­‐01808].	  All	  chapters	  reflect	  original,	  unpublished	  work	  by	  the	  author,	  who	  had	  primary	  responsibility	  for	  all	  aspects	  of	  each	  study,	  including	  design,	  data	  collection,	  analysis,	  and	  manuscript	  composition.	  Chapter	  3	  was	  developed	  in	  discussion	  with	  Dr.	  Andrew	  Baron,	  who	  also	  provided	  input	  into	  study	  design	  and	  data	  collection.	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	   v	  Table	  of	  contents	  Abstract	  .......................................................................................................................................................	  ii	  Preface	  ……………………………………………………………………………………………………………...iv	  Table	  of	  contents	  ……………………………………………………………………………………………...	  v	  List	  of	  tables	  …………………………………………………………………………………………………….vii	  List	  of	  figures	  ………………………………………………………………………………………………….viii	  Acknowledgements	  ………………………………………………………………………………………….ix	  	   	  Chapter	  one:	  Introduction	  …………………………………………………………………………..........1	  	   1.1	   General	  overview	  …………………………………………………………………………...1	  	   1.2	   Two	  modes	  of	  learning	  ……………...………………………………….………………...2	  	   1.3	   Selective	  cultural	  learning……………………………………………………………….5	  	   1.4	  	   Learning	  in	  preverbal	  infants	  ………………………………………………………..11	  	   1.5	  	   Evaluation	  of	  sources	  of	  cultural	  knowledge	  ………………………………….19	  	   1.6	   Cultural	  differences	  in	  cultural	  learning?	  ………………………………….…...25	  	   1.7	  	   Thesis	  rationale	  ……………………………………………………………………………28	  	  	  	  	  	  	  	  	  Chapter	  two:	  	  Preverbal	  infants’	  context	  dependent	  preferences	  for	  	  conformists	  vs.	  innovators	  …...................…………………………………………………...30	  	   2.1	  	   Introduction	  ………………………………………………………………………………...30	  	   2.2	   Study	  1	  ………………………………………………………………………………………...39	  	   2.3	   Study	  2	  ………………………………………………………………………………………...45	  	   2.4	   General	  discussion	  ……………………………………………………………………….48	  	  Chapter	  three:	  Preschoolers	  like	  and	  learn	  from	  nonconformists…………………53	  	   3.1	   Introduction	  ………………………………………………………………………………...53	  	   3.2	   The	  present	  study	  ………………………………………………………………………...60	  	  Chapter	  four:	  Infant	  social	  referencing	  differs	  by	  culture	  and	  by	  domain……..76	  	   4.1	   Introduction	  ………………………………………………………………………………...76	  	   4.2	   The	  present	  studies	  ……………………………………………………………………...80	  	   4.3	   Study	  1a	  –	  social	  targets	  ……………………………………………………………….84	  	   4.4	   Study	  1b	  –	  undirected	  gaze	  …………………………………………………………..96	  	   4.5	   Study	  1c	  –	  behavioural	  coding	  …………………………………………………….100	  	   4.6	   Study	  2	  –	  object	  targets	  ………………………………………………………………104	  	   4.7	  	   General	  discussion	  ……………………………………………………………………..109	  	   vi	  	  Chapter	  five:	  Parents’	  affective	  communications	  across	  cultures	  ………………..114	  	   5.1	   Introduction	  ………………………………………………………………………………114	  	   5.2	   The	  present	  study	  ………………………………………………………………………117	  	  Chapter	  six:	  General	  discussion	  …………………………...………………………………............138	  	  References	  …………………………………………………………………………………………………….	  148	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	   vii	  List	  of	  tables	  	  	  Table	  3.1	  Children’s	  probabilities	  of	  liking	  the	  novel-­‐dance	  Protagonist,	  by	  age	  and	  by	  condition	  ……………………………………………………………………………………………………………….	  66	  	  	  Table	  3.2	  Children’s	  probabilities	  of	  learning	  from	  the	  novel-­‐dance	  Protagonist,	  by	  age	  	  	  	  and	  by	  condition	  ………………………………………………………………………………………....................69	  	  Table	   3.3	   Preference	   and	   Learning	   predicted	   by	   age,	   sex,	   condition,	   and	   age-­‐by-­‐condition	  interaction	  ………………………………………………………………………………………..........	  70	  	  Table	   3.4	   Learning	   predicted	   by	   Liking,	   age,	   condition,	   and	   age-­‐by-­‐condition	  interaction	  ………………………………………………………………………………………………………..........72	  	  Table	  4.1	  Logistic	  regression	  models	  predicting	  Protagonist	  choice	  in	  Study	  1a.............92	  	  Table	  4.2	  Logistic	  regression	  models	  comparing	  results	  of	  Study	  1a	  and	  1b……………...98	  	  Table	   5.1	   Comparison	   of	   parents’	   communications	   about	   good	   and	   bad	   targets,	   by	  ethnicity	  ………………………………………………………………………………………………………………126	  	  Table	   5.2	   Comparison	   of	   East	   Asian	   and	   Euro-­‐Canadian	   parents’	   communications	  about	  good	  and	  bad	  targets,	  by	  communication	  type………………………………………………130	  	  	  Table	  5.3	  Parent-­‐infant	  attentional	  states,	  by	  ethnicity	  and	  attention	  type	  ……………..133	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	   viii	  List	  of	  figures	  	  	  Figure	  2.1	  Stimuli	  for	  Study	  1……………………………………………………………………….…….42	  	  Figure	  2.2	  Infants’	  Protagonist	  choices	  by	  condition	  in	  Study	  1……………………..……..44	  	  Figure	  2.3	  Stimuli	  for	  Study	  2,	  Consensus	  condition…………………………………………....46	  	  Figure	  2.4	  Infants’	  Protagonist	  choices	  by	  condition	  in	  Study	  2………………………..….47	  	  	  	  	  Figure	  3.1	  Children’s	  probabilities	  of	  liking	  the	  novel-­‐dance	  Protagonist,	  by	  age	  and	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  by	  condition…………………………………………………………………………………………….…………66	  	  	  Figure	  3.2	  Children’s	  probabilities	  of	  learning	  from	  the	  novel-­‐dance	  Protagonist,	  by	  	  age	  and	  by	  condition………………………………………………………………………………………….	  69	  	  Figure	  4.1	  Video	  stimuli	  for	  Study	  1a	  …………………………………………………….………......88	  	  	  	  	  Figure	  4.2	  Infants	  Protagonist	  choices	  by	  Protagonist	  valence,	  ethnicity,	  and	  age	  for	  	  Study	  1a…………………………………………………………………………………………………………….90	  	  	  Figure	  4.3	  Video	  stimuli	  for	  Study	  1b	  ………………………………………………………………...97	  	  Figure	  4.4	  Infants’	  Protagonist	  choice,	  by	  ethnicity,	  in	  Study	  1b………………………......97	  	  Figure	  4.5	  Infants’	  object	  choice	  by	  ethnicity,	  and	  by	  age	  in	  Study	  2	  ………………….106	  	  Figure	  4.6	  Infants’	  object	  choice	  by	  ethnicity,	  and	  by	  age	  in	  Study	  2…………………..107	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	   ix	  	  	  Acknowledgements	  	  	   It	  really	  does	  take	  a	  village.	  	  I	   owe	  my	  deepest	   thanks	   to	  my	  advisor	  Kiley	  Hamlin,	   for	  her	  mentorship,	   her	  relentless	   curiosity	   and	   tireless	   pursuit	   of	   the	   big	   questions,	   and	   her	   unwavering	  support	  throughout	  this	  journey.	  	  	   Andy	  Baron,	  who	  is	  always	  willing	  to	  Skype	  from	  anywhere	  in	  the	  world:	  thank	  you	  for	  believing	  in	  the	  Smurfs.	  	  	   Thank	   you	   to	   my	   committee	   members,	   Janet	   Werker,	   Steve	   Heine,	   and	   Andy	  Baron,	   for	   supporting	   me	   through	   this	   process;	   my	   work	   benefitted	   in	   rigor	   and	   in	  scope	  from	  your	  invaluable	  feedback.	  	  	   I	   was	   fortunate	   to	   have	   been	   surrounded	   by	   a	   community	   of	   inspiring	  individuals	  at	  UBC	  who	  made	  this	  process	  so	  enjoyable:	  Doan	  Le,	  Conor	  Steckler,	  Enda	  Tan,	  Julia	  Van	  de	  Vondervoort,	  Setareh	  Nourani,	  Janine	  Slevinski,	  Shannon	  Bridson,	  and	  the	   many	   wonderful	   research	   assistants	   at	   the	   Centre	   for	   Infant	   Cognition.	   Hyemin	  Choo	   and	   Leigh-­‐Ann	   Bong	   were	   diligent	   and	   tireless	   coders	   of	   parent-­‐infant	  interactions,	  and	  Aida	  Sepehr,	  of	  infant	  facial	  expressions.	  Matthew	  Shin,	  Mimi	  Sun,	  Mu	  Yang,	  Nadia	  Hui,	  and	  Sheila	  Tse	  were	  invaluable	  to	  recruiting	  East	  Asian	  families.	  Special	   thanks	   to	   Aiyana	  Willard	   for	   making	   sure	   I	   didn’t	   resort	   to	   too	  much	  canned	  soup;	  Maciek	  Chudek	  for	  challenging	  my	  thinking	  just	  by	  being	  himself;	  Michael	  Muthukrishna,	   for	   the	   illuminating	   debates;	   Sarah	   Klain,	   A-­‐Team	   and	   fellow	   Green	  College	  friends,	  for	  their	  enriching	  friendships	  and	  ideas.	  	  	   x	  	   This	   thesis	   would	   not	   have	   been	   possible	   without	   the	   many	   families	   who	  participated	  in	  studies	  at	  the	  Centre	  for	  Infant	  Cognition,	  and	  the	  Living	  Lab	  at	  the	  Telus	  World	  of	  Science.	  They	  have	  my	  deepest	  gratitude.	  Thank	  you	  to	  my	  parents,	  for	  leading	  me	  to	  life	  outside	  of	  the	  fishbowl,	  that	  inspired	  these	  questions.	  Daniel	  Randles,	  for	  his	  daily	  support	  of	  all	  my	  endeavours.	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	   1	  1.	  	  Introduction	  	  1.1	  	   General	  overview	  Human	  cultural	  diversity	  is	  a	  defining	  characteristic	  of	  our	  species.	  Adapting	  to	   the	  wide	   range	   of	   ecologies	   that	   our	   species	   inhabit,	   from	   tropics	   to	   tundra	   to	  marshland,	  would	  be	  impossible	  without	  the	  capacity	  to	  acquire	  know-­‐how,	  built	  up	  and	  refined	  over	  generations,	  for	  living	  in	  the	  local	  environment.	  From	  subsistence	  (how	   to	  hunt,	   tend	   to	   crops,	   or	   to	   shop	   for	   produce),	   and	   treating	  disease	   (which	  plant,	  or	  root,	  or	  medication	  to	  compound	  or	  use),	   to	  coexisting	  with	  social	  others	  (how	  to	  engage	  in	  courtship,	  conduct	  warfare,	  or	   implement	  values	  such	  as	  liberty	  and	  equality),	  and	  navigating	  terrain	  (reading	  the	  position	  of	  the	  sun	  and	  stars,	  use	  maps,	   or	   the	   GPS),	   “culture”	   encompasses	   all	   “information	   capable	   of	   affecting	  individuals’	   behaviours	   that	   they	   acquire	   from	   other	   members	   of	   their	   species”	  (Richerson	   &	   Boyd,	   2005).	   Human	   newborns	   do	   not	   possess	   knowledge	   of	   any	  particular	  culture,	  but	  rather,	  must	  learn	  these	  from	  other	  humans	  over	  the	  course	  of	  their	  lives.	  The	  youngest	  humans,	  then,	  must	  be	  equipped	  with	  a	  common	  set	  of	  tools	   that	   allows	   them	   to	   acquire	   cultural	   knowledge.	   What	   might	   this	   common	  toolkit	   contain?	   And	   how	   does	   the	   process	   of	   cultural	   learning	   unfold	   during	   the	  first	  years	  of	  life?	  By	  the	  time	  they	  start	  school,	  children	  already	  possess	  a	  great	  deal	  of	  cultural	  knowledge.	   Most	   typically	   developing	   children	   have	   command	   of	   their	   native	  language,	   express	   culturally	   specific	   food	   preferences,	   and	   can	   label	   and	   use	  common	  objects.	  They	  also	  generally	   abide	  by	   their	   societies’	   social	  norms,	  which	  prescribe	   how	   to	   speak	   and	   take	   turns,	   and	   proscribe	   certain	   kinds	   of	   aggression	  	   2	  against	   others.	   And	   yet,	   despite	   children	   clearly	   demonstrating	   many	   forms	   of	  cultural	  knowledge,	  relatively	  little	  is	  understood	  about	  how	  children	  become	  such	  accomplished	   cultural	   learners	   in	   such	   a	   brief	   time.	   Indeed,	  much	  more	   is	   known	  about	  what	  children	  know	  at	  different	  ages	  than	  about	  how	  they	  traverse	  from	  one	  kind	  of	  understanding	  to	  another	  (Needham	  &	  Woodward,	  2009,	  p.	  xviii).	  	  What	   do	   earlier	   forms	   of	   cultural	   learning	   look	   like?	   Are	   preverbal	   infants	  capable	   of	   cultural	   learning,	   and	  do	   they	   exhibit	   selectivity	   in	   doing	   so?	  What	   are	  some	   dimensions	   along	   which	   infants	   evaluate	   potential	   sources	   of	   cultural	  knowledge?	   This	   thesis	   focuses	   on	   social	   evaluative	   abilities	   during	   infancy	   and	  early	  childhood	  that	  help	  cultural	  novices	   identify	  and	  learn	   local	  culture,	   focusing	  in	  particular	  on	  evaluating	  and	  learning	  social	  conventional	  behaviours.	  	  1.2	   Two	  modes	  of	  learning	  There	   are	   numerous	   possible	  ways	   to	   learn	   about	   the	  world.	  Humans,	   like	  other	  species,	  can	  discover	  objective	  facts	  about	  the	  world	  through	  first-­‐person	  trial	  and	  error,	  aided	  by	  reasoning	  about	  cause	  and	  effect.	  For	  example,	  an	  individual	  can	  pick	  between	  several	  possible	  tools	  by	  trying	  each	  one	  and	  evaluating	  the	  efficacy	  of	  the	  outcomes.	  This	  initial	  trial-­‐and-­‐error	  will	  allow	  the	  individual	  to	  select	  the	  best	  tool	  for	  achieving	  his	  or	  her	  goal	  in	  the	  present,	  as	  well	  as	  to	  retain	  the	  conclusion	  for	   later	  use.	  While	  trial	  and	  error	   learning	   is	  certainly	  a	  beneficial	  mechanism	  for	  acquiring	  some	  forms	  of	  knowledge,	  there	  are	  plentiful	  situations	  in	  which	  learning	  by	   trial-­‐and-­‐error	   proves	   less	   beneficial	   than	   another	   common	   learning	   mode:	  	   3	  cultural	   learning1,	  whereby	  knowledge	   is	   acquired	   from	  others	   via	   observation	  or	  teaching.	   Indeed,	   cultural	   learning	   provides	  many	   advantages	   over	   trial-­‐and-­‐error	  learning.	  For	  instance,	  it	  may	  be	  dangerous	  or	  otherwise	  costly	  to	  learn	  by	  trial	  and	  error	  (e.g.	  touching	  a	  hot	  stove).	  It	   is	  often	  more	  expedient	  to	  learn	  from	  others	  in	  situations	   in	  which	  how	   things	  work	   is	   not	   obvious	  or	   immediately	   available	   (e.g.	  the	   relationship	   between	   being	   exposed	   to	   viruses	   and	   subsequently	   showing	   flu	  symptoms).	   In	   these	   situations,	   cultural	   learning	   can	   save	   individuals	   time	   and	  effort	  by	  observing	  others’	  successes	  and	  mistakes.	  	  According	   to	   recent	   estimates,	   humans’	   adaptation	   to	   our	   diverse	  circumstances	   was	   primarily	   driven	   by	   cultural	   learning,	   while	   individual	  resourcefulness	  plays	  a	  lesser	  role	  (Mathew	  &	  Perreault,	  2015).	  Humans’	  proclivity	  for	   cultural	   learning	   has	   been	   documented	   extensively	   in	   many	   domains	   of	   life.	  Children	  and	  adults	   learn	  acts	  of	  aggression,	  prosocial	  giving,	  alcohol	  consumption	  and	   littering	   from	  others,	  especially	   in	  unfamiliar	  situations	  when	   the	  appropriate	  behaviour	  is	  unknown	  or	  ambiguous	  (Bandura	  &	  McDonald,	  1963;	  Baron,	  Vandello,	  &	   Brunsman,	   1996;	   Goldstein,	   Cialdini,	   &	   Griskevicius,	   2008;	   Prentice	   &	   Miller,	  1993).	   Knowledgeable	   adults	   facilitate	   the	   cultural	   learning	   process	   by	  demonstrating	  culturally	  desirable	  behaviours,	  and	  providing	  justifications	  for	  those	  behaviours	  at	  a	  level	  that	  is	  appropriate	  for	  the	  learner’s	  abilities.	  In	  turn,	  children’s	  reasoning	   and	   judgment	   resemble	   that	   of	   their	   cultural	   models	   (Holstein,	   1972;	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  1	  Rather	  than	  the	  common	  sense	  understanding	  of	  “culture”,	  in	  reference	  to	  “pop”	  or	  “high”	  culture,	  here	  cultural	  learning	  refers	  more	  broadly	  to	  any	  information	  learned	  from	  a	  conspecific;	  it	  is	  also	  commonly	  referred	  to	  as	  social	  learning	  in	  the	  literature.	  	   4	  Bandura,	  Grusec,	  &	  Menlove,	  1967;	  Brody	  &	  Henderson,	  1977),	  both	  in	  the	  types	  of	  arguments	  used	  to	  justify	  their	  judgments,	  and	  the	  content	  of	  judgments	  themselves.	  	  Infants	   in	  their	  first	  year	  of	   life	   imitate	  a	  range	  of	  behaviours	  performed	  by	  adult	   models.	   Starting	   as	   early	   as	   an	   hour	   after	   birth,	   infants	   can	   imitate	   facial	  postures,	  including	  tongue	  protrusions,	  lip	  pursing,	  and	  finger	  movements	  (Meltzoff	  &	  Moore,	  1983).	  By	  the	  end	  of	  the	  first	  year	  and	  into	  the	  second	  year,	  infants	  reliably	  imitate	  many	  novel	  actions	  (Buttelmann,	  Zmyj,	  Daum,	  &	  Carpenter,	  2013;	  DiYanni	  &	  Kelemen,	   2008;	   Herrmann,	   Call,	   Hernandez-­‐Lloreda,	   Hare,	   &	   Tomasello,	   2007;	  Nielsen,	   2006;	   Zmyj	   &	   Daum,	   2009;	   Zmyj	   et	   al,	   2002);	   and	   are	   even	   able	   to	  successfully	  imitate	  failed,	  or	  incomplete	  actions	  (Meltzoff,	  1995).	  The	   clearest	   evidence	   that	   infants	   learn	   from	   others	   is	   in	   the	   language	  domain,	  since	  there	  is	  no	  way	  for	  infants	  to	  identify	  the	  meaning	  of	  words	  except	  by	  hearing	  others	  speak.	  By	  9	  months,	  infants	  already	  recognize	  many	  words	  referring	  to	   common	   household	   items	   and	   parts	   of	   the	   body,	   outperforming	   even	   their	  parents	  reports	  of	   their	  word	  knowledge	  (Bergelson	  &	  Swingley,	  2012).	  By	  twelve	  months	   of	   age,	   infants	   are	   adept	   at	   discriminating	   phonemes	   (speech	   sounds)	   of	  their	   native	   language,	   but	   have	   lost	   the	   ability	   to	   discriminate	   foreign	   language	  phonemes	   that	   they	   were	   able	   to	   differentiate	   a	   few	   months	   earlier	   (Kuhl	   et	   al.,	  2006;	  Werker	  &	  Tees,	   2005).	   These	   results	   suggest	   that	   twelve-­‐month-­‐old	   infants	  have	   learned	   to	   recognize	   the	   phonemes	   of	   their	   native	   language.	   Finally,	   the	  complexity	   and	   variation	   in	   caregivers’	   speech	   predicts	   infants’	   vocabulary	   and	  language	   processing	   (Fernald,	  Marchman,	   &	  Weisleder,	   2012),	   further	   supporting	  the	  view	  that	  infants	  learn	  from	  their	  caregivers.	  	   5	  One	   aspect	   of	   the	   current	   discussion	   that	   stands	   out	   is	   the	   variability	   in	  solutions	  offered	  by	  cultures:	   there	  are	  over	  6000	   languages	   spoken	  worldwide,	  a	  plethora	  of	  possible	  actions	  for	  any	  given	  physical	  context,	  and	  competing	  religious	  systems	  that	  each	  purport	  to	  be	  the	  most	  authentic.	  Humans	  are	  clearly	  addicted	  to	  culture,	  but	  how	  is	  a	  learner	  to	  ensure	  that	  she	  is	  learning	  the	  “right	  stuff”?	  	  1.3	   Selective	  cultural	  learning	  Our	  heavy	  dependence	  on	  knowledge	  learned	  from	  others	  requires	  humans	  to	   be	   somewhat	   credulous.	   To	   believe	   and	   to	   rely	   on	   culturally	   transmitted	  information,	  we	  must	  trust	  that	  what	  others	  tell	  us	  are	  in	  large	  part	  truthful	  (Jaswal,	  2013;	   Koenig	   &	   Sabbagh,	   2013).	   This	   tendency	   to	   trust	   is	   evident	   in	   adults	   and	  children,	  who	  readily	  learn	  and	  act	  on	  knowledge	  that	  is	  neither	  accessible	  by	  direct	  observation	   nor	   easily	   verifiable,	   such	   as	   the	   shape	   of	   the	   earth	   (Vosniadou	   &	  Brewer,	  1994),	   the	   life	  cycle	  of	  animals	  (Giménez	  &	  Harris,	  2002),	   the	  existence	  of	  microbial	  organisms,	  and	  how	  the	  brain	  works	  (Gottfried,	  Gelman,	  &	  Schultz,	  1999).	  This	   tendency	   to	   trust	   is	   so	   strong	   that	   learners	   will	   sometimes	   trust	   another	  person’s	   claims	   even	   if	   they	   conflict	   with	   their	   own	   observations	   (Jaswal,	   2010;	  Jaswal,	  Croft,	  Setia,	  &	  Cole,	  2010).	  However,	   learners	   are	   not	   entirely	   vulnerable	   to	   the	   whims	   and	  inconsistencies	  of	  potential	   informants.	  Even	  young	   learners	  are	  able	   to	  avoid	   the	  pitfalls	  of	  erroneous	  information	  and	  learn	  only	  some	  information	  from	  others.	  For	  example,	   extensive	   research	   has	   shown	   that	   given	   a	   choice	   between	   potential	  models	   from	   which	   to	   learn	   new	   information,	   young	   children	   are	   sensitive	   to	   a	  number	   of	   dimensions	   on	   which	   informants	   differ,	   and	   preferentially	   acquire	  	   6	  information	   from	   individuals	   who	   are	   likely	   to	   provide	   information	   that	   is	   more	  accurate,	   reliable,	   well-­‐intentioned,	   and	   relevant	   (Koenig	   &	   Sabbagh,	   2013;	   Mills,	  2013).	  I	  will	  address	  each	  dimension	  below.	  	  1.3.1	  	   Accuracy	  and	  reliability.	  	  A	   first	   challenge	   to	   successful	   cultural	   learning	   is	   to	   identify	   accurate	   and	  knowledgeable	   informants.	   Indeed,	   potential	   models	   may	   be	   unreliable	   for	   a	  number	  of	  reasons:	  1)	  they	  may	  be	  poorly	  informed,	  overall	  (a	  sign	  that	  this	  may	  be	  the	  case	  is	  if	  the	  model	  was	  wrong	  in	  the	  past,	  or	  they	  may	  seem	  uncertain	  or	  show	  hesitation);	  2)	   they	  may	   lack	  situationally-­‐relevant	   information	  (if	   they	  are	  absent	  from,	  or	  lack	  visual	  access	  to	  an	  event);	  or	  3)	  they	  may	  have	  intentions	  to	  deceive	  or	  withhold	  accurate	   information	  (they	  show	  signs	  of	  being	  uncooperative).	  Learners	  must	  safeguard	  against	  these	  bad	  models	  by	  being	  choosy	  about	  whom	  to	  trust	  and	  pay	  attention	  to.	  Research	  suggests	  that	  far	  from	  blindly	  learning	  from	  any	  potential	  source	  of	  information,	   young	   children	   show	   sensitivity	   to	   differences	   in	   informants’	   past	  reliability.	  Three-­‐year-­‐olds	  assess	  past	  reliability	  of	  models	  in	  word	  learning—they	  are	  more	  likely	  to	   learn	  novel	  words	  from	  informants	  who	  have	  previously	  named	  familiar	   objects	   correctly,	   compared	   to	   those	   who	   have	   named	   them	   incorrectly	  (Birch,	   Vauthier,	   &	   Bloom,	   2008;	   Jaswal	   &	   Malone,	   2007;	   Jaswal	   &	   Neely,	   2006;	  Koenig	   &	   Harris,	   2005;	   Koenig,	   Clément,	   &	   Harris,	   2004;	   Pasquini,	   Corriveau,	  Koenig,	  &	  Harris,	  2007;	  Scofield	  &	  Behrend,	  2008).	  By	  their	  4th	  year,	  children	  track	  the	   degree	   of	   informants’	   past	   reliability	   in	   deciding	   whom	   to	   trust,	   trusting	  individuals	   who	   have	   been	   relatively	   more	   accurate	   over	   those	   who	   have	   been	  	   7	  relatively	  less	  accurate	  (Pasquini	  et	  al.,	  2007).	  Although	  3-­‐year-­‐olds	  prefer	  to	  learn	  from	  those	  who	  are	  consistently	  accurate	  and	  reliable,	  unlike	  4-­‐year-­‐olds	  they	  fail	  to	  differentiate	  between	  degrees	  of	  accuracy	  and	  error;	  for	  example,	  they	  are	  no	  more	  likely	  to	  trust	  someone	  who	  made	  one	  mistake	  and	  several	  accurate	  judgments	  than	  someone	  who	  made	  several	  mistakes.	  	  Children’s	  sensitivity	  to	  reliability	  reflects	  a	  direct	  judgment	  of	  the	  quality	  of	  an	  informant’s	  information	  that	  is	  based	  on	  the	  child’s	  own	  prior	  knowledge.	  Other	  cues	  help	  learners	  assess	  informant	  quality	  in	  a	  more	  indirect	  fashion,	  even	  if	  they	  possess	   no	   prior	   knowledge.	   One	   such	   cue	   is	   confidence,	   which	   portrays	   the	  informant’s	  relative	  certainty	  of	  the	  quality	  of	  the	  information	  they	  are	  transmitting.	  Confidence	   is	   a	   useful	   cue,	   as	   it	   tends	   to	   be	   correlated	   with	   actual	   accuracy:	  confident	   informants	   tend	  to	  be	  more	  skillful	  and	  offer	  more	  accurate	   information	  than	   uncertain/hesitant	   informants	   (e.g.	   Sporer,	   Penrod,	   Read,	   &	   Cutler,	   1995).	  Two-­‐year-­‐olds	   are	   sensitive	   to	   nonverbal	   cues	   of	   confidence;	   they	   prefer	   to	   learn	  from	   someone	   who	   confidently	   performs	   an	   action,	   compared	   to	   someone	   who	  performs	  it	  unconfidently	  (Birch,	  Akmal,	  &	  Frampton,	  2010).	  They	  are	  also	  sensitive	  to	  verbal	  cues	  of	  confidence:	  they	  prefer	  to	  learn	  from	  a	  confident	  informant	  (who	  declares	  “This	  is	  a	  spoon”),	  to	  an	  informant	  who	  expresses	  uncertainty	  (who	  says	  “I	  think	  this	  is	  a	  spoon”	  (Jaswal	  &	  Malone,	  2007).	  	  Another	   indirect	   cue	   to	   informant	   accuracy	   is	   prestige,	   or	   the	   amount	   of	  respect	   a	   given	   individual	   has	   been	   granted	   by	   others.	   This	   is	   often	  measured	   by	  how	  much	  attention	  is	  allocated	  to	  an	  individual:	  those	  with	  more	  respect	  are	  often	  attended	   to	  more.	  Prestige	   is	   likely	  an	   indirect	   cue	   to	  past	  accuracy:	  an	   informant	  	   8	  who	   has	   proven	   to	   be	   reliable	   in	   the	   past	   is	   likely	   to	   have	   more	   respect	   and	   be	  attended	  to	  more	  than	  someone	  who	  was	  proven	  unreliable.	  A	  learner	  can	  thus	  use	  the	  amount	  of	   attention	  directed	  at	   an	   informant	  as	  an	   indirect	   gauge	  of	  previous	  demonstrated	  accuracy	  or	  reliability.	  Indeed,	  3-­‐	  and	  4-­‐	  year-­‐olds	  are	  more	  likely	  to	  learn	   from	   those	  who	   others	   attend	   to,	   particularly	   in	   the	   domain	   in	  which	   their	  prestige	  was	  originally	  observed	  (Chudek,	  Heller,	  Birch,	  &	  Henrich,	  2012).	  	  Related	  to	  prestige,	  another	  indirect	  cue	  to	  informant	  accuracy	  is	  informant	  age.	   As	   older	   individuals	   are	   likely	   to	   have	   more	   knowledge,	   it	   would	   benefit	  learners	  to	  generally	  learn	  from	  older	  individuals;	   indeed,	  children	  are	  more	  likely	  to	   learn	   from	  older	   informants,	  unless	   those	   informants	  have	  proven	  unreliable	   in	  the	  past	  (Jaswal	  &	  Neely,	  2006;	  Rakoczy,	  Hamann,	  Warneken,	  &	  Tomasello,	  2010).	  This	   tendency	   to	   learn	   from	  older	   informants	   is	   itself	   selective,	  however:	   children	  tend	  to	  choose	  similarly	  aged	  informants	  when	  learning	  about	  toys	  (VanderBorght	  &	  Jaswal,	  2009).	  	  1.3.2	  	   Situational	  access.	  	  In	   addition	   to	   being	   sensitive	   to	   who	   might	   be	   more	   knowledgeable	   in	  general,	   children	   are	   also	   sensitive	   to	   the	   fact	   that	   an	   individual	   may	   have	  knowledge	  of	  some	  situations	  and	  not	  others.	  Three-­‐year-­‐olds	  show	  selective	  trust	  for	  an	  informant	  who	  has	  demonstrated	  the	  source	  of	  their	  knowledge	  (by	  looking	  inside	   a	   container)	   over	   an	   informant	   who	   has	   indicated	   probable	   ignorance	   (by	  guessing	  or	  pretending	  that	  something	  is	  in	  a	  container;	  Koenig,	  2012).	  Similarly,	  4-­‐	  and	  5-­‐year-­‐old	  children	  are	  more	  likely	  to	  trust	  a	  speaker	  who	  has	  visual	  access	  to	  	   9	  an	   object	   over	   a	   speaker	   whose	   vision	   is	   limited	   when	   learning	   object	   labels	  (Brosseau-­‐Liard	  &	  Birch,	  2011).	  	  1.3.3	   Helpful	  intentions.	  	  Even	   if	   one	   can	   determine	   whether	   or	   not	   a	   potential	   informant	   is	  knowledgeable,	   it	   would	   be	   inadvisable	   to	   learn	   from	   informants	   who	   intend	   to	  deceive.	   Research	   suggests	   that	   in	   the	   preschool	   years,	   young	   children	   come	   to	  recognize	  that	  all	  informants	  may	  not	  be	  equally	  well	  meaning.	  For	  example,	  3-­‐	  to	  5-­‐year	   olds	   preferentially	   trust	   benevolent	   informants	   (Mascaro	   &	   Sperber,	   2009).	  This	  ability	  may	  not	  be	  robust	  in	  3-­‐	  and	  4-­‐year	  olds,	  but	  by	  5	  years	  of	  age	  children	  reliably	  take	  advice	  from	  a	  benevolent	  informant	  (Vanderbilt,	  Liu,	  &	  Heyman,	  2011).	  When	   both	   accuracy	   and	   benevolence	   cues	   are	   available,	   3-­‐	   to	   5-­‐year-­‐olds	  selectively	   learn	   from	  a	  person	  described	  as	  an	  expert	  vs.	  a	  non-­‐expert	  only	  when	  they	  were	   also	   described	   as	   nice,	   rather	   than	  mean	   (Landrum,	  Mills,	   &	   Johnston,	  2013).	  Finally,	  5-­‐	  and	  6-­‐year	  olds	  are	  able	  to	  discern	  differently	  valenced	  intentions	  by	  observing	  past	  behaviour;	  they	  put	  greater	  trust	  into	  informants	  who	  previously	  tried	   to	  help	  others	  more	   than	   informants	  who	  previously	   tried	   to	  deceive	  others,	  regardless	  of	  whether	   their	  help	   led	   to	  a	  good	  or	  bad	  outcome	   (Liu,	  Vanderbilt,	  &	  Heyman,	  2013).	  1.3.4	  	   Relevance.	  	  One	   type	   of	   challenge	   that	   learners	   face	   is	   discerning	   what	   information	   is	  applicable	   to	   their	   situation,	   including	   its	   relevance	   to	   their	   local	   sociocultural	  context.	   Social	   rules	   pertaining	   to	   how	   to	   use	   objects,	   how	   to	   behave,	   and	  which	  languages	   are	   spoken	   in	   a	   given	   culture	   apply	   to	   members	   of	   particular	   groups.	  	   10	  When	   presented	  with	   a	   choice	   of	   informants,	   young	   learners	   show	   sensitivity	   on	  dimensions	  that	  are	  indicative	  of	  geographical	  and	  cultural	  proximity	  (stable	  group	  membership).	   One	   such	   dimension	   is	   whether	   the	   learner	   already	   knows	   the	  individual;	   children	   prefer	   familiar	   individuals	   as	   informants	   over	   strangers	  (Corriveau	  &	  Harris,	  2009;	  Harris	  &	  Corriveau,	  2011).	  Beyond	  the	  immediate	  family,	  learners	   are	   sensitive	   to	   which	   individuals	   share	   their	   particular	   culture.	   For	  example,	   4-­‐	   and	   5-­‐year	   old	   children	   preferentially	   reproduce	   object	   functions	  demonstrated	   by	   native	   accented	   speakers	   of	   their	   own	   language	   (Corriveau,	  Kinzler,	  &	  Harris,	  2013).	  Thus,	  despite	  a	  degree	  of	  credulity	  necessary	  to	  learn	  from	  other	  people,	  by	  4	  years	   of	   age,	   children	   begin	   to	   demonstrate	   selectivity	   along	   dimensions	   of	  reliability,	   benevolence,	   and	   relevance	   in	   their	  decisions	   regarding	  whom	   to	   learn	  from.	   Three	   year	   olds,	   in	   comparison,	   are	   less	   reliable	   at	   evaluating	   along	   these	  dimensions,	  suggestive	  of	  developmental	  change	  taking	  place	  over	  the	  3rd	  year.	  How	  does	   such	   selectivity	  develop?	  Do	   children	  achieve	   such	   competency	   in	  evaluating	  potential	   knowledge	   sources	   by	   toiling	   through	   repeated	   trials	   of	   success	   and	  failure?	  Or	  are	  they	  equipped	  with	  some	  evaluative	  tools	  from	  the	  outset?	  What	  do	  earlier	   forms	  of	   cultural	   learning	  mechanisms	   look	   like,	   and	  what	   are	   the	   relative	  roles	   of	   built-­‐in	   constraints,	   and	   domain-­‐general	   learning	   mechanisms	   in	   the	  development	   of	   learning	   abilities	   from	   their	   early	  manifestations	   to	   their	  mature	  forms?	  