1   1   Assessment  of  attachment  behaviour  to  human  caregivers  in  wolf  pups  (Canis  lupus   2   lupus).   3   Nathaniel  J.  Hall1,  Kathryn  Lord2,  Anne-­‐Marie  K.  Arnold3,  Clive  D.L.  Wynne4,  Monique   4   A.R.  Udell5   5   1Department  of  Psychology,  University  of  Florida,  Gainesville.     6   2Department  of  Biology,  Gettysburg  College   7   3Department  of  Animal  Ecology,  University  of  Utrecht,  Netherlands.   8   4Department  of  Psychology,  Arizona  State  University,  Tempe.   9   5Animal  and  Rangeland  Sciences,  Oregon  State  University,  Corvallis.   10     11   Corresponding  author   12    Nathaniel  J.  Hall   13   njhall1@ufl.edu   14   Phone:  (352)  392-­‐0601   15   Fax:  (352)  392-­‐7985   16   Department  of  Psychology,  University  of  Florida     17   Gainesville,  FL  32611   18         19   2   Abstract   20   Previous  research  suggested  that  16-­‐week  old  dog  pups,  but  not  wolf  pups,   21   show  attachment  behaviour  to  a  human  caregiver.  Attachment  to  a  caregiver  in  dog   22   pups  has  been  demonstrated  by  differential  responding  to  a  caregiver  compared  to   23   a  stranger  in  the  Ainsworth  Strange  Situation  Test.  We  show  here  that  3-­‐7  week  old   24   wolf  pups  also  show  attachment-­‐like  behaviour  to  a  human  caregiver  as  measured   25   by  preferential  proximity  seeking,  preferential  contact,  and  preferential  greeting  to   26   a  human  caregiver  over  a  human  stranger  in  a  modified  and  counterbalanced   27   version  of  the  Ainsworth  Strange  Situation  Test.  In  addition,  our  results  show  that   28   preferential  responding  to  a  caregiver  over  a  stranger  is  only  apparent  following   29   brief  isolation.  In  initial  episodes,  wolf  pups  show  no  differentiation  between  the   30   caregiver  and  the  stranger;  however,  following  a  2-­‐min  separation,  the  pups  show   31   proximity  seeking,  more  contact,  and  more  greeting  to  the  caregiver  than  the   32   stranger.  These  results  suggest  intensive  human  socialization  of  a  wolf  can  lead  to   33   attachment-­‐like  responding  to  a  human  caregiver  during the first two months of a 34   wolf pup's life.     35   Keywords:  Attachment,  wolves,  Canis  lupus,  Strange  Situation  Test,  domestication   36         37   3   1. Introduction 38   Attachment  behaviour  refers  to  any  “affectional  tie”  that  one  individual,  be  it   39   human  or  non-­‐human  animal,  displays  towards  another  specific  individual   40   (Ainsworth  and  Bell,  1970).  According  to  Ainsworth  and  Bell  (1970,  p.  50)  “The   41   behavioural  hallmark  of  attachment  is  seeking  to  gain  and  to  maintain  a  certain   42   degree  of  proximity  to  the  object  of  attachment,  which  ranges  from  close  physical   43   contact  under  some  circumstances  to  inter-­‐action  or  communication  across  some   44   distance  under  other  circumstances.”  To  help  explain  the  origins  and  function  of   45   attachment  behaviour,  Bowlby  and  Ainsworth  formulated  a  framework  for   46   attachment  that  posited  the  attachment  to  a  caregiver  is  critical  for  the  survival  of   47   infants  of  many  species  since  caregiver  proximity  can  function  as  protection  against   48   predators  (Bowlby,  1958;  Bowlby,  1982;  for  a  review  see  Bretherton,  1992  or   49   Kraemer,  1997).  This  perspective  incorporated  the  findings  from  the  primate   50   literature  that  highlighted  the  importance  of  mother  care  for  the  healthy   51   development  of  rhesus  monkeys  and  the  readiness  with  which  infant  monkeys  will   52   form  attachments  even  to  inanimate  mother  surrogates  (Harlow,  1971;  Kraemer,   53   1997).     54       55   the  dog  human-­‐caregiver  relationship.    Topál  et  al.  (1998)  were  the  first  to  adapt  the   56   Ainsworth’s  Strange  Situation  Test  (SST;  Ainsworth  &  Bell,  1970)  to  assess  whether   57   adult  pet  dogs  show  attachment  to  their  human  owners.    In  the  SST,  the  subject  is   58   brought  into  a  novel  room.  Then,  in  a  series  of  brief  episodes,  the  presence  of  the   More  recent  attachment  research  has  extended  the  attachment  framework  to     4   59   caregiver  and  a  stranger  is  systematically  manipulated.  A  brief  isolation  episode  also   60   occurs  approximately  halfway  through  the  test,  which  typically  leads  to  mild   61   distress.  Observers  then  score  the  subject’s  response  to  the  presence  and  absence  of   62   the  stranger  and  caregiver  to  assess  attachment-­‐related  behaviours  towards  the   63   human  caregiver.  Topál  et  al.  (1998)  recorded  the  amount  of  physical  contact   64   between  the  dog  and  owner  and  dog  and  stranger  in  addition  to  how  often  the  dog   65   engaged  in  play,  exploration,  passive  behaviour,  or  waiting  at  the  door  in  the   66   owner’s  or  stranger’s  absence.  They  found  that  the  dog-­‐owner  relationship  could  be   67   described  in  terms  of  attachment  between  the  dog  and  owner,  as  some  dogs  showed   68   the  secure-­‐base  effect  in  which  exploration  increased  in  the  presence  of  the  owner   69   compared  to  the  stranger.    In  addition,  dogs  were  shown  to  span  a  variety  of   70   attachment  styles  along  the  secure-­‐insecure  dimension,  which  is  similar  to  human   71   child  attachment  classifications  (Topál  et  al.,  1998).       72   Topál  et  al.  (2005)  explored  the  possible  effects  of  domestication  on  dogs’   73   formation  of  attachment  to  human  caregivers  by  comparing  the  attachment   74   behaviour  of  16-­‐week  old  hand-­‐reared  wolves,  hand-­‐reared  dogs,  and   75   conventionally  reared  dogs  (i.e.  mother  nursed  in  human  homes)  during  an  SST.   76   Dogs  that  were  raised  in  human  homes  (conventionally  reared  or  hand-­‐reared)   77   showed  greater  responding  to  a  human  caregiver  than  a  stranger,  whereas  hand-­‐ 78   reared  wolf  pups  showed  equal  responding  to  the  caregiver  and  stranger.  