The effect of conditional probability of chord progression in Western music corpus on brain...

74
THE EFFECT OF CONDITIONAL PROBABILITY OF CHORD PROGRESSION IN WESTERN MUSIC CORPUS ON BRAIN RESPONSE: AN MEG STUDY 2010-09-11 Department of Brain and Cognitive Sciences | SNU Seung-Goo KIM

Transcript of The effect of conditional probability of chord progression in Western music corpus on brain...

THE EFFECT OF CONDITIONAL PROBABILITY OF CHORD PROGRESSION IN WESTERN MUSIC

CORPUS ON BRAIN RESPONSE: AN MEG STUDY

2010-09-11Department of Brain and Cognitive Sciences | SNU

Seung-Goo KIM

THE EFFECT OF CONDITIONAL PROBABILITY OF CHORD PROGRESSION IN WESTERN MUSIC

CORPUS ON BRAIN RESPONSE: AN MEG STUDY

2010-09-11Department of Brain and Cognitive Sciences | SNU

Seung-Goo KIM

INTRODUCTION

NEURONAL CURRENT

http://www.cnsforum.com/content/pictures/imagebank/hirespng/placement_EEG_leads.png

NEURONAL CURRENT

http://www.cnsforum.com/content/pictures/imagebank/hirespng/placement_EEG_leads.png

NEURONAL CURRENT

http://www.cnsforum.com/content/pictures/imagebank/hirespng/placement_EEG_leads.png

http://www.nmr.mgh.harvard.edu/meg/talks.php

NEURONAL CURRENT

http://www.cnsforum.com/content/pictures/imagebank/hirespng/placement_EEG_leads.png

EEG & MEG

Prof.  Chang-­‐Hwan  Im,  2010,  Class

EEG & MEG

Prof.  Chang-­‐Hwan  Im,  2010,  Classhttp://www.nmr.mgh.harvard.edu/meg/talks.php

INSTRUMENTS

http://www.calvin.edu/academic/psych/experiential/research.html

http://www.virtualmedicalcentre.com/healthinvestigations.asp?sid=11

http://www.nmr.mgh.harvard.edu/meg/talks.php

rMusical  scale§ Tonal  systems  within  an  octave  such

as  major  or  minor

rChord§ Simultaneous  tones  such  as  (C,E,G)

rChord  function§ The  function  of  a  chord  within  a  scale

• (C,E,G)  is  Tonic  (I)  in  C  Major  and  also  Dominant  (V)  in  F  Major.

rChord  progression§ Changes  of  chord  functions§ “Syntactic”  rules  govern  this  progression  in  Western  music

CHORD PROGRESSION

rMusical  scale§ Tonal  systems  within  an  octave  such

as  major  or  minor

rChord§ Simultaneous  tones  such  as  (C,E,G)

rChord  function§ The  function  of  a  chord  within  a  scale

• (C,E,G)  is  Tonic  (I)  in  C  Major  and  also  Dominant  (V)  in  F  Major.

rChord  progression§ Changes  of  chord  functions§ “Syntactic”  rules  govern  this  progression  in  Western  music

CHORD PROGRESSION

PROCESS OF MUSICAL SYNTAX

rKoelsch,  2000,  J.  Cog.  Neurosci.§ Regular  (50%),  N6  at  the  3rd  pos  (25%),  N6  at  the  5th  pos  (25%)

C:  I  I6  IV  V7    I                C:  I    I6  N6  V7  I              C:  I  I6  IV  V7  N6

Koelsch  et  al.,  2000,  J.  Cog.  Neurosci.

PROCESS OF MUSICAL SYNTAX

rKoelsch,  2000,  J.  Cog.  Neurosci.§ Regular  (50%),  N6  at  the  3rd  pos  (25%),  N6  at  the  5th  pos  (25%)

C:  I  I6  IV  V7    I                C:  I    I6  N6  V7  I              C:  I  I6  IV  V7  N6

Koelsch  et  al.,  2000,  J.  Cog.  Neurosci.

NEURAL GENERATORS OF ERAN

rLocated  usingMEG  &  fMRI

rSeveral  studieslocated  IFG

rSometimesbilaterallythus  simply  EAN(esp.  in  women)

Maess  et  al,  2001,  NN;  Koelsch  et  al.,  2005,  TICS;  Koelsch  et  al.,  2003,  Neuroreport.