To	  answer	  these	  questions,	  populations	  with	  less	  experience	  in	  learning	  are	  particularly	   informative	   to	   study.	   In	  particular,	   infants	   in	   their	   first	  months	  of	   life	  have	  had	  less	  time	  to	  learn	  techniques	  for	  learning,	  and	  also	  do	  not	  have	  the	  benefit	  	   11	  of	  symbolically	  based	  learning	  strategies	  that	  are	  facilitated	  by	  language.	  Next,	  I	  turn	  to	  the	  foundations	  of	  such	  selectivity	  in	  learning	  during	  infancy.	  	  	  1.4	   Learning	  in	  preverbal	  infants.	  	  How	  do	  preverbal	  infants	  solve	  the	  vastly	  complex	  problem	  of	  knowing	  what,	  how,	  and	  from	  whom	  to	  learn?	  To	  learn	  successfully,	  attention	  must	  be	  constrained,	  rendering	   some	  aspects	  of	   the	  world	   easier	   to	   attend	   to	   and	   remember.	  With	   few	  opportunities	  to	  learn	  what	  aspects	  of	  the	  environment	  are	  important	  to	  attend	  to,	  how	   might	   young	   infants	   constrain	   their	   attention	   and	   learning?	   Humans	   seem	  particularly	  likely	  to	  retain	  certain	  kinds	  of	  information—that	  is,	  they	  are	  biased	  in	  terms	  of	  what	  they	  learn.	  1.4.1	  	   Content	  biases	  in	  learning	  Some	  have	  proposed	  that	  humans	  possess	  content	  biases	  regarding	  features	  of	   the	   ancestral	   environment	   that	   consistently	   posed	   a	   threat	   to	   survival.	   For	  example,	   snake,	   spider,	   and	   height	   phobias	   are	   all	   extremely	   common	   today,	  perhaps	  because	  although	  it	  is	  rare	  to	  die	  from	  spider	  bites	  today,	  individuals	  who	  feared	   these	   threats	   during	   human	   evolution	   were	   more	   likely	   to	   survive	   and	  reproduce.	   Indeed,	   these	   common	  phobias	   are	   often	   compared	   to	  modern	   threats	  like	   automobile	   travel,	   which	   kill	   many	   more	   people	   each	   year	   than	   snakes	   or	  spiders,	  and	  yet	  almost	  no	  one	  is	  afraid	  to	  get	  in	  a	  car.	  	  Preverbal	   infants’	   reasoning	   appears	   to	   be	   constrained	   by	   domain-­‐specific	  rules	  and	   limits.	  Research	  with	  preverbal	   infants	  suggests	   that	  content	  biases	  may	  be	  present	  in	  infants’	  learning	  from	  the	  first	  year	  of	  life.	  For	  example,	  content	  biases	  for	   learning	  about	   the	  edibility	  of	  plants	  have	  been	  shown	   in	  6-­‐	  and	  18-­‐month-­‐old	  	   12	  infants	  (Wertz	  &	  Wynn,	  2014).	  In	  this	  work,	  infants	  preferentially	  reached	  for	  food	  from	   a	   plant	   vs.	   from	   an	   artifact,	   but	   showed	   no	   such	   selectivity	   when	   choosing	  between	   things	   to	   use.	   Critically,	   this	   preference	   disappeared	   in	   the	   absence	   of	   a	  social	  model	  demonstrating	   that	  a	  plant	   is	  good	   to	  eat.	  This	  biased	   learning	  about	  edibility	   of	   plants	   may	   be	   due	   to	   the	   important	   risk	   posed	   by	   plant	   toxins	   over	  evolutionary	   history,	   such	   that	   infants	   will	   prefer	   only	   plants	   that	   others	   have	  demonstrated	  to	  be	  safe.	  	  Similar	  content	  biases	  appear	   to	  exist	   for	   infants’	   learning	  of	   fear	  of	  snakes	  (DeLoache	  &	  LoBue,	  2009).	   In	   this	   study,	   infants	   looked	   longer	   toward	  a	   film	  of	   a	  snake	  in	  motion	  than	  a	  film	  of	  other	  animals	  in	  motion,	  but	  only	  when	  the	  films	  were	  accompanied	  by	  a	  fearful	  adult	  voice.	  In	  contrast,	  infants	  did	  not	  show	  preferential	  looking	   when	   the	   animals	   were	   presented	   without	   an	   accompanying	   fear	   voice.	  These	  results	   suggest	   that	   rather	   than	  possessing	  an	   innate	   fear	  of	   snakes,	   infants	  may	   be	   flexibly	   prepared	   to	   easily	   acquire	   fear	   of	   snakes;	   for	   instance,	   in	   the	  presence	   of	   another	   human	   who	   indicates	   that	   snakes	   are	   threatening.	   Critically,	  infants’	   readiness	   to	   learn	   fear	   applies	   specifically	   to	   snakes,	   and	   not	   to	   various	  comparison	  animals.	  Together,	  the	  above	  examples	  support	  the	  existence	  of	  content	  biases	   in	   infant	   cultural	   learning,	   pointing	   to	   constraints	   in	   the	   relative	   ease	  with	  which	   certain	   types	   of	   content	   can	   be	   learned.	   In	   addition,	   they	   highlight	   the	  importance	   of	   humans	   as	   sources	   of	   information,	   as	   infants	   showed	   preferential	  learning	  only	  in	  the	  presence	  of	  a	  human	  informant.	  In	  addition	   to	  having	  a	  bias	   for	   learning	  particular	  kinds	  of	  content,	  human	  learners	   are	   also	   biased	   in	   terms	   of	  who	  or	  what	   they	   learn	   from.	  A	   challenge	   all	  	   13	  learners	  face	  is	  to	  identify	  sources	  of	  cultural	  information,	  interpret	  others’	  actions	  and	   infer	   the	   intent	   behind	   their	   communications.	   To	   do	   so	   effectively	   requires	   a	  number	   of	   basic	   cognitive	   capacities,	   including	   but	   not	   limited	   to	   an	   interest	   in	  	  people,	  the	  ability	  to	  infer	  goals	  (in	  order	  to	  identify	  the	  model’s	  intended	  target	  or	  outcome	  of	  their	  modeled	  actions),	  and	  to	  understand	  communicative	  gestures	  and	  evaluative	  states	  such	  as	  pointing,	  eye	  gaze,	  and	  emotions	  (Bruner,	  1983,	  1993).	  1.4.2	  	   Source	  biases	  in	  learning	  From	   early	   on,	   young	   humans	   appear	   to	   categorize	   the	   world	   into	   agents	  who	  possess	  goals	  and	  desires,	  and	  non-­‐agents	  that	  adhere	  to	  physical	  laws	  but	  lack	  psychological	  states	  such	  as	  goals	  and	  preferences.	  Newborn	  infants	  prefer	  to	  look	  at	   goal-­‐directed	   actions	   (grasping	   a	  ball)	   over	  non-­‐goal	   oriented	  movement	   just	   2	  days	   after	   birth	   (Craighero,	   Leo,	   Umiltà,	   &	   Simion,	   2011).	   Six-­‐month-­‐old	   infants	  attribute	   preferences	   and	   goals	   to	   agents,	   but	   not	   to	   inanimate	  mechanical	   claws	  (Woodward,	   1998).	   They	   also	   attribute	   valence	   to	   actions	   (good	   or	   bad),	   when	  interactions	   are	   between	   animate-­‐like	   agents,	   but	   not	   to	   similar	   interactions	  between	   non-­‐agentive	   shapes	   (Hamlin,	   Wynn,	   &	   Bloom,	   2007).	   An	   additional	  expectation	  that	  human	  infants	  seem	  to	  possess	  is	  for	  other	  people	  to	  be	  sources	  of	  information.	   From	   shortly	   after	   birth,	   infants	   are	   very	   interested	   in	   other	   people.	  They	  are	  especially	  attentive	  to	  faces,	  human	  voices,	  and	  especially	  faces	  that	  engage	  them	   in	   eye	   contact	   (DeCasper	   &	   Fifer,	   1980;	   Farroni,	   Csibra,	   Simion,	   &	   Johnson,	  2002;	   Goren,	   Sarty,	   &	   Wu,	   1975).	   These	   preferences	   serve	   infants’	   learning	   by	  attending	   to	   expressive	   parts	   of	   the	   body	   that	   are	   likely	   to	   communicate	  information.	  	  	   14	  Faces.	  Within	  hours	  after	  birth,	  human	  neonates	  show	  a	  preference	  for	  human	  faces	  and	  face-­‐like	  stimuli	  (Cassia,	  Turati,	  &	  Simion,	  2004;	  Farroni	  et	  al.,	  2005;	  Valenza,	  Simion,	  Cassia,	  &	  Umiltà,	  1996),	  as	  well	  as	  for	  faces	  they	  have	  been	  exposed	  to	  (Walton	  &	  Bower,	  1993;	  Walton,	  Armstrong,	  &	  Bower,	  1997).	  Furthermore,	  they	  appear	  to	  use	  this	  visual	  information;	  averaging	  across	  faces	  they	  have	  encountered	  to	  form	  a	  composite	  “prototype”	  (Langlois	  &	  Roggman,	  1990).	  Neonates	  are	  able	  to	  form	  prototypes	  very	  rapidly,	  capable	  of	  creating	  a	  composite	  after	  a	  single,	  one-­‐minute	  exposure	  to	  each	  of	  several	  faces	  (Walton	  &	  Bower,	  1993).	  Important	  to	  the	  task	  of	  being	  a	  discriminating	  learner,	  prototypes	  are	  used	  as	  a	  basis	  for	  preference;	  strangers’	  faces	  that	  are	  similar	  to	  the	  prototypical	  average	  are	  preferred	  over	  those	  that	  deviate	  (Langlois	  &	  Roggman,	  1990;	  Pascalis,	  de	  Schonen,	  Morton,	  Deruelle,	  &	  Fabre-­‐Grenet,	  1995;	  Walton,	  Bower,	  &	  Bower,	  1992).	  While	  this	  preference	  for	  strangers	  with	  prototypical	  appearances	  seems	  puzzling,	  in	  real	  life	  it	  may	  serve	  to	  help	  neonates	  develop	  a	  preference	  for	  attending	  to	  individuals	  who	  spend	  the	  most	  time	  in	  close	  proximity	  to	  them,	  such	  as	  their	  primary	  caretaker,	  as	  well	  as	  those	  who	  look	  like	  him/her	  (kin).	  Indeed,	  neonates	  do	  show	  a	  preference	  for	  their	  mother’s	  face	  compared	  to	  a	  stranger’s	  by	  just	  48	  hours	  after	  birth	  (Bushnell,	  Sai,	  &	  Mullin,	  1989;	  Pascalis	  et	  al.,	  1995;	  Walton	  et	  al.,	  1992).	  By	  three	  months,	  infants	  also	  exhibit	  a	  preference	  for	  faces	  of	  the	  same	  gender	  as	  their	  primary	  caregiver	  (Quinn,	  Yahr,	  Kuhn,	  Slater,	  &	  Pascalis,	  2002)	  and	  the	  racial	  group	  in	  their	  social	  environment	  (Bar-­‐Haim,	  Ziv,	  Lamy,	  &	  Hodes,	  2006;	  Kelly	  et	  al.,	  2005),	  showing	  early	  generalization	  of	  experiences	  with	  individual	  faces	  to	  groups	  of	  	   15	  people	  who	  share	  similar	  traits.	  These	  preferences	  might	  allow	  infants	  to	  direct	  preferential	  attention	  to	  potential	  sources	  of	  relevant	  cultural	  information.	  Speech	   sounds.	   Newborn	   infants	   also	   prefer	   speech	   sounds	   compared	   to	  non-­‐speech	  sounds	  of	  similar	  pitch	  and	  amplitude	  (Vouloumanos	  &	  Werker,	  2007).	  Infant-­‐directed	   speech,	   characterized	   by	   higher	   and	   broader	   pitch,	   greater	  amplitude	   variation	   and	   slower	   speed	   than	   adult-­‐oriented	   speech	   also	   increases	  infants’	  attention	  and	   interest	   toward	   the	  speaker	   (Cooper	  &	  Aslin,	  1990;	  Fernald,	  1985;	  Pegg,	  Werker,	  &	  McLeod,	  1992;	  Werker,	  Pegg,	  &	  McLeod,	  1994).	  	  Eye	  gaze.	  An	  assumption	  that	  infants	  appear	  to	  hold	  is	  that	  others	  are	  willing	  and	  able	  to	  share	  information	  with	  them.	  As	  early	  as	  9-­‐months,	  infants	  look	  at	  adults	  while	   simultaneously	   making	   communicative	   gestures	   such	   as	   pointing	   and	  reaching	  in	  bids	  for	  help;	  these	  efforts	  increase	  with	  age	  and	  with	  difficulty	  of	  tasks	  (Goubet,	   Rochat,	   Maire	   Leblond,	   &	   Poss,	   2006).	   Infants	   appear	   to	   be	   especially	  attuned	   to	   cues	   indicating	   that	   something	   is	   being	   taught,	   including	   ostensive	  pedagogical	   cues	   and	   child-­‐directed	   speech	   (Csibra	   &	   Gergely,	   2006;	   Gergely	   &	  Csibra,	  2005).	  Ostensive	  communication	  such	  as	  overt	  speech	  and	  eye	  contact	  set	  up	  “teaching	   moments”	   that	   prepare	   young	   learners	   for	   demonstrations	   intended	   to	  transfer	   knowledge.	   These	   attention	   biases	   for	   social	   sources	   of	   information	  may	  help	  learners	  learn	  both	  from	  and	  about	  other	  people.	  1.4.3	  	   Inferring	  content	  of	  actions	  On	   top	   of	   having	   a	   desire	   to	   attend	   to	   cultural	   informants,	   learners	   must	  identify	  what	  is	  being	  taught.	  A	  fundamental	  aspect	  of	  social	  learning	  is	  to	  recognize	  that	  others’	  actions	  are	  carried	  out	  in	  the	  service	  of	  underlying	  goals:	  Alice	  reached	  	   16	  out	   because	   she	   wanted	   to	   grab	   the	   cup.	   Learning	   to	   achieve	   desired	   outcomes	  (getting	   a	   cup)	   requires	   accurately	   identifying	   the	   relevant	   sequence	   of	   actions	   a	  model	   employed	   (reaching	   out	   at	   an	   appropriate	   angle,	   with	   a	   specific	   hand	  formation	   suitable	   for	   grasping)	   to	   accomplish	   their	   goal	   (a	   particular	   cup),	   and	  evaluating	  the	  outcome	  (did	  the	  model	  get	  the	  cup?).	  Since	  goals	  are	  invisible	  mental	  states,	  their	  content	  must	  be	  inferred	  from	  overt	  behaviours.	  	  Numerous	   studies	   show	   that	   from	   early	   on,	   infants	   can	   infer	   goals	   from	  observable	   actions	   in	   a	   number	   of	   ways.	   Five-­‐month-­‐old	   infants	   are	   able	   to	   infer	  others’	   goals	   from	   repeated	   reaching	   behaviour	   (Luo	   &	   Baillargeon,	   2005;	  Woodward,	  1998).	  By	  12	  months,	   they	  also	  use	   cues	  of	   equifinality—reaching	   the	  same	   destination	   via	   different	   paths	   (Gergely,	   Nádasdy,	   Csibra,	   &	   Bíró,	   1995),	  engaging	   in	   contingent	   behaviour	   (Johnson,	   2003;	   Johnson,	   Alpha	   Shimizu,	   &	   Ok,	  2007;	  Shimizu	  &	  Johnson,	  2004),	  and	  rationality—reaching	  a	  destination	  efficiently	  (Gergely	   et	   al.,	   1995;	   Gergely,	   Bekkering,	   &	   Király,	   2002;	   Kamewari,	   Kato,	   Kanda,	  Ishiguro,	  &	  Hiraki,	  2005)	  to	  infer	  the	  content	  of	  others’	  goals.	  	  The	  connection	  between	  an	  agent	  and	  their	  goal	  does	  not	  have	  to	  be	  physical	  for	   infants	   to	   infer	   their	   content.	   By	   their	   first	   birthday,	   infants	   are	   capable	   of	  inferring	   distal	   goals	   through	   visual	   attention	   (Johnson	   et	   al.,	   2007;	   Moll	   &	  Tomasello,	   2004;	   Woodward,	   2003),	   failed	   actions	   (Brandone	   &	   Wellman,	   2009;	  Hamlin,	  Hallinan,	  &	  Woodward,	  2008),	  and	  compound	  goals	  that	  require	  2	  or	  more	  contingent	  steps	  to	  complete	  (Sommerville	  &	  Woodward,	  2005).	  They	  will	  even	  use	  pointing	   as	   a	  way	   to	   communicate	   about	   and	  draw	  others’	   attention	   to	   their	   own	  distant	   goals	   (Tomasello,	   Carpenter,	   &	   Liszkowski,	   2007;	   Liszkowski,	   Carpenter,	  	   17	  Henning,	  Striano,	  &	  Tomasello,	  2004).	  Thus,	  preverbal	  infants	  are	  equipped	  to	  infer	  the	  content	  of	  other	  people’s	  proximal	  and	  distal	  goals	  via	  a	  variety	  of	  means	  by	  the	  end	  of	  their	  first	  year	  of	  life.	  	  1.4.4	   Using	  emotions	  	  Before	   infants	   can	   understand	   or	   produce	   language,	   they	   are	   sensitive	   to	  affective	   communications	   and	   what	   they	   reveal	   about	   others’	   mental	   states	   and	  their	  relationship	  to	  the	  external	  world.	  Newborns	  have	  been	  found	  to	  differentiate	  happy	   to	   sad	  voices	   in	   their	   own	   language	   (Mastropieri	  &	  Turkewitz,	   1999).	  By	  4	  months	  after	  birth,	   infants	  reliably	  differentiate	  between	  and	  sometimes	  recognize	  facial	   expressions	   of	   anger,	   fear	   and	   happiness	   (Haviland	   &	   Lelwica,	   1987;	  LaBarbera,	  Izard,	  Vietze,	  &	  Parisi,	  1976;	  Montague	  &	  Walker-­‐Andrews,	  2001).	  Five-­‐month-­‐olds	   can	   discriminate	   among	   vocal	   expressions	   of	   emotion	   presented	   in	  infant-­‐directed	   speech	   in	   their	   native	   or	   a	   foreign	   language	   (Fernald,	   1993),	   or	  adult-­‐directed	   speech	   accompanied	   by	   a	   facial	   expression	   (Walker-­‐Andrews	   &	  Grolnick,	  1983;	  Walker-­‐Andrews	  &	  Lennon,	  1991).	  Five-­‐month-­‐olds	  are	  also	  adept	  at	   discriminating	   between	   dynamically	   presented	   expressions	   of	   emotion	   (Caron,	  Caron,	  &	  MacLean,	  1988).	  By	   5.5	   months,	   infants	   have	   been	   shown	   to	   regulate	   their	   own	   behaviour	  toward	   objects	   that	   other	   individuals	   had	   previously	   emoted	   towards;	   infants	   are	  more	   likely	   to	   touch	   an	   object	   that	   had	   received	   a	   happy	   reaction	   compared	   to	   a	  fearful	   reaction,	   so	   long	   as	   there	   are	   redundant	   facial	   and	   vocal	   emotional	  expressions	  available	  (Vaillant-­‐Molina	  &	  Bahrick,	  2012).	  By	  the	  end	  of	  the	  first	  year,	  infants	  reliably	  regulate	  their	  behaviours	  in	  response	  to	  emotional	  communications	  	   18	  through	   a	   single	   modality	   (Klinnert,	   Emde,	   Butterfield,	   &	   Campos,	   1986;	   Hornik,	  Risenhoover,	   &	   Gunnar,	   1987;	   Mumme,	   Fernald,	   &	   Herrera,	   1996;	   Sorce,	   Emde,	  Campos,	   &	   Klinnert,	   1985).	   While	   a	   combination	   of	   vocal	   and	   facial	   expressions	  continues	  to	  be	  most	  potent	  for	  affecting	  infants’	  behaviour,	  vocal	  cues	  alone	  appear	  to	   be	  more	   effective	   than	   facial	   cues	   alone	   (Mumme	   et	   al.,	   1996;	   Vaish	  &	   Striano,	  2004).	  	  Also	  toward	  the	  end	  of	  the	  first	  year,	  infants	  begin	  to	  actively	  seek	  out	  other	  people	  for	  emotional	  information	  regarding	  an	  ambiguous	  event	  (Striano	  &	  Rochat,	  2000;	  Walden	  &	  Ogan,	  1988).	  Around	  the	  same	  time,	  infants	  demonstrate	  an	  ability	  to	  use	  an	  emoter’s	  affective	  displays	  toward	  objects	  to	  predict	  her	  subsequent	  object	  choice;	  they	  fail	  to	  do	  so	  if	  the	  emoter	  was	  looking	  away	  from	  the	  object,	  suggesting	  that	  they	  understand	  emotions	  as	  referential	  communications	  about	  a	  target	  (Flom	  &	  Johnson,	  2011;	  Phillips,	  Wellman,	  &	  Spelke,	  2002;	  Repacholi,	  1998).	  	  The	   above	   discussion	   indicates	   that	   preverbal	   infants	   are	   demonstrably	  selective	  in	  knowledge	  content	  and	  source,	  preferentially	  learning	  adaptive-­‐relevant	  content	   from	   social	   sources.	   Furthermore,	   they	   are	   able	   to	   perceive	   the	   intended	  outcomes	   of	   others’	   actions	   and	   reactions,	   through	   direct	   instruction	   as	   well	   as	  through	   observation.	   These	   abilities	   set	   the	   stage	   for	   cultural	   learning	   in	   infancy,	  channeling	   young	   humans’	   attention	   toward	   cultural	   conspecifics,	   and	   equipping	  them	  to	  understand	  conspecifics’	  behaviours.	  With	  these	  abilities	  alone,	  infants	  may	  be	  capable	  of	  identifying	  causally	  effective	  actions	  for	  achieving	  a	  number	  of	  desired	  physical	  outcomes,	  such	  as	  reaching	  a	  toy,	  or	  opening	  a	  jar.	  However,	  by	  themselves,	  these	  abilities	  are	  insufficient	  for	  learning	  about	  conventional	  behaviours,	  for	  these	  	   19	  vary	  by	  culture,	  yet	  typically	  are	  not	  evaluable	  by	  physical	  causality.	  To	  learn	  about	  conventional	  behaviours,	  learners	  must	  rely	  on	  identifying	  knowledgeable	  sources,	  or	   by	   analyzing	   the	   conventionality	   of	   the	   behaviours	   themselves.	   Next,	   I	   turn	   to	  infants’	  capacities	  for	  identifying	  good	  sources	  of	  cultural	  knowledge.	  1.5	   Evaluating	  sources	  of	  cultural	  knowledge	  	  Successful	   cultural	   learning	   requires	   identifying	   appropriate	   models;	  something	   that,	   as	   reviewed	   above,	   young	   children	   are	   already	   adept	   at.	   Are	  preverbal	   infants	   also	   sensitive	   to	   cues	   that	   may	   be	   indicative	   of	   relevant	   and	  accurate	  information	  and	  sources?	  There	  is	  mounting	  evidence	  that	  even	  very	  young	  infants	   discriminate	   between	   potential	   sources	   on	   a	   number	   of	   dimensions,	   and	  moderate	  their	  behaviours	  based	  on	  these	  differential	  evaluations.	  	  1.5.1	  	   Evaluation	  by	  behavioural	  valence	  Within	  one’s	  cultural	  group,	  individuals	  vary	  in	  the	  extent	  to	  which	  they	  are	  able	   and	  willing	   to	   share	  knowledge.	  As	  discussed	  above,	  preschool	   aged	   children	  are	   capable	   of	   determining	   whether	   someone	   is	   a	   competent	   and/or	   benevolent	  informant.	   This	   sensitivity	   appears	   to	   be	   ontogenetically	   precocious,	   as	   even	  preverbal	   infants	   form	   evaluations	   based	   on	   the	   degree	   to	   which	   an	   individual’s	  actions	  are	  cooperative	  or	  prosocial.	  As	  early	  as	  3-­‐months	  after	  birth,	  and	  robustly	  through	   the	   first	   year,	   infants	   appear	   sensitive	   to	   the	   valence	   of	   others’	   actions,	  preferring	  those	  who	  behave	  prosocially	  vs.	  antisocially	  (Hamlin	  et	  al.,	  2007;	  Hamlin	  &	  Wynn,	  2011).	  In	  addition	  to	  informing	  their	  evaluations,	  benevolence	  appears	  to	  also	   influence	   infants’	   choice	   of	   cultural	  models:	   16	  month	   olds	   fail	   to	  match	   the	  	   20	  food	   preferences	   of	   a	   previously	   mean	   individual,	   but	   readily	   imitate	   the	   food	  preferences	  of	  nice	  and	  neutral	  individuals	  (Hamlin	  &	  Wynn,	  2012).	  Infants	   also	   prefer	   to	   learn	   from	   informants	  who	  were	   reliable	   in	   the	   past	  (see	  Harris	  &	  Lane,	  2014	  for	  a	  review).	  For	  example,	  8-­‐month-­‐old	  infants	  selectively	  look	   in	   the	  direction	  cued	  by	  a	  previously	   reliable	   informant,	  but	  not	  a	  previously	  unreliable	   informant	   (Tummeltshammer,	  Wu,	   Sobel,	  &	  Kirkham,	  2014b).	   Fourteen	  month	  olds	  are	  more	  likely	  to	  imitate	  individuals	  who	  reacted	  appropriately	  in	  the	  past,	  and	  discount	   individuals	  who	  displayed	   inappropriate	  emotional	  reactions	  to	  the	   contents	   of	   a	   container,	   in	   whether	   to	   follow	   their	   eye	   gaze	   (Chow,	   Poulin-­‐Dubois,	  &	  Lewis,	  2008),	   and	   to	   imitate	   their	  unusual	  method	  of	   turning	  on	  a	   light	  (Poulin-­‐Dubois,	   Brooker,	   &	   Polonia,	   2011).	   They	   also	   tend	   to	   imitate	   competent	  informants	  who	  use	  objects	  correctly	  compared	  to	  incompetent	  informants	  who	  use	  objects	  incorrectly	  (Zmyj,	  Buttelmann,	  Carpenter,	  &	  Daum,	  2010).	  	  The	   research	   reviewed	   above	   on	   infants’	   social	   evaluations	   and	   learning	  suggest	  that	   infants	  distinguish	  potential	   informants	  based	  on	  their	  past	  reliability	  and	  helpfulness,	  both	  of	  which	  are	  relevant	  attributes	  of	  a	  good	  cultural	  informant.	  Next,	  I	  will	  focus	  on	  informant	  qualities	  that	  are	  important	  to	  learning	  about	  cultural	  conventional	   behaviours	   from	   individuals	   infants	   have	  no	  prior	   knowledge	   about:	  they	   neither	   displayed	   inaccurate	   or	   nonsensical	   behaviours,	   nor	   were	   their	  behaviours	   nice	   or	   mean.	   I	   will	   review	   suggestive	   evidence	   that	   even	   young,	  preverbal	   infants	   may	   be	   capable	   of	   learning	   about	   and	   evaluating	   inherently	  neutral	  actions,	  using	  existing	  skills	  such	  as	  their	  sensitivity	  to	  group	  membership,	  statistical	  regularity,	  and	  others’	  emotional	  reactions.	  	  	   21	  1.5.2	  	   Evaluation	  by	  group	  membership	  One	   way	   in	   which	   preverbal	   infants’	   sensitivity	   to	   differential	   informant	  quality	   has	   been	   studied	   is	   by	   looking	   at	   their	   attention	   and	   liking	   for	   a	   pair	   of	  individuals	  differing	  on	  a	  relevant	  dimension,	  such	  as	   familiarity,	  or	  gender.	  While	  these	   studies	   do	   not	   directly	   measure	   learning,	   selective	   attention	   and	   liking	  arguably	   create	   opportunities	   for	   learning,	   as	   learners	   pay	   more	   attention	   to	  positively	   evaluated	   targets,	   and	   in	   doing	   so	   foregoing	   exposure	   to	   alternative	  models.	   Using	   these	   preference	   methods,	   infants	   in	   their	   first	   six	   months	   of	   life	  evaluate	  others	  based	  on	  stable	   indicators	  of	   social	  group	  membership,	  preferring	  individuals	  of	   their	   same	   racial	   group,	   and	   those	  who	   speak	   their	  native	   language	  (Bar-­‐Haim	  et	   al.,	   2006;	  Kinzler,	  Dupoux,	  &	   Spelke,	   2007).	  By	   the	   end	  of	   their	   first	  year,	   infants	   are	   also	  more	   likely	   to	   accept	   toys	   and	   food	   offered	   by	   a	   speaker	   of	  infants’	  own	  dialect,	  but	  not	   from	  someone	  of	   their	  own	  race	  (Kinzler	  et	  al.,	  2007;	  Shutts,	   Kinzler,	   McKee,	   &	   Spelke,	   2009).	   Infants	   even	   form	   evaluations	   based	   on	  stable	   behavioural	   markers	   such	   as	   food	   preferences,	   namely,	   whether	   a	   puppet	  shares	   infants’	   own	   expressed	   food	   preference	   (Mahajan	   &	   Wynn,	   2012).	   Like	  dialect	   and	   morphology,	   food	   preferences	   co-­‐vary	   with	   social	   group,	   and	   are	  strongly	   influenced	   by	   prenatal	   exposure	   to	   maternal	   diet	   and	   postnatal	  observational	   learning	   (Addessi,	   Galloway,	   Visalberghi,	   &	   Birch,	   2005;	   Baeyens,	  Vansteenwegen,	   De	   Houwer,	   &	   Crombez,	   1996;	   Mennella,	   Jagnow,	   &	   Beauchamp,	  2001).	   Infants’	  evaluations	  are	  not	   limited	   to	   those	  who	  share	  or	  differ	   from	  their	  food	  preferences;	  they	  also	  extend	  to	  3rd	  parties	  who	  interact	  with	  those	  individuals,	  preferring	  individuals	  who	  help	  those	  who	  share	  their	  food	  preferences,	  as	  well	  as	  	   22	  individuals	   who	   harm	   those	   who	   do	   not	   (Hamlin,	   Mahajan,	   Liberman,	   &	   Wynn,	  2013).	   Thus,	   preverbal	   infants	   seem	   quite	   capable	   of	   identifying	   individuals	   who	  may	   be	   in	   their	   social	   in-­‐group	   and	  who	  will	   likely	   be	   good	   sources	   of	   culturally	  relevant	  information.	  	  1.5.3	   Evaluation	  by	  statistical	  information	  	  Statistical	   regularity	   is	   an	   important	   component	   to	   infants’	   early	   visual	  preferences.	  Three-­‐month-­‐old	  infants	  prefer	  the	  faces	  of	  the	  gender	  of	  their	  primary	  caregiver	  (Quinn	  et	  al.,	  2002).	  Six-­‐month-­‐old	  infants	  prefer	  faces	  that	  more	  closely	  resemble	   the	   statistical	   average	   of	   faces	   they	   have	   encountered	   (Rubenstein,	  Kalakanis,	   &	   Langlois,	   1999).	   The	   influence	   of	   statistical	   regularity	   extends	   to	  preverbal	  infants’	  evaluations	  of	  informants	  on	  the	  basis	  of	  their	  behaviours.	  Young	  infants	   preferentially	   attend	   to	   informants	   who	   act	   with	   a	   high	   degree	   of	  consistency	   (Tummeltshammer,	   Mareschal,	   &	   Kirkham,	   2014a).	   Infants	   in	   their	  second	   year	   also	   evaluate	   reliability	   by	   the	   extent	   to	   which	   a	   model	   uses	   a	  conventional	   label	   or	   demonstrates	   the	   conventional	   usage	   for	   a	   familiar	   object	  (Poulin-­‐Dubois	  et	  al.,	  2011;	  Zmyj	  et	  al.,	  2010).	  This	  result	  suggests	  that	  they	  are	  able	  to	   track	   how	   people	   typically	   behave	   in	   a	   given	   situation,	   and	   compare	   new	  behaviours	   against	   a	   prototype	   of	   past	   experiences	   that	   held	   similar	   relevant	  features.	  Could	  young	  preverbal	   infants	  apply	  their	   facility	  with	  tracking	  statistical	  regularities	   to	   evaluate	   and	   learn	   about	   intrinsically	   neutral	   targets?	   Given	   the	  highly	   consistent	   nature	   of	   conventional	   actions,	   infants	  may	   be	   able	   to	   use	   their	  sensitivity	   to	   statistical	   information	   to	   evaluate	   people	   who	   engage	   in	   different	  behaviours	   that	   are	   inherently	   neither	   good	   nor	   bad,	   but	   are	   nevertheless	  	   23	  differentially	   valued.	   This	   question	   will	   be	   examined	   in	   Chapters	   2	   and	   3	   of	   this	  dissertation.	  1.5.4	   Evaluation	  by	  emotional	  cues.	  As	   reviewed	   above,	   infants’	   early	   emerging	   understanding	   of	   emotions	   as	  communicative	   signals	   allow	   them	   to	   tap	   into	   other	   people’s	   evaluations	   of	   an	  ambiguous	   situation	   to	   inform	   their	   own	   responses	   accordingly.	   Do	   they	   also	   use	  others’	  emotional	  reactions	  as	  a	  tool	  for	  learning	  about	  unfamiliar	  people	  and	  their	  actions?	  	  To	   date,	   three	   studies	   have	   examined	   relevant	   aspects	   of	   infants’	  observational	  learning	  using	  emotional	  cues.	  Repacholi	  &	  Meltzoff	  (2007)	  found	  that	  18-­‐month-­‐olds	  are	  able	  to	  learn	  evaluations	  of	  objects	  based	  on	  emotional	  reactions.	  In	   the	   study,	   infants	  watched	   an	   individual	  manipulate	   objects,	  while	   an	   observer	  expressed	   either	   angry	   or	   neutral	   affect	   toward	   these	   actions.	   When	   given	   the	  opportunity	  to	  play	  with	  the	  objects,	  infants	  who	  had	  observed	  the	  observer’s	  anger	  avoided	  the	  objects,	  whereas	  those	  who	  had	  observed	  a	  neutral	  reaction	  were	  more	  likely	   to	   manipulate	   the	   objects.	   However,	   in	   this	   study,	   infants’	   learning	   was	  specific	  to	  the	  individual	  who	  displayed	  their	  emotional	  reaction,	  and	  behaved	  in	  a	  manner	  consistent	  with	  the	  emoter’s	  emotional	  expression	  only	  when	  he	  or	  she	  was	  present.	  Infants	  did	  not	  regulate	  their	  behaviours	  across	  contexts	  when	  the	  original	  emoter	   was	   absent,	   suggesting	   they	   did	   not	   infer	   that	   positive	   and	   negative	  reactions	   were	   due	   to	   objective	   characteristics	   of	   the	   objects	   (i.e.	   Mom	   had	   a	  negative	  reaction	  to	  the	  object	  because	  it	  is	  bad,	  or	  dangerous).	  	  	   24	  While	   learning	   individual	   preferences	   is	   one	   desirable	   outcome	   of	   cultural	  learning,	   amassing	   shared,	  more	   generalizable	   knowledge	   about	  what	   objects	   are	  good	  to	  use	  and	  which	  individuals	  are	  safe	  to	  interact	  with	  can	  help	  infants	  navigate	  a	  broader	  range	  of	  situations.	  Eighteen-­‐month-­‐old	  infants	  do	  appear	  to	  infer	  object-­‐specific	  evaluations	  when	  it	  was	  preceded	  by	  ostensive	  signals	  (i.e.	  the	  emoter	  first	  looked	   at	   the	   infant	   before	   evaluating	   the	   object;	   Egyed,	   Kiraly,	  &	  Gergely,	   2013).	  When	  ostensive	  signals	  were	  lacking,	  they	  interpreted	  the	  evaluation	  to	  be	  a	  person-­‐specific	  preference,	   and	  did	  not	  use	   the	  emotional	   reactions	   to	   regulate	   their	  own	  interactions	  with	  the	  object	  (i.e.	  they	  did	  not	  generalize	  the	  evaluations	  to	  pertain	  to	  their	  own	  interactions	  with	  the	  object).	  	  