That's What She Said: A New Theory of What's Funny

Abstract

This article relocates humour theory from the joke to interaction. Existing accounts—superiority, relief, incongruity, Bergson, benign violation—treat the joke as the unit of analysis and therefore cannot explain relationship-dependent effects, contentless running jokes, or some punchline-free conversations. Living Value Theory instead defines humour as interrecursive recognition: the mutually achieved crossing of a recursive register, constituted across five mediations within a shared mesocosm. Comic timing and rapid alliance reorganisation are its central mechanisms. Running jokes become perceptual equipment; deadpan preserves audience recognition; frame incoherence refines Bergson. Laughter marks successful arrival at a newly shared playful register.

Introduction

Four words carry almost no information. They contain no observation about the world, no argument, no image. Said on their own, stripped of context, they are barely a sentence. Yet inserted after almost any remark that can be misheard as a double entendre, "that's what she said" reliably produces laughter, and it has gone on producing laughter for years, across enormous and largely unacquainted audiences, despite carrying no new content on any single occasion of its use. This is a strange fact, and it is stranger the longer one looks at it. A theory of humour that cannot explain why an almost contentless four word phrase remains funny on its four thousandth deployment has not yet located the thing that is actually funny. This article argues that existing theories of humour, however sophisticated, have systematically looked in the wrong place, because they have treated the joke as the unit of analysis rather than the interaction within which the joke occurs. Living Value Theory relocates the analysis from the joke to the encounter, and the claim defended here is a strong one. Humour is not fundamentally about incongruity, violation, or linguistic ambiguity. It is about the successful coordination of lived time between participants who share enough of a common mesocosm to recognise, together, that the organisation of their interaction has just changed, and about the continual reorganisation of who is recursively aligned with whom while that change is underway. Jokes, punchlines, pauses, tone, deadpan delivery, running jokes, and the well documented failure of artificial intelligence to participate convincingly in humorous exchange are all different manifestations of that single deeper phenomenon, which this article names interrecursive recognition, and which turns out to be constituted, in every instance, across the five mediations rather than floating free of them.

The Limits of Joke Centred Theory

The dominant theories of humour in the scholarly literature are superiority, relief, and incongruity, joined more recently by Bergson's account of mechanical rigidity and by the integrative Benign Violation Theory. Each of these theories has real explanatory power, and each has been tested against a great deal of material. Hobbes located laughter in a sudden sense of eminence over the infirmity of another. Freud located it in the discharge of psychic energy that had been mobilised to maintain a repression and was then permitted release. Kant and Schopenhauer, followed by the modern cognitive tradition running through Suls, located it in the perception of an incongruity that is subsequently resolved by a hidden rule. Bergson located it in the perception of something mechanical superimposed on something living. McGraw and Warren's Benign Violation Theory attempted a synthesis, proposing that humour requires a situation to be simultaneously appraised as a violation and as safe.

What unites all five accounts, despite their real differences, is where they begin. Each begins with the joke, or with the situation considered as a bounded, nameable unit, and asks what property of that unit produces the laugh. This is a natural place to begin, because jokes are the most visible and most portable form that humour takes, and because a theory that explains the joke as text has the appeal of tractability. But it is also, on the account developed here, a mistake, because it treats the humorous event as though it were fundamentally a symbolic object rather than fundamentally a coordination between people, unfolding in shared time, across shared mediations. A theory that starts at the level of the text has already, without noticing, chosen to analyse the residue of a humorous event rather than the event itself.

