By Gordon Hull
Last time, I offered a quick synopsis of Bernard Dionysius Geoghegan’s excellent new book Code. Here, I’d like to track one specific Foucault reference in it. Geoghegan takes Lévi-Strauss’s Savage Mind as a central text in the ambivalence French theorists came to feel about American communication theories, and he notes that the book “occasioned a broader reassessment of the human sciences marked by a new ascent of ‘coding’ as a key concept poised to dislocate and perhaps dissolve, existing scientific hierarchies” (152). He adds:
“Learning to code – that is, to cast cultural objects in terms of codes, relays, patterns, and systems – did more than reframe existing knowledge in cybernetic jargon. It also reflected a growing cynicism toward existing cultural and scientific nodes. From the 1960s onward, the semiotic task of deciphering obscure ‘codes’ in culture, politics, and science overtook the structuralist project. This crypto-structuralism shifted emphasis from the neutral connotations of ‘communication’ to antagonistic notions of code …. If these terms furthered the technocratic project of US foundations, they also set in motion a radical critique of scientific neutrality. Beneath the neutral science, something ‘savage’ lurked.” (152-3).
Geoghegan cites Lacan, Barthes and the Tel Quel group (on which see Danielle Marx-Scouras’s excellent study). He also quietly footnotes Foucault’s “Message ou bruit [message or noise]” “for a critical discussion of these same terms by Foucault” (215n81).
Foucault’s essay is short; it originally appeared in 1966 and there is relatively little on it. Stuart Elden provides some context, noting that it appeared in a medical journal in the context of discussion of Birth of the Clinic. Read with the additional context provided by Geoghegan, it provides some insight into what Foucault thought he was doing in his works of the mid 1960s. Birth of the Clinic appeared in 1963, and as Elden notes, received almost no attention when it was released (Archaeology of Foucault, 6-10). It was revised and republished in 1972, and Elden notes that “most of the structural language is removed” even though “the findings in the analysis enabled [by it] remain intact” (Archaeology, 8-9). The 1966 piece isn’t about structuralism, but in the context of Savage Mind (1962) it seems instructive, because in it, Foucault is undermining the idea that structures could be read out of nature (or that the structures one reads out could be universal).
Foucault begins by discussing how medicine has traditionally been situated among other forms of knowledge according to “linear schemas” (D&E #44 I, 585-8 in the two volume 2001 edition I have). For example, the soul is located above the body and tissue below the level of the organism. Against this, Foucault proposes that the problems in medicine “seem isomorphic to those which one can encounter elsewhere, if only in the disciplines which are occupied either with language, or those which function as a language.” The topics are different, he suggests, but are perhaps structurally analogous when understood as theory and practice.
He proposes that “one says, one repeats, since Balint, that disease sends one or several ‘messages’ that the doctor listens to and interprets.” [side note: Balint is a reference – the only one in D&E, according to the index; I also can’t find a reference either in History of Madness or Birth of the Clinic – to Michael Balint, an émigré Hungarian psychoanalyst who lived in London (no, I don’t know any more than this about him, but would welcome learning more)]. What strikes me as interesting though is the immediate next sentence: “this permits many blessings of humanism on the doubtful theme of ‘disease-doctor couple’ [Cela permet bien des humanisms bénisseurs sur le thème douteux du ‘couple médecin-malade’]” Even here, in other words, Foucault has as his target humanism: his general framing in the mid-1960s is to oppose “humanism” and the “science of man” in all its forms; “Message ou bruit” appeared the same year as Order of Things. But it also suggests that the language of communication is also a form of humanism, structuralist trappings to the contrary notwithstanding (more about this some other time; there is evidence that at least Foucault and Pierre Macherey think that the wrong kind of structuralism is tantamount to humanism).
Foucault proceeds. “In order to have a ‘message’ it is necessary that “first there is noise (in the case of medicine, this primordial noise is the ‘non-silence of the organs’ [qu’il y ait d’abord du bruit (dan le cas de la médicine, ce bruit primordial, c’est le ‘non-silence des organes’].”
This noise then has to involve discrete elements which can be associated with other elements that then constitute their meaning. Foucault then says that "disease doesn’t send a ‘message,’ since a message depends on a ‘code’ established according to prior [précedéntes] rules. There is no code in nature, however deformed [dénaturée] it is. Disease is content to make noise, and it is already beautiful. Everything else, it is medicine that does that; in fact it does even more than it lets itself believe.” No matter how “unnatural” or “denatured’ one’s understanding of nature is, there is no code inherent to it. In other words, the message that medicine decodes is made intelligible by medicine itself, because it is medicine that has decided what kinds of noise indicate what kind of message. There are two steps to this process: the “constitution of a code” and “listening to the message.”
