cannaïveobserversdisnguisha$ … · cannaïveobserversdisnguisha$...

38
Can naïve observers dis.nguish a violinist's solo from an ensemble performance? A pilot study 3 rd workshop on Social Behaviour in Music ACM ICMI 2012 – 26 th October Donald Glowinski 1 , Kim TorresEliard 2 , Carlo Chiorri 3 Antonio Camurri 1 and Didier Grandjean 2 1 CasaPaganini Infomus (UNIGE, Italy) 2 NEAD (UNIGE, Switzerland 3 DISFOR (UNIGE, Italy)

Upload: vungoc

Post on 01-May-2018

218 views

Category:

Documents


3 download

TRANSCRIPT

Can  naïve  observers  dis.nguish  a  violinist's  solo  from  an  ensemble  

performance?  A  pilot  study  

3rd  workshop  on  Social  Behaviour  in  Music    ACM  ICMI    2012  –  26th  October    

Donald  Glowinski1,  Kim  Torres-­‐Eliard2,  Carlo  Chiorri3  

Antonio  Camurri1  and  Didier  Grandjean2  

1  CasaPaganini  -­‐  Infomus  (UNIGE,  Italy)  2NEAD  (UNIGE,  Switzerland    3  DISFOR  (UNIGE,  Italy)  

 

Solo  or…  

Ensemble  

Test:    Solo  or  Ensemble?  

Test:  Solo  or  Ensemble?  

Research  QuesYon  

•  Are  they  non-­‐verbal  behavioural  variables  that  enable  to  disYnguish  between  performing  an  acYon  alone  or  jointly  in  a  group?  

Hypothesis  •  Playing  music  jointly  with  others  may  affect  individual  behaviour.    –  Joint  performance  requires  strategies  to  cope  with  others’  acYons  and  to  adapt  one’s  behaviour  accordingly.  Knoblich  et  al.  2011  

•  Auditory  and  visual  non-­‐verbal  behavioural  cues  may  help  in  disYnguishing  between  a  solo  and  a  group  performance  – Non  verbal  cues,  potenYally  used  by  the  audience  characterize  the  social  behavior  and  the  emoYonal  reacYons  of  musicians.  Keller  et  al.  2010  

!

CommunicaYon  performers/observers  

Musicians  

!

Perceived  cues   AaribuYon  of  emoYon  and  social  states  

Non-­‐Verbal  Behavioural  Cues    (audio,  movement)  

Adapted  Lens  Model  Juslin  &  Lindstrom,  2010  

!

CommunicaYon  performers/observers  

Musicians  

!

Perceived  cues   AaribuYon  of  emoYon  and  social  states  

Behavioural  Cues    (audio,  movement)  

Adapted  Lens  Model  Juslin  &  Lindstrom,  2010    

Behavioural  cues  in  music  performance    

1)  key  gestures  using  upper-­‐body  parts  [examples]  -­‐Examples  

-­‐head’s  nod  to  indicate  a  synchronous  start  Davidson  2005  -­‐gaze  interacYon  to  capture  co-­‐performers’  aaenYon  Thompson  2005  

-­‐  dynamic  aspects  of  moYon  features  are  complementary  to  postural  and  gesture  shape-­‐  related  informaYon    Castellano  2008  

-­‐  self-­‐explanatory  gestures  (e.g.,  nods)  Vs.  specific  gestures  typical  of  musicians  Davidson  2005  

Behavioural  cues  in  music  performance    

2)  long-­‐range  behavioural  variaYons  -­‐gradual  and  not  salient  (different  from  key  gestures)    -­‐refer  to  implicit  adaptaYon  and  co-­‐ordinaYon  processes  of  musicians  during  the  performance    Glowinski  et  al.  2010,  Varni  et  al.  2011,  Panos  et  al.  2012      

!

CommunicaYon  performers/observers  

Musicians  

!

Perceived  cues   AaribuYon  of  emoYon  and  social  states  

Behavioural  Cues    (audio,  movement)  

Adapted  Lens  Model  Juslin  &  Lindstrom,  2010    

Perceptual  Experiments  

•  MulYmodal  features  are  used  in  the  evaluaYon  of  music  performance  expressivity  [examples]  -­‐  Audio  features  (e.g.,  Yming)  Juslin  2005  -­‐  Video  features  (e.g.,  energy)  Vines  et  al.  2006    

•  Specific  contribuYons  of  body  parts  and  kinemaYc  informaYon  for  communicaYng  expressivity  -­‐  Video-­‐processed  material  Dahl  et  al  2007  -­‐  Point  light  display  Keller  et  al.  2010  

What  about  the  effect  of  playing  with  an  ensemble?  

•  How  joint  acYvity  of  musician  may  impact  upon  the  observer’s  percepYon?  

>  Revisi.ng  adapted  Lens  Model  under  a  social  perspecYve  

!

Adapted  Lens  Model  (Juslin  &  Lindstrom,  2010)  

 

Musicians  

!

Perceived  cues   AaribuYon  of  emoYon  and  social  states  

Behavioural  Cues    (audio,  movement)  

! !

Musicians   Perceived  cues   AaribuYon  of  emoYon  and  social  states  

Behavioural  Cues    (audio,  movement)  

Revisited  Lens  Model  

Social    Context  (e.g.,  joint  ac.vity)  

M1

M2

Mn

X1

Y2

Zn

XY12

YZ2n

XZ1n

L

PCx1

PCxy12

PCxz1n

PCy2

PCyz2n

PCzn

PCnn

Musicians   Behavioural  cues  (audio,  movement)  

Social  context  (e.g.  joint  acYvity)  

Perceived  cues  

AaribuYon  of  social  states  

Revisited  Lens  Model  

SYmulus  

!