NEURAL GENERATORS OF ERAN

rLocated  usingMEG  &  fMRI

rSeveral  studieslocated  IFG

rSometimesbilaterallythus  simply  EAN(esp.  in  women)

Maess  et  al,  2001,  NN;  Koelsch  et  al.,  2005,  TICS;  Koelsch  et  al.,  2003,  Neuroreport.

AUTOMATICITY & ATTENTION

rAttended  condition§ detect  the  intensity,  

not  the  deviant  chord§ (Task-­‐irrelevant

-­‐target  condition:  to  avoid  P300b)

rUnattended  condition§ ignoring  all  chord  

with  ‘hard’  reading  comprehension  task

Loui  et  al.,  2005,  Cog.  Brain  Res.

MUSICAL TRAINING

Koelsch  et  al.,  2002,  Psychophysiol.

rERAN  was  greater  in  experts§ Not  in  deviant  instrument  MMN

MUSICAL TRAINING

Koelsch  et  al.,  2002,  Psychophysiol.

rERAN  was  greater  in  experts§ Not  in  deviant  instrument  MMN

SHORT-TERM-EXPERIENCE: HABITUATION

r2  Hours  of  chords  with  silent  movie» Koelsch,  Jentschke,  2008,  Brain  Research.

§ Target(supertonic)-­‐irrelevant  task(timber  detection)

rDeclined;  but  not  abolished

Koelsch  et  al.,  2008,  Brain  res.

OTHER IN/OUT-OF-KEY CHORDS

rNeapolitan  6th  is  out-­‐of-­‐key  chord§ What  if  with  other  chords?

Koelsch  et  al.,  2007,  Psychophysiol.

OTHER IN/OUT-OF-KEY CHORDS

rNeapolitan  6th  is  out-­‐of-­‐key  chord§ What  if  with  other  chords?

Koelsch  et  al.,  2007,  Psychophysiol.

EAN VS. MMN

rMismatch  Negativity  (MMN)§ Well-­‐known  ERP  response  to  the  rare  stimuli  (ex:  S,  S,  S,  D,  S,  S,  ….)

• Similar  in  terms  of  latency,  related  neural  generators,  

§ The  rarity  depends  to  the  short-­‐term  rarity• Same  stimuli  could  be  either  a  standard  or  a  deviant

Opitz  et  al.,  2002,  Neuroimage.

EAN VS. MMN

rEAN:  Invariant  to  the  short-­‐term  rarity§ The  irregular  chords  § The  rarity  of  the  progression  is  rather  dependent§ MMN  can  be  elicited  by  roving  paradigm

• EAN  can  be  evoked  even  with  equiprobability

Garrido  et  al.,  2008,  Neuroimage.

Roving  Paradigm:Firstly,  it’s  oddballLastly,  it’s  standard

SHORT-TERM PROBABILITY & MMN

Javitt  et  al.,  1998,  Biol  Psy.;  Haenschel  et  al.,  2005,  J.  Neurosci.

NEWLY LEARNED PITCH PATTERNS

rUsing  artificial  scales  (Bohlen-­‐Pierce)§ Standard  pitch  pattern  (70%)  &  Deviant  pitch  pattern  (20%)§ 100  sequences  X  10  runs  =  total  40  min

Loui  et  al.,  2009,  J.  Neurosci.

NEWLY LEARNED PITCH PATTERNS

rUsing  artificial  scales  (Bohlen-­‐Pierce)§ Standard  pitch  pattern  (70%)  &  Deviant  pitch  pattern  (20%)§ 100  sequences  X  10  runs  =  total  40  min

Loui  et  al.,  2009,  J.  Neurosci.

NEWLY LEARNED PITCH PATTERNS

rUsing  artificial  scales  (Bohlen-­‐Pierce)§ Standard  pitch  pattern  (70%)  &  Deviant  pitch  pattern  (20%)§ 100  sequences  X  10  runs  =  total  40  min

Loui  et  al.,  2009,  J.  Neurosci.

NEWLY LEARNED PITCH PATTERNS

rUsing  artificial  scales  (Bohlen-­‐Pierce)§ Standard  pitch  pattern  (70%)  &  Deviant  pitch  pattern  (20%)§ 100  sequences  X  10  runs  =  total  40  min

Loui  et  al.,  2009,  J.  Neurosci.

