Welcome to The Blackthorn Hedge. Today’s letter is another exploration of a common metaphor or idiom, this time the “Overton Window,” like before I’ve looked at hive minds, mind viruses, and a horror variant of political horseshoe theory. This is also a sort of reversal of the last letter on Sunday: if that one was about ways to get out of discourse, this one is about staying in it.
As always, I welcome all respectful comments, messages, and other engagement.
It’s commonplace knowledge that there are things we cannot safely and politely say, and it’s common to express this spatially in terms of “bounds” on speech. Speech is “out of bounds” if it induces unacceptable responses, for instance, or if it makes a bad impression.
The origins of these spatial metaphors are varied and highly redundant, coming from everything from sports practice to coloring books. It’s an extremely natural abstraction from any number of well-motivated and culturally stable basic skill practices.
Among specific spatial metaphors for bounds on acceptable speech, over the last couple of decades “The Overton Window” has become a pervasive idiom for talking about acceptable political speech. It began as a specific concept for coordinating fundraising by the libertarian think tank the Mackinac Center for Public Policy, in the late 1990s in conversations led by Joseph Overton: these conversations introduced a simple scale for judging how palatable policy recommendations would be to an audience, from the most extreme, popularly unthinkable socialist recommendations to most extreme, popularly unthinkable free market recommendations. Imagine “98% income taxes to fund gulags” on one side and “legal death pools and contract killing at the same time” on the other,1 with current policies in the middle and the most popular left wing policy recommendations and right wing policy recommendations adjacent to it on each side. Since they were libertarians, they themselves wrote it as a vertical axis rather than left-right and put “more free” on top and “less free” on bottom, but that framing didn’t stick.
The point for these inventors was to sort political messages according to whether they would be politically viable. It was a mapping concept, meant to help them aim their messages at realistic targets. Their idea was that the populace had a window of recommendations they could accept and politicians needed to craft their own messages to land in that window; as befit a libertarian ideology, the idea was that the politicians should be supplying messages in response to some pre-existing, autonomous popular demand.
That also didn’t stick, to put it mildly. As anyone who has seen the trillion-dollar revenues in advertising knows,2 market demand isn’t really pre-existing and autonomous. Demand is made just as much as supply. The same is true for political demand. Thus people have been talking about politicians “shifting the Overton window” or “widening the Overton window” or “smashing the Overton window” for as long as the idiom has been popular, even though it’s exactly against the libertarian spirit that the inventors intended to realize through their concept.3
The work the idiom does now in private and public conversation is widely varied. Sometimes it’s a window on a free market to socialist axis, sometimes it’s on a right to left axis, sometimes it’s on a based to woke axis. Sometimes it’s used like a map of a fixed territory, as the libertarians intended; other times it’s used like a map declaring contested territorial claims, as the “wideners” and “smashers” repurposed it. In any case, it’s pervasive among people I talk to, and I thought that I should take some time to write about some distinctive ways I use it myself.
Specifically, I’ll talk about the multiplicity and overlap of the windows, their connectivity and dimensionality, and their convexity and smoothness or roughness on their boundaries. I find all these shape concepts are useful in practice: they’re good for opening up frustratingly stereotypical and antagonistic back-and-forth cliche tennis conversations into creative conversations with more collaborative potential. I also just enjoy shape talk; I had a great time with Jordan Ellenberg’s recent book Shape: The Hidden Geometry of Information, Biology, Strategy, Democracy, and Everything Else right after it came out in paperback.
The Blackthorn Hedge is not a project to widen the broad public’s Overton Window, but it is a project to create a small public social center with a distinctive range of topics that are free and open to discuss, a sort of small public Overton Window of its own – that’s where I’ll start with “multiplicity.” I’ve discussed that goal most directly and at greatest length in “Motte and Bailey; Hedge and Field,” but I repeat it often in my other letters as well. I think most public online communities implicitly do create their own small public Overton Windows, their own bounds of local political recommendation, spontaneously as they grow. I’ll have much more to say on that spontaneous growth in Thursday’s letter; here I’ll just focus on some properties of what it is that emerges in the process of growth.
