exploring musical patterns of community on harry smith’s anthology of american folk music.
Comments prepared for the second gathering of the NEH Advanced Research Workshop HIPSTAS, High Performance Sound Technologies for Access and Scholarship, which focuses on using the digital sound analysis tool ARLO, Slides below too.
I have been using ARLO iteratively to search for unperceived connections and contrasts within a small dataset: the famous Anthology of American Folk Music put together by Harry Smith in 1952. Consisting of 84 tracks, divided into three categories—Ballads, Social Music, and Songs—this collage of US “roots” music recordings was a kind of mystical remix of old commercial recordings of hillbilly sounds, race records, and other ethnic musics that existed on the periphery of the emerging twentieth-century American commercial recording industry.
Smith was an avant-garde filmmaker, autodidact, ethnographer, collector of esoteric items, quirky bohemian—also a drug addict and hustler—who claimed to be a modern-day cultural alchemist. Here he is in a photograph from the 1980s taken by poet Allen Ginsberg, who playfully describes him “transforming milk into milk.”
On the Anthology, Smith reorganized the sounds he had collected from commercial 78 rpm recordings of the 1920s and 30s and illegally bootlegged into a long-playing proto-box set release for the famous Folkways label, once owned by radical lefty impresario Moe Asch and now a part of the Smithsonian Institution.
What Smith released into the world was, in essence, a remix before the term existed. He erased categories of race, class, and region in the US, the kinds of classifications that were typically used to organize folk music. In their place, he offered what many in the burgeoning folk music revival of the 1950s and 60s came to imagine as a dreamt-up integrated community of American vernacular life in sound.
This was a visionary musical collage of democratic longing, what the cultural critic Greil Marcus has called “Smithville,” a fictitious American every-town. It was what folk music scholar Robert Cantwell describes as “Smith’s Memory Theater,” an effort to summon in one compressed sonic form the range of American culture and spirit at the grassroots. It became a sacred text, a sort of bible and reference manual, within the folk music revival during the 1950s and thereafter, partly because the Anthology not only preserved traditional sounds, but also rearranged them in such a way as to suggest to folkies new ways of comprehending the world.
In its strange, collaged form, surrounded by quasi-mystical imagery and humorous yet scholarly liner notes, it was not only a compendium of folk song, but also a kind of guide for self-transformation and enhanced living for many folk revivalists. What I want to know is if ARLO can help us discern the patterns that Smith himself claimed to be chasing after with the Anthology: patterns of commonality and community, individuality and difference, all expressed musically and brought into being as much through comparative collage and pattern-making as in the discrete sounds of the Anthology themselves. These patterns include not only melodic and rhythmic similarities and elaborations, but also harmonic tonalities and sonorities that ARLO might be able to discern and re-represent.
Smith himself remarked of the Anthology in a 1968 interview with John Cohen in Sing Out! magazine: “I was just fiddling with those ideas then…Everything is computerized now.” He saw potential in the use of the computer to listen better to musical expression and its cultural—even its political—implications when remixed, re-mediated, and heard anew in reoriented relationships. Now that we are developing sonic analysis tools such as ARLO, can we use them to continue the project that Smith began, which was to see if music rearranged and recontextualized could unleash new understandings of those who made—and also those who continue to listen to—these important vernacular American sounds?
I think the jury is still out on this question. What leaps out at me about ARLO is that it is clearly well-suited for questions of accessing collections, particularly large ones. This is something that computational automation is good at: seeking out patterns based on precise parameters, processing large amounts of data, turning back search queries, pinpointing the needles of material lodged in haystacks of data. But what about analysis, interpretation, the making of new meaning from sources, especially of music?
The challenges here between access and analysis are not completely different at the level of ARLO’s parameters: we are in the same territory of seeking out similarities and differences across sets of sonic materials. But the question of what one wants to do with those parameters becomes more complex when it comes to analysis. Matters not only of what but also of why start to leap to the fore. What counts, as it were, as evidence for particular arguments becomes an issue. And what questions we wish to ask—which questions of ARLO’s parameters and functions might lead to new findings—this becomes the most pressing dilemma of all.
