- PDF Document

Presentation Transcript

  1.   Detecting  Alliance  Ruptures  and  Rupture  Repair  with  the     Segmented  Working  Alliance  Inventory—Observer  Form  (S-­‐WAI-­‐O)     Elizabeth  A.  Berk,  M.A.                 ABSTRACT:  A  series  of  studies  have  shown  the  quality  of  the  therapeutic  alliance  to   be  a  strong  predictor  of  outcome  across  a  wide  range  of  modalities  (Castonguay,   Constantino  &  Holtforth  2006;  Martin,  Garske,  &  Davis,  2000;  Horvath  &  Symonds,   1991)  and  demonstrated  that  when  therapists  can  work  through  negative  process   and  repair  ruptures  in  the  alliance,  their  patients  may  have  better  treatment   outcomes  (Bordin,  1994;  Henry  &  Strupp,  1994;  Horvath,  1995;  Safran,  Muran  &   Eubanks-­‐Carter,  2011).    This  paper  presents  a  study  that  used  the  Segmented   Working  Alliance  Inventory-­‐Observer  Form  (S-­‐WAI-­‐O),  an  observer-­‐based  method   of  detecting  ruptures  and  their  repair,  and  provides  psychometric  validity  and   reliability  for  the  measure.    Observers  applied  the  S-­‐WAI-­‐O  coding  method  to   therapy  sessions  from  22  therapeutic  dyads,  whose  therapists  underwent   specialized  rupture  resolution  training.    The  S-­‐WAI-­‐O  was  used  to  detect  changes  in   the  quality  of  the  therapeutic  alliance  before  and  after  this  specialized  training.     While  no  differences  were  found  between  the  pre-­‐  and  post-­‐training  phases  of   therapy,  the  S-­‐WAI-­‐O  was  found  to  be  a  valid  and  reliable  measure.    This  paper  is   broken  down  into  two  parts:  (1)  a  review  of  the  principal  theoretical  concepts   surrounding  the  study,  and  (2)  the  empirical  study  itself.    

  2.               Detecting  Alliance  Ruptures  and  Rupture  Repair  with  the     Segmented  Working  Alliance  Inventory—Observer  Form   (S-­‐WAI-­‐O)                     by         Elizabeth  A.  Berk,  M.A.             April  2013                   Submitted  to  The  New  School  for  Social  Research  of  the  New  School  in  partial   fulfillment  of  the  requirements  for  the  degree  of  Doctor  of  Philosophy.       Dissertation  Committee:   Dr.  Jeremy  Safran   Dr.  J.  Christopher  Muran   Dr.  Christopher  Christian  

  3.   DEDICATION   I  would  like  to  dedicate  this  dissertation  to  my  family,  who  have  always  supported   me  in  whatever  interests  I  have  wanted  to  pursue.    I  am  especially  indebted  to  my   fiancé  Jake,  who  at  the  very  least  has  earned  an  associates  degree  in  psychology  for   all  that  he  has  learned  through  helping  me  to  study,  editing  my  papers  and   supporting  me  through  the  entire  process  of  graduate  school.      

  4.     iii   ACKNOWLEDGEMENTS   I  would  like  to  thank  Drs.  Jeremy  Safran  and  Chris  Muran.    Their  research   mentorship  has  been  invaluable  and  has  permanently  shaped  the  way  that  I  think   about  psychotherapy  and  psychotherapy  research.         I  would  also  like  to  acknowledge  the  contributions  of  my  coders:  Amy  Withers,   Jessica  Kraus,  Matt  Blanchard,  Stacy  Rozsa,  Rebecca  Kronwith,  Lesley  Solomon,   Mary  Minges,  Orly  Caduri,  Paige  Safyer  and  Eleonora  Calvaca.    This  dissertation   would  not  have  been  completed  without  their  hard  work  and  long  hours  coding.         Finally,  I  have  to  thank  the  Brief  Psychotherapy  Research  Program  and  all  of  its   members.    The  Brief  has  always  been  a  source  of  support,  a  place  to  explore   research  ideas  and,  most  importantly,  a  social  club.                              

  5.     iv   TABLE  OF  CONTENTS     DEDICATION  ............................................................................................................................  iii   ACKNOWLEDGEMENTS  ........................................................................................................  iii   LIST  OF  TABLES  AND  FIGURES  ............................................................................................  v   Chapter  1:  Theoretical  Basis  ................................................................................................  1   I.  The  Therapeutic  Alliance  ...........................................................................................................  1   II.  Ruptures  in  the  Alliance  ............................................................................................................  5   III.  The  Resolution  of  Alliance  Ruptures  ...................................................................................  8   IV.  Detecting  Ruptures  and  their  Repair   ...................................................................................  9   V.  Direct  Self-­‐Report  ........................................................................................................................  9   VI.  Indirect  Self-­‐Report  .................................................................................................................  11   VII.  Observer  Report  ......................................................................................................................  15   VIII.  Concluding  Remarks  ............................................................................................................  19   Chapter  2:  The  Empirical  Study  .......................................................................................  20   I.  Literature  Review  .......................................................................................................................  20   a.  The  Working  Alliance  .............................................................................................................................  20   b.  Ruptures  in  the  Therapeutic  Alliance  .............................................................................................  21   c.  Detection  of  Ruptures  in  the  Therapeutic  Alliance  ....................................................................  23   II.  Hypotheses  ..................................................................................................................................  26   III.  Method  .........................................................................................................................................  26   a.  Data  Collection   ...........................................................................................................................................  26   b.  Participants   .................................................................................................................................................  27   c.  Therapists  ....................................................................................................................................................  27   d.  Patients  .........................................................................................................................................................  27   e.  Diagnostic  Assessment  ..........................................................................................................................  28   f.  Treatment  Model  and  Therapist  Training  Procedure  ...............................................................  29   g.  Cognitive  Behavioral  Therapy  ............................................................................................................  29   h.  Rupture  Resolution  Training  ..............................................................................................................  30   i.  Process  Measures  ......................................................................................................................................  32   j.  Case  and  Session  Selection  ....................................................................................................................  35   k.  S-­‐WAI-­‐O  Coding  and  Scoring  ...............................................................................................................  35   IV.  Results  ..........................................................................................................................................  38   a.   Psychometric  Validity  of  S-­‐WAI-­‐O  ................................................................................................  38   b.  The  Effects  of  Rupture  Resolution  Training  .................................................................................  43   V.  Control  Chart  Examples  ...........................................................................................................  44   VI.  Discussion  ...................................................................................................................................  50   References   ...............................................................................................................................  54   APPENDIX  A  ............................................................................................................................  63            

  6.     v   LIST OF TABLES AND FIGURES   Figure  1.  Control  Chart…………………………………………………………………………….…..37   Figure  2.  S-­‐WAI-­‐O  Factor  Analysis………………………………………………………………..39   Figure  3.  Descriptive  Statistics……………………………………………………………………..43   Figure  4.  Changes  in  Ratings  by  Modality  and  Time  of  Switch………………………..44   Figure  5.  Resolved  CBT  Rupture,  Session  15…………………………………………………44   Figure  6.  Unresolved  CBT  Rupture,  Session  15………………………….…………………..46   Figure  7.  Resolved  RR  Rupture,  Session  23…………………………………………………...47   Figure  8.  Unresolved  RR  Rupture,  Session  15………………………………………………..49   Figure  9.  Post-­‐Session  Questionnaire,  Patient  Version…………………………………...63   Figure  10.  Post-­‐Session  Questionnaire,  Therapist  Version……………………………..65   Figure  11.  Segmented  Working  Alliance  Inventory—Observer  Form……………..68                                      

  7.   1   Chapter 1: Theoretical Basis I.  The  Therapeutic  Alliance     "At  last  the  Dodo  said,  'Everybody  has  won,  and  all  must  have  prizes.'"                                                -­‐Lewis  Carroll,  Alice  in  Wonderland     Rosenzweig's  (1936)  Dodo  bird  verdict  was  a  proposition  that  all     psychotherapies,  regardless  of  theoretical  orientation  or  technique,  are  basically   equally  effective—all  psychotherapies  have  "won,"  so  comparisons  of  different   treatment  modalities  are  meaningless.    Luborsky,  Singer  and  Luborsky  (1974)   provided  empirical  support  for  the  Dodo  bird  verdict  with  one  of  the  first  large-­‐scale   comparative  studies,  which  revealed  few  significant  differences  in  outcome  between   treatment  types.    Subsequent  research  has  shown  therapy  to  be  an  effective  form  of   treatment,  but  there  is  little  evidence  to  support  the  superiority  of  one  type  of   therapy  or  technique  over  another  (Lambert  &  Bergin,  1994;  Stiles,  Shapiro  &  Elliot,   1986).    While  some  researchers  are  still  concerned  with  comparing  types  of   treatment,  there  has  been  a  shift  in  psychotherapy  research  toward  studying  the   process  of  therapy  and  thereby  discovering  common  mechanisms  of  change  across   psychotherapies  (Rice  &  Greenberg,  1984).         One  of  the  most  widely  investigated  common  factors  is  the  therapeutic   working  alliance,  a  psychoanalytic  concept  popularized  in  psychotherapy  research   by  Bordin  (1979).    A  large  body  of  research  has  consistently  shown  a  positive   relationship  between  therapeutic  alliance  and  treatment  outcome  (Hartley  &   Strupp,  1983;  Horvath  &  Symonds,  1991;  Martin,  Garske  &  David,  2000).    For   example,  Hartley  and  Strupp's  (1983)  research  at  Vanderbilt  University  found  that  

  8.   2   an  early  establishment  of  a  therapeutic  alliance  (i.e.,  by  the  third  session)  predicted   success  in  treatment.    Horvath  and  Symonds  (1991)  completed  the  first  meta-­‐ analysis  comparing  the  therapeutic  alliance  and  outcome.    Their  analysis  included   24  studies  and  found  a  moderate  effect  size  (0.26)  linking  the  quality  of  the  alliance   and  outcome.    In  2000,  Martin  et  al.  completed  a  similar  meta-­‐analysis  with  an   additional  55  studies;  as  in  previous  studies,  their  results  confirmed  the  presence  of   a  moderate  and  consistent  relationship  between  alliance  and  outcome.    This  body  of   research  indicates  that  the  alliance  is  a  critical  mechanism  of  change  in   psychotherapy  (Safran  &  Muran,  2000).    Researchers  are  now  turning  their   attention  to  the  way  an  alliance  develops  and  is  maintained.     The  first  origins  of  the  therapeutic  alliance  come  from  Freud  (1912,  1913),   who  described  the  patient's  attachment  to  the  analyst  as  "effective  transference"   and  believed  the  analyst's  first  goal  in  treatment  was  to  "attach"  the  patient  to   himself  by  showing  consistent  sympathy  and  interest,  clearing  away  early  resistance   and  avoiding  mistakes.    While  Freud  considered  the  patient's  positive  reactions  and   transference  towards  the  analyst  necessary  components  of  a  successful  analysis,  he   also  regarded  this  attachment  as  almost  inevitable,  requiring  little  work  from  the   therapist.     Although  often  attributed  to  Zetzel,  Sterba  (1934)  first  introduced  the  term   "therapeutic  alliance."    Like  Freud,  Sterba  focused  on  the  patient's  contributions  to   the  therapeutic  alliance  and  saw  the  analyst's  role  as  mostly  passive.    Unlike  Freud,   however,  Sterba  believed  the  patient's  alliance  with  the  analyst  emerged  from  the   rational  part  of  the  patient's  ego,  which  could  overcome  resistances  and  instinctual  