Multiplicity & Overlap
Whether or not they should, people use “Overton Window” to describe limits of any number of kinds of speech in any number of contexts: political or sexual, public or private, groupchat or forum, banquet or protest, bull session or debate. Some people will refer to an “Overton Window” of almost anything.
Not all of these are equally useful, but I’ve never seen much hope in the short term for policing “proper use,” not with currently prevailing online social norms against this sort of normative conceptual authority. Instead I’ve tried to be ready to work with any of these uses: whatever anyone tries to say with the idiom, I’ll try to hear it.
There is one kind of multiplicity to the Overton Window that is certainly appropriate to use: temporal multiplicity. The Mackinac Center inventors have always emphasized that the Overton Window changes. The Overton Window of today and the Overton Window of ten years from now will probably not be the same: though many unthinkable things will remain unthinkable and many policies will remain unchanged by reform, there will be some policies that change in a way that makes reversing the change again now unthinkable (Roe v. Wade comes to mind as a controversial example) and there will be some things that were unthinkable that become policy (for instance major financial reforms in the wake of recessions).
People say “the Window widens” when a later Window includes more of the fringe recommendations that weren’t quite in the Window before, and they say it “shifts” when it includes more of one side’s fringe and less of the other side’s fringe, but it’s an odd metaphor. There’s also another natural possibility, often skipped over: the recommendations have moved into the Window rather than the Window moving to the recommendations. The political positioning of a recommendation can change: it’s not actually a fixed, objective attribute of the recommendation.
This brings up a second kind of multiplicity to the Overton Window that is also certainly appropriate to use: axial multiplicity. Various versions of “The X-Y Overton Window” where X and Y are political opposites. The “more free” versus “less free” original form of the Window was never going to be conceptually stable because “free” is an essentially contested concept in politics. Different groups would mean different things with the same labels “more free” and “less free”, and they would therefore interpret the Window differently even if they had stuck to the same words… which they certainly didn’t. Every political dualism can have its own Overton Window.
These different dualisms correspond to different groups of political interpreters, bringing up a third kind of multiplicity: interpretational multiplicity. The “Window according to X,” where X is anyone or any group that has decided to arrange political recommendations on some dualistic spectrum where the center is fully accepted and the extremes are popularly unthinkable.
These can be shifted or widened compared to one another in the same way that the temporally different Windows can shift or widen, and they can also overlap in new ways, more like “tilting” than shifting or widening, related to the similarity of the axes and interpreters: a right wing vs. left wing axis will have some clear similarity to a free market vs socialist axis, for instance, and the political positioning judgments of the right wing Heritage think tank will probably overlap significantly with those of the free market Mackinac Center.
Finally, the fourth kind of multiplicity I’ll mention, another that arises again and again, is audience multiplicity. Different audiences accept different recommendations. “The Window for X,” where X is any audience, possibly not even a human audience if it’s an automated bot, and possibly not even a real audience since people have no trouble imagining and then analyzing audiences that don’t actually exist (including audiences of robots that will never exist). These Windows overlap insofar as the audiences are similar to one another.
These multiplicities are useful for countering temptations or pressures to treat claimed Overton Windows as more fixed, more fundamental, more objective, and more universal than they are. In fact they change, they’re conceptually flexible, they’re interpretively subjective, and they’re particular to certain audiences. Always thinking of each Window as some “opposition X-Y Window for audience Z according to interpreters W in timeframe T” is extra work up front, but it cuts through confusion and in the end that saves time and energy.
_by_Sentinel-2.jpg){kind=link}
Connectivity and Dimensionality
One of the most implausible aspects of the Overton Window metaphor is that it is constructed as if recommendations are more acceptable the more centrist they are. In conditions of serious political polarization, this isn’t true at all: the populace wants to see a fight. They want people to express commitment to one principle or another and they mistrust the “mushy” unprincipled middle.