-­‐2  

-­‐1.5  

-­‐1  

-­‐0.5  

0  

0.5  

1  

1.5  

2  take09   take10   take11   take12  

z-­‐score  

Expressivity  

1st  violin  

-­‐2  

-­‐1.5  

-­‐1  

-­‐0.5  

0  

0.5  

1  

1.5  

2  

take09   take10   take11   take12  z  score  

Sa.sfac.on  

1st  violin  

SYmulus  •  60  Selected  Audio/Video  samples  represent  a  

broad  range  of  expressive  performance  (see  post-­‐performance  raYngs)  

Experimental  Design  

•  CondiYon  (Solo  Vs  Ensemble)                                                      

Percep.on  •  Perceived  CondiYon  (Solo  Vs  Ensemble)  •  Level  of  Confidence    (0  -­‐>  100)  

Emo.on  •  9  GEMS  dimension  (0  -­‐>  100)      Wonder  

 Tenderness    Tension    Sadness    Transcendence    Joyful    AcYvaYon    Nostalgia    Power    Peacefulness  

1  

2  

3  

4  

5  

Factors  (IV)   Measure  (DV)  

•  Music  Segment  

     

Body  features    •  Final  Report    

ParYcipants  •  20  parYcipants  (5  males)  from  the  University  

of  Geneva  •  Mean  age  23.3±2.9  years,  range  18−31  

•  Experiment  duraYon  :  1h20  

Results  •  Did  parYcipants  successfully  dis.nguish  between  Solo  

Vs  Ensemble  performance?              >  no  

!

Fisher’s  exact  test  showed  no  significant  associaYon  of  CondiYon  (Solo  vs  Ensemble)  with  the  Perceived  CondiYon    

Results  •  Was  it  possible  to  correctly  disYnguishing  between  

the  two  condiYons?            >  yes  

! !

Received  Operator  Characteris.cs  (ROC)  curves  were  employed  to  assess  each  parYcipant’s  "diagnosYc"  accuracy  

Successful  parYcipant  (AUC  =  .783)   Not  successful  parYcipant  (AUC=  .513)  

Results  •  Analysis  of  the  parYcipants’  raYngs  suggest  they  may  

adopt  strategies  to  decode  social  behaviour            

Results  

Generalized  Linear  Mixed  Model  (GLMM)  to  study  fixed  effects  of  CondiYon  on  Level  of  Confidence,  Music  Segment  and  Emo8on    

•  Analysis  of  the  parYcipants’  raYngs  suggest  they  may  adopt  strategies  to  decode  social  behaviour  

         

Level  of  Confidence  

!

Level  of  Confidence  

Level  of  Confidence  on  perceived  condiYon  (solo  vs  ensemble)  

Perceived  Solo                                  Perceived  Ensemble  

•  Higher  confidence  level  when  reporYng  first  violinist’s  performance  as  being  within  an  Ensemble  

Music  Segment  

Music  Segment  

Bar  Chart  of  the  Music  Segment  Vs  Perceived  CondiYon  

•  Music  Segment  effect  on  parYcipants’  judgments.  

!Music  Segment  

1                      2                          3                      4                    5  

Music  Segment  

Bar  Chart  of  the  Music  Segment  Vs  Perceived  CondiYon  

•  Music  Segment  effect  on  parYcipants’  judgments.  

!

Perceived  Solo   Perceived  Ensemble  

Music  Segment  

1                      2                          3                      4                    5  

Music  Segment  •  Music  Segment  effect  on  parYcipants’  judgments.  

!

Perceived  Solo   Perceived  Ensemble  

Music  Segment  

1                      2                          3                      4                    5  

Segment  4  fugato  wriYng  style  sets  all  musicians  at  the  same  level    Segment  5  first  violinist,  prominent  role      

EmoYons  

EmoYons  •  EmoYons  effect  on  parYcipants’  judgments:    ü  Joy  and  Serenity  more  frequently  associated  with  

Solo  condiYon  

ü  Significant  posiYve  correlaYon  between  Tenderness  and  level  of  confidence  when  correctly  recognizing  Ensemble  condiYon  

ü  Higher  raYngs  of  Nostalgia  and  Sadness  when  correctly  recognizing  Solo  and  Ensemble  condiYon  

 

EmoYons  •  EmoYons  effect  on  parYcipants’  judgments:    ü  Joy  and  Serenity  more  frequently  associated  with  

Solo  condiYon  =>  emoYon  as  a  social  marker?  

ü  Significant  posiYve  correlaYon  between  Tenderness  and  level  of  confidence  when  correctly  recognizing  Ensemble  condiYon  

ü  Higher  raYngs  of  Nostalgia  and  Sadness  when  correctly  recognizing  Solo  and  Ensemble  condiYon      =>  arousal  of  emoYon  related  to  self-­‐confidence  

in  social  judgments?    

Conclusion  

•  A  first  aaempt  to  invesYgate  percepYon  of  social  behaviour  in  music  performance    

-­‐  original  experimental  procedure  (Solo  Vs  Ensemble)  

-­‐  RevisiYng  model  of  communicaYon  between  observer  /  performers  

•  Reveal  possible  strategies  of  non-­‐expert  to  decode  social  behaviour  based  on  non-­‐verbal  cues  

 

Future  work  

•  Possible  tracks  for  future  research  include:  (i)  addressing  one  modality  at  a  Yme  

(ii)  addressing  experts  (creaYon  of  focus  group)  

(iii)  correlaYng  the  results  of  the  perceptual  experiments  with  automated  behavioural  analysis  of  musicians.