A Slate article from two and bit years ago drew my attention to some research published in Nature1 (also two and a bit years ago) of which I should have already been aware. The Slate article claimed that the research found empirical evidence indicating that, much like your imaginary parents have always said, popular music really is getting louder and stupider. Slate then goes on to illustrate why this straw man claim (one the authors of the study did not actually make) cannot be adequately born out by empirical research. And I am left, sitting here, writing this, about reading that, contorted into an uncomfortable cringe about the whole discussion. I will explain why that is.
The research takes as its data set 464,441 distinct recordings from the Western popular musical cannon from between 1955–2010. It ran these through some crazily sophisticated modes of digital analysis to examine the music by pitch (mostly looking to harmony and changes therein), timbre (sound color), and loudness. What this showed was that over the set period, the complexity and diversity of harmonic constructions and changes had reduced, as had the range of timbral diversity. It also showed that loudness had increased (which simultaneously decreased the range and subtlety of dynamic variation). It required some pretty complex math to draw out these tendencies, on which I am not qualified to comment. To the extent that I understand this sort of work, I am led to largely agree with the findings overall. Especially as the research paper makes clear that these trends are somewhat dependent on the analytical models that are employed, and are not indisputable. My points of contention on this stuff — and really the research field in general — constitute a side note to the broader argument being made here, but nonetheless are relevant to understanding the potential pitfalls of empirical research on music.2
Two and a bit years ago Music Information Retrieval and sound file analysis was the research I wanted to do. I had pretensions of how I would add to this data a breakdown of sales income or stream counts of particularly bland songs in order to prove something about market influence and audience conformity that is actually already blindingly obvious to a lot of people. (That anything resembling the old form record industry has survived for long enough to now be feeding on the carbuncular vessel of Spotify is a testament to how much our listening and our basic relationship to music is structured around these corporate edifices as gatekeepers.) So I’m glad that this work has been done, but at the same time I feel induced to cringe. I cringe partly for the reasons elucidated in the mini-essay I’ve called a footnote, but also because this study represents a great deal of work that concludes with what many have been arguing as their starting point for years: that industrially producing culture is a very strange thing to do.
Beyond the most obvious technical reasons (the reduction of the recording studio to the presets of Pro-Tools, producing tracks as file types that have already cut out any frequency that “does not transmit the information” and mastering software that allows for the sort of compression that produces the loudest, flattest results), we seem to just be left in the wilderness of market forces without a wider consideration of what that means. What does it mean to have your ears awash with products? What does it mean when these particular songs — often heightened exaltations of extraordinary emotion — are the products? Can I express anything that is not already a product? Who am I, if everyone who hears this song is also singing along through the tears?
Most of what gets written about music comes from people listening to it. But that is not how most of the music that’s around is experienced. We seem to be stuck with this notion that music is consumed by Adorno’s expert listener, when we are in fact way more passive than that. In her recent book, Anahid Kassabian argues that we are now in a condition of ubiquitous listening (borrowing from the term ubiquitous computing that describes the damage many of us are currently inflicting on the upper portions of our spines as we drag to refresh Facebook, again). Thus when the Nature researchers describe their dataset as 1200 days of continuous music, it evokes an image of someone sitting before a pair of speakers for years. But we’re not Alex before the screen, eyelids forced open and regular drops provided, being conditioned towards the bland. Pop music is instead the soup we swim in, the fabric of your genes brushing on your knees. So looking at music, to find out how music works, can end up telling you very little about how music works.3
For example: sometime someone you know, perhaps even trust or love, will play you a song and you will have the uncanny feeling that you have heard it before, but you can’t remember where. Not just that you heard it before, but that hearing it now stirs up feelings and associations that seem fully formed yet also unanchored. It could well be because you have heard a lot of things like this, but it’s also likely, if you trace it back, that you have heard this exact song before. You heard it in a context where noticing that you were listening to a pleasing piece of pop wasn’t on the agenda. Perhaps this was at the mall, in a café, or given that the feelings this piece stirred up are of long-term and deep anxiety, it was probably featured on an oddly unskippable advert for a bank loan, on that YouTube video you actually wanted see so you waited through it, the whole time wondering if you could actually afford to take on another loan, different from the one being advertised. This song has claimed a place in your heart.