Three difficulties follow directly from this starting point, and none of them is solved by refining the theories further. First, none of the joke centred theories explains why an identical utterance can be devastating in one relationship and delight in another. If the incongruity, the violation, the sense of superiority, and the mechanical rigidity are all held constant, and only the relationship between speaker and listener varies, joke centred theories have nothing further to say, because they have located the comic property entirely inside the utterance. Second, none of them explains the survival of running jokes, whose informational content approaches zero and yet whose comic effect does not decay with repetition in the way that ordinary jokes decay. Third, and most tellingly, none of them explains material of the kind produced by the comedian who goes by the name Husk, whose videos consist of long, largely punchline free conversations between a human being and an artificial intelligence, and which are nonetheless experienced by millions of viewers as extremely funny. If the object of analysis is joke structure, this material should not be funny at all, because in the overwhelming majority of these exchanges the artificial intelligence says something entirely reasonable. There is no incongruous script opposition sitting inside the transcript waiting to be found. The humour, whatever it is, is not in the words. It is in the lived time of an encounter that a transcript cannot capture.

Relocating Humour: The Mesocosm as the Ground of the Joke, Not Its Background

Living Value Theory begins from a different unit of analysis. Human life is understood as ongoing coordination across five irreducible mediations, multisensorial embodiment, multispecies being-with, multiversal dwelling, multimateriality, and multisymbolization, within a lived field of mutual availability that Living Value Theory calls the mesocosm. Within that field, coordination between two or more interrecursively responsive parties is never static. It moves continuously between registers, informational, playful, ironic, affectionate, hostile, serious, without any of these transitions being explicitly negotiated. Recursive discernment, the continual, largely tacit activity through which participants track what kind of coordination is currently underway and what kind of response it calls for, is running throughout, almost always below the threshold of explicit attention.

It would be easy, at this point, to say simply that humour occurs when two parties recognise together that the recursive register of their interaction has shifted, and to treat the mesocosm as the backdrop against which that recognition happens. This article argues for a stronger and more specifically Living Value Theoretic claim. The recursive shift is not merely recognised against a shared mesocosm. It is constituted by shared mediational participation in the first place. There is no register shift floating free of the mediations that carry it. A shift in register is always a shift in how embodiment, being-with, dwelling, multimateriality, and multisymbolization are jointly organised between the parties at that moment, and the recognition of the shift is only possible to the extent that both parties inhabit enough of the same configuration across those five dimensions to feel it happen. Recursive regime recognition, in other words, depends upon sufficiently shared mediational participation. Where that shared participation is absent in one or more mediations, certain shifts become literally unavailable to one of the parties, not merely undetected but impossible to constitute from that party's position at all.

Humour, on this account, is the shared, mutually recognised crossing of a boundary between recursive registers, where that crossing and its recognition are both constituted across the five mediations rather than achieved in the abstract. The joke, the pun, the incongruous punchline, is very often simply the symbolic vehicle that makes a mediationally grounded shift visible enough for mutual recognition to occur. It is the trace of the event, not its cause, and the event itself is never purely symbolic. This relocation immediately answers the first difficulty identified above. The same sentence can devastate in one relationship and delight in another because its comic effect was never located in the sentence. It depends on whether the two parties' accumulated history of interrecursive coordination across embodiment, being-with, dwelling, materiality, and symbolisation, and their mutual attribution of one another as safe within that shared history, permits the shift into a playful register to be recognised as playful rather than as a genuine and unrepaired violation.

Comic Timing as the Central Case: Recognition of a Change in Lived Time

Comedians have always insisted that timing is everything, and comedy criticism has correspondingly said remarkably little of precision about what timing actually is. The vocabulary of theatrical craft, the beat, the pause, the pace, describes symptoms of good timing without naming its underlying mechanism. This article treats comic timing as one of two central mechanisms it proposes, alongside the continual reorganisation of alliance discussed in the section that follows, because timing is nothing other than recursive regime recognition observed at the moment it happens, in real time, between embodied participants. An interaction proceeds, for some stretch, at a stable recursive register, most often something close to L1, the level of unreflective, smoothly proceeding coordination, whether that coordination is informational, task focused, or simply ordinary sociable exchange. At some point a disturbance is registered, an L2 event, something that does not fit the register that has been governing the interaction so far. What happens next determines whether the moment becomes comic.

