One of the central problems in philosophical methodology generally is determining the set of texts which are legitimate to use in the pursuit of research. Scholars from a host of different traditions and approaches have radically disagreed about which texts ought to be included in respectable research on any given topic. This equally applies to research on Foucault's philosophy. This disagreement in Foucault scholarship3 is framed by the pre-theoretical, and often tacit, assumption by Foucault scholars that there is a coherent demarcation between Foucault's published and unpublished work. However, I will argue that such a demarcation is internally inconsistent and not historically defensible. That is, Foucault's publication history is a very messy one and when we pay close attention to that history, the dichotomy between published and unpublished turns out to be merely an artifact of scholarly convention and tacit practice of textual scholarship. Imposing such a strict demarcation cannot be historically supported.

Scholars make a demarcation between Foucault's published and unpublished work based upon the pre-theoretical background assumptions presupposed by their scholarly projects. Some texts show up as legitimate while others do not. This kind of approach presupposes that within Foucault's writings there exists a system of sorts which could be misrepresented if we do not, as best we can, distinguish between the published and unpublished work.4

In Foucault scholarship there is a central argument by which scholars situate their methodology. The argument goes as follows:

The first claim is simply assumed as a tacit practice of textual scholarship in general. It can be seen to be assumed and practically implemented, though perhaps tacitly, by almost all Foucault scholars and editors who weigh in on this question. That is, to weigh in on the question of the importance of the published vs. the unpublished work necessitates the assumed existence of a demarcation between the two. The second claim (P2), and its radicalization (P2a), are held, sometimes tacitly, by most editors and many Foucault scholars. Scholars and editors often disagree about claim two and the conclusion but leave the first claim intact. However, I will argue that even though scholars seem to agree about the first claim, their tacit and often unacknowledged demarcation criteria are different. I demonstrate four of these common demarcation criteria (publication, publication*, authorization, and publicness). This means that within scholarly discourse there is not one set of published texts but many unacknowledged sets.

Instead of searching for abstract and universally applicable demarcation criteria, I argue that scholars should instead become more explicit about how their projects presuppose certain pre-theoretical values that allow particular texts to be prioritized. It is at this time, when all of Foucault's notes are in the process of being transcribed by artificial intelligence, that we ought to become clear about these long overdue methodological problems.5 That is, we must first answer the question how do we read Foucault and for what purpose?6

The case of Foucault's work is a complex one because of a tension between Foucault's own methodological insights and his final wishes. On the one hand, Foucault is very clear that when we constitute an author's œuvre, the complete works, the decision about which texts to include is always a matter of interpretation. The texts that make up an author's œuvre are neither an immediate nor certain unity but, rather, an interpreted construction. An œuvre is always united by a particular function that makes use of the name of an author to achieve that unity. In a work published in 1969 Foucault states, "this 'author function' […] results from a complex operation whose purpose is to construct the rational entity we call an author."7 The author function then plays a central role in determining which texts are included or excluded as proper to an œuvre. Foucault writes in the first edition of The Archaeology of Knowledge,

According to Foucault, which texts are included in an œuvre is always a matter of interpretation. Foucault is very explicit that these functions are complex. He writes,

According to Foucault's own writings, the establishment of an œuvre is based upon certain kinds of decisions that require interpretation and evaluation. Not only that, but these choices and demarcation criteria are, according to Foucault, difficult to explicate or even formulate.

However, in Foucault scholarship this comes into conflict with the history of Foucault's own life. Foucault was very careful about his public image as an author and exercised a quite stunning and obsessive degree of control over his writings. This was particularly the case when it came to how his work would be treated after his death. Foucault is said to have told his friends, "Don't pull the Max Brod Kafka-trick on me."10 Kafka wrote what were essentially two short wills in which he asked Brod to burn his diaries, manuscripts, letters, drawings and, interestingly, journal publications.11 Max Brod, against these wishes, went on to edit and publish Kafka's literary estate including his manuscripts, diaries, and letters. While it is a matter of debate in Kafka scholarship whether or not Kafka meant what he wrote in those wills,12 it is far clearer in the case of Foucault. It is also clear that Foucault participated in playing a similar 'trick' on Friedrich Nietzsche.13