A	   recent	   study	   by	   Vaish,	   Grossmann,	   &	   Woodward	   (2015)	   suggest	   that	  infants’	  willingness	  to	  generalize	  evaluations	  may	  be	  moderated	  by	  their	  emotional	  valence.	   The	   study	   found	   that	   2-­‐year-­‐olds	   were	   more	   likely	   to	   generalize	   an	  emoter’s	  negative	  evaluations	  of	  an	  object	  to	  new	  individuals,	  whereas	  they	  did	  not	  tend	   to	   do	   so	   for	   positive	   emotions.	   Thus,	   negative	   emotional	   reactions	   may	   be	  attributed	   to	   the	   object	   itself,	   whereas	   positive	   emotional	   reactions	   may	   be	   an	  attitude	  held	  by	  a	  specific	  individual.	  	  Could	   younger	   infants	   as	   bystanders	   extract	   target-­‐centered	   information	  from	  other	  people’s	   emotional	   reactions	   and	  use	   it	   to	   guide	   their	  own	  behaviours	  with	  regards	  to	  people,	  and	  to	  objects?	  This	  question	  will	  be	  examined	  in	  Chapter	  4	  of	  this	  dissertation.	  	  	  	  	   25	  1.6	   Cultural	  differences	  in	  cultural	  learning?	  	  The	   discussion	   thus	   far	   has	   focused	   heavily	   on	   universal	   aspects	   of	   early	  social	   cognition	   underlying	   humans’	   impressive	   ability	   to	   learn	   culture.	  Conspicuously	   absent	   is	   a	   treatment	   of	   cultural	   variation	   itself.	   Indeed,	   cultural	  differences	  are	  plain	  by	  early	  childhood,	  and	  only	  grow	  thereafter.	  But	  when	  exactly	  do	   these	   differences	   begin	   to	   emerge?	   Do	   infants	   and	   toddlers	   across	   the	   world	  more	  or	   less	  resemble	  one	  another	  until	   they	  enter	  school	  age?	  Or	  do	  they	  exhibit	  differences	  much	   sooner?	   And	  more	   specifically,	   how	   similar	   or	   different	   are	   the	  ways	  in	  which	  infants	  growing	  up	  in	  different	  cultures	  engage	  in	  cultural	  learning?	  Do	  young	  humans	  engage	  in	  cultural	  learning	  using	  a	  universal	  set	  of	  mechanisms,	  applied	   in	   uniform	   ways,	   or	   does	   cultural	   learning	   set	   learners	   on	   divergent	  trajectories	  early	  on?	  	  At	   the	   broadest	   level,	   the	   social	   and	   economic	   arrangements	   of	   the	   larger	  society	  in	  which	  development	  unfolds	  impacts	  infants’	  earliest	  experiences.	  Infants	  born	   in	   traditional	  societies	  are	  typically	  surrounded	  by	  peers	  and	  by	  members	  of	  the	   extended	   family,	   many	   of	   who	   actively	   contribute	   to	   child	   rearing,	   and	  participate	   extensively	   in	   the	   infant’s	   life	   (Chen,	   French,	   Schneider,	   2006;	   LeVine,	  1988).	   In	   contrast,	   infants	   growing	   up	   in	   societies	   consisting	   predominantly	   of	  nuclear	   families	   are	   typically	   raised	   by	   one	   or	   2	   parents,	   along	   with	   any	   older	  siblings.	  Whether	  an	  infant	  is	  raised	  by	  a	  community,	  or	  in	  a	  nuclear	  family	  context	  has	   been	   shown	   to	   result	   in	   differences	   of	   attachment	   quality:	   infants	   who	   have	  many	  alloparents	  experience	  attachment	  with	  several	  individuals,	  but	  less	  intensely	  with	  each	  individual,	  than	  do	  infants	  growing	  up	  with	  a	  single	  caregiver.	  Attachment	  	   26	  relationships	   have	   implications	   for	   learning:	   infants	   raised	   by	   a	   community	   of	  caregivers	  are	  exposed	  to	  more	  potential	  cultural	  models	   than	  are	   those	  raised	  by	  parents	  in	  nuclear	  families,	  and	  also	  have	  more	  opportunities	  to	  exercise	  selectivity	  in	  model	  choice.	  	  Infants	   develop	   in	   contexts	   structured	   by	   primary	   socialization	   agents	  (usually	  parents),	  who	  form	  the	  earliest	  sources	  of	  cultural	  transmission.	  Parenting	  behaviours	  are	  typically	  informed	  by	  valued	  developmental	  outcomes,	  and	  aimed	  to	  prepare	   children	   for	   the	   local	   cultural	   environment	   (Keller,	   2007).	   Differential	  emphasis	   on	   aspects	   of	   the	   environment	   has	   been	   documented	   in	   several	   cross-­‐cultural	   studies	   comparing	   parent	   and	   parent	   interactions.	   For	   instance,	   North	  American	   parents	   view	   autonomy	   and	   independence	   as	   valued	   outcomes	   of	  development,	  whereas	  relatively	  more	  interdependent	  Japanese	  parents	  value	  social	  relatedness	   as	   an	   ideal	   outcome.	   Consistent	  with	   these	   values,	   American	  mothers	  have	   been	   found	   to	   be	   more	   responsive	   when	   their	   children	   orient	   to	   physical	  objects	  in	  the	  environment,	  while	  Japanese	  mothers	  are	  more	  responsive	  when	  their	  infants	  oriented	   to	   them	  (Fernald	  &	  Morikawa,	  1993;	  Fogel,	  Toda,	  &	  Kawai,	  1988;	  Bornstein	   et	   al,	   1992;	   Tamis-­‐LeMonda,	   Bornstein,	   Cyphers,	   Toda,	   &	  Ogino,	   1992).	  Implicit	   in	   these	   interactions	  are	  differential	  valuation	  of	   learning	  about	  objects	   in	  the	   environment,	   and	   social	   routines	   in	   these	   two	   cultures.	   Through	   these	  reinforcement	  events,	  infants	  in	  North	  American	  cultures	  may	  learn	  that	  accurately	  predicting	   the	  behaviour	  and	   functions	  of	  objects	   is	   important,	  whereas	   infants	   in	  East	   Asian	   cultures	   may	   learn	   that	   accurately	   interpreting	   the	   social	   cues	   and	  empathizing	  with	  one’s	  interaction	  partners	  is	  more	  important.	  	  	   27	  Indeed,	   even	   the	   extent	   to	   which	   members	   of	   a	   culture	   engage	   in	   active	  teaching	  is	  not	  universal.	   In	  Western	  cultures,	  adults	  often	  heavily	  scaffold	  infants’	  exploration	   and	   learning	   about	   objects	   (Bornstein,	   2002;	   Keller	   et	   al.,	   2009).	  However,	   in	   many	   non-­‐Western	   societies,	   learning	   experiences	   are	   often	   more	  collaborative	   (Rogoff	   et	   al.,	   1993),	   and	   may	   occur	   in	   the	   absence	   of	   adult	  participation	   altogether	   (Bakeman,	   Adamson,	   Konner,	   &	   Barr,	   1990).	   In	   lieu	   of	  active	  pedagogy	  by	  adults,	  early	  cultural	  learning	  in	  many	  societies	  takes	  the	  form	  of	  observational	  learning	  (Heyes,	  2012;	  Lancy,	  Bock,	  &	  Gaskins,	  2010;	  Odden	  &	  Rochat,	  2004).	   For	   example,	   Mayan	   children	   tend	   to	   learn	   through	   overhearing	  conversation,	   whereas	   American	   children	   more	   often	   participate	   directly	   in	  language	  exchanges	  (Shneidman	  &	  Goldin-­‐Meadow,	  2012).	  Caregivers	   also	   employ	   culturally	   variable	   methods	   for	   communicating	  information,	   particularly	   for	   fostering	   their	   children’s	   learning	   (Bornstein,	   Cote,	  Haynes,	   Suwalsky,	   &	   Bakeman,	   2012;	   Herrmann,	   Legare,	   Harris,	   &	   Whitehouse,	  2013;	  Legare	  &	  Nielsen,	  in	  press;	  Mathew	  &	  Perreault,	  2015).	  One	  example	  of	  such	  cross-­‐cultural	   difference	   is	   the	   extent	   to	   which	   parents	   rely	   on	   at-­‐a-­‐distance	  communication	  using	   facial	   and	   vocal	   utterances	   (in	   visual	   joint	   attention)	   versus	  physical	   manipulations	   of	   their	   child’s	   bodies	   (physical	   joint	   attention)	   to	   teach	  about	  objects	   (Little,	  Carver,	  &	  Legare,	  2015).	  The	   former	  pattern	  of	   interaction	   is	  predominantly	   characteristic	   of	  Western	   urban	   settings	   (Duranti,	   2009;	   Richman,	  Miller,	  &	  LeVine,	  1992),	  whereas	  the	  latter	  is	  more	  common	  in	  traditional	  societies.	  These	   differences	   may	   be	   a	   manifestation	   of	   the	   broader	   socialization	   values	  described	  earlier:	  extensive	  bodily	  contact	  and	  direct	  physical	  manipulation	  tend	  to	  	   28	  foster	   reliance	   on	   the	   caregiver	   and	   facilitate	   bonds	   of	   closeness,	   while	   physical	  separation	  and	  at-­‐a-­‐distance	  communication	  fosters	  independence	  and	  autonomous	  judgment	  and	  decision-­‐making	  (Keller,	  2007).	  	  We	   might	   expect	   early	   differences	   in	   social	   and	   emotional	   experiences	   to	  have	   cascading	   downstream	   effects	   on	   social	   cognition.	   Early	   pathways	   may	  influence	   developmental	   outcomes	   by	   canalizing	   and	   constraining	   subsequent	  attention,	   preferences,	   and	   learning.	   For	   example,	   preference	   for	   one’s	   native	  language,	  though	  arrived	  at	  similarly	  for	  infants	  across	  cultures	  (through	  exposure	  and	   familiarity),	   channels	   their	   subsequent	   attention	   to	   native	   language	   speakers	  (Kinzler	   et	   al.,	   2007;	   Moon,	   Cooper,	   Fifer,	   1993).	   Attentional	   biases	   earlier	   in	  development	  may	  influence	  which	  groups	  of	  people	  infants	  are	  more	  likely	  to	  accept	  a	  toy	  from	  by	  10	  months	  (Kinzler,	  2007),	   learn	  from	  by	  14	  months	  (Buttelmann	  et	  al.,	   2013),	   and	   friendship	   choices	   by	   5	   years	   of	   age	   (Kinzler	   et	   al.,	   2007).	   Thus,	  differences	   in	   early	   cultural	   learning	   experiences	   may	   create	   diverging	   outcomes	  with	  lasting	  impact	  on	  development.	  This	  question	  of	  cultural	  differences	  in	  cultural	  learning	  will	  be	  explored	  in	  Chapter	  5.	  	  1.7	   Thesis	  rationale	  and	  outline	  of	  subsequent	  chapters	  	   The	   chapters	   within	   this	   dissertation	   examine	   how	   the	   youngest	   humans	  engage	   in	   cultural	   learning	   about	   conventional	   behaviours.	   In	   particular,	   they	   ask	  whether	  preverbal	  infants	  and	  preschool	  aged	  children	  have	  the	  means	  available	  to	  evaluate	  others	  using	  cues	  such	  as	  who	  performs	  common	  actions,	  how	  others	  react	  emotionally	  to	  actions	  and	  individuals,	  and	  via	  culturally	  specific	  pedagogical	  social	  interactions.	   	   In	   Chapter	   2,	   infants’	   preference	   for	   individuals	   who	   perform	   a	  	   29	  consensus	  action	  vs.	  an	  oft-­‐repeated	  action	  is	  examined,	  and	  evidence	  suggests	  that	  sensitivity	   to	   the	   distribution	   of	   actions	   emerges	   early.	   Chapter	   3	   extends	   on	  Chapter	   2	   by	   investigating	   the	   developmental	   trajectory	   of	   using	   consensus	   to	  inform	  subsequent	  learning	  of	  new	  information	  in	  pre-­‐school	  aged	  children.	  Chapter	   4	   examines	   the	   developmental	   trajectory	   of	   preverbal	   infants’	  observational	  learning	  about	  people	  and	  objects	  via	  bystander	  reactions.	  It	  presents	  evidence	  that	  not	  only	  are	  interpretations	  of	  emotional	  cues	  constrained	  by	  domain	  (specifically,	  infants	  use	  bystander’s	  emotional	  reactions	  differently	  when	  they	  refer	  to	   social	   vs.	   inanimate	   targets),	   but	   also	   that	   cultural	   differences	   in	   the	   use	   of	  emotional	  cues	  for	  social	  learning	  are	  present	  by	  6	  months	  of	  age.	  	  Chapter	   5	   investigates	   how	   parents	   convey	   evaluative	   messages	   about	  targets	   during	   interactions	   with	   infants,	   and	   explore	   cultural	   differences	   in	   these	  pedagogical	   interactions.	   In	  particular,	   the	  study	   focuses	  on	  cultural	  differences	   in	  reliance	  on	  at-­‐a-­‐distance	  vocal	  communications	  and	  direct	  physical	  manipulation.	  	  	  	  	  	  	  	  	  	  	   30	  2.	  Preverbal	  infants’	  context	  dependent	  preferences	  for	  conformists	  vs.	  innovators	  	  2.1	  	   Introduction	  The	  capacity	  to	  evaluate	  potential	  social	  partners—to	  tell	  friend	  from	  foe,	  and	  expert	  from	  novice—is	  foundational	  to	  human	  social	  life.	  Friends	  confer	  numerous	  benefits,	  from	  sharing	  knowledge	  and	  resources	  to	  cooperatively	  completing	  large	  tasks	  that	  are	  impossible	  to	  tackle	  individually.	  Social	  others	  are	  also	  an	  efficient	  source	  of	  knowledge,	  as	  learning	  from	  others	  eliminates	  much	  of	  the	  costs	  of	  trial-­‐and-­‐error	  learning,	  and	  learning	  social	  conventions	  makes	  it	  possible	  to	  coordinate	  group	  activities.	  Humans	  evaluate	  others	  intuitively	  and	  effortlessly,	  along	  dimensions	  such	  as	  skill,	  ethnicity,	  gender,	  age,	  and	  appearance	  (Quinn	  et	  al.,	  2002;	  Rakoczy	  et	  al.,	  2010;	  Rubenstein	  et	  al.,	  1999).	  Just	  what	  qualities	  make	  a	  good	  social	  partner	  at	  any	  given	  moment,	  however,	  may	  vary	  depending	  on	  one’s	  interactional	  goals	  and	  situational	  needs.	  What	  are	  the	  ontogenetic	  origins	  of	  context	  dependent,	  flexible	  social	  evaluations?	  Studies	  with	  children	  have	  demonstrated	  flexibility	  in	  learning	  early	  on	  in	  development,	  being	  sensitive	  to	  contextual	  factors	  and	  situationally	  salient	  motivations.	  In	  choosing	  how	  to	  learn,	  children	  use	  both	  imitation—learning	  faithfully	  from	  others,	  and	  innovation—recombining	  extant	  behaviours	  in	  one’s	  repertoire	  into	  novel	  forms.	  Children	  may	  favour	  imitation	  when	  learning	  social	  conventions	  that	  are	  particular	  to	  their	  social	  group	  (e.g.	  dancing	  the	  macarena),	  and	  which	  requires	  a	  high	  degree	  of	  fidelity	  to	  reproduce	  (Kenward,	  2012;	  Legare	  &	  Nielsen,	  in	  press).	  In	  contrast,	  children	  may	  favour	  innovation	  when	  learning	  	   31	  instrumental	  behaviours	  targeted	  at	  effectively	  reaching	  a	  functional	  outcome	  (e.g.	  making	  the	  biggest	  sandcastle;	  (Legare,	  Wen,	  Herrmann,	  &	  Whitehouse,	  2015).	  	  Furthermore,	  the	  fidelity	  with	  which	  children	  imitate	  may	  differ	  with	  context.	  Children	  are	  at	  times	  rational	  imitators,	  effective	  at	  extracting	  causally	  relevant	  aspects	  of	  behaviours	  for	  achieving	  a	  particular	  goal	  (Carpenter,	  Akhtar,	  &	  Tomasello,	  1998;	  Gergely	  et	  al.,	  2002;	  Meltzoff,	  1995);	  at	  other	  times,	  children	  are	  over-­‐imitators,	  copying	  actions	  that	  are	  causally	  irrelevant	  to	  their	  desired	  end	  goal	  (Gergely	  &	  Csibra,	  2006;	  Horner	  &	  Whiten,	  2005;	  Lyons,	  Young,	  &	  Keil,	  2007).	  Children’s	  flexibility	  in	  switching	  between	  rational	  and	  over-­‐imitation	  appear	  to	  be	  moderated	  by	  contextually	  salient	  goals:	  efficient	  learning	  motivation	  leads	  to	  rational	  learning,	  whereas	  affiliation	  with	  social	  others	  is	  more	  likely	  to	  lead	  to	  over-­‐imitation	  (Nielsen,	  2006;	  Over	  &	  Carpenter,	  2012).	  	  While	  there	  does	  not	  appear	  to	  be	  similarly	  strong	  evidence	  of	  context-­‐dependent	  learning	  in	  the	  first	  year	  of	  life,	  results	  from	  a	  number	  of	  studies	  suggest	  that	  even	  preverbal	  infants	  may	  make	  flexible,	  context-­‐dependent	  evaluations	  of	  others’	  actions	  for	  the	  purpose	  of	  identifying	  good	  social	  partners.	  There	  appear	  to	  be	  2	  primary	  avenues	  for	  doing	  so:	  evaluating	  behaviours	  and	  assessing	  group	  membership.	  One	  aspect	  of	  behaviour	  that	  has	  important	  implications	  for	  social	  interaction	  is	  the	  degree	  to	  which	  a	  person	  is	  cooperative	  and	  trustworthy.	  Cooperative	  acts	  include	  working	  towards	  a	  common	  goal	  or	  resource	  (e.g.	  paying	  taxes,	  or	  not	  littering	  to	  keep	  the	  streets	  clean),	  not	  exploiting	  others	  when	  opportunities	  arise	  (e.g.	  stealing	  unattended	  laptops	  in	  a	  coffee	  shop),	  teaching	  useful	  skills	  and	  knowledge	  (e.g.	  helping	  a	  child	  to	  ride	  a	  bike),	  and	  not	  	   32	  deceiving/misleading	  others.	  From	  this	  perspective,	  good	  social	  partners	  are	  individuals	  who	  help	  rather	  than	  harm	  others	  (i.e.	  are	  prosocial).	  	  A	  second	  avenue	  for	  evaluating	  social	  partners	  is	  by	  a	  person’s	  group	  membership.	  Behaviours	  do	  not	  take	  place	  in	  a	  vacuum,	  but	  instead,	  in	  the	  sociocultural	  context	  of	  a	  social	  group.	  Successful	  cooperation	  gave	  rise	  to	  humans	  living	  in	  cultural	  groups	  that	  share	  a	  language,	  customs,	  beliefs	  and	  attitudes	  (Henrich	  &	  Henrich,	  2007).	  These	  shared	  behaviours	  facilitate	  social	  interaction	  through	  shared	  understanding,	  reducing	  the	  burden	  of	  having	  to	  negotiate	  the	  terms	  of	  engagement	  in	  each	  new	  interaction.	  Group	  membership	  is	  thus	  useful	  for	  identifying	  good	  social	  partners	  as	  individuals	  whose	  attitudes	  and	  behaviours	  we	  can	  understand	  and	  predict,	  and	  thereby	  interact	  with	  successfully.	  	  Recent	  research	  on	  the	  development	  of	  social	  evaluation	  has	  found	  that	  humans	  do	  evaluate	  potential	  social	  partners	  on	  the	  basis	  of	  their	  behaviours	  from	  an	  early	  age.	  As	  early	  as	  five	  months,	  human	  infants	  prefer	  those	  who	  help	  over	  those	  who	  hinder	  others’	  from	  achieving	  their	  goals	  (Hamlin	  &	  Wynn,	  2011).	  Consistent	  support	  for	  such	  prosocial	  preferences	  has	  been	  found	  with	  3-­‐month-­‐old	  infants,	  who	  look	  longer	  (interpreted	  as	  preference)	  at	  individuals	  who	  help	  others	  achieve	  their	  goals	  than	  individuals	  who	  hinder	  others	  from	  achieving	  their	  goals	  (Hamlin,	  Wynn,	  &	  Bloom,	  2010).	  	  However,	  these	  preferences	  are	  not	  absolute,	  as	  young	  infants’	  evaluations	  are	  sensitive	  to	  the	  contexts	  in	  which	  behaviours	  take	  place.	  When	  the	  past	  history	  of	  the	  target	  is	  known,	  5-­‐month-­‐old	  infants	  take	  the	  overall	  moral	  balance	  into	  account	  in	  their	  evaluations,	  preferring	  an	  agent	  who	  helps	  a	  previously	  helpful	  	   33	  target	  and	  an	  agent	  who	  hinders	  a	  previously	  anti-­‐social	  target.	  Nine-­‐	  and	  14-­‐month-­‐old	  infants	  further	  take	  into	  account	  the	  target’s	  similarities	  to	  oneself	  to	  help	  them	  evaluate	  their	  behaviours	  (Hamlin	  et	  al.,	  2013).	  Similarity	  to	  self	  may	  predict	  similarity	  along	  other	  dimensions	  that	  pave	  the	  way	  to	  friendship	  (people	  who	  listen	  to	  the	  same	  music	  are	  also	  likely	  to	  enjoy	  the	  same	  food).	  Underlying	  such	  similarities	  may	  be	  common	  cultural	  group	  membership	  from	  which	  we	  develop	  our	  palate	  and	  acquire	  our	  music	  preferences.	  	  Evidence	  also	  exists	  for	  young	  infants’	  evaluations	  on	  the	  basis	  of	  group	  membership.	  In	  the	  first	  few	  months	  of	  life,	  infants	  appear	  capable	  of	  discriminating	  between	  individuals	  based	  on	  dialect	  (Kinzler,	  Corriveau,	  &	  Harris,	  2007),	  ethnic	  appearance	  (Bar-­‐Haim	  et	  al.,	  2006)	  and	  similarity	  of	  food	  preferences	  (Mahajan	  &	  Wynn,	  2012).	  These	  preferences	  may	  help	  learners	  identify	  individuals	  who	  likely	  belong	  to	  the	  same	  cultural	  group,	  and	  who	  are	  therefore	  “good	  bets”	  to	  learn	  from;	  as	  infants	  who	  prefer	  speakers	  of	  their	  native	  dialect	  are	  also	  likely	  to	  correctly	  identify	  in-­‐group	  members	  who	  possess	  knowledge	  of	  their	  culture.	  Furthermore,	  since	  those	  who	  live	  in	  physical	  proximity	  are	  also	  likely	  to	  possess	  locally	  relevant	  cultural	  knowledge,	  a	  preference	  for	  individuals	  with	  familiar	  appearances	  should	  tend	  to	  increase	  the	  likelihood	  that	  infants	  will	  easily	  identify	  and	  prefer	  to	  attend	  to	  members	  of	  their	  own	  culture.	  Finally,	  humans	  learn	  what	  is	  good	  to	  eat	  through	  observational	  learning,	  promoting	  a	  strong	  group-­‐based	  food	  preference	  (Addessi	  et	  al.,	  2005;	  Baeyens	  et	  al.,	  1996;	  Shutts	  et	  al.,	  2009).	  Therefore,	  liking	  others	  who	  share	  similar	  food	  preferences	  can	  also	  help	  infants	  identify	  in-­‐group	  members.	  	  	   34	  As	  with	  behaviours,	  evaluations	  of	  group	  membership	  are	  also	  context	  dependent.	  Children	  and	  adults	  flexibly	  assign	  temporary	  group	  membership	  to	  others	  in	  our	  language	  community	  when	  travelling	  abroad,	  yet	  at	  a	  moment’s	  notice,	  will	  reform	  group	  boundaries	  based	  on	  sex	  when	  locating	  the	  correct	  washroom.	  Even	  19-­‐month-­‐old	  infants	  show	  such	  flexibility	  in	  making	  assessments	  of	  group	  membership	  in	  3rd	  party	  contexts	  (Sloane,	  Baillargeon,	  &	  Premack,	  in	  prep).	  They	  expect	  individuals	  with	  dissimilar	  appearances	  to	  receive	  different	  treatment	  based	  on	  group	  membership,	  yet	  can	  flexibly	  reimagine	  these	  dissimilar-­‐looking	  individuals	  to	  be	  part	  of	  the	  same	  group,	  deserving	  of	  similar	  treatment,	  following	  a	  brief	  episode	  of	  coordinated	  play.	  This	  latter	  finding	  suggests	  that	  infants	  can	  make	  complex	  analyses	  of	  identifying	  members	  of	  a	  community	  through	  their	  behaviours,	  and	  further,	  evaluate	  the	  appropriateness	  of	  a	  3rd	  party’s	  behaviours	  towards	  those	  individuals	  based	  on	  their	  group	  membership.	  	  The	  ability	  to	  evaluate	  goal-­‐oriented	  social	  behaviours	  and	  facets	  of	  group	  membership	  can	  serve	  a	  cultural	  novice	  quite	  well	  in	  identifying	  good	  social	  partners.	  However,	  a	  vast	  number	  of	  human	  social	  behaviour	  are	  not	  based	  on	  physical	  causality,	  nor	  are	  clearly	  marked	  in	  terms	  of	  group	  membership.	  Instead,	  they	  are	  governed	  by	  social	  conventions—shared	  rules,	  or	  a	  group’s	  characteristic	  ways	  of	  doing	  things	  that	  are	  arbitrary	  in	  nature.	  They	  include	  rules	  such	  as	  when	  and	  where	  one	  must	  wear	  clothing,	  whether	  it	  is	  acceptable	  to	  litter,	  and	  what	  utensils	  (or	  which	  hand)	  should	  be	  used	  for	  eating	  (Bickman,	  1972;	  Elster,	  1989;	  (Searle,	  2010;	  Tomasello	  &	  Rakoczy,	  2003).	  	   35	  One	  cue	  that	  could	  hint	  at	  conventionality	  for	  cultural	  novices	  is	  that	  the	  prevalence	  of	  behaviours	  tend	  to	  correlate	  positively	  with	  how	  accepted	  they	  are	  (Cialdini,	  Reno,	  &	  Kallgren,	  1990).	  For	  instance,	  the	  socially	  accepted	  manner	  of	  eating	  a	  sandwich	  is	  with	  one’s	  hands,	  which	  corresponds	  with	  the	  observation	  that	  most	  people	  eating	  sandwiches	  can	  be	  seen	  eating	  them	  with	  their	  hands.	  In	  contrast,	  ravioli	  is	  eaten	  with	  cutlery,	  and	  the	  majority	  of	  ravioli	  eaters	  do	  so,	  with	  few	  people	  eating	  it	  with	  their	  hands.	  For	  a	  naïve	  individual	  buying	  a	  sandwich	  at	  a	  deli,	  a	  quick	  look	  to	  other	  deli	  patrons	  may	  indicate	  to	  him	  or	  her	  that	  the	  accepted	  and	  safe	  thing	  to	  do	  is	  to	  eat	  a	  sandwich	  with	  his	  or	  her	  hands,	  rather	  than	  with	  fork	  and	  knife	  or	  chopsticks.	  Thus,	  when	  it	  comes	  to	  conventional	  behaviours,	  a	  mechanism	  for	  identifying	  a	  “good”	  social	  partner	  may	  involve	  assessing	  the	  degree	  to	  which	  someone’s	  actions	  are	  prevalent	  in	  the	  environment.	  	  A	  recent	  study	  by	  Powell	  and	  Spelke	  (2013)	  suggest	  that	  there	  might	  be	  innate	  expectations	  for	  conventionality	  in	  infancy.	  The	  study	  showed	  that	  8	  to	  12	  month	  old	  infants	  expected	  members	  of	  social	  groups	  to	  behave	  as	  fellow	  group	  members	  do.	  This	  expectation	  existed	  only	  for	  animate,	  social	  agents	  (self-­‐propelled	  shapes	  with	  eyes),	  and	  not	  for	  inanimate	  objects	  (similar	  shapes	  without	  eyes,	  and	  propelled	  by	  an	  external	  force).	  Infants	  made	  attributions	  of	  same-­‐group	  membership	  for	  individuals	  who	  engaged	  in	  similar	  actions,	  even	  if	  their	  appearances	  differed.	  This	  study	  supports	  the	  notion	  that	  infants	  may	  possess	  inherent	  expectations	  that	  group	  members	  produce	  similar	  behaviours.	  That	  is,	  they	  may	  have	  some	  notion	  that	  behaviours	  are	  bound	  by	  conventions.	  	  	   36	  Given	  infants’	  apparent	  sensitivity	  to	  conventional	  behaviours,	  can	  they	  use	  this	  to	  evaluate	  social	  others	  who	  appear	  to	  conform	  to	  them?	  Positively	  evaluating	  and	  associating	  with	  individuals	  who	  perform	  conventional	  acts	  may	  serve	  a	  naïve	  learner	  well,	  particularly	  in	  situations	  where	  individuals’	  group	  identity	  is	  uncertain.	  However,	  when	  one	  is	  in	  familiar	  environs,	  associating	  with	  individuals	  who	  may	  be	  more	  knowledgeable	  may	  be	  preferable.	  More-­‐knowledgeable	  individuals	  may	  deviate	  from	  conventional	  behaviours	  and	  appearances	  because	  they	  have	  access	  to	  techniques	  and	  skills	  that	  perform	  better	  than	  average.	  For	  example,	  while	  the	  majority	  of	  children	  can	  identify	  a	  green	  leafy	  organism	  as	  a	  “plant”,	  some	  children	  might	  have	  the	  additional	  ability	  to	  identify	  the	  plant	  as	  “stinging	  nettle”	  and	  know	  of	  its	  poisonous	  properties.	  Thus,	  while	  labeling	  the	  organism	  as	  a	  “plant”	  in	  English	  rather	  than	  in	  Spanish	  is	  a	  good	  indicator	  of	  group	  membership	  in	  an	  English	  speaking	  culture,	  labeling	  that	  plant	  as	  "stinging	  nettle"	  puts	  one	  in	  a	  knowledgeable	  minority.	  Associating	  with	  the	  latter	  group	  of	  more-­‐knowledgeable	  individuals	  would	  be	  beneficial	  to	  a	  naïve	  learner.	  	  Are	  infants	  sensitive	  to	  the	  distribution	  of	  frequent	  behaviours	  (widely	  distributed	  frequent	  behaviours	  being	  a	  sign	  of	  conventionality)?	  And	  can	  they	  use	  such	  distributional	  information	  to	  inform	  their	  evaluations	  of	  individuals	  who	  perform	  the	  actions,	  and	  do	  so	  in	  a	  context-­‐flexible	  manner?	  Infants	  have	  already	  demonstrated	  sensitivity	  to	  the	  frequency	  of	  exposure	  in	  learning	  contexts	  in	  past	  studies.	  Preverbal	  infants	  prefer	  people,	  faces,	  gender,	  flavours	  and	  languages	  to	  which	  they	  have	  received	  more	  exposure	  and	  are	  therefore	  more	  familiar	  (Kelly	  et	  al.,	  2005;	  Kinzler	  et	  al.,	  2007;	  2010;	  Langlois	  &	  Roggman,	  1990;	  Mennella	  et	  al.,	  	   37	  2001;	  Ramsey,	  Langlois,	  &	  Marti,	  2005).	  However,	  not	  all	  frequent	  behaviours	  are	  equivalent.	  Whereas	  an	  individual	  repeatedly	  performing	  an	  action	  can	  provide	  a	  predictable	  signal	  that	  facilitates	  learning,	  or	  suggest	  that	  a	  particular	  individual	  has	  a	  preference	  for	  performing	  the	  behaviour,	  it	  is	  only	  when	  several	  individuals	  perform	  the	  same	  behaviour	  that	  it	  is	  possible	  to	  assume	  that	  the	  behaviour	  is	  conventional	  amongst	  a	  group.	  	  In	  the	  set	  of	  studies	  presented	  here,	  we	  examined	  whether	  7-­‐month-­‐old	  infants	  use	  actors’	  performance	  of	  prevalent	  behaviours	  to	  inform	  their	  social	  evaluations.	  We	  investigated	  whether	  infants’	  evaluations	  differ	  according	  to	  contextual	  information	  provided;	  specifically,	  the	  distribution	  of	  frequent	  behaviours,	  and	  whether	  the	  targets	  of	  evaluation	  demonstrated	  prior	  knowledge	  of	  the	  frequent	  behaviours.	  Study	  1	  presented	  infants	  with	  videos	  in	  which	  several	  animated	  shapes	  with	  faces	  (Demonstrators)	  performed	  physical	  movements	  that	  resemble	  dances.	  Demonstrators	  exited	  after	  their	  dance,	  followed	  by	  the	  appearance	  of	  2	  Protagonists;	  one	  performed	  the	  same	  movement	  sequence	  as	  the	  Demonstrators,	  and	  the	  other	  completed	  a	  novel	  sequence	  of	  movements	  (from	  here	  on	  in,	  referred	  to	  as	  a	  “dance”).	  	  Study	  1	  created	  a	  scenario	  where	  the	  Protagonist	  lacked	  opportunity	  to	  learn	  the	  Demonstrators’	  dance.	  Opportunity	  to	  learn	  the	  dance	  was	  manipulated	  by	  Protagonists’	  physical	  presence	  vs.	  absence	  onscreen	  during	  the	  dance	  demonstration.	  Protagonists’	  absence	  during	  the	  demonstration	  was	  intended	  to	  convey	  their	  lack	  of	  visual	  access	  to	  the	  performance,	  and	  thus	  lack	  of	  opportunity	  to	  learn	  the	  physical	  dance	  sequence.	  Previous	  studies	  have	  found	  infants	  to	  be	  	   38	  sensitive	  to	  visual	  access,	  taking	  into	  account	  barriers	  and	  blindfolds	  that	  can	  obstruct	  one’s	  view	  (Brooks	  &	  Meltzoff,	  2002;	  Luo	  &	  Baillargeon,	  2007),	  and	  thus,	  access	  to	  situationally	  relevant	  information.	  For	  example,	  16-­‐month-­‐olds	  are	  more	  surprised	  by	  agents	  who	  incorrectly	  label	  a	  familiar	  object	  when	  they	  are	  facing	  the	  object,	  than	  when	  they	  are	  facing	  away	  from	  the	  object	  (Koenig	  &	  Echols,	  2003).	  Thus,	  Protagonists’	  absence	  is	  intended	  to	  convey	  their	  lack	  of	  visual	  access,	  and	  therefore	  opportunity	  to	  learn	  the	  Demonstrators’	  dance,	  and	  performance	  of	  the	  same	  dance	  is	  due	  to	  the	  Protagonist’s	  prior	  knowledge.	  	  Half	  of	  study	  participants	  saw	  the	  Consensus	  condition	  stimuli	  described	  above;	  the	  other	  half	  saw	  the	  Repetition	  condition	  stimuli	  in	  which	  the	  dance	  is	  repeatedly	  performed	  by	  a	  single	  Demonstrator,	  rather	  than	  by	  all	  4	  Demonstrators.	  This	  was	  designed	  to	  contrast	  a	  scenario	  in	  which	  the	  Demonstrators’	  dance	  was	  a	  shared,	  conventional	  dance.	  In	  contrast,	  the	  high	  frequency	  dance	  did	  not	  have	  the	  characteristics	  of	  appearing	  conventional	  in	  the	  Repetition	  condition.	  If	  infants	  form	  preferences	  based	  on	  a	  group-­‐level	  analysis,	  they	  should	  only	  show	  a	  preference	  in	  the	  Consensus,	  and	  not	  the	  Repetition	  condition.	  However,	  if	  infants	  perform	  a	  frequency-­‐based	  analysis,	  then	  they	  should	  show	  a	  preference	  in	  both	  Consensus	  and	  Repetition	  conditions	  and	  disregard	  the	  distribution	  of	  behaviours	  across	  Demonstrators.	  	  