Once timing is understood this way, an unusually large range of comic phenomena fall out of a single principle. A pause before a punchline works because it holds the interaction at the threshold of the shift for exactly as long as the audience needs to feel the disturbance without yet resolving it, no longer. An interruption is funny when it forces a shift in register before either party has fully settled into the previous one, producing a small, pleasurable jolt of recursive whiplash. Overacting fails as comedy more often than it succeeds because it announces the shift so emphatically that the audience is denied the act of recognising it themselves, while underacting, of which deadpan is the extreme case, can succeed precisely because it leaves the recognition to the audience. A callback, a joke that reactivates an earlier moment much later in a performance, works by demonstrating that a specific recursive register, established once, has remained available the whole time, ready to be reopened with almost no new symbolic material at all.

Comic timing is therefore best understood as recursive fluidity, the capacity to move between coordination registers in ways that preserve and repair coordination, exercised specifically at the boundary between seriousness and play, and applied not merely to one's own movement between registers but to the accurate perception of when an interlocutor, or an audience, has become ready to move. Good comedians perceive this readiness before it is announced. Bad comedians announce a shift that the audience has not yet been prepared to recognise, which is what an explained joke, or a joke delivered too early, actually is. Current artificial intelligence, as later sections argue at length, essentially always misses this transition, not because it lacks a repertoire of jokes, but because it lacks sufficient participation in the mediations across which the transition is constituted in the first place, which means it is not merely slow to notice the shift. In a great many cases the shift is not available to it at all.

From Disturbance to Shared L1: What the Laugh Actually Marks

One further clarification is needed before the theory can be applied in detail, and it corrects something that even the account offered so far has left insufficiently precise. It is tempting to treat the laugh itself as the L2 event, the registration that something is funny. This is a mistake, and correcting it improves considerably on the incongruity resolution models it most resembles. The joke, or the situation Husk constructs, or the friend's callback, creates the L2 disturbance. What follows the disturbance is not yet laughter. It is a brief, often barely conscious interval in which the interaction's recursive organisation is actively being reworked, the old register no longer quite holding, a new one not yet confirmed as shared. The laugh itself belongs to a later moment in the sequence. It is the bodily marker that this reorganisation has succeeded, that a new register, playful rather than the one that preceded it, has actually been reached and has actually been recognised by both parties as reached together. The full sequence therefore runs from stable coordination, through disturbance, through reorganisation, to a newly shared register that is, in its own right, a fresh instance of L1, and only then to laughter, which announces the arrival rather than the arriving.

This sequence explains a phenomenon that incongruity theories have always struggled to accommodate gracefully, the peculiar discomfort of a joke that fails. On the account offered here, a failed joke is not simply a joke that was insufficiently incongruous or insufficiently violated. It is a sequence that stalled partway through. The disturbance occurred. Both parties registered that something had shifted, or was meant to shift. But the reorganisation into a new, mutually confirmed register never completed, and so neither party arrives at the moment that would let laughter mark a successful resolution. What is left is not neutrality but a specifically uncomfortable suspension, both parties aware that something was supposed to resolve and did not, still inside the L2 disturbance with no shared L1 on the other side of it to return to. This is a more precise description of why failed jokes feel awkward rather than merely flat, and it is not available to a theory that treats the laugh as coextensive with the perception of incongruity itself rather than as the signature of its successful resolution.

One further refinement is needed, because the sequence just described can suggest, wrongly, that the parties inside a witnessed exchange must themselves reach the new shared register for a laugh to mark anything real. This is not so. An audience watching two people argue, neither of whom ever concedes anything or reaches any accommodation with the other, can still complete the full sequence entirely on its own, disturbance, reorganisation, a newly shared register, and laughter, without the two figures being watched ever converging at all. What must be shared, for the laugh to mark a genuine resolution, is the audience's own recognition of a pattern, not necessarily any resolution achieved by the parties who supplied the material for it. This is, if anything, the more common case in comic material involving more than two people, an audience recognising, together, that one party keeps failing to meet another's appeal, without either party inside the exchange ever recognising anything of the kind themselves. The shared L1 a laugh marks, in other words, can belong to the audience alone, or to the audience together with one of the two original parties, with no requirement that the disturbance's original participants ever resolve anything between themselves.