In American politics, Eric Hoffer’s The True Believer has an iconic description of this dynamic:
When people are ripe for a mass movement, they are usually ripe for any effective movement, and not solely for one with a particular doctrine or program. In pre-Hitlerian Germany it was often a toss-up whether a restless youth would join the Communists or the Nazis. In the overcrowded pale of Czarist Russia the simmering Jewish population was ripe for both revolution and Zionism. In the same family, one member would join the revolutionaries and the other the Zionists. Dr. Chaim Weizmann quotes a saying of his mother in those days: “Whatever happens, I shall be well off. If Shemuel [the revolutionary son] is right, we shall all be happy in Russia; and if Chaim [the Zionist] is right, then I shall go to live in Palestine.” (Hoffer, The True Believer Part 1, Section III, Number 14 (1951))
This book has been recommended by figures as different as Dwight D. Eisenhower, Hillary Clinton, and Marc Andreessen: it represents a common thread in American political traditions. These traditions do not take for granted that the center holds.4
In conditions of serious binary polarization, the middle might become unacceptable in a way that the Overton Window can’t directly capture unless it’s split into two: one window on one side of the middle, one window on the other. When people talk about “smashing the Overton Window,” I usually find that this is what they mean: they mean that the center has not held and a once-unitary Window has now split into disconnected pieces.
However, there are some puzzles here: first of all, are both of the split halves of old Window right to call “acceptable” to the full populace? In some cases this does hold: the two sides genuinely respect and recognize one another as antagonists, so the populace as a whole can genuinely respect and accept recommendations from each of the two; they’re simply two separate Panes of a single popular Window. More often, however, the populace has also split into separate audiences that accept different recommendations: the two Panes are the Windows of two different audiences.
Even more often, there is some complex mixture of both: there is part of the populace that respects both sides, as an audience, and there are two more parts that respect either of each of the two sides but not the other. And usually there is also a part of the audience that rejects the whole setup and accepts neither!
Most often, there is not a total split of the Window and there is a center with some acceptance but possibly less than either of the polarized parties. This sort of situation, which seems to match our contemporary American politics to me, can superficially look like one where the overall Window metaphor works. However, it’s much more informative to split out the different audiences if you’re doing political messaging work in these cases: sending the same message to the entire populace will result in very different responses from its different parts.
To split them, the most natural thing to do is often to add another axis: if the original axis was left-right, the added axis could be polarization. Then what would emerge, most often, is a Λ-shaped sort of structure: there is a bipartisan center window that is viable for the issues that are not polarized at the same time there are two opposed partisan window panes that are viable for the most polarized issues. The opposed panes get closer to one another as the issues become less polarized, and if the center is not entirely gone, then there is some level of polarization at which the split windows merge.
If the center is gone, then it’s not a “Λ” but rather just some kind of “| |” or “. .” figure. “| \” if one side is more willing to compromise on bipartisan issues than the other but still not enough to form a center. “| .” if one of the sides has recommendations they’ll accept about the bipartisan issues but the other is entirely focused on the polarized issues (arguably our situation today).
And here we might also want to start giving some credit to more minority viewpoints and extreme outlier parties – we could start seeing Window shapes like “. Λ” or even “| . Λ | .” when some but not all of the further extremists have their own distinctive platforms for what the central partisans would agree on. If there are hyper-partisan issues that unify a normally divided middle against the wildest extremists, the Window could look like “. U .” and I leave it to the reader to imagine as many more of these two-dimensional figures as they would like.
Each of the disconnected elements of these figures represents a distinctly different part of the populace’s set of acceptable political recommendations. The more complex elements like the “Λ” and “U” represent mixed parts of the populace in partial agreement with one another, not fully politically disconnected but also not fully unified.
Once disconnections and multidimensionality are taken into account, these don’t seem very Window-like anymore, to me. They’re much more like islands of political viability, and I find this an effective alternative in practice: when people want to talk to me about populaces with more complex political heterogeneity, I ask them to drop the Window idiom and start talking about Overton Islands. A chain of Overton Islands like “| . Λ .” is, naturally, an Overton Archipelago.