  9.   3   drives  to  form  a  pact  with  the  analyst  and  engage  in  treatment;  the  patient's   motivation  for  treatment  was  not  a  libidinal  drive,  but  a  rational  desire  for  health.     For  Sterba,  this  rational  desire  allowed  the  patient  to  maintain  a  "split  in  the  ego"   that  preserved  the  alliance  with  the  therapist  in  the  face  of  negative  transference.       Zetzel  (1956),  an  ego  psychologist,  later  built  upon  both  Sterba  and  Freud's   ideas  about  the  therapeutic  alliance,  defining  it  as  the  patient's  identification  or   positive  transference  with  the  therapist.    For  Zetzel,  as  for  Freud  and  Sterba  before   her,  the  therapeutic  alliance  was  a  necessary  component  of  therapy  that  enabled  the   patient  to  work  collaboratively  with  the  analyst  in  the  face  of  conflictual  feelings   about  the  analyst  and  the  process.    However,  in  contrast  to  both  Freud  and  Sterba,   Zetzel  promoted  a  more  active  role  for  the  therapist,  and  believed  that  the  analyst   must  sometimes  actively  intervene  to  bolster  reality  testing  and  ego  resources  and   to  contain  regressive  transference  manifestations  (Zetzel,  1970).    In  Zetzel's   conception,  the  alliance  requires  a:     Capacity  to  tolerate  anxiety  and  frustration,  to  accept  certain  reality   limitations,  and  to  differentiate  between  mature  and  infantile  aspects   of  mental  life.    This  relationship  acts,  on  the  one  hand,  as  a  barrier  to   significant  ego  regression,  and,  on  the  other,  as  a  fundamental  feature   of   the   analytic   situation   against   with   the   fantasies,   memories   and   emotions  evoked  by  the  transference  neurosis  can  be  measured  and   contrasted  (Zetzel,  1956,    p.  185).         Zetzel  also  differentiated  between  positive  transference,  which  contributes  to  a   therapeutic  alliance  and  is  necessary  for  effective  analytic  treatment,  and  a   transference  neurosis,  which  is  part  of  the  patient's  resistance  and  requires  further   analysis.      

  10.   4     In  1965,  Greenson  gained  more  attention  for  the  working  alliance,  defining   it  as  "the  relatively  non-­‐neurotic,  rational  rapport  which  the  patient  has  with  his   analyst"  (p.  157).    While  he  did  not  discount  the  effect  of  transference  entirely,  and   acknowledged  that  it  may  have  some  influence  on  the  development  of  the   therapeutic  relationship,  Greenson  agreed  with  Sterba  that  the  core  of  the  working   alliance  is  the  patient's  rational  desire  to  cooperate  with  his  analyst,  to  utilize  the   analyst's  interventions  and  interpretations,  and  ultimately  to  become  healthy   (Greenson,  1967).    Like  Zetzel,  Greenson  believed  that  the  therapist  actively   contributes  to  the  formation  and  maintenance  of  the  working  alliance;  however,  he   emphasized  that  the  role  of  the  therapist  is  not  to  promote  reality  testing  and  ego   functions,  but  to  act  with  an  appropriate  level  of  "humanness,"  which  he  defined  as   "understanding  and  insight  conveyed  in  an  atmosphere  of  serious  work,   straightforwardness,  compassion,  and  restraint"  (Greenson,  1965,  p.  179).    This   attitude  of  humanness  is  a  balance  between  gratification  and  neutrality  and  helps   build  the  "real"  relationship  that  Greenson  considered  the  key  mediating  factor  in   the  therapeutic  process.      Bordin  (1979),  drawing  from  the  work  of  both  Zetzel  and  Greenberg,  argued   that  the  working  alliance  is  central  to  therapeutic  change  in  all  types  of  therapies.     Concerned  about  the  growing  number  of  theoretical  orientations  and   psychotherapies,  Bordin  identified  the  working  alliance  as  a  change  agent  that  is   common  across  treatment  models;  while  each  kind  of  therapy  makes  different   demands  of  the  patient  and  therapist,  they  all  have  "embedded"  working  alliances.     Bordin  proposed  that  the  success  of  therapy  is  due,  possibly  entirely,  to  the  strength  

  11.   5   of  the  working  alliance,  and  that  this  strength  is  determined  by  the  goodness  of  fit   between  the  demands  of  the  working  alliance  required  by  a  particular  kind  of   therapy  and  the  personal  characteristics  of  the  patient  and  therapist.    He  also  broke   down  the  working  alliance  into  three  distinct  but  interrelated  components:  "An   agreement  on  goals,  an  assignment  of  tasks  or  a  series  of  tasks,  and  the  development   of  a  bond"  (Bordin,  1979,  p.  253).    The  goals  of  therapy  are  what  the  patient  and   therapist  want  to  achieve  through  the  therapy  (e.g.,  better  self-­‐understanding  or  a   reduction  in  panic  attacks),  and  the  tasks  of  therapy  are  the  means  by  which  these   goals  are  achieved  (e.g.,  transference  interpretations  or  thought  records).    A  bond  is   an  affective  relationship  established  between  the  patient  and  therapist  that  makes   therapeutic  work  possible.    Unlike  some  of  his  predecessors,  Bordin  argues  that   both  the  patient  and  therapist  contribute  to  the  establishment  and  maintenance  of  a   working  alliance.         II.  Ruptures  in  the  Alliance     As  mentioned  above,  there  is  a  great  body  of  research  dedicated  to  the   therapeutic  alliance,  which  has  shown  that  the  quality  of  the  alliance  is  one  of  the   best  predictors  of  psychotherapy  outcome  (see,  e.g.,  Hartley  &  Strupp,  1983;   Horvath  &  Symonds,  1991;  Martin,  Garske  &  David,  2000).    Over  the  last  twenty   years,  a  second  wave  of  therapeutic  alliance  research  has  emerged,  with  an   emphasis  on  exploring  the  factors  that  establish  and  maintain  an  alliance  (Safran,   Muran  &  Eubanks-­‐Carter,  2010).    This  research  has  been  particularly  devoted  to   examining  the  processes  involved  in  identifying  and  repairing  strains  or  ruptures  in   the  alliance  (Bordin,  1980;  Safran,  Muran,  Samstag  &  Stevens,  2002).    

  12.   6   Safran  and  Muran  (1996)  defined  a  rupture  in  the  therapeutic  alliance  as  a     deterioration  in  the  relationship  between  the  patient  and  the  therapist  or  a  difficulty   in  establishing  such  a  relationship;  ruptures  are  unavoidable  events  in  therapy  and   can  range  from  explosive  blowouts  to  minor,  almost  imperceptible  shifts.    The  most   severe  ruptures  may  lead  to  premature  termination,  while  mild  ruptures  may  never   be  addressed  during  the  course  of  treatment.    Safran  and  Muran  (2000),  drawing   from  relational  psychoanalytical  theory,  stated  that  both  the  patient  and  therapist   contribute  to  the  development  of  a  rupture.    In  other  words,  a  rupture  is  not  caused   exclusively  by  a  patient's  neuroses  or  a  therapist's  actions;  it  is  an  interactive,   dynamic  process  with  contributions  from  both  the  patient  and  the  therapist.     Ruptures  are  often  important  opportunities  to  work  through  problems,  strengthen   the  alliance  and  correct  maladaptive  interpersonal  processes  (Safran,  Crocker,   McMain  &  Murray,  1990;  Safran  &  Muran,  1996).       One  can  conceptualize  ruptures  as  a  disturbance  in  one  or  more  of  Bordin's   (1979)  three  components  of  the  alliance:  goals,  tasks  and  the  therapeutic  bond.    The   tasks  and  goals  of  therapy  are  in  a  constant  state  of  negotiation,  and  this  negotiation   allows  for  interpersonal  conflicts  and  ruptures.    For  instance,  the  task  and  bond  may   be  interrupted  when  a  therapist  brings  up  a  topic  that  the  patient  is  unwilling  to   discuss  (i.e.,  a  task)  and  the  patient  responds  with  hostility  (thus  disturbing  the   bond).         Harper  (1989a,  1989b)  conceptualized  ruptures  in  the  therapeutic  alliance  as   being  divided  into  two  types  of  markers:  withdrawal  and  confrontation.    While  she   did  not  discount  therapists'  contributions  to  ruptures,  her  manual  for  identifying  

  13.   7   ruptures  focused  on  patient  behaviors.    In  Harper's  model,  a  withdrawal  rupture   occurs  when  the  patient  disengages  from  the  therapist,  the  therapy  or  his  own   internal  experience;  confrontation  ruptures  occur  when  the  patient  moves  against   the  therapist  or  some  aspect  of  the  therapeutic  process  in  an  aggressive,  hostile   manner  by  expressing  resentment  or  dissatisfaction.    While  confrontation  markers   are  usually  less  subtle  and  can  be  easily  imagined,  withdrawal  markers  are   sometimes  harder  to  detect  and  define.    A  wonderful  example  of  a  withdrawal   rupture  comes  from  an  interview  with  a  therapist  at  the  Brief  Psychotherapy   Research  Program  in  New  York  City.   Interviewer:   Have   you   experienced   any   moments   of   conflicts,   disagreements,   misunderstandings,   or   tension   in   your   relationship   with  your  patient?     Therapist:  Yes  there  is  a  level  at  which  it  is  still  below  the  surface  in   some  ways,  and  it  tends  to  come  up  around  this  issue  of  control  and   my   asking   her   about   her   felt   experience   and   then   her   being   uncomfortable  with  that  and  moving  away  from  it  and  then  my  trying   to  pull  her  back.    And  initially  it  felt  like  we  were  trying  to  smooth   things  over  and  not  looking  at  that.  We’ve  never  had  a  conflict  that   was   a   really   clear   confrontation   where   she   was   openly   angry   or   anything  of  that  nature  but  its  more  this  sense  of  her  having  reactions   to  things  I’ve  said  and  I’ve  asked  about  it  directly  and  we’ve  explored   it  but  it  feels  like  there  is  more.       Interviewer:  But  there  hasn’t  been  a  salient  moment  kind  of  moment   where  she  has  rejected  something  you’ve  said.       Therapist:  No  there  have  been,  we’ve  had  conversations  where  she   has  acknowledged  feeling  pushed  by  me  and  controlled  in  some  ways   but  I  can’t  say  that  there  was  a  moment  that  there  was  a  clear  conflict.     More  that  there  were  these  little  ripples  that  happen  fairly  regularly.       This  example  shows  how  withdrawal  ruptures  can  subtly  affect  the  therapeutic   process  and  alliance  throughout  the  course  of  treatment.      

  14.   8   III.  The  Resolution  of  Alliance  Ruptures     A  number  of  studies  have  shown  that  alliance  ruptures  present  an   opportunity  for  psychological  transformation  because  they  allow  a  therapist  to   repair  a  patient's  maladaptive  interpersonal  schemas  (Safran  &  Muran,  2000;   Safran,  Muran,  Samstag  &  Stevens,  2002;  Safran  &  Segal,  1990).    Resolving  ruptures   may  also  help  prevent  early  termination  from  treatment  (Martin  et  al.,  2000;   Samstag,  Batchelder,  Muran,  Safran,  &  Winston,  1998).       To  further  understand  the  process  of  rebuilding  the  alliance  after  a  rupture,   Safran  and  Muran  (1996)  created  a  rupture  resolution  model  that  explains  the   process  of  resolution.    Drawing  from  Harper's  (1989a,  1989b)  paradigm,  they   created  two  models  of  resolution:  one  for  resolving  confrontation  ruptures  and  one   for  resolving  withdrawal  ruptures  (Safran  et  al.,  1990;  Safran  &  Muran,  1996;  2000).     When  a  withdrawal  rupture  occurs,  the  resolution  process  involves  addressing   interpersonal  fears,  expectations  and  internalized  criticism  that  impede  the   patient's  ability  to  directly  express  negative  feelings;  the  therapist  slowly   encourages  the  patient  to  self-­‐assert  and  express  underlying  hopes  and  wishes.     When  a  confrontation  rupture  occurs,  the  resolution  process  involves  exploring  the   fears  of  self-­‐criticism  that  create  aggressive  and  assertive  feelings  towards  the   therapist  or  therapy  and  impede  the  expression  of  underlying  needs.    The   therapist's  task  is  to  encourage  the  patient  to  express  vulnerable  feelings.    Although   Safran  and  Muran  defined  two  different  resolution  models,  they  acknowledge  that   elements  of  both  types  of  ruptures  are  usually  present  in  a  rupture  episode,  so  the   therapist  may  utilize  both  models  of  resolution  in  one  rupture  episode.        