The   79   authors  of  this  study  suggested  that,  through  domestication,  dogs  might  have   80   evolved  “a  capacity  for  attachment  to  humans  that  is  functionally  analogous  to  that     5   81   present  in  human  infants”  (Topál  et  al.,  2005,  pp.  1373),  whereas  wolf  pups  did  not   82   appear  to  form  this  same  attachment  to  their  human  caregiver.     83   However,  recent  research  has  brought  to  light  the  importance  of  socialization   84   procedures,  and  experimental  methodology  in  behavioural  comparisons  between   85   dogs  and  wolves.  For  example,  adult  wolves,  once  thought  to  be  incapable  of   86   following  human’s  points,  are  now  known  to  be  as  responsive  to  human  gestures   87   and  attentional  state  as  pet  dogs  given  equivalent  rearing  and  testing  conditions   88   (Gacsi  et  al.,  2009;  Udell  et  al.,  2008).  Thus  the  hypothesis  that  dogs  display  a  unique   89   attachment  mechanism  to  form  attachments  to  humans,  distinct  from  that  displayed   90   by  other  mammals  (e.g.  Cairns,  1966;  Harlow,  1971;  Kraemer,  1997)  warrants   91   further  investigation.     92   Human  infants  start  to  use  their  mother  as  a  secure  base  when  exploring  the   93   environment  at  the  age  of  eight  months;  however,  from  the  second  year  on,  their   94   attachment  behaviour  becomes  more  flexible  and  they  will  be  less  dependent  on  the   95   presence  of  their  mother  when  interacting  with  others  (Bowlby,  1969).  Perhaps   96   wolves  may  be  more  likely  to  show  a  caregiver  preference  in  a  novel  situation  at  a   97   younger  age  than  the  16  weeks  tested  by  Topál  et  al  (2005).  It’s  unclear  whether  a   98   wolf’s  attachment  to  a  human  changes  with  age,  but  if  wolves  do  form  attachments   99   to  a  human  caregiver,  it  may  be  most  apparent  at  a  younger  age  when  the  wolves   100   may  require  the  presence  of  a  caregiver  to  be  comfortable  and  explore  a  novel   101   situation.  Thus  attachment  in  wolves  may  be  most  apparent  when  wolves  are  first     6   102   starting  to  emerge  from  the  den  around  three  weeks  of  age  (Packard,  Mech,  &  Ream,   103   1992).       104   In  addition,  it  is  also  important  to  note  that  at  the  time  of  testing,  the  wolves   105   tested  by  Topál  et  al.  (2005)  were  no  longer  living  with  their  human  caretaker,  but   106   had  been  relocated  to  a  private  wolf  farm  between  2-­‐4  months  of  age  (see  Virányi  et   107   al.,  2008).    As  a  result,  at  the  time  of  testing,  interactions  with  their  caretaker  had   108   been  reduced  to  half  a  day  twice  per  week  (Virányi  et  al.,  2008).    Reduced  levels  of   109   caretaker-­‐wolf  contact  may  have  altered  the  attachment  relationship  during  this   110   period,  which  may  have  contributed  to  the  study’s  findings  (Udell  &  Wynne,  2010).   111   The  aim  of  the  present  study  was  to  investigate  whether  human-­‐raised  wolf   112   pups,  still  experiencing  around  the  clock  interactions  with  their  primary  caregiver,   113   would  show  an  attachment  response  to  that  caregiver  on  the  SST.  Recent  research   114   with  dogs  in  the  SST  has  introduced  a  counterbalanced  version  of  the  SST   115   controlling  for  the  order  in  which  the  owner  and  stranger  entered  and  exited  the   116   room  (episode  order;  Palmer  &  Custance,  2008).  While  Palmer  and  Custance  (2008)   117   confirmed  that  adult  dogs  show  attachment  behaviours  towards  their  owners,  it  was   118   also  found  that  episode  order  could  significantly  influence  a  dog’s  response  towards   119   their  owner.  Rehn  et  al.  (2013)  further  investigated  order  effects  within  the  SST  in   120   dogs  by  implementing  a  control  condition  in  which  two  equally  unfamiliar   121   individuals  entered  and  exited  the  room  as  they  would  in  the  normal  SST.  Here,  the   122   only  difference  between  the  two  individuals  was  the  order  in  which  they  entered   123   and  exited  the  room.  Rehn  et  al.  found  that  dogs  displayed  attachment-­‐like     7   124   behaviour  to  one  of  the  unfamiliar  people  simply  as  a  function  of  the  order  in  which   125   the  unfamiliar  persons  entered  and  exited.  However,  exploration  was  more   126   susceptible  to  this  order  effect  than  proximity-­‐seeking  behaviours  such  as  initiating   127   contact.             128   In  the  present  study,  we  therefore  use  a  counterbalanced  version  of  the  SST   129   to  test  10  human-­‐reared  wolf  pups’  attachment-­‐like  behaviour  to  a  human   130   caregiver.  Pups  were  tested  three  times,  once  each  at  3,  5  and  7  weeks  of  age,   131   throughout  which  time  the  pups  were  receiving  near  24-­‐hour  care  from  a  human   132   caregiver.       133   2. Methods 134   2.1  Subjects   135   Ten  wolf  pups  (Canis  lupus)  from  two  litters  (one  litter  of  four  and  one  litter   136   of  six)  participated  in  the  present  experiments.  They  were  removed  from  the  den   137   when  they  were  approximately  10  days  of  age  and  hand-­‐reared  according  to  the   138   procedures  outlined  in  Klinghammer  and  Goodman  (1987)  by  two  human-­‐ 139   caregivers  at  Wolf  Park  in  Battle  Ground,  IN  (see  Table  1  for  subject  information).   140   The  hand-­‐rearing  procedure  involved  the  presence  of  a  human  caregiver  in  an   141   indoor  room  for  24  hours  a  day  with  the  pups  for  the  first  1.5-­‐  2  months  of  life,  at   142   which  point  the  caregivers  were  present  for  approximately  16  hours  a  day.   143   Caregivers  were  also  responsible  for  bottle-­‐feeding  the  pups  every  4-­‐6  hours  until     8   144   the  pups  were  able  to  eat  solid  foods.  