In 1977 Foucault told a friend, "When I die, I will leave no manuscripts" and in a letter found in his apartment, written eighteen months before his death, Foucault wrote, "I leave my apartment and all it contains to Daniel Defert. No posthumous publications."14 Further, reports suggest that Foucault asked Hervé Guibert to destroy some of his unfinished work.15 While Kafka's intentions may be disputed by scholars, Foucault's intentions were clear because he himself carried out some of this work. Not only did Foucault leave the letter detailing no posthumous publications but actively tried to make this impossible. Shortly before his death Foucault began destroying hundreds of pages of notebooks, letters, and manuscripts including a manuscript about the painter Manet.16 As Foucault's friend Gilbert put it, Foucault left behind, "only polished bones around a black diamond, gleaming and impenetrable, securely closed in on its secretes."17

The fact that Foucault suggested that an author's œuvre is a matter of discourse and interpretation with no final answer is in tension with the fact that Foucault tried, and eventually failed, to exercise tight control over his estate. This tension is made very clear by David Macy who writes,

If one takes Foucault's last wishes seriously, then none of his unpublished works ought to be used. If one follows his writings on method and his approach to Nietzsche scholarship, all of Foucault's work is fair game.19

Foucault argues that the use of the term œuvre changes when discussing different authors.20 What Foucault says about the œuvre of different authors, can be made even more particular when looking at his own work. Although different scholars and editors talk about the published work of Foucault, the published œuvre, they have in mind very different sets of textual objects.

The first key to understanding why scholars differ so radically is to see it as both a legal disagreement and as a methodological disagreement. As Clare O'Farell argues,

These battles have been a matter of legality for Foucault's executors. There have also been parallel methodological battles among editors and scholars. This division between those who use or exclude the unpublished works in Foucault scholarship was aptly summarized by Brad E. Stone. In his 2004 review of Foucault's lectures in Foucault Studies, Stone writes,

One scholar that holds such a view is Mark G. E. Kelly. Kelly suggests that we ought to prioritize the published works that Foucault authorized over his unauthorized unpublished works generally. He writes,

Despite Kelly's strong assertion, he is quick to qualify this "canon." He suggests that some material which is not published should not be strictly considered unpublished because it is canonical. That is, although it is unpublished, it should be given privileged status analogous to the published work. He writes,

Those who hold that Foucault's unpublished works should not be used often appeal to moralistic principles of disloyalty or embarrassment to justify their point. For example, David Macey writes,

More often than not, these kinds of moralistic ethical objections do not lead to methodological discussions. Scholars simply sidestep these objections altogether. For example, Ben Golder writes, "Putting aside for a moment certain ethical considerations pertaining to the use of this serendipitous Nachlass, the lecture courses evidently constitute an invaluable resource for Foucault scholars."26 Others simply embrace the political use of Foucault's work and refuse to be embarrassed by it. Bernard E. Hartcourt, for example, writes,

Most scholars that take a less radical track tend to begin by defending themselves from a moral objection over the use of the unpublished work. For example, Lynne Huffer writes, "I know that my insistence on citing an unpublished interview Foucault disliked will be seen by some as disloyal or, at the very least, in bad taste. But I see it differently."28 Huffer then goes on to defend the prioritization of unpublished texts in her work. This can also be seen in the work of James Miller when he justifies including Foucault's interviews. Miller argues that we ought to prioritize some unpublished material because, "A number of these interviews were carefully edited, corrected—and in a few cases, substantially rewritten—by Foucault himself."29 However, Miller goes on to cite Deleuze saying that even those interviews that were not edited by Foucault are no less important because "it is an improvised conversation."30 These claims are in tension with Miller's claim that it is Foucault's published texts which should be our "primary source" for research.31

So, what we see here is accent to the basic framework of prioritizing the published work over the unpublished work and then making exceptions, or excuses, for why certain unpublished material is acceptable to use. Golder is another example of a scholar who understands the published work as primary while regarding the lecture courses as supplements. He writes,