Study	  2	  used	  the	  same	  video	  stimuli,	  save	  for	  1	  difference:	  in	  this	  study,	  the	  Protagonists	  were	  present	  during	  the	  Demonstrators’	  dance.	  This	  alteration	  created	  a	  scenario	  in	  which	  Protagonists	  could	  imitate	  the	  Demonstrators	  on	  the	  spot,	  and	  therefore	  provided	  no	  information	  regarding	  Protagonists’	  prior	  knowledge	  of	  the	  	   39	  group	  dance.	  In	  this	  altered	  context,	  the	  Protagonist	  who	  performed	  the	  more	  common	  behaviour	  would	  merely	  be	  a	  copycat,	  rather	  than	  someone	  who	  was	  “in	  the	  know”.	  	  Given	  everyone’s	  equal	  standing	  in	  access	  to	  the	  group	  dance,	  the	  question	  prompted	  in	  Study	  2	  may	  be	  “who	  knows	  more?”	  As	  with	  Study	  1,	  this	  question	  requires	  a	  group-­‐level	  analysis,	  since	  “knowing	  more”	  requires	  an	  assessment	  of	  the	  average	  knowledge	  level	  of	  the	  group.	  A	  novel	  dance	  may	  be	  perceived	  as	  a	  new	  piece	  of	  dance	  if	  everyone	  else	  performed	  the	  same	  dance	  (the	  average	  knowledge	  level	  is	  the	  one	  piece	  of	  dance);	  however,	  a	  novel	  dance	  performed	  in	  the	  context	  of	  an	  oft-­‐repeated	  dance	  by	  a	  single	  Demonstrator	  does	  not	  provide	  any	  information	  about	  how	  much	  more	  or	  less	  the	  dancer	  knows	  relative	  to	  everyone	  else	  in	  the	  group.	  When	  there	  is	  information	  about	  average	  knowledge	  in	  a	  group,	  infants	  who	  prefer	  more	  knowledgeable	  individuals	  may	  show	  a	  preference	  for	  the	  maverick,	  or	  the	  novel-­‐dance	  Protagonist.	  Infants’	  preferences	  in	  both	  studies	  were	  measured	  by	  their	  choice	  between	  same-­‐dance	  and	  novel-­‐dance	  Protagonists.	  	  2.2	  	   Study	  1	  The	  present	  study	  examined	  whether	  infants	  prefer	  those	  who	  have	  knowledge	  of	  common	  actions.	  	  2.2.1	  	   Method	  	   Sixty-­‐four	   full	   term	   7-­‐month-­‐old	   infants	   participated	   in	   this	   study.	   Thirty	  girls	   (47%),	   34	   boys	   (M	   age	   =	   7.0;	   range	   =	   6.3	   –	   7.6)	   participated.	   Data	   from	   7	  additional	   infants	   were	   discarded	   because	   of	   fussiness	   (n	   =	   4),	   failure	   to	  make	   a	  choice	  (n	  =	  1),	  parental	  interference	  (n	  =	  1),	  and	  equipment	  failure	  (n	  =	  1).	  Infants	  	   40	  were	   from	  a	  range	  of	  ethnic	  backgrounds	  (51%	  European,	  18.5%	  East	  Asian,	  11%	  South	  and	  South	  East	  Asian,	  5%	  North	  American,	  including	  First	  Nations,	  3%	  Central	  and	  Latin	  American,	  1.5%	  African,	  and	  10%	  from	  other	  ethnicities,	  and	  15%	  were	  born	   to	   families	   from	   2	   or	   more	   cultures).	   Infants	   were	   recruited	   through	   the	  database	   maintained	   by	   the	   University	   of	   British	   Columbia	   Early	   Development	  Research	  Group,	  and	  were	  given	  a	  token	  gift	  for	  their	  participation.	  Infants	  watched	  these	  videos	  while	  seated	  in	  their	  parent’s	  lap	  in	  front	  of	  a	  TV	  screen.	  Parents	  were	  instructed	  not	  to	  communicate	  or	  to	  point	  during	  the	  presentation	  of	  the	  stimuli.	  	  Videos	  depicted	  animated	  characters	  engaging	  in	  neutral,	  arbitrary	  action	  sequences	  (“dances”),	  and	  which	  did	  not	  contain	  intrinsic	  indicators	  of	  quality	  or	  correctness.	  Two	  Demonstrator	  dances	  were	  used,	  a	  somersault	  dance,	  and	  a	  jumping	  dance,	  and	  2	  novel	  dances	  were	  used,	  a	  rolling	  dance,	  and	  a	  jittering	  dance.	  Dances	  were	  equated	  for	  amount	  of	  physical	  displacement	  and	  rate	  of	  movement,	  so	  as	  to	  appear	  similar	  in	  attractiveness.	  In	  one	  video,	  the	  Demonstrator(s)	  performed	  the	  somersault	  dance	  as	  the	  frequent	  dance;	  the	  same-­‐dance	  Protagonist	  also	  performed	  the	  somersault	  dance,	  and	  the	  novel-­‐dance	  Protagonist	  performed	  the	  rolling	  dance.	  In	  the	  other	  video,	  Demonstrator(s)	  performed	  the	  jumping	  dance	  as	  the	  frequent	  dance;	  here,	  the	  same-­‐dance	  Protagonist	  also	  performed	  the	  jumping	  dance,	  whereas	  the	  novel-­‐dance	  Protagonist	  performed	  the	  jittering	  dance.	  Each	  infant	  was	  familiarized	  to	  2	  repetitions	  of	  both	  videos	  (presentation	  order	  of	  videos	  was	  counterbalanced),	  for	  a	  total	  of	  4	  one-­‐minute	  videos.	  Thus,	  every	  infant	  saw	  Demonstrators	  perform	  both	  the	  somersault	  and	  jumping	  dances	  as	  frequent	  	   41	  dances;	  to	  evaluate	  the	  Protagonists,	  they	  had	  to	  extract	  the	  similarity	  of	  their	  dances	  to	  the	  Demonstrators’	  across	  both	  videos.	  	  Each	  video	  contained	  3	  phases:	  an	  entrance	  phase,	  a	  demonstration	  phase,	  and	  a	  performance	  phase.	  The	  entrance	  phase	  begins	  with	  4	  Demonstrators	  entering	  from	  the	  left	  side	  the	  screen.	  Demonstrators	  are	  simple	  geometric	  shapes	  with	  eyes,	  each	  bearing	  a	  unique	  colour.	  These	  	  “Demonstrators”	  proceeded	  to	  the	  right	  side	  of	  the	  screen,	  arrange	  themselves	  in	  a	  vertical	  column,	  and	  directed	  their	  eye	  gaze	  toward	  the	  centre	  of	  the	  screen.	  During	  the	  demonstration	  phase,	  each	  Demonstrator	  moved	  to	  the	  centre	  of	  the	  screen,	  performed	  the	  somersault	  (or	  jumping)	  dance,	  and	  then	  returned	  to	  its	  place	  in	  the	  right	  column.	  At	  the	  end	  of	  the	  demonstration	  phase,	  all	  of	  the	  Demonstrators	  exited	  the	  screen	  on	  the	  top	  right.	  	  To	  examine	  the	  effect	  of	  knowledge	  on	  social	  preference,	  the	  targets	  of	  evaluation	  (Protagonists)	  were	  absent	  during	  the	  Demonstration	  phase.	  This	  is	  to	  convey	  that	  the	  Protagonists	  do	  not	  have	  an	  opportunity	  to	  learn	  the	  somersault	  (or	  jumping)	  dance	  on	  the	  spot,	  and	  their	  subsequent	  performance	  of	  the	  same	  dance	  is	  a	  product	  of	  prior	  knowledge	  (presumably	  because	  they	  are	  part	  of	  the	  in-­‐group).	  Once	  the	  Demonstrators	  leave,	  two	  “Protagonists”	  enter	  together	  from	  the	  left	  side,	  arrange	  themselves	  into	  a	  vertical	  column	  on	  the	  left,	  and	  direct	  their	  gaze	  toward	  the	  centre	  of	  the	  screen.	  Finally,	  during	  the	  performance	  phase,	  the	  2	  Protagonists	  on	  the	  left	  side	  of	  the	  screen	  take	  turns	  performing	  dances.	  One	  Protagonist	  performs	  the	  somersault	  dance	  (same	  dance	  as	  the	  Demonstrators)	  suggesting	  it	  has	  prior	  knowledge	  of	  the	  Demonstrators’	  actions,	  and	  the	  other	  performs	  the	  rolling	  dance	  (novel	  dance)	  suggesting	  that	  it	  may	  not	  have	  prior	  knowledge	  of	  the	  	   42	  Demonstrators’	  actions.	  The	  order	  in	  which	  Protagonists	  performed	  their	  dances	  was	  counterbalanced	  across	  subjects.	  See	  Figure	  1	  for	  stills	  of	  the	  video	  sequences.	  	  In	   order	   to	   compare	   the	   effect	   of	   consensus	   frequency	   vs.	   repetitive	  frequency	   on	   infants’	   preferences,	   half	   of	   our	   participants	   were	   assigned	   to	   a	  Repetition	  condition.	  In	  contrast	  to	  the	  Consensus	  condition	  described	  above,	  only	  a	  single	  Demonstrator	  performed	  the	  same	  dance	  4	   times	  during	   the	  demonstration	  phase,	   while	   the	   rest	   remained	   stationary.	   The	   stimuli	   and	   procedures	   were	  otherwise	   identical	   across	   the	   two	   conditions.	   We	   predicted	   that	   infants	   would	  prefer	   the	   Protagonist	   who	   performed	   the	   same	   action	   as	   the	   Demonstrators,	  exhibiting	  prior	  knowledge	  of	  that	  group’s	  actions,	  in	  the	  Consensus,	  but	  not	  in	  the	  Repetition	  condition.	  	  	  Figure	   2.1	   Stimuli	   for	   Study	   1:	   only	   Demonstrators	   enter	   the	   screen	   (top	   left).	  Demonstrators	   perform	   same	   action	   sequence	   one	   after	   another	   (top	   right).	   After	  Demonstrators	  leave	  stage,	  Protagonists	  enter	  stage;	  one	  performs	  the	  same	  dance	  as	  Demonstrators,	  the	  other	  Protagonist	  performs	  a	  new	  dance.	  	  	  	   43	  Following	  presentation	  of	  the	  stimuli,	  parents	  were	  asked	  to	  close	  their	  eyes.	  A	  research	  assistant	  blind	  to	  the	  condition	  (Consensus	  or	  Repetition	  condition)	  and	  to	  the	  identity	  of	  the	  Protagonists	  (who	  performed	  the	  same	  vs.	  novel	  dance)	  presented	  the	  infant	  with	  a	  choice	  of	  the	  Protagonists,	  and	  asked,	  “Would	  you	  like	  to	  play	  with	  one?”	  A	  successful	  choice	  was	  noted	  when	  the	  infant	  looked	  at	  both	  choices,	  then	  reached	  for	  one	  of	  them	  via	  a	  visually	  guided	  reach.	  	  2.2.2	   Results	  Infants	  in	  the	  Consensus	  condition	  preferred	  the	  Protagonist	  who	  performed	  the	  same	  dance	  as	  the	  Demonstrators,	  suggestive	  of	  prior	  knowledge	  of	  the	  group’s	  actions	  (23	  vs.	  9;	  binomial	  probability	  test,	  p	  =	  .01).	  In	  contrast,	  infants	  in	  the	  Repetition	  condition	  chose	  the	  same-­‐dance	  and	  novel-­‐dance	  Protagonists	  equally;	  suggesting	  that	  simply	  performing	  a	  frequently	  observed	  dance	  is	  insufficient	  to	  generate	  a	  preference	  (17	  vs.	  15;	  binomial	  probability	  test,	  p	  =	  .43).	  The	  difference	  between	  these	  conditions	  was	  marginally	  significant	  in	  the	  predicted	  direction	  (one-­‐tailed	  Fisher’s	  Exact	  probability	  test	  =	  .098).	  There	  was	  no	  effect	  of	  order	  of	  events,	  type	  of	  dance,	  color	  or	  side	  of	  Protagonist	  on	  any	  comparison.	  	  2.2.3	   Discussion	  Seven-­‐month-­‐old	  infants	  preferred	  actors	  who	  performed	  an	  action	  that	  was	  previously	  performed	  by	  a	  4-­‐member	  consensus.	  However,	  they	  did	  not	  show	  such	  a	  preference	  for	  someone	  who	  performed	  an	  equally	  frequent,	  but	  non-­‐consensus	  behaviour.	  By	  holding	  the	  frequency	  of	  dances	  constant	  across	  conditions,	  we	  were	  able	  to	  rule	  out	  familiarity	  preference	  as	  the	  reason	  for	  infants’	  choice,	  since	  infants	  received	  equal	  exposure	  to	  the	  Demonstrators’	  dance	  in	  the	  Repetition	  condition	  as	  	  	   44	  Figure	  2.2	  Infants’	  Protagonist	  choices	  by	  condition	  in	  Study	  1.	  	  	  in	  the	  Consensus	  condition,	  without	  a	  commensurate	  increase	  in	  preference	  for	  the	  same-­‐dance	  Protagonist.	  This	  suggests	  that	  infants	  used	  the	  distribution	  of	  frequency	  information	  to	  evaluate	  actions,	  and	  those	  who	  perform	  them.	  Furthermore,	  infants’	  choices	  in	  the	  Consensus	  condition	  are	  consistent	  with	  our	  prediction	  that	  infants	  prefer	  individuals	  who	  know	  the	  group’s	  shared	  dance.	  	  A	  question	  that	  remains	  unaddressed	  by	  the	  present	  study	  is	  the	  possibility	  that	  rather	  than	  liking	  those	  “in	  the	  know”,	  infants	  prefer	  individuals	  who	  imitate	  others.	  For	  example,	  infants	  could	  have	  inferred	  the	  presence	  of	  the	  Protagonists	  off-­‐stage	  during	  the	  Demonstrators’	  performance,	  and	  subsequently	  mimicked	  their	  dances	  after	  their	  departure.	  Indeed,	  humans	  are	  quite	  sensitive	  to	  imitation,	  and	  respond	  well	  to	  being	  imitated.	  In	  experimental	  settings,	  people	  show	  greater	  liking,	  higher	  levels	  of	  trust,	  and	  more	  prosocial	  helping	  to	  those	  who	  imitate	  us	  (Bargh	  &	  Chartrand,	  1999;	  van	  Baaren,	  Holland,	  Steenaert,	  &	  van	  Knippenberg,	  2003).	  Could	  Study	  1	  results	  be	  explained	  by	  a	  consistent	  preference	  for	  imitators	  to	  mavericks?	  Consensus	   Repetition	  novel	  dance	   9	   15	  same	  dance	   23	   17	  0%	  25%	  50%	  75%	  100%	  Protagonist	  choice	  	   45	  There	  are	  reasons	  to	  believe	  that	  this	  is	  not	  necessarily	  the	  case.	  While	  those	  who	  behave	  conventionally	  signal	  that	  they	  are	  members	  of	  the	  in-­‐group,	  the	  average	  Joe	  is	  also	  limited	  as	  a	  source	  of	  new	  insight.	  If	  infants	  are	  motivated	  to	  identify	  good	  sources	  of	  cultural	  knowledge	  and	  can	  flexibly	  moderate	  their	  motivations	  and	  criteria	  for	  “good	  social	  partner”	  according	  to	  situational	  need,	  they	  might	  prefer	  to	  associate	  with	  extra-­‐knowledgeable	  individuals	  who	  deviate	  from	  the	  group	  because	  they	  have	  better-­‐than-­‐average	  information	  (as	  in	  the	  plant	  vs.	  poison	  ivy	  example	  given	  earlier).	  	  2.3	   Study	  2	  To	  determine	  whether	  infants’	  preferences	  are	  driven	  by	  identifying	  good	  sources	  of	  knowledge,	  or	  by	  identifying	  imitators,	  we	  modified	  Study	  1	  stimuli	  so	  that	  the	  Protagonists	  are	  present	  during	  the	  Demonstration	  phase,	  and	  thus	  the	  state	  of	  their	  prior	  knowledge	  of	  the	  Demonstrators’	  dance	  is	  explicitly	  ambiguous:	  Protagonists	  could	  have	  prior	  knowledge	  of	  the	  dance,	  but	  it	  is	  also	  possible	  that	  they	  learned	  it	  from	  observing	  the	  Demonstrators	  through	  imitation.	  If	  infants	  like	  those	  who	  follow	  conventions,	  or	  are	  willing	  imitators,	  we	  predict	  a	  similar	  pattern	  of	  results	  to	  Study	  1.	  If	  on	  the	  other	  hand,	  infants	  seek	  out	  extra-­‐knowledgeable	  individuals,	  we	  should	  see	  the	  opposite	  pattern,	  one	  where	  the	  maverick	  is	  preferred.	  2.3.1	   Methods	  A	  separate	  group	  of	  64	  full	  term	  7-­‐month-­‐old	  infants	  participated	  in	  this	  study.	  Twenty-­‐seven	  girls	  (42%),	  37	  boys	  (M	  age	  =	  7.07;	  range	  =	  6.53	  –	  7.57)	  participated.	  Infants	  were	  from	  a	  diverse	  range	  of	  ethnicities	  (52%	  European,	  23%	  	   46	  East	  Asian,	  9%	  South	  and	  Southeast	  Asian,	  5%	  North	  American,	  including	  First	  Nations,	  2%	  African,	  and	  3%	  from	  other	  ethnicities,	  and	  16%	  were	  born	  to	  families	  that	  identifies	  with	  2	  or	  more	  cultures).	  Data	  from	  7	  additional	  infants	  were	  discarded	  because	  of	  procedural	  error	  (n	  =	  2),	  fussiness	  (n	  =	  1),	  equipment	  failure	  (n	  =	  3),	  and	  parental	  interference	  (n	  =	  1).	  As	  in	  Study	  1,	  infants	  were	  recruited	  through	  the	  database	  maintained	  by	  the	  University	  of	  British	  Columbia	  Early	  Development	  Research	  Group,	  and	  were	  given	  a	  token	  gift	  for	  their	  participation.	  	  The	  procedures	  are	  identical	  to	  Studies	  1.	  The	  only	  difference	  is	  an	  alteration	  in	  the	  stimuli	   shown.	   In	   the	   video,	   Protagonists	   enter	   the	   screen	   immediately	   after	   the	  Demonstrators,	  and	  are	  present	  during	  the	  demonstration	  phase	  and	  observe	  their	  dances	  in	  real	  time.	  In	  this	  case,	  performing	  the	  same	  action	  as	  the	  Demonstrators	  does	  not	  require	  prior	  knowledge,	  as	  it	  could	  be	  learned	  during	  the	  demonstration.	  	  Figure	   2.3	   Stimuli	   for	   Study	   2,	   Consensus	   condition:	   Demonstrators	   and	  Protagonists	  enter	  the	  stage	  (top	  left).	  Demonstrators	  perform	  identical	  dances	  one	  after	   another	   (top	   right).	   After	   Demonstrators	   exit	   the	   screen,	   one	   Protagonist	  performs	  the	  same	  dance,	  the	  other	  Protagonist	  performs	  a	  novel	  dance.	  	  	  	   47	  2.3.2	   Results	  Unlike	  in	  Study	  1,	  infants	  in	  the	  Consensus	  condition	  of	  Study	  2	  significantly	  preferred	  the	  novel-­‐dance	  Protagonist	  to	  the	  same-­‐dance	  Protagonist	  (23	  of	  32	  infants,	  binomial	  probability	  test,	  p	  =	  .01).	  Replicating	  Study	  1,	  infants	  in	  the	  Repetition	  condition	  of	  Study	  2	  were	  equally	  likely	  to	  choose	  the	  same-­‐dance	  Protagonist	  as	  the	  novel-­‐dance	  Protagonist	  (15	  of	  32	  infants,	  binomial	  test,	  p	  =	  .58),	  again	  supporting	  our	  hypothesis	  that	  infants	  are	  sensitive	  to	  the	  distribution	  of	  behaviours,	  and	  not	  just	  to	  their	  frequency.	  The	  difference	  in	  choice	  patterns	  between	  Consensus	  and	  Repetition	  conditions	  was	  marginally	  significant	  (Fischer’s	  Exact	  Test,	  p	  =	  .098).	  There	  was	  no	  effect	  of	  order	  of	  events,	  color	  or	  side	  of	  Protagonist	  on	  any	  comparison.	  	  	  Figure	  2.4	  Infants’	  Protagonist	  choices	  by	  condition	  in	  Study	  2.	  	  	  	  Consensus	   Repetition	  novel	  dance	   23	   17	  same	  dance	   9	   15	  0%	  25%	  50%	  75%	  100%	  Protagonist	  choice	  	   48	  2.3.3	   Discussion	  	  As	  with	  Study	  1,	  7-­‐month-­‐old	  infants	  were	  sensitive	  to	  the	  distribution	  of	  frequency	  information,	  showing	  a	  clear	  preference	  in	  the	  Consensus	  condition	  and	  no	  clear	  preference	  in	  the	  Repetition	  condition.	  In	  contrast	  to	  Study	  1,	  however,	  infants	  showed	  an	  opposite	  pattern	  of	  preferences,	  preferring	  the	  maverick	  (novel-­‐dance)	  Protagonist,	  to	  the	  imitator	  (same-­‐dance)	  Protagonist.	  These	  results	  do	  not	  support	  a	  general	  preference	  for	  imitators,	  but	  rather	  a	  context-­‐specific	  preference	  for	  those	  “in	  the	  know”	  when	  the	  targets	  of	  evaluation	  lacked	  opportunities	  to	  acquire	  group	  behaviours	  on	  the	  spot,	  and	  a	  preference	  for	  mavericks	  when	  opportunities	  for	  imitation	  were	  plentiful.	  This	  view	  is	  further	  strengthened	  by	  null	  results	  in	  the	  Repetition	  condition:	  a	  simple	  preference	  for	  imitators	  should	  lead	  infants	  to	  prefer	  anyone	  who	  repeats	  another’s	  actions,	  rather	  than	  a	  selective	  preference	  for	  those	  who	  introduce	  novel	  dances	  to	  a	  group	  where	  everyone	  can	  produce	  the	  same	  dance.	  	  2.4	  	   General	  discussion	  Infants’	  choice	  patterns	  across	  the	  2	  studies	  support	  context-­‐sensitivity	  in	  social	  evaluation.	  Whereas	  infants	  predominantly	  preferred	  the	  same-­‐dance	  Protagonist	  in	  Study	  1,	  they	  showed	  a	  preference	  for	  the	  novel-­‐dance	  Protagonist	  in	  Study	  2.	  	  In	  addition	  to	  infants’	  differing	  visual	  access	  to	  the	  Protagonists	  (the	  manipulated	  difference	  between	  Studies	  1	  and	  2),	  a	  few	  others	  factors	  could	  be	  relevant	  to	  explaining	  this	  difference.	  First,	  the	  number	  of	  characters	  present	  on	  screen	  and	  to	  which	  infants	  could	  potentially	  attend	  during	  the	  demonstration	  	   49	  phase	  is	  greater	  in	  Study	  2	  than	  in	  Study	  1.	  The	  presence	  of	  2	  additional	  individuals	  during	  the	  demonstration	  phase	  could	  have	  imposed	  a	  greater	  cognitive	  load	  on	  infants.	  Past	  research	  indicates	  that	  infants	  tend	  to	  prefer	  to	  allocate	  their	  attention	  toward	  the	  familiar	  in	  more	  complex	  tasks	  and	  toward	  the	  novel	  in	  simpler	  tasks,	  perhaps	  as	  a	  way	  of	  managing	  cognitive	  load	  (Kidd,	  Piantadosi,	  &	  Aslin,	  2012).	  The	  difference	  in	  the	  number	  of	  individuals	  present	  at	  the	  outset	  could	  have	  had	  an	  effect	  on	  infants’	  subsequent	  attention,	  leading	  them	  to	  prefer	  the	  familiar	  under	  higher	  cognitive	  load,	  and	  the	  novel	  under	  lower	  cognitive	  load.	  	  The	  presence	  of	  the	  Repetition	  conditions,	  however,	  renders	  this	  explanation	  less	  likely.	  As	  Demonstrators	  are	  matched	  to	  the	  Consensus	  conditions	  in	  the	  number	  of	  characters	  present,	  the	  number	  of	  times	  an	  action	  is	  performed,	  and	  the	  spatial	  grouping	  of	  the	  individuals.	  The	  pattern	  of	  results	  from	  these	  conditions	  suggest	  that	  it	  is	  unlikely	  for	  the	  cognitive	  demands	  arising	  from	  the	  presence	  or	  absence	  of	  the	  imitator/observers	  to	  explain	  the	  differences	  in	  the	  Consensus	  conditions,	  since	  infants’	  choices	  were	  always	  at	  chance	  in	  the	  Repetition	  conditions.	  However,	  the	  Repetition	  condition	  may	  be	  less	  cognitively	  taxing	  overall	  due	  to	  the	  action	  being	  concentrated	  in	  a	  single	  Demonstrator,	  cutting	  down	  on	  the	  demands	  of	  tracking	  the	  actions	  of	  several	  individuals.	  Another	  alternative	  account	  for	  the	  results	  is	  that	  rather	  than	  using	  the	  presence	  or	  absence	  of	  the	  Protagonists	  as	  a	  signal	  of	  knowledge,	  infants	  used	  physical	  proximity	  as	  a	  cue	  to	  group	  membership,	  founded	  on	  a	  prepared	  system	  for	  reasoning	  about	  groups	  and	  social	  behaviours	  (Powell	  &	  Spelke,	  2013).	  Due	  to	  their	  locations	  on	  either	  side	  of	  the	  screen,	  and	  the	  separate	  timing	  of	  their	  entrances	  in	  	   50	  Study	  1,	  infants	  may	  have	  perceived	  the	  Demonstrators	  and	  Protagonists	  as	  two	  distinct	  social	  groups,	  each	  presumed	  to	  have	  their	  respective	  conventional	  dance.	  Infants’	  preference	  for	  the	  maverick	  in	  Study	  2	  could	  reflect	  an	  aversion	  towards	  those	  who	  follow	  another	  group’s	  conventions	  (the	  imitator).	  However,	  the	  lack	  of	  consistency	  in	  such	  preferences	  across	  the	  two	  studies	  renders	  this	  explanation	  less	  plausible.	  If	  infants	  always	  prefer	  those	  who	  perform	  the	  different,	  in-­‐group	  dance,	  and	  are	  using	  physical	  proximity	  as	  a	  cue	  to	  group	  membership,	  we	  would	  expect	  to	  see	  a	  preference	  for	  the	  novel-­‐dance	  Protagonist	  for	  both	  studies.	  Instead,	  the	  reversal	  of	  preference	  from	  Study	  1	  to	  Study	  2	  suggests	  that	  the	  Protagonists’	  presence	  during	  the	  Demonstrators’	  performance	  likely	  created	  two	  different	  contexts	  in	  infants’	  minds.	  Infants’	  preferences	  for	  the	  absent	  same-­‐dance	  Protagonist	  in	  Study	  1	  and	  the	  present	  novel-­‐dance	  Protagonist	  in	  Study	  2	  are	  most	  parsimoniously	  explained	  by	  a	  preference	  for	  individuals	  who	  are	  knowledgeable.	  In	  Study	  1,	  when	  the	  Protagonists’	  relationship	  to	  the	  Demonstrators	  was	  unclear,	  infants	  preferred	  the	  individual	  who	  behaved	  like	  an	  in-­‐group	  member.	  In	  Study	  2,	  when	  Protagonists	  were	  present	  to	  observe	  and	  mimic	  the	  Demonstrators’	  dance,	  infants	  preferred	  the	  Protagonist	  who	  introduced	  a	  novel	  dance.	  This	  is	  not	  surprising,	  given	  that	  previous	  research	  has	  found	  infants	  to	  take	  contextual	  information	  into	  account	  when	  evaluating	  agents’	  valenced	  actions	  (Hamlin,	  2014).	  In	  particular,	  visual	  access	  (and	  presumably	  situational	  knowledge)	  of	  protagonists	  plays	  an	  important	  role	  in	  informing	  infants’	  evaluations	  (Hamlin,	  Ullman,	  Tenenbaum,	  Goodman,	  &	  Baker,	  2013).	  If	  our	  interpretation	  is	  correct,	  the	  current	  studies	  build	  on	  this	  growing	  	   51	  understanding	  of	  early	  context-­‐sensitivity	  in	  learning,	  and	  suggest	  that	  infants’	  hedge	  their	  bets	  (or	  control	  their	  exposure	  to	  risk)	  when	  it	  comes	  to	  evaluating	  social	  partners.	  When	  there	  is	  a	  great	  deal	  of	  uncertainty	  surrounding	  the	  target	  identities,	  infants	  prefer	  the	  safer	  bet	  who	  demonstrates	  their	  knowledge	  of	  group	  behaviours.	  On	  the	  other	  hand,	  when	  group	  identity	  is	  less	  in	  question,	  infants	  eschew	  the	  imitator	  for	  someone	  who	  demonstrates	  knowledge	  of	  novel	  behaviours.	  	  	  We	  have	  interpreted	  our	  results	  in	  Study	  1	  as	  infants	  liking	  those	  in	  the	  know.	  However,	  due	  to	  the	  forced	  binary	  choice	  format	  of	  the	  current	  study,	  and	  without	  a	  comparison	  between	  the	  same-­‐dance	  and	  novel-­‐dance	  Protagonists	  with	  a	  neutral	  character,	  we	  are	  unable	  to	  draw	  conclusions	  regarding	  which	  is	  driving	  the	  preference:	  are	  infants’	  choices	  of	  the	  same-­‐dance	  Protagonist	  motivated	  by	  liking	  of	  that	  individual,	  or	  an	  aversion	  of	  the	  novel-­‐dance	  Protagonist?	  In	  Study	  2,	  our	  ability	  to	  conclude	  that	  infants	  prefer	  the	  more-­‐knowledgeable	  individual	  is	  similarly	  constrained	  by	  lack	  of	  a	  neutral	  comparison.	  Are	  infants	  choosing	  the	  maverick	  because	  they	  want	  they	  prefer	  individuals	  who	  are	  different,	  or	  because	  they	  are	  motivated	  to	  avoid	  imitators	  (possibly	  imitators	  of	  out-­‐group	  conventions)?	  Follow-­‐up	  studies	  are	  needed	  to	  help	  disentangle	  the	  directionality	  of	  these	  preferences.	  	  Different	  result	  patterns	  between	  Consensus	  and	  Repetition	  conditions	  in	  both	  studies	  support	  the	  proposal	  that	  infants’	  are	  sensitive	  to	  the	  distribution	  of	  frequently	  observed	  behaviours.	  This	  is	  consistent	  with	  the	  interpretation	  that	  infants	  identify	  prevalent	  behaviours	  as	  shared	  amongst	  members	  of	  a	  community,	  and	  thus,	  conventional.	  In	  real	  human	  groups,	  behaviours	  and	  ideas	  are	  distributed	  imperfectly	  across	  individuals	  (less	  than	  100%)	  for	  a	  number	  of	  reasons.	  For	  	   52	  example,	  hand	  washing	  after	  using	  the	  bathroom	  is	  a	  very	  prevalent	  behaviour	  in	  many	  societies	  with	  access	  to	  indoor	  plumbing.	  However,	  even	  within	  these	  societies,	  there	  are	  some	  individuals	  who	  do	  not	  consistently	  wash	  their	  hands	  after	  using	  the	  bathroom.	  Yet	  adults	  who	  observe	  a	  day’s	  activities	  in	  a	  public	  bathroom	  would	  easily	  make	  the	  inference	  that	  hand	  washing	  is	  a	  group	  norm,	  even	  with	  imperfect	  agreement	  across	  observations.	  To	  be	  a	  useful	  mechanism	  for	  identifying	  knowledgeable	  members	  of	  the	  in-­‐group,	  some	  tolerance	  for	  imperfect	  levels	  of	  agreement	  in	  behaviours	  should	  be	  expected.	  Future	  studies	  could	  identify	  the	  boundary	  conditions	  on	  infants’	  sensitivity	  to	  the	  level	  of	  consensus	  required	  across	  group	  members	  for	  conducting	  the	  group-­‐level	  analyses	  seen	  in	  the	  studies	  above.	  	   Finally,	  the	  studies	  in	  this	  chapter	  present	  preliminary	  evidence	  that	  infants	  may	  be	  motivated	  to	  evaluate	  social	  others	  in	  order	  to	  identify	  knowledgeable	  individuals.	  It	  was	  difficult	  to	  directly	  address	  the	  question	  of	  learning	  with	  a	  preverbal	  population.	  In	  the	  following	  chapter,	  we	  discuss	  a	  study	  in	  which	  we	  examine	  2-­‐	  to	  6-­‐year-­‐olds’	  use	  of	  frequency	  information	  to	  inform	  both	  preference	  and	  learning.	  	  	  	  	  	  	  	  	  	   53	  3.	  Preschoolers	  like	  and	  learn	  from	  nonconformists	  	  	  3.1	   Introduction	  Cultural	  learning	  presents	  a	  conundrum:	  to	  learn	  good	  cultural	  information,	  a	  learner	  must	  identify	  high	  quality	  sources.	  Yet	  learners	  are	  typically	  at	  a	  disadvantage,	  lacking	  sufficient	  expertise	  to	  evaluate	  the	  quality	  of	  content	  offered	  by	  alternative	  sources.	  A	  discriminating	  learner	  must	  successfully	  identify	  characteristics	  of	  knowledge	  sources	  that	  are	  correlated	  with,	  or	  indicative	  of,	  information	  quality.	  While	  recent	  work	  has	  found	  children	  to	  be	  astute	  choosers	  of	  knowledge	  sources,	  these	  studies	  shared	  the	  assumption	  that	  young	  children	  make	  “epistemically	  adaptive”	  safe	  bets	  such	  as	  following	  a	  consensus	  (Koenig,	  2013).	  The	  only	  cases	  to	  the	  contrary	  occur	  when	  children	  are	  able	  to	  successfully	  gauge	  that	  a	  minority	  is	  more	  successful	  than	  the	  group.	  However,	  children	  hold	  many	  irrational	  beliefs.	  Is	  this	  because	  children	  are	  not	  astute	  judges	  after	  all?	  Or	  might	  there	  be	  a	  rational	  reason	  why	  despite	  their	  best	  efforts	  to	  evaluate	  their	  knowledge	  sources,	  children	  still	  end	  up	  with	  “out	  there”	  beliefs?	  	  There	  is	  a	  growing	  literature	  showing	  that	  children	  are	  selective	  about	  whom	  to	  learn	  from.	  Though	  they	  may	  lack	  the	  ability	  to	  evaluate	  the	  quality	  of	  content	  itself,	  young	  children	  appear	  capable	  of	  assessing	  the	  quality	  of	  sources	  it	  comes	  from.	  Three-­‐year-­‐olds	  are	  more	  likely	  to	  learn	  from	  previously	  accurate	  sources	  (Birch	  et	  al.,	  2008;	  Clément,	  Koenig,	  &	  Harris,	  2004;	  Jaswal	  &	  Malone,	  2007;	  Jaswal	  &	  Neely,	  2006;	  Koenig	  &	  Harris,	  2005;	  Pasquini	  et	  al.,	  2007;	  Scofield	  &	  Behrend,	  2008),	  rejecting	  information	  from	  informants	  who	  have	  shown	  ignorance	  of	  labels	  for	  	   54	  familiar	  objects	  (Koenig	  &	  Harris,	  2005;	  Sabbagh	  &	  Baldwin,	  2001).	  By	  their	  4th	  year,	  children	  track	  the	  degree	  of	  informants’	  past	  reliability	  in	  deciding	  whom	  to	  trust	  (Pasquini	  et	  al.,	  2007).	  	  This	  tendency	  to	  evaluate	  informant	  past	  accuracy	  is	  clearly	  adaptive,	  as	  it	  can	  help	  children	  avoid	  learning	  inaccurate	  information	  from	  ignorant	  or	  misinformed	  individuals.	  However,	  it	  requires	  children	  to	  have	  prior	  knowledge	  on	  which	  to	  base	  such	  assessments	  of	  competence.	  	  A	  dimension	  that	  does	  not	  require	  direct	  evaluation	  of	  competence	  is	  an	  individual’s	  willingness	  to	  share	  knowledge.	  Benevolence	  is	  an	  important	  attribute	  of	  good	  sources,	  since	  informants	  could	  be	  stingy	  about	  sharing	  knowledge,	  or	  give	  misleading	  information.	  Three	  to	  five	  year	  olds	  have	  demonstrated	  an	  ability	  to	  evaluate	  informants’	  benevolence	  in	  experimental	  settings,	  preferring	  to	  learn	  from	  previously	  helpful	  informants	  to	  unhelpful	  ones	  (Liu	  et	  al.,	  2013).	  However,	  even	  evaluations	  of	  benevolence	  require	  some	  knowledge	  of	  an	  individual’s	  past	  behaviour,	  specifically,	  whether	  they	  have	  acted	  with	  good	  intentions	  in	  the	  past.	  