NEWLY LEARNED PITCH PATTERNS

rUsing  artificial  scales  (Bohlen-­‐Pierce)§ Standard  pitch  pattern  (70%)  &  Deviant  pitch  pattern  (20%)§ 100  sequences  X  10  runs  =  total  40  min

Loui  et  al.,  2009,  J.  Neurosci.

NEWLY LEARNED PITCH PATTERNS

rUsing  artificial  scales  (Bohlen-­‐Pierce)§ Standard  pitch  pattern  (70%)  &  Deviant  pitch  pattern  (20%)§ 100  sequences  X  10  runs  =  total  40  min

Loui  et  al.,  2009,  J.  Neurosci.

MOTIVATIONS

rPossible  underlying  mechanism  of  EAN  has  been  suggested  as  ‘Probabilistic  learning’  depending  on  the  previous  exposures.§ Unlike  the  MMN  depending  on  the  present  situation

» Koelsch,  2009,  Psychophy.

§ Learning  of  chord  pattern  could  evoke  EAN» Loui  et  al.,  2009,  J.  Neurosci.

rHowever,  the  direct  relation  of  the  frequency  of  exposure  in  terms  of  chord  progression  and  the  physiological  responses  has  not  been  investigated  yet.

AIMS OF THE PRESENT STUDY

rIf  the  EANm  responses  are  dependent  on  the  probability  in  previous  exposure

rIf  this  effect,  if  any,  interacts  with  the  musical  expertise  and  the  individual  ability  of  discriminating  chord  progressions

METHODS

APPROXIMATION OF FREQUENCYrConditional  Probability  from  J.  S.  Bach’s  Chorale

§ Empirical  probability  as  an  approximation  of  frequencies  in  exposure  of  music  listeners  in  Western  classical  context

P(I|V)=0.75; P(vi|V)=0.11; P(ii|V)=0.05Rohrmeier,  2007,  Sound  and  Music  Computing  Conference,  Greece.

TonicP(I|V)=0.75

SubmediantP(vi|V)=0.11

SupertonicP(ii|V)=0.05

I vi ii V

MATERIAL

•  1  chord=0.6  sec  (100  BPM)•Modulated  to  12  keys•  Consecutively  presented

•  Note  that  (vi)  and  (ii)  occurred  before  the  last  chord  to  control  on-­‐line  rarity  of  chord  itself

TonicP(I|V)=0.75

SubmediantP(vi|V)=0.11

SupertonicP(ii|V)=0.05

I vi ii V

MATERIAL

•  1  chord=0.6  sec  (100  BPM)•Modulated  to  12  keys•  Consecutively  presented

•  Note  that  (vi)  and  (ii)  occurred  before  the  last  chord  to  control  on-­‐line  rarity  of  chord  itself

TonicP(I|V)=0.75

SubmediantP(vi|V)=0.11

SupertonicP(ii|V)=0.05

I vi ii V

MATERIAL

•  1  chord=0.6  sec  (100  BPM)•Modulated  to  12  keys•  Consecutively  presented

•  Note  that  (vi)  and  (ii)  occurred  before  the  last  chord  to  control  on-­‐line  rarity  of  chord  itself

TonicP(I|V)=0.75

SubmediantP(vi|V)=0.11

SupertonicP(ii|V)=0.05

I vi ii V

MATERIAL

•  1  chord=0.6  sec  (100  BPM)•Modulated  to  12  keys•  Consecutively  presented

•  Note  that  (vi)  and  (ii)  occurred  before  the  last  chord  to  control  on-­‐line  rarity  of  chord  itself

PARTICIPANTS

r11  Non-­‐musicians  &  9  Musicians  (N=20)§ after  exclusion  for  excessive  MEG  or  EOG  artifacts§ As  no  male  musicians  could  be  recruited,  to  control  the  sex  ratio  

between  groups,  all  participants  were  female§ Other  2  participants  were  discarded  for  exceed  EOG  rejections

Mean  (SD) N Age  (Yr) Handedness  (EI)

Musical  training  (Yr)

Practice  (Hour/Week)

Non-­‐musicians 11 24.93  (2.5) 91.49  (9.27) 3.63  (1.63) 1.0  (3.0)

