Machinic Ballads: Alan Lomax’s Global Jukebox and the Categorization of Sound Culture
Today, SO! continues its series reconsidering the life and work of Alan Lomax in his centenary year, edited by Tanya Clement of The University of Texas at Austin. We started out with Mark Davidson‘s reflections on what it means to raise questions about the politics behind Lomax’s efforts to record and collect folk music, and continued a few weeks later with Parker Fishel‘s consideration of Lomax’s famous “Southern Journey” and how it has been appropriated by musicians more recently.
With Clement’s own article below, the series begins to rethink Lomax as a touchstone in current and continuing drives to collect, measure and compute sonic cultures, something that seems hot all of a sudden (see, for instance, coverage of recent digital analysis of trends in pop music at Queen Mary University of London). In her thoughtful, illuminating and inspiring article below, Clement challenges us to consider the politics behind these efforts to search, retrieve and analyze audio, something that the case of Lomax throws into stark relief.
— Special Editor Neil Verma
When the Association for Cultural Equity, an organization that Alan Lomax founded in 1983, announced the release of 17,000 music tracks from Lomax’s fieldwork collections, the New York Times heralded the release as a manifestation of Lomax’s Global Jukebox project, a computational experiment for accessing and studying his vast multimedia collection of the world’s culture. The Times piece likens Lomax’s project to Pandora, which allows the listener to search for music “like” music she has already found. Lomax’s biographer, John Szwed, also makes this comparison but modifies his description by proclaiming that unlike Pandora’s recommendations which are “based on personal taste” and “tend to lead sideways . . . to production style,” Lomax’s Global Jukebox idea held the potential to point a listener to “deeper principles of cultural and musical organization” (The Man Who Recorded the World 391).
Gobsmacked by whizbang possibilities, neither the Times nor Szwed discuss the deeper principles behind Lomax’s attempt to represent culture as a global search engine. In the context of the powerful work being accomplished in the Music Information Retrieval (MIR) community and my own project (HiPSTAS) to develop software for making sound collections searchable and accessible, In this article I will argue that how we build systems for searching and retrieving and browsing cultural artifacts as data is a profoundly political act. Recognizing such politics suggests that Lomax’s Global Jukebox project serves as a cautionary tale for how social and cultural contexts — or what Donna Haraway calls our “ways of being” — are reflected in the systems we develop.
The Singer with the Song
The year that Alan Lomax was born (1915), his father John Alan Lomax published a landmark piece heralding seven new types of American ballads for study. American ballads, he argues “reveal the mode of thinking, the character of life, and the point of view, of the vigorous, red-blooded, restless Americans, who could no more live life contented shut in by four walls than could Beowulf and his clan, who sailed the seas around the coasts of Norway and Sweden” (“Some Types of American Folk-Song”, 3). Unlike any other collection of ballads, John’s “American ballad” included the ballads of the miner, the lumbermen, the inland sailor, the soldier, the railroader, “the ballads of the negro; and the ballads of the cowboy . . . [and] the songs of the down-and-out classes, — the outcast girl, the dope fiend, the convict, the jail-bird, and the tramp” (3). Governed by a laudable goal to record the songs of folk cultures at the fringes of mainstream society, the senior Lomax’s view of the communities where he would collect his songs (including jails and state farms), was complex, and can fairly be called both progressive as well as racist (Porterfield 170).
John and Alan went on seven collecting trips together between 1934 and 1936 and co-authored five books on their return. On these trips, they collected songs from people on the street in cities like New Orleans and people in the country, from both church-goers and prisoners. While John held romanticized views of the “noble” southern black man, Alan, on the other hand, indicated a more nuanced understanding of the complexities inherent to his father’s attempt to generalize patterns of “folk” for study. Alan linked “the singer with the song” and was interested in the politics behind prisoners made to sing with guns at their backs and in the cultural lives of people that were so poor in means but so rich in “beautiful harmony, with enormous volume, with total affection” (Szwed 49). While Alan maintained that he was interested in the individual’s story, John believed that “a genuine ballad has no one author. It is therefore the expression of no one mind: it is the product of the folk . . . It might have been written by any one” (“Some Types of American Folk-Song”, 1).
