Behavioral and DTI Studies on Normal and Impaired Learning of Musical Structure
A talk for CogSci 2013 in Berlin, August 1, 2013
Youtube video is here: http://www.youtube.com/watch?v=h1PnbDIhOXA
3. Tone-deafness: a disorder of pitch
perception and action
Congenital amusia
Inability to sing in tune
Incidence: 4 – 17%
Montreal Battery of
Evaluation for Amusia
Inability to discriminate pitch
>1 semitone threshold
(musicianbrain.com/pitchtes
t)
4. What is the source of musical
knowledge?
Frequency
Probability
Pitch
Harmony
Melody
Perception
5. Existing musical systems confound learning
with memory
Test learning with new frequencies &
probabilities
New musical system
Tone-deafness
We need a system to assess
implicit music learning
6. Bohlen-Pierce
A new tuning system – the BP scale
F = 220 * 2 n/12
F = 220 * 3 n/13
200
300
400
500
600
700
0 1 2 3 4 5 6 7 8 9 10 11 12 13
increments (n)
frequency(Hz)
Western
Loui, Wessel, & Hudson Kam, 2010, Music Perception
7. The BP scale can form chords
200
300
400
500
600
700
0 1 2 3 4 5 6 7 8 9 10 11 12 13
increments (n)
frequency(Hz)
F = 220 * 3 n/13
Bohlen-Pierce
3 : 5 : 7
Loui, Wessel, & Hudson Kam, 2010, Music Perception
8. Composing in the Bohlen-Pierce
scale
10 7 10 10
6 4 7 6
0 0 3 0
F = 220 * 3 n/13
Krumhansl, 1987;
Loui, Wessel, & Hudson Kam, 2010, Music Perception
9. Composing melody from harmony –
applying a finite-state grammar
10 7 10 10
6 4 7 6
0 0 3 0
Loui, Wessel, & Hudson Kam, 2010, Music Perception
11. Can we learn the B-P scale?
General design of behavioral studies:
1. PRE-TEST
assess baseline
2. EXPOSURE to melodies in one grammar
~30 minutes
3. POST-TESTS
assess learning
12. Learning a musical system:
Probability sensitivity
Can we remember old melodies?
2-AFC test of recognition
Can we learn new melodies?
2-AFC test of generalization
16. Learning a new musical system:
Frequency sensitivity
Can we learn to expect frequent tones?
Probe tone ratings test
Probe tone profiles reflect frequencies of compositions
Krumhansl, 1990
19. Correlating ratings with exposure
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Pre
Correlation(r)
Post
Exposure
Loui, Wessel & Hudson Kam, 2010, Music Perception
**
** p <
0.01
20. Participants:
15 tone-deaf, 20 control
Matched for age, sex, number of years of musical training
Pre-test 30-min. exposure Post-test
Pre- vs. Post-Exposure
Tone-deaf vs. Controls
Probe tone test: Melody Tone
Probe tone profiles reflect frequencies in musical
compositions (Krumhansl 1990)
Statistical learning in tone-deaf individuals
(In progress)
Jan Iyer
25. What are the neural substrates
of music learning?
STG IFG
Superior Temporal Gyrus (STG)
Inferior Frontal Gyrus (IFG)
Mandell et al, 2007; Hyde et al, 2007
29. Tract volume reflects individual
differences in learning
Volume of right ventral arcuate
fasciculus is correlated with
generalization score, but not with
recognition score.
r = 0.53, p = 0.03
0
5
10
15
20
25
0 0.5 1 1.5
RIFG–RMTG
Tractvolume(103mm3)
Generalization
(proportion correct)
Loui, Li, & Schlaug (2011) NeuroImage
r = 0.054, n.s.