  15.   9   IV.  Detecting  Ruptures  and  their  Repair     Given  the  importance  that  maintaining  a  strong  alliance  and  repairing   ruptures  has  for  treatment  outcome  and  retention,  researchers  should  strive  to   create  methodologies  and  instruments  for  detecting  ruptures  and  rupture  repair   episodes.    Once  rupture  and  repair  episodes  are  successfully  identified,  researchers   can  then  begin  to  study  how  ruptures  are  successfully  and  unsuccessfully  resolved.     A  number  of  methods  have  been  employed  to  study  these  events.    In  the  naturalistic   method  of  observing  ruptures,  researchers  observe  the  natural  occurrence  of   ruptures  and  rupture  resolution  in  psychotherapy  and  examine  their  relationship   with  outcome  (Eubanks-­‐Carter,  Muran  &  Safran,  2010).    There  are  three  methods   for  identifying  ruptures  and  rupture  resolution  in  a  naturalistic  way:  direct  patient   or  therapist  self-­‐report  of  ruptures  and  resolution,  indirect  self-­‐report  based  on   measures  of  the  alliance,  and  observer-­‐based  measures  of  ruptures  and  resolution.       V.  Direct  Self-­‐Report   Eames  and  Roth  (2000)  compared  patient  and  therapist  report  of  rupture   and  patient  and  therapist  alliance  scores  to  patients'  attachment  styles,  and   hypothesized  that  an  insecure  attachment  style  may  hinder  the  early  formation  of   the  alliance  and  contribute  to  more  ruptures  in  later  therapy.    After  each  session,   they  administered  the  Post  Session  Questionnaire  (PSQ;  Muran,  Samstag  &  Winston,   1992)  to  30  patients  and  the  11  therapists  in  the  study.    The  PSQ  has  self-­‐report   measures  of  the  alliance  as  well  as  questions  relating  to  the  occurrence  of  ruptures,   rupture  intensity  and  the  extent  to  which  ruptures  were  resolved.    Overall,   therapists  reported  twice  as  many  ruptures  as  their  patients.    The  researchers  

  16.  10   conceded  that  this  discrepancy  between  patient  and  therapist  report  of  rupture   could  not  be  accounted  for  in  their  study,  but  theorized  that  it  may  have  been  due  to   differences  in  awareness  of  tension,  differences  in  attribution,  or  reporting  bias  (i.e.,   that  the  therapists  knew  the  researchers  were  studying  ruptures  and  therefore   focused  more  on  their  presence  than  they  would  have  otherwise.    They  also  found   that  a  patient's  Preoccupied  attachment  style  was  associated  with  more  frequent   report  of  ruptures  by  the  therapist,  whereas  a  Dismissing  attachment  style  was   associated  with  fewer  therapist-­‐reported  ruptures.    They  did  not  find  a  significant   relationship  between  attachment  style  and  patient-­‐reported  ruptures.     Muran  et  al.  (2009)  studied  the  relationship  of  early  alliance  ruptures  and   their  resolution  to  process  and  outcome  measures  in  a  sample  of  128  patients  in   three  time-­‐limited  therapies.    They  also  used  the  PSQ  to  measure  alliance,  rupture   presence,  rupture  intensity  and  resolution,  and  found  frequent  report  of  ruptures  in   the  first  six  sessions  of  therapy  from  both  patients  (37%  of  sessions)  and  therapists   (56%  of  sessions).      High  rupture  intensity,  as  reported  by  both  the  patient  and   therapist,  was  associated  with  poor  outcome  on  measures  of  interpersonal   functioning.    However,  successful  resolution  of  these  ruptures  was  associated  with   higher  treatment  retention  rates.         There  are  several  problems  with  direct  self-­‐report  of  ruptures  and  rupture   repair  episodes.    Regan  and  Hill  (1992)  exposed  one  such  problem  when  they  asked   patients  and  experienced  therapists  about  the  thoughts  and  feelings  that  they  were   unwilling  or  unable  to  express  in  treatment.    They  found  that  for  both  patients  and   therapists,  most  of  these  unsaid  things  were  negative.    They  also  found  that  

  17.  11   therapists  were  only  aware  of  17%  of  the  things  patients  had  left  unsaid.    In  a   subsequent  study,  Rhodes,  Hill,  Thompson  and  Elliot  (1994)  found  that  patients   who  felt  uncomfortable  discussing  misunderstandings  with  their  therapists  were   able  to  conceal  these  feelings  and  the  misunderstandings  were  left  unaddressed,   which  often  led  to  early  treatment  termination.    Patients  and  therapists  are   therefore  not  equally  aware  of  or  willing  to  admit  to  problems  in  the  relationship.     This  is  reflected  in  the  discrepant  reports  of  ruptures  between  patient  and  therapist   that  are  found  in  many  studies  (e.g.,  Muran  et  al.,  2009,  Eames  &  Roth,  2000).     Patients  may  also  be  unwilling  to  report  problems  with  the  relationship  because   they  have  the  perception  that  it  will  negatively  impact  their  therapist  in  some  way   or  that  their  therapists  may  view  these  ratings  even  when  patients  are  assured  that   this  is  not  the  case.       VI.  Indirect  Self-­‐Report   Because  patients  and  therapists  have  such  a  discrepancy  in  reporting   ruptures,  some  researchers  question  the  utility  of  direct  self-­‐report  as  a  method  for   identifying  ruptures  and  have  developed  indirect  methods  of  identifying  ruptures  as   an  alternative.    These  indirect  methods  involve  collecting  patient  and  therapists'   measures  of  the  overall  alliance  and  then  using  fluctuations  in  the  alliance  scores  to   detect  ruptures  and  rupture  repair  episodes.       An  early  example  of  this  method  comes  from  Safran  et  al.  (1990,  1994),  who   asked  patients  to  answer  six  questions  from  the  Working  Alliance  Inventory  (WAI;   Horvath  &  Greenberg,  1986)  after  each  session.    Patients  were  instructed  to  answer   these  questions  for  the  beginning,  middle  and  end  of  the  therapy  session.    The  

  18.  12   researchers  chose  sessions  that  showed  a  lower  alliance  score  in  the  middle  of   the  session  and  higher  scores  at  the  beginning  and  end,  which  they  identified  as   rupture  and  repair  sequences.    They  then  used  these  segments  to  complete  a  task   analysis  of  the  rupture  and  repair  process.    This  relatively  simple  investigation   within  single  therapy  sessions  has  sparked  more  complicated  and  sophisticated   attempts  to  indirectly  measure  rupture  and  repair  episodes  across  the  course  of  a   treatment.       For  example,  Kivlighan  and  Shaughnessy  (2000)  used  patient  alliance  ratings   collected  after  each  therapy  session  in  a  time-­‐limited  treatment  to  determine   patterns  in  the  alliance  and  associate  these  patterns  with  outcome.    They  identified   three  distinctive  patterns:  a  stable  alliance  pattern,  a  linear  growth  pattern  and  a   quadratic  growth  or  "U-­‐Shaped"  pattern.    They  found  that  patients  with  the   quadratic  growth  pattern  (i.e.,  high  alliance  in  early  sessions,  then  a  lower  alliance  in   the  middle  of  treatment,  then  a  return  to  high  alliance  in  final  sessions)  had   significantly  better  outcome  than  the  other  two  growth  patterns.    This  suggests  that   an  early  high  alliance,  with  subsequent  problems  and  repairs  in  the  alliance,  is   associated  with  greater  change  in  the  patient's  outcome  than  a  stable  or  linear   growth  pattern.       Stiles  et  al.  (2004)  tried  to  replicate  this  work  by  using  patient-­‐rated  alliance   scores  on  the  Agnew  Relationship  Measure  (ARM;  Agnew-­‐Davies,  Stiles,  Hardy,   Barkham  &  Shapiro,  1998).    They  studied  79  patients  who  were  being  treated  in   either  a  cognitive  or  psychodynamic  therapy  for  depression.    The  researchers  found   four  distinctive  patterns  of  alliance  development,  only  two  of  which  matched  

  19.  13   Kivlighan  and  Shaughnessy's.    They  did  not  find  the  U-­‐shaped  quadratic  growth   pattern  to  be  associated  with  better  outcome;  instead,  they  identified  a  "V-­‐shaped"   pattern,  with  smaller,  more  discrete  rupture  and  repair  episodes,  that  was   associated  with  higher  outcome.    While  both  Stiles  et  al.  and  Kivlighan  and   Shaughnessy  found  an  association  between  rupture  repair  episodes  and  higher   outcome,  the  U-­‐Shaped  pattern  was  not  associated  with  good  outcome  in  both   studies.    Therefore  the  results  of  these  studies  cannot  be  generalized.         Strauss  et  al.  (2006)  developed  yet  another  method  for  detecting  rupture  and   repair  episodes  by  examining  changes  in  alliance  scores  over  the  course  of  therapy   for  30  patients  with  Avoidant  and  Obsessive-­‐Compulsive  Personality  Disorders.     Researchers  measured  alliance  scores  by  administering  the  California   Psychotherapy  Alliance  Scale  (CALPAS;  Marmar,  Weiss  &  Gaston,  1989)  to  patients   after  each  session.    They  then  devolved  criteria  to  define  sessions  where  rupture   and  resolution  occurred.    For  example,  a  session's  alliance  score  had  to  differ  by  at   least  one  standard  deviation  (as  calculated  across  the  entire  data  set)  from  the   patient's  other  alliance  scores  in  order  to  be  considered  a  rupture  and  resolution   session.    The  researchers  found  that  patients  whose  therapy  included  rupture  repair   episodes  experienced  greater  symptom  reduction  than  those  who  did  not.   Stevens,  Muran,  Safran,  Gorman  and  Winston  (2007)  developed  a  similar   system  for  finding  fluctuations  in  patient-­‐rated  WAI  scores,  using  a  sample  of  44   patients  with  cluster  C  personality  disorders.    They  defined  ruptures  as  decreases   by  at  least  one  point  on  the  WAI,  and  deemed  these  ruptures  resolved  if  the  alliance   score  rose  to  within  0.25  points  of  the  pre-­‐rupture  score  in  three  to  five  sessions.    