The Global Jukebox project demonstrates an almost complete reversal in Alan’s concerns. The studies behind the Global Jukebox include Alan’s Cantometrics and Choreometrics, in which he produces taxonomies for studying song and dance and his Parlametrics project, an “experiment in metalinguistics,” which Alan and his collaborators describe as a taxonomy of “patterns of style” in speech based on dynamic changes in pitch, loudness, speed, spacing, rhythm, and timbre (“A stylistic analysis of speaking”). These taxonomies show that Alan’s early consideration for the individual performer gave way to a desire to make folk study more scientific as a cultural mapping like what his father espoused rather than what Szwed and others have seen as Alan’s concerns with the situated politics of individuals.
Alan’s Parlametric study serves as good example. Approaching delegates from the United Nations and soliciting mail-in samples from regions not covered by the U.N. volunteers, Alan and his team collected representative recordings of 114 languages. Then, in order to study the “generally neglected meta-communicational level” in these recordings, the team designed a rating system including 50 codes that (1) “described the distinctive features of each recording,” and (2) “tended to cluster the recordings into sets of similars” that Alan maintains anyone could “readily use” to record “salient differences in conversation style” (19). These clusters pointed to 14 factors that Alan and his team would use to categorize the cultures from which they received samples:
- Speech length
- Descending cadence
Using these factors, Alan makes some broad assertions. The association of clear syllabification” (the degree to which syllables run together) “is most strongly predicted among gardeners with domesticated animals” and “[t]he association of clear syllabification to feminine autonomy is suggested by the discovery that this mode of speaking predicts and is predicted by permissive rather than restrictive premarital sexual mores” (27). Further, “Dominance vs. Sharing of conversation space” is strongly correlated with settlement size and severity of sexual sanctions,” a statement that Alan immediately rationalizes by noting that “this relation between a more crowded social space, high sexual tension and increased rate of interaction seems to make good sense, even if it does not account for every possibility” (31).
These spurious and broad generalizations were what Lomax hoped to facilitate for all with his Global Jukebox as the access point for “the first numerical models of the full range of global cultural variation in holistic form” for “the scientist, the layman, and the student to explore, experience, and manipulate the broad universe of culture and creativity in a systematic fashion, with audio-visual illustrations at every turn of the road” (“The Global Jukebox,” 318). By leveraging his taxonomies of song, dance, and speech in the computer age, Alan could suddenly associate and differentiate cultures holistically and en masse.
Machinic Methods / Humanistic Questions
As someone who works in the liminal spaces between the humanities and technology, between cultural studies and critique and the machines that increasingly function both as access points and barriers to our cultural artifacts, I see Alan’s switch to generalizable taxonomies as par for the course in the digital age. My own >HiPSTAS project’s primary objective is to develop a virtual research environment in which users can better access and analyze spoken word collections of interest to humanists. We understand that in order for us to search digital sound artifacts, we have to create taxonomies, metadata, keywords and other generalizable frameworks that facilitate discovery.
At the same time that we are using machinic methods, however, we can still ask humanistic questions that open up rather than close down debates and dialogues. In a recent test for the HiPSTAS project, for example, we used machine learning to analyze the recordings in the UT Folklore Center Archives, which comprises 219 hours of field recordings collected by John and Alan Lomax, Américo Paredes, and Owen Wilson, among others (UT Folklore Center Archives, ca. 1928-1981, Dolph Briscoe Center for American History, University of Texas at Austin, Box 2.325/R). In our attempt to predict the presence of different sonic patterns including instrumental music, singing, and speech, the results of our analysis are noteworthy as the visualization shown in this brief movie demonstrates.
from Tanya Clement on Vimeo
Within the results, we see a visualization of how many seconds comprise each file (in blue) and how many of those seconds for each file our software has predicted the presence of instruments (green), speech (red), and song (purple). A subtle yet striking difference emerges in the comparison between the Lomax recordings (created 1926-1941), which are the oldest in the collection, and the others, which were created up until 1968. The Lomax recordings (primarily created by John Lomax) consistently contain the least amount of speech in comparison to what the other files contain.
