Mapping Music: Cluster Analysis Of Song-Type Frequencies Within And Between Cultures

Understanding cross-cultural patterns of musical diversity requires some method of visualizing these patterns using maps. The traditional methods of cross-cultural comparison have been criticized for ignoring the rich diversity of musical styles that exists within each culture. We present a compromise solution in which we map the relative frequencies of different “cantogroups” (stylistic song-types) both within and between cultures. Applying this method to 259 traditional group songs from twelve indigenous peoples of Taiwan, we identified five major cantogroups, the frequencies of which varied across the twelve groups. From this information, we were able to create musical maps of Taiwan. (This article refers to a supplementary speadsheet that can be found at http://neuroarts.org/pdf/Savage_Brown_2014_Supplement.xls) E have an ambivalent relationship with maps. On the one hand, maps provide an essential tool for understanding the cross-cultural diversity of musical styles, which has been one of the primary goals of ethnomusicology since its beginnings in the comparative musicology of the late nineteenth and early twentieth centuries (Nettl 2005). On the other hand, ethnomusicologists have increasingly become aware of and interested in intra-cultural diversity, which is more difficult to visualize using maps. Many ethnomusicologists have tried—either visually or verbally—to create maps of the major musical regions of the world following a model we will term here the “one culture = one music” model, in which the basic musical style of a This content downloaded from 130.113.111.210 on Sat, 1 Mar 2014 12:34:29 PM All use subject to JSTOR Terms and Conditions 134 Ethnomusicology, Winter 2014 culture is classified and mapped onto its geographical boundaries. This has often been inspired by similar attempts of anthropologists to create maps of culture areas or Kulturkreis (culture circles). Nettl (2005) offers a thorough review of major attempts at musical mapping and their various strengths and weakness, notably by Sachs (1929) for musical instruments of the world, Roberts (1936) and Nettl (1954) for Native America, Merriam (1959) for Africa, McLean (1976) for Oceania, and Lomax (1968) for world song style. Of these, Lomax’s analysis of over 4000 songs from more than 200 cultures using his Cantometric classification scheme and its distillation into a map of ten major song-style regions (Lomax 1976) was certainly the most comprehensive and arguably the “least unsatisfactory” (Nettl 2005:330) attempt to create a musical map of the world. Leaving aside numerous methodological and political issues involved in generating and interpreting such maps (e.g., Maranda 1970; Erickson 1978, Nettl and Bohlman 1991; Nettl 2005; Savage and Brown 2013), there remains the fundamental problem of how to represent musical diversity on a map. As stated above, most previous attempts at creating musical maps used a one culture=one music model where the basic musical style of a culture was mapped onto its geographical boundaries. Although Lomax (1976) sub-divided some geographic regions—notably Taiwan and Oceania—into several song-style regions, his map did not allow for multiple styles within a single song-style region. McLean’s (1979) map came perhaps the closest to achieving this goal by using dashed and solid lines, and allowing these boundaries to overlap one another. While these maps provided some sense of intra-cultural musical diversity, they offered only a limited sense of the relative frequencies of the different musical styles within a given region. In recent times, ethnomusicologists have come to focus more on intracultural diversity than on cross-cultural diversity, leading some to question the value of musical maps. While earlier scholars who created musical maps were aware of the reductionism inherent in assigning a single musical style to a culture, they also believed that minimal information was lost in the process, with Lomax (1968) in particular arguing that cultures’ musical styles are relatively homogeneous. Others, using their own fieldwork experience however, argued that music was so heterogeneous as to make the mapping of a single “favored song style” (Lomax 1968:33) onto a geographic region all but impossible (Henry 1976; Feld 1984). Recently, an analysis of traditional group songs from Taiwan and the Philippines lent support to these claims, finding that 98 percent of the musical variability of a cross-cultural sample was accounted for by differences within cultures rather than differences between them, a profile strikingly similar to the partitioning of genetic diversity within and between cultures (Rzeszutek, Savage, and Brown 2012).1 While this analysis succeeded in quantifying musical diversity, it did not attempt to provide a better means of mapping it. This content downloaded from 130.113.111.210 on Sat, 1 Mar 2014 12:34:29 PM All use subject to JSTOR Terms and Conditions Population geneticists, however, have long been aware of the issue of intracultural diversity and have developed phylogeographic methods for mapping genes onto geographic regions (Cavalli-Sforza, Menozzi, and Piazza 1994). The most influential method has been to measure and map the frequency of occurrence of various genetic types within and between populations. Unique DNA sequences are referred to as haplotypes, and groups of related haplotypes are referred to as haplogroups. Many cultures contain a diverse mix of individuals representing different genetic haplogroups, and mapping the relative frequencies of these different haplogroups has proven to be a powerful tool for visualizing genetic diversity and tracing prehistoric human migrations (Cavalli-Sforza et al. 1994). To take a simple hypothetical example, a particular world region might contain three haplogroups: A, B, and C. Thirty percent of individuals from one area might possess haplogroup A, 30 percent haplogroup B, and 40 percent haplogroup C, whereas 50 percent of individuals from a second area might possess haplogroup A, 40 percent haplogroup B, 10 percent haplogroup C. Thus, while many genetic types might be shared between the groups in the region, the relative