dwight krehbiel, professor of psychology bethel college, north newton, ks
DESCRIPTION
Biopsychological Responses to Music Chosen by a Computer: Validation of a Music Search Engine July 31, 2009. Dwight Krehbiel, Professor of Psychology Bethel College, North Newton, KS. Acknowledgments Professor Bill Manaris, Department of Computer Science, College of Charleston, Charleston, SC - PowerPoint PPT PresentationTRANSCRIPT
Biopsychological Responses to Music Chosen by a Computer: Validation of
a Music Search Engine
July 31, 2009
Dwight Krehbiel, Professor of PsychologyBethel College, North Newton, KS
This material is based upon work supported by the National Science Foundation under Grant No. 0849499 and No. 0736480 from the Division of Information and Intelligent Systems and Grant No. 0511082 from the Division of Undergraduate Education.
AcknowledgmentsProfessor Bill Manaris, Department of Computer Science, College of Charleston, Charleston, SC
and his students:Patrick Roos Luca PellicoroThomas Zalonis J.R. Armstrong
who created the search engine used in the experiments
And Bethel student collaborators:Aimee Siebert Tim BurnsJosé Rojas Erin WhiteSonia Barrera Becky BuchtaYue Yu Brittany BakerSierra Pryce Elizabeth FriesenNaomi Graber Lisa Penner
Natural Patterns in Music(and Many Other Phenomena)
Zipf’s law:
The probability of an event of rank f is inversely proportional to that rank f raised to some power n, and n is close to 1.
or
P(f) ~ 1/fn
Example events in music:• pitches, durations, harmonic intervals, melodic intervals• but also pairs of intervals, sets of three intervals, etc.
Basic finding: Music that is Zipfian is generally judged to be more pleasant than is non-Zipfian music.
Web User Interface of the Search Engine
Do listeners like what the search engine finds?
Liking Ratings Excerpt Set A (n=25) Genre Familiarity Ratings
Liking Ratings Excerpt Set B (n=25) Genre Familiarity Ratings
O MS MS2 MD2 MD
O MS MS2 MD2 MD
O MS MS2 MD2 MD
O MS MS2 MD2 MD
O: original
MS: most similar
MS2: 2nd most similar
MD2: 2nd most dissimilar
MD: most dissimilar
Experimental Design & Procedure
Experimental Design (cont)Two sets of excerpts (one min/excerpt)Set A - set of 7 presented to all participants: O = original, MS = most similar, MS2 = 2nd most similar, MS3 = 3rd most similar, MD3= 3rd most dissimilar, MD2 = 2nd most dissimilar, MD = most dissimilarSet B - set of 5 unique to each participant (chosen from one of their three favorite genres): O, MS, MS2, MD2, MDRandom order of all 12 excerpts for each participant40 participantsInstrumental music only
Emotion Rating Instrument
One of Our Participants
Psychological Ratings – Set A
Psychological Ratings – Set B
Posterior Frontal Asymmetry – Set A
Asymmetry Near the Central Sulcus – Set A
n = 38
Skin Conductance Changes during the Music – Set A
(Individual Participants' Data)
Heart Interbeat Intervals
(IBI)
- SetsA & B
(means and standard
deviations across 60
sec of listening
All Psycho-physiological
Measures– Set A
(means across 60 sec of listening)
Summary & Conclusions• A search engine based on aesthetic similarity
can find music that we like, perhaps by finding music with sound patterns that are already familiar.
• Affective responses to similar music found by the search engine (pleasantness, activation, liking) are clearly different from those to dissimilar music.
• Similarity judgments by human participants show clear agreement with search engine ratings.
Summary & Conclusions (cont) Hemispheric asymmetry measures (alpha
power) show significant differences between similar and dissimilar music when all participants listen to the same music, but not when preferences are controlled (i.e. asymmetry is not closely correlated with consciously reported affective responses).
Peripheral psychophysiological responses also display significant differences between similar and dissimilar music. Heart rate differences do appear to be correlated with affective responses.