The so-called “loudness wars” refer to the increasing overall volume of audio files, facilitated by contemporary mastering technology, at the expense of dynamic range. It has been something like an arms race, as tracks needed to increase in volume in order to stand out against the competition. I think the same could be said for the reduction of harmonic complexity. The hook has to stand out too and not be lost in ambiguous tonality. It is not that our tastes have become simpler, but that most of the music we hear (indeed most of contemporary Western popular music) is not for you to listen to but actually just needs to be present enough to organize your experience into glorious little sound-tracked chunks that could potentially leave you teetering on the edge of hysteria.
The authors of the Nature paper state explicitly that by treating music like a linguistic source, they are decoding “musical discourse”: music as text. But this has already been shown to be an inadequate way to approach music and especially inadequate in the case of popular music. Popular music (as indeed anything else with an audience greater than zero) is defined, rightly or wrongly, through its interactions in the world, not through its essential objectness. The musical discourse is tied up in these interactions. When the authors conclude that, based on their research, loud simple chords within a limited timbre spectrum may signify musical newness to the pop music consumer, they are leaping a chasm between what music sound files contain (sound information, meta-data dates) and what music does (providing the quantized lattice fabric on which we unconsciously structure experience). Far from newness, these same signifiers could as easily be about stability, group inclusion, and fear of the other. They might be discussed in terms of “hot new tracks” but to think that the new is what listeners are after is a rather shallow way of looking at pop. Kasabian argues that ubiquitous cultural productions can not be understood as individuals assessing each “text” but rather as an example of “distributed subjectivity”. Put simply, she describes how when we look at music, the whole notion of a rational individual actor becomes a silly rubric for understanding human behavior. Things we consider to be intrinsic parts of ourselves (taste, for example) actually exist between people through forms of mediation. So when I look at this kind of work, especially as it is recounted in the media, even while recognizing its validity, I cringe.
I cringe because it seems that there is so much work yet to do in order to shift this discussion from where it currently and persistently rests: that what we hear is what people really want. That because we are the rational individual actors of the enlightenment subject and right-wing economics, all critique of popular culture is elitist. However, what we appear to want did not simply and only come from inside ourselves with the marketplace happy to oblige. The distinct, separated “rational” individual is a political fabrication. What we are comforted, inspired, and entertained by is constructed from the foundation of mammalian physical systems but the rest of the structure has been substantially influenced by nearly a century of cultural industrial power and the society that this ferments.
1. Authored by Joan Serrá, Álvaro Corral, Marián Boguñá, Martín Haro & Josep Ll. Arcos, who, due to their numerousness, shall hereafter be referred to as the authors.
2. Put simply, a lot of this research takes place without much consideration for music as it relates to politics and without a particularly sophisticated ontology of music, especially in the analysis of popular music. This is not universally true, but it is done often enough to make me wince. This particular piece of research ran up against the old problem of providing a static picture of musical facts while describing music as a temporal phenomenon that relies on the experience of internal difference and change over time. Standard time measurements (minutes, hours) alone are irrelevant in musical experience, so these researchers used what they ambiguously term “the beat” (ambiguity easily removed by using the term pulse or meter). But also, so as to avoid the ridiculous simplification of one beat merely being compared to those immediately preceding and proceeding, some sophisticated interpolation takes place that gives a fuller picture of the changes over time. In this study this interpolation not only restricted the songs as discrete bodies, but also to the works of surrounding years. That’s all interesting and cool, but the earlier ambiguity belies another problem. Why was the use of beat and beats not considered as an area for analysis? You could suggest that this be examined by somehow tying together the findings of timbre and dynamics, but that would give a lousy, incomplete fudge of a picture, and anyway, it hasn’t been done. This has ontological implications, as the exclusion is tied to Western classical musical understandings of what hierarchical elements constitutes the essence of music, which is itself pretty political. It also has more overt political implications, in that the use of and focus on beats in Western popular music is tied to discussions of class and race. Its exclusion here is a glaring inadequacy, as the way beat infuses with the other examined elements could paint a wildly different picture of the apparent reduction in complexity. These problems don’t mean that the research is wrong, just grand in title (‘Measuring the Evolution of Contemporary Western Popular Music’) and sample, but myopic in understanding.
3. I would have had far less of a problem with this paper had it simply been about showing how to discover trends. Instead it kept trying to relate it to the real world, which is effort I appreciate but, unlike complex database signal analysis, is harder than it looks.