Music categorization


Published on

If our hypotheses are met, then they support the similarity based view as the predominant conceptual principle for how we recognize music especially those we have heard over and over again. However, training may introduce rule based cognition process to allow wider range of processing in order to achieve cognitive economy.

  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Music categorization

  1. 1. How do you know what kind of music you are hearing on your iPod? Cognitive Psychology San Francisco State University Yovanni Antonelli Spring 2007
  2. 2. Concepts and categorization is an abstract topic to many of us. However, as you can see on this iPod, categorization has useful daily function. How is music recognized and categorized? Our group wants to test the conceptual framework we learned and see how it applies to music. 1
  3. 3. Objectives <ul><li>To determine how music is categorized cognitively </li></ul><ul><li>To evaluate whether similarity based view or explanation based view can better explain the structure and acquisition of unfamiliar forms of music </li></ul><ul><li>To study the influence of prior training in music in the method of conception </li></ul>1
  4. 4. Background (I) <ul><li>The ways in which music can illuminate fundamental issues in cognition have been under-examined, or even dismissed as epiphenomenal. </li></ul><ul><li>However, cognition in music is more and more acknowledged as fundamental to our understanding of cognition as a whole, hence music cognition should be able to contribute both conceptually and methodologically to cognitive science. </li></ul>Honing, Henkjan (2006). &quot;On the growing role of observation, formalization and experimental method in musicology.&quot; Empirical Musicological Review] 1 (1) 2-5. 1
  5. 5. Background (II) <ul><li>Some past research address </li></ul><ul><ul><li>How does the mind recognize a piece of music or a musical style? </li></ul></ul><ul><ul><li>What makes a performance expressive? How is music processed in the brain </li></ul></ul><ul><ul><li>how does it change the brain in the process?” </li></ul></ul><ul><li>Music like another sensory input must be categorized based on similar cognitive concepts </li></ul><ul><li>Little research has been done to specifically address which conceptual view better address the recognition of music </li></ul>1 Zbikowski, LM. Conceptualizing Music: Cognitive Structure, Theory and Analysis. Oxford University Press, 2005.
  6. 6. Background (III) 1 Posner ML, Keele SW (1968). On the Genesis of Abstract Ideas. Journal of Experimental Psychology , 77(3,1), 353-363. Hintzman DL (1986). “Schema Abstraction” in a multple-trace memory model. Psychological Review, 93, 411-428. Murphy GL (1993). Theories and concept formation. In I.Van Mechelen etc : Categories and concepts: Theoretical views and inductive data analysis pp173-200. London, Academic Press. Similarity Based View Explanation Based View Prototype Theory : Based on central tendencies of a category using probabilistic model Exemplar Theory : Drew from past exemplars of a category weighted by similarity to the new object Theory Based, Rule Based, Some understand of the basic P : Not correlative and no context E : More correlative and more context sensitive Sensitive to context, knowledge, theory, experience
  7. 7. Background (IV) <ul><li>These two apparently distinct views have also been seen as special cases of a more general Bayesian learning framework. </li></ul><ul><li>Tenenbaum tested this generalized theory using a number concept game or a word learning task. </li></ul><ul><li>He found that if similarity-liked generalization occurs when one example was given but it rapidly converged to rule based view after as few as three examples were given. </li></ul>Tenenbaum JB (2000). Rules and Similarity in Concept Learning. In Solia A, Leen TK et al Advances Neural Information Processing Systems 12 , 59-65. MIT Press. 1
  8. 8. Hypothesis <ul><li>Music is mainly categorized by similarity based view </li></ul><ul><li>Unfamiliar forms of music are sorted first by explanation based view with further sorting based on familiarity </li></ul><ul><li>Persons with music background used more of explanation based view to categorize new music </li></ul>2
  9. 9. Methods <ul><li>30 subjects (age 18 to 25) with no prior exposure to psychology research will be recruited into the study </li></ul><ul><ul><li>10 of the subjects = no prior music training, Group Novice ( N ) </li></ul></ul><ul><ul><li>10 subjects = formal music training for at least 1 year but not professional, Group Experienced ( E ) </li></ul></ul><ul><ul><li>10 subjects = professional musicians, Group Professional ( P ) </li></ul></ul><ul><li>50 pieces of music (abbreviated to approximately 1 min each) are selected. </li></ul><ul><ul><li>The pieces are chosen from: Classical, R&B, Rock, Hip hop, and Jazz. This forms test set (I), n=15. Each category will have three pieces and all are familiar to most young adults. </li></ul></ul><ul><ul><li>A second test set (II) are chosen from less heard of pieces, from other categories (e.g. classic tunes play in rock), other cultures etc, n=35. </li></ul></ul>3