  20.  14   While  rupture-­‐repair  episodes  were  found  to  be  common,  appearing  in  50%  of   the  cases,  these  episodes  were  not  related  to  treatment  outcome.       Eubanks-­‐Carter,  Gorman  and  Muran  (2012)  borrowed  methods  from  a   number  of  fields,  including  epidemiology,  manufacturing,  climatology  and   economics,  to  detect  changes  in  the  alliance  as  measured  by  patient-­‐rated  WAI   scores.    They  identified  four  methods  for  detecting  ruptures:  criterion-­‐based   methods,  control  chart  methods,  partitioning  methods  and  regression  methods.     They  illustrated  these  methods  and  discussed  the  strengths  and  weaknesses  of  each   one  using  a  case  example.    While  none  of  the  methods  produced  identical  results   with  respect  to  demarcating  ruptures,  there  were  a  number  of  points  where  all   methods  identified  a  weakness  in  the  alliance.    The  researchers  identified  Shewhart   control  charts  as  the  most  useful,  pragmatic  way  to  identify  ruptures  because  this   method  does  not  require  a  large  number  of  data  points,  can  be  used  in  common   statistical  software  and  can  track  the  alliance  in  real-­‐time,  which  may  be  useful  for   informing  therapists  of  problems  in  the  relationship.    Future  research  with  larger   samples  sizes  would  help  to  clarify  the  utility  of  these  different  methods.       While  indirect  methods  of  detecting  rupture  and  rupture  repair  episodes  are   able  to  combat  the  problems  associated  with  direct  patient  and  therapist  self-­‐report   of  ruptures,  there  are  a  number  of  limitations  to  these  methods.    One  such  limitation   is  that  most  of  these  methods  provide  global  ratings  over  the  course  of  a  therapy,   and  therefore  do  not  add  to  the  understanding  of  the  in-­‐session  process  of  rupture   and  repair.    Another  issue  is  that  each  study  produces  a  different  method  for  

  21.  15   detecting  ruptures  and  repair  episodes,  and  the  results  from  each  method  have   not  been  adequately  replicated.       VII.  Observer  Report     Because  of  the  above-­‐mentioned  problems  using  direct  and  indirect  self-­‐ report  methods  of  detecting  rupture  and  rupture  repair  episodes,  researchers  have   begun  to  develop  observer-­‐based  methods  for  detecting  these  events  with  the  hopes   of  mitigating  problems  with  the  other  methods.       Lansford  (1986)  did  an  exploratory  study  of  "weakenings"  in  the  therapeutic   alliance  by  reviewing  the  audiotapes  of  six  cases  of  time-­‐limited  dynamic  therapy   and  assigning  researchers  to  code  the  sessions  for  different  criteria  related  to   weakenings  in  the  patient-­‐therapist  relationship.    Lansford  described  weakenings  as   "A  negative  response  to  therapy,  fear  of  the  therapist's  critical  judgment  or   disapproval,  problems  with  coming  to  or  talking  in  therapy,  or  problems  with   termination"  (p.  364).    He  reported  a  75%  to  100%  reliability  between  coders,  but   did  not  note  how  this  was  calculated.    The  results  of  the  study  were  that  successfully   resolved  weakenings  were  associated  with  better  overall  outcome  (although,  again,   Lansford  did  not  explain  how  outcome  was  determined).    This  early  study  was  an   important  investigation  into  the  relationship  between  ruptures  and  outcome,  but   many  of  the  methods  used  in  the  study  were  not  adequately  described,  so  the  results   must  be  interpreted  with  caution.       Sommerfeld,  Orlach,  Zim  and  Mikulincer  (2008)  compared  observer-­‐rated   ruptures  to  patient  report  of  ruptures  and  patient  Core  Conflictual  Relationship   Themes  (CCRTs).    To  accomplish  this,  the  researchers  first  identified  ruptures  by  

  22.  16   having  two  graduate  students  code  transcripts  using  the  withdrawal  and   confrontation  rupture  markers  identified  by  Harper  (1989a,  1989b).    The  coders'   reliability  was  fair  to  good;  they  had  an  average  reliability  of  0.45  for  confrontation   markers  and  0.59  for  withdrawal  markers,  though  their  study  does  not  note  how   reliability  was  calculated.    After  the  transcripts  were  coded,  each  session  was   assigned  a  separate  "Yes"  or  "No"  value  for  the  presence  of  withdrawal  and   confrontation  markers.    Observer-­‐rated  ruptures  were  found  in  77%  of  all  sessions.     After  each  session,  patients  filled  out  the  Post-­‐Session  Questionnaire,  which  includes   two  Likert-­‐rated  questions  about  the  presence  of  ruptures  (Samstag,  Batchelder,   Muran,  Safran  &  Winston,  1998).    When  researchers  compared  the  observer  ratings   of  ruptures  to  patient  ratings,  they  did  not  find  a  relationship;  however,  they  found   that  the  emergence  of  ruptures  was  associated  with  the  emergence  of  CCRTs   directed  at  the  therapist.    These  findings  suggest  that  ruptures  occur  when  patients'   dysfunctional  interpersonal  schemas  are  activated,  and  are  thus  an  important   opportunity  for  working  on  patients'  maladaptive  interpersonal  patterns.    One   major  limitation  for  this  study  is  the  use  of  transcripts  instead  of  audio  or  video   recordings:  Transcripts  are  costly  and  time  consuming  to  produce,  and  they  omit   subtle  interactions  between  the  patient  and  therapist,  including  body  language  and   tone,  which  are  important  clues  to  the  interpersonal  process.       Colli  and  Lingiardi  (2009)  also  used  transcripts  to  identify  in-­‐session   ruptures  and  resolution  with  a  measure  that  they  created,  the  Collaborative   Interaction  Scale  (CIS).    Drawing  from  the  work  of  Safran  and  Muran,  (Safran  et  al.,   1990;  Safran,  Muran  &  Samstag,  1994,  etc.)  they  divided  the  scale  into  patient  and  

  23.  17   therapist  contribution,  with  three  subscales  for  the  patient  (the  collaborative   processes  scale,  direct  rupture  markers  scale  and  indirect  rupture  markers  scale)   and  two  subscales  for  the  therapist  (the  positive  intervention  scale  and  the  negative   intervention  scale).    In  this  study,  coders  rated  in  pairs  and  achieved  good  reliability,   with  an  average  reliability  of  0.68-­‐0.76  per  scale  as  measured  with  an  Intraclass   Correlation  Coefficient  (ICC).    They  found  a  significant  correlation  between  therapist   negative  interventions  and  patient  alliance  ruptures.    They  also  found  a  significant   correlation  between  therapist  positive  interventions,  patient  collaborative   processes  and  indirect  rupture  markers.    While  they  found  good  reliability  and   interesting  correlations,  the  researchers  did  not  validate  their  measure  by   comparing  it  to  other  measures  of  the  alliance,  ruptures  or  outcome.    Another   shortcoming  of  the  study  is  that  the  measure  is  very  time  consuming:  Each  session   has  to  be  transcribed,  then  broken  down  into  speech  utterances,  then  each  utterance   rated  by  two  independent  observers.    As  noted  above,  the  use  of  transcripts  is   problematic  because  many  interpersonal  cues  are  lost.         Citing  problems  with  the  use  of  transcripts  and  highly  trained  coders,   Eubanks-­‐Carter,  Muran  and  Safran  (2009)  developed  a  coding  system  that  can  be   used  by  beginning  graduate  students  with  videotaped  sessions.    This  system,  The   Rupture  Resolution  Rating  System  (3RS)  draws  on  Harper's  (1989a,  1989b)  manual   for  coding  confrontation  and  withdrawal  ruptures  and  the  Rupture  Resolution  Scale   developed  by  Samstag,  Muran  and  Safran  (2004).    3RS  scores  give  each  session  an   overall  Likert  rating  for  confrontation  ruptures,  withdrawal  ruptures,  resolution   attempts  and  the  success  of  resolution  attempts.    This  measure  is  still  in  

  24.  18   development,  but  a  preliminary  study  found  a  large  discrepancy  between  patient-­‐ reported  ruptures  (reported  in  35%  of  sessions)  and  observer-­‐rated  ruptures   (reported  in  75%  of  sessions);  this  discrepancy  is  similar  to  Sommerfeld,  et  al.'s   (2008)  findings.    One  potential  problem  with  the  measure  is  that  it  gives  an  overall   rating  for  the  session  instead  of  defining  where  in  the  session  ruptures  occurred,   which  could  be  useful  information  for  examining  rupture  and  repair  episodes  in   more  detail.         While  past  research  has  found  interesting  links  between  ruptures,  repair   episodes  and  patient  outcome,  the  studies'  methods  for  detecting  ruptures  were   often  problematic.    Many  of  these  studies  are  limited  by  a  very  small  sample  size.     Most  of  the  existing  observer-­‐based  measures  (Lansford,  1986;  Sommerfeld,  et  al.,   2008;  Colli  &  Lingiardi,  2009)  use  transcripts,  which  are  time-­‐consuming  to  produce   and  lose  many  subtle  interpersonal  cues.    The  patient  and  therapist's  tone  of  voice   (e.g.,  whether  a  statement  was  made  in  a  hostile,  sarcastic  or  warm  tone)  and  body   movements  (e.g.,  avoiding  eye  contact  or  turning  away  from  each  other)  can  be   important  clues  about  the  interpersonal  processes  between  them.    Another  problem   with  some  of  the  measures  (e.g.,  Sommerfeld  et  al.;  Eubanks-­‐Carter,  Safran  &  Muran,   2009)  is  that  the  coding  yields  a  global  score  for  the  session  instead  of  demarcating   where  in  the  session  the  rupture  occurred.    If  other  researchers  wanted  to  study  the   session,  there  might  be  a  lot  of  "noise"  in  the  session  from  non-­‐rupture  material,   which  could  skew  their  results.    Future  researchers  should  concentrate  on  making   an  easy-­‐to-­‐use  observer-­‐based  measure  that  can  be  used  with  videotaped  sessions   to  demarcate  rupture  and  repair  episodes  within  sessions.          

  25.  19   VIII.  Concluding  Remarks     Because  there  has  been  strong  evidence  to  suggest  that  a  good  alliance  is   positively  associated  with  outcome,  alliance  research  has  shifted  from  comparing   the  alliance  and  outcome  to  studying  the  process  by  which  the  alliance  is  created,   repaired  and  maintained.    One  area  of  particular  interest  is  ruptures  and  rupture   repair  episodes.    There  has  been  evidence  to  suggest  that  repairing  alliance  ruptures   is  associated  with  more  gains  in  therapy  and  higher  treatment  retention.    While   many  methods  have  been  employed  to  detect  ruptures  and  rupture  repair  episodes,   none  are  without  their  limitations.    Direct  report  of  ruptures  from  the  patient  and   therapist  are  limited  by  patient  awareness  and  willingness  to  disclose  information   about  ruptures.    Indirect  methods  for  reporting  ruptures  combat  problems  with   direct  report  of  rupture,  but  are  limited  by  the  large  number  of  methods  and   inability  to  replicate  results.    Observer-­‐based  methods  for  detecting  ruptures  and   rupture  repair  episodes  are  limited  by  the  common  use  of  transcripts  and  the   inability  to  determine  where  the  rupture  occurs  within  the  session.    The  following   empirical  article  reviews  the  development  and  validation  of  an  observer-­‐based   method  of  detecting  rupture  and  rupture  repair  episodes,  the  Segment  Working   Alliance  Inventory—Observer  Form  (S-­‐WAI-­‐O),  that  attempts  to  combat  the  issues   laid  out  in  this  literature  review.                  