Testing  procedures  were  approved  by  the   145   University  of  Florida  Institutional  Animal  Care  and  Use  Committee.     146   2.2  General  procedure     147   Wolf  pups  were  given  a  modified  version  of  the  Ainsworth  Strange  Situation   148   Test  (detailed  below)  during  their  3rd,  5th  and  7th  week  of  life  (see  Table  1  for   149   exact  ages).  At  each  age,  a  novel  testing  room  and  a  novel  stranger  were  used.    The   150   caregiver  remained  the  same  across  ages.     151   In  total,  nine  subjects  were  tested  during  week  3,  nine  during  week  5,  and  ten   152   during  week  7.  One  subject  was  ill  during  week  3  and  5  and  was  only  tested  at  7   153   weeks  of  age.  One  additional  subject’s  last  two  episodes  from  week  3  were  excluded   154   due  to  an  experimenter  error  in  which  the  episode  order  was  inverted  for  the  last   155   two  sessions.     156   Each  novel  testing  room  was  an  indoor  space  (approximately  18  m2)  to  which   157   the  pups  had  never  previously  been  exposed.  In  each  testing  room,  two  2m-­‐ 158   diameter  non-­‐overlapping  circles  were  marked  on  the  floor  with  tape.  The  marked   159   circles  were  used  to  code  proximity  to  the  caregiver  or  stranger  by  having  the   160   stranger  and  caregiver  sit  in  the  centre  of  each  circle.    Approximately  six  toys  were   161   distributed  between  the  two  circles.  Toys  were  not  included  for  the  testing  at  5  and   162   7  weeks  of  age  for  litter  two  due  to  experimenter  error.       163   164   9   2.3 Modified Strange Situation Test The  pup  was  brought  to  a  novel  room  where  the  presence  and  absence  of  a   165   caregiver  and  stranger  were  manipulated  over  six  episodes  each  lasting  2  minutes.   166   Each  time  the  stranger  and/or  caregiver  entered  the  room;  they  opened  and  closed   167   the  door,  slowly  walked  to  the  centre  of  one  of  the  circles  and  sat  down  on  the  floor.   168   The  circle  the  caregiver  or  stranger  sat  in  was  determined  randomly  prior  to  the   169   start  of  the  test.  During  the  episode,  the  pups  were  free  to  move  about  the  room   170   without  restriction.  If  the  pup  approached  the  caregiver  or  stranger  in  the  circle  and   171   initiated  contact,  the  caregiver  or  stranger  would  pet  the  pup.  If  the  pup  initiated   172   play  by  bringing  a  toy  to  either  the  caregiver  or  stranger,  the  caregiver  or  stranger   173   could  engage  in  play.    The  stranger  and  caregiver,  however,  were  instructed  not  to   174   move  outside  their  circle  during  an  episode.  To  exit  at  the  end  of  an  episode,  the   175   caregiver  or  stranger  stood  up,  turned  to  the  door,  and  slowly  walked  towards  it.   176   Upon  reaching  the  door,  the  caregiver  or  stranger  said  "goodbye"  and  exited.     177   To  control  for  potential  order  effects,  two  counterbalanced  sequences  of  the   178   entering  and  exiting  of  the  stranger  and  caregiver  were  utilized.  Table  2  outlines   179   these  two  episode  orders  and  indicates  whether  the  stranger  alone,  caregiver  alone,   180   stranger  and  caregiver,  or  neither  was  in  the  room  with  the  pup.    For  Episode  Order   181   1,  the  caregiver  sat  alone  in  the  room  with  the  pup  for  the  first  episode.  After  2  mins,   182   a  stranger  entered  the  room  and  sat  in  the  adjacent  circle  to  the  caregiver  for  the   183   second  episode.  Next,  the  caregiver  left  the  room,  leaving  the  stranger  and  pup  alone   184   in  the  room  for  Episode  3.    For  Episode  4,  the  strange  left  the  room,  leaving  the  pup     10   185   alone.  In  Episode  5  the  stranger  returned  to  the  room.  In  the  final  episode,  the   186   caregiver  entered  so  the  caregiver  and  stranger  were  present  with  the  pup.  Episode   187   Order  2  followed  a  similar  pattern  except  that  it  counterbalanced  Episode  Order  1   188   (see  table  2).  The  episode  order  assigned  for  each  pup  was  pseudo-­‐randomly   189   determined  so  that  at  each  age,  half  of  the  pups  were  tested  with  each  order.  In   190   addition,  the  order  each  pup  was  tested  with  was  changed  across  the  three  testing   191   weeks  so  that  each  pup  was  tested  once  with  one  episode  order,  and  twice  with  the   192   other  episode  order.     193   2.4 Behaviour coding 194   During  each  episode,  the  pups’  behaviour  was  recorded  on  video  for   195   subsequent  analysis.  The  behaviours  scored,  the  behavioural  definition,  and   196   observer  agreement  scores  are  listed  in  Table  3.    Briefly,  during  each  episode  we   197   scored  the  amount  of  time  the  pup  spent  in  proximity  (within  the  2m  circle)  and   198   within  physical  contact  of  the  caregiver  and  stranger.  These  two  behaviours  were   199   not  mutually  exclusive:  a  pup  could  be  in  contact  while  also  in  proximity.  Both   200   proximity  and  contact  behaviours  were  scored  to  assess  the  approach  and   201   investigative  behaviour  (proximity)  as  well  as  close  contact  seeking  (contact).  Both   202   are  important,  as  pups  could  prefer  to  approach  and  investigate  one  individual  (high   203   proximity),  but  prefer  not  to  be  touched  (low  contact).  This  would  indicate  a  fearful   204   curiosity,  whereas  a  high  proximity  high  contact  would  be  more  indicative  of   205   comfort  seeking.  We  also  scored  whether  the  pups  greeted  and  followed  the   206   caregiver  and  stranger  when  entering  or  exiting  the  room,  with  a  zero  indicating  no     11   207   greeting  or  following,  a  one  indicating  a  calm  greeting  or  following,  and  a  two   208   indicating  an  excited  greeting  or  follow  (see  Table  3).  