  10. 10. Methods-Design I <ul><li>The 50 pieces of music played back via a head phone in random order. Each piece of music is played back once. </li></ul><ul><li>For each song, the following question will be displayed: </li></ul><ul><li> This piece of music is best classified as : </li></ul><ul><ul><li>Classical </li></ul></ul><ul><ul><li>Rock </li></ul></ul><ul><ul><li>R & B </li></ul></ul><ul><ul><li>Hip hop </li></ul></ul><ul><ul><li>Jazz </li></ul></ul><ul><li>The subjects are asked to click the choice as quickly as possible and not to wait for the piece to finish. The time from the start of the music to clicking of the answer is recorded. The answer cannot be changed once it is selected. </li></ul>3
  11. 11. Methods-Design II <ul><li>The subjects are also asked to pick a reason why they chose the particular category based on certain attributes of the music: </li></ul><ul><li> You choose this category because of : </li></ul><ul><ul><li>Melody/Tune </li></ul></ul><ul><ul><li>Rhythm/Beat </li></ul></ul><ul><ul><li>Types of instruments/vocal involved </li></ul></ul><ul><ul><li>Artist </li></ul></ul><ul><ul><li>Guess </li></ul></ul>3
  12. 12. Analysis <ul><li>End points : </li></ul><ul><li>The time to answer and accuracy of the three test groups (N,E,P) when the pieces of music (test set I or II) are played. </li></ul><ul><li>The major reasons for picking the category chosen will be correlated to the time and accuracy. </li></ul>4 <ul><li>Independent Variables : Group N,E,P and Music Test Set I and II </li></ul><ul><li>Dependent Variables : time to answer and type of music and reason </li></ul>
  13. 13. Results (expected) Time (sec) % Correct (Reason) Test Set I Test Set I Test Set II Test Set II (M, R) (M, A) (M, R) (R,G) (R,T) (R,G) 4 N E P N E P N E P N E P
  14. 14. Results (read only) <ul><li>The time to answer and accuracy of the three test groups (N,E,P) are not different when the familiar pieces of music (test set I) are used. </li></ul><ul><li>However, it takes longer and becomes less accurate with test set II in each group. </li></ul><ul><li>Also, there is increase time and decrease accuracy in the test set II when we compared the groups (N, E, P) </li></ul><ul><li>The novice seems to rely on the melody to determine but switch to rhythm and guess when the music becomes unfamiliar </li></ul><ul><li>The expert seems to rely on rhythm and type of instruments (timbre) </li></ul>4
  15. 15. Conclusions (I) <ul><li>Whichever mechanism the novice or expert used, it produced a similar outcome (time and accuracy) in categorization when the music is familiar. </li></ul><ul><li>The reason used is predominantly “melody”. </li></ul><ul><li>Since each melody is unique, it is likely that the categorization used is similarity-based view. It is also more consistent with the exemplar model since the pieces provided are indeed exemplars of the category. </li></ul>5
  16. 16. Conclusions (II) <ul><li>When the unfamiliar test set (II) is used, the deterioration in time and accuracy suggest a different or additional categorizing mechanism is in play. </li></ul><ul><li>The Professional performs significantly better than the other two showing that prior experience significantly helps. </li></ul><ul><li>Since the reason given predominantly is rhythm which is not unique and is based on rules—theory. Thus, likely it is that rule based view seems to be used when unfamiliar music is encountered. </li></ul>5
  17. 17. Unexpected Results <ul><li>The experiments may not work since the two music sets may not be sufficiently different to distinguish the underlying mechanisms of cognition. </li></ul><ul><li>The unfamiliar pieces can be confusing to classify and time to classification may denotes confusion predominantly. </li></ul>5
  18. 18. Discussion (I) <ul><li>If our hypotheses are met , then they support the similarity based view as the predominant conceptual principle for how we recognize music especially those we have heard over and over again. However, training may introduce rule based cognition process to allow wider range of processing in order to achieve cognitive economy. </li></ul><ul><li>Detail analysis using the Bayesian model may also shed light on the Tenenbaum hypothesis . </li></ul>5
  19. 19. Discussion (II) <ul><li>If our hypotheses are NOT met , then the experiments may suggest there are no predominant cognitive views in music cognition. </li></ul><ul><li>This experiment will help us understand better how we recognize music and may also delineate the training effect. </li></ul><ul><li>However, this study may have used the wrong categorizations (i.e. classical, R&B etc) since these are really artificial industry standards. The brain may categorize music as enjoyable, not enjoyable or repulsive. Nevertheless this experiment may shed light on how quickly music is recognized and what factors play into its recognition. </li></ul>5