  26.  20   Chapter 2: The Empirical Study I. Literature Review a. The Working Alliance The  relationship  between  a  patient  and  his  therapist  has  been  conceptualized   in  a  number  of  ways.    One  of  the  most  popular  models  of  this  relationship  is  Bordin’s   (1979)  working  alliance,  which  has  been  widely  accepted  across  treatment   modalities.        Bordin’s  pan-­‐theoretical  model  of  the  working  alliance  comprises  three   parts:  bond,  tasks,  and  goals.    He  defines  bond  as  the  affective  connection  between   patient  and  therapist,  goals  as  the  long-­‐term  objectives  for  treatment,  and  tasks  as   the  specific  activities  that  the  dyad  engages  in  to  gain  benefits  from  treatment  and   work  toward  the  patient’s  goals.    While  the  focus  of  these  three  components  is  on   the  patient’s  experience  and  desires,  Bordin  emphasized  that  both  the  patient  and   therapist  contribute  to  the  formation  and  quality  of  the  working  alliance.               A  number  of  meta-­‐analyses  have  shown  the  therapeutic  working  alliance  to   be  a  robust  predictor  of  outcome  in  psychotherapy  across  a  variety  of  treatment   modalities  (Castonguay,  Constantino  &  Holtforth  2006;  Martin,  Garske,  &  Davis,   2000;  Horvath  &  Symonds,  1991).    Several  studies  have  suggested  that  a  strong   alliance  is  a  necessary  element  for  positive  change  in  psychotherapy  (Safran,  Muran,   Samstag,  &  Stevens,  2002;  Hartley  &  Strupp,  1983),  while  a  poor  therapeutic  alliance   predicts  patient  dropout  from  treatment  (Martin  et  al.,  2000;  Samstag,  Batchelder,   Muran,  Safran,  &  Winston,  1998).    Because  the  quality  of  the  alliance  is  such  a  strong   predictor  of  outcome  and  dropout  rate,  researchers  have  begun  to  focus  on  the  

  27.  21   process  of  building  a  strong  alliance  and  working  through  impasses  (See   Eubanks-­‐Carter,  Muran  &  Safran,  2010  for  a  review  of  such  studies).   b. Ruptures in the Therapeutic Alliance   The  therapeutic  alliance  is  in  a  constant  state  of  negotiation.    Patients  and   therapists  must  work  together  to  ensure  that  tasks  and  goals  are  appropriate,  all   while  maintaining  an  affective  connection.    A  rupture  in  the  therapeutic  alliance   occurs  when  negotiation  breaks  down  or  there  is  tension  in  the  relationship  (Safran,   Muran,  &  Eubanks-­‐Carter,  2011).    Ruptures  range  in  intensity  from  minor  tensions   to  explosive  blowouts,  and  are  most  likely  present  in  all  therapeutic  relationships.     For  example,  in  a  2009  study  by  Muran  et  al.,  37%  of  patients  and  56%  of  therapists   across  a  range  of  treatment  modalities  reported  ruptures  in  the  first  six  sessions.    In   this  study,  over  half  the  therapists  reported  ruptures  early  on  in  the  treatment,  and   one  also  imagines  additional  ruptures  of  which  patients  and  therapists  were  either   unaware  or  unwilling  to  report.    Hill  et  al.  (1996)  found  that  therapists  are  often  not   aware  of  patients’  dissatisfaction  with  treatment;  when  therapists  do  not  effectively   address  or  resolve  this  dissatisfaction,  patients  are  more  likely  to  terminate   treatment  (Rhodes,  Hill,  Thompson  &  Elliot,  1994).    Given  the  importance  of  the   alliance  to  treatment  outcome  and  retention,  it  is  crucial  that  therapists  be  able  to   identify,  negotiate  and  repair  ruptures  in  the  therapeutic  relationship.     A  number  of  studies  have  suggested  that  ruptures  in  the  alliance  per  se  are   not  detrimental  to  the  therapeutic  relationship,  as  the  process  of  resolving  them   provides  an  opportunity  for  exploration  and  modification  of  a  patient’s  maladaptive   interpersonal  schemas  (Muran,  2001,  Muran  &  Saffran,  2002).      Therefore,  ruptures  

  28.  22   and  their  successful  resolution  may  be  an  important  part  of  the  change  process  in   psychotherapy.         Safran  and  Muran  (2000)  created  a  rupture  resolution  model  that  explains   the  process  of  rebuilding  the  alliance  after  a  rupture.    The  model  includes  two  types   of  ruptures:  confrontation  and  withdrawal  (Harper,  1989a;  Harper,  1989b;  Safran  &   Muran,  2000).    A  confrontation  rupture  is  characterized  by  a  patient  taking  a  hostile,   angry,  or  accusatory  stance  toward  the  therapist  or  some  aspect  of  the  therapeutic   process.    By  contrast,  a  withdrawal  rupture  manifests  when  the  patient  disengages   from  the  therapist,  the  therapy  process,  or  his  own  internal  experience.    Although  a   given  patient  may  be  more  likely  to  present  with  one  type  of  rupture  than  another,   both  types  are  likely  to  emerge  over  the  course  of  treatment.    Resolving  a   confrontation  rupture  involves  exploring  the  patient’s  fears  of  criticism  from  the   therapist  and/or  self-­‐criticism,  which  get  in  the  way  of  expressing  underlying  needs,   and  moving  gradually  toward  the  patient  expressing  more  vulnerable  feelings.    The   process  of  resolving  a  withdrawal  rupture  involves  noting  the  withdrawal  behavior   and  exploring  interpersonal  fears  and  internalized  criticism,  which  get  in  the  way  of   expressing  negative  feelings.    This  process  builds  slowly  toward  self-­‐assertion  and   the  expression  of  underlying  wishes  and  hopes.         Brief  Relational  Therapy  (BRT)  was  developed  according  to  Safran  and   Muran’s  model  of  rupture  resolution  (Safran  &  Muran,  1996).    BRT  is  an   experientially  and  relationally  based  modality  with  an  intensive  focus  on  the  “here   and  now”  of  therapy.    This  focus  allows  patient  and  therapist  to  explore  and   examine  the  patient’s  maladaptive  emotions  as  they  occur  over  the  course  of  a  

  29.  23   rupture,  which  in  turn  helps  to  repair  the  rupture  and  create  a  stronger   relationship  between  the  patient  and  therapist.           Safran,  Muran  and  Eubanks-­‐Carter  (2011)  recently  conducted  a  meta-­‐ analysis  examining  the  relationship  between  rupture-­‐repair  episodes  and  outcome.       They  found  that  the  presence  of  successful  rupture-­‐repair  episodes  was  positively   correlated  with  good  outcome.    However,  only  three  studies  were  included  in  their   meta-­‐analysis,  suggesting  that  research  on  rupture-­‐repair  episodes  and  treatment   outcome  is  still  sparse.       c. Detection of Ruptures in the Therapeutic Alliance   Because  ruptures  may  be  subtle  interpersonal  interactions,  of  which  either   the  patient  or  therapist  may  not  be  fully  aware,  researchers  have  employed  several   methods  for  detecting  ruptures  and  rupture-­‐repair  episodes  (Safran,  Muran,  &   Eubanks-­‐Carter,  2011).  The  three  main  methods  for  identifying  ruptures  are  direct   self-­‐report,  indirect  self-­‐report,  and  observer-­‐based  methods.         One  method  of  identifying  ruptures  is  through  direct  self-­‐report  by  patients   or  therapists.    For  example,  at  the  Brief  Psychotherapy  Research  Program  at  Beth   Israel  Medical  Center  in  New  York  City,  immediately  following  a  session,  patients   and  therapists  are  administered  session  impact  questionnaires,  which  require  them   to  rate  the  quality  of  the  alliance,  the  presence  of  ruptures,  and  the  degree  of   resolution  (PSQ;  Muran,  Safran,  Samstag,  &  Winston,  2004).    However,  there  are  a   number  of  limitations  to  self-­‐report  of  ruptures.    Patients’  and  therapists’  reports  of   rupture  seldom  coincide;  therapists  often  report  more  ruptures  than  their  patients   (Muran  et  al.,  2009).    Also,  patients  and  therapists  may  have  reservations  about  

  30.  24   directly  reporting  ruptures.    Patients  may  avoid  or  deny  problems  in  the  alliance,   because  these  problems  threaten  the  hope  that  the  therapist  can  help  him  or  her;   therapists  may  be  self-­‐conscious  of  their  abilities  and  therefore  deny  any  problems   in  the  relationship.    Because  ruptures,  especially  withdrawal  ruptures,  may  be   subtle,  patients  and  therapists  may  not  report  these  ruptures  because  they  are  not   even  aware  of  them.       One  way  to  counter  the  problems  associated  with  direct  report  of  ruptures  is   to  indirectly  measure  problems  in  the  relationship  through  repeated  post-­‐session   patient  ratings  of  the  alliance.    Researchers  tend  to  use  patient  ratings  of  the  alliance   because  they  are  more  predictive  of  outcome  than  therapist  ratings  (Horvath  &   Symonds,  1991).    For  example,  Stiles  et  al.  (2004)  tracked  changes  in  patient  post-­‐ session  alliance  ratings  over  the  course  of  treatment.    They  found  that  patients   whose  alliance  ratings  had  V-­‐shaped  patterns,  in  which  the  alliance  ratings  dropped   and  then  quickly  returned  to  their  baseline,  made  the  greatest  gains  in  treatment.     Patients  with  U-­‐shaped  patterns,  in  which  alliance  ratings  took  many  sessions  to   return  to  baseline,  made  fewer  gains  in  treatment.    However,  there  are  limitations  to   this  methodology  as  well.    Patients  may  still  hesitate  to  rate  the  therapeutic   relationship  negatively,  even  if  they  are  not  asked  directly  about  problems  within   the  therapeutic  dyad.    In  a  meta-­‐analytic  review  of  79  studies,  Martin,  Garske,  and   Davis  (2000)  found  that  patients  tend  to  rate  the  alliance  as  stable  across  treatment,   while  therapists  and  observers  note  more  change  in  the  alliance.    Thus  it  is  more   difficult  to  observe  changes  in  the  alliance  over  time  through  patient  ratings.     Another  limitation  of  this  methodology  is  that  sessions  in  which  ruptures  have  been  

  31.  25   successfully  resolved  may  be  rated  as  having  a  high  alliance  by  patients,  because   they  were  able  to  successfully  work  through  difficulties.      Researchers  using  this   methodology  may  therefore  miss  these  important  rupture-­‐repair  episodes   altogether.             The  last  method  for  detecting  ruptures  in  the  alliance  is  through  observer-­‐ based  measures.    This  approach  addresses  the  problem  of  patients  and  therapists   underreporting  ruptures  because  of  discomfort  or  lack  of  awareness.    However,   observer-­‐based  methodologies  often  require  the  use  of  session  transcripts,  which   miss  subtle  interpersonal  cues,  and  may  also  require  the  use  of  highly  experienced   clinicians  as  raters  (e.g.  Colli  &  Lingiardi,  2009;  Harper,  1989a,  1989b).    One   exception  is  The  Rupture  Resolution  Rating  System  (3RS;  Eubanks-­‐Carter,  Muran  &   Safran,  2009),  which  follows  Harper’s  (1989a,  1989b)  unpublished  manual  for   coding  withdrawal  and  confrontation  ruptures.    3RS  uses  video  data  instead  of   transcripts,  and  graduate  students  (as  opposed  to  highly  trained  clinicians)  can   become  reliable  coders.  This  coding  system  denotes  the  presence  of  both  types  of   ruptures  and  the  intensity  of  these  markers,  but  cannot  detect  where  in  a  session   ruptures  have  occurred  or  whether  they  were  successfully  resolved.         To  address  the  above-­‐mentioned  methodological  problems,  a  new  measure   was  created:  The  Segmented  Working  Alliance  Inventory—Observer  Form  ,  or  S-­‐ WAI-­‐O  (S-­‐WAI-­‐O;  Berk,  Safran,  Muran  &  Eubanks-­‐Carter,  2010).    Based  on  previous   observer  forms  of  the  Working  Alliance  Inventory,  the  S-­‐WAI-­‐O  is  an  observer-­‐based   coding  system  which  tracks  changes  in  the  alliance  over  the  course  of  a  single   session.    S-­‐WAI-­‐O  codes  can  then  be  used  to  detect  ruptures  and  rupture  repair  

  32.  26   episodes.    An  initial  pilot  study  for  the  S-­‐WAI-­‐O  found  good  inter-­‐rater  reliability   (ICC=.83),  and  S-­‐WAI-­‐O’s  detection  of  rupture  correlated  moderately  with  patient   report  of  rupture  (χ²(1)=4.02,  p=.05)  (Berk,  Safran,  Muran,  Eubanks-­‐Carter,  2010).     The  primary  purpose  of  this  study  was  to  validate  the  S-­‐WAI-­‐O  as  a  measure   of  in-­‐session  variations  in  the  working  alliance.    In  light  of  the  importance  of   negotiating  and  repairing  ruptures  in  the  therapeutic  alliance,  this  study  also  aimed   to  investigate  the  impact  of  specialized  Rupture  Resolution  (RR)  training  on   therapists’  ability  to  successfully  negotiate  alliance  ruptures  and  increase  the   therapeutic  alliance  over  the  course  of  treatment.    Ruptures,  rupture  resolution  and   the  working  alliance  were  measured  using  the  Segmented  Working  Alliance   Inventory—Observer  Form  (S-­‐WAI-­‐O),  which  was  designed  to  track  changes  in  the   quality  of  the  working  alliance  over  the  course  of  a  single  psychotherapy  session   (Berk,  Safran,  Muran  &  Eubanks-­‐Carter,  2010).       II. Hypotheses   The  aim  of  this  study  was  to  establish  psychometric  validity  and  reliability   for  the  S-­‐WAI-­‐O.    The  two  corollary  hypotheses  were  that  therapists  who  undergo   Rupture  Resolution  Training  (RR)  would  be  more  able  to  successfully  resolve   alliance  ruptures  than  those  who  have  not  undergone  this  training,  and  that  working   alliance  ratings  would  increase  after  therapists  undergo  rupture  resolution  training.   III. Method a. Data Collection   All  data  for  this  study  was  collected  from  archival  data  at  the  Brief   Psychotherapy  Research  Program  at  Beth  Israel  Medical  Center  in  New  York  City.  