Scored  episode  times  were   209   approximately  2  min;  however,  due  to  minor  variations  in  time  taken  for  the  human   210   to  enter,  sit  down  or  exit  across  episodes,  all  behaviours  are  reported  as  a   211   proportion  of  the  episode  time,  except  for  greeting  and  following  which  were  rated   212   categorically.  A  second  observer  scored  37%  of  the  videos.  Percent  agreement  was   213   calculated  for  the  continuous  behaviours  and  Cohen’s  Kappa  was  calculated  for  the   214   categorical  scale  by  comparing  the  two  raters’  scores  on  an  episode-­‐by-­‐episode   215   basis.  For  the  continuous  behaviours,  an  agreement  was  defined  as  both  observers   216   scoring  within  two  seconds  (or  5%  of  the  scored  time)  of  each  other.  Any  larger   217   discrepancy  was  scored  as  a  percent  agreement  by  dividing  the  smaller  scored  time   218   by  the  larger.     219   2.5 Statistical analyses 220   Data  were  analysed  using  the  statistical  package  R  (R  Core  Team,  2012)  and   221   plotted  with  the  R  package  ggplot2  (Wickham,  2009).  Linear  mixed  effects  models   222   using  the  lme4  package  (Bates,  Maechler,  Bolker    &  Walker,  2013)  were  used  to   223   assess  the  effects  of  the  presence  and  absence  of  the  caregiver  and  stranger  on  the   224   subjects’  behaviour.  P-­‐values  for  ANOVA  tests  and  t-­‐tests  were  generated  from  the   225   LmerTest  package  (Kuznetsova,  Brockhoff  &  Christensen,  2013)  using  a   226   Satterthwaite  approximation  for  the  degrees  of  freedom.  Each  model  included   227   random  intercepts  for  the  subject  and  litter  variables.  The  subject  term  was  nested   228   in  the  litter  term.     229   230   12   3. Results Linear  mixed  effects  models  were  run  for  each  scored  behaviour  to   231   investigate  the  effect  of  the  episode  and  condition  on  that  behaviour.  To  test   232   whether  pups  engaged  in  differential  levels  of  a  scored  behaviour  towards  the   233   caregiver  compared  to  the  stranger,  linear  models  included  a  dummy  coded  variable   234   to  indicate  whether  the  behaviour  occurred  towards/  in  the  presence  of  the   235   caregiver  or  the  stranger.     236   3.1 Proximity seeking to the caregiver and stranger 237   Figure  1  shows  the  overall  patterns  of  proximity  to  the  caregiver  and   238   stranger  across  episodes  separated  by  age  and  episode  order,  and  an  overall   239   averaged  summary  across  weeks.  Overall,  similar  patterns  of  responding  were   240   observed  at  each  age  of  testing  (3,  5,  7  weeks);  however,  differences  appeared  in   241   proximity  to  the  stranger  and  caretaker  across  the  episodes.       242   3.1.1.  Episodes  2  &  6:  During  Episode  2  and  Episode  6,  both  the  caregiver  and   243   stranger  were  present  in  the  room.  A  linear  mixed  effect  model  was  used  to  test   244   whether  pups’  proximity  to  a  person  was  predicted  by  the  familiarity  of  that  person   245   (caregiver  vs.  stranger),  the  pups’  age  (3,  5  or  7  weeks),  the  episode  order  (Order  1   246   or  Order  2),  a  2-­‐way  interaction  between  the  episode  (2  vs.  6)  and  the  person   247   (caregiver  vs.  stranger),  and  a  2-­‐way  interaction  between  the  episode  and  age.     248   There  was  no  interaction  between  age  and  episode  (F(1,98.95)  =  0.05,  p  =  0.82),  but   249   there  was  a  significant  interaction  between  the  episode  and  proximity  to  the     13   250   caregiver  and  the  stranger  (  F(1,98.95)  =  12.60,  p  <  0.001),  indicating  that  preference   251   for  the  caregiver  and  stranger  changed  from  Episode  2  to  Episode  6.     252   We  therefore  explored  the  pups’  preference  between  the  caregiver  and   253   stranger  during  Episode  2  and  Episode  6.  For  Episode  2,  there  was  no  indication   254   that  pups  had  different  preferences  for  approaching  the  caregiver  over  the  stranger   255   depending  on  the  episode  order  (Person  by  episode  order  Interaction:  F(1,  41.22)  =   256   0.11,  p  =  0.74).  There  was  also  no  effect  of  age  (F(1,  44.02)  =  0.40,  p=  0.53),  episode   257   order  (F(1,  47.75)  =  0.70,  p  =  0.41),  or  difference  between  the  caregiver  and  stranger   258   (F(1,  41.22)  =  2.61,  p  =  0.11).    This  contrasts  the  findings  in  Episode  6,  which  showed  a   259   significant  interaction  between  the  episode  order  and  the  person  the  pup   260   approached  (F(1,  46.98)  =  7.77,  p  <  0.01),  indicating  that  pups’  preference  for  the   261   caregiver  depended  on  the  episode  order.  When  looking  at  each  episode  order   262   separately,  there  was  a  significant  preference  for  proximity  to  the  caregiver  over  the   263   stranger  (F(1,  24.99)  =  16.73,  p  <  0.0001)  in  Episode  Order  1.  For  Episode  Order  2   264   where  the  caregiver  was  already  present  in  Episode  5,  there  was  no  significant   265   difference  in  time  spent  with  the  caregiver  and  stranger  in  Episode  6  (F(1,  20.99)  =   266   0.05,  p  =  0.82).  In  both  episode  orders,  there  was  again,  no  effect  of  age  (Order  1:  F(1,   267   24.99)  =  0.05,  p  =  0.83;  Order  2:  F(1,  20.99)  =  0.60,  p  =  0.45).     268   3.1.2.  Episodes  1,  3  &  5:  Given  that  there  was  no  indication  of  an  age  effect   269   across  Episodes  2  and  6,  the  data  were  averaged  across  age  to  provide  a  complete   270   within-­‐subject  data  set.  There  was  a  significant  effect  of  episode  (F(2,  50.98)  =  8.00,  p  <   271   0.001)  showing  that  pups  spent  more  time  in  proximity  to  both  the  caregiver  and     14   272   stranger  in  Episode  5  than  Episode  1  (t50.98  =  3.99,    p  <  0.001).  