Jacob S. Fisher also shares this supplement intuition when he writes,

Some argue for this supplemental approach on the grounds of reconstruction. For example, in order to reconstruct Foucault's "Christian book," Jeremy R. Carrette, in the absence at the time of the full published work, puts together a variety of what he calls "fragments." He does this in order to "bring together a selection of central documents, including a course outline, lecture transcripts and published extracts."34 This is then further supplemented with "fragments from interviews and lectures in the early 1980s which further supplement these central texts but on the whole the other pieces only replicate or elaborate material contained in this selection."35 In addition to the interviews, it is also often claimed that Foucault's lectures can provide such a supplement. Take, for example, the position put forward by Brad Elliott Stone who writes,

This supplemental approach implicitly accepts that there is a clear and concise distinction between the published and unpublished works. Only after such a demarcation is accepted, these authors argue that the unpublished works should only be used to supplement the published works.

One strand of this line of thinking questions whether or not we ought to constitute Foucault as the author of a published œuvre at all. Jonathan Simons comes down in favor of constituting a Foucauldian œuvre. He writes,

Some scholars suggest, or simply assume, that Foucault has a single œuvre.38 Some interpreters, such as J. G. Bird and Michael Walzer, have no qualms with trying to establish the definitive and final meaning of Foucault's texts.39 Walzer even explicitly states that he is making Foucault into a subject and assumes he is "an author in the conventional sense" in that he bears the responsibility of his texts.40 On the other side of the spectrum, Mike Gane rejects such interpretations of Foucault that function like post mortem after the death of the author. An interpretation that takes Foucault to be an author in the conventional sense "requires [Foucault] to be pinned down and examined."41 That is, they require one Foucault which, after his death, does not change. David Armstrong claims that there are "many different Foucaults" and the kind of reading that tries to find the 'real Foucault' is simply "some medieval hegemonic gesture" that attempts to grasp what he really meant. Additionally, Armstrong claims that the unity of authorship makes no difference whatsoever to their interpretation. If it turned out there was no historical Foucault at all, or many Foucaults, no disastrous consequences follow. In establishing Foucault's influence, his goal is to "explore the connections between some texts that bear his imprimatur and the reader."42 Despite his radicality, Armstrong still relies on the idea that some of Foucault's texts are published and that this is an important fact.43

This quick overview has shown that there are methodological differences within Foucault scholarship even about what texts are legitimate. However, what seems to be clear is that there is almost universal agreement that a distinction between the published and unpublished work is a coherent idea in itself. What we find in practice, however, is that scholars draw this line of demarcation in very different ways based on the needs of their research projects and the tacit demarcation criteria they accept.

In order to explicate some of the problems developed and discussed within scholarship, I will turn to the question of what counts as the published work. When scholars populate the set of text that they categorize as published, they use different criteria and thus generate non-coextensive sets of texts. Since these criteria are often tacit, this allows disagreement to preserve a veneer of agreement. However, the agreement about how to treat the term "published" hides the fact that scholars and editors are using it in several mutually exclusive ways. I will highlight four different ways in which scholars use the term (published vs. published*, published vs. authorized, and published vs. public). All four of these face historical challenges in the publication history of Foucault's work that disrupt their smooth application and use as universal demarcation criteria.

A final example explicates how these demarcation criteria come into conflict. In 1966 Foucault gave a radio broadcast entitled "Utopies et hétérotopies" to France Culture, France's foremost cultural radio station.82 Ionel Schein, who heard the lecture, invited Foucault to give the same lecture at the Cercle d'études architecturales (Circle of Architectural Studies). A stenographer made a copy of Foucault's lecture which was later distributed to members.83 Foucault at the time was reluctant to publish it.84 However, excerpts of the talk appeared in a journal in 1968.85

It wasn't until shortly before Foucault's death that he allowed the manuscript of the talk, unreviewed by Foucault, to be displayed to the public at the Internationale Bauaustellung Berlin in 1984.86 Later in 1984, a fuller version of this talk was published as "Des espaces autres. une conférence inédite de Michel Foucault" in Architecture, Mouvement, Continuité, which was later translated and republished.87

The editors of Diacritics, who published a translation of it, argue it is not part of Foucault's official corpus because it was not reviewed and published by Foucault [publication*],88 Defert and Anthony Vidler argue it is legitimate because it was authorized [authorization],89 Michiel Dehaene and Lieven De Cauter also argue it was legitimate but emphasize the fact that it was made public [publicness].90 However, this disagreement has tended to ignore the original radio presentation by which these ideas first became public. A consistent emphasis on publicness would also prioritize the broadcast.