Strategies	  that	  rely	  on	  having	  historical	  knowledge	  may	  be	  insufficient	  to	  help	  learners	  decide	  between	  potential	  informants,	  as	  the	  ultra-­‐sociality	  of	  human	  groups	  present	  many	  situations	  in	  which	  one	  must	  assess	  and	  learn	  from	  strangers.	  How	  can	  naïve	  learners	  rely	  on	  contextual	  information	  to	  make	  an	  informed	  guess	  regarding	  which	  individual	  is	  a	  better	  cultural	  informant	  in	  terms	  of	  quality	  and	  benevolence?	  Lacking	  the	  ability	  to	  directly	  assess	  the	  quality	  of	  information,	  and	  the	  opportunity	  to	  assess	  informants’	  past	  behaviours,	  confidence	  provides	  a	  measure	  of	  the	  speaker’s	  own	  assessment	  of	  accuracy.	  There	  is	  some	  evidence	  that	  2-­‐	  and	  3-­‐	   55	  year-­‐old	  children	  prefer	  to	  learn	  from	  confident	  actors	  and	  speakers,	  compared	  to	  hesitant	  ones	  (Birch	  et	  al.,	  2010;	  Jaswal	  &	  Malone,	  2007),	  providing	  one	  way	  for	  learners	  to	  evaluate	  the	  competence	  of	  complete	  strangers.	  However,	  there	  is	  accumulating	  evidence	  that	  people	  tend	  to	  err	  on	  overconfidence	  in	  their	  own	  abilities	  (Kruger	  &	  Dunning,	  1999;	  Robins	  &	  Beer,	  2001;	  Svenson,	  1981;	  Williams	  &	  Gilovich,	  2008).	  Thus,	  while	  a	  useful	  metric,	  relying	  on	  informants’	  own	  assessments	  may	  be	  prone	  to	  error,	  and	  result	  in	  misinformation.	  Ideally,	  learners	  could	  deploy	  additional	  metrics	  to	  assess	  the	  quality	  of	  potential	  informants.	  	  Paying	  attention	  to	  the	  prevalence,	  or	  frequency	  of	  occurrence,	  of	  behaviours	  is	  another	  useful	  metric	  for	  learning	  behaviours	  that	  lead	  to	  good	  outcomes.	  Learners	  could	  prefer	  to	  learn	  commonly	  seen	  behaviours,	  or,	  alternatively,	  choose	  to	  learn	  rarely	  seen	  behaviours.	  Both	  strategies	  are	  “frequency-­‐dependent	  learning”,	  because	  they	  are	  decision	  rules	  based	  on	  the	  property	  of	  frequency	  of	  occurrence	  of	  behaviours,	  rather	  than	  their	  content	  or	  outcomes.	  In	  other	  words,	  frequency-­‐dependent	  learning	  does	  not	  require	  evaluating	  the	  content	  or	  payoff	  of	  possible	  behaviours,	  but	  instead,	  focuses	  on	  characteristics	  of	  the	  situation,	  namely	  the	  proportion	  of	  people	  performing	  it	  (Boyd	  &	  Richerson,	  1985;	  Henrich	  &	  McElreath,	  2003;	  Laland,	  2004).	  Both	  variants	  of	  frequency	  dependent	  learning	  play	  a	  role	  in	  cultural	  learning,	  and	  each	  can	  lead	  to	  favourable	  learning	  outcomes	  in	  different	  situations.	  It	  may	  be	  beneficial	  to	  learn	  commonly	  occurring	  behaviours	  because	  they	  have	  been	  tried	  and	  tested.	  This	  strategy	  is	  particularly	  useful	  when	  physical	  causality	  between	  actions	  and	  outcomes	  are	  relatively	  opaque,	  and	  the	  efficacy	  	   56	  difficult	  to	  assess	  immediately.	  In	  these	  situations,	  looking	  to	  the	  majority	  lets	  other	  people	  do	  the	  work	  of	  filtering	  out	  bad	  strategies	  that	  lead	  to	  poor	  outcomes.	  	  Several	  studies	  show	  that	  children	  use	  information	  about	  whether	  an	  individual	  formed	  part	  of	  a	  majority	  to	  inform	  their	  choice	  of	  whom	  to	  learn	  from.	  	  In	  one	  study,	  4	  year	  olds	  were	  presented	  with	  novel	  objects	  and	  given	  the	  task	  to	  learn	  their	  labels.	  One	  informant	  provided	  labels	  while	  bystanders	  assented	  with	  smiles	  and	  nods,	  thus	  forming	  a	  consensus	  with	  the	  informant.	  A	  second	  informant	  provided	  labels	  while	  bystanders	  dissented	  with	  frowns	  and	  head	  shakes,	  thus	  putting	  them	  in	  the	  minority.	  Lacking	  direct	  means	  to	  assess	  the	  accuracy	  of	  the	  labels,	  children	  took	  into	  account	  whether	  labels	  were	  met	  with	  assent	  or	  dissent	  from	  observing	  bystanders,	  preferring	  to	  learn	  from	  the	  informant	  who	  received	  bystanders’	  assent	  (Fusaro	  &	  Harris,	  2008).	  In	  another	  study,	  3	  and	  4	  year	  olds	  looked	  to	  consensus	  information,	  preferring	  to	  learn	  from	  those	  who	  belonged	  to	  a	  consensus	  rather	  than	  a	  lone	  dissenter	  when	  deciding	  whether	  or	  not	  to	  trust	  an	  informant’s	  testimony	  (Corriveau,	  Fusaro,	  &	  Harris,	  2009).	  One	  study	  further	  showed	  that	  while	  children	  were	  hesitant	  to	  trust	  an	  informant	  who	  was	  a	  member	  of	  an	  out-­‐group,	  they	  were	  willing	  to	  suspend	  their	  skepticism	  when	  these	  informants	  formed	  a	  consensus	  (Chen,	  Corriveau,	  &	  Harris,	  2012).	  Thus,	  a	  number	  of	  studies	  point	  to	  children	  being	  more	  persuaded	  by	  consensus	  information,	  sometimes,	  even	  over	  their	  own	  better	  judgment.	  	  On	  the	  other	  hand,	  despite	  the	  critical	  importance	  of	  learning	  from	  and	  associating	  with	  the	  majority	  in	  human	  social	  life,	  innovation	  is	  also	  an	  ubiquitous	  aspect	  of	  cultural	  learning,	  and	  can	  lead	  to	  true	  improvement	  upon	  existing	  	   57	  knowledge	  (Legare	  &	  Nielsen,	  in	  press).	  While	  it	  is	  a	  safe	  bet	  to	  look	  to	  the	  majority	  when	  outcomes	  are	  difficult	  to	  assess	  directly,	  by	  definition	  it	  results	  in	  only	  average	  payoffs.	  Experts	  may	  deviate	  from	  the	  majority	  of	  individuals	  in	  their	  domain	  of	  expertise	  because	  of	  greater	  efficacy	  or	  ability	  to	  innovate	  improvements	  to	  the	  existing	  process	  (Muthukrishna	  et	  al,	  2015).	  	  Recent	  studies	  showed	  that	  children	  are	  willing	  to	  endorse	  a	  successful	  but	  unconventional	  informant	  over	  conventional	  but	  unsuccessful	  ones	  (Scofield,	  Gilpin,	  Pierucci,	  &	  Morgan,	  2013).	  Similarly,	  Seston	  and	  Keleman	  (2013)	  found	  that	  children	  followed	  consensus	  when	  majority	  and	  minority	  opinions	  were	  equally	  plausible,	  however,	  they	  followed	  minority	  opinion	  if	  it	  was	  deemed	  more	  plausible.	  Finally,	  children	  observing	  demonstrations	  of	  how	  to	  open	  novel	  puzzle	  boxes	  were	  equally	  likely	  to	  learn	  from	  an	  individual	  as	  a	  group.	  However,	  they	  were	  more	  likely	  to	  learn	  from	  a	  successful	  individual	  than	  an	  unsuccessful	  group	  (Wilks,	  Collier-­‐Baker,	  &	  Nielsen,	  2014).	  Together,	  these	  studies	  suggest	  that	  when	  able,	  children	  will	  choose	  to	  learn	  from	  a	  minority	  if	  they	  trust	  that	  they	  will	  be	  more	  successful.	  Despite	  evidence	  supporting	  children’s	  own	  tendencies	  to	  use	  frequency-­‐dependent	  learning	  strategies,	  little	  work	  has	  been	  done	  looking	  at	  whether	  children	  prefer	  others	  who	  perform	  uncommon	  and	  widespread	  behaviours,	  and	  their	  tendencies	  to	  learn	  from	  such	  individuals.	  For	  the	  same	  reasons	  that	  learners	  may	  at	  times	  prefer	  mavericks	  and	  at	  others,	  conventional	  actors,	  they	  may	  prefer	  others	  who	  do	  similarly.	  To	  the	  extent	  that	  conventional	  behaviours	  indicate	  that	  an	  actor	  is	  a	  part	  of	  one’s	  ingroup,	  interacting	  with	  such	  individuals	  should	  facilitate	  coordination.	  When	  there	  are	  multiple	  potential	  models	  whose	  group	  memberships	  	   58	  are	  unclear,	  interacting	  with	  someone	  who	  appears	  to	  be	  in	  one’s	  ingroup	  will	  also	  insure	  against	  accidentally	  learning	  outgroup	  behaviours.	  On	  the	  other	  hand,	  in	  homogeneous	  group	  environments	  where	  the	  occurrence	  of	  such	  mistakes	  are	  relatively	  unlikely,	  interacting	  with	  unconventional	  individuals	  who	  may	  be	  extra-­‐skillful	  could	  be	  advantageous	  both	  in	  the	  resources	  they	  may	  gain	  through	  proximity,	  and	  in	  the	  higher	  quality	  skills	  they	  can	  learn	  from	  such	  individuals.	  	  The	  link	  between	  an	  individual’s	  likeability	  and	  suitability	  as	  a	  cultural	  model	  is	  an	  important	  one	  because	  in	  real	  life,	  we	  don’t	  always	  get	  opportunities	  to	  evaluate	  others	  in	  a	  group	  setting	  at	  the	  moments	  when	  we	  want	  to	  learn.	  Generating	  a	  conclusion	  about	  the	  quality	  of	  a	  potential	  model	  beyond	  the	  evaluative	  context	  is	  an	  important	  feature	  of	  cultural	  learning.	  Children	  and	  adults	  tend	  to	  form	  such	  generalizations,	  from	  positive	  evaluation	  due	  to	  competence	  in	  one	  domain,	  to	  imitation	  in	  unrelated	  domains.	  Such	  generalizations	  explain	  why	  celebrity	  endorsements	  work,	  and	  why	  Olympic	  figure	  skaters	  are	  recruited	  to	  sell	  toothpaste.	  Do	  children	  generalize	  from	  positive	  evaluations	  in	  one	  domain	  to	  suitability	  as	  a	  cultural	  model	  in	  an	  unrelated	  domain?	  	  There	  are	  at	  least	  3	  routes	  by	  which	  liking	  someone	  can	  increase	  one’s	  propensity	  to	  learn	  from	  them,	  including	  proximity,	  trust,	  and	  perceived	  competence.	  For	  one,	  we	  may	  favourably	  evaluate	  someone	  specifically	  because	  they	  are	  competent	  in	  a	  content	  area	  that	  we	  deem	  important,	  and	  we	  want	  to	  associate	  with	  them	  to	  acquire	  their	  skills	  (consciously	  motivated,	  or	  not).	  In	  this	  case,	  liking	  is	  a	  direct	  assessment	  of	  whether	  an	  individual	  is	  good	  to	  learn	  from.	  	   59	  However,	  skill	  alone	  may	  be	  insufficient,	  if	  an	  individual	  is	  unwilling	  to	  impart	  their	  knowledge,	  or	  intentionally	  distorts	  it	  to	  their	  advantage.	  	  Another	  important	  reason	  why	  we	  may	  prefer	  to	  learn	  from	  those	  we	  like	  is	  because	  we	  have	  deemed	  them	  to	  be	  trustworthy.	  Generally,	  we	  tend	  to	  prefer	  individuals	  who	  uphold	  their	  promises	  to	  others,	  and	  ourselves,	  and	  who	  do	  not	  appear	  to	  be	  deceitful.	  Trustworthiness	  is	  the	  foundation	  to	  smooth	  social	  interactions,	  as	  we	  expect	  others	  to	  fulfill	  their	  explicitly	  stated	  intentions	  (to	  arrive	  at	  7pm,	  or	  to	  provide	  food	  at	  a	  party),	  as	  well	  as	  to	  uphold	  implicitly	  shared	  social	  obligations	  (to	  not	  sit	  in	  an	  elevator,	  or	  to	  return	  lost	  belongings).	  It’s	  a	  good	  idea	  to	  preferentially	  learn	  from	  trustworthy	  individuals	  because	  bad	  information	  can	  lead	  to	  costly	  errors,	  and	  verifying	  it	  can	  be	  difficult,	  so	  we	  should	  avoid	  people	  who	  may	  intentionally	  mislead	  us.	  	  In	  addition	  to	  qualities	  that	  are	  directly	  relevant	  to	  learning,	  we	  may	  like	  someone	  for	  a	  variety	  of	  other	  reasons:	  they	  may	  be	  funny,	  or	  own	  a	  game	  we	  like	  to	  play.	  	  Regardless	  of	  why	  we	  do	  so,	  liking	  can	  lead	  to	  spending	  more	  time	  in	  close	  proximity	  to	  that	  individual,	  and	  adopting	  the	  person’s	  behaviours	  and	  attitudes	  due	  to	  regular	  exposure.	  Research	  on	  attitude	  change	  find	  that	  we	  are	  indeed	  more	  likely	  to	  be	  persuaded	  to	  change	  our	  beliefs	  and	  attitudes	  by	  people	  we	  like	  (Chaiken,	  1980;	  Petty	  &	  Cacioppo,	  1981).	  At	  the	  same	  time,	  by	  spending	  time	  with	  one	  individual,	  one	  forgoes	  the	  opportunity	  to	  learn	  from	  other	  potential	  models	  during	  this	  time.	  Thus,	  in	  the	  broadest	  sense,	  liking	  someone	  can	  lead	  to	  learning	  from	  them	  through	  mere	  exposure,	  and	  a	  willingness	  to	  be	  like	  them.	  	  	  	   60	  Existing	  evidence	  is	  suggestive	  that	  children	  extend	  social	  evaluations	  to	  inform	  model	  preferences	  for	  food,	  and	  toy	  choices	  (Hamlin	  &	  Wynn,	  2012;	  Kinzler	  et	  al.,	  2010).	  Can	  children	  use	  frequency	  information	  for	  picking	  informants?	  Studies	  in	  the	  previous	  chapter	  demonstrated	  that	  in	  a	  3rd	  party	  context,	  7-­‐month	  old	  infants	  are	  sensitive	  to	  the	  similarity	  of	  an	  individual’s	  actions	  to	  a	  group’s,	  suggesting	  that	  young	  humans	  are	  at	  least	  capable	  of	  forming	  an	  evaluation	  on	  these	  grounds.	  However,	  what	  does	  such	  preferences	  say	  about	  one’s	  propensity	  to	  learn	  from	  a	  positively	  regarded	  individual?	  	  	  3.2	   The	  present	  study	  The	  present	  study	  extends	  on	  Chapter	  2	  by	  looking	  at	  the	  developmental	  trajectory	  of	  frequency-­‐based	  social	  evaluations	  in	  2-­‐6	  year	  olds,	  and	  examines	  the	  additional	  question	  of	  whether	  children	  use	  3rd	  party	  frequency-­‐based	  analysis	  for	  choosing	  informants.	  Specifically,	  we	  ask:	  do	  children	  evaluate	  others	  based	  on	  their	  performance	  of	  a	  commonly	  seen	  behaviour?	  And,	  do	  children	  use	  frequency	  information	  to	  guide	  informant	  choices	  in	  a	  different	  domain?	  	  Study	  methods	  from	  Chapter	  2	  were	  adapted	  to	  be	  more	  suitable	  for	  preschool-­‐age	  children.	  We	  created	  a	  live	  action	  dance	  show	  using	  generic	  Smurf	  plush	  toys	  (lacking	  identifying	  features	  of	  the	  main	  characters),	  while	  maintaining	  the	  basic	  structure	  of	  the	  show.	  The	  experiment	  was	  conducted	  following	  the	  recent	  release	  of	  a	  Smurfs	  movie,	  so	  the	  toys	  were	  familiar	  and	  engaging	  to	  many	  children.	  Importantly,	  Smurfs	  look	  like	  members	  of	  a	  distinct	  social	  group,	  and	  were	  introduced	  as	  such	  by	  the	  Experimenter,	  by	  saying	  “Do	  you	  know	  who	  these	  guys	  are?	  They	  are	  Smurfs!”	  	  	   61	  Four	  identical	  Smurfs	  were	  “Demonstrators”	  and	  2	  were	  “Protagonists”.	  Protagonists	  were	  distinguishable	  from	  each	  other	  by	  wearing	  vertically	  striped	  vs.	  horizontally	  striped	  hats;	  they	  were	  otherwise	  identical.	  Smurfs	  were	  manipulated	  by	  the	  Experimenter	  to	  perform	  dances.	  To	  control	  for	  familiarity	  and	  novelty	  effect,	  we	  maintained	  the	  2	  (between-­‐subjects)	  conditions:	  in	  the	  Consensus	  condition,	  4	  Demonstrators	  each	  performed	  the	  same	  dance	  once;	  in	  the	  Repetition	  condition,	  the	  number	  and	  type	  of	  dances	  was	  kept	  equivalent	  to	  that	  in	  the	  Consensus	  condition,	  however,	  all	  4	  were	  performed	  by	  a	  single	  Demonstrator.	  In	  both	  conditions,	  after	  the	  4th	  and	  final	  dance,	  all	  4	  Demonstrators	  were	  removed	  from	  the	  stage.	   Replicating	  the	  design	  of	  Study	  1	  from	  the	  previous	  chapter,	  Protagonists	  were	  absent	  during	  the	  demonstration	  phase,	  and	  were	  introduced	  only	  after	  Demonstrators	  had	  exited	  the	  stage.	  Despite	  the	  superficial	  similarity	  in	  structure	  between	  the	  current	  and	  previous	  Chapter’s	  studies,	  an	  important	  alteration	  was	  made.	  Whereas	  group	  membership	  was	  never	  explicitly	  addressed,	  but	  may	  have	  been	  implied	  by	  the	  Protagonists’	  presence	  vs.	  absence	  in	  the	  infant	  studies,	  the	  Smurfs’	  common	  group	  membership	  was	  explicitly	  stated	  during	  the	  introduction	  of	  the	  current	  study.	  If	  children	  prefer	  and	  wish	  to	  learn	  from	  an	  extra-­‐knowledgeable	  Smurf,	  they	  may	  choose	  the	  novel-­‐dance	  Protagonist,	  as	  in	  Study	  2.	  However,	  if	  they	  prefer	  a	  conventional	  Smurf,	  they	  may	  choose	  the	  same-­‐dance	  Protagonist,	  as	  in	  Study	  1.	  	  	  	   62	  3.2.1	   Method	  One	  hundred	  and	  ninety-­‐eight	  children	  participated	  in	  the	  study	  (Mean	  age	  =	  3.98,	  44.6%	  female,	  range	  =	  2	  years,	  0	  days	  –	  6	  years,	  0	  days).	  Data	  from	  19	  children	  were	  excluded	  due	  to	  parental	  interference,	  or	  to	  providing	  no	  choice	  on	  both	  the	  dependent	  measures.	  Participants	  were	  recruited	  during	  a	  visit	  to	  the	  local	  Science	  Museum.	  Parents	  were	  approached	  with	  a	  description	  of	  the	  study,	  and	  if	  they	  were	  interested,	  were	  invited	  to	  visit	  a	  specific	  area	  of	  the	  museum	  where	  there	  were	  designated	  testing	  rooms.	  The	  majority	  of	  participants	  were	  White	  and	  all	  were	  English	  speaking	  (though	  not	  necessarily	  as	  a	  first	  language),	  though	  a	  range	  of	  ethnicities	  and	  SES	  backgrounds	  were	  represented.	  Children	  were	  tested	  individually	  in	  a	  testing	  room,	  seated	  across	  a	  table	  from	  the	  Experimenter.	  To	  introduce	  the	  study,	  the	  Experimenter	  gestured	  to	  4	  Demonstrator	  Smurfs	  seated	  in	  a	  group	  to	  the	  left	  and	  2	  Protagonist	  Smurfs	  seated	  in	  a	  group	  to	  the	  right,	  all	  across	  from	  participants	  on	  the	  table,	  and	  asked,	  “Do	  you	  know	  who	  these	  guys	  are?	  That’s	  right,	  they’re	  Smurfs!	  We’re	  going	  to	  see	  these	  Smurfs	  do	  a	  dance	  today.”	  After	  the	  introductions,	  Protagonists	  were	  removed	  from	  the	  table	  and	  placed	  out	  of	  sight,	  while	  Demonstrators	  remained	  seated	  on	  the	  table	  in	  their	  positions.	  Next,	  the	  Experimenter	  transitioned	  to	  the	  demonstration	  phase,	  and	  in	  the	  Consensus	  condition,	  proceeded	  to	  animate	  each	  Demonstrator	  in	  turn	  to	  perform	  a	  dance.	  Each	  dance	  was	  preceded	  by	  an	  excited	  “My	  turn!”	  as	  the	  Smurf	  was	  being	  animated,	  followed	  by	  moving	  to	  the	  centre	  of	  the	  stage.	  Demonstrators	  performed	  one	  of	  2	  dances:	  the	  Jumping	  dance	  and	  the	  Swaying	  dance.	  The	  Jumping	  dance	  	   63	  consisted	  of	  Smurfs	  jumping	  up	  and	  down	  4	  times,	  and	  the	  Swaying	  dance	  consisted	  of	  swaying	  side	  to	  side	  4	  times;	  both	  dances	  were	  performed	  at	  the	  same	  rhythm,	  for	  the	  same	  total	  duration,	  and	  Smurfs	  moved	  approximately	  the	  same	  distance	  (up	  or	  side	  to	  side)	  from	  their	  starting	  places	  during	  each	  one.	  After	  the	  Demonstrators’	  performance,	  the	  Experimenters	  said,	  “ok	  Bye!	  See	  you	  later!”	  and	  were	  removed	  together	  from	  the	  table.	  In	  the	  Repetition	  condition,	  all	  procedures	  were	  identical,	  but	  rather	  than	  4	  Demonstrators	  performing	  in	  sequence,	  a	  single	  Demonstrator	  performed	  the	  same	  dance	  4	  times.	  Between	  each	  performance,	  the	  Demonstrator	  travelled	  back	  to	  its	  starting	  position,	  exclaimed	  “my	  turn!”	  and	  proceeded	  to	  the	  centre	  of	  the	  table	  to	  perform	  its	  dance.	  Its	  physical	  actions	  and	  words	  thus	  closely	  matched	  that	  of	  the	  Consensus	  condition.	  	  Following	  the	  demonstration	  phase,	  the	  Protagonists	  were	  reintroduced	  to	  the	  table.	  The	  Experimenter	  said,	  “Let’s	  see	  what	  these	  guys	  do”,	  and	  one	  of	  the	  Protagonists	  performed	  the	  same	  dance	  as	  the	  Demonstrators,	  and	  the	  other	  Protagonist	  performed	  the	  novel	  dance.	  For	  example,	  when	  the	  Demonstrators	  performed	  the	  Jumping	  dance,	  the	  same-­‐dance	  Protagonist	  also	  performed	  the	  Jumping	  dance,	  while	  the	  novel-­‐dance	  Protagonist	  performed	  the	  Swaying	  dance.	  Dances	  performed	  by	  the	  Demonstrator(s),	  performance	  orders,	  and	  Protagonist	  type	  (whether	  they	  performed	  the	  same	  or	  novel	  dance)	  were	  counterbalanced	  across	  subjects.	  	  After	  each	  child	  viewed	  the	  dances,	  they	  were	  presented	  with	  the	  2	  Protagonists	  side-­‐by-­‐side	  in	  the	  centre	  of	  the	  table	  and	  asked,	  “Which	  one	  do	  you	  like	  more?”	  If	  the	  child	  did	  not	  provide	  a	  choice	  after	  3	  seconds,	  they	  were	  prompted	  	   64	  by	  the	  Experimenter,	  “Do	  you	  like	  one	  of	  these	  guys	  more	  than	  the	  other?”	  A	  small	  number	  of	  children	  (n	  =	  9,	  4.3	  %	  of	  the	  sample)	  claimed	  to	  like	  the	  2	  Protagonists	  equally;	  their	  responses	  for	  Liking	  were	  excluded	  from	  the	  analyses.	  Following	  their	  response	  for	  Liking,	  an	  unfamiliar	  object	  (a	  metal	  thermos	  cap)	  was	  introduced.	  The	  Experimenter	  held	  the	  object	  and	  rotated	  it	  in	  different	  angles,	  then	  placed	  it	  on	  the	  table	  in	  front	  of	  the	  child.	  Children	  were	  asked	  if	  they	  knew	  what	  it	  was;	  none	  did.	  The	  Experimenter	  then	  said,	  “These	  guys	  have	  different	  names	  for	  this	  object,	  let’s	  hear	  what	  they	  think	  it’s	  called.”	  The	  Experimenter	  then	  picked	  up	  each	  of	  the	  Protagonists	  in	  turn	  to	  point	  at	  the	  cap	  and	  label	  it;	  one	  said,	  “it’s	  a	  pavo!”	  and	  the	  other	  said,	  “it’s	  a	  loba!”	  Children	  were	  then	  asked,	  “What	  do	  you	  think	  it’s	  called?”	  Children’s	  responses	  were	  recorded,	  and	  all	  participants	  were	  thanked	  and	  given	  a	  sticker	  for	  their	  participation.	  	  	  3.2.2	   Results	  Liking	  In	  response	  to	  the	  question	  “who	  do	  you	  like	  more?”	  children	  picked	  the	  novel-­‐dance	  Protagonist	  more	  often	  in	  the	  Consensus	  condition	  (57	  of	  81,	  or	  70.3%,	  binomial	  probability	  test,	  p	  <.001,	  2-­‐tailed),	  but	  did	  not	  show	  a	  preference	  in	  the	  Repetition	  condition	  (51	  of	  90,	  or	  56.6%,	  binomial	  probability	  test,	  p	  =	  .246,	  2-­‐tailed).	  This	  supports	  our	  prediction	  that	  children’s	  social	  preferences	  may	  be	  informed	  by	  the	  distribution	  of	  observed	  behaviours.	  Furthermore,	  their	  preference	  for	  the	  novel-­‐dance	  Protagonist	  corroborates	  our	  explanation	  that	  when	  individuals’	  group	  identity	  is	  clear,	  children	  prefer	  individuals	  who	  introduce	  novel,	  rather	  than	  	   65	  conventional,	  behaviours.	  However,	  these	  preferences	  show	  marked	  differences	  by	  age.	  	  Effect	  of	  age	  	  Two-­‐	  and	  3-­‐year-­‐old	  children	  do	  not	  show	  significant	  preferences	  for	  either	  Protagonist	  in	  either	  Consensus	  or	  Repetition	  conditions.	  The	  proportion	  of	  2-­‐year-­‐olds	  who	  preferred	  the	  novel-­‐dance	  Protagonist	  was	  53%	  in	  the	  Consensus	  condition	  (binomial	  probability	  test,	  p	  =	  1),	  and	  62%	  in	  the	  Repetition	  condition	  (binomial	  prob.	  test,	  p	  =	  .27),	  and	  proportion	  of	  3-­‐year-­‐olds	  was	  65%	  in	  the	  Consensus	  condition	  (binomial	  prob.	  test,	  p	  =	  .21)	  and	  69%	  in	  the	  Repetition	  condition	  (binomial	  prob.	  test,	  =	  .21).	  Furthermore,	  there	  was	  no	  significant	  difference	  between	  conditions	  at	  either	  age	  (Fisher’s	  Exact	  Test,	  p	  =	  .76	  for	  2	  year-­‐olds,	  and	  p	  =	  1	  for	  3	  year	  olds).	  Children	  started	  to	  show	  a	  significant	  preference	  for	  the	  novel-­‐dance	  character	  at	  age	  4	  (proportion	  choosing	  novel-­‐dance	  Protagonist	  =	  76%,	  p	  =	  .016,	  2-­‐tailed),	  and	  did	  so	  only	  in	  the	  Consensus,	  but	  not	  the	  Repetition,	  condition	  (Fisher’s	  Exact	  Test,	  p	  =	  .089,	  2	  tailed).	  This	  pattern	  becomes	  more	  pronounced	  by	  age	  5,	  where	  88%	  of	  children	  in	  the	  Consensus	  condition	  chose	  the	  novel-­‐dance	  Protagonist	  (p	  =	  .006),	  compared	  to	  44%	  in	  the	  Repetition	  condition	  (p	  =	  .81).	  The	  difference	  between	  conditions	  is	  significant	  by	  a	  Fisher’s	  Exact	  Test	  (p	  =	  .013).	  In	  summary,	  the	  overall	  pattern	  described	  earlier	  was	  due	  to	  the	  4	  and	  5	  year	  olds	  both	  differentiating	  between	  Repetition	  and	  Consensus	  conditions,	  and	  preferring	  the	  novel-­‐dance	  Protagonist	  in	  the	  Consensus	  condition.	  	  	  	  	  	   66	  	  	  Figure	  3.1	  Proportion	  of	  children	  who	  liked	  the	  novel-­‐dance	  Protagonist,	  by	  	  	  	  	  	  	  	  	  age	  and	  by	  condition.	  	  	  	  	  	  	  	  	  	  	  	  Table	  3.1	  Children’s	  probabilities	  of	  liking	  the	  novel-­‐dance	  Protagonist,	  by	  age	  and	  by	  condition.	  	  	  Magnitude	  estimates	  for	  liking	  We	  employed	  a	  second	  analytic	  strategy	  to	  examine	  the	  magnitude	  of	  effects	  of	  age	  and	  condition	  on	  children’s	  likelihood	  of	  preferring	  the	  novel-­‐dance	  Protagonist.	  For	  this	  analysis,	  a	  binary	  logistic	  regression	  was	  run	  using	  combined	  	  	   Consensus	   Repetition	  difference	  between	  conditions	  	  	  Age	  group	   Pr(novel)	   p-­‐value*	   Pr(novel)	   p-­‐value*	   p-­‐value**	   n	  2	  to	  3	   0.53	   1	   0.62	   0.27	   0.76	   46	  3	  to	  4	   0.65	   0.21	   0.69	   0.21	   1	   40	  4	  to	  5	   0.76	   0.016	   0.52	   1	   0.089	   54	  5	  to	  6	   0.88	   0.006	   0.44	   0.81	   0.013	   34	  *Binomial	  probability	  test	  (2-­‐tailed)	  	  **Fisher's	  Exact	  Test	  	  	  0	  0.1	  0.2	  0.3	  0.4	  0.5	  0.6	  0.7	  0.8	  0.9	  1	  2	  to	  3	   3	  to	  4	   4	  to	  5	   5	  to	  6	  Proportion	  of	  children	  who	  chose	  the	  novel-­‐dance	  Protagonist	  Age	  group	  Consensus	  Repetition	  *	   67	  data	  from	  participants	  of	  all	  age	  groups.	  In	  the	  binary	  logistic	  regression	  model,	  condition	  (Repetition,	  Consensus),	  age	  (centred	  on	  sample	  mean	  of	  3.98),	  sex	  (female,	  male),	  and	  a	  condition	  by	  age	  interaction	  term	  were	  entered	  as	  model	  predictors	  for	  likelihood	  of	  choosing	  the	  novel-­‐dance	  Protagonist.	  	  An	  omnibus	  test	  of	  the	  model	  was	  significant	  (𝜒2	  (4)	  =	  10.977,	  p	  =	  .027),	  improving	  our	  ability	  to	  predict	  infants’	  Protagonist	  choices	  on	  3%	  of	  cases.	  Together,	  the	  coefficients	  explained	  approximately	  8.2%	  of	  the	  variance	  in	  target	  choice	  (Nagelkerke	  R2	  =	  .082).	  Logistic	  Regression	  coefficients	  and	  standard	  errors	  for	  each	  predictor	  variable	  are	  shown	  in	  Table	  3.1.	  	  Looking	  at	  individual	  predictors,	  analyses	  revealed	  that	  being	  in	  the	  Consensus	  condition	  predicted	  children	  being	  1.7	  times	  as	  likely	  to	  pick	  the	  novel-­‐dance	  Protagonist,	  compared	  to	  the	  Repetition	  condition	  (logistic	  regression	  coefficient	  =	  -­‐.546,	  p	  =	  .096,	  Odds	  Ratio	  =	  .579).	  Sex	  was	  a	  significant	  predictor,	  such	  that	  boys	  were	  nearly	  twice	  as	  likely	  to	  prefer	  the	  novel-­‐dance	  Protagonist	  as	  girls	  (logistic	  regression	  coefficient	  =	  -­‐.673,	  p	  =	  .042,	  Odds	  Ratio	  =	  .51)	  regardless	  of	  condition.	  Age	  alone	  was	  not	  a	  significant	  predictor;	  however,	  children’s	  likelihood	  of	  differentiating	  their	  choice	  by	  condition	  increased	  with	  age,	  indicating	  that	  for	  every	  1	  year	  increase	  in	  age,	  children	  were	  1.8	  times	  as	  likely	  to	  prefer	  the	  novel-­‐dance	  Protagonist	  in	  the	  Consensus	  condition	  as	  compared	  to	  the	  Repetition	  condition	  (or	  .57	  times	  as	  likely	  to	  prefer	  the	  same-­‐dance	  Protagonist;	  logistic	  regression	  coeff	  =	  -­‐.583,	  p	  =	  .055,	  Odds	  Ratio	  =	  .558).	  	  	  	  	   68	  Learning	  Overall,	  children	  were	  more	  willing	  to	  adopt	  the	  label	  for	  the	  unfamiliar	  object	  from	  the	  novel-­‐dance	  Protagonist	  in	  the	  Consensus	  condition	  (60.5%	  or	  49	  of	  81	  children,	  binomial	  probability	  test,	  p	  =	  .075,	  2-­‐tailed),	  but	  not	  in	  the	  Repetition	  condition	  (39	  of	  89,	  or	  43.8%,	  binomial	  probability	  test,	  p	  =	  .289,	  2-­‐tailed).	  Consistent	  with	  our	  results	  for	  liking,	  children	  appeared	  sensitive	  to	  the	  distribution	  of	  observed	  behaviours	  for	  making	  informant	  choices.	  In	  particular,	  children	  preferred	  to	  adopt	  the	  unfamiliar	  object	  label	  from	  a	  Smurf	  who	  performed	  a	  novel	  dance,	  after	  having	  seen	  a	  group	  of	  Smurfs	  first	  perform	  a	  shared	  dance.	  As	  with	  learning,	  children’s	  informant	  preference	  becomes	  increasingly	  pronounced	  with	  age.	  	  Effects	  of	  age	  Two-­‐,	  3-­‐,	  and	  4-­‐year	  old	  children	  were	  equally	  likely	  to	  learn	  from	  the	  novel-­‐dance	  Protagonist	  as	  the	  same-­‐dance	  Protagonist	  in	  both	  Consensus	  and	  Repetition	  conditions	  (see	  Table	  3	  for	  probabilities	  of	  learning	  from	  the	  novel-­‐dance	  Protagonist	  and	  associated	  binomial	  probability	  test	  p-­‐values	  at	  each	  age).	  Only	  5-­‐year-­‐olds	  made	  a	  significantly	  different	  choice	  of	  informant	  in	  the	  Consensus	  condition	  than	  from	  the	  Repetition	  condition,	  preferring	  to	  learn	  from	  the	  novel-­‐dance	  Protagonist	  in	  the	  Consensus	  condition	  (binomial	  probability	  test,	  p	  =	  .024,	  2-­‐tailed).	  The	  difference	  in	  choice	  patterns	  between	  Consensus	  and	  Repetition	  conditions	  was	  significant	  by	  a	  Fisher’s	  Exact	  Test	  (p	  =	  .005).	  	  	  	  	   69	  Magnitude	  estimates	  for	  learning	  Using	  the	  same	  analytic	  approach	  as	  for	  the	  liking	  measure,	  we	  conducted	  a	  binary	  logistic	  regression	  to	  examine	  the	  magnitude	  of	  difference	  in	  likelihood	  by	  age	  and	  by	  condition.	  Condition,	  age,	  sex,	  and	  age-­‐by-­‐condition	  interaction	  term	  were	  	  Table	  3.2	  Children’s	  probabilities	  of	  learning	  from	  the	  novel-­‐dance	  Protagonist,	  by	  age	  and	  by	  condition.	  	  	  	  	   Consensus	   Repetition	  difference	  between	  conditions	   	  	  Age	  group	   Pr(novel)	   p-­‐value*	   Pr(novel)	   p-­‐value*	   p-­‐value**	   n	  2	  to	  3	   0.53	   1	   0.54	   1	   1	   46	  3	  to	  4	   0.55	   1	   0.5	   1	   1	   40	  4	  to	  5	   0.58	   0.55	   0.41	   0.44	   0.27	   54	  5	  to	  6	   0.81	   0.024	   0.29	   0.14	   0.005	   34	  *Binomial	  probability	  test	  (2-­‐tailed)	  	  **Fisher's	  Exact	  Test	  	  	  Figure	  3.2	  Children’s	  probabilities	  of	  learning	  from	  the	  novel-­‐dance	  Protagonist,	  by	  age	  and	  by	  condition.	  	  0	  0.1	  0.2	  0.3	  0.4	  0.5	  0.6	  0.7	  0.8	  0.9	  2	  to	  3	   3	  to	  4	   4	  to	  5	   5	  to	  6	  Probability	  of	  choosing	  novel-­‐dance	  Protagonist	  	  Age	  group	  Consensus	  Repetition	  **	  	   70	  entered	  as	  model	  predictors	  for	  likelihood	  of	  choosing	  the	  novel-­‐dance	  Protagonist.	  An	  omnibus	  test	  of	  the	  model	  was	  significant	  (𝜒2(4)=10.997,	  p	  =	  .027),	  improving	  our	  ability	  to	  predict	  infants’	  informant	  choice	  on	  9.2%	  of	  cases.	  Together,	  the	  coefficients	  explain	  approximately	  8.1%	  of	  the	  variance	  in	  informant	  choice	  (Nagelkerke	  R2	  =	  .081).	  	  	  