Of course, there are a number of ways you can read these results. Given the conversation above, one could hypothesize that perhaps the Lomaxes were primarily interested in their participants’ songs rather than their stories. One could also think about it in terms of recording capabilities across time. When the Lomaxes were first recording, John Lomax writes, “The amplifier weighed more than one hundred pounds; the turntable case weighed another one hundred; two Edison batteries weighed seventy-five pounds each. The microphone, cable, the tools, etc., accounted for sufficient weight to make the total five hundred pounds. . . . In order to carry them in the car I tore out the back seat . . .” Even in 1967, forty years later, good recorders still weighed 70 pounds and required a car battery, but tapes were longer and costs were less. More tape and more time at less cost both financially and physically had a big impact on what researchers recorded. At the same time, the data shows that the later recordings are not much longer, but do seem to have more seconds of speech.
There is a danger in these kinds of machine-generated generalities. We employed taxonomies (instrumental, sung, speech) to teach the machine to categorize these patterns, but why these patterns? Are there others? Or did I choose these based on what I already wanted to say about the Lomaxes’ practices? And, I haven’t even mentioned here the subjective practices inherent to choosing algorithms for such work.
These kinds of questions require more research, and more contextualization than this aggregated data set can show. Just as the ballads that John and Alan Lomax once collected were written and sung by someone, so were the communities that Alan interpreted through his Parlametrics made up of individuals, not types. Perhaps Alan’s desire “to record the world” was just and Google, the collector, categorizer, and interface for all things on the Internet, isn’t evil. But the Global Jukebox Project serves as a cautionary tale about the politics behind the speed and efficiency that machinic methods seem to promise, a politics that needs to be far less opaque about its deeper principles and problems.
Tanya Clement is an Assistant Professor in the School of Information at the University of Texas at Austin. She has a PhD in English Literature and Language and an MFA in fiction. Her primary area of research is scholarly information infrastructure. She has published widely on digital humanities and digital literacies as well as scholarly editing, modernist literature, and sound studies. Her current research projects include High Performance Sound Technologies in Access and Scholarship (HiPSTAS).
Featured image: “Day 21 – Waveform” by Flickr user evil_mel, CC BY-NC 2.0
REWIND! . . .If you liked this post, you may also dig:
Revising the Future of Music Technology — Aaron Trammell
SO! Amplifies: Carleton Gholz and the Detroit Sound Conservancy — Carleton Gholz
Sounds of Science: The Mystique of Sonification — Margaret Anne Schedel
6 responses to “Machinic Ballads: Alan Lomax’s Global Jukebox and the Categorization of Sound Culture”
Trackbacks / Pingbacks
- July 28, 2017 -
- August 22, 2016 -
- June 8, 2015 -
Reblogged this on Electronic Music Production.
It was inevitable that Alan Lomax would follow lines of generalization as the subject matter was the music of the world and he was not of the micro school of doing anything . Much like other cultures that have been blessed and cursed with new techknowledgy , he ran with his 500 lbs. of equipment like it was a wonderous toy and then weapon when he realised the power it gave him over others hearts and minds . And of course had to ‘sell’ himself in this most competitive of worlds , that of of recorded sound . He seems to have had one foot in the future , seeing relationships that other ethnomusicologists could not because of their own cultural blind spots , and one foot dragging the past and seeming trapped in his own ego determinants for failing to comprehend the soul of these sounds as somthing beyond a promising product . Ah, but we are all only human and as vast as his organisational skills and intellect were he was certian to prove human . Tis the fate of catagorizers to miss the point of the story being told and think it is there for them to analyze and not meant to be simply lived .
Hi Tanya —
Interesting to see application of ARLO to Lomax materials around issues of access, or maybe more accurately around ways of assessing what is in a collection of materials from a bird’s eye algorithmic view.
I have various quibbles with your characterization of Lomax here, but what I’m must curious about is whether you could walk us through the vimeo video more. I can’t quite “read” it, I have to say. What patterns are there other than the fact that there is more instrumental music than spoken word?