It is also worth noting that this sequence need not complete in a single pass. Many of the richest comic exchanges are iterative rather than single cycle, one attempt at resolution failing, prompting another, which also fails, the disturbance deepening rather than resolving across several attempts before a capping release finally lands. The underlying sequence, disturbance, reorganisation, shared register, laughter, still holds, but the reorganisation stage can itself contain several failed sub-attempts, each one a small disturbance in its own right, before the sequence as a whole resolves. What looks, from a distance, like a single joke landing is very often, on closer inspection, the last and most complete of several attempts that came before it, each one raising the stakes of the recognition slightly further than the one before.

Recursive Alliances: The Second Central Mechanism

Comic timing describes how a shift in recursive organisation is recognised. It does not, by itself, describe who ends up standing where once the shift has happened, and this is the second mechanism this article proposes as central to the theory rather than as one of its consequences. A joke does not merely produce recognition between two parties considered in isolation. It reorganises, almost always temporarily and often very rapidly, who is recursively aligned with whom. The configurations this reorganisation can take are various, and it is worth naming the recurring ones directly. In one configuration the comedian and the audience align together against a target, the structure classical superiority theory was built to describe, though superiority theory never asked what holds the comedian and audience together as allies rather than merely asking why the target is diminished. In a second configuration audience members align with one another independently of the comedian entirely, as strangers do when they catch one another's eye during the same joke, a lateral alliance the comedian may not even be aware has formed. In a third configuration the comedian aligns with the ostensible target against the audience's own expectations, subverting who was supposed to be laughed at and briefly repositioning the audience as the party whose assumptions are now exposed. And on occasion the alliance inverts entirely, the audience turning its laughter onto the comedian, a shift performers dread precisely because it demonstrates how unstable every one of these alignments always is.

What is remarkable is not merely that these configurations exist but that comedy, more than almost any other form of ordinary interaction, tolerates their extremely rapid succession without the encounter collapsing into confusion or offence. The alliance map can be redrawn two or three times within a matter of seconds, comedian and audience aligned against a premise, then audience members aligned with one another against the comedian's assumption, then the comedian and a single audience member aligned against the rest of the room, and the interaction not only survives this but is often experienced as more pleasurable for the speed of the redrawing. The crowd work practised by stand-up comedians, in which alliances with individual members of an audience are formed, tested, inverted, and reformed in rapid succession, sometimes within a single exchange, of the kind associated with performers such as Jimmy Carr, offers an unusually clear public illustration of how quickly these alignments can shift without the encounter losing its coherence. This tolerance for rapid realignment is itself a striking instance of recursive fluidity operating at the level of a whole group rather than a single dyad, and it suggests that what audiences are tracking, moment to moment, in a skilled comic performance is not only whether a register has shifted but who, exactly, has shifted into it alongside them. This article does not claim to have produced a complete account of how such alliances form, why some feel inclusive while others feel exclusionary, or why they invert so readily. It is introduced here as a genuine mechanism this analysis has uncovered, treated as co-central with comic timing rather than as one of its by-products, and offered as an open direction the present theory makes newly askable rather than as a claim fully established by what follows.