For instance, with the Anthology, we might ask questions about similarities of melody, of pitch. But we might also, thanks to the design of ARLO, ask questions about attack, energy, harmonic timbre. In terms of music’s temporality and rhythmic dimensions, we have to decide what units of measurement are worth… measuring as a discrete entity: a single note? A lyrical line? A whole verse? A particular “lick” on an instrument? One voice singing? Two voices? Instruments in relationship to each other? Silence between notes? They are all potentially crucial aspects of the pattern-making that is the Anthology of American Folk Music. Which permutations of what matter in musical expression, even on these scratchy old records, maybe especially on these scratchy old records. The complexity is endless. There is no “objectivity” here, really, in the shallowest sense of that term. There is only choice, selection, exploration, iteration, comparison, and the mingling of computer-heard sound and human-heard sound, different ears, if you will, intermingled in responses to sonic registers.
Given these challenges, my early experiments using ARLO have been tentative. I’ve been most interested in seeing at first if Smith’s remix of vernacular American folk sounds possesses sonic similarities of pitch, on the one hand, and energy, or intensity, on the other. So I have toyed with the pitch parameter and the energy spectra parameter the most just to see if the same tagged section of a song—say the chorus of the final song on the Anthology, Henry Thomas’s “Fishin’ Blues,” produces similar or different results.
What I have found, as a preliminary result, is that ARLO confirms the ways in which little snippets of similar melody, in terms of interval relationships, are scattered all across the Anthology, in unlikely places, producing a kind of subtle sense of interlinked connection across the collection of songs. This isn’t altogether surprising of course. It is a reminder that the different genres of vernacular American music that Smith chose to include in his remix—string bands from Appalachia and blues from the Mississippi Delta, Cajun fiddle tunes and Midwestern polkas, sweetly fingerpicked songster tales of redemption and justice and harsh, powerful frailing banjo plaints of sin and sorrow—share many fairly narrow musical traits: the melodies in particular are based on variations of the pentatonic scale, a shared musical style of the British isles and West Africa, which were not the only but perhaps the crucial streams that fed into American folk music represented on the Anthology. But it is nice to see ARLO catch melodic similarities where even a well-trained ear might not catch them, not just between Thomas’s “Fishin’ Blues” and his version of “Old Dog Blue” but also between “Fishin’ Blues” and fellow Texan Blind Lemon Jefferson’s “See That My Grave Is Kept Clean,” between Thomas’s song and the similar songster stylings of Mississippi John Hurt’s “Spike Driver Blues,” and even between “Fishin’ Blues” and far more distant seeming songs from the ballad tradition, such as “John Hardy” by the Carter Family, and even between “Fishin’ Blues” and the Cajun fiddle and accordion song, “Le Vieux Soulard Et Sa Femme.”
At one level this is just work of confirmation, not of revelation. Which is to say it provides more evidence that Smith had good ears, that he used more fluid sonic similarities and variations to guide his remixes rather than seemingly static categories of race, ethnicity, region. And it more deeply documents that he created a collection that was richly suggestive rather than rigidly definitive. But the findings also suggest something else: with more research, we might begin to bring into view—more accurately, into hearing—new findings and understandings, new analysis and interpretations. These would involve not only fresh interpretations of Smith’s work, but also revealing dimensions of the contours and qualities of American folk music itself. ARLO as a computational ear proposes that Smith’s human ears only began to perceive the deep interlinkages within American vernacular song. If we are willing to go fishin’ a bit more, even within this small sampling of songs on the Anthology, this micro-dataset, and if we use the bait and tackle, the lures and lines, of our ARLO tuning forks, visualizations, libraries, projects, media files, tag sets, unsupervised tag discoveries, and k-analysis clusters to do so, we may be able to discern far more of this suggestiveness, this interrelatedness within diversity, in the great ocean of American folk traditions. Even more, perhaps, than what Smith, that crazed remix alchemist, himself heard and harvested.