Turning	  to	  the	  individual	  predictors,	  children	  in	  the	  Consensus	  condition	  were	  nearly	  twice	  as	  likely	  to	  endorse	  the	  novel-­‐dance	  Protagonist’s	  label	  for	  the	  bottle	  cap	  as	  those	  in	  the	  Repetition	  condition	  (logistic	  regression	  coeff	  =	  -­‐.579,	  p	  =	  .065,	  OR	  =	  .56).	  Age	  was	  a	  marginally	  significant	  predictor,	  such	  that	  older	  children	  were	  1.45	  times	  more	  likely	  to	  prefer	  the	  same-­‐dance	  Informant	  (logistic	  regression	  coeff	  =	  .334,	  p	  =	  .076,	  OR	  =	  1.397).	  However,	  a	  significant	  Condition	  by	  Age	  interaction	  indicates	  that	  with	  every	  year	  increase	  in	  age,	  children	  in	  the	  Consensus	  condition	  were	  2.1	  times	  as	  likely	  to	  endorse	  the	  novel-­‐dance	  Protagonist’s	  label	  for	  the	  bottle	  cap,	  compared	  to	  the	  Repetition	  condition	  (logistic	  regression	  coeff	  =	  	  -­‐.783,	  p	  =	  .008,	  OR	  =	  .457).	  Unlike	  the	  preference	  measure,	  sex	  was	  not	  a	  significant	  covariate	  for	  which	  Protagonist’s	  label	  children	  endorsed.	  	  Table	  3.3	  Liking	  (Preference)	  and	  Learning	  predicted	  by	  age,	  sex,	  condition,	  and	  age-­‐by-­‐condition	  interaction	  term	  Predictors	   Preference	   Learning	  Age	  (centered)	   .068	  (.184)	   .372	  (.192)†	  	  Sex	   -­‐.664	  (.331)*	   .256	  (.323)	  Condition	   -­‐.560	  (.329)†	   -­‐.554	  (.314)†	  Age	  *	  Condition	  	   -­‐.563	  (.305)†	   -­‐.783	  (.293)**	  Observations	  (n)	   176	   174	  †p	  <	  .10,	  *p	  <	  .05,	  **p	  <	  .01,	  ***p	  <	  .001	  Logistic	  regression	  coefficients	  are	  the	  natural	  log	  (ln)	  of	  odd	  ratios	  for	  each	  predictor.	  Standard	  errors	  are	  presented	  in	  parentheses.	  	  	   71	  Liking	  predicts	  learning.	  	  Children's	  liking	  for	  a	  Protagonist	  significantly	  predicted	  whom	  they	  wanted	  to	  learn	  from.	  In	  a	  separate	  logistic	  regression	  model	  using	  Liking	  to	  predict	  Informant	  choice,	  children	  who	  reported	  liking	  a	  Protagonist	  were	  5	  times	  as	  likely	  to	  learn	  from	  that	  same	  Protagonist	  than	  were	  children	  who	  did	  not	  report	  the	  same	  preference	  (logistic	  regression	  coeff	  =	  1.605,	  p	  <	  .001;	  OR	  =	  4.98).	  In	  this	  model,	  Condition	  moderated	  by	  Age	  continues	  to	  be	  a	  significant	  predictor	  in	  this	  model	  (logistic	  regression	  coeff	  =	  -­‐.642,	  p	  =	  .039,	  OR	  =	  .526).	  That	  is,	  children	  increasingly	  differentiated	  their	  preferences	  across	  study	  conditions,	  showing	  a	  preference	  to	  learn	  from	  the	  novel-­‐dance	  Informant	  in	  the	  Consensus	  condition,	  and	  no	  clear	  preference	  in	  the	  Repetition	  condition.	  In	  this	  analysis,	  we	  removed	  Sex	  as	  a	  covariate,	  since	  it	  was	  a	  non-­‐significant	  predictor	  in	  the	  full	  model,	  and	  including	  it	  greatly	  hampers	  the	  model's	  predictions	  fit	  to	  the	  observed	  data.	  Hosmer	  and	  Lemeshow	  test	  indicate	  that	  the	  predicted	  data	  did	  not	  significantly	  differ	  from	  the	  observed	  data	  (𝜒2  (8)	  =	  5.594,	  p	  =	  .693),	  indicating	  good	  model	  fit.	  Together,	  preference	  (same-­‐dance	  Protagonist,	  novel-­‐dance	  Protagonist),	  condition,	  age,	  and	  an	  age-­‐by-­‐condition	  interaction	  term	  accounted	  for	  22.8%	  of	  variability	  in	  children’s	  Informant	  choices	  (Nagelkerke	  R2	  =	  .228)	  and	  also	  improved	  predictions	  of	  those	  choices	  on	  19%	  of	  cases.	  	  	  	  	  	  	  	  	  	  	   72	  Table	  3.4	  Learning	  predicted	  by	  Liking	  (Preference),	  Age,	  Condition,	  and	  Age	  x	  Condition	  interaction	  term	  Predictors	   Learning	  Age	  (centered)	   .339	  (.202)	  †	  Condition	   -­‐.441	  (.340)	  Age	  *	  Condition	   -­‐.642	  (.310)*	  Preference	  	   1.605	  (.362)***	  Observations	  (n)	   169	  †p	  <	  .10,	  *p	  <	  .05,	  **p	  <	  .01,	  ***p	  <	  .001	  Logistic	  regression	  coefficients	  are	  the	  natural	  log	  (ln)	  of	  odd	  ratios	  for	  each	  predictor.	  Standard	  errors	  are	  presented	  in	  parentheses.	  	  	  3.2.3	   Discussion	  	  In	  both	  preference	  and	  learning	  measures,	  children’s	  choices	  differed	  by	  age.	  The	  youngest	  tested	  groups	  (2	  and	  3	  year	  olds)	  did	  not	  differ	  in	  their	  choice	  of	  Protagonist	  across	  Consensus	  and	  Repetition	  conditions—it	  appears	  that	  they	  were	  insensitive	  to	  the	  distribution	  of	  information	  across	  individuals.	  In	  contrast,	  4	  and	  5	  year	  olds	  were	  influenced	  by	  behavioural	  consensus	  across	  individuals	  (they	  preferred	  the	  Protagonist	  who	  did	  a	  novel	  dance),	  but	  not	  repetitive	  actions	  by	  a	  single	  individual	  (in	  which	  they	  chose	  the	  2	  Protagonists	  equally);	  this	  effect	  is	  more	  pronounced	  in	  older	  children.	  The	  transitional	  age	  at	  which	  children	  in	  our	  sample	  differentiated	  between	  Consensus	  and	  Repetition	  conditions	  occurs	  around	  4	  years	  of	  age	  for	  preference,	  and	  a	  year	  later	  for	  Informant	  choice.	  Five-­‐year-­‐olds	  in	  our	  sample	  preferentially	  learned	  a	  novel	  object	  label	  from	  the	  novel-­‐dance	  Protagonist,	  but	  were	  equally	  likely	  to	  learn	  from	  the	  same-­‐dance	  Protagonist	  and	  novel-­‐dance	  Protagonist	  in	  the	  Repetition	  condition.	  	  The	  above	  preference	  results	  provided	  a	  conceptual	  replication	  of	  studies	  in	  Chapter	  2,	  where	  7-­‐month-­‐old	  infants	  preferred	  a	  novel-­‐dance	  Protagonist	  in	  the	  	   73	  Consensus,	  but	  not	  Repetition	  condition.	  While	  common	  group	  membership	  was	  addressed	  explicitly	  in	  the	  current	  study,	  it	  may	  have	  been	  implied	  in	  the	  infant	  studies	  by	  Protagonists’	  presence	  and	  absence	  during	  the	  Demonstrators’	  dances.	  Thus,	  there	  appear	  to	  be	  developmental	  continuity	  in	  preference	  for	  mavericks	  when	  there	  are	  opportunities	  to	  imitate	  the	  group	  dance,	  and	  when	  individuals’	  group	  identities	  are	  known.	  This	  may	  be	  due	  to	  an	  assumption	  that	  when	  everyone	  can	  imitate	  on	  the	  spot,	  or	  everyone	  is	  a	  member	  of	  the	  same	  group,	  that	  there	  is	  a	  common	  core	  of	  shared	  knowledge.	  However,	  it	  is	  also	  possible	  that	  2	  independent	  motivations	  drove	  infants’	  and	  children’s	  preferences	  for	  the	  maverick.	  	  Two	  and	  3-­‐year-­‐olds	  did	  not	  show	  differential	  preferences	  across	  conditions	  in	  the	  present	  study.	  One	  important	  difference	  between	  the	  infant	  and	  child	  studies	  is	  the	  nature	  of	  the	  dependent	  measures:	  while	  the	  infant	  studies	  required	  nonverbal,	  spontaneous	  responses	  through	  reaching,	  the	  children	  study	  used	  an	  elicited,	  verbal	  measure	  of	  preference.	  Such	  U-­‐shaped	  relationships	  between	  age	  and	  ability,	  in	  which	  young	  infants	  show	  an	  ability	  on	  implicit	  measures,	  and	  on	  which	  young	  children	  fail	  on	  explicit	  versions,	  only	  to	  succeed	  a	  couple	  of	  years	  later	  have	  been	  found	  in	  several	  other	  areas	  in	  social	  cognitive	  development	  (e.g.	  ToM;	  Baillargeon,	  Scott,	  &	  He,	  2010;	  Clements	  &	  Perner,	  1994;	  Onishi	  &	  Baillargeon,	  2005).	  It	  is	  plausible	  that	  a	  similar	  mechanism	  is	  in	  operation	  in	  this	  case,	  and	  that	  the	  change	  in	  methodology	  underlies	  the	  apparent	  disappearance	  of	  an	  earlier	  emerging	  ability,	  rather	  than	  true	  conceptual	  change.	  	  Children’s	  tendency	  to	  learn	  from	  the	  novel-­‐dance	  Protagonist	  in	  the	  Consensus	  condition	  appears	  inconsistent	  with	  previous	  findings	  that	  children	  trust	  	   74	  informants	  who	  were	  part	  of	  a	  consensus	  over	  a	  maverick	  (Corriveau	  et	  al.,	  2009).	  However,	  differences	  between	  the	  studies	  may	  account	  for	  this.	  One	  important	  design	  component	  of	  our	  study	  involved	  establishing	  consensus	  in	  one	  domain	  (dance),	  and	  followed	  by	  a	  learning	  task	  in	  a	  different	  domain	  (object	  labeling).	  Thus,	  children	  were	  initially	  introduced	  to	  the	  potential	  Informants	  in	  a	  context	  where	  learning	  was	  not	  a	  relevant	  objective.	  Children’s	  subsequent	  desire	  to	  learn	  from	  an	  Informant	  may	  be	  informed	  by	  positive	  feelings	  towards	  the	  individual	  formed	  during	  the	  dance	  phase,	  rather	  than	  a	  direct	  assessment	  of	  their	  skill	  in	  word	  labeling.	  	  The	  age	  patterns	  in	  our	  results	  provide	  some	  support	  for	  this	  interpretation:	  4	  year	  olds	  in	  our	  sample	  reliably	  showed	  a	  preference	  for	  a	  Protagonist,	  a	  full	  year	  before	  they	  as	  a	  group	  reliably	  learned	  from	  a	  Protagonist.	  The	  timing	  of	  these	  effects,	  together	  with	  the	  strong	  relationship	  between	  children’s	  expressed	  preference	  and	  their	  subsequent	  choice	  of	  Informant,	  suggests	  that	  children	  may	  first	  form	  a	  favorable	  impression	  of	  a	  Protagonist,	  which	  informs	  who	  they	  choose	  to	  learn	  from	  in	  a	  different	  context.	  Such	  willingness	  to	  generalize	  trust	  in	  an	  individual	  beyond	  the	  context	  in	  which	  favourable	  attitudes	  were	  formed	  may	  be	  one	  way	  in	  which	  children	  mistakenly	  acquire	  less-­‐than-­‐optimal	  knowledge,	  and	  yet	  is	  clearly	  an	  important	  aspect	  of	  cultural	  transmission.	  	  In	  our	  particular	  experimental	  set-­‐up,	  children	  were	  invited	  to	  play	  a	  game,	  and	  likely	  assessed	  it	  to	  be	  a	  situation	  in	  which	  uniqueness	  and	  self-­‐expression	  are	  acceptable	  responses.	  This	  could	  be	  possible	  because	  there	  were	  no	  obvious	  repercussions	  for	  learning	  from	  the	  “wrong”	  model.	  Future	  studies	  should	  	   75	  investigate	  the	  effect	  of	  incentives/stakes	  for	  learning	  outcomes	  to	  further	  clarify	  contextual	  features	  that	  promote	  learning	  from	  a	  consensus	  and	  ones	  that	  encourage	  learning	  from	  a	  maverick.	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	   76	  4.	  Infant	  social	  referencing	  differs	  by	  culture	  and	  by	  domain	  	  4.1	   Introduction	  Young	  humans	  possess	  an	  exceptional	  capacity	  to	  learn	  from	  others	  (Herrmann	  et	  al.,	  2007).	  One	  of	  the	  earliest	  means	  at	  their	  disposal	  for	  accessing	  knowledge	  in	  other	  people’s	  minds	  is	  sensitivity	  to	  emotional	  information.	  Emotions	  rapidly	  convey	  evaluations	  such	  as	  approval	  and	  disapproval,	  as	  well	  as	  appraisals	  of	  threat	  and	  safety.	  These	  evaluations	  can	  be	  spontaneous	  reactions	  without	  an	  intended	  recipient,	  or	  intentional	  communicative	  displays	  and/or	  vocalizations.	  Can	  preverbal	  infants	  exploit	  the	  rapid,	  nonverbal	  and	  universal	  nature	  of	  basic	  emotions	  for	  cultural	  learning?	  By	  the	  middle	  of	  their	  first	  year,	  preverbal	  infants	  can	  both	  accurately	  distinguish	  among	  some	  basic	  emotions,	  and	  show	  an	  incipient	  ability	  to	  use	  them	  appropriately	  to	  modify	  their	  own	  behaviours	  towards	  targets	  of	  emotional	  reactions.	  Infants	  are	  able	  to	  perceive	  and	  distinguish	  among	  anger,	  fear,	  and	  happiness	  by	  4	  months	  of	  age	  (Haviland	  &	  Lelwica,	  1987;	  LaBarbera	  et	  al.,	  1976;	  Montague	  &	  Walker-­‐Andrews,	  2001).	  By	  5	  months,	  they	  can	  discriminate	  among	  emotional	  vocalizations	  presented	  in	  infant-­‐directed	  speech,	  and	  adult-­‐directed	  speech	  when	  accompanied	  by	  a	  facial	  expression	  (Fernald,	  1993).	  Already,	  at	  5.5	  months,	  infants	  are	  able	  to	  regulate	  their	  behaviours	  towards	  unfamiliar	  objects	  based	  on	  other	  people’s	  fearful	  or	  happy	  reactions,	  when	  the	  reactions	  are	  communicated	  both	  visually	  and	  vocally	  (Vaillant-­‐Molina	  &	  Bahrick,	  2012).	  	   77	  From	  the	  second	  half	  of	  the	  first	  year	  and	  through	  the	  second	  year,	  infants	  become	  competent	  users	  of	  emotional	  information	  for	  learning	  about	  the	  physical	  world.	  Between	  10	  and	  12	  months	  of	  age,	  infants	  begin	  to	  reliably	  moderate	  their	  behaviours	  toward	  novel	  objects	  or	  environments	  in	  response	  to	  emotional	  reactions,	  preferring	  to	  approach	  objects	  associated	  with	  a	  happy	  reaction,	  and	  avoiding	  objects	  associated	  with	  a	  fear	  reaction	  (Mumme	  et	  al.,	  1996;	  Sorce	  et	  al.,	  1985;	  Walden	  &	  Ogan,	  1988).	  By	  18	  months	  of	  age,	  infants	  can	  use	  a	  Bystander’s	  neutral	  vs.	  angry	  reaction	  toward	  objects	  to	  make	  inferences	  about	  his/her	  enduring	  object	  preferences	  (Repacholi	  &	  Meltzoff,	  2007):	  they	  are	  less	  likely	  to	  play	  with	  an	  object	  toward	  which	  an	  Bystander	  has	  previously	  expressed	  anger	  rather	  than	  a	  neutral	  expression.	  Critically,	  18-­‐month-­‐olds’	  likelihood	  to	  play	  with	  an	  object	  that	  is	  the	  target	  of	  anger	  depends	  on	  whether	  the	  original	  Bystander	  is	  present	  during	  infants’	  exploration	  of	  the	  object,	  suggestive	  that	  their	  attribution	  is	  person-­‐specific.	  That	  said,	  infants	  also	  appear	  capable	  of	  using	  additional	  cues	  present	  in	  emotional	  displays	  to	  infer	  the	  content	  of	  shared	  attitudes:	  Eighteen-­‐month-­‐olds	  infer	  shared	  attitudes	  (held	  by	  individuals	  other	  than	  the	  original	  Bystander)	  when	  emotional	  reactions	  are	  preceded	  by	  infant-­‐directed	  eye	  contact	  (an	  ostensive	  signal	  of	  intentional	  teaching),	  but	  infer	  person-­‐specific	  attitudes	  in	  the	  absence	  of	  eye	  contact	  (Egyed	  et	  al.,	  2013).	  Together,	  these	  studies	  show	  a	  sophisticated	  use	  of	  emotional	  information	  in	  which	  infants	  differentiate	  between	  individual	  and	  shared	  attitudes	  towards	  unfamiliar	  aspects	  of	  the	  physical	  world.	  	  Although	  relatively	  large	  quantities	  of	  evidence	  have	  accumulated	  for	  infants’	  ability	  to	  use	  emotions	  to	  learn	  about	  unfamiliar	  physical	  objects	  or	  environments,	  	   78	  considerably	  less	  has	  been	  done	  to	  investigate	  whether	  infants	  use	  others’	  emotions	  to	  also	  learn	  about	  the	  social	  world.	  Indeed,	  the	  social	  world	  poses	  its	  own	  set	  of	  opportunities	  and	  threats	  to	  naïve	  learners;	  individuals	  vary	  in	  their	  level	  of	  benevolence	  and	  cooperativeness	  during	  social	  interactions,	  as	  well	  as	  in	  their	  level	  of	  skill	  and	  ability	  in	  navigating	  the	  environment.	  Although	  identifying	  cooperative	  and	  knowledgeable	  individuals	  to	  associate	  with	  and	  learn	  from	  may	  be	  critical	  for	  optimal	  development,	  it	  is	  not	  necessarily	  trivial	  to	  understand	  who	  these	  individuals	  are,	  and	  the	  risks	  of	  misidentification	  may	  be	  severe.	  Therefore,	  it	  would	  greatly	  benefit	  young	  humans	  to	  be	  able	  to	  learn	  the	  value	  of	  potential	  interaction	  partners	  by	  looking	  to	  the	  reactions	  of	  other,	  more	  experienced	  individuals	  in	  their	  environment,	  and	  to	  use	  those	  reactions	  to	  infer	  whom	  should	  be	  approached	  and	  whom	  avoided.	  By	  engaging	  in	  this	  social	  referencing	  about	  social	  targets,	  Bystanders	  can	  significantly	  increase	  the	  likelihood	  of	  associating	  with	  benevolent	  and	  skilled	  individuals.	  	  Some	  evidence	  exists	  suggesting	  that	  infants	  can	  use	  emotional	  information	  to	  learn	  about	  social	  targets.	  Ten-­‐month-­‐old	  infants	  are	  more	  likely	  to	  approach	  and	  offer	  toys	  to	  a	  stranger	  after	  their	  mothers	  speak	  to	  them	  about	  the	  stranger	  in	  a	  positive	  tone	  compared	  to	  when	  mothers	  spoke	  in	  a	  neutral	  tone	  (Feinman	  &	  Lewis,	  1983).	  However,	  these	  results	  do	  not	  hold	  if	  the	  infants	  merely	  observe	  mothers	  speak	  in	  a	  positive	  tone	  (vs.	  a	  neutral	  tone)	  to	  the	  stranger,	  suggestive	  that	  at	  10	  months	  the	  capacity	  to	  engage	  in	  ‘social’	  social	  referencing	  may	  be	  weak.	  In	  another	  study,	  15-­‐month-­‐old	  infants	  smiled	  more	  and	  were	  also	  more	  willing	  to	  accept	  toys	  from	  a	  stranger	  after	  observing	  their	  mothers	  interact	  positively	  with	  her,	  compared	  	   79	  to	  when	  they	  had	  no	  interaction	  (Feiring,	  Lewis,	  &	  Starr,	  1984).	  However,	  while	  its	  use	  of	  naturalistic	  manipulation	  of	  positive	  vs.	  neutral	  interaction	  strengthens	  ecological	  validity,	  the	  latter	  study	  did	  not	  control	  for	  factors	  outside	  of	  emotions,	  such	  as	  physical	  distance	  or	  amount	  of	  eye	  contact,	  which	  could	  have	  affected	  infants’	  willingness	  to	  interact	  with	  the	  stranger.	  Furthermore,	  each	  of	  the	  studies	  described	  above	  focused	  on	  comparing	  the	  effect	  of	  a	  positive	  vs.	  a	  neutral	  (or	  lack	  of)	  interaction;	  neither	  examined	  how	  infants	  respond	  to	  observing	  a	  Bystander	  display	  a	  negative	  emotion	  toward	  a	  social	  target.	  To	  our	  knowledge,	  only	  one	  study	  has	  been	  conducted	  explicitly	  examining	  12-­‐month-­‐old	  infants’	  social	  referencing	  for	  a	  social	  target,	  by	  examining	  the	  impact	  of	  maternal	  emotions	  on	  infants’	  subsequent	  behaviour.	  The	  study	  found	  that	  infants	  who	  observed	  their	  mothers	  displaying	  behavioural	  cues	  of	  anxiety	  when	  interacting	  with	  a	  stranger	  were	  significantly	  less	  likely	  to	  approach	  the	  stranger	  themselves	  than	  infants	  whose	  mothers	  exhibited	  no	  anxious	  symptoms	  (de	  Rosnay,	  Cooper,	  Tsigaras,	  &	  Murray,	  2006).	  None	  of	  the	  above	  studies	  just	  described	  compared	  the	  effects	  of	  both	  positive	  and	  negative	  emotions	  on	  infants’	  behaviour	  within	  the	  same	  study.	  While	  each	  study	  had	  a	  control	  comparison	  to	  neutral	  or	  absence	  of	  interaction,	  this	  does	  not	  rule	  out	  the	  explanation	  that	  physiological	  arousal	  from	  experiencing	  any	  emotion	  could	  have	  caused	  subsequent	  changes	  in	  behaviour.	  Another	  important	  issue	  pertains	  to	  the	  studies’	  inability	  to	  speak	  to	  the	  mechanism	  by	  which	  mothers’	  emotions	  influenced	  infants’	  behaviour:	  were	  infants	  more	  inhibited	  around	  the	  stranger	  due	  to	  emotional	  contagion	  from	  their	  mothers,	  or	  did	  infants	  understand	  	   80	  the	  informational	  content	  of	  emotions,	  without	  personal	  experience	  of	  emotions	  being	  necessary?	  The	  latter	  mechanism	  would	  provide	  more	  support	  for	  emotional	  understanding	  as	  a	  means	  for	  cultural	  learning,	  as	  it	  permits	  the	  learner	  to	  hold	  learned	  attitudes	  even	  when	  the	  original	  emoter	  is	  absent,	  thus	  generalizing	  beyond	  the	  learning	  context.	  	  4.2	  	   The	  present	  studies	  The	  current	  studies	  set	  out	  to	  examine	  infants’	  ability	  to	  use	  social	  referencing	  to	  learn	  about	  social	  targets.	  In	  particular,	  it	  aimed	  to	  address	  4	  goals.	  Our	  first	  goal	  was	  to	  examine	  the	  effect	  of	  seeing	  observers	  emote	  both	  positively	  and	  negatively	  about	  distinct	  social	  targets,	  to	  rule	  out	  physiological	  changes	  associated	  with	  general	  arousal	  (rather	  than	  a	  particular	  discrete	  emotion	  per	  se)	  as	  responsible	  for	  infants’	  tendency	  to	  avoid	  targets	  of	  their	  mothers’	  negative	  emotions	  in	  past	  studies.	  Two	  of	  the	  earliest	  distinguishable	  and	  recognized	  emotional	  expressions	  are	  positively	  valenced	  expressions	  (smiles)	  corresponding	  to	  happiness	  and	  negatively	  valenced	  expressions	  (frowns)	  corresponding	  to	  anger	  (LaBarbera	  et	  al.,	  1976;	  Montague	  &	  Walker-­‐Andrews,	  2001).	  We	  hypothesized	  that,	  by	  attending	  to	  positively	  and	  negatively	  valenced	  emotional	  reactions,	  preverbal	  infants	  could	  accurately	  use	  emotional	  information	  referentially	  and	  moderate	  their	  own	  behaviours	  toward	  the	  target	  of	  these	  emotions.	  These	  emotions	  might	  provide	  crude,	  but	  adequate	  evaluations	  of	  people	  and	  object	  targets.	  We	  expected	  infants	  to	  prefer	  targets	  of	  positive	  emotional	  reactions	  to	  targets	  of	  negative	  emotional	  reactions.	  	  	   81	  Second,	  we	  wanted	  to	  investigate	  whether	  infants	  use	  social	  referencing	  in	  similar	  ways	  for	  social	  and	  asocial	  targets.	  As	  summarized	  above,	  many	  studies	  show	  sensitivity	  to	  emotional	  information	  for	  objects	  targets	  by	  10	  months.	  Given	  the	  paucity	  of	  studies	  on	  social	  targets,	  infants’	  ability	  to	  use	  others’	  emotional	  displays	  to	  regulate	  their	  behavior	  toward	  social	  targets	  is	  less	  well	  understood.	  Given	  that	  social	  targets	  operate	  via	  invisible	  mental	  states	  that	  cannot	  be	  directly	  assessed,	  that	  they	  can	  self-­‐locomote	  and	  carry	  out	  behaviours	  according	  to	  invisible	  intentions,	  it	  may	  be	  that	  it	  is	  more	  difficult	  to	  reliably	  link	  others’	  emotional	  displays	  to	  social	  than	  to	  asocial	  targets.	  On	  the	  other	  hand,	  compared	  to	  object	  targets,	  social	  agents	  can	  potentially	  cause	  greater	  harm	  to	  a	  vulnerable	  infant,	  and	  are	  important	  potential	  sources	  of	  cultural	  information.	  The	  potential	  for	  both	  benefit	  and	  harm	  from	  unknown	  social	  others	  makes	  it	  critical	  that	  young	  humans	  can	  reliably	  discern	  their	  value.	  Therefore,	  one	  might	  expect	  infants	  to	  be	  more	  sensitive	  to	  evaluative	  emotional	  information	  about	  social	  versus	  asocial	  targets.	  	  Third,	  we	  intended	  to	  investigate	  whether	  infants	  understand	  emotions	  as	  referential,	  or	  whether	  their	  modulation	  of	  behaviours	  is	  a	  result	  of	  mood	  contagion.	  	  As	  discussed	  above,	  observed	  emotions	  may	  exert	  influence	  on	  Bystanders	  by	  reproducing	  the	  same	  emotional	  arousal	  and	  subsequent	  action	  tendencies	  in	  them,	  as	  in	  the	  original	  emoter.	  Alternatively,	  the	  Bystander	  could	  perceive	  and	  interpret	  the	  implications	  of	  observed	  emotions	  without	  personally	  experiencing	  the	  physiological	  arousal	  or	  behavioural	  and	  cognitive	  effects.	  What	  is	  the	  nature	  of	  the	  link	  between	  observed	  emotions	  and	  infants’	  behaviours?	  	   82	  Finally,	  we	  were	  interested	  in	  whether	  infants	  being	  raised	  in	  different	  cultural	  milieus	  would	  exhibit	  differences	  in	  their	  understanding	  and	  use	  of	  emotional	  information.	  On	  the	  one	  hand,	  universal	  basic	  emotions	  may	  emerge	  spontaneously,	  independent	  of	  cultural	  input,	  and	  serve	  as	  a	  mechanism	  by	  which	  events	  are	  subsequently	  interpreted	  and	  learned.	  On	  the	  other	  hand,	  sociocultural	  theories	  argue	  that	  differences	  in	  display	  rules,	  emotional	  expressivity	  and	  recognition	  of	  emotions	  have	  been	  noted	  in	  adults	  across	  cultures,	  so	  culture	  must	  play	  a	  role	  in	  constructing	  individuals’	  understanding	  of	  emotion	  concepts.	  Group	  differences	  in	  the	  emergence	  of	  basic	  cognitive	  abilities	  resulting	  from	  different	  experience	  are	  not	  uncommon.	  For	  instance,	  Salomo	  and	  Liszkowski	  (2013)	  documented	  earlier	  pointing	  in	  triadic	  (joint	  attention)	  situations	  in	  Chinese	  infants	  compared	  to	  Dutch	  and	  Mayan	  infants,	  corresponding	  to	  the	  amount	  of	  pointing	  that	  caregivers	  engaged	  in	  with	  their	  infants.	  Infants’	  rate	  of	  word	  acquisition	  also	  varies	  by	  SES,	  differing	  by	  parents’	  use	  of	  gesture	  (Rowe	  &	  Goldin-­‐Meadow,	  2009),	  and	  by	  the	  amount	  of	  time	  spent	  in	  joint	  attention	  (Tomasello	  &	  Farrar,	  1986).	  It	  remains	  an	  open	  question,	  however,	  how	  early	  cultural	  learning	  starts	  producing	  these	  differences	  in	  even	  basic	  emotional	  understanding.	  Indeed,	  despite	  the	  early	  emergence	  of	  emotional	  understanding	  documented	  in	  North	  American	  infants,	  the	  amount	  of	  experience	  infants	  receive	  regarding	  emotion-­‐laden	  exchanges	  likely	  varies	  across	  cultures.	  Researchers	  have	  documented	  differences	  in	  how	  much	  face-­‐to-­‐face	  vs.	  side-­‐by-­‐side	  experience	  infants	  have	  with	  their	  mothers	  as	  a	  product	  of	  the	  socialization	  values	  of	  autonomy	  vs.	  relatedness	  (Keller	  et	  al.,	  2007).	  Cultures	  that	  value	  autonomy	  tend	  to	  structure	  	   83	  interactions	  with	  infants	  in	  a	  physically	  autonomous	  arrangement,	  involving	  more	  face	  to	  face,	  and	  at-­‐a-­‐distance	  interactions.	  In	  contrast,	  cultures	  that	  value	  relatedness	  tend	  to	  structure	  interactions	  that	  produce	  more	  physical	  proximity,	  involving	  more	  touching	  and	  holding.	  These	  physical	  behaviours	  may	  have	  implications	  for	  the	  amount	  of	  experience	  infants	  have	  receiving	  emotional	  communications,	  which	  provide	  information	  efficiently	  from	  a	  distance.	  For	  example,	  a	  parent	  who	  is	  physically	  holding	  an	  infant	  can	  keep	  him	  or	  her	  safe	  by	  staying	  away	  from	  potential	  hazards,	  or	  moving	  away	  from	  it.	  However,	  when	  separated	  by	  physical	  distance,	  a	  parent	  must	  communicate	  prohibition	  or	  warning	  in	  another	  way,	  by	  tone	  of	  voice	  or	  facial	  expression.	  It	  follows	  that	  infants	  socialized	  in	  a	  culture	  that	  values	  autonomy	  might	  have	  more	  practice	  and	  must	  begin	  to	  rely	  on	  emotional	  communications	  sooner	  and	  to	  a	  greater	  extent	  than	  an	  infant	  socialized	  in	  a	  culture	  of	  relatedness.	  	  To	  examine	  these	  questions,	  we	  used	  a	  choice	  method	  to	  measure	  infants’	  preference	  for	  targets	  of	  positive	  vs.	  negative	  emotional	  reactions.	  Across	  several	  studies,	  infants	  observed	  the	  same	  6	  individuals:	  4	  Bystanders	  and	  2	  Protagonists,	  and	  the	  same	  two	  objects.	  In	  each	  event,	  the	  Bystanders	  observed	  one	  Protagonist	  move	  toward	  and	  pick	  up	  one	  of	  the	  two	  objects,	  and	  the	  Bystanders	  responded	  either	  positively	  or	  negatively	  (smiled	  with	  raised	  eyebrows	  or	  frowned	  with	  lowered	  eyebrows).	  All	  infants	  observed	  a	  series	  of	  4	  events,	  in	  which	  each	  Protagonist	  picked	  up	  each	  object	  once.	  Across	  4	  between-­‐subjects	  conditions	  (described	  in	  detail	  below),	  Bystanders	  directed	  their	  positive	  and	  negative	  emotions	  either	  consistently	  toward	  single	  social	  targets	  (smiling	  whenever	  	   84	  Protagonist	  A	  chose	  either	  object;	  frowning	  whenever	  Protagonist	  B	  chose	  either	  object)	  or	  consistently	  toward	  single	  objects	  (smiling	  whenever	  either	  Protagonist	  chose	  object	  A	  and	  frowning	  whenever	  either	  Protagonist	  chose	  object	  B).	  We	  reasoned	  that	  if	  infants	  can	  engage	  in	  both	  social	  and	  asocial	  social	  referencing,	  they	  should	  prefer	  the	  targets	  of	  positive	  emotions	  in	  all	  conditions,	  choosing	  Protagonist	  A	  over	  Protagonist	  B	  and	  object	  A	  over	  object	  B.	  	  In	  order	  to	  control	  the	  magnitude	  and	  consistency	  of	  information	  across	  events	  and	  conditions,	  stimuli	  were	  animated	  videos	  in	  which	  Bystanders	  and	  Protagonists	  consisted	  of	  different	  coloured	  geometrical	  shapes	  and	  animated	  eyes.	  The	  use	  of	  similar	  simple	  animations	  has	  been	  used	  successfully	  in	  past	  work	  examining	  infants’	  understanding	  of	  social	  others	  (Csibra,	  Bıró,	  Koós,	  &	  Gergely,	  2003;	  Geraci	  &	  Surian,	  2011;	  Kuhlmeier,	  Bloom,	  &	  Wynn,	  2004),	  and,	  given	  that	  subtle	  differences	  in	  expressive	  style	  between	  members	  of	  different	  cultural	  groups	  may	  make	  it	  more	  difficult	  to	  decode	  expressions	  by	  cultural	  out-­‐group	  members	  (e.g.,	  Elfenbein	  &	  Ambady,	  2002;	  Elfenbein,	  Mandal,	  &	  Ambady,	  2004;	  Kilbride	  &	  Yarczower,	  1983;	  Markham	  &	  Wang,	  1996),	  reduced	  the	  possibility	  that	  an	  in-­‐group	  recognition	  bias	  would	  influence	  results.	  Specifically,	  by	  using	  models	  without	  a	  particular	  ethnicity,	  animated	  emotions	  should	  be	  equally	  (un)familiar	  to	  all	  infants.	  	  4.3	  	   Study	  1a	  –	  social	  targets	  4.3.1	   Methods	  	  Sixty-­‐five	  6-­‐month-­‐olds	  (48.5%	  female;	  M	  =	  6.14,	  SD	  =	  .307,	  range	  =	  5.33	  –	  6.9)	  and	  fifty-­‐eight	  12-­‐month-­‐olds	  (49%	  female;	  M=11.95,	  SD=.308,	  range	  =	  11.26	  –	  12.63)	  participated.	  We	  chose	  to	  study	  both	  6-­‐	  and	  12-­‐month-­‐olds	  because	  5.5	  	   85	  months	  is	  the	  earliest	  social	  referencing	  toward	  objects	  that	  has	  been	  demonstrated	  (Vaillant-­‐Molina	  &	  Bahrick,	  2012)	  and	  12	  months	  is	  the	  average	  age	  at	  which	  infants	  can	  reliably	  use	  social	  referencing.	  	  	  Because	  we	  were	  interested	  in	  whether	  there	  are	  any	  cultural	  differences	  in	  infants’	  tendency	  to	  socially	  reference,	  parents	  were	  asked	  to	  report	  their	  infants’	  ethnicity	  and	  linguistic	  exposure	  on	  a	  provided	  demographic	  questionnaire.	  For	  ethnicity,	  parents	  were	  asked	  to	  check/list	  all	  that	  applied	  to	  their	  child:	  36	  infants	  were	  identified	  as	  Western	  European	  (eighteen	  6-­‐month-­‐olds,	  eighteen	  12-­‐month-­‐olds),	  36	  as	  East	  Asian	  (Chinese,	  Japanese	  and	  Korean;	  eighteen	  6-­‐month-­‐olds,	  eighteen	  12-­‐month-­‐olds),	  28	  of	  mixed	  East	  Asian	  and	  European	  ethnicities	  (thirteen	  6-­‐month-­‐olds,	  fifteen	  12-­‐month-­‐olds),	  and	  23	  of	  other	  ethnicities	  (including	  Caribbean,	  Fijian,	  Ukrainian,	  Turkish,	  Thai,	  Vietnamese	  and	  Filipino	  Canadian).	  