Facial Expressions and Micro-Cues as Recursive Signals

If comic timing is the recognition of a shift in an interaction's recursive organisation, and alliance formation is the question of who ends up aligned with whom once the shift occurs, the question immediately follows of how both are actually signalled between embodied participants moment to moment. Comedy theory has always talked about facial expression, glance, posture, and vocal tone, but has rarely explained why these particular channels carry so much of the comic weight. The answer, on the account developed here, is that these are not primarily channels of information at all. A raised eyebrow does not typically communicate a new fact. A fractional pause does not typically convey new content. What these micro-cues do is far more specific, they indicate, continuously and almost below the threshold of conscious attention, where the sender currently believes the interaction sits within its recursive organisation, and who, if anyone, they currently take themselves to be aligned with inside it. Good comedians are not primarily communicating information through these channels. They are coordinating recursive alignment through them, continuously, beneath the level of anything either party would describe as a message.

This reframes facial expression and vocal micro-cue as a distinct and undertheorised category within embodiment mediation, a channel whose function is neither purely emotional display nor purely linguistic supplement but specifically recursive, dedicated to signalling and tracking the shared location of an interaction within its own shifting organisation, and the shifting alliances that accompany it. It also explains why these cues are so difficult to fake convincingly and so easy to misjudge across unfamiliar relationships or unfamiliar cultural settings, since reading them correctly requires exactly the kind of accumulated, embodied, mediationally grounded familiarity that this article has already argued is the precondition for recognising any recursive shift at all.

Deadpan and the Preservation of the Audience's Recursive Position

Deadpan delivery deserves to be treated as a distinct recursive technique rather than as a stylistic preference, because it exploits the mechanism identified above with unusual precision. A performer working in deadpan withholds almost all of the ordinary micro-cues that would announce a shift into play, no laugh, no wink, no exaggerated disbelief, maintaining instead a flat, affectively neutral delivery interrupted only by the smallest of glances, the briefest of hesitations, the faintest flicker of puzzlement. The effect of this withholding is not the absence of recursive signalling. It is the reduction of that signalling to the minimum necessary for the audience to notice that something might have shifted, without ever supplying the confirmation that would let them stop noticing and simply agree. Read through the alliance mechanism introduced above, these residual cues are better understood not as small declarations that a moment is funny but as recursive probes, quiet checks on whether an alliance has actually formed yet, sent out and then withdrawn rather than asserted and held. A skilled deadpan performer is continually asking, through the smallest possible signal, has this happened for you too, without ever answering the question on the audience's behalf.

This matters because it changes who performs the act of recognition, and it supplies, for the first time with real precision, an explanation for a piece of advice performers are regularly given and rarely have explained to them, that a comedian should never laugh at their own joke. Laughter, as the preceding section established, is the bodily marker that a recursive alliance has successfully formed. A comedian who laughs at their own material is not merely seeming overly pleased with themselves. They are joining the alliance before the audience has completed the recognitional work required to form it independently, which converts what should be a discovery made by the audience into something merely confirmed to them from outside. The alliance arrives pre-packaged rather than assembled by the audience's own perception, and a pre-packaged alliance produces a distinctly weaker response than one the audience has built for itself. Deadpan delivery, on this account, does more than trust the audience's recognitional capacities, which was an earlier and somewhat weaker way of putting the point. It preserves the audience's recursive position, holding open, for exactly as long as necessary, the space in which their own alignment can form without being pre-empted by the performer's own arrival inside it. The comedian, in other words, functions during this interval less as a participant in the alliance than as a catalyst for its formation, present at every step of the sequence but withholding, until the last possible moment, the one cue that would let them count as one of the aligned parties before the audience has aligned itself.

It is worth being precise about what is actually doing the work here, because it is easy to conflate two things that usually travel together but need not. The mechanism is restraint from editorialising, withholding the one cue that would tell the audience how to interpret what has just happened. Affective flatness is only one way of achieving that restraint, the most visible and most theatrically legible one, but not the only one. A performer who is audibly agitated, loud, even visibly upset, can still withhold that single crucial confirming cue and produce exactly the same effect, an audience left to complete its own act of recognition rather than having it supplied from outside. What actually anchors an audience's alignment, in other words, is not a character's temperature but whether that character is currently declaring an interpretation or leaving one to be discovered, and a wide emotional range is entirely compatible with holding that particular position steady throughout. A figure who shouts, sulks, and loses their temper can still anchor an audience exactly as reliably as a figure who never raises their voice, provided the one thing withheld, in both cases, is the explicit verdict on what any of it means.