When  proximity  was   273   averaged  across  Episodes  1,  3  and  5,  pups  overall  spent  more  time  in  proximity  to   274   the  stranger  (t50.99  =  2.07,  p  =  0.043).    When  considering  Episode  5  alone  (the  first   275   reunion  following  isolation),  however,  there  was  no  difference  in  proximity  between   276   the  stranger  and  caregiver  (t6.83  =  0.989,  p  =  0.36).   277   3.2 Contact between the caregiver and stranger 278     Figure  2  shows  the  proportion  of  each  episode  the  pup  made  physical  contact   279   (i.e.  petting)  with  the  caregiver  and  stranger  across  episodes  for  both  episode   280   orders  across  all  three  weeks  of  testing.  The  overall  patterns  are  similar  to   281   proximity  seeking,  with  contact  changing  as  a  function  of  the  episode.  Again,  there   282   appears  to  be  little  effect  of  testing  across  ages.     283     284   proximity  seeking.  During  Episode  2,  pups  showed  no  preference  between  the   285   caregiver  and  stranger  (F(1,48.98)  =  3.07,  p  =  0.09  ).  During  Episode  6,  however,   286   preference  between  the  caregiver  and  stranger  depended  on  the  episode  order   287   (Episode  by  person  interaction:  F(1,46.98)  =  6.93,  p  =  0.01).  Pups  in  episode  Order  1   288   significantly  preferred  the  caregiver  (F(1,24.98)  =  20.79,  p  <  0.001),  whereas  there  was   289   no  difference  in  time  with  the  caregiver  and  stranger  for  Order  2  when  pups  were   290   already  re-­‐united  with  the  caregiver  in  Episode  5  (F(1,21)  =  0.05,  p  =  0.83).    Similar  to   291   proximity,  there  was  no  effect  of  age  (F(1,  98.95)  =  0.11,  p  =  0.74),  or  interaction   292   between  age  and  episode  (F(1,  98.95)  =  0.05,  p  =  0.82).       3.2.1  Episodes  2  &  6:  Overall,  contact  seeking  showed  an  identical  pattern  to     293   15   3.2.2.  Episodes  1,  3  &  5:    Pups  overall  showed  the  same  pattern  of  contact   294   with  the  stranger  and  caregiver  as  they  did  for  proximity.  Pups  overall  showed   295   significantly  more  contact  in  Episode  5  then  Episode  1  (t50.97=  4.44,  p<  0.0001)  and   296   more  contact  with  the  stranger  when  averaged  across  episodes  1,3,  and  5  than  with   297   the  caregiver  (t50.98=2.05,  p=0.046).  When  looking  at  Episode  5  alone  (the  first   298   reunion  following  isolation),  however,  there  was  no  significant  difference  in  time   299   spent  in  contact  with  the  caregiver  or  stranger    (t8.90  =  0.30,  p  =  0.77).     300   3.3. Greeting and following the caregiver and stranger 301     302   which  occurred  during  Episodes  2,  5  and  6.  Following  was  scored  when  the   303   caregiver  or  stranger  exited  a  room,  which  occurred  during  Episodes  3  and  4.  A   304   mean  greeting  and  following  score  was  computed  for  each  pup  across  all  ages.   305   Figure  3  shows  the  mean  score  for  following  and  greeting  both  the  stranger  and   306   caregiver.  Overall,  pups  were  not  more  likely  to  follow  the  caregiver  over  the   307   stranger  (F(1,  9.00)=  1.18,  p  =  0.31).  The  pups,  however,  did  show  more  excited   308   greetings  to  the  caregiver  than  they  did  to  the  stranger  (F(1,  9.00)=  6.40,  p  =  0.03)   309   upon  their  return.     310   4. Discussion 311   Greetings  were  scored  when  the  caregiver  or  stranger  entered  the  room,   Overall,  the  pups  showed  differential  responding  to  the  caregiver  when   312   compared  to  the  stranger.  Pups  were  more  likely  to  greet  the  caregiver  with  whines   313   and  ears  back  upon  reunion  than  they  did  the  stranger.  In  addition,  pups  showed  an     16   314   effect  of  reunion  in  Episode  6,  seeking  greater  proximity  and  physical  contact  with   315   the  caregiver  than  the  stranger.    However,  this  effect  was  only  evident  when  pups   316   were  tested  with  episode  Order  1.  This  is  likely  due  to  the  fact  that  the  caregiver  was   317   already  present  during  the  post-­‐isolation  Episode  5  in  Order  2.  This  provided  time   318   for  the  pups  to  engage  in  reunion  behaviour  during  Episode  5,  followed  by  increased   319   exploration  of  other  environmental  features,  including  the  unfamiliar  human,  in   320   Episode  6.  In  fact,  this  is  what  is  predicted  when  a  secure  attachment  is  present,   321   known  as  the  secure  base  effect  (Ainsworth  &  Bell,  1970).    However  in  Order  1,   322   Episode  6  was  the  first  reunion  with  the  caregiver,  which  led  to  a  highly  significant   323   bias  for  the  caregiver-­‐  an  outcome  also  predicted  in  previous  attachment  literature   324   (Ainsworth  &  Bell,  1970).         325     326   and  the  stranger  prior  to  isolation  (Episode  4).  Instead,  pups  were  indifferent  in   327   Episode  2  when  given  a  choice  between  these  two  people,  and  in  fact  showed  a   328   slight  stranger  preference  for  contact  across  Episodes  1,  3  and  5.  However,  after  a   329   brief  2-­‐minute  isolation  phase,  the  pups  showed  a  strong  caregiver  preference  in   330   Episode  6  if  the  caregiver  returned,  but  the  pups  showed  no  preference  for  the   331   stranger  if  the  stranger  returned  in  Episode  6.  Interestingly,  however,  we  did  not   332   observe  any  differences  in  proximity  and  contact  between  the  stranger  and   333   caregiver  during  Episode  5,  the  episode  immediately  after  isolation.  One  possibility   334   is  that  because  this  comparison  was  across  episode  orders  (Order  1  vs.  Order  2)  and   335   averaged  across  ages  allowing  order  effects  or  minor  age  effects  to  potentially  mask   336   an  effect.  