https://d2ng68cf36kxsk.cloudfront.net/store/public/post/908/post_images/Slide325.JPG">

“What characterizes a symphony when experienced in immediate listening … is a particular intensity and concentration,” he says. “The ‘surrounding’ function of music also disappears … partly because of the monoaural conditions of radio … The sound is no longer ‘larger’ than the individual.” These are some of the ways in which “symphonies brought to the overburdened hypothetical farmer in the Middle West are … deteriorated by radio transmission,” – and phonograph.³⁰⁷ This was before the application of place-dimensionality to music with stereophonic recordings, and its extension with the precedence effect.^§AS1.4.2

Then digitized music – the first commercial digital recording was made by Nippon Columbia in 1971, a jazz recording, “Something” by Steve Marcus. Adorno would have been doubly displeased. The Compact Disc, a large-capacity laser disc for digital recordings, was brought to market in 1982.³⁰⁸ The first commercial release was ABBA’s album, The Visitors. Then, with the Internet and the need to compress digital sound files without loss of quality, came the MP3 format.

In 1988 Hiroshi Yasuda of the Visual Media Lab at Nippon Telegraph and Telephone in Tokyo and Leonardo Chiariglione of Telecom Italia’s research labs, commenced an initiative, the Motion Picture Expert Group (MPEG) under the auspices of the International Standardization Organization. Its brief: to develop an international standard for audio and video encoding. The original group of twenty-five people consisted of representatives of the recording, telecommunications, and media industries. Again, we find corporate interests rising above the competitive fray of the market to create another of the universal frameworks of our contemporary meaning.

MPEG-1 was released in 1993 and MPEG-2 in 1997. Layer I is basic data mapping, Layer II bit allocation, and Layer III audio frequency resolution and file compression, or “MP3” for short. Since its beginnings with the phonograph, sound recording worked by registering the amplitude of sound vibration over time. Digital recording required recording in numbers, encoding, storing, then decoding and converting the numbers back into sound. This is how sound-making became a domain of calculation, on principles envisaged by Ada Lovelace^§MS1.3.2a and Alan Turing.^§MS1.3.2b The transposition runs like this: sound > recording in calculable number (text) > recalculation for the re-rendering as sound.

MP3 recordings sample sound at a rate of 44,100 records per second (44.1 kHz). The human ear can discriminate sound up to about 20 kHz, so the sampling rate is plenty for two-channel sound. The data at this rate of sampling is big – it was too big for home computing devices in the 1980s, hence the invention of the Compact Disc laser reader. Even with the CD’s larger storage capacity, part of the trick was to compress files by removing redundant information. Combining the innovation of the disc compression software, a CD could hold about seventy minutes of music. With the rise of larger storage capacities on hard drives or solid state memory, and the data pipeline of the internet rapidly expanding, the CD was a doomed technology. Now the MP3 format and its slightly modified descendants rule supreme.³⁰⁹

If the phonograph and radio transformed the cultural practices of music, the medium of digitization adds to these some new social functions of meaning. A logic of mass production and economies of scale had been built into analogue record manufacture and wireless radio. Setting up a record studio for analogue recordings and pressing records only made sense on the economies of scale of mass production. These technologies only worked on an industrial scale. So, the business model for music production became long runs for mass markets. Radio bandwidths were limited and transmission facilities were large and expensive, so in any one geographic area there could be no more than a handful of options. Popular music was reduced to best-selling “hits,” and “classical” music to a limited repertoire of old European masters. This massification of culture and the creation of passive audiences was what Adorno and Horkheimer railed against in their critical theory of the “culture industries.”³¹⁰

But with music digitization and streaming, by the end of the 2010s, Apple Music offered about 45 million songs, and Spotify 35 million – including pieces by Cazden,^§AS1.4.2a Adorno,^§AS1.4.2b and Avraamov^§AS1.4.2c that could never have made it into the top forty, that would be uneconomical to press as records, and that local stores would never stock. The result: no two music libraries, no two playlists, are quite the same. This affords a radical shift in the social designs of interpretation in late modernity, from homogenizing-massifying of interest, to differentiating-dividing. It’s not that this changes the basic structures of inequality, rather that the dynamics of the maintenance of inequality become quite different.^§AS2.2.2

At the same time, and perhaps not so paradoxically, there has been a huge concentration of control in the means of production and distribution of our musical meanings. There are just two comprehensive streaming services, or perhaps four if you add Amazon and YouTube to Apple Music and Spotify. Not only do these providers archive music on a massive scale; they also record the minutiae of every user’s listening habits in ways not previously possible. All-knowing megacorporations track our musical proclivities in meticulous detail. We are complicit when their “artificial intelligence” makes a new music suggestion that is uncannily prescient, a piece you should have known about but didn’t, or that somehow you have forgotten.

In reality, the intelligence is not so clever as such flashes of recognition might trick us into believing. The smarts are not in the 44,100 digital samples per second, nor their calculation for the purposes of encoding and decoding. Only the simplest of mechanical sense can be made of the math. Most of its genius is simply to name and record a myriad of sequential moments of sound amplitude, an avalanche of detail that is useful for manufacture but mindless for humans in its detail.

The innovation, rather, is to record the range of self-declared human affinities named as the product numbers^§MS1.1.1c and track these numbers in a database – “recommender systems,” they are called.³¹¹ This affords what Norman Cazden called “realism in music,”^§AS1.4.2a a social realism when someone whose finely calibrated tastes are calculated to be close to yours. Of the tens of millions of options, another user may have a certain two pieces in their playlist, and you have just one, so the machine offers you their second as a suggestion. Recommendation software works simply by recording data that it was until now impractical to record, in this case, the minutiae of every user’s listening. The machine listens to its listeners. The meaning is social. And these megacorporations now know our meanings.

---

³⁰⁵ Michael Chanan, 1995, Repeated Takes: A Short History of Recording and Its Effects on Music, London: Verso, pp. 1–3.

³⁰⁶ Marc Raboy, 2016, Marconi: The Man Who Networked the World, Oxford: Oxford University Press.

³⁰⁷ Theodor W. Adorno, 2002, Essays on Music, translated by S.H. Gillespie, Berkeley: University of California Press, pp. 245, 257, 253.

³⁰⁸ Thomas Fine, 2008, “The Dawn of Commercial Digital Recording,” ARSC Journal 39(1):1–17.

³⁰⁹ Hans Georg Musmann, 2006, “Genesis of the Mp3 Audio Coding Standard,” IEEE Transactions on Consumer Electronics 52(3):1043–49; Marina Bosi and Richard E. Goldberg, 2003, Introduction to Digital Audio Coding and Standards, New York: Springer Science.

³¹⁰ Theodor W. Adorno and Max Horkheimer, 1944 [2002], Dialectic of Enlightenment, translated by E. Jephcott, Stanford, CA: Stanford University Press, pp. 94–136.

³¹¹ Joseph A. Konstan and John Riedl, 2012, “Recommender Systems: From Algorithms to User Experience,” User Modeling and User-Adapted Interaction 22(1–2):101–23. doi: 10.1007/s11257-011-9112-x.

---

Kalantzis, Mary, and Bill Cope. 2020. Adding Sense: Context and Interest in a Grammar of Multimodal Meaning. Cambridge UK: Cambridge University Press. pp. 120–123. https://doi.org/10.1017/9781108862059

Hiroshi Yasuda and Leonardo Chiariglione’s MP3 §AS1.4.2d