Musicians 9 23.50  (3.4) 88.64  (12.78) 18.67  (4.0) 10.22  (6.74)

MEG RECORDING

r1  Block  =  100  trails  (6  min)§ 3  kinds  of  sequences  x  30  =  90  trials§ 3  kinds  of  “staccato”  sequences  =  10  trials

rTask:  click  the  left  mouse  button  on  “staccato”  chord  (randomly  at  the  2nd  ~5th  position)§ To  preserve  the  arousal  level  of  the  participants

rTotal  6  Blocks  with  breaks  (App.  1  Hour)§ 180  trails  were  presented  for  each  condition

rElekta  Neuromag  306-­‐channel  whole-­‐head  MEG§ Anatomical  landmarks,  HPI  coils  and  points  upon  the  head  digitized§ Sampling  rate=600Hz;  BPF=0.1~200  Hz

MEG RECORDING

r1  Block  =  100  trails  (6  min)§ 3  kinds  of  sequences  x  30  =  90  trials§ 3  kinds  of  “staccato”  sequences  =  10  trials

rTask:  click  the  left  mouse  button  on  “staccato”  chord  (randomly  at  the  2nd  ~5th  position)§ To  preserve  the  arousal  level  of  the  participants

rTotal  6  Blocks  with  breaks  (App.  1  Hour)§ 180  trails  were  presented  for  each  condition

rElekta  Neuromag  306-­‐channel  whole-­‐head  MEG§ Anatomical  landmarks,  HPI  coils  and  points  upon  the  head  digitized§ Sampling  rate=600Hz;  BPF=0.1~200  Hz

BEHAVIORAL TEST

rIdentify  the  last  chord  (Tonic/  Submediant/  Supertonic)  of  the  sequences  using    a  keypad  in  the  same  environment  with  MEG  recording

r1  Block=  3  kinds  x  12  keys  =  36  questionsrTotal  3  blocks  (App.  15  min)

rExamples  for  each  chord  progressions  were  presented  before  each  block

SIGNAL PROCESSING

rSpatio-­‐temporal  filter  applied  to  eliminate  environmental  noise  (MaxFilterTM)

rAll  blocks  transformed  to  the  head  position  of  the  first  block

rEpoch:  -­‐200  ~  +500  msec  to  the  target  onset§ Epochs  with  EOG  artifacts  discarded  (<10%)

rAveraged  across  blocks  for  each  condition§ Mean  averaged  trials  for  each  condition  >150

rBand  pass  filter§ 1~20  Hz  with  zero  phase  for  ECD  analysis

EQUIVALENT CURRENT DIPOLE MODEL

http://www.nmr.mgh.harvard.edu/meg/talks.php

EQUIVALENT CURRENT DIPOLE MODEL

http://www.nmr.mgh.harvard.edu/meg/talks.php

EQUIVALENT CURRENT DIPOLE MODEL

http://www.nmr.mgh.harvard.edu/meg/talks.php

MULTIPLE ECD ANALYSIS W/ BESA

rMultiple  ECDs  were  fitted  as  followings:§ Generators  of  P2m  activities  were  fitted  using  genetic  algorithm  

implemented  in  BESA®  (Brain  Electrical  Source  Analysis;  NeuroScan)• For  the  latency  of  180-­‐190  ms  in  the  temporal  regions

§ Additional  two  bilateral  ECDs  for  EANm  responses  were  fitted• The  time-­‐window  of  140-­‐220  ms  after  onset• Seeding  from  the  frontal  regions  bilaterally  for  the  condition  of  

Supertonic  as  the  frontal  activities  were  shown  consistently  in  the  previous  literatures  (accumulated  goodness  of  fitting  was  over  80%)

§ The  activities  of  ECDs  were  estimated  for  each  condition

MULTIPLE ECD ANALYSIS W/ BESA

rMultiple  ECDs  were  fitted  as  followings:§ Generators  of  P2m  activities  were  fitted  using  genetic  algorithm  

implemented  in  BESA®  (Brain  Electrical  Source  Analysis;  NeuroScan)• For  the  latency  of  180-­‐190  ms  in  the  temporal  regions

§ Additional  two  bilateral  ECDs  for  EANm  responses  were  fitted• The  time-­‐window  of  140-­‐220  ms  after  onset• Seeding  from  the  frontal  regions  bilaterally  for  the  condition  of  