Another  explanation  is  that  isolation  in  Episode  4  was  sufficiently  stressful   We  did  not  observe  many  differences  in  preference  between  the  caregiver     17   337   that  it  activated  general  proximity  seeking  to  any  available  person,  even  if  the  only   338   available  person  was  not  a  caretaker.  Importantly,  however,  if  the  caregiver  was   339   present  is  Episode  5,  the  pups  showed  indifference  between  the  stranger  and   340   caregiver  in  Episode  6.  If  the  stranger  was  present  in  Episode  5,  the  pups  showed  a   341   dramatic  caregiver  preference  in  Episode  6  indicating  that  the  effect  of  the  presence   342   of  the  caregiver  or  stranger  in  Episode  5  was  not  equivalent.  Only  if  the  caregiver   343   was  present  in  Episode  5,  did  the  pups  show  the  same  pattern  that  was  shown   344   before  isolation  in  Episode  2.  This  suggests  that  despite  the  pups  showing  similar   345   time  in  contact  and  proximity  with  the  caregiver  and  stranger  in  Episode  5,  only   346   contact  and  proximity  with  the  caregiver  in  Episode  5  functioned  to  return  the  pup   347   to  baseline  preferences.     348    The  fact  that  caregiver  preferences  only  occurred  in  Episode  6  exemplifies   349   Ainsworth  and  Bell’s  (1970)  hypothesis  that  attachment  behaviour  is  heightened  in   350   situations  perceived  as  threatening,  which  in  the  present  case,  was  isolation.  One   351   difference;  however,  in  our  findings  from  those  observed  with  human  infants  and   352   dogs  is  that  separation  from  the  attachment  figure  alone  was  not  sufficient  to   353   activate  attachment  behaviours  (e.g.  Ainsworth  &  Bell,  1970,  Palmer  &  Custance,   354   2005;  Topál  et  al.,  1998;  Topál  et  al.,  2005).  In  the  present  study,  complete  isolation   355   (Episode  4)  was  required  to  activate  attachment  behaviours.  However,  this  could   356   have  been  a  by-­‐product  of  the  rearing  and  socialization  practices  employed  with  the   357   pups  used  in  this  study.  The  pups  were  regularly  introduced  to  novel  humans  and   358   environments  (Klinghammer  &  Goodman,  1987),  making  it  unsurprising  that  being   359   in  a  room  with  a  novel  human  was  not,  in  itself,  a  strange  or  stressful  situation.       360   18   Thus,  this  study  provides  evidence  that  the  behaviour  demonstrated  by   361   hand-­‐raised  wolf  pups  towards  humans  can  be  categorized  as  attachment   362   (Ainsworth  and  Bell,  1970)  in  some  cases,  given  early  socialization  to  humans   363   (Klinghammer  and  Goodman,  1987),  with  continued  contact  with  the  caregiver   364   through  the  time  of  testing,  and  the  implementation  of  methodological  controls  for   365   known  order  effects  (Palmer  &  Custance,  2008;  Rehn  et  al.,  2013).    This  of  course   366   does  not  mean  that  all  wolf  pups  will  necessarily  show  strong  attachment  behaviour   367   towards  humans  (e.g.,  Topal  et  al.,  2005),  as  early  rearing  history  and  differences  in   368   caretaker  behaviour  are  known  to  influence  both  the  initial  formation  of  the   369   attachment  bond  and  the  attachment  style  that  develops  between  an  individual  and   370   their  caretaker  (Ainsworth  and  Bell,  1970).  The  fact  that  the  wolves  tested  by  Topál   371   et  al.  (2005)  were  older,  less  human  dependent,  and  no  longer  living  with  their   372   human  caretaker  at  the  time  of  testing  could  have  altered  their  attachment   373   behaviour.     374   At  present,  however,  we  cannot  determine  which  if  any  of  these  factors   375   contributed  to  the  differences  between  the  present  study  and  Topál  et  al  (2005).   376   Age  may  be  a  significant  factor  (16  weeks  vs.  3,5  &  7  weeks),  however,  differences  in   377   the  length  of  time  spent  with  the  pups  on  a  daily  basis,  time  spent  overall  during  the   378   subject’s  lifetime,  or  other  unknown  rearing  differences  could  have  contributed  to   379   the  differences  between  the  two  studies.  It  is  unknown  what  effect  age  may  have  on   380   attachment  to  humans.  In  our  limited  age  range  of  testing,  we  saw  no  effect;   381   however,  we  may  have  observed  an  effect  had  testing  been  carried  out  until  16   382   weeks  of  age.    Future  studies  are  necessary  to  determine  the  typical  developmental     19   383   stages  of  wolf  attachment  to  humans  and  the  rearing  factors  that  may  influence  it.   384   Early  socialization  and  life  experiences  may  likely  influence  attachment  test   385   performance  for  candis.  In  fact,  many  feral  dogs  actively  avoid  human  contact  in  the   386   absence  of  early  socialization  (Ortolani,  Vernooij  &  Coppinger,  2009). Future  studies   387   on  the  development  of  attachment  bonds  in  canids  may  carefully  detail  the  ontogeny   388   of  attachment  formation  and  the  conditions  that  lead  to  its  development  and   389   maintenance  in  later  life  allowing  for  further  comparisons  between  wolves  and   390   dogs.     391   Overall,  the  results  show  that  wolf  pups  will  form  attachments  to  their   392   human  caregivers.  This  is  an  important  finding,  as  it  suggests  that  capacity  to  form   393   attachments  to  humans  is  not  itself  a  product  of  domestication.  This  suggests  that   394   young  non-­‐domesticated  canids  can  form  attachments  to  humans.  Our  study  also   395   suggests  that  the  conditions  under  which  attachment  behaviour  is  displayed,  for   396   example,  that  isolation  is  required  to  elicit  attachment  like  responding,  may  differ   397   between  wolves  and  what  is  seen  with  dogs.  Another  area  for  future  investigation  is   398   looking  into  the  maintenance  of  attachment  into  adulthood.  Although  it’s  unclear   399   whether  the  differences  in  results  from  Topál  et  al.  (2005)  and  the  present  one  is   400   due  to  the  age  of  the  subjects  at  testing,  it’s  possible  that  domestication  influenced   401   how  attachments  are  maintained  throughout  development  and  into  adulthood.   402   When  tested,  most  adult  dogs  typically  show  attachments  to  their  owners;  it’s  not   403   clear  whether  this  would  be  the  case  for  wolves  or  even  all  breeds  of  dogs.  Thus,  our   404   results  highlight  that  the  ability  to  form  attachments  to  humans  likely  proceeded   405   domestication,  but  domestication  may  have  changed  the  ease  at  which  these     20   406   attachments  could  be  formed,  the  conditions  under  which  they  are  shown,  and  how   407   they  are  maintained  as  adults.     408   Acknowledgments 409     410   throughout  the  project.  We  thank  Pat  Goodman,  Dana  Drenzek,  Monty  Sloan,  the   411   many  volunteers,  and  the  many  interns  that  helped  in  this  project.     412     We  thank  the  staff  at  Wolf  Park  in  Battleground  Indiana  for  assistance       21   413   References 414   Ainsworth, M. D., & Bell, S. M. (1970). Attachment, exploration and separation: 415   Illustrated by the behaviour of one-year-olds in a strange situation. Child 416   Development, 41(1), 49–67. 417   Ainsworth, M. D. S. 1978. Patterns of attachment; A Psychological Study of the 418   Strange Situation. Lawrence Erlbaum Associates, Inc., Publishers. 419   420   Bates, D., Maechler, M., Bolker, B., & Walker, S. (2013). lme4: 421   Linear mixed-effects models using Eigen and S4. R package version 1.0-5. 422   http://CRAN.R-project.org/package=lme4 423   424   425   426   427   428   429   430   Bowlby, J. 1958, "The Nature of the Childs Tie to His Mother", International Journal of Psycho-Analysis, vol. 39, no. 5, pp. 350-373. Bowlby, J. 1969. Attachment and Loss: Volume 1 Attachment. Penguin Books Ltd. 1981. Bowlby, J. 1982, "Attachment and Loss - Retrospect and Prospect", American Journal of Orthopsychiatry, vol. 52, no. 4, pp. 664-678. Bretherton, I. (1992). The origins of attachment theory: John Bowlby and Mary Ainsworth. Developmental Psychology, 28(5), 759–775.   431   432   22   Cairns, R. B. (1966). Development, maintenance, and extinction of social attachment behavior in sheep. Journal of Comparative and Physiological Psychology, 62(2), 433   298–306. doi:10.1037/h0023692 434   Gacsi, M., Gyori, B., Viranyi, Z., Kubinyi, E., Range, F., Belenyi, B. & Miklosi, A. 435   (2009). Explaining Dog Wolf Differences in Utilizing Human Pointing Gestures: 436   Selection for Synergistic Shifts in the Development of Some Social Skills. PLOS 437   ONE, 4 438   Harlow, H. F., Harlow, M. K., & Suomi, S. J. (1971). From Thought to Therapy: Lessons 439   from a Primate Laboratory: How investigation of the learning capability of rhesus 440   monkeys has led to the study of their behavioral abnormalities and rehabilitation. 441   American Scientist, 59(5), 538–549. 442   Klinghammer, E. & Goodmann, P. A. (1987). Socialization and management of wolves 443   in captivity. In: Frank, H. (ed.) Man and Wolf: Advances, Issues, and Problems 444   in Captive Wolf Research. W. Junk, Dordrecht 445   Kraemer, G. W. (1997). Psychobiology of Early Social Attachment in Rhesus 446   Monkeys Clinical Implications. Annals of the New York Academy of 447   Sciences, 807(1), 401–418. doi:10.1111/j.17496632.1997.tb51935.x 448   Kuznetsova, A., Brockhoff, B., & Christensen, H.B. (2013). lmerTest: Tests for random 449   and fixed effects for linear mixed effect models (lmer objects of lme4 450   package). R packageversion 2.0-3. http://CRAN.R 451   project.org/package=lmerTest   452   23   Ortolani, A., Vernooij, H., & Coppinger, R. (2009). Ethiopian village dogs: Behavioural 453   responses to a stranger’s approach. Applied Animal Behaviour Science, 119(3-4), 454   210–218. doi:10.1016/j.applanim.2009.03.011 455   Packard, J. M., Mech, L. D., & Ream, R. R. (1992). Weaning in an arctic wolf pack: 456   behavioral mechanisms. Canadian Journal of Zoology, 70(7), 1269–1275. 457   doi:10.1139/z92-177 458   Palmer, R., & Custance, D. (2008). A counterbalanced version of Ainsworth’s Strange 459   Situation Procedure reveals secure-base effects in dog–human relationships. Applied 460   Animal Behaviour Science, 109(2-4), 306–319. doi:10.1016/j.applanim.2007.04.002 461   462   463   Range, F. & Virányi, Z. (2011). Development of Gaze Following Abilities in Wolves (Canis Lupus). PLoS ONE, 6, e16888. R Core Team (2012). R: A language and environment for statistical computing. R 464   Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, 465   URL: http://www.R-project.org/. 466   Rehn, T., McGowan, R. T. S., & Keeling, L. J. (2013). Evaluating the Strange Situation 467   Procedure (SSP) to Assess the Bond between Dogs and Humans. PLoS ONE, 468   8(2), e56938. doi:10.1371/journal.pone.0056938 469   Topál, J., Gácsi, M., Miklósi, Á., Virányi, Z., Kubinyi, E., & Csányi, V. (2005). 470   Attachment to humans: a comparative study on hand-reared wolves and differently 471   socialized dog puppies. Animal Behaviour, 70(6), 1367–1375. 472   doi:10.1016/j.anbehav.2005.03.025   473   24   Topal, J., Miklosi, A., Csanyi, V., & Dka, A. (1998). Attachment Behaviour in Dogs 474   (Canis familiaris ): A New Application of Ainsworth’s ( 1969 ) Strange Situation 475   Test. Journal of Comparative Psychology, 112(3), 219–229. 476   Udell, M. A., & Wynne, C. D. (2010). Ontogeny and phylogeny: both are essential to 477   human-sensitive behaviour in the genus Canis. Animal Behaviour, 79(2), e9 478   e14. 479   Udell, M. A. R., Dorey, N. R. & Wynne, C. D. L. (2010). What did domestication do to 480   dogs? A new account of dogs’ sensitivity to human actions. Biological Reviews, 481   85, 327–345. 482   Udell, M. A. R., Dorey, N. R. & Wynne, C. D. L. (2008). Wolves outperform dogs in 483   following human social cues. Animal Behaviour, 76, 1767–1773. 484   Udell, M. A. R., Spencer, J. M., Dorey, N. R. & Wynne, C. D. L. (2012). Human 485   Socialized Wolves Follow Diverse Human Gestures… And They May Not Be 486   Alone. International Journal of Comparative Psychology, 25, 97–117. 