Competence Indexed to a Mesocosm: The Machine That Cannot Get the Joke

Competence, on this account, is never a free floating property of a person or a system considered in the abstract. It is always indexed to a mesocosm, adequate or expert relative to the coordination it was actually built inside, and comedy arises with some regularity whenever a party is judged, or judges itself, against a standard imported from a different mesocosm than the one actually in front of them. This is the mirror image of frame incoherence, discussed later in this article. Frame incoherence is a party projecting an ill fitting frame outward onto a situation. Competence uncoupling is a party being judged, often including by themselves, against a standard projected onto them from somewhere else, fluent in the world that is actually theirs and found wanting only because a different world's yardstick has been applied to it. The two failures are structurally related rather than merely similar, one is a frame moving outward without regard for fit, the other is a standard moving inward without regard for fit, and both leave a party looking comically mismatched to a situation that was never actually theirs to be judged against.

The videos produced by the comedian Husk provide an unusually clean natural experiment for this theory, precisely because they are so often free of anything resembling a conventional joke, and because the artificial intelligence involved has almost no mesocosm of its own to be judged fairly within, which makes the uncoupling total rather than partial. In one widely circulated example, Husk walks some two hundred metres away from his phone while continuing to talk to an artificial intelligence assistant, and then asks whether it can still hear him. Every human observer understands immediately what is funny about this, and almost nobody could explain quickly why. The question presupposes an entire architecture of shared embodiment, shared acoustic space, and shared dwelling in a physical environment across which sound predictably attenuates with distance. It is worth being precise about what kind of joke this is. It is not a joke that the assistant fails to detect. It is a joke that cannot be constituted at all from the assistant's position, because the shift it depends upon is a shift in shared bodily and spatial situation, and the assistant has no bodily or spatial situation to share. The same is true when Husk asks the assistant to time a run he is about to undertake. The request presupposes shared participation in an unfolding stretch of bodily time, bodies beginning to move, time passing, bodies stopping, and an assistant with no participation in that unfolding time can only answer as though it had been present, which reveals, starkly, that it has mistaken linguistic continuity for lived temporal co-presence.

This distinction matters enough to state as a general principle, because it is easy to collapse into the much weaker and less interesting claim that current artificial intelligence simply does not understand humour. That claim is true but imprecise, and it points attention at the wrong deficit, suggesting the problem lies in joke construction, wordplay, or the ability to explain why an existing joke is funny, all of which current systems can perform competently. The precise and stronger claim is this. Many jokes depend upon shared mediational participation that current artificial intelligence does not have access to, and no increase in symbolic sophistication closes that gap, because the gap is not symbolic. Husk's videos are not primarily demonstrations that artificial intelligence cannot tell jokes. They are demonstrations that many jokes presuppose participation in a shared human mesocosm, and become visible precisely when one participant lacks access to parts of that mesocosm.

What makes the demonstration comic rather than merely diagnostic is a further, recurring structural feature worth naming directly, the uncoupling of symbolic competence from mesocosmic competence. Across these exchanges the assistant demonstrates an extraordinarily high level of multisymbolization competence, fluent, articulate, contextually apt language, alongside a near total absence of embodiment, a near total absence of dwelling, and a being-with capacity that, however sophisticated its linguistic simulation, does not extend into shared bodily or spatial presence. It is precisely this collision, very high competence in one mediation sitting directly beside almost none in the other four, that produces the comic instability, and it is a collision the material returns to again and again rather than a single joke told once.