Once that complexity is on the table, it’s also easy to switch up the dimensions: instead of left/right and polarized/bipartisan, one can use authoritarian/libertarian and left/right or Islamic/secular and left/right or any of the many other political spectra anyone has come up with. And there’s also no problem going to three dimensions, though now instead of Islands one has something like more like Clouds and instead of an Overton Archipelago it’s an Overton Nebula.
If you keep stretching your imagination into higher dimensions, into hyperspace, then you’re implicitly starting to explore general vector embeddings and you’re well on your way to learning the fundamentals of modern AI mathematics.
Convexity and Roughness
The issue of connectivity above was intertwined with the common illusion (or pretense) that a more centrist, more moderate position is always more acceptable to the public than a more extreme position. In higher dimensions, this becomes an even more important issue: between any two somewhat-acceptable or somewhat-popular recommendations, are straight-line compromises at least as acceptable or popular?
It’s not true for the disconnected islands of those simple Overton Archipelagos I used as examples above. It’s also not true for the “U” shaped or “Λ” shaped Islands. It is true for the simple “.” and “|” Islands, including the slanted “\” Islands.
The geometrical property that corresponds to this intuitive property is “convexity,” and while it’s technical, it’s also important. The arguments for possible disconnection in one dimension are sufficient arguments for possible non-convexity in higher dimensions, because convexity is no more or less than the property that every cross-section of a shape by any ray passing through the shape is simply connected. Circles and triangles are convex, while stars and trees are not.
Stars do have a different property somewhat like convexity: they have simply connected cross-sections on any ray passing through the center of the star. Some political coalitions are like this: they have several different partisan factional components, corresponding to arms of the star, that can each form many acceptable compromises with the center – but the populace finds it more difficult to approve of compromises directly between any two adjacent partisan factions. This is common in imperial politics with a dominant but fragile center, where peripheral alliances may seem threatening to the general order of things so the factions of the periphery are kept divided. In this way Overton Star Islands are an important type. So are near-stars where the arms curve around somewhat as they extend.
Convex objects do not have spikes, by contrast. A single spike might be convex, but no object with spikes is. If we take the imaginary Overton Star Island and then imagine the alternative where all the jutting political arms of the coalition could make compromises among one another acceptably, then the star would become a sort of spikeless pancake shape. This construction, adding all these “missing” compromises, is called a “convex hull” in geometry. The boundaries are straight except at corners corresponding to the points of the star’s arms. The convex hull of the “Λ” is a triangle, and the convex hull of the “U” is the bullet-shaped filled “U”.
This brings me back to an issue I skimmed past in the prior section on dimension: in the original one dimensional case, the Overton Window has several degrees from “unthinkable” to “implemented,” and every successive degree corresponds to a more centrally nested one-dimensional window interval. Success implies popularity, popularity implies acceptability, and acceptability implies thinkability.5 If the windows are more complex than unitary intervals, for instance two-paned pairs of intervals, then they can nest more interestingly but they must still nest consistently: some panes of the Overton Window might just be full of acceptable recommendations with no contents that are popular or none that are successful, for instance.
In the Overton Island metaphor, we can think of these as different topographic elevations: acceptability might be the shoreline, popularity a ridge height, success a mountain threshold, thinkability a shallow reef depth. In two dimensions this becomes more interesting because the nested shapes can have different shapes: a common example, building on the empire example before, is a star-shaped shoreline for acceptability with a nested star of popularity ridges inside, all those ridges coming off of a circular central mountain of implemented policy, all surrounded by shallow bays and lagoons out to a broad pancake-shaped underwater limit for the thinkable possibilities (a pancake shaped a lot like the convex hull of the star-like shoreline). It’s a common kind of island and a common kind of Overton Island.