0
5
10
15
20
25
0 0.5 1
Recognition
(proportion correct)
30. Crucial junction of arcuate fasciculus
predicts learning behavior
Search for Fractional Anisotropy correlates of
generalization performance
FA (white matter integrity) of temporal-parietal junction
predicts individual differences in pitch-related learning.
p < 0.05 FWE
Loui, Li, & Schlaug (2011) NeuroImage
31. Behavioral implications of individual
differences in structural connectivity in
statistical learning
Normal Tone-deaf
Tracts from right STG
Loui, Alsop, & Schlaug, 2009, Journal of Neuroscience
33. Experiments now available for
download
http://figshare.com/articles/Bohlen_Pierc
e_scale_artificial_grammar_learning_expe
riment/75772
Also at
http://psycheloui.com/publications/down
loads
Max/MSP format
Several versions with melodies included
34. Acknowledgements
Gottfried Schlaug
BIDMC, HMS
Music and Neuroimaging Lab
(http://musicianbrain.com)
Katy Abel
Rob Ellis
Anja Hohmann
Jan Iyer
Charles Li
Berit Lindau
Christoph Mathys
Sang-Hee Min
Matthew Sachs
Catherine Wan
Jasmine Wang
Anna Zamm
Xin Zheng
David Alsop
BIDMC, HMS
Carol Krumhansl
Cornell University
University of California at Berkeley
David Wessel
Center for New Music & Audio
Technologies
Erv Hafter
Auditory Perception Lab
Carla Hudson Kam
Language & Learning Lab
Bob Knight
Helen Wills Neuroscience Institute
35. Take-home
Much of what we know and love about music is acquired
via statistical sensitivity to the frequency and probability
of occurrence of events in the auditory environment.
This statistical learning mechanism relies on intact white
matter connectivity between temporal and frontal lobe
regions, and may subserve multiple auditory-motor
functions including language as well as music.
Editor's Notes
I want to know how this happens.
TONE-DEAFNESS: A DISORDER OF PITCH PERCEPTION – AND ACTIONCongenital amusia(Grant-Allen, 1878)Incidence: 4 – 17%Inability to sing in tuneInability to discriminate pitch >1 semitone threshold
These questions really get at the core of music cognition and neuroscience more generally, which is what are the core sources of knowledge in music? What exactly needs to be learned in order to be musical? Many here have asked that question and I think we can all agree that pitch is a fundamental source of musical information. So, part of musical competence is the ability to perceive pitch. And we already know that pitch perception is disrupted at least in some tone-deaf individuals. But we also know that pitches don’t exist in isolation. Pitches are strung together to form musical structure. Pitches that are important in a piece occur at a higher frequency, and this gives rise to harmony and tonality. Pitches that are highly probable given other pitches gives rise to melodic structures such as motifs. So to understand musical structure, it’s really the frequencies and probabilities, and how the brain learns to compute them implicitly, that we need to try to understand.
So how do we go about trying to understand how the brain learns frequencies and probabilities of pitches? Well, as we said, most people have already had so much exposure to Western music that even people without musical training show implicit knowledge of the frequencies and probabilities of Western musical sounds. What we really need is a new system of pitches with new frequencies and probabilities that are different from Western music. And this would give us a high degree of experimental control, so that we can systematically manipulate what frequencies and probabilities they get exposed to. To that end, in the past few years we have developed an “alien” or a “Martian” musical scale based on an alternative musical system known as the Bohlen-Pierce scale. Then bycomparing tone-deaf people and matched controls in the way they learn the statistics of music, we can really get at the degree to which different types of musical knowledge might be learnable.
Previous studies from our lab and others have found differences between tone-deaf people and controls in both the structure and the function of the brain. In particular, these brain differences center around two brain regions: the superior temporal gyrus (STG) and inferior frontal gyrus (IFG).
Diffusion of water Infer connectivity about biological matterArcuate fasciculus
Brain-behavioral correlations:If we believe that the right superior AF is predictive of tone-deafness, then people who are more tone-deaf should have less of that connection…What about that inferior branch of the right AF? Superior AF correlates with perceptual thresholdInferior AF correlates with perception-production mismatch
N = 16, high-resolution DTI and learning and memory tasks
Synaptic pruning or abnormal neuronalmigrationInsights into a pitch disability: perception and productionIf we know that this bundle of connections is directly related to language ability, then mapping its relationship with music is revolutionizing how we think about language and its disorders, and how we can rehabilitate them.
Music requires learning from multiple sources of knowledgeMuch of what we know and love about music is acquired via statistical sensitivity to the frequency and probability of occurrence of events in the auditory environment. This statistical learning mechanism relies on intact white matter connectivity between temporal and frontal lobe regions, and may subserve multiple auditory-motor functions including language as well as music. By combining the approaches of an artificial musical grammar with musical disorders, the hope is to understand nature vs. nurture interactions in music in the brain.