  33.  27   The  research  program,  which  has  been  running  continuously  since  the  1980's  and   has  received  National  Institute  of  Mental  Health  (NIMH)  funding,  studies  outcome   and  process  variables  related  to  the  therapeutic  relationship  within  short-­‐term   treatments  with  adults  with  personality  disorders.    The  research  program  also   serves  as  a  training  site  for  psychology  externs  and  interns  and  psychiatry  residents.           b. Participants   This  study  comprised  22  therapeutic  dyads  (n=22  patients  and  n  =22   therapists)  sampled  from  therapy  cases  at  the  Brief  Psychotherapy  Research   Program.         c. Therapists The  therapists  used  for  this  study  included  7  males  and  15  females  ranging  in   age  from  24  to  38  (M=29.5,  SD=3.41).      All  therapists  were  psychology  externs  at  the   Brief  Psychotherapy  Research  Program  at  the  time  of  therapy.    Of  the  therapists,   81.8  %  were  Caucasian,  9.1  %  were  Asian,  and  9.1  %  were  Latino.    Clinical   experience  averaged  1.7  years.    The  therapists  provided  informed  consent  with   respect  to  the  parameters  of  the  research  protocol.             d. Patients   The  patients  used  for  this  study  included  10  men  and  12  women,  ranging  in   age  from  26  to  69  (M=45.13,  SD=10.23).    Of  the  patients,  72.72%  were  Caucasian,   9.09%  were  Asian,  and  9.09%  were  African  American.    Most  of  the  patients   (95.92%)  attended  some  college  or  had  a  degree.    Patients  presented  with  a  variety   of  difficulties  related  to  depression,  anxiety,  and  interpersonal  functioning  and  were   assessed  with  the  Structured  Clinical  Interview  for  DSM-­‐IV  (SCID),  Axis  I  &  II  to  

  34.  28   establish  a  diagnosis  (First,  Spitzer,  Gibbon  Williams,  1995).    With  respect  to  Axis   I  disorders,  63.63%  met  for  a  primary  Mood  Disorder  diagnosis,  while  another   27.27%  met  for  an  Anxiety  Disorder;  9.09%  did  not  meet  for  a  disorder  on  Axis  I  of   the  DSM-­‐IV-­‐TR  (American  Psychological  Association,  2000).    Patients  also  met  for  a   Cluster  C  Personality  Disorder  or  Personality  Disorder  Not  Otherwise  Specified   (PDNOS)  on  Axis  II  of  the  DSM-­‐IV-­‐TR;  31.81%  met  criteria  for  a  primary  diagnosis  of   PD  NOS,  31.8%  for  Avoidant  PD,  22.72%  for  Obsessive-­‐Compulsive  PD,  and  4.54%   for  Dependent  PD.    Approximately  50%  of  patients  met  for  multiple  Axis  II   diagnoses.     Exclusion  criteria  included  evidence  of  psychosis,  mania,  an  organic  brain   syndrome,  mental  retardation,  impulse  control  and/or  compulsive  disorders,  as  well   as  any  current  substance  abuse  disorders  or  active  suicidal  behavior.    Patients  must   have  been  stabilized  on  psychiatric  medications  for  at  least  three  months  prior  to   their  intake  assessment,  and  patients  had  to  agree  not  to  begin  medications  during   the  treatment.         Patients  were  recruited  primarily  through  newspaper  advertisements,   hospital  referrals,  doctor  referrals,  and  the  program’s  website.    They  provided   informed  consent  to  the  parameters  of  the  research  protocol  and  paid  a  nominal   per-­‐session  fee  based  on  an  income-­‐sensitive  sliding  scale,  but  were  not  assessed  a   fee  for  the  initial  diagnostic  evaluation.             e. Diagnostic Assessment Patient  eligibility  for  the  study  was  established  through  an  intensive  intake   process,  including  an  initial  phone  screen  and  the  administration  of  the  Structured  

  35.  29   Clinical  Interview  for  DSM-­‐IV,  Axis  I  &  II,  which  was  used  to  determine  diagnostic   status  (First,  Spitzer,  Gibbon  Williams,  1995).    Research  assistants,  who  are  first-­‐  to   fourth-­‐year  graduate  students  in  clinical  psychology,  administered  the  interview.     Training  for  the  research  assistants  included  attending  a  one-­‐day  training  seminar,   viewing  a  training  video,  role-­‐playing,  observation  of  a  live  demonstration,  and   completion  of  an  inter-­‐rater  reliability  test.    This  test  consisted  of  rating  various   videotaped  samples  of  previous  interviews  conducted  by  trained  interviewers;  the   standard  for  completing  training  was  an  intraclass  coefficient  of  >  .70  on  both  the   Axis  I  &  II  sections  of  the  reliability  test.       f. Treatment Model and Therapist Training Procedure Dyads  participated  in  a  30-­‐session,  one-­‐session-­‐per-­‐week  format.    All   sessions  were  videotaped  per  the  requirements  of  the  research  protocol,  and  all   sessions  were  conducted  at  Beth  Israel  Medical  Center  or  The  New  School  in  New   York  City.    The  cases  used  in  this  study  were  sampled  from  those  cases  that  began  in   Cognitive  Behavioral  Therapy  (CBT)  and  switched  into  Rupture  Resolution  Training   (RR)  at  a  predetermined  midpoint.   g. Cognitive Behavioral Therapy   The  central  tenant  of  CBT  is  that  distorted  or  dysfunctional  thinking  affects  a   person’s  mood  and  their  behavior;  the  goal  of  therapy  is  to  correct  this  distorted   way  of  thinking  (Beck,  1995).    Thinking  realistically  and  rationally  will  in  turn   change  a  person’s  behavior  and  mood,  thus  decreasing  psychiatric  symptoms.    The   two  specific  tasks  of  therapy  are  cognitive  restructuring  (i.e.  changing  thinking   patterns)  and  behavior  activation  (i.e.  participating  in  new  or  pleasurable  activities).  

  36.  30   Therapists  at  the  Brief  Psychotherapy  Research  Program  participated  in  a   yearlong  CBT  didactics  seminar,  which  explored  the  basic  tenets  of  CBT  and  focused   more  specifically  on  adapting  those  tenets  to  patients  with  Personality  Disorders.     Therapists  were  taught  a  variety  of  CBT  techniques  including  the  use  of  thought   records,  in-­‐session  role-­‐playing,  exposure  therapy  and  homework  tasks.    Each   therapist  then  completed  a  30-­‐session  CBT  treatment  case.    During  this  case,   therapists  attended  weekly  group  supervision  with  a  highly  experienced  CBT   supervisor.    Therapists’  sessions  were  regularly  monitored  for  adherence  to  the  CBT   model.       h. Rupture Resolution Training RR  training  focuses  on  the  ongoing,  collaborative  exploration  and  negotiation   of  the  therapeutic  alliance,  with  a  specific  emphasis  on  attending  to  and  repairing   ruptures  in  the  therapeutic  alliance  (Safran  &  Muran,  2000).    Therapists  are  trained   to  focus  on  the  here-­‐and-­‐now  of  the  therapeutic  relationship  and  to  consider  their   own  contributions  to  interactions  with  patients.       Once  therapists  completed  their  first  CBT  case  and  established  their   adherence  to  the  CBT  model,  they  were  given  a  second  case,  which  also  began  in  the   CBT  modality.    At  a  predetermined  midpoint  (after  either  session  eight  or  session   16),  therapists  stopped  attending  CBT  supervision  and  switched  into  RR  supervision   for  the  remainder  of  the  30  sessions.    RR  group  supervision  was  led  by  one  of  three   highly  experienced  supervisors,  who  were  very  familiar  with  the  RR  model.   Therapists’  adherence  to  both  the  CBT  and  RR  treatment  model  was  assessed   through  regular  adherence  checks.  

  37.  31   RR  supervision  is  designed  to  help  therapists  foster  an  awareness  of  their   own  feelings  and  reactions  so  that  they  can  use  them  to  develop  personally  and  as   therapists  (Safran,  Muran,  Stevens  &  Rothman,  2007).    RR  supervision  includes  four   main  components:  explicitly  establishing  an  experiential  focus,  active  self-­‐ exploration  on  the  part  of  the  trainee,  focusing  on  the  relational  context  of   supervision  and  the  supervisory  working  alliance,  and  using  the  supervisor  as  a   model  for  the  trainees.       A  typical  RR  supervision  begins  with  a  mindfulness  exercise,  which  helps  to   set  the  tone  for  supervision,  and,  with  time,  helps  to  increase  trainees'  awareness  of   subtle  thoughts,  feelings  and  fantasies  that  arise  while  working  with  the  patient.     Increasing  mindfulness  also  helps  trainees  to  become  more  tolerant  and  accepting   of  a  full  range  of  emotions  and  internal  experiences,  which  can  then  be  used  to  guide   the  treatment.    Next,  a  trainee  will  present  a  moment  or  moments  in  a  specific  case   where  he  felt  unsure  or  confused.    At  the  Brief  psychotherapy  Research  Program,  all   sessions  are  video  recorded,  so  these  moments  are  presented  to  the  group  by   watching  the  tape.    While  watching  these  moments,  the  supervisor  encourages  the   trainee  to  reconstruct  his  feelings  during  the  session.    Other  members  of  the  group   provide  feedback  about  the  patient's  impact  on  them  (e.g.,  I  felt  very  bored  when  the   client  started  to  tell  this  story),  and  the  supervisor  models  his  technique  by   imagining  he  is  in  the  therapeutic  situation  and  talking  about  what  his  thoughts,   feelings  and  intuitions  might  be  during  this  moment  in  therapy.    After  discussing  the   videotape,  the  trainee  engages  in  an  awareness-­‐oriented  role-­‐play  to  promote   emotional  self-­‐awareness  and  to  ground  the  training  at  an  experiential  level.    The  