Each of the boundaries might be fairly complex at a fine grain, as well: just like there are jagged hillsides on some islands, there can be jagged boundaries to political acceptability. “Put one foot wrong and you’re falling off the cliff.” These arise naturally when there are high stakes and high erosion. If there are spies or gatekeepers listening for minor signs of disloyalty, for instance, the whole society will tend to roughen its discursive boundaries. More will be hinted rather than said, and also more that is said will be judged closely for hints. Here we get into the territory of fatal Soviet and Fascist doctrinal disputes, Leo Strauss’s Persecution and the Art of Writing, Alexandre Koyré’s “The Political Function of the Modern Lie,” or de Santillana’s The Crime of Galileo.
Argonauts of the Overton Archipelago
There are as many possible Overton Islands as there are possible Overton Windows: temporal, axial, interpretational, and audience multiplicity all apply just the same way except that the axes and audiences will be more complex.
A journey between Overton Islands can be as simple as going between rooms of a politically mixed party. I often recall one School of the Art Institute of Chicago student Halloween party, around 2011, where in one room everyone was gathered around a couple of keffiyeh-wearing guests talking about Palestine and Edward Said while in the kitchen the organizers were talking about the culture industry and Theodor Adorno; the lines were stark and hostile, and later in the night there was a fight between two leaders of the two groups.
Sometimes these Islands suddenly and catastrophically sink into the depths or explode into jagged chunks. They may also suddenly rise up like new volcanos or slowly accrete like coral atolls. The signs any new Island has risen or fallen may draw visitors curious about the new shape and new possibilities in the aftermath (or just curious to learn what to do if their own Islands are in danger). How will the abrupt rise of the Abundance platform affect the seas? How about the sinking of Urbit? How many islands could the flood of LLM slop drown? What new Island shapes could good uses of LLMs raise?
The Blackthorn Hedge is rising as a sort of ring-atoll (without any LLM use). I’m first accreting a border of essays that raise controversial topics without (I hope) inflaming controversy, and I’m not in a hurry to raise up any forbidding central mountain of authority.
At the same time, I’m enjoying my time on Substack getting to know the other Islands of the local Archipelago: I’m not here primarily for my own writing! I’m doing my best to explore some political matters that I’ve put off for a long time, in particular: actually coming in from shore, unusually for me, given how happy I usually am being just marginally thinkable or acceptable. I’m liking a literary side of the platform, as well, from the Blavatsky and Yeats milieux at
to the Adam Smith and Austen of to the Caro and Conan Doyle of .I’m still finding my bearings and seeing where I’m best suited. If I’ve been saying things you find unacceptable or otherwise making myself unwelcome on any of our shared Islands, please feel free to message me. I don’t take spontaneous speech for granted and I’m generally happy to learn better etiquette.
Assume that somehow the contract killers do not violate the nonaggression principle.
Or, more darkly: “as anyone who has followed the Purdue Pharma opiate scandals knows.”
In other words from Jean Baudrillard on Marx in The Mirror of Production, the current symbolic exchange value of the idiom is disconnected from its originally intended use value.
For a particularly libertarian-friendly exploration, see Albert O. Hirschman’s classic Exit, Voice, and Loyalty and its discussion of “Hotelling’s Law.”
Naturally this is an idealistic presentation of how the sausage gets made, and a cynic will know that “popular” is a codeword here if it’s supposed to be a requirement for policy implementation.
Thanks for this letter! I was excited by this as potentially a good entry point to share with a few particular people. A few musings I had about the coming shapes of windows and islands, shared for fun and in case they happen to resonate: personally, I expect waves of numerous meme windows as we move towards more hyper-personalization and then waves of increased uniformity as certain ai-generated windows (frames for looking at society) emerge as particularly salient and powerful. Maybe in your metaphor, the numerous small windows are part of the archipelago, and the large salient movements are tsunamis or tropical storms that occasionally submerge and remap them. Very evocative, and to me reassuring that if one isn’t finding a home on one island, there are wide seas to explore.
really appreciated the thoroughness with which you attend to the metaphors we use to conceptualize this topic. very nicely framed.