In the context of the first, Foucault points out that the code “never ceases changing.” The least common but most interesting change is when we “define the elements of signal [message], where we only heard noise, [then] medicine adds new domains.” Example? “Freud made verbal statements of illnesses, considered until then as noise, something that ought to be treated as a message.” That is, they became symptoms, or perhaps (in a somewhat more current idiom) information and not just data. This decision creates a form of medicine. Freud’s innovation was important and wide-reaching: “from now on … diverse forms of medicine have heard, as messages, the verbalizations of diseases.”
Foucault also emphasizes the difficulty in stabilizing this apparatus and locking down a “unique code.” “Since disease has nothing to say, there is no reason that only one code arrives to ‘inform [informer]’ all this noise.” Again, there is contemporary relevance.
As Colin Koopman emphasizes, there is a moment of formatting that is part of the translation of data into information. Koopman returns to the early 20th century, and notes that “The success of birth registration in America depended not only on its official trio of justifications, but it also came to depend on a trio of information technologies.” Thus:
“Two of these technologies were already under design by 1903: the standard birth certificate blank and a protocol for registration administration. Seen through the lens of contemporary technologies, we can view these as a file standard and a filing protocol. Together they enact the work of formatting in the sense I give that term here-a technique that serves to fasten that which passes through it. Standardized blank certificates and registration bureaucracies promised to fasten the babies that would be formatted according to their terms.” (43, I am leaving out his discussion of information auditing here).
That is, natural events have to be made legible to information systems by devising ways to understand them, and then these mechanisms serve to filter “signal” (location of birth) from “noise” (vaginal or caesarian birth, perhaps) in establishing the identity of the baby. As Koopman notes at the outset of the chapter, this can create its own perversities; the birth certificate for a California baby born in the car on the way to the hospital listed “location” as “automobile,” nevermind that an automobile is a location in a completely different sense of the word than “Los Angeles.” The birth certificate also had to document proof that the baby was born alive. In the absence of a doctor to certify it, the agency accepted bringing the baby to the registration appointment. This basic apparatus of formatting is required prior to the interpretation of individual data; in Freud’s case, the interpretation of dreams requires the prior decision both that dreams are signal amenable to interpretation and some sort of logic governing that interpretation (Oedipal complex, etc).
Current bureaucratic and automated systems are all about producing “signal” and differentiating that from “noise.” But as Foucault’s point about how nature isn’t communicating some sort of code underscores, there’s at least two kinds of risks associated with this constitution of code. One – famously documented at the Spurious Correlations website – is that one’s signal is no useful signal at all. For example, from 2000 to 2009, per capita cheese consumption correlates with the number of people who died by becoming tangled in their bedsheets at a 94% rate. These problems give the lie to the claim that if you just had enough data, the patterns and correlations would somehow legitimate themselves – the “n=everything” approach, as one famous paean to big data put it.
Another risk is that the process of declaring data to be signal is somehow unfair. Affect recognition technology, for example, is based on bad pseudoscience, and so it is not only epistemically worthless, but it also is unfair to the people it assesses. I complain about this and similar cases in machine learning as examples of epistemic injustice in a just-published paper (preprint here), but there’s plenty to choose from in this general area. Ari Waldman, for example, shows how gender binarism is enforced, sometimes inadvertently, at nearly all levels of bureaucratic process.
Foucault’s statement thus lines up with what Geoghegan calls “crypto-structuralism:” it sounds sort of structuralist, but it pulls the rug out from under it by undermining the universality on which its epistemic status depends. By way of closing, let me note that Foucault’s invocation of signal and noise lines up with another underappreciated aspect of this work in the 1960s, his reading of more pragmatic accounts of language in theorists like JL Austin and Wittgenstein. As Elden documents, we know that by the fall of 1966, as he arrived in Tunisia, Foucault was reading a range of work including Austin’s Philosophical Papers, Wittgenstein’s Philosophical Investigations and various works by Dewey (Archaeology of Foucault, 158-9).
Next time I’ll say something about the political implications of signal and noise and try to connect them to some of Foucault’s remarks in the 1970s.
Recent Comments