487   Virányi, Z., Gácsi, M., Kubinyi, E., Topál, J., Belényi, B., Ujfalussy, D., & Miklósi, Á. 488   (2008). Comprehension of human pointing gestures in young human-reared 489   wolves (Canis lupus) and dogs (Canis familiaris). Animal Cognition, 11(3), 373 490   387. 491   Wickham, H. (2009). ggplot2: elegant graphics for data analysis. Springer New York, 492   493   2009.     494   25     Subject Sex Age in days Age in days Age in days (Week 3) (Week 5) (Week 7) Litter Kanti M 2 23 37 50 Bicho M 2 23 36 50 Mowgli M 2 25 35 53 Pigeon F 2 24 37 51 Bigboy M 2 25 36 51 Fiona F 2 24 35 50 Dharma F 1 23+ 35 47 Devra1 F 1 NA NA 47 Gordon M 1 21 35 47 Tilly F 1 22 35 47 495   Table  1.  Subject  information.  Table  gives  sex  and  exact  age  at  each  testing  week.   496   1Devra  was  unable  to  be  tested  at  3  and  5  weeks  due  to  illness.    +Last  two  episodes   497   were  excluded  due  to  experimenter  error.     498         499   26     Order Episode 1 Episode 2 Episode 3 Episode 4 Episode 5 Episode 6 1 Caregiver Caregiver + Stranger Isolation Stranger Stranger + Returns Caregiver Caregiver Caregiver+ Returns Stranger Stranger 2 Stranger Stranger + Caregiver Caregiver Isolation 500     501   Table  2.    Outline  of  the  two  episode  orders.  Each  cell  displays  whether  the   502   caregiver,  the  stranger  or  both  were  present  in  the  testing  room  with  the  pup.     503         504   27     Behaviour Definition Per cent Cohen’s Agreement Kappa     Behaviours During an Episode Proximity seeking caregiver Proportion of the episode in which the pup had at 95.7% least 2 paws within the 2 m circle the Proximity seeking stranger caregiver/stranger was siting in.   Contact caregiver Proportion of the episode in which the pup 93.4% engaged in physical contact with the Contact stranger caregiver/stranger (not mutually exclusive with proximity) Behaviour Between Episodes   A score from 0-2 on the type of greeting that Greeting Caregiver occurred within 15 s of the caregiver/stranger entering the room and sitting: 0: “No greeting- did not approach” 0.71 1: “Calm Greeting-approached but did not display Greeting Stranger ears back or whining” 2: “Excited greeting- approached with ears back and whining”   28   A score from 0-2 on the type of following that Following Caregiver occurred within 15 s of the caregiver/stranger exiting the room. 0: “No following”- did not approach leaving person or door     0.63 1: “Calm follow”-followed leaving person but did not try to follow through door, jump on door or Following Stranger whine 2: “Excited follow”- followed person and tried to exit through door, jump at door or whined 505     506   Table  3.  Behavioural  coding:  definitions  for  each  behaviour.  For  continuous   507   variables,  per  cent  agreement  is  reported  from  36%  of  the  videos  double  coded  (10   508   of  28  videos).  For  categorical  variables,  Cohen’s  Kappa  is  reported  for  the  36%  of   509   double  coded  videos.   510     511     512       29   Order 1 Week 3 1.00 Order 2 0.75 0.50 0.25 0.00 Week 5 0.75 0.50 0.25 0.00 Caregiver 1.00 Week 7 Proportion of time in proximity 1.00 Stranger 0.75 0.50 0.25 0.00 Average 1.00 0.75 0.50 0.25 0.00 1 513   2 3 4 5 6 1 2 3 4 5 6 Episode   514   Figure  1.  Proportion  of  time  in  each  episode  spent  in  proximity  to  the  caregiver  and   515   stranger  in  the  Ainsworth  SST  procedure.  The  left  column  represents  subjects   516   experiencing  Episode  Order  1,  the  right  column  subjects  experiencing  episode  order   517   2.  Bars  indicate  the  mean  and  dots  show  each  data  point.    Each  row  shows  the   518   results  for  a  different  age  and  the  final  row  shows  the  results  averaged  across  all   519   ages.  Episodes  2  &  6  indicate  proximity  to  both  the  caregiver  and  stranger  as  both   520   were  in  the  room  (2  bars).  Episodes  1,  3  and  5  indicate  proximity  to  only  the   521   caregiver  or  stranger,  as  only  one  was  present  and  the  person  present  depended  on   522   the  episode  order.  Episode  4  is  blank,  as  neither  the  caregiver  nor  stranger  was   523   present  (i.e.  isolation).  The  Average  row  indicates  each  pups  proportion  in   524   proximity  averaged  for  the  episode  order  in  which  they  received  twice.       30   Order 1 Week 3 1.00 Order 2 0.75 0.50 0.25 0.00 Week 5 0.75 0.50 0.25 0.00 Caregiver 1.00 Week 7 Proportion of time in contact 1.00 Stranger 0.75 0.50 0.25 0.00 Average 1.00 0.75 0.50 0.25 0.00 1 525   2 3 4 5 6 1 2 3 4 5 6 Episode   526   Figure  2.  Proportion  of  episode  time  spent  in  contact  with  the  caregiver  and   527   stranger  in  the  Ainsworth  SST  procedure.  The  left  column  represents  subjects   528   experiencing  Episode  Order  1,  the  right  column  subjects  experiencing  Episode   529   Order  2.  Bars  indicate  the  mean  and  dots  indicate  each  data  point.    Each  row  shows   530   the  results  for  a  different  age  and  the  final  row  shows  the  results  averaged  across  all   531   ages.  Episodes  2  &  6  indicate  contact  with  both  the  caregiver  and  stranger  as  both   532   were  in  the  room  (2  bars).  Episodes  1,  3  and  5  indicate  proximity  to  only  the   533   caregiver  or  stranger,  as  only  one  was  present  and  the  person  present  depended  on   534   the  episode  order.  Episode  4  is  blank,  as  neither  the  caregiver  nor  stranger  was   535   present  (i.e.  isolation).  The  Average  row  indicates  each  pups  proportion  in   536   proximity  averaged  for  the  episode  order  in  which  they  received  twice.     31   2.0 Mean Rating 1.5 1.0 0.5 0.0 Follow Caregiver Follow Stranger Greet Caregiver Greet Stranger 537     538   Figure  3.  Mean  rating  for  following  the  caregiver  and  stranger  as  well  as  greeting   539   the  caregiver  and  stranger.  See  Table  3  for  the  scoring  system.  Bars  show  the  mean   540   score  and  error  bars  show  the  95%  confidence  interval.