Different criteria include or exclude different documents in the history of this text. According to "publication*," the transcribed excerpts from the lecture would count but the radio broadcast, the public presentation in Berlin and the transcription published after his death would not count. According to "authorization," the publication of the text in 1984 would be acceptable but the radio broadcast, the 1968 publication of excerpts, and the public presentation of the text in Berlin would not count. Under the "publicness" criteria, the radio show and the Berlin piece would count since both were public presentations. Additionally, the transcript of the lectures would also count, but, only after 1984 when it was made public. The texts of the lecture that circulated among the architects after the lecture, however, would not count. This means that even regarding the history of a single work, different criteria demand that scholars work with different texts. These texts vary considerably.

As Kelvin Knight and Vidler point out, the radio broadcast contains different material than any of the published work.91 From the perspective of those holding the publicness criteria, the radio broadcast was the foundation upon which the later work developed. Knight concludes that this work was not, as it seems in the published work, about urban architectural space but about fictional space within literature as it was in the broadcast.92 This makes clear that depending on which criteria one makes use of, one is given different texts and these different texts presuppose very different philosophies.

I have demonstrated that scholars often implicitly accept that there exists a clear demarcation between the published and the unpublished work. However, scholars make use of different criteria to justify where this demarcation is made. As I have shown, these all contain historical difficulties such that no single criteria can demarcate all possible historical cases. I have shown that these often tacit demarcation criteria lead to different sets of texts being considered as published.

This should lead us to a central question: Why do we read Foucault and what for? My contention is not that scholars ought to agree on necessary and sufficient conditions of a certain piece of text being part of the published work. I do not think this would be practically achievable given the historical complexities of Foucault's work.

Different projects find different pieces of evidence important. There is no reason to think there should be only one kind of project. I see no reason why a biography of Foucault should be barred from referring to his letters. A work on the history of the publication of Foucault's texts might find the book contracts Foucault signed of immense value. A work on speech writing for public philosophy in France would find the text of Foucault's public lectures of immeasurable value. The point is, that projects frame the kind of methodological concerns which make individual texts valuable. It is perfectly natural that different projects will find different evidence important to prioritize for various reasons. Expecting universal agreement about the value of particular texts would be to make a kind of category error. This is, in fact, one of the ways which Foucault's very analysis of authorship frees us.

However, and I cannot emphasize this enough, I am not saying that facts about the publication history are superfluous. It is an important fact that the interview published in La Quinzaine Littéraire was not published as agreed upon. It is an important fact that Foucault did not attach his name to some work he published. It is important that Foucault changed texts from edition to edition. These are important concrete historical facts about particular texts. What is not important, in my view, is finding a single and universal methodological principle that allows us to decide in every case whether a piece of text is part of the set designated published or the set designated unpublished.

As scholars of Foucault and Foucault's philosophy, one must admit there is no single unchallenged demarcation between the published work and the unpublished. Each criteria, applied universally, is inconsistent and irreconcilable with historical evidence. Multiple criteria applied simultaneously generates mutually exclusive and non-coextensive sets.

We ought to become explicit about why a piece of text, in its material particularity, is valuable to our projects. In particular, we should follow Foucault's own example when he said of Nietzsche's works that we ought to return to the manuscripts themselves.93 In this sense, one would no longer write about the "work" Madness and Civilization in abstract terms but instead refer specifically to those editions that are of import to the project. Such an approach would require an explanation of how and why they are important.

For example, we can imagine a project being interested in the philosophy to which Foucault publicly attached his imprimatur. If such a scholarly project is interested in the final form of Foucault's philosophy, then it should start with the individual texts to which Foucault publicly attached his imprimatur. Such a project should, however, not neglect the complex historical and contextual evidence surrounding the status of Foucault's imprimatur on those material objects that constitute individual texts. If one's project values Foucault's public imprimatur, then the set of valuable texts is not coextensive with the set of published texts. It would necessarily exclude his contributions and publications in the left press that remained unsigned as well as his published, but anonymous, conversations with Thierry Voeltzel.94 Simply because Foucault's words are published during his life does not mean Foucault publicly attached his imprimatur to those words. By focusing on Foucault's imprimatur, one no longer has need of a strict "published" vs. "unpublished" dichotomy. One might even go further to specify not just works and editions, but particular material objects.