Supertonic  as  the  frontal  activities  were  shown  consistently  in  the  previous  literatures  (accumulated  goodness  of  fitting  was  over  80%)

§ The  activities  of  ECDs  were  estimated  for  each  condition

MULTIPLE ECD ANALYSIS W/ BESA

All  in-­‐key  chords

rMultiple  ECDs  were  fitted  as  followings:§ Generators  of  P2m  activities  were  fitted  using  genetic  algorithm  

implemented  in  BESA®  (Brain  Electrical  Source  Analysis;  NeuroScan)• For  the  latency  of  180-­‐190  ms  in  the  temporal  regions

§ Additional  two  bilateral  ECDs  for  EANm  responses  were  fitted• The  time-­‐window  of  140-­‐220  ms  after  onset• Seeding  from  the  frontal  regions  bilaterally  for  the  condition  of  

Supertonic  as  the  frontal  activities  were  shown  consistently  in  the  previous  literatures  (accumulated  goodness  of  fitting  was  over  80%)

§ The  activities  of  ECDs  were  estimated  for  each  condition

MULTIPLE ECD ANALYSIS W/ BESA

All  in-­‐key  chords

rMultiple  ECDs  were  fitted  as  followings:§ Generators  of  P2m  activities  were  fitted  using  genetic  algorithm  

implemented  in  BESA®  (Brain  Electrical  Source  Analysis;  NeuroScan)• For  the  latency  of  180-­‐190  ms  in  the  temporal  regions

§ Additional  two  bilateral  ECDs  for  EANm  responses  were  fitted• The  time-­‐window  of  140-­‐220  ms  after  onset• Seeding  from  the  frontal  regions  bilaterally  for  the  condition  of  

Supertonic  as  the  frontal  activities  were  shown  consistently  in  the  previous  literatures  (accumulated  goodness  of  fitting  was  over  80%)

§ The  activities  of  ECDs  were  estimated  for  each  condition

MULTIPLE ECD ANALYSIS W/ BESA

rMultiple  ECDs  were  fitted  as  followings:§ Generators  of  P2m  activities  were  fitted  using  genetic  algorithm  

implemented  in  BESA®  (Brain  Electrical  Source  Analysis;  NeuroScan)• For  the  latency  of  180-­‐190  ms  in  the  temporal  regions

§ Additional  two  bilateral  ECDs  for  EANm  responses  were  fitted• The  time-­‐window  of  140-­‐220  ms  after  onset• Seeding  from  the  frontal  regions  bilaterally  for  the  condition  of  

Supertonic  as  the  frontal  activities  were  shown  consistently  in  the  previous  literatures  (accumulated  goodness  of  fitting  was  over  80%)

§ The  activities  of  ECDs  were  estimated  for  each  condition

MULTIPLE ECD ANALYSIS W/ BESA

Supertonic

rMultiple  ECDs  were  fitted  as  followings:§ Generators  of  P2m  activities  were  fitted  using  genetic  algorithm  

implemented  in  BESA®  (Brain  Electrical  Source  Analysis;  NeuroScan)• For  the  latency  of  180-­‐190  ms  in  the  temporal  regions

§ Additional  two  bilateral  ECDs  for  EANm  responses  were  fitted• The  time-­‐window  of  140-­‐220  ms  after  onset• Seeding  from  the  frontal  regions  bilaterally  for  the  condition  of  

Supertonic  as  the  frontal  activities  were  shown  consistently  in  the  previous  literatures  (accumulated  goodness  of  fitting  was  over  80%)

§ The  activities  of  ECDs  were  estimated  for  each  condition

MULTIPLE ECD ANALYSIS W/ BESA

Supertonic

rMultiple  ECDs  were  fitted  as  followings:§ Generators  of  P2m  activities  were  fitted  using  genetic  algorithm  

implemented  in  BESA®  (Brain  Electrical  Source  Analysis;  NeuroScan)• For  the  latency  of  180-­‐190  ms  in  the  temporal  regions

§ Additional  two  bilateral  ECDs  for  EANm  responses  were  fitted• The  time-­‐window  of  140-­‐220  ms  after  onset• Seeding  from  the  frontal  regions  bilaterally  for  the  condition  of  

