Examining the Role of National Music Styles in the Works of Non-Native Composers Katherine Vukovics Daniel Shanahan Louisiana State University
The Normalized Pairwise Variability Index Grabe and Low (2000) Provides a metric of rhythmic variability in an utterance Image taken from Patel, 2008
Image taken from Grabe and Low, 2002
Image taken from Daniele, 2008
Image taken from Patel, 2008
Image taken from Daniele and Patel, 2013
Image taken from Hansen, Sadakata, and Pearce 2016
Image taken from Daniele, 2016a
Plan: Using a larger dataset with melodies from a larger number of countries, can we replicate some of these effects? What can we learn about npvi from looking at composers from countries frequently underrepresented in such research? Can we use this expanded dataset to address questions that go beyond nationality, and address aspects of intended nationality?
Cleaning The RISM Dataset RISM (Répertoire International des Sources Musicales) Dataset: Collection of incipits encoded for bibliographic purposes in libraries across the world. Curated by Shanahan, Sapp, and Bell.
Cleaning The RISM Dataset Musical Notation **kern Notation Plaine and Easie (RISM-World) Encoding
The Corpus 1,269,908 incipits!
The Corpus Getting rid of duplicates: 1,004,372 incipits
Composer Nationality
The Corpus Diminishing returns: composers that only had one or two melodies were extremely difficult to find. When we only look at pieces for which we were able to (somewhat) easily grab the nationality of the composer, we have: 668,350 pieces. Today, we ll only be looking at 225,279
Study 1: Examining the Music and Language Relationship
Replicating a Music and Language Comparison Hypothesis 1: Using the RISM corpus of instrumental melodic incipits, we expect melodic npvi to be a significant predictor of composer nationality. More or less replicating Patel and Daniele, 2003 Specifically, we will examine French, German, Italian, Czech, and English composers.
Replicating a Music and Language Comparison Multivariate Logistic Regression: How well can nationality be predicted from an npvi rating in French, German, Italian, Czech and English melodies? (Nationality ~ npvi rating) Significant results: p <.001 (n=29,095) However, the effect size is quite small The odds ratio for the predictors is roughly 1 (1.002, with a 95% CI of 1.000-1.004) This means that the odds of the composer being a certain nationality, given an npvi rating are roughly the same (only very slightly higher) than a composer of a different nationality with a different npvi rating.
Multidimensional Scaling As we are getting significant results, but a small effect size, it might be worth looking at these as visual relationships.
Re-Examining Daniele and Patel, 2013
Image taken from Hansen, Sadakata, and Pearce 2016
Study 2: Examining Intended National Style
What is intended national style? Although a composer may have come from a certain part of Europe, musical styles, such as French keyboard music or Italian opera often transcended nationality. Bukofzer writes that one of the aspects of Bach s genius was his ability to fuse these national styles (Bukofzer, 1947; p.258) Would the intended national style carry more weight than a composer s nationality?
Two Test Cases: Mozart and Handel Can npvi be a significant predictor of Mozart s Italian operatic style and Handel s English oratorios?
Case Study 1: Handel s English Oratorios 993 vocal melodic lines from 58 works (taken from the RISM dataset): 488 Italian 505 English Again, we ran a logistic regression to see: How well can intended national style be predicted from an npvi rating in English and Italian?
Results Again, there were significant results: p <.001 The effect size still continues to be quite small: The odds ratio for the predictors:.99 (with the 95% CI being between.985 and 995) Given a specific npvi, a Handel oratorio is only slightly more likely to be labeled as a certain intended national style.
Case Study 2: Mozart s Italian Operas 1,389 vocal melodic lines from 77 works (taken from the RISM dataset): 840 German 549 Italian Again, we ran a logistic regression to see: How well can intended nationality be predicted from an npvi rating in German and Italian?
Results The results were not significant: p =.735 The effect size, however, continues to be quite small: The odds ratio for the predictors:.99 (with the 95% CI being between.995 and 1.003) Given a specific npvi, a Mozart opera is not significantly more likely to be labeled as a certain intended national style.
Conclusion We hypothesized that npvi would be a significant predictor of intended national style. Results were significant with Handel s vocal music, but the effect size was quite small. Results were not significant with Mozart s vocal music. There are other stylistic features at play: Handel s English works are sacred oratorios, and his Italian works are largely secular operas. Non-rhythmic elements are quite likely a factor. It could be a result from a shift away from national styles.
Discussion npvi is a significant predictor of composer nationality (replicating many earlier findings, most notably Patel and Daniele, 2003) The effect size, however, is quite small. A post-hoc MDS and hierarchical clustering reveals relationships that we would not expect, and this relationship seems to change over time in similarly unexpected ways. The effect of npvi on predicting national styles is inconsistent (being significant for Handel but not for Mozart) The effect size was again quite small. This could point to a shift away from national styles in the post-baroque era.
Thank you!
Questions?
Outtakes
Hypotheses
1700-1709
1710-1719
1720-1729
1730-1739
1740-1749
1750-1759
1760-1769
1770-1779
1780-1789 Belgian, Austrian, Czech, Dutch, English, French
1790-1799
Image taken from Hansen, Sadakata, and Pearce 2016