frequencies of these types might vary significantly among them, thus providing clues about their histories. Applying these concepts to music, we introduce the new notion of a cantogroup or stylistic song-type.2 By analogy with genetics, unique combinations of musical features will be referred to as cantotypes, and groups of similar cantotypes as cantogroups. Our method takes its lead from population genetics in that, instead of assigning a single musical style to each geographic region or ethnolinguistic grouping, we characterize a specific area as a mosaic with respect to the frequency of the cantogroups that make up the group of populations under consideration. So, just as a geneticist represents a population of people in terms of the relative frequencies of haplogroups contained within it, and just as different populations vary with respect to the relative proportions of haplogroups, so too can we represent a population in terms of the relative frequencies of cantogroups that make up its musical repertoire. Further, we can represent differences between groups in terms of the varying relative proportions of the cantogroups within each group. The basic idea of comparing relative frequencies of musical features across cultures is not new, being implicit in even the earliest comparative musicological descriptions and being subject to relatively sophisticated statistical analyses by the 1950’s (Freeman and Merriam 1956). Some additional underlying concepts for this analysis have been developed to varying degrees in publications by Lomax (1968, 1976, 1980), Aarden and Huron (2001), Leroi and Swire (2006), and Grauer (2011), but we have synthesized them here for the first time into a comprehensive framework. In order to illustrate this novel method of comparative analysis for music, we used music from twelve groups of indigenous peoples living in Taiwan3—the Savage and Brown: Mapping Music 135 This content downloaded from 130.113.111.210 on Sat, 1 Mar 2014 12:34:29 PM All use subject to JSTOR Terms and Conditions 136 Ethnomusicology, Winter 2014 most diverse region in Lomax’s map—as a test case. We classified 259 songs using the CantoCore classification system (Savage, Merritt, Rzeszutek, and Brown 2012), applied cluster analysis to identify major cantogroups across the twelve populations, and then created maps of relative cantogroup frequencies for each population. Compared to the standard one culture=one music approach, our method permits a quantification and mapping of cross-cultural differences, while at the same time respecting the diversity of musical styles within each group. Indigenous Music In Contemporary Taiwan Taiwan provides an excellent example for a case study in mapping musical diversity. Although it is a small island, it has some of the highest levels of diversity in the world, not only musically but also in other domains, such as linguistics and genetics. While these populations, like indigenous peoples in most parts of the world, have been greatly affected by colonialism and globalization, most have managed to preserve substantial amounts of their musical, linguistic, and genetic heritages even as they have adapted to changing lifestyles. Importantly, Taiwan’s musical diversity has led it to be well-documented by ethnomusicologists from both inside and outside of Taiwan. There are currently fourteen officially recognized groups of indigenous peoples in Taiwan, numbering about 500,000 in total, or about 2 percent of Taiwan’s modern-day population. All of the indigenous peoples originally spoke Austronesian languages and are thought to share descent from a proto-Austronesian culture that first occupied Taiwan at least 5,000 years ago (Blust 1999; Bellwood and Dizon 2008; Gray et al. 2009). Taiwan’s high degree of linguistic diversity is one of the strongest pieces of evidence that has led to it being widely regarded as the primary homeland of the more than 1,000 Austronesian-speaking peoples that are spread throughout the Pacific and Indian Oceans, as far west as Madagascar, as far east as Rapanui (Easter Island), and as far south as New Zealand.4 Nine of the populations (Amis, Atayal, Bunun, Paiwan, Puyuma, Rukai, Saisiyat, Tao, and Tsou) have long been recognized as distinct from one another. Between 2000–2008, five populations that were previously lumped together with nearby groups (Thao, Kavalan, Truku, Sakizaya, and Sediq) were officially recognized as distinct by the government. Many groups, most notably the various Siraya populations in the western plains, still remain unrecognized. Beginning in the seventeenth century, successive colonizations by the Netherlands, China, Japan, and the Republic of China (Taiwan), resulted in the introduction of new musical styles and contexts, such as Christian hymns, folk and classical music from China, and enka and karaoke from Japan. Colonialism also resulted in the suppression of many indigenous musical traditions—particularly those involving headhunting, which was formerly widespread but is no longer This content downloaded from 130.113.111.210 on Sat, 1 Mar 2014 12:34:29 PM All use subject to JSTOR Terms and Conditions practiced today (Kurosawa 1973; Loh 1982; Hsu 2002; Tan 2008). Due to these influences, as well as more general economic, demographic, and cultural trends accompanying globalization, contemporary indigenous Taiwanese music spans a vast variety of forms, including traditional music, with minimal cross-cultural influence; contemporary Chinese, Japanese, and Western music performed and enjoyed by indigenous Taiwanese; and much music with traditional roots that has been substantially influenced by cross-cultural contact. This last category includes not only obvious hybrids, such as Siraya songs sung in Chinese or Christian hymns sung in Bunun style, but also newly-composed urban songs with traditional roots or songs influenced by inter-tribal contact, such as the pan-aboriginal style often marketed to tourists. Tan (2008) described some of the complexities and politics of these interactions, focusing on the famous example of the unauthorized remix of a recording of a traditional Amis weeding song (Enigma’s “Return to Innocence”) that became famous worldwide when it was used to promote the 1996 Atlanta Olympics.