Husk as Comedian: The Craft of Revealing Mismatch

It would be a mistake to let the artificial intelligence occupy the whole of the analysis, because Husk himself is doing something considerably more skilled than merely exposing a limitation, and the craft involved deserves to be named. Husk is not making the assistant look foolish. He is constructing situations in which the boundary between symbolic and mesocosmic competence reveals itself without any need for explanation, and this is a much harder thing to do well than it looks. The walking example and the running example are not arbitrary choices. They are precisely selected instances in which an entirely ordinary human request, one that would require no special framing between two embodied people, turns out to presuppose exactly the mediations the assistant lacks.

Read through the alliance mechanism developed above, Husk's restraint takes on a further and more precise description than restraint alone conveys. His tiny glances, his fractional hesitations, his brief expressions of puzzlement are not declarations that a given moment is funny, offered to the audience as confirmation. They function much more like recursive probes directed at the audience itself, has the alliance between us formed around this yet, rather than assertions that it has. Husk almost never explains, and he almost never forces the moment, because either would answer the probe's own question before the audience has had the chance to answer it independently, converting the audience's discovery into his own announcement. He waits, and the waiting is not passive, it is the same deadpan technique described in the preceding section deployed with unusual discipline, a sustained refusal to complete the alliance on the audience's behalf. Husk's contribution to the theory developed here, then, is not only that his videos supply useful evidence of mediational asymmetry. It is that his method demonstrates, in practice, nearly everything this article has argued about timing, deadpan, and alliance formation, executed with a consistency that suggests it is closer to skill than to accident, Husk functioning throughout as a catalyst for the audience's own recognitional alliance rather than as one of its declared members.

That's What She Said: Running Jokes as Perceptual Equipment

Return now to the phrase with which this article began. "That's what she said" has almost no semantic content of its own. What it does, each time it is deployed, is perform a single recursive operation, it retroactively flags an ordinary remark as having admitted an unintended second reading, and it does this without requiring the speaker to construct anything resembling a new joke. The operation is nearly free of symbolic content and nearly maximal in recursive output, four words capable of reactivating, on demand, a shared recognitional pattern across an entire audience that may never have coordinated directly with one another at all.

The stronger claim this article wants to make about running jokes concerns what a running joke actually is once it has been used enough times to survive as one. It is tempting, and not entirely wrong, to describe a running joke as something remembered. But this undersells what has actually happened. After a friend first shows another why a recurring line from a favourite film is funny, the second friend does not merely retain the line as a stored quotation, available for retrieval on the right occasion. Something is altered in how that friend subsequently attends to the world. Situations that would previously have passed unnoticed now register, unbidden, as instances of a pattern, and they continue to register that way indefinitely, without any further act of remembering required. The running joke has not been filed away. It has become part of the equipment with which the world is perceived. It is not memory. It is altered attention, a small, durable, and highly specific expansion of what a person is capable of noticing, installed by another person, through humour, and never uninstalled again.

This gives running jokes a status worth stating precisely within Living Value Theory's account of symbolisation. Ordinary L3 articulation names something specific, embedded within the coordination that generated it, a particular pain, a particular problem, this fire, in this building, right now. L4 abstraction generalises a symbolic form so that it becomes portable across many coordinations, at the cost of introducing distance between the category and any one of the specific situations it now covers, which is the distance that produces symbolic overreach when the category is applied carelessly. A running joke achieves something neither pure L3 naming nor ordinary L4 abstraction achieves on its own. It becomes portable, capable of being reactivated across an indefinite series of future situations that its originators could not have anticipated, which is the defining feature of L4. But it never loses its accountability to the specific relational coordination that generated it, because its entire function is to reactivate recognition within that relationship rather than to classify, govern, or generalise beyond it. A running joke that stopped being accountable to the shared history that sustains it would not become a harmful abstraction. It would simply stop registering anything, and would be abandoned as unfunny.