Here are some quibbles:
-Are you making too strong a binary between individual and collective? I’m not downplaying the many (and there are many!) problematic dimensions to both John and Alan Lomax’s folklore work and song collecting. But both had more sophisticated understandings of the relationship of the singular, individual musician and some larger collective body of human knowledge and expressive power that all tapped into and drew upon. Or put another way, for the Lomaxes there was Leadbelly—and then there’s the song culture that Leadbelly both helped to shape and from which he drew (and all draw) as if it were a wellspring. There’s ownership and property of culture that the Lomaxes did a poor (sometimes mediocre, and once in a while quite noble) effort to recognize and there’s the quality of culture as something human, collective, arising from the singular but feeding into a larger pool of knowledge and feeling. Of course, that’s a very Romantic position to have, but it isn’t as simple as positioning the individual against the collective. For the Lomaxes those two categories intersected and related in more rich and complicated ways. It seems to me your digital experiments raise awareness about how taxonomies threaten to reduce our understanding of that dynamic relationship between singularity and multiplicity. We need to steer the digital and the way we use it toward deepening our understanding of that interplay, not reducing it.
-Similarly, I wonder if you are too strongly distinguishing between “participants’ songs rather than their stories.” Aren’t their songs precisely the vehicle for their stories? Couldn’t an instrument express their stories as effectively—neigh, even more effectively!—than their voices? Why privilege spoken language over other modes of expression?
-I thought the story with Pandora was that it’s founders actually started out with the coding design of Lomax’s cantometrics system in the Global Jukebox. Which is to say that Pandora is in some sense an extension of the GJ? Do I have this wrong?
-I think Cantometrics is quite a bit different from John Lomax’s ideas about balladry. J. Lomax thought that there was much closer to Herder, Romantic notion of a nationalistic Volk culture; it was fundamentally conservative and about the content of songs. A. Lomax thought that the coding of social science could reveal both the diversity of global song style (style rather than content, a big shift that was shaped by A. Lomax’s studies in paralinguistic in early 60s) and then a computational search for where similarities might lurk among cultures.
Of course, he ran into all the problems you mention: where are the boundaries between culture groups? Can we link culture to material practices (all that old anthropological effort to distinguish between agricultural societies, hunter societies, etc.) and if so how do we make those links: are they direct? homological? stranger than that?
So it’s not that Cantometrics and the Global Jukebox are not problematic in the ways in which they are embedded in now-outmoded anthropological cross-cultural comparative analysis (Lomax did this work just as the field of anthropology was shifting fundamentally away from that kind of approach toward Geertzian thick descriptive and from there toward ever more self-reflective interrogations of the power relations in the ethnographic approach). It’s not that the project isn’t flawed ethically or epistemologically. It’s that there is a quite different politics at work in A. Lomax’s social scientific efforts to grapple with culture and cultural *diversity* than J. Lomax’s older 19th century Romantic mode. There’s the trace of romanticism there, but also newer notions of cultural difference and singularity and diversity present in A. Lomax’s work.
It is true that A. Lomax did think he had come across correlations between song *style* and cultural practices and beliefs, that the kinds of pitches and tonalities of certain groups matched shared group attitudes toward sexuality, social control, gender, power. That you could hear those attitudes in, say, the tightness or tension of a singing style or, with choreometrics, that you could correlate economic practices and social mores to certain movements in the realm of expressive culture such as dancing. At one level it was essentialist and quite absurd; at another level, it was quite productive.
What it makes me wonder about lately is this: despite all its problems, there still is value in trying to figure out where/how/why/when singular, individual modes of expression intersect with collective bodies of cultural knowledge. We should still be looking for the patterns produced from those intersections, just as A. (and even J.) Lomax did in their ambitious but imperfect ways. If we treat computational approaches as non-positivistic efforts to do this wondering, if we refine our interpretations out of our examinations of digitally-produced patterns and do so in multiple ways, at multiple scales, through the ability to shift data around to perceive artifacts in their multiplicity, then computers and digital data might help us better understand, theorize, question, and analyze culture and its relationship to issues of individuality, power, commonality, and difference making.