The failure of the published or unpublished dichotomy to capture the world is demonstrated best by focusing on the rich histories of individual texts as material objects. Consider, for example, that Foucault gave a first edition copy of his Raymond Roussel to the radio journalist Alain Trutat with a very short interpretation, or at least description of the function, of the text itself (Figure 1).

He writes, "For Alain Trutat, because he allowed me to talk about Roussel – with best regards."96 The point here is not that this is a discourse changing interpretation of Foucault's own work. However, what it does demonstrate is that this text exemplifies physio-graphic characteristics different from other texts of the same edition and these differences are part of Foucault's writing.

An argument can be made that the text originally printed here in the first edition (1963) is published despite the bewildering history of its publication.97 When we focus on individual material objects, rather than the abstraction of "the work" or even "edition," the question of the division between the published and the unpublished can be radically questioned. Should we categorize this particular text, this material object, as published or unpublished? Should this text be considered authorized? Should the writing in this text be considered public or private?98 While the type on the page released to the public in 1963 may have been made public, the hand-written annotation by Foucault's hand certainly was not. When we focus on this unique material object, which bears the traces of Michel Foucault, it becomes clear it is both published and unpublished.

The traces of Foucault's work left after his death create a problematic situation. One might think that the source of the problem is only Foucault himself. However, closer scrutiny of scholarly practices in philosophy more generally reveal that this is not an idiosyncrasy of Foucault's life alone. As Foucault himself noted, after editing the complete works of Nietzsche in French, these were also difficulties for editing Nietzsche's work. Foucault's arguments have proven the test of time. The tacit demarcation criteria I have articulated among Foucault scholars are also present among Nietzsche scholars as I have argued elsewhere.99

The tacit assumption that there can be a clear distinction between published and unpublished work is not only found in scholarship on Foucault and Nietzsche but is a ubiquitous presupposition within standard commentaries on Descartes, Locke, Berkeley, Bentham, Kant, and even Derrida.100 Publication itself is messy and such a widespread acceptance of the dichotomy between published and unpublished works points to a larger problem in historical scholarship to which we ought to attend. However, this goes beyond the scope of this essay.

A second objection might be that many scholars argue there exists not one Foucault but many Foucaults. One of the freeing things about the ideas expressed in Foucault's texts is that authorial intention and even an author are not necessary for discourse formation. I would readily admit there are many ways in which to read Foucault and, as I have demonstrated in this essay, many ways to demarcate between published and unpublished texts. When we include not only scholarly use, but the varied use of Foucault for artistic and applied purposes these expand considerably. However, Foucault scholars, those interested in interpreting his texts, all admit that there are better and worse ways to read Foucault.101 Foucault himself suggested that certain forms of editing and interpretation can deform (déformations) texts and lead to errors (erreurs) and omissions (omissions).102 The implicit idea here, shared by most scholars, is not only that there are many ways to interpret Foucault, many Foucaults, but that some are more plausible than others in certain contexts. If there is to be any answer to these questions, if an answer could be possible at all, my argument holds that we first need to ask; "how do we read Foucault and for what purpose?"

I have argued that the dichotomy between the published or unpublished offers an insufficient framework to answer the question of how we read Foucault. Further, this framework actively conceals the deep methodological disagreements that find their source in tacit criteria that privilege non-coextensive sets of texts. Instead, I argue that by focusing on the values of the projects at hand and the individual texts in question, we can avoid needless abstraction and overgeneralization. We ought to reject any criteria that claims to universally demarcate between the false dichotomy of published and unpublished. Each textual artifact is materially unique and carries the burden of a deep and complex history to which we ought to bear witness in our scholarship. Explicating how we are using particular texts and why those texts are valuable to our particular projects would allow us to avoid historically falsifiable abstractions and focus on the material objects and their history. Let us return, as Foucault himself suggests, to the manuscripts themselves.