  38.  32   supervision  session  ends  with  a  group  debriefing  to  allow  the  group  to  give  final   impressions  and  to  check  in  with  the  presenting  trainee.           i. Process Measures In  addition  to  the  administration  of  measures  of  global  outcome  of  treatment,   the  Post  Session  Questionnaire  (PSQ;  Muran,  Safran,  Samstag,  &  Winston,  2002)  was   administered  and  completed  by  therapists  and  patients  following  each  session  (see   Appendix  A,  Figures  9  and  10).    The  PSQ  consists  of  several  self-­‐report  scales   assessing  session  impact  and  the  therapeutic  relationship,  including  the  Session   Evaluation  Questionnaire  (SEQ;  Stiles  &  Snow,  1984),  the  Working  Alliance   Inventory  (WAI;  Horvath  &  Greenberg,  1989),  and  the  Rupture  Resolution   Questionnaire  (RRQ;  Winkelman,  Safran,  &  Muran,  1996).      The  PSQ  also  includes  six   global  questions  designed  to  assess  rupture  and  resolution  within  each  session:  (1)   whether  the  rater  experienced  any  “tension,  problems,  misunderstandings  or   conflicts”  with  the  other  in  the  dyad;  (2)  to  what  extent  the  rater  was  “overly   accommodating  or  overly  protective,”  “making  nice  or  smoothing  things  over,”  or   “holding  back;”  (3)  a  brief  description  of  the  problem;  (4)  the  extent  to  which  the   problem  was  addressed;  (5)  the  degree  to  which  the  problem  was  resolved;  and  (6)   a  brief  description  of  what  contributed  to  the  resolution  of  the  problem.    Questions   1,  2,  4,  and  5  were  assessed  on  a  five-­‐point  Likert  Scale.         The  Working  Alliance  Inventory  (WAI;  Horvath  &  Greenberg,  1989;  WAI-­‐ 12:  Tracey  &  Kokotovic,  1989)  is  a  well-­‐established  psychotherapy  self-­‐report   research  measure  of  the  therapeutic  alliance  between  patient  and  therapist.    It  is   conceptually  based  on  Bordin’s  (1979)  trans-­‐theoretical  model  of  the  working  

  39.  33   alliance  and  comprises  three  subscales,  which  measure  bond,  agreement  on  task,   and  agreement  on  goals.    This  study  will  use  a  12-­‐item  version  of  the  WAI,  which  is   completed  by  both  patient  and  therapist.    The  12  items  are  rated  on  a  seven-­‐point   Likert-­‐type  scale,  where  1  =  “never”  and  7  =  “always.”          The  Session  Evaluation  Questionnaire  (SEQ;  Stiles  &  Snow,  1984)  is  a  self-­‐ report  measure  that  assesses  the  impact  of  a  psychotherapy  session  on  two   subscales,  Depth  and  Smoothness.    The  SEQ  consists  of  12  bipolar  adjective  scales   presented  in  7-­‐point  semantic  differential  format  that  yields  the  two  subscales.    The   Depth  subscale  measures  how  powerful  or  weak  a  particular  session  was;  the   Smoothness  subscale  measures  how  relaxed  or  tense  the  session  was.           The  Segmented  Working  Alliance  Inventory—Observer  Form  (S-­‐WAI-­‐O;   Berk,  Safran,  Muran  &  Eubanks-­‐Carter,  2010)  is  an  observer-­‐based  measure   designed  to  identify  rupture  and  resolution  events  within  a  psychotherapy  session   (See  Appendix  A,  Figure  11).        It  consists  of  12  items,  which  make  up  two  subscales:   task  and  bond.    The  S-­‐WAI-­‐O  is  based  on  the  fourth  revision  of  the  Working  Alliance   Inventory-­‐Observer  Form  (WAI-­‐O)  (Darchuk,  Wang,  Weibel,  Fende,  Anderson  &   Horvath,  2000).    The  items  and  anchors  for  the  S-­‐WAI-­‐O  were  sampled  directly  from   Darchuk  et  al.’s  measure  and  modified  to  suit  the  nature  of  this  coding  system.     Items  were  retained  for  the  S-­‐WAI-­‐O  if  they  showed  good  variance  in  a  pilot  study   (Berk,  Safran,  Muran  &  Eubanks-­‐Carter,  2010).    This  particular  version  of  the  WAI-­‐O   was  selected  because  of  its  unique  coding  guidelines:  typically,  observer-­‐based  WAI   coding  systems  instruct  the  coders  to  assume  a  good  alliance  and  to  subtract  from   their  scores  only  when  evidence  of  a  rupture  is  present,  but  Darchuk  et  al.  included  

  40.  34   more  detailed  anchors  for  each  item  and  instructions  for  coders  to  assume  an   average  alliance  and  deviate  from  this  score  only  when  there  is  evidence  for  or   against  an  item.    These  alterations  counter  the  ceiling  effect  often  seen  in  observer   versions  of  the  WAI  (Raue,  Goldfried  &  Barkham,  1997).       The  S-­‐WAI-­‐O  is  unique  from  other  versions  of  the  WAI-­‐O  because  it  was   developed  to  measure  change  in  the  quality  of  the  working  alliance  over  the  course   of  a  therapy  session.  To  accomplish  this,  S-­‐WAI-­‐O  ratings  are  made  by  coders  every   five  minutes  throughout  the  therapy  session.    During  initial  coding  with  items  from   the  task,  bond,  and  goal  sections  of  the  WAI-­‐O,  very  little  variation  was  found  in  the   goal  items.    Further  investigation  revealed  that  coders  were  consistently  coding  “No   Evidence”  because  the  client  and  therapist  are  not  likely  to  address  goal-­‐related   issues  every  five  minutes.    Because  task  and  goal  items  are  traditionally  very  highly   correlated  and  the  concepts  tend  to  overlap,  the  goal  items  were  removed  from  the   measure.    This  left  S-­‐WAI-­‐O  with  12  items,  six  for  task  and  six  for  bond.             A  pilot  study  was  conducted  to  establish  initial  reliability  and  validity  (Berk,   Safran,  Muran  &  Eubanks-­‐Carter,  2010).    In  this  study  the  S-­‐WAI-­‐O  was  used  to  code   23  CBT  sessions  from  eight  therapeutic  dyads.    Coders  were  able  to  establish  good   reliability  per  segment  with  an  average  intraclass  correlation  coefficient  (ICC)  of   0.82.    S-­‐WAI-­‐O’s  detection  of  ruptures  had  a  moderate  and  significant  correlation   with  patient  report  of  rupture  (χ²(1)=4.02,  p=0.05),  but  did  not  correlate  with   therapist  report  of  rupture.       Procedure  

  41.  35   j. Case and Session Selection   The  dataset  (N  =22)  comprised  CBT/RR  cases  with  sufficient  PSQ  and  video   data.    Half  of  the  cases  (n=11)  switched  from  CBT  to  RR  supervision  after  session  8   was  completed,  and  the  other  half  (n=11)  switched  to  RR  supervision  after  session   16.    All  cases  had  to  have  video  data  from  sessions  5-­‐8,  14-­‐16  and  22-­‐24  to  be   selected.    Six  sessions  were  selected  from  each  case:  two  sessions  from  the  early   phase  of  treatment,  two  from  the  middle  phase  of  treatment,  and  two  from  the  later   phase  of  treatment.    Sessions  were  selected  to  take  advantage  of  the  multiple   baseline  design  of  the  study,  which  accounts  for  the  effects  of  time  and  maturation.   k. S-WAI-O Coding and Scoring To  measure  ruptures  and  their  resolution  within  each  session,  coders  were   trained  by  the  primary  author  until  they  reached  a  group  reliability  of  0.70  or  higher   as  measured  by  a  single  measures  intraclass  correlation  coefficient  (ICC)  on  all   segments  for  three  consecutive  sessions.    Once  the  coders  reached  reliability,  two   coders  coded  each  session  independently,  and  their  reliability  for  each  segment  was   monitored.    If  coders’  reliability  was  below  0.60,  a  third  coder  was  added  and  the   two  most  reliable  coders’  scores  were  used.    A  third  coder  was  added  for  six   sessions;  reliability  was  usually  low  for  these  sessions  because  there  was  little   variance.       After  a  session  was  coded,  S-­‐WAI-­‐O  scores  for  each  segment  were  averaged,   creating  a  global  working  alliance  score.    The  S-­‐WAI-­‐O  scores  were  then  analyzed   using  a  method  called  control  charting.    Walter  Shewhart  (1931)  developed  control   charting  as  an  engineer  at  Western  Electric,  a  manufacturing  arm  of  the  Bell  

  42.  36   Telephone  Company.    Shewhart  designed  the  control  chart  to  distinguish   between  normal,  predictable  variation  and  unpredictable  variation,  which  would   signal  a  need  for  intervention.    Control  charts  are  still  used  today  in  manufacturing   and  are  part  of  the  Six  Sigma  quality  control  movement  that  began  at  Motorola   Corporation  (Pande,  Neuman,  &  Cavanagh,  2000).    Eubanks-­‐Carter,  Gorman  &   Muran  (2010)  demonstrate  how  Shewhart  control  charts  can  be  used  to  detect   changes  in  patient  ratings  of  the  alliance  over  a  30-­‐session  treatment.    Their  work   was  the  inspiration  for  using  control  charts  with  the  S-­‐WAI-­‐O.       The  Shewhart  control  chart  is  a  line  graph  with  the  independent  variable   (time)  on  the  X-­‐axis  and  the  dependent  variable  (in  this  case  the  S-­‐WAI-­‐O  average)   on  the  Y-­‐axis  (see  Figure  1  as  an  example).    A  horizontal  line  represents  the  mean  of   the  dependent  variable;  additional  horizontal  lines  are  placed  equidistantly  above   and  below  the  mean  to  represent  the  upper  control  limit  (UCL)  and  lower  control   limit  (LCL),  respectively.    These  control  limits  describe  a  statistically  defined   confidence  interval,  in  this  case  two  standard  deviations  from  the  mean  in  each   direction  (a  95%  confidence  interval).    Although  industry  commonly  uses  three   standard  deviations  (99.7%  confidence  interval),  two  standard  deviations  are   recommended  when  there  are  fewer  than  20  observations,  as  is  the  case  in  this   study  (Wild  &  Seber,  2000).    

  43.  37   Figure 1. Control Chart   As  noted  above,  control  charts  are  used  to  detect  significant  variation  from   the  mean.    Several  rules  have  been  established  to  determine  whether  variation  is   significant  or  “out  of  control”;  the  most  commonly  cited  rules  are  based  on  Western   Electric  Company’s,  and  can  be  remembered  as  “ones,  runs  and  trends”  (Western   Electric,  1956).    Western  Electric  defined  ones  as  any  data  point  that  falls  either   above  or  below  the  control  limits;  runs  as  seven  consecutive  points  above  or  below   the  mean;  and  trends  as  seven  or  more  consecutive  points,  moving  up  or  down,  that   bisect  the  mean.         These  rules  have  been  modified  for  specific  datasets  in  previous  studies  (e.g.,   Wheeler  &  Chambers,  1992).    Because  the  sessions  coded  in  this  dataset  typically   have  only  10  data  points,  specific  rules  were  developed  to  fit  this  study  as  well.  In   this  study  a  negative  deviation  from  the  mean,  as  detected  by  control  charting,  

  44.  38   constitutes  a  rupture  in  the  therapeutic  alliance.    The,  “ones,  runs  and  trends”  that   demarcate  a  rupture  are:   Ones:  any  data  point  that  falls  below  the  lower  control  limit     Runs:  three  or  more  consecutive  points  below  the  mean   Trends:  three  or  more  consecutive  points  moving  down  that  bisect  the  mean   A  rupture  is  considered  resolved  once  the  ones,  runs  or  trends  are  corrected  (in   other  words,  when  the  following  point  returns  within  the  control  limits,  moves  to  or   above  the  mean,  or  begins  to  ascend,  respectively).    An  entire  session  may  be   considered  a  rupture  session  if  the  S-­‐WAI-­‐O  scores  are  all  below  4,  which  indicates  a   poor  alliance,  even  if  there  are  no  ones,  runs  or  trends  present.    Also,  if  the  first   segment  of  the  session  falls  below  the  control  limits,  the  total  S-­‐WAI-­‐O  score  must   be  below  four,  as  four  constitutes  no  evidence,  which  is  often  selected  when  the   patient  and  therapist  are  beginning  a  session.       IV. Results a.Psychometric Validity of S-WAI-O   An  exploratory,  principle  components  factor  analysis  with  a  forced,  two-­‐ factor  solution  and  a  varimax  rotation  was  run  to  establish  the  Task  and  Bond   subscales  (See  figure  2).    The  results  yielded  two  factors  with  most  of  the  items   loading  onto  their  theoretical  components.    However,  Bond  item  seven  (There  is  a   sense  of  discomfort  in  the  relationship)  loaded  highly  onto  both  components  and   more  strongly  onto  the  Task  component.    