For	  linguistic	  exposure,	  parents	  were	  asked	  to	  list	  all	  languages	  that	  their	  child	  hears,	  and	  to	  provide	  an	  estimated	  percentage	  of	  time	  that	  each	  is	  heard	  in	  their	  infants’	  daily	  life.	  We	  reasoned	  that	  because	  we	  were	  completing	  our	  research	  in	  the	  multi-­‐cultural	  city	  with	  several	  generations	  of	  immigrant	  groups	  represented,	  language	  exposure	  may	  provide	  a	  more	  fine-­‐grained,	  continuous	  measure	  of	  variability	  in	  acculturation	  than	  does	  reported	  ethnicity.	  	  Data	  from	  26	  additional	  infants	  were	  discarded	  because	  of	  fussiness	  (n=10),	  no	  choice	  (n	  =	  4),	  parental	  interference	  (n=5),	  procedural	  error	  (n=6),	  and	  equipment	  malfunction	  (n=1).	  Furthermore,	  the	  sample	  size	  of	  infants	  who	  were	  not	  Western	  European	  and/or	  East	  Asian	  (n=23)	  was	  too	  small	  to	  inform	  conclusions	  about	  how	  infants	  from	  those	  ethnic	  backgrounds	  behaved.	  Therefore	  they	  were	  	   86	  excluded	  from	  the	  analyses	  reported	  below.	  In	  total,	  100	  infants	  were	  used	  for	  analyses.	  Infants	  were	  recruited	  through	  a	  database	  maintained	  by	  the	  University	  of	  British	  Columbia	  Early	  Development	  Research	  Group.	  Participating	  families	  were	  given	  a	  token	  gift	  for	  their	  participation.	  	  Infants	  sat	  in	  a	  square	  testing	  room	  on	  their	  parent’s	  lap,	  approximately	  1.5	  metres	  from	  a	  50-­‐inch	  screen	  TV;	  the	  room	  was	  otherwise	  unadorned.	  Parents	  were	  instructed	  to	  refrain	  from	  speaking	  or	  directing	  their	  infants’	  attention	  in	  any	  way.	  	  Introductory	  trials.	  In	  all	  conditions,	  infants	  were	  first	  shown	  a	  36	  second	  familiarization	  video	  in	  which	  6	  animated	  characters	  of	  different	  colours	  and	  shapes	  with	  eyes	  were	  introduced	  moving	  around	  a	  screen	  to	  background	  music.	  Infants’	  on-­‐screen	  and	  off-­‐screen	  looking	  time	  during	  the	  video	  was	  measured	  using	  an	  eye-­‐gaze	  coding	  program	  (jHab).	  Infants	  who	  reached	  a	  minimum	  looking	  criterion	  of	  75%	  of	  the	  entire	  trial	  duration	  proceeded	  to	  habituation	  trials;	  otherwise	  the	  introductory	  trial	  was	  repeated	  until	  the	  criterion	  was	  reached.	  	  Habituation	  trials.	  Infants	  watched	  videos	  of	  two	  Protagonists	  (a	  blue	  square	  and	  a	  yellow	  triangle	  with	  eyes)	  each	  alternately	  approaching	  2	  objects	  (a	  bottle	  and	  a	  shell)	  for	  a	  total	  of	  4	  unique	  videos.	  At	  the	  beginning	  of	  each	  video,	  four	  Bystanders	  are	  arranged	  across	  the	  top	  of	  the	  screen	  on	  a	  black	  and	  white	  checkered	  background,	  and	  two	  Protagonists	  sit	  below	  the	  Bystander	  ‘gallery’,	  on	  either	  side	  in	  the	  centre	  of	  the	  screen.	  The	  2	  objects	  sit	  at	  the	  bottom	  of	  the	  screen.	  One	  of	  the	  two	  Protagonists	  moves	  to	  the	  centre	  of	  the	  screen,	  and	  appears	  to	  deliberate	  between	  the	  two	  objects:	  It	  first	  briefly	  approaches	  one	  object,	  and	  then	  returns	  to	  the	  middle	  of	  the	  screen,	  and	  then	  gazes	  toward	  and	  briefly	  approaches	  	   87	  the	  second	  object,	  and	  then	  returns	  to	  the	  centre	  again.	  From	  the	  gallery	  above,	  the	  Bystanders’	  eyes	  move	  so	  that	  they	  are	  following	  along	  with	  the	  Protagonist’s	  movements;	  to	  ensure	  that	  infants	  see	  the	  Bystanders’	  eye	  movements,	  the	  checkerboard	  flickers	  to	  attract	  their	  attention.	  After	  the	  Protagonist	  has	  checked	  out	  each	  object	  in	  turn,	  it	  then	  gazes	  at	  and	  moves	  to	  contact	  the	  object	  it	  first	  approached,	  moving	  the	  object	  a	  few	  inches	  toward	  the	  side	  of	  the	  display	  and	  pausing,	  as	  though	  choosing	  that	  object	  rather	  than	  the	  other	  one.	  Following	  the	  Protagonist’s	  choice,	  the	  Bystanders	  all	  simultaneously	  express	  either	  a	  positive	  expression	  (smiling)	  or	  a	  negative	  expression	  (frowning)	  while	  continually	  directing	  their	  gaze	  towards	  the	  Protagonist.	  Smiling	  is	  depicted	  dynamically	  with	  a	  short	  horizontal	  line	  transforming	  into	  an	  upturned	  mouth	  and	  eyebrows	  that	  turned	  up	  at	  the	  inner	  corners	  by	  15	  degrees.	  	  Frowning	  is	  depicted	  using	  identical	  but	  opposite	  movements:	  mouths	  widened	  and	  turned	  downward,	  and	  the	  inner	  corners	  of	  the	  eyebrows	  turned	  down	  by	  15	  degrees.	  Looking	  time	  coding	  for	  each	  video	  began	  after	  the	  Protagonist’s	  object	  choice	  was	  made	  and	  the	  Bystanders	  emoted,	  and	  continued	  until	  the	  infant	  looked	  away	  for	  2	  consecutive	  seconds,	  or	  after	  30	  seconds	  elapsed.	  	  In	  Study	  1a,	  one	  Protagonist	  was	  always	  smiled	  at,	  regardless	  of	  which	  object	  it	   picked;	   the	   other	  was	   always	   frowned	   at.	   For	   example,	   one	   infant	  may	   see	   the	  yellow	  triangle	  Protagonist	  smiled	  at	  both	  after	  picking	  the	  shell	  in	  trial	  1,	  and	  after	  picking	   the	   glue	   bottle	   in	   trial	   2;	   the	   same	   infant	   would	   see	   the	   blue	   square	  Protagonist	   always	   frowned	   at,	   both	   after	   picking	   the	   shell	   in	   trial	   1,	   and	   after	  picking	  the	  glue	  bottle	   in	  trial	  2.	  This	  design	  was	  utilized	  to	   imply	  that	  Bystanders	  	   88	  differentially	   valued	   the	   Protagonists,	   but	   not	   the	   objects	   the	   picked.	  Bystanders’	  differential	  valuations	  were	  depicted	  across	  4	  different	  videos,	  for	  each	  combination	  of	  Protagonist	  and	  object	  (2	  positive	  and	  2	  negative	  reactions).	  Infants	  were	  shown	  the	   stimuli	   repeatedly	   until	   they	   habituated,	   or	   until	   they	   viewed	   3	   complete	  iterations	   of	   the	   4	   videos	   for	   a	   total	   of	   12	   trials.	   The	   habituation	   criterion	   was	  established	  after	  the	  first	  4	  trials:	  Infants	  were	  considered	  habituated	  when	  looking	  time	  during	  their	  last	  4	  trials	  were	  less	  than	  half	  of	  that	  of	  their	  first	  4	  trials,	  which	  could	   happen	   for	   the	   first	   time	   on	   trial	   8.	   Protagonist	   receiving	   the	   positive	  evaluation	   (yellow	   triangle,	   blue	   square),	   Protagonist	   side	   (left,	   right),	   order	  (positive	   first,	   negative	   first),	   and	   Bystander	   expression	   valence	   were	  counterbalanced	  across	  participants	  (across	  16	  different	  script	  versions).	  	  	  Figure	   4.1	  Video	   stimuli	   for	   Study	   1:	   Bystanders	   reacting	   positively	   to	   the	   blue	  square	  picking	  the	  shell	  (left).	  Bystanders	  reacting	  negatively	  to	  the	  yellow	  triangle	  picking	   the	   shell	   (right).	   Not	   depicted	   are	   2	   additional	   videos	   depicting	   the	   blue	  square	   and	   yellow	   triangle	   picking	   the	   glue	   bottle,	   and	   again	   eliciting	   different	  reactions	  from	  the	  Bystanders	  (positive	  for	  the	  blue	  square,	  negative	  for	  the	  yellow	  triangle).	   Together,	   the	   4	   videos	   show	   2	   different	   contexts	   in	   which	   Bystanders	  differentially	  evaluate	  the	  Protagonists	  for	  taking	  identical	  actions.	  	  	  	   89	  Choice	  measure.	  After	  the	  habituation	  criterion	  was	  reached,	  infants	  were	  presented	  with	  a	  choice	  measure.	  Parents	  were	  instructed	  to	  turn	  their	  chair	  90	  degrees	  to	  the	  right,	  so	  that	  they	  were	  no	  longer	  facing	  the	  TV	  screen,	  and	  to	  close	  their	  eyes.	  An	  Experimenter	  who	  was	  blind	  to	  the	  infant’s	  assigned	  condition	  presented	  a	  white	  board	  with	  the	  choices	  presented	  25	  cm	  apart	  (foam	  replicas	  of	  Protagonists	  shown	  in	  stimuli	  videos),	  initially	  out	  of	  the	  infant’s	  reach.	  Infants	  were	  required	  to	  have	  looked	  at	  both	  possible	  choices	  before	  the	  choice	  board	  was	  moved	  within	  reach.	  The	  Experimenter	  recorded	  the	  first	  Protagonist	  that	  the	  infant	  intentionally	  touched,	  defined	  as	  visually	  guided	  contact	  with	  a	  Protagonist.	  The	  side	  of	  the	  positively	  and	  negatively	  evaluated	  Protagonist	  was	  counterbalanced	  during	  choice.	  	  	  4.3.2	   Results	  As	  a	  group,	  infants	  did	  not	  appear	  to	  exhibit	  a	  preference	  for	  the	  targets	  of	  positive	  emotional	  reactions	  (57	  of	  100;	  binomial	  probability	  =	  .194),	  either	  at	  12	  months	  alone	  	  (29	  of	  51;	  binomial	  probability	  test,	  p	  =	  .4)	  or	  at	  6	  months	  alone	  (28	  of	  49;	  binomial	  probability	  test,	  p	  =	  .39).	  However,	  when	  broken	  down	  by	  the	  ethnicity	  of	  the	  infants,	  Protagonist	  choices	  were	  significantly	  influenced	  by	  bystanders’	  emotional	  reactions	  in	  a	  subset	  of	  the	  infants.	  Twelve-­‐month-­‐olds.	  	  Infants’	  choices	  were	  significantly	  influenced	  by	  bystanders’	  emotional	  reactions.	  Supporting	  our	  predictions,	  15	  of	  18	  Euro-­‐Canadian	  infants	  chose	  the	  Protagonist	  whose	  actions	  were	  followed	  by	  positive	  bystander	  reactions	  (binomial	  probability	  test,	  p	  =	  .01,	  2-­‐tailed).	  In	  contrast,	  East	  Asian	  infants	  did	  not	  show	  a	  clear	  preference	  for	  either	  Protagonist	  (8	  of	  18	  chose	  	   90	  the	  positively	  evaluated	  Protagonist,	  binomial	  probability	  test,	  p	  =	  .81,	  2-­‐tailed).	  The	  difference	  between	  European	  and	  East	  Asian	  infant	  choices	  is	  significant	  by	  a	  Fisher’s	  Exact	  Test	  (p	  <	  .035,	  2-­‐tailed).	  Similarly,	  Mixed	  infants	  did	  not	  reliably	  prefer	  targets	  of	  either	  emotional	  reaction	  (6	  of	  15	  chose	  positive,	  binomial	  probability	  test,	  p=	  .60).	  Six-­‐month-­‐olds.	  Consistent	  with	  the	  predicted	  direction,	  twelve	  of	  eighteen	  6-­‐month-­‐old	  Euro-­‐Canadian	  infants	  chose	  the	  Protagonist	  who	  received	  positive	  bystander	  reactions	  (binomial	  probability	  test,	  p	  =	  .238,	  2-­‐tailed).	  Unexpectedly,	  only	  4	  of	  18	  East	  Asian	  Canadian	  infants	  chose	  the	  positively	  evaluated	  Protagonist;	  a	  significant	  majority	  of	  infants	  chose	  the	  Protagonist	  who	  evoked	  a	  negative	  emotional	  reaction	  from	  the	  Bystanders	  (binomial	  probability	  test,	  p	  =	  .03,	  two-­‐tailed).	  Infants	  of	  mixed	  East	  Asian	  and	  European	  descent	  showed	  a	  similar	  choice	  pattern	  to	  Euro-­‐Canadians—they	  did	  not	  show	  a	  reliable	  preference	  for	  either	  Protagonist	  (9	  of	  13	  infants	  chose	  the	  positively	  evaluated	  Protagonist;	  binomial	  	  probability	  test,	  p	  =	  .267,	  2-­‐tailed).	  Figure	  4.2	  Infants	  	  choices	  by	  Protagonist	  valence,	  ethnicity,	  and	  age	  for	  Study	  1a.	  	  0	  2	  4	  6	  8	  10	  12	  14	  16	  Positively	  evaluated	  Protagonist	   Negatively	  evaluated	  Protagonist	   Positively	  evaluated	  Protagonist	   Negatively	  evaluated	  Protagonist	  6-­‐month-­‐olds	   12-­‐month-­‐olds	  Euro-­‐Canadian	   Mixed	   East	  Asian	  	   91	  Likelihood	  of	  choosing	  the	  positively	  evaluated	  Protagonist,	  by	  culture	  We	  used	  a	  second	  analytic	  strategy	  to	  investigate	  the	  magnitude	  of	  the	  difference	  in	  likelihood	  of	  choosing	  the	  Protagonist	  who	  evoked	  positive	  emotional	  reactions,	  compared	  to	  Protagonist	  who	  evoked	  negative	  reactions.	  Two	  binary	  logistic	  regressions	  were	  run,	  a	  full	  model	  including	  all	  predictors	  and	  covariates,	  and	  a	  partial	  model	  with	  all	  non-­‐significant	  predictors	  in	  the	  full	  model	  removed	  (“Using	  multivariate	  statistics,”	  2001),	  and	  the	  amount	  of	  variance	  explained	  were	  compared.	  In	  the	  full	  model,	  ethnicity	  (European,	  East	  Asian,	  and	  Mixed),	  age,	  sex	  (male,	  female),	  Protagonist	  that	  elicited	  the	  emotional	  reaction	  (yellow	  triangle	  Protagonist	  or	  blue	  square	  Protagonist),	  and	  reaction	  order	  (whether	  negative	  or	  positive	  emotional	  reaction	  trials	  were	  presented	  first)	  were	  entered	  as	  model	  predictors	  for	  likelihood	  of	  choosing	  the	  positive	  Protagonist.	  An	  omnibus	  test	  of	  the	  full	  model	  was	  significant	  (Chi-­‐square(6)	  =	  14.038,	  p	  =	  .029),	  improving	  our	  ability	  to	  predict	  infants’	  positive	  vs.	  negative	  Protagonist	  choices	  on	  11%	  of	  cases.	  Together,	  the	  coefficients	  explained	  approximately	  17%	  of	  the	  variance	  in	  target	  choice	  (Nagelkerke	  R-­‐square	  =	  .172).	  Looking	  at	  the	  individual	  predictors,	  the	  analyses	  revealed	  that	  an	  infant’s	  ethnicity	  is	  the	  only	  variable	  that	  helps	  predict	  whether	  an	  infant	  will	  pick	  the	  positive	  Protagonist.	  Specifically,	  an	  infant	  identified	  by	  parents	  as	  being	  of	  European	  ethnicity	  (including	  East	  Asian-­‐European	  mixed	  infants)	  significantly	  predicted	  a	  3.187	  times	  increase	  in	  likelihood	  of	  picking	  the	  positive	  Protagonist,	  compared	  to	  infants	  who	  do	  not	  have	  any	  European	  ethnicity	  (logistic	  regression	  coefficient	  =	  1.159,	  p	  =	  .036,	  Odds	  Ratio	  =	  3.187).	  Infants’	  sex,	  age,	  the	  order	  of	  stimuli	  presentation	  did	  not	  significantly	  	   92	  contribute	  to	  the	  model’s	  predictive	  power,	  nor	  was	  there	  a	  colour	  or	  shape	  preference	  in	  infants’	  choices.	  	  A	  partial	  model	  was	  also	  run,	  containing	  European	  ethnicity	  dummy	  variable	  (the	  only	  significant	  predictor	  in	  the	  full	  model)	  as	  a	  predictor,	  and	  age	  as	  a	  covariate.	  With	  only	  these	  predictors,	  the	  partial	  model	  is	  a	  better	  fit	  to	  the	  data	  than	  the	  full	  model:	  it	  significantly	  improves	  prediction	  of	  infants’	  choices	  by	  15%	  (omnibus	  test	  chi-­‐square	  (2)	  =	  12.098,	  p	  =	  .002),	  and	  explains	  15%	  of	  the	  variance	  in	  infants’	  choices	  (Nagelkerke	  R-­‐square	  =	  .15).	  In	  this	  model,	  being	  ethnically	  European	  increases	  an	  infant’s	  likelihood	  of	  choosing	  the	  positive	  Protagonist	  4.121	  times,	  compared	  to	  infants	  of	  other	  ethnicities	  (logistic	  regression	  coefficient	  =	  1.416,	  p	  =	  .002,	  Odds	  Ratio	  =	  4.121).	  The	  probability	  of	  picking	  the	  positive	  Protagonist	  increases	  with	  age,	  but	  was	  not	  significant	  at	  conventional	  levels	  (coefficient	  =	  .107,	  p	  =	  .149,	  Odds	  Ratio	  =	  1.112).	  Table	  4.1	  Logistic	  regression	  models	  predicting	  Protagonist	  choice	  in	  Study	  1a.	  Full	  model	  uses	  Ethnicity,	  age,	  sex,	  protagonist,	  and	  order	  to	  predict	  Protagonist	  choice.	  Partial	  model	  uses	  only	  significant	  predictors	  from	  Full	  Model	  to	  predict	  Protagonist	  choice.	  Predictors	  Full	  Model	   Partial	  Model	  Choice	  coefficient	  (B)	  Standard	  Error	  (SE)	  Choice	  coefficient	  (B)	  Standard	  Error	  (SE)	  European	   1.159*	   0.553	   1.416**	   0.456	  East	  Asian	   -­‐0.517	   -­‐0.517	  	  	  	  Age	  (centered)	   0.107	   0.077	   0.107	   0.074	  Sex	   -­‐0.356	   0.436	  	  	  	  Protagonist	   0.094	   0.454	  	  	  	  Reaction	  order	   -­‐0.237	   0.432	   	  	   	  	  †p	  <	  .10,	  *p	  <	  .05,	  **p	  <	  .01,	  ***p	  <	  .001	  	  We	  were	  also	  interested	  in	  using	  language	  exposure	  as	  a	  continuous	  measure	  of	  infants’	  exposure	  to	  different	  cultural	  practices.	  In	  a	  separate	  logistic	  regression	  	   93	  model,	  the	  amount	  of	  an	  East	  Asian	  language	  that	  an	  infant	  is	  exposed	  to	  significantly	  predicts	  their	  choice	  of	  target	  (model	  chi-­‐square	  (1)	  =	  10.449,	  p	  =	  .001,	  Nagelkerke	  R2	  =	  .13).	  In	  this	  model,	  each	  percentage	  increase	  in	  reported	  exposure	  to	  an	  East	  Asian	  language	  is	  associated	  with	  being	  1.02	  times	  as	  likely	  to	  choose	  the	  negative	  target	  (coefficient	  =	  -­‐.019,	  p	  =	  .002,	  OR	  =	  .981).	  To	  illustrate,	  based	  on	  our	  model	  we	  can	  calculate	  the	  likelihood	  that	  a	  6-­‐month-­‐old	  will	  choose	  the	  positive	  emotion	  target	  at	  0%,	  50%	  and	  100%	  East	  Asian	  language	  exposure,	  each	  of	  which	  were	  commonly-­‐reported	  language	  percentages	  in	  our	  sample.	  According	  to	  the	  model,	  with	  100%	  exposure	  to	  an	  East	  Asian	  language,	  the	  child	  has	  a	  20.2%	  probability	  of	  picking	  the	  positive	  emotion	  target;	  at	  50%	  exposure,	  the	  likelihood	  rises	  to	  42.02%	  (an	  increase	  of	  22%);	  at	  0%	  exposure	  (in	  this	  sample,	  0%	  East	  Asian	  language	  exposure	  is	  strongly	  correlated	  with	  full	  time	  exposure	  to	  English)	  the	  probability	  of	  picking	  the	  positive	  target	  is	  67.44%.	  Thus,	  compared	  to	  infants	  who	  receive	  no	  exposure	  to	  English,	  those	  who	  are	  exposed	  only	  to	  English	  are	  47%	  more	  likely	  to	  pick	  the	  target	  of	  positive	  emotional	  reactions,	  regardless	  of	  their	  ethnicity.	  	  4.3.3	   Discussion	  Six-­‐month-­‐old	  European	  infants	  in	  our	  sample	  did	  not	  reliably	  pick	  positively	  evaluated	  targets,	  while	  their	  East	  Asian	  counterparts	  preferred	  the	  negatively	  evaluated	  target.	  By	  12-­‐months,	  both	  East	  Asian	  and	  European	  infants	  showed	  a	  shift	  towards	  preferring	  positively	  evaluated	  targets.	  Twelve-­‐month-­‐old	  European	  infants	  chose	  targets	  of	  positive	  emotional	  reactions,	  suggesting	  that	  they	  do	  reliably	  use	  social	  referencing.	  East	  Asian	  infants	  were	  at	  chance	  and	  no	  longer	  	   94	  preferred	  the	  negative	  target.	  Overall,	  results	  of	  the	  binary	  logistic	  regression	  indicate	  that	  at	  both	  6	  and	  12	  months	  of	  age,	  being	  raised	  in	  a	  primarily	  European	  cultural	  upbringing	  predicts	  greater	  likelihood	  of	  an	  infant	  preferring	  the	  positively	  evaluated	  target.	  Mixed	  infants	  as	  a	  group	  did	  not	  show	  clear	  use	  of	  social	  referencing	  at	  either	  age.	  They	  showed	  choice	  patterns	  that	  were	  somewhat	  varied:	  more	  similar	  to	  European	  infants	  at	  6	  months,	  and	  more	  similar	  to	  East	  Asian	  infants	  at	  12	  months.	  These	  patterns	  may	  be	  due	  to	  the	  greater	  heterogeneity	  of	  socialization	  experiences	  mixed	  infants	  encounter,	  as	  reflected	  in	  the	  greater	  variability	  in	  their	  language	  exposure.	  One	  way	  to	  address	  this	  heterogeneity	  in	  enculturation	  is	  to	  employ	  an	  analytic	  method	  that	  treats	  ethnicity	  as	  a	  continuous	  predictor.	  The	  amount	  of	  exposure	  to	  languages	  can	  serve	  both	  as	  a	  proxy	  for	  degree	  of	  enculturation	  to	  different	  cultures	  represented	  in	  our	  sample.	  Higher	  percentages	  of	  exposure	  to	  an	  East	  Asian	  language	  should	  correspond	  to	  greater	  exposure	  to	  East	  Asian	  cultural	  practices	  that	  give	  rise	  to	  infants’	  differential	  response	  to	  emotional	  stimuli,	  and	  vice	  versa	  with	  English	  and	  mainstream	  North	  American	  culture.	  Consistent	  with	  this	  interpretation,	  more	  exposure	  to	  an	  East	  Asian	  language	  predicted	  greater	  likelihood	  of	  choosing	  the	  negative	  emotional	  target,	  consistent	  with	  the	  pattern	  of	  choice	  using	  ethnicity	  as	  a	  predictor.	  	  Cultural	  differences	  in	  infants’	  social	  preferences	  may	  be	  due	  to	  several	  possible	  differences	  in	  infants’	  experience	  with	  emotional	  expressions,	  including	  differential	  abilities	  in	  interpreting	  our	  stimuli,	  in	  interpreting	  emotional	  expressions,	  and	  differing	  interpretations	  of	  the	  situational	  demands	  of	  our	  study.	  	  	   95	  For	  example,	  European	  infants	  may	  be	  more	  familiar	  with	  both	  positive	  and	  negative	  emotions,	  and	  further,	  understand	  their	  significance	  and	  employ	  them	  effectively	  to	  regulate	  their	  own	  actions.	  Our	  understanding	  of	  the	  development	  of	  social	  referencing	  is	  based	  on	  predominantly	  European-­‐American	  samples,	  who	  are	  exposed	  to	  different	  emotional	  experiences	  than	  infants	  growing	  up	  in	  other	  cultural	  contexts	  (Bornstein	  et	  al.,	  2012;	  Fernald	  &	  Morikawa,	  1993;	  Keller	  et	  al.,	  2006).	  East	  Asian	  infants	  may	  have	  failed	  to	  differentiate	  their	  behaviours	  towards	  positively	  and	  negatively	  evaluated	  targets	  as	  a	  result	  of	  less	  familiarity	  with	  the	  significance	  of	  emotional	  communications,	  or	  less	  ability	  to	  regulate	  their	  own	  behaviours	  in	  response	  to	  observed	  emotions.	  That	  more	  infants	  of	  all	  ethnicities	  reached	  for	  the	  negative	  target	  at	  6	  months	  may	  also	  reflect	  this	  lower	  level	  of	  understanding	  in	  both	  European	  and	  East	  Asians	  infants	  at	  this	  age,	  which	  gives	  way	  to	  improved	  understanding	  and	  a	  shift	  in	  preference	  towards	  positive	  targets	  at	  12	  months.	  Although	  our	  results	  suggest	  that	  European-­‐Canadian	  infants	  understand	  and	  use	  the	  evaluative	  significance	  of	  emotional	  expressions,	  an	  alternative	  possibility	  is	  that	  an	  association	  between	  the	  salient	  target	  (i.e.	  Protagonist	  who	  was	  actively	  doing	  the	  choosing	  just	  before	  the	  emotional	  display)	  and	  a	  contiguous	  emotional	  expression	  could	  account	  for	  infants’	  choices.	  This	  account	  does	  not	  require	  infants	  to	  interpret	  the	  interaction	  between	  Bystanders	  and	  their	  target	  as	  an	  evaluative	  social	  situation	  in	  order	  to	  develop	  a	  preference;	  it	  merely	  requires	  infants	  to	  associate	  a	  target	  with	  a	  preferred	  emotion	  (through	  emotional	  contagion,	  temporal	  contiguity,	  or	  statistical	  relation).	  On	  the	  other	  hand,	  if	  infants	  understand	  emotions	  as	  being	  about	  a	  target,	  rather	  than	  merely	  associated	  with	  one,	  then	  they	  should	  not	  	   96	  show	  any	  preference	  for	  targets	  that	  are	  accompanied	  by	  emotions	  that	  are	  directed	  away	  from	  them	  (for	  example,	  if	  Julie	  smiles	  immediately	  after	  Gary	  makes	  a	  funny	  face	  behind	  her	  back,	  we	  would	  infer	  that	  since	  she	  couldn’t	  see	  Gary,	  Julie	  is	  likely	  smiling	  about	  something	  other	  than	  Gary’s	  funny	  face).	  Study	  1b	  was	  designed	  to	  examine	  the	  nature	  of	  infants’	  understanding	  of	  the	  relationship	  between	  emotional	  reactions	  and	  their	  targets,	  by	  looking	  at	  whether	  they	  are	  sensitive	  to	  Bystanders’	  visual	  access	  to	  the	  actions	  of	  their	  targets.	  	  4.4	   Study	  1b	  	  -­‐	  undirected	  gaze	  4.4.1	   Methods	  Fifty-­‐one	  full-­‐term	  infants	  participated	  in	  this	  study	  (M	  age=	  7	  months,	  range	  =	  5;15	  –	  12;15).	  Eighteen	  were	  Western	  European,	  15	  were	  East	  Asian,	  and	  12	  were	  East	  Asian-­‐European	  mixed.	  As	  in	  Study	  1a,	  data	  from	  6	  infants	  of	  other	  ethnicities	  were	  not	  included	  in	  the	  analyses	  because	  there	  were	  not	  enough	  participants	  from	  the	   same	   ethnic	   background	   to	   sufficiently	   draw	   conclusions	   about	   the	   group.	  Twelve	  infants	  were	  excluded	  from	  the	  analyses	  due	  to	  fussiness	  (n=10),	  procedure	  error	  (n=1),	  and	  no	  choice	  (n=1).	  Identical	  procedures	  were	  followed	  for	  Study	  1.	  The	  only	  stimuli	  modification	  was	  that	  at	  the	  beginning	  of	  each	  video,	  rather	  than	  gazing	  down	  at	  the	  Protagonists	  the	  Bystanders	  instead	  gaze	  upward,	  away	  from	  the	  scene,	  and	  remained	  fixed	  pointing	  upward	   for	   the	  remainder	  of	  each	  vignette.	  Through	   this	  modification,	  Bystanders	  have	  no	   visual	   access	   to	   the	   actors’	   object	   choices.	   Thus,	   although	   the	  Bystanders	  make	  the	  same	  emotional	  displays	  at	  the	  very	  same	  time,	  their	  expressions	  do	  not	  appear	  to	  refer	  to	  the	  scene	  below.	  	   97	  Figure	  4.3.	  Video	  stimuli	  for	  Study	  1b.	  Bystanders’	  gaze	  is	  directed	  away	  from	  the	  scene	   below	   so	   they	   have	   no	   knowledge	   of	   the	   yellow	   triangle’s	   actions	   (left).	  Bystanders	   express	   negative	   emotions	   away	   from	   the	   yellow	   character’s	   choice	  (right).	  	  	  4.4.2	   Results	  Infants	  were	  equally	  likely	  to	  choose	  targets	  associated	  with	  positive	  and	  negative	  emotional	  expressions,	  regardless	  of	  their	  ethnicity	  (7	  of	  18	  European	  infants,	  7	  of	  15	  East	  Asian,	  and	  5	  of	  12	  Mixed	  infants	  chose	  the	  positively	  associated	  target).	  Infants’	  age	  and	  sex	  did	  not	  affect	  their	  choice	  (see	  Table	  4.2	  for	  coefficients).	  	  	  Figure	  4.4	  Infants’	  Protagonist	  choice,	  by	  ethnicity,	  in	  Study	  1b.	  	  0	  2	  4	  6	  8	  10	  12	  Euro-­‐Canadian	   Mixed	   East	  Asian	  Protagonist	  choice	  Ethnicity	  Positively	  evaluated	  Protagonist	  Negatively	  evaluated	  Protagonist	  	   98	  4.4.3	  	   Comparing	  Study	  1a	  and	  1b	  The	  pattern	  of	  results	  for	  this	  study	  differs	  significantly	  from	  that	  of	  Study	  1a,	  in	  which	  Bystanders’	  gaze	  was	  directed	  at	  the	  Protagonists	  and	  their	  actions.	  	  Consistent	  with	  results	  of	  the	  simple	  binomial	  probability	  tests	  above,	  analyses	  using	  a	  binary	  logistic	  regression	  combining	  data	  across	  the	  2	  studies	  reveal	  that	  there	  is	  a	  significant	  interaction	  between	  Ethnicity	  (European	  vs.	  non-­‐European)	  and	  Study	  (directed	  gaze	  vs.	  undirected	  gaze;	  logistic	  regression	  coefficient	  =	  1.64,	  p	  =	  .037,	  Odds	  Ratio	  =	  5.156).	  That	  is,	  European	  infants	  differed	  significantly	  in	  their	  choices	  between	  Studies	  1a	  and	  1b,	  while	  East	  Asian	  infants	  did	  not.	  	  Similarly,	  using	  percent	  exposure	  to	  English	  as	  a	  predictor	  supports	  the	  interpretation	  that	  there	  is	  a	  significant	  interaction	  between	  infants’	  cultural	  background	  and	  their	  choice	  preferences	  across	  the	  2	  studies	  (coeff	  =	  .027,	  p	  =	  .01,	  OR	  =	  1.027).	  	  Table	  4.2	  Logistic	  regression	  models	  comparing	  results	  of	  Study	  1a	  and	  1b.	  Model	  1:	  Study	  and	  Ethnicity	  predicting	  Protagonist	  choice	  (left	  2	  columns),	  and	  Model	  2:	  Study	  and	  %	  English	  exposure	  predicting	  Protagonist	  choice	  (right	  2	  columns)	  Predictors	  ETHNICITY	  predictors	  model	   LANGUAGE	  predictors	  model	  Choice	  coefficient	  (B)	  Standard	  Error	  (SE)	  Choice	  coefficient	  (B)	  Standard	  Error	  (SE)	  Study	   -­‐0.065	   0.463	   -­‐1.321	   0.813	  European	   -­‐0.288	   0.645	  	  	  	  European*Study	   1.64*	   0.785	  	  	  	  %	  English	  	   	  -­‐0.011	   0.009	  %	  English	  *	  Study	   	  	   	  	   0.027*	   0.01	  †p	  <	  .10,	  *p	  <	  .05,	  **p	  <	  .01,	  ***p	  <	  .001	  4.4.4	   Discussion	  The	  current	  study	  found	  that	  infants	  did	  not	  show	  consistent	  preferences	  for	  Protagonists	  whose	  actions	  were	  merely	  temporally	  associated	  with	  positive	  versus	  	   99	  negative	  expressions	  from	  social	  others	  in	  a	  display,	  but	  that	  were	  not	  interpretable	  as	  referring	  to	  the	  Protagonists	  themselves.	  Unlike	  in	  Study	  1a,	  infants’	  preferences	  did	  not	  differ	  based	  on	  ethnicity.	  In	  contrast,	  when	  Bystanders’	  gazes	  were	  directed	  towards	  the	  Protagonists	  (Study	  1a),	  European	  infants,	  as	  well	  as	  Mixed	  and	  East	  Asian	  infants	  with	  large	  amounts	  of	  exposure	  to	  English,	  showed	  a	  preference	  for	  the	  positively	  evaluated	  target.	  Thus,	  earlier	  than	  previously	  demonstrated	  (14	  months;	  Luo	  &	  Beck,	  2010;	  Sodian,	  Thoermer,	  &	  Metz,	  2007),	  infants	  appear	  to	  be	  sensitive	  to	  the	  relationship	  between	  visual	  access	  and	  relevance	  of	  bystanders’	  emotional	  displays:	  they	  only	  took	  into	  account	  their	  emotional	  reactions	  when	  bystanders	  had	  visual	  access	  to	  the	  Protagonists’	  actions,	  but	  not	  when	  their	  gaze	  was	  directed	  away.	  	  Overall,	  the	  difference	  in	  choice	  patterns	  across	  Studies	  1a	  and	  1b	  supports	  the	  interpretation	  that	  infants	  with	  high	  levels	  of	  exposure	  to	  European-­‐Canadian	  culture	  find	  emotional	  expressions	  to	  be	  informative	  when	  they	  are	  directed	  at	  a	  referent,	  but	  not	  when	  they	  are	  merely	  spatially	  and	  temporally	  contiguous	  with	  its	  actions.	  In	  contrast,	  East	  Asian	  infants	  with	  lower	  levels	  of	  acculturation	  were	  less	  likely	  to	  differentiate	  between	  directed	  emotional	  reactions	  and	  undirected	  emotional	  expressions,	  and	  to	  use	  the	  former	  to	  inform	  their	  social	  evaluations.	  	  Another	  possible	  explanation	  remains	  for	  how	  infants’	  may	  be	  influenced	  by	  Bystanders’	   emotional	   reactions,	   without	   true	   conceptual	   understanding	   of	   the	  informational	   value	  of	   emotions.	  An	  observed	  emotion	  may	  affect	  behaviour	   via	   a	  process	  of	  emotional	  contagion,	  whereby	  infants	  exposed	  to	  others’	  emotional	  states	  then	   internally	   reproduce	   the	   physical	   states	   corresponding	   to	   those	   emotional	  	   100	  states,	  accompanied	  by	  the	  same	  behavioural	  tendencies	  provoked	  by	  the	  emotions.	  	  Infants	   who	   successfully	   used	   others’	   emotional	   expressions	   to	   guide	   their	  subsequent	   choices	   in	   Study	  1a	  may	  have	   personally	   experienced	   the	  Bystanders’	  emotions	   due	   to	   contagion,	   accompanied	   by	   the	   corresponding	   approach	   and	  avoidance	   behavioural	   tendencies,	   or	   they	  may	   have	   been	   able	   to	   understand	   the	  significance	   of	   emotional	   expressions	   without	   personally	   experiencing	   those	  emotions	  themselves.	  