Supertonic  as  the  frontal  activities  were  shown  consistently  in  the  previous  literatures  (accumulated  goodness  of  fitting  was  over  80%)

§ The  activities  of  ECDs  were  estimated  for  each  condition

MULTIPLE ECD ANALYSIS W/ BESA

rMultiple  ECDs  were  fitted  as  followings:§ Generators  of  P2m  activities  were  fitted  using  genetic  algorithm  

implemented  in  BESA®  (Brain  Electrical  Source  Analysis;  NeuroScan)• For  the  latency  of  180-­‐190  ms  in  the  temporal  regions

§ Additional  two  bilateral  ECDs  for  EANm  responses  were  fitted• The  time-­‐window  of  140-­‐220  ms  after  onset• Seeding  from  the  frontal  regions  bilaterally  for  the  condition  of  

Supertonic  as  the  frontal  activities  were  shown  consistently  in  the  previous  literatures  (accumulated  goodness  of  fitting  was  over  80%)

§ The  activities  of  ECDs  were  estimated  for  each  condition

MULTIPLE ECD ANALYSIS W/ BESATonic

rMultiple  ECDs  were  fitted  as  followings:§ Generators  of  P2m  activities  were  fitted  using  genetic  algorithm  

implemented  in  BESA®  (Brain  Electrical  Source  Analysis;  NeuroScan)• For  the  latency  of  180-­‐190  ms  in  the  temporal  regions

§ Additional  two  bilateral  ECDs  for  EANm  responses  were  fitted• The  time-­‐window  of  140-­‐220  ms  after  onset• Seeding  from  the  frontal  regions  bilaterally  for  the  condition  of  

Supertonic  as  the  frontal  activities  were  shown  consistently  in  the  previous  literatures  (accumulated  goodness  of  fitting  was  over  80%)

§ The  activities  of  ECDs  were  estimated  for  each  condition

MULTIPLE ECD ANALYSIS W/ BESATonic

Submediant

rMultiple  ECDs  were  fitted  as  followings:§ Generators  of  P2m  activities  were  fitted  using  genetic  algorithm  

implemented  in  BESA®  (Brain  Electrical  Source  Analysis;  NeuroScan)• For  the  latency  of  180-­‐190  ms  in  the  temporal  regions

§ Additional  two  bilateral  ECDs  for  EANm  responses  were  fitted• The  time-­‐window  of  140-­‐220  ms  after  onset• Seeding  from  the  frontal  regions  bilaterally  for  the  condition  of  

Supertonic  as  the  frontal  activities  were  shown  consistently  in  the  previous  literatures  (accumulated  goodness  of  fitting  was  over  80%)

§ The  activities  of  ECDs  were  estimated  for  each  condition

MULTIPLE ECD ANALYSIS W/ BESATonic

SubmediantSupertonic

STATISTICAL ANALYSES

rRepeated  measure  ANOVA§ Within-­‐subject:  Chord  (3  levels);  Between-­‐subject:  Group  (2  levels)

rGeneral  Linear  Model  (GLM)§ Additional  quadratic  model  to  test  the  effects  of  training

RESULTS

ECD LOCATIONS

Mean  (SEM)  in  head  coordinatex  (mm) y  (mm) z  (mm)

lHG -­‐42.43  (1.16) -­‐0.52  (1.57) 46.61  (1.25)

rHG 46.06  (1.30) 6.35  (1.90) 47.69  (1.29)

lIFG -­‐36.27  (2.15) 18.13  (4.97) 65.77  (3.52)

rIFG 41.99  (2.42) 19.89  (4.89) 63.56  (3.54)

Hotelling’s  t2  :  t2(3,  36)=  39.07,  p<  0.0001);  Right:  t2  (3,  36)=  21.62,  p<  0.001)

ECD ACTIVITIES (N=20)

Shading  indicates  the  time-­‐window  of  [140,  220]  ms  after  onsetActivities  didn’t  differ  across  hemispheres  (t(19)=  -­‐0.63,  p>  0.5)  thus  pooled

EFFECTS OF CHORD & TRAINING

rEANm  response  was  defined  as  the  mean  of  absolute  values  in  the  time  window  [0.14,  0.22]

rRepeated  measure  ANOVA  on  the  EANm  responses  rThe  effect  of  chord  functions  

§ F(1.188,  21.389)=  34.636,  p<  0.0001  (Greenhouse-­‐Geisser’s  corr.)

rThe  effect  of  musical  training  with  an  interaction  § F(1.188,  21.389)=  4.304,  p=  0.0044  (Greenhouse-­‐Geisser’s  corr.)