  45.  39   Figure 2. S-WAI-O Factor Analysis Item   Task  1   Task  2   Task  3   Task  4   Task  5   Task  6   Bond  7   Bond  8   Bond  9   Bond  10   Bond  11   Bond  12     Component  1:  Task   0.76   0.82   0.81   0.85   0.81   0.81   0.64   0.47   0.34   0.47   0.29   0.47   Component  2:  Bond   0.43   0.36   0.33   0.34   0.34   0.39   0.58   0.76   0.86   0.79   0.88   0.76   To  ensure  that  the  S-­‐WAI-­‐O  is  a  psychometrically  sound  instrument,  both   reliability  and  validity  needed  to  be  established.    In  other  words,  the  measure   should  be  consistent  and  should  measure  the  constructs  it  is  intended  to  measure.     In  this  study,  good  segment  inter-­‐rater  reliability  was  established  (Average  ICC   M=0.79,  SD=0.07,  range=0.64-­‐1.00)  (See  Figure  3  for  descriptive  statistics).      A   dependent  samples  t-­‐test  found  no  significant  differences  in  reliability  scores  for   CBT  and  RR  sessions  (t(1307)=1.39,  p=0.67).    Construct  validity  for  the  measure   was  established  by  finding  evidence  of  convergent  and  divergent  validity.     Convergent  validity  was  found  by  comparing  the  S-­‐WAI-­‐O’s  report  of  rupture  and   resolution  to  that  of  the  patient  and  therapist  on  the  PSQ;  comparing  S-­‐WAI-­‐O   scores  to  patient  and  therapist  report  of  the  working  alliance;  and  comparing  S-­‐ WAI-­‐O  scores  to  the  smoothness  index  of  the  SEQ.      

  46.  40   To  establish  a  relationship  between  S-­‐WAI-­‐O,  patient  report  of  rupture   and  therapist  report  of  rupture,  both  the  S-­‐WAI-­‐O  and  the  patient  and  therapist   PSQs’  report  of  rupture  and  resolution  were  converted  to  a  binary  variable  (either   “Yes,  there  was”  or  “No,  there  was  not”  a  rupture)  and  a  chi-­‐square  regression  was   run.    There  was  a  significant,  positive  relationship  between  S-­‐WAI-­‐O’s  report  of   rupture  and  patient  report  of  rupture  (χ²  (1)=  9.08,  p=0.003,  Φ=0.23).  (Note:  the  phi   coefficient  (Φ)  is  used  to  detect  effect  size  for  chi-­‐square  tests  and  can  be   interpreted  as  follows:  0.1-­‐0.2  is  a  weak  association,  0.2-­‐0.4  is  a  moderate   association,  0.4-­‐0.6  is  a  relatively  strong  association,  0.6-­‐0.8  is  a  strong  association,   and  0.8-­‐1.0  is  a  very  strong  association  (Rea  &  Parker,  2005)).    There  was  not  a   significant  relationship  between  S-­‐WAI-­‐O  and  therapist  report  of  rupture  (χ²  (1)=   1.72,  p=0.19,  Φ=0.13)  or  between  patient  and  therapist  report  of  rupture  (χ²   (1)=0.04,  p=0.43,  Φ=0.22).    These  results  are  consistent  with  the  S-­‐WAI-­‐O  pilot   study.    Interestingly,  when  the  two  rupture  questions  on  the  PSQ  were  compared   between  the  patient  and  the  therapist,  there  was  a  significant  association  between   patient  and  therapist  report  of  the  first  rupture  question,  which  asks  about   problems  or  tension  (χ²  (1)=  11.54,  p=0.  001,  Φ=0.29)  but  no  association  between   patient  and  therapist  report  of  the  second  rupture  question,  which  asks  about   accommodation  or  "holding  back"  (χ²  (1)=  0.02,  p=0.88,  Φ=-­‐0.01).    This  finding   makes  sense,  given  that  these  behaviors,  which  one  might  associate  with   withdrawal  ruptures,  are  often  very  subtle,  so  either  the  patient  or  therapist  might   not  be  aware  of  them.        

  47.  41   Similarly,  for  ratings  of  resolution,  S-­‐WAI-­‐O,  patient  and  therapist  report   of  resolution  was  converted  to  a  binary  variable  (either  “Yes,  there  was”  or  “No,   there  was  not”  resolution)  and  a  chi-­‐square  regression  was  run.    Again,  there  was  a   significant  relationship  between  S-­‐WAO-­‐I’s  detection  of  resolution  and  patient   report  of  resolution  (χ²  (1)=  7.01,  p=  0.008,  Φ=0.23),  and  no  significant  relationship   between  S-­‐WAI-­‐O’s  report  of  resolution  and  therapist  report  of  resolution  (χ²   (1)=1.78,  p=0.18,  Φ=0.12)  or  patient  and  therapist  report  of  resolution  (χ²  (1)=0.75,   p=0.38,  Φ=0.08).   To  further  establish  convergent  validity,  a  simple  regression  was  run   between  the  S-­‐WAI-­‐O  session  average  and  patient  and  therapist  report  of  the   working  alliance  as  measured  by  the  WAI-­‐S.    Since  the  S-­‐WAI-­‐O  is  based  on  the  WAI,   there  should  be  some  relationship  between  the  measures.    However,  this   relationship  has  not  been  studied  thoroughly;  Tichenor  and  Hill  (1989)  did  not  find   a  relationship  between  client  and  therapist  versions  of  the  WAI  and  an  observer   version  of  the  WAI.    Because  the  S-­‐WAI-­‐O  measures  changes  in  the  alliance  and  is   not  just  an  overall  measure  of  the  alliance,  the  correlations  were  not  expected  to  be   exceptionally  high.    To  run  this  comparison,  S-­‐WAI-­‐O  segment  scores  were  averaged   to  create  a  single  score  for  each  session  and  then  a  correlation  was  run.    There  was  a   moderate  to  high,  significant  correlation  between  the  S-­‐WAI-­‐O  session  average  and   patient  (r=0.49,  p<0.001,  d=0.78)  and  therapist  (r=  0.50,  p<0.001,  d=.05)  reported   WAI-­‐S  scores.    There  was  also  a  significant  correlation  between  patient  and   therapist  WAI  scores  (r=0.58,  p<0.001,  d=0.78).      

  48.  42   The  final  test  of  convergent  validity  was  to  examine  the  relationship   between  the  Smoothness  subscale  of  the  SEQ  and  the  presence  of  ruptures  as   measured  by  the  S-­‐WAI-­‐O.    Smoothness  scores  were  expected  to  drop  as  ruptures   were  detected.    A  point-­‐biserial  correlation  was  run,  resulting  in  a  significant,   negative  relationship  between  S-­‐WAI-­‐O’s  report  of  rupture  and  patient  (rpb=-­‐0.29,   p=0.01)  and  therapist  (rpb=-­‐0.20,  p=0.01)  report  of  session  smoothness.       To  establish  divergent  validity,  a  point-­‐biserial  correlation  was  run  between   the  S-­‐WAI-­‐O’s  report  of  rupture  and  patient  and  therapist  ratings  of  Depth  on  the   SEQ.    This  variable  was  chosen  to  demonstrate  divergent  validity  because,  while   some  ruptures  may  initiate  greater  session  depth  and  exploration,  others  may  be   dealt  with  in  a  very  superficial  manner.    Therefore,  the  concepts  overlap  but  should   not  correlate  significantly;  this  hypothesis  was  confirmed  as  there  was  not  a   significant  relationship  between  S-­‐WAI-­‐O’s  report  of  rupture  and  patient  (rpb=0.005,   p=0.98)  and  therapist  (rpb=0.04,  p=0.69)  report  of  session  depth.   Interestingly,  when  the  relationship  between  S-­‐WAI-­‐O's  report  of  resolution   and  patient  and  therapists'  ratings  of  Smoothness  and  Depth  were  examined,  there   was  a  significant,  negative  relationship  between  rupture  repair  and  Smoothness  for   both  the  patient  (rpb=-­‐0.28,  p=0.002)  and  the  therapist  (rpb=-­‐0.21,  p=0.01).     However,  there  was  no  relationship  between  S-­‐WAI-­‐O  ratings  of  repair  and  Depth   for  either  the  patient  (rpb=0.03,  p=0.76)  or  therapist  (rpb=0.05,  p=0.59).      

  49.  43   Figure 3. Descriptive Statistics   S-­‐WAI-­‐O  Session  Average,  Mean  (SD)   Rupture  Frequency       S-­‐WAI-­‐O     Patient       Therapist   Resolution  Frequency       S-­‐WAI-­‐O     Patient     Therapist   SEQ  Smoothness,  Mean  (SD)     Patient     Therapist   SEQ  Depth,  Mean  (SD)     Patient     Therapist       CBT   RR   Entire  Therapy   5.05  (0.76)     62.9%   19.7%   38.6%     51.5%   16.7%   20.5%     4.58(1.27)   4.09  (1.04)     5.24  (1.09)   4.56  (0.98)     5.12  (0.65)     62.1%   19.7%   30.3%     54.5%   13.6%   16.7%     4.69  (1.30)   4.09  (0.98)     5.26  (1.04)   4.46  (0.88)     5.00  (0.85)     63.6%   19.7%   47%     48.5%   19.7%   24.2%     4.47  (1.23)   4.09  (1.11)     5.22  (1.17)   4.66  (1.06)     b. The Effects of Rupture Resolution Training   After  establishing  the  psychometric  validity  of  the  S-­‐WAI-­‐O,  two  corollary   hypotheses  were  tested:  (1)  that  therapists  who  undergo  Rupture  Resolution   Training  (RR)  will  be  more  able  to  successfully  resolve  alliance  ruptures  than  those   who  have  not  undergone  this  training,  and  (2)  that  S-­‐WAI-­‐O  ratings  will  increase   after  therapists  undergo  rupture  resolution  training.    To  test  these  hypotheses,  a   Generalized  Estimating  Equation  (GEE)  was  run  with  S-­‐WAI-­‐O  ratings  of  rupture   and  repair,  and  overall  session  S-­‐WAI-­‐O  scores  as  the  dependent  variables.    Modality   (CBT  or  RR)  and  the  time  of  the  switch  to  RR  training  (either  session  8  or  16)  were   the  independent  variables.    It  was  expected  that  reports  of  rupture  would  decrease   immediately  after  RR  training  was  introduced  and  that  working  alliance  ratings   would  increase  after  RR  training  was  introduced,  because  the  training  emphasizes   repairing  ruptures  and  creating  a  stronger  therapeutic  alliance.    There  were  no   significant  differences  between  the  two  modalities  (See  figure  4  for  results).          

  50.  44   Figure 4. Changes in Ratings by Modality and Time of Switch Variable   Wald  Chi-­‐Square  (df)   S-­‐WAI-­‐O  Rupture   S-­‐WAI-­‐O  Repair   S-­‐WAI-­‐O  Average   Significance   0.91   0.90   0.53   0.01  (1)   0.01  (1)   0.40  (1)   V. Control Chart Examples   The  following  section  will  describe  the  content  of  four  rupture  sessions  by   using  the  sessions'  control  charts  to  illustrate  an  unresolved  and  resolved  rupture   from  the  CBT  portion  of  therapy  and  the  RR  portion  of  therapy.       Figure 5. Resolved CBT Rupture, Session 15   Although  neither  the  patient  nor  the  therapist  reported  a  rupture  in  this  CBT   session,  S-­‐WAI-­‐O  detected  a  resolved  rupture  beginning  in  segment  seven  and