In Defense of Noise

Email a Friend

Last week’s episode of OTM featured an interview with Nate Silver discussing some of the central themes and ideas in his new book,The Signal and the Noise: Why So Many Predictions Fail but Some Don't. Silver suggests in the book’s introduction that the problem of discerning signal from noise is one of particular importance in an age of “Big Data,” an era of increasingly powerful publishing and communications technologies. “[If] the quantity of information is increasing by 2.5 quintillion bytes per day,” he writes, “the amount of useful information almost certainly isn’t. Most of it is just noise, and the noise is increasing faster than the signal.”

Silver is careful to point out that he is using a metaphorical distinction between signal and noise somewhat loosely to describe the hunt for needles of relevance in growing haystacks of information. But the problem is more than just a matter of managing and processing huge quantities of data. There’s an important aesthetic dimension as well. For the most part, after all, we tend to encounter noises as things that we can feel and sense concretely, not just as abstractions. And I’d like to suggest that the noises we hear at those more personal scales also have some very important things to tell us if we listen carefully.

Noise, in the sense that I’m using the term here, can be a slippery, slightly paradoxical object. It’s the part of the signal that isn’t supposed to be part of the signal. It’s the fuzz and hiss that seem to simultaneously add to and detract from signals. Noises interfere with signals, but they also provide subtle textures and meaningful contextual cues. The scratches on a vinyl record are just distortions at one level, but at another they’re also audible evidence of physical album’s age and wear. Noises of this sort ground signals in specific times, places and media.

This ambiguous relationship among media, signal and noise is what makes the opening scene from “Poltergeist” so captivating and disturbing.

The noise and static on the screen after the station’s signoff don’t represent a total absence of signal: they’re just the sum of all transmissions minus the earthly television broadcasts. Since the static isn’t from any human source, that leaves only the natural… and supernatural ones.

More importantly, though, what counts as noise changes dramatically across time and social contexts. A few decades ago, media producers and consumers generally tried to reduce things like static, surface noise, film grain, and similar sorts of artifacts and byproducts. But today those noisy physical traces of analog media are themselves signals, and valuable ones at that. Instagram, a service that was acquired by Facebook in April for about a billion dollars, is specifically designed to add all the messy textures and “imperfections” of past analog photo technologies — oversaturated colors, ragged frames, graininess, uneven exposure — to your digital images. The aim is to produce a photo that resembles one that’s been developed in chemicals, printed on paper and touched by human hands, rather than a digital file that was processed by software and transmitted bit-for-bit through a computer network. And Instagram is hardly an exception. Today almost every video and audio editing program available includes features that allow you to add the hiss of a vinyl record or the traces of scratches on film to your digital recordings. Nostalgia for noise — noisestalgia? — is big business.

(a catalog of different instagram filters)

So far, though, I’ve described fairly trivial examples. There are often much more profound implications to the matter of where, how and by whom the lines are drawn between signal and noise. Consider, for instance, the case of “alternating sounds” in the late 19th century. American and European researchers had observed that in some Native American languages, it appeared that the same word could be pronounced using different sounds that fluctuated from moment to moment. As David Garrison Brinton described the problem, “The same person pronounces the same word differently; and when his attention is called to it, will insist that it is the same.” The thoroughly racist explanation proposed by Brinton and others was that Native Americans’ “primitive speech” simply had not evolved beyond “vague and fluctuating” phonetic elements to the point of achieving stable sounds.

But, as anthropologist Franz Boas showed in 1889, these “alternating sounds” never actually existed. In a short but devastating critique, Boas explained that languages aren’t merely a set of sounds, but a set of meaningful distinctions and perceptual boundaries between one sound and another. Being able to communicate in any given language means being attuned to the contrasts that matter to other speakers of that language. Confronted with a different set of contrasts, a different set of boundaries between signal and noise, non-Native American researchers produced fluctuating transcripts. As Boas succinctly put it, “I maintain that there is no such phenomenon as synthetic or alternating sounds, and that their occurrence is in no way a sign of primitiveness of the speech in which they are said to occur; that alternating sounds are in reality alternating apperceptions of one and the same sound.”

You may want to keep the power of distinctions between signal and noise in mind as we wade through the last weeks of the presidential race. Political campaigns and media outlets are deeply invested in contests to determine what feels like static, interference and distortion to different voters and different audiences. Have a look, for instance, at how’s Greg Sargent concluded his recap of the implications of last Wednesday’s Presidential debate for the newspaper’s liberal “Plum Line” blog: 

Romney took command. Obama didn’t fight. Beyond the noise of the policy back and forth, voters sense this kind of thing on a basic level. More of this from Obama will demoralize supporters and feed a bad developing press narrative about Romney as the hungry challenger on offense, gaining momentum daily. It has to change.

Take a moment to consider what it means to be able to refer to the noise of the policy back and forth. This sort of figure-ground shift doesn’t happen by itself; it takes effort and action. Sargent isn’t wrong to attribute great political significance to impressions of authority and command. And ironically, the “bad developing press narrative” that Sargent complains about is developed precisely through the sorts of decisions Sargent himself makes in the same paragraph about how and where to split the candidates’ actions into signal and noise.