We	  address	  these	  possibilities	  in	  Study	  1c,	  by	  coding	  infants’	  affective	  displays	  and	  behaviours	  during	  the	  study	  session.	  	  4.5	   Study	  1c	  –	  Behavioural	  coding	  Emotional	  states	  can	  be	  transferred	  between	  people	  via	  emotional	  contagion,	  and	  affect	  people’s	  attitudes	  and	  behaviours	  (Barsade,	  2002;	  Cacioppo,	  &	  Rapson,	  1994).	  In	  the	  absence	  of	  infants’	  ability	  to	  report	  on	  their	  affective	  experiences,	  an	  expedient	  method	  to	  examine	  whether	  infants’	  choices	  were	  driven	  by	  “caught”	  affective	  states	  is	  to	  look	  at	  whether	  their	  expressions	  mirrored	  the	  Bystanders’	  expressions	  in	  the	  stimuli.	  Preverbal	  infants	  generally	  have	  limited	  capacity	  to	  regulate	  or	  mask	  their	  subjective	  experiences,	  and	  are	  thus	  expected	  to	  reveal	  their	  affective	  states	  through	  observable	  behaviours.	  If	  emotional	  contagion	  is	  at	  work	  in	  driving	  infants’	  social	  preferences,	  we	  would	  expect	  infants	  to	  express	  more	  positive	  affect	  on	  trials	  in	  which	  a	  Protagonist	  is	  positively	  evaluated	  by	  the	  Bystanders,	  and	  to	  express	  more	  negative	  affect	  on	  trials	  containing	  negative	  emotional	  reactions	  (i.e.	  correspondence	  between	  trial	  valence	  and	  infants’	  affective	  behaviours).	  	  In	  addition	  to	  affective	  expressions,	  we	  were	  interested	  in	  whether	  infants’	  looking	  behaviours	  could	  help	  explain	  their	  target	  choices.	  Previous	  research	  on	  	   101	  gaze	  has	  shown	  it	  to	  be	  a	  reliable	  indicator	  of	  self-­‐regulation	  (Stifter,	  2002;	  Stifter	  &	  Braungart,	  1995).	  Infants	  are	  able	  to	  control	  their	  exposure	  to	  stimuli	  by	  averting	  their	  eyes	  during	  arousing	  situations	  (Stern,	  1974;	  Stifter	  &	  Moyer,	  1991),	  both	  positive	  (Stifter	  &	  Moyer,	  1991)	  and	  negative	  (Field,	  2002).	  Along	  this	  line	  of	  reasoning,	  we	  wanted	  to	  measure	  infants’	  gaze,	  including	  when	  they	  engaged	  in	  watching	  the	  stimuli,	  and	  when	  they	  turned	  to	  reference	  their	  caregiver.	   4.5.1	   Methods	  	   To	  this	  end,	  10-­‐second	  clips	  featuring	  12-­‐month-­‐old	  infants	  who	  participated	  in	  Study	  1a	  described	  above	  were	  created.	  We	  reasoned	  that	  12-­‐month-­‐olds	  greater	  reliability	  in	  producing	  affective	  expressions	  would	  produce	  greater	  reliability	  in	  coding,	  and	  hence,	  more	  useful	  data.	  Each	  clip	  began	  immediately	  after	  infants	  saw	  the	  Bystanders’	  emotional	  expressions	  at	  the	  end	  of	  each	  trial,	  and	  ended	  when	  infants	  looked	  away	  from	  the	  screen	  for	  2	  seconds	  or	  more.	  Videos	  were	  centered	  on	  the	  infants’	  head	  and	  torso	  while	  seated	  on	  a	  caregiver’s	  lap,	  giving	  a	  clear	  view	  of	  their	  face	  and	  movements.	  Each	  clip	  was	  labeled	  using	  a	  series	  of	  numbers	  and	  letters	  so	  as	  to	  avoid	  providing	  coders	  with	  identifying	  information.	  A	  West	  Asian	  coder	  who	  was	  blind	  to	  participants’	  condition,	  and	  trial	  valence	  coded	  each	  video.	  Coding	  Scheme	  Each	  video	  was	  coded	  for	  positivity-­‐negativity,	  frequency	  of	  referencing	  the	  caregiver,	  and	  frequency	  of	  looks	  away	  from	  the	  screen.	  	  Positivity-­‐negativity	  was	  coded	  on	  a	  5-­‐point	  Likert	  scale,	  ranging	  from	  -­‐2	  (very	  negative)	  to	  +2	  (very	  positive).	  Zero,	  the	  midpoint,	  denoted	  neutral	  affect.	  Coders	  were	  instructed	  to	  use	  a	  holistic	  assessment	  of	  facial,	  vocal,	  and	  bodily	  	   102	  posture	  to	  code	  each	  video	  (Aknin,	  Hamlin,	  &	  Dunn,	  2012).	  Naïve	  coder	  ratings	  have	  been	  shown	  to	  correlate	  highly	  with	  validated	  emotion	  coding	  systems,	  such	  as	  Baby	  FACS	  (Oster,	  2003).	  Referencing.	  Infants	  who	  turn	  to	  immediately	  look	  at	  their	  caregiver	  were	  considered	  to	  engage	  in	  referencing	  (which	  may	  be	  due	  to	  desire	  to	  for	  reassurance,	  or	  to	  seek	  out	  information).	  The	  total	  number	  of	  times	  an	  infant	  referenced	  the	  caregiver	  was	  counted	  and	  recorded.	  	  Reliability	  Reliability	  between	  the	  primary	  and	  a	  second	  coder	  was	  established	  on	  an	  independent	  set	  of	  50	  videos	  (from	  another	  study)	  before	  coding	  on	  the	  current	  study	  began.	  Agreement	  was	  89%	  on	  the	  positivity-­‐negativity	  scale,	  and	  96%	  for	  frequency	  of	  caregiver	  referencing	  and	  looks	  away	  from	  screen	  between	  the	  2	  coders.	  Discrepant	  coding	  was	  discussed	  and	  resolved	  before	  coding	  on	  the	  current	  set	  commenced.	  	  4.5.2	   Results	  Ratings	  from	  the	  primary	  coder	  are	  reported	  below.	  Overall,	  infants	  did	  not	  differ	  in	  their	  expressions	  across	  positive	  and	  negative	  trials	  (mean	  difference	  =	  -­‐.027,	  p	  =	  .817).	  	  To	  investigate	  whether	  the	  lack	  of	  association	  between	  trial	  valence	  and	  infant	  affect	  applies	  equally	  across	  ethnicities,	  a	  generalized	  linear	  mixed	  model	  was	  used.	  This	  method	  was	  chosen	  because	  it	  can	  account	  for	  non-­‐independent	  errors	  on	  repeated	  measures	  (since	  each	  infants	  has	  4	  trials	  that	  have	  non-­‐independent	  errors).	  Using	  trial	  valence,	  ethnicity,	  and	  ethnicity-­‐by-­‐positivity	  interaction	  to	  	   103	  predict	  infants’	  positivity,	  we	  found	  no	  significant	  main	  effects	  of	  either	  valence	  (estimate	  of	  fixed	  effect	  =	  -­‐.132,	  p	  =	  .753)	  or	  ethnicity	  (estimate	  of	  fixed	  effect	  (East	  Asian)=	  .241,	  p	  =	  .339;	  estimate	  of	  fixed	  effect	  (European)	  =	  .145,	  p	  =	  .551).	  However,	  there	  was	  a	  significant	  interaction	  of	  ethnicity	  by	  valence,	  such	  that	  Mixed	  infants	  were	  significantly	  more	  positive	  on	  positively	  valenced	  trials,	  and	  more	  negative	  on	  negatively	  valenced	  trials	  than	  both	  East	  Asian	  and	  European	  infants	  (estimate	  (East	  Asian)	  =	  -­‐.528,	  p	  =	  .025;	  estimate	  (European)	  =	  -­‐.450,	  p	  =	  .047;	  Mixed	  ethnicity	  is	  the	  reference	  group).	  This	  is	  an	  unexpected	  interaction	  that	  we	  did	  not	  have	  any	  a	  priori	  prediction	  or	  theoretical	  explanation	  for.	  Given	  the	  lack	  of	  main	  effect	  of	  ethnicity	  and	  valence,	  it	  is	  possible	  that	  it	  is	  an	  artifact	  of	  capitalizing	  on	  chance	  (especially	  given	  the	  small	  sample	  of	  Mixed	  infants).	  	  Finally,	  we	  examined	  whether	  trial	  valence	  predicts	  the	  number	  of	  times	  an	  infant	  references	  his	  or	  her	  parent.	  We	  found	  no	  relationship	  between	  the	  valence	  of	  the	  stimuli	  and	  the	  number	  of	  times	  infants	  look	  to	  their	  parent.	  However,	  there	  is	  a	  small	  marginal	  effect	  of	  trial	  number,	  such	  that	  infants	  are	  more	  likely	  to	  reference	  his	  or	  her	  parents	  with	  later	  trials,	  regardless	  of	  valence	  (estimate	  =	  .071,	  p	  =	  .084).	  This	  result	  is	  likely	  due	  to	  increasing	  boredom,	  as	  infants	  approach	  the	  habituation	  criteria.	  4.5.3	   Discussion	  Infants’	  expressions	  and	  behaviours	  are	  not	  associated	  with	  the	  valence	  of	  each	  stimulus	  trial,	  and	  thus,	  we	  did	  not	  find	  evidence	  of	  emotional	  contagion	  influencing	  infants’	  target	  choices.	  This	  preliminary	  step	  does	  not	  rule	  out	  the	  	   104	  possibility	  that	  infants	  did	  experience	  emotional	  contagion.	  However,	  we	  were	  unable	  to	  detect	  it	  through	  their	  facial	  and	  behavioural	  expressions.	  	  So	  far,	  Study	  1	  has	  examined	  preverbal	  infants’	  understanding	  of	  emotional	  reactions	  in	  association	  with	  social	  targets.	  Study	  1b	  ruled	  out	  associative	  learning	  as	  a	  mechanism	  for	  infants’	  choices,	  and	  Study	  1c	  helps	  discount	  emotional	  contagion	  as	  another	  potential	  mechanism.	  Taken	  together,	  it	  is	  plausible	  European-­‐Canadian	  infants’	  preference	  for	  the	  target	  of	  positive	  emotional	  reactions	  was	  due	  to	  an	  understanding	  of	  the	  meaning	  conveyed	  by	  positive	  reactions,	  rather	  than	  a	  result	  of	  affective	  contagion,	  or	  a	  temporal	  association	  between	  salient	  physical	  characters	  and	  positive	  emotional	  expressions.	  	  Next,	  we	  turn	  to	  address	  the	  second	  question	  outlined	  earlier,	  comparing	  infants’	  use	  of	  social	  referencing	  for	  asocial	  targets	  with	  social	  targets.	  While	  numerous	  studies	  have	  examined	  infants’	  use	  of	  social	  referencing	  for	  asocial	  targets,	  showing	  that	  by	  10	  months,	  infants	  reliably	  use	  observed	  emotions	  to	  regulate	  behaviours	  towards	  novel	  toys	  and	  objects,	  none	  has	  included	  social	  referencing	  for	  social	  targets	  in	  the	  same	  study	  (Mumme	  et	  al.,	  1996;	  Sorce	  et	  al.,	  1985;	  Walden	  &	  Ogan,	  1988).	  Thus,	  we	  were	  interested	  in	  using	  a	  standardized	  methodology	  to	  conduct	  such	  a	  comparison.	  We	  turn	  to	  this	  question	  in	  Study	  2.	  4.6	   Study	  2	  –	  Object	  targets	  4.6.1	  Methods	  One	  hundred	  and	  three	  full	  term	  infants	  participated	  in	  this	  study	  (53	  6-­‐month-­‐olds,	  50	  12-­‐month-­‐olds;	  M	  age=	  9.23	  months,	  range	  =	  5;15	  –	  12;16).	  Thirty-­‐six	  were	  Western	  European,	  34	  were	  East	  Asian,	  and	  18	  were	  East	  Asian-­‐European	  	   105	  mixed.	  Fifteen	  infants	  of	  other	  ethnicities	  also	  participated	  in	  the	  study.	  However,	  due	  to	  large	  number	  of	  cultural	  backgrounds	  represented,	  there	  were	  insufficient	  participants	  from	  the	  same	  ethnic	  background	  to	  draw	  conclusions	  about	  the	  broader	  cultural	  group.	  Thus,	  they	  were	  excluded	  from	  results	  presented	  below.	  An	  additional	  46	  infants	  were	  excluded	  from	  the	  analysis	  due	  to	  fussiness	  (n	  =	  15),	  no	  choice	  (n	  =	  15),	  parental	  interference	  (n	  =	  8),	  and	  procedural	  error	  (n	  =	  8).	  Procedures	  were	  identical	  to	  Study	  1.	  As	  with	  Study	  1,	  infants	  watch	  videos	  of	  two	  characters	  (a	  blue	  square	  and	  a	  yellow	  triangle	  with	  eyes)	  alternately	  choosing	  between	  2	  objects	  (here,	  a	  bottle	  and	  a	  highlighter).	  The	  only	  difference	  is	  in	  modifications	  to	  the	  stimuli	  such	  that	  across	  the	  set	  of	  4	  habituation	  trials,	  one	  object	  is	  always	  smiled	  at,	  regardless	  of	  which	  character	  chooses	  it;	  the	  other	  is	  always	  frowned	  at.	  This	  design	  is	  utilized	  to	  imply	  that	  Bystanders	  differentially	  value	  the	  objects,	  but	  not	  the	  characters.	  If	  infants	  can	  learn	  about	  objects	  as	  bystanders	  to	  others’	  emotional	  reactions,	  then	  their	  object	  choices	  should	  be	  reliably	  affected	  by	  the	  emotions	  they	  witnessed	  being	  expressed	  towards	  the	  objects.	  Furthermore,	  they	  should	  prefer	  and	  reach	  for	  objects	  that	  were	  the	  targets	  of	  positive	  emotional	  reactions	  more	  than	  objects	  that	  were	  targets	  of	  negative	  emotional	  reactions.	  	  	  	  	  	  	  	  	  	  	   106	  Figure	   4.5	  Stimuli	   for	   Study	   2:	   Bystanders	   reacting	   positively	   to	   the	   blue	   square	  picking	   the	   blue	   bottle	   (left).	   Bystanders	   reacting	   negatively	   to	   the	   blue	   square	  picking	  the	  purple	  highlighter	  (right).	  Not	  shown	  here:	  same	  scenarios	  depicting	  the	  yellow	  triangle	  making	  the	  same	  choices	  as	  the	  blue	  square,	  and	  eliciting	  the	  same	  reactions	  from	  the	  Bystanders.	  	  	  4.6.2	   Results	  Six-­‐month-­‐old	  infants	  did	  not	  show	  a	  significant	  preference	  for	  either	  positively	  or	  negatively	  evaluated	  object	  (12	  of	  18	  European	  infants,	  8	  of	  18	  East	  Asian,	  and	  4	  of	  9	  Mixed	  infants	  chose	  the	  positively	  associated	  target).	  12-­‐month-­‐old	  infants	  in	  our	  sample	  similarly	  showed	  a	  lack	  of	  distinct	  object	  preference	  based	  on	  the	  valence	  of	  Bystanders’	  emotional	  reactions.	  Ten	  out	  of	  18	  European	  infants,	  9	  out	  of	  16	  East	  Asian	  infants,	  and	  3	  out	  of	  9	  Mixed	  infants	  picked	  the	  positively	  evaluated	  object	  target.	  Repeating	  our	  analytic	  strategy	  from	  Study	  1,	  we	  conducted	  a	  binary	  logistic	  regression	  combining	  the	  2	  age	  groups	  to	  predict	  infants’	  target	  choice.	  Infants’	  ethnicity,	  age	  and	  sex	  did	  not	  affect	  their	  choices.	  	  	  	   107	  Figure	  4.6	  Infants’	  object	  choice	  by	  ethnicity,	  and	  by	  age	  in	  Study	  2.	  	  Next,	   we	   compared	   results	   across	   Studies	   1a	   and	   1b	   in	   order	   to	   examine	  whether	  6	  and	  12-­‐month-­‐old	   infants	  use	  social	  referencing	  similarly	   for	  social	  and	  asocial	  targets.	  Conducting	  a	  binary	  logistic	  regression	  model	  using	  Study	  (social	  vs.	  asocial),	  ethnicity	  (European	  vs.	  non-­‐European),	  and	  Study-­‐by-­‐ethnicity	  interaction	  term	   to	   predict	   infants’	   target	   choices	   based	   on	   valence	   of	   Bystanders’	   emotional	  reactions,	   we	   found	   that	   overall,	   infants	   did	   not	   choose	   differently	   across	   the	   2	  studies	  (B=	  .44,	  p	  =	  .186,	  OR=1.552).	  This	  is	  due	  to	  a	  significant	  cultural	  difference	  in	  how	   infants	  used	  social	   referencing:	  European	   infants	  were	  3.87	   times	  as	   likely	   to	  pick	  targets	  of	  positive	  emotional	  reactions,	  compared	  to	  non-­‐European	  infants	  (B	  =	  1.352,	  p	  =	   .002,	  OR=	  3.867).	  Furthermore,	   the	  extent	   to	  which	  European-­‐Canadian	  infants	  used	  social	   referencing	  varied	  by	  Study:	   their	  preference	   for	   the	  positively	  evaluated	  Social	   target	  was	  nearly	  4	   times	  as	   large	  as	   for	   the	  positively	   evaluated	  Asocial	  target	  (B=	  -­‐1.342,	  p	  =	  .003,	  OR	  =	  .261).	  	  	  0	  2	  4	  6	  8	  10	  12	  14	  Positively	  evaluated	  object	   Negatively	  evaluated	  object	   Positively	  evaluated	  object	   Negatively	  evaluated	  object	  6-­‐month-­‐olds	   12-­‐month-­‐olds	  Euro-­‐Canadian	   Mixed	   East	  Asian	  	   108	  4.6.3	   Discussion	  Unlike	   previous	   studies	   of	   social	   referencing,	   results	   of	   the	   present	   study	  found	   that	   infants	   who	   observed	   bystander	   evaluations	   of	   novel	   objects	   did	   not	  reliably	  use	  those	  evaluative	  expressions	  to	  guide	  their	  behaviours	  with	  regards	  to	  the	   objects.	   This	   also	   stands	   in	   contrast	   to	   Euro-­‐Canadian	   infants’	   demonstrated	  ability	   to	   use	   the	   same	   emotional	   evaluations	   to	   guide	   their	   social	   evaluations	   in	  Study	   1.	   A	   few	   design	   differences	   between	   the	   current	   and	   previous	   studies	  may	  account	  for	  these	  contrasting	  results.	  	  Previous	   studies	   of	   social	   referencing	   with	   infants	   in	   their	   first	   year	   have	  only	  shown	  direct	  emotional	  expressions	  towards	  objects,	  and	  not	  towards	  people	  interacting	  with	  objects.	   It	   is	  possible	   that	  social	   targets,	  as	   intentional	  agents,	  are	  privileged	   as	   targets	   of	   social	   referencing.	   And	  while	   doable,	   social	   referencing	   of	  objects	  may	   involve	  more	  complex	  cognitions,	  which	  proved	   to	  be	   too	  challenging	  for	   even	   the	   oldest	   infants	   tested	   here	   (12.5	   months).	   The	   only	   published	   study	  showing	  successful	  use	  of	  emotional	  reactions	   to	  an	   individual	   interacting	  with	  an	  object	  occurs	  at	  18	  months	  (Repacholi	  &	  Meltzoff,	  2007),	  several	  months	  older	  than	  infants	  in	  our	  sample.	  That	  infants	  may	  privilege	  affective	  reactions	  to	  social	  targets	  over	   object	   targets	   is	   corroborated	   by	   the	   fact	   that	   extracting	   Bystanders’	  evaluations	   of	   object	   targets	   in	   Study	   2	   required	   the	   exact	   same	   process	   of	  aggregating	   evaluative	   information	   across	   4	   trials	   as	   to	   identify	   the	   favoured	  Protagonist	   in	  Study	  1a.	  Hence,	   rather	   than	  greater	  salience	  of	   social	   compared	   to	  object	   targets	   in	   these	   stimuli	   in	  particular,	   the	  difference	   in	   infants’	   performance	  	   109	  across	   the	   two	   studies	  may	   be	   due	   to	   infants’	   differential	   sensitivity	   to	   social	   vs.	  asocial	  targets	  more	  broadly.	  	  4.7	   General	  discussion	  Our	  hypotheses	  regarding	  the	  influence	  of	  observed	  valenced	  emotional	  expressions	  on	  infants’	  behaviours	  were	  partially	  supported.	  European	  infants	  preferred	  social	  targets	  that	  received	  social	  approval	  in	  the	  form	  of	  positive	  emotional	  reactions	  over	  those	  who	  received	  social	  rejection	  in	  the	  form	  of	  negative	  emotional	  reactions	  in	  Study	  1a;	  their	  choices	  did	  not	  appear	  to	  be	  driven	  by	  a	  mere	  temporal	  association	  between	  a	  salient	  action	  and	  a	  contiguous	  emotional	  expression	  (Study	  1b),	  nor	  due	  to	  affective	  contagion	  from	  the	  Bystanders	  to	  the	  infant	  (Study	  1c).	  In	  contrast,	  the	  very	  same	  emotional	  reactions	  did	  not	  influence	  European	  infants’	  preferences	  for	  asocial	  targets	  (Objects);	  at	  both	  6-­‐	  and	  12-­‐months	  infants	  chose	  equally	  between	  positively	  and	  negatively	  evaluated	  object	  targets	  (Study	  2).	  	  Unlike	  Euro-­‐Canadian	  infants,	  East	  Asian	  infants	  did	  not	  prefer	  the	  positively	  evaluated	  Protagonist.	  At	  6	  months	  of	  age,	  East	  Asian	  infants	  preferred	  the	  negatively	  evaluated	  Protagonist,	  and	  by	  12	  months	  chose	  equally	  between	  Protagonists	  who	  provoked	  positive	  versus	  negative	  emotional	  reactions	  (Study	  1a).	  As	  with	  European	  infants,	  East	  Asian	  infants	  also	  did	  not	  use	  others’	  emotional	  reactions	  to	  asocial	  targets	  to	  inform	  their	  preferences,	  choosing	  at	  chance	  in	  response	  to	  observing	  positive	  and	  negative	  emotional	  reactions	  to	  objects	  at	  both	  6	  and	  12	  months	  of	  age	  (Study	  2).	  Furthermore,	  their	  choices	  do	  not	  appear	  to	  be	  driven	  by	  temporal	  contiguity	  (Study	  1b)	  or	  affective	  contagion	  (Study	  1c).	  	  	   110	  From	  a	  social	  referencing	  perspective,	  we	  expected	  infants	  to	  choose	  the	  target	  that	  had	  elicited	  positive	  emotional	  reactions	  from	  the	  Bystanders	  because	  it	  communicates	  that	  the	  referent	  is	  valued.	  There	  are	  a	  variety	  of	  reasons	  why	  infants	  would	  prefer	  the	  valued	  Protagonist.	  First,	  socially	  vetted	  individuals	  tend	  to	  have	  a	  history	  of	  being	  prosocial,	  and	  cooperative,	  thus	  rendering	  interactions	  with	  such	  individuals	  relatively	  low	  risk,	  and	  potentially	  beneficial.	  Another	  reason	  is	  that	  through	  interacting	  with	  popular	  individuals,	  one’s	  reputation	  might	  improve	  by	  association,	  through	  the	  logic	  of	  a	  friend	  of	  a	  friend	  is	  my	  friend.	  Finally,	  by	  paying	  attention	  to	  the	  valued	  individuals,	  one	  may	  learn	  skills	  that	  led	  their	  achieving	  social	  esteem	  in	  the	  first	  place	  (Foulsham,	  Cheng,	  Tracy,	  Henrich,	  &	  Kingstone,	  2010;	  Henrich	  &	  Gil-­‐White,	  2001).	  The	  results	  observed	  in	  Euro-­‐Canadian	  infants	  are	  consistent	  with	  this	  explanation.	  We	  did	  not	  compare	  positively	  and	  negatively	  evaluated	  target	  against	  neutral	  target,	  thus	  it	  is	  unclear	  what	  motivated	  infants’	  choices:	  avoidance	  of	  the	  negatively	  evaluated	  target,	  or	  approach	  of	  the	  positively	  evaluated	  target.	  Previous	  work	  suggests	  that	  infants	  may	  exhibit	  a	  negativity	  bias	  in	  emotional	  processing	  (see	  Vaish,	  Grossmann,	  &	  Woodward,	  2008	  for	  a	  review).	  The	  greater	  weight	  given	  to	  negative	  information	  may	  extend	  to	  other	  areas	  as	  well,	  such	  as	  sociomoral	  evaluations.	  For	  example,	  Hamlin	  and	  colleagues	  (Hamlin	  et	  al.,	  2010)	  found	  that	  3-­‐month-­‐olds	  prefer	  neutral	  to	  antisocial	  targets,	  but	  did	  not	  differentiate	  their	  response	  to	  helpful	  and	  neutral	  targets,	  suggesting	  that	  avoidance	  of	  potential	  threat	  may	  be	  more	  important	  than	  seeking	  out	  beneficial	  opportunities.	  Future	  work	  	   111	  would	  need	  to	  be	  conducted	  to	  see	  whether	  avoidance	  of	  the	  negatively	  evaluated	  target	  is	  similarly	  driving	  infants’	  choices	  here.	  As	  noted	  in	  the	  introduction	  of	  this	  chapter,	  there	  have	  been	  documented	  differences	  in	  parental	  behaviours	  across	  cultures.	  One	  such	  difference	  is	  in	  the	  amount	  of	  physical	  contact	  parents	  have	  with	  infants	  in	  relatively	  autonomous	  European	  and	  North	  American	  cultures,	  compared	  to	  relatively	  interdependent	  cultures	  in	  East	  Asia	  and	  Latin	  America	  (Keller	  et	  al.,	  2006;	  Salomo	  &	  Liszkowski,	  2013).	  One	  consequence	  of	  being	  in	  direct	  physical	  contact	  with	  the	  caregiver	  is	  that	  infants	  may	  receive	  less	  practice	  with	  interpreting	  remote,	  affective	  cues	  from	  parents	  (since	  physical	  contact	  usually	  entails	  side-­‐by-­‐side,	  rather	  than	  face-­‐to-­‐face	  interaction).	  As	  a	  consequence,	  East	  Asian	  infants	  may	  gain	  proficiency	  at	  using	  others’	  emotions	  to	  self-­‐regulate	  at	  a	  slower	  rate	  than	  their	  European-­‐Canadian	  counterparts.	  	  A	  related	  explanation	  is	  that	  accuracy	  of	  perception	  of	  negative	  emotions	  may	  be	  lower	  amongst	  East	  Asian	  compared	  to	  Europeans	  (see	  Ekman,	  Friesen,	  et	  al,	  1987;	  Izard,	  1977;	  Matsumoto,	  1992).	  This	  is	  a	  variant	  of	  the	  above	  explanation,	  where	  less	  familiarity	  may	  have	  led	  to	  confusion	  between	  similar	  expressions,	  such	  as	  anger	  and	  sadness,	  2	  emotions	  that	  have	  lower	  cross	  cultural	  recognition	  rates	  even	  among	  adults	  (Calvo,	  Gutiérrez-­‐García,	  Fernández-­‐Martín,	  &	  Nummenmaa,	  2014;	  Nelson	  &	  Russell,	  2013).	  By	  this	  account,	  East	  Asian	  infants	  may	  have	  perceived	  the	  negative	  faces	  to	  be	  sad	  rather	  than	  angry.	  Furthermore,	  studies	  have	  documented	  cultural	  differences	  in	  the	  degree	  to	  which	  parents	  encourage	  the	  development	  of	  other-­‐regarding	  concern	  vs.	  mastery	  of	  the	  physical	  environment	  	   112	  (Tamis-­‐LeMonda,	  Katz,	  &	  Bornstein,	  2002).	  Parental	  encouragement	  to	  prioritize	  other-­‐regarding	  concern,	  combined	  with	  interpreting	  the	  stimuli	  faces	  as	  expressions	  of	  sadness	  could	  have	  activated	  a	  desire	  to	  approach	  and	  comfort	  the	  negatively	  evaluated	  target	  Protagonist,	  superseding	  a	  desire	  to	  avoid	  a	  potentially	  bad	  social	  partner.	  	  In	  observations	  of	  Japanese	  and	  American	  mother-­‐infant	  dyads,	  Japanese	  mothers	  tended	  to	  put	  more	  emphasis	  on	  rehearsing	  social	  routines	  with	  their	  infants	  and	  instilling	  empathy	  for	  others’	  feelings,	  while	  American	  mothers	  tended	  to	  emphasize	  individual	  autonomy	  and	  direct	  infants’	  attention	  to	  extra-­‐dyadic	  objects	  (Fernald	  &	  Morikawa,	  1993;	  Tamis-­‐LeMonda,	  Bornstein,	  Cyphers,	  Toda,	  &	  Ogino,	  1992).	  For	  example,	  Fernald	  and	  Morikawa	  (1993)	  documented	  a	  two-­‐fold	  difference	  in	  the	  frequency	  with	  which	  Japanese	  and	  American	  mothers	  led	  their	  infants	  in	  social	  routines	  that	  practiced	  social	  exchanges,	  empathy	  and	  greetings.	  The	  implication	  of	  such	  differences	  in	  emphasis	  during	  socialization	  is	  that	  Euro-­‐Canadian	  infants	  may	  learn	  to	  prioritize	  interpreting	  emotional	  cues	  to	  accurately	  identify	  the	  best	  social	  partner	  (crucial	  to	  navigating	  the	  world	  as	  an	  autonomous	  agent	  and	  avoiding	  danger),	  while	  East	  Asian	  infants	  may	  learn	  to	  prioritize	  social	  harmony	  (approaching	  the	  socially	  rejected	  target	  to	  provide	  comfort	  and	  affiliation)	  over	  avoidance	  of	  negatively	  evaluated	  targets.	  	  Are	  there	  cultural	  differences	  in	  parents’	  reliance	  on	  emotional	  communications	  to	  convey	  evaluative	  information	  about	  people	  and	  objects?	  If	  so,	  what	  other	  means	  of	  communications	  are	  used	  to	  do	  so?	  In	  order	  to	  better	  understand	  infants’	  experiences	  with	  evaluative	  communications	  in	  our	  sample,	  we	  	   113	  use	  structured	  observations	  to	  investigate	  how	  parents	  in	  our	  sample	  interact	  with	  their	  infants	  to	  convey	  information	  about	  approval	  and	  disapproval	  toward	  2	  different	  targets.	  This	  study	  is	  outlined	  in	  the	  following	  chapter.	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	  	   114	  5.	  Parents’	  affective	  communications	  across	  cultures	  	  5.1	   Introduction	  The	   current	   chapter	   investigates	   differences	   and	   similarities	   in	   parenting	  practices	   across	  East	  Asian	   and	  European-­‐Canadian	   cultures	   that	  may	   account	   for	  differences	  in	  infants’	  use	  of	  social	  referencing	  in	  Chapter	  4.	  	  Parents	   constitute	   a	   vital	   aspect	   of	   infants’	   developmental	   context.	   Parents	  play	  a	  significant	  role	  in	  structuring	  infants’	  interactions	  by	  shaping	  basic	  aspects	  of	  cognition,	   including	   directing	   attention	   (Tomasello	   &	   Todd,	   1983),	   co-­‐regulating	  emotions	   (Murray	   &	   Trevarthen,	   1985;	   Reddy,	  Murray,	   &	   Trevarthen,	   1997),	   and	  shaping	  expectations	  for	  social	  relationships	  (Johnson	  et	  al.,	  2007;	  2010).	  They	  are	  also	  primary	  agents	  of	  cultural	  transmission	  in	  early	  life,	  modeling	  and	  encouraging	  culturally	  valued	  behaviours,	  emotions	  and	  attitudes,	  and	  responding	  negatively	  to	  counter-­‐normative	  ones.	   Indeed,	   although	   there	  are	  undoubtedly	   idiosyncrasies	   in	  the	   details	   of	   behaviours	   that	   are	   modeled	   across	   individual	   parent-­‐child	   dyads,	  these	   are	   likely	   to	   be	   yoked	   to	   a	   specific	   culture	   and	   its	   particular	   set	   of	   social	  norms.	  	  	  	  Parenting	  practices	  share	  both	  important	  similarities	  and	  differences	  across	  cultures,	  reflecting	  local	  perspectives	  of	   ideal	  end	  points	  in	  development	  (Keller	  et	  al.,	  2006).	  Individualistic	  cultures,	  typified	  by	  North	  Americans	  of	  European	  descent,	  perceive	  the	  ideal	  end	  point	  of	  development	  to	  include	  achieving	  autonomy	  and	  well	  defined	   personal	   preferences,	   attitudes,	   and	   goals.	   Correspondingly,	   parenting	  practices	  in	  individualistic	  cultures	  (e.g.	  American	  mothers)	  propagate	  these	  values	  by	   encouraging	   physical	   autonomy,	   tending	   to	   let	   their	   infants	   take	   the	   lead	   in	  	   115	  initiating	  activities	  and	  interactions	  (Kuchner,	  1981).	  Parents	  also	  foster	  self-­‐esteem	  by	  focusing	  on	  children’s	  personal	  attributes,	  preferences	  and	  judgments	  (Mullen	  &	  Yi,	  1995;	  Wang,	  Leichtman,	  &	  Davies,	  2000);	  for	  example,	  casting	  their	  child	  as	  the	  central	   character	   in	   narratives	   of	   past	   events,	   and	   focusing	   conversations	   on	   the	  child’s	  thoughts	  and	  feelings,	  rather	  than	  other	  features	  of	  the	  situation,	  such	  as	  how	  others	  were	  feeling,	  or	  discussing	  contextually	  normative	  behaviours.	  Individualistic	  cultures	  also	  tend	  to	  hold	  the	  view	  that	  emotions	  are	  an	  authentic	  expression	  of	  the	  self	  (Markus	  &	  Kitayama,	  1995),	  reflecting	  one’s	  true	  feelings,	  rather	  than	  culturally	  prescribed	  scripts	  for	  one’s	  particular	  role	  in	  a	  situation.	  	  In	  contrast,	  interdependent	  cultures	  common	  in	  East	  Asia	  and	  Latin	  America	  tend	   to	   emphasize	   relatedness	   among	   group	   members	   and	   social	   harmony	  (Kagitcibasi,	   2005;	  Keller	   et	   al.,	   2006;	  Markus	  &	  Kitayama,	   1991).	   Consistently,	   in	  interdependent	  cultures,	  parenting	  behaviours	  emphasize	  social	  routines	  that	  show	  greater	   tolerance	   for	   other-­‐reliance,	   and	   promote	   relatedness	   and	   concern	   for	  others	  (Dennis,	  Talih,	  Cole,	  Zahn-­‐Waxler,	  &	  Mizuta,	  2007).	  	  Studies	   comparing	   parental	   behaviours	   across	   American	   and	   East	   Asian	  cultures	  have	  documented	  both	  universality	  and	  cultural	  variability	  (Callaghan	  et	  al.,	  2011).	  In	  naturalistic	  observations	  of	  mothers	  and	  infants	  freely	  interacting	  in	  their	  homes,	   Japanese	   and	   Amerian	  mothers	   appeared	   to	   spend	   similar	   proportions	   of	  time	   engaging	   their	   infants	   in	   social	   and	   pedagogical	   interactions	   (Bornstein,	  Azuma,	   Tamis-­‐LeMonda,	   &	   Ogino,	   1990).	   Mothers	   are	   similar	   in	   overall	   levels	   of	  maternal	   expressiveness,	   touching,	   and	   vocalizing	   (Fogel	   et	   al.,	   1988).	   However,	  studies	  have	  also	  documented	  mean	  differences	  in	  the	  amount	  of	  maternal-­‐initiated	  	   116	  interactions	   across	   American	   and	   East	   Asian	   samples.	   For	   instance,	   Japanese	  mothers	  speak	  more	  to	  their	  infants	  than	  do	  American	  mothers;	  consistent	  with	  the	  above	   observation	   that	   American	  mothers	   tend	   to	   let	   their	   infants	   take	   the	   lead.	  Similarly,	  Japanese	  mothers	  initiate	  more	  interactions,	  especially	  exchange	  routines	  (used	  in	  offering,	  requesting,	  and	  accepting	  objects	  politely)	  and	  empathy	  routines	  (used	  to	  encourage	  infants	  to	  behave	  positively	  towards	  a	  person	  or	  object)	  at	  every	  age	   of