CORRELATION WITH PROB. & TRAINING

The  quadratic  model  considering  the  effects  of  musical  training  and  interaction  explained  significantly  better  (F(2,  56)=  6.4390,  p=  0.0030)

CORRELATION WITH BEHAVIORAL MEASURES

The  correlation  between  the  correct  rates  for  Submediants  and  normalized  EANm  activities  was  significantly  positive  (r=  0.6244,  p=0.0400)

DISCUSSION

EFFECT OF CONDITIONAL PROBABILITY

rEnding  chord  which  has  the  lower  conditional  probability  elicited  the  greater  EANm  response§ F(1.188,  21.389)=  34.636,  p<  0.0001  (Greenhouse-­‐Geisser’s  corr.)

rIt  is  not  due  to  acoustic  deviance  but  syntactic§ All  chord  functions  were  played  within  the  sequence  before  ending

rAlso  the  EANm  responses  correlates  with  the  probability  interacting  with  musical  training§ Musicians:  r=  0.4440;  Non-­‐musicians:  r=0.3929

IFG AND PROBABILITY

sgKIM  (2010)  ThesisEAN/  MEG  

Maess  et  al.,  (2001)  NNERAN/  MEG  

Opitz  et  al.,  (2002)  NIMMN/  fMRI

Koelsch  et  al.,  (2005)  NIERAN/  fMRI

RR R

VENTRAL PREFRONTAL CORTEX & PROB.

↓ Repeated  “X”

Casey  et  al.,  (2001)  HBM

↑ Other  than  “X”  

INDIVIDUAL DIFFERENCES

rFor  experts,  the  behavioral  scores  were  not  normally  distributed  § most  of  them  reached  the  maxima  (“Ceiling  effect”)

rFor  the  normal  group,  Submediant  scores  highly  correlate  with  the  overall  scores  (r=  0.91)

rThe  EANm  could  be  considered  as  reflecting  the  individual  competence  of  identifying  subtle  aberration§ Thus  it  could  be  suggested  as  the  physiological  index  of  the  

probabilistic  representation  of  individuals

LIMITATIONS

rmusicians  showed  enhanced  responses  to  the  short-­‐term  deviants  &  irregular  chords  (ERAN)

rThis  might  implicate  the  facilitated  sensitivity  to  musical  syntax  or  the  familiarity  to  the  corpus  in  the  present  study

Huron,  2006,  Sweet  AnticipationBaroque  Corpora Popular  music  Corpora

LIMITATIONS

rEffects  of  the  melody  &  chord  function§ Different  top  voices  might  have  confounded  the  effect  of  chord  

functions§ But  not  likely,

rDifferent  melody§ Doesn’t  affect  the

time-­‐windowlater  than  100  ms.

Koelsch  &  Jentschke  (2010)  J.  Cog  Neurosci.

CONCLUSION

rThe  effects  of  the  probability  of  chord  progression  on  the  brain  responses  were  found§ The  negative  correlation  between  the  conditional  probability  

and  the  corresponding  neural  response§ Also  it  was  found  to  be  more  facilitated  by  the  musical  training

rThe  current  results  suggest  the  physiological  response  as  a  reflection  of  the  probabilistic  representations  on  the  musical  syntax.  

rMoreover,  the  results  indicate  that  the  probabilistic  representation  is  related  to  the  musical  training  as  well  as  individual  sensitivity.

CONCLUSION

rThe  effects  of  the  probability  of  chord  progression  on  the  brain  responses  were  found§ The  negative  correlation  between  the  conditional  probability  

and  the  corresponding  neural  response§ Also  it  was  found  to  be  more  facilitated  by  the  musical  training

rThe  current  results  suggest  the  physiological  response  as  a  reflection  of  the  probabilistic  representations  on  the  musical  syntax.  

rMoreover,  the  results  indicate  that  the  probabilistic  representation  is  related  to  the  musical  training  as  well  as  individual  sensitivity.