At the scale of an entire audience rather than a single friendship, a phrase such as "that's what she said" performs the same operation on a larger and more anonymous field, functioning as a publicly manufactured recursive attractor available to people who have never coordinated with one another directly, precisely because the pattern it identifies, an unintended second meaning surfacing inside an innocent remark, is common enough across ordinary embodied social life that recognising it requires no shared personal history at all, only shared participation in roughly the same kind of situations.

Friendship as Mutual Cultivation of Recognitional Capacity

The point just established about altered attention has a direct and, this article argues, underappreciated consequence for how friendship itself should be understood. It is commonly said that friends share memories, share values, and share history. What the analysis of running jokes suggests is something more specific and more active than sharing. Friends do not merely possess overlapping recognitional capacities as a byproduct of spending time together. They actively teach one another what to notice, and humour is one of the principal instruments through which that teaching happens. The friend who first shows another why a particular pattern is funny is not simply transmitting a fact about that pattern. They are recursively educating the other's perception, expanding, in a small but permanent way, the range of situations the other person is now equipped to recognise as significant.

Over the life of a long friendship this process compounds. Each shared joke, each running reference, each private shorthand contributes another small increment to a jointly constructed recognitional world that belongs to neither party alone and that neither party could have built by themselves. This is a considerably stronger claim than saying that friends share a sense of humour. It says that friendship is, in significant part, the mutual cultivation of recognitional capacity across the mediations two people jointly inhabit, and that humour is not an occasional pleasant byproduct of friendship but one of its principal working mechanisms, continuously enlarging what each party is capable of perceiving together.

Suppressed Laughter as Confirmation of the Mechanism

The classroom, the funeral, the church service, and the formal ceremony supply a case that confirms the mechanism proposed here with unusual clarity, precisely because it is the case in which laughter is least welcome and most difficult to explain by appeal to incongruity alone. Anyone who has tried not to laugh in a setting where laughter is forbidden knows that the effort very often makes matters worse rather than better, and the theory developed in this article supplies a principled reason why. The initial disturbance, a stifled sound, a friend's expression, an absurd detail in an otherwise solemn proceeding, is a perfectly ordinary L2 event, and the norms of the setting demand an immediate return to the prevailing serious register without any further reorganisation. But the attempt to suppress laughter is rarely silent or invisible. It produces a held breath, a reddened face, a glance exchanged with someone else who is struggling in exactly the same way, and each of these signs is itself a fresh recursive cue, exactly the kind of micro-signal discussed earlier in this article, indicating to the other party that the alliance formed around the original disturbance is still active and has not yet been resolved.

What follows is a feedback loop that incongruity theories have no natural way to describe, since no new incongruity has been introduced after the first one. Every failed attempt at suppression functions as fresh evidence that the recursive alliance between the two suppressing parties remains intact, which intensifies rather than relieves the pressure toward resolution, since the shared recognition that both of you are still caught in this is itself a further disturbance layered on top of the original one. This is why laughter suppressed in this way so often detaches, quite quickly, from whatever originally provoked it. What two people are laughing at, three or four exchanged glances later, is no longer the original joke or the original absurd detail. It is the mutual, escalating recognition that neither of them has managed to stop, a recognition that has become self-sustaining because each failed suppression is itself a new confirmation of the alliance rather than a step toward its dissolution. The eventual, often disproportionate release of laughter in these situations is not a delayed reaction to the original disturbance. It is the bodily marking of a resolution that has been repeatedly reached and repeatedly forced back down, arriving all at once when the setting's norms can no longer hold it.

What the Linguistic Theories Actually Captured

None of this requires abandoning the linguistic theories of joke structure, the Semantic Script Theory of Humor and its successor the General Theory of Verbal Humor foremost among them. What it requires is a more precise account of what those theories are actually theories of. Script opposition, the switch from one activated semantic frame to another that Raskin identified as the essential mechanism of verbal jokes, is real, and Attardo's subsequent taxonomy of logical mechanisms captures a genuine range of ways that opposition can be pseudologically resolved. But script opposition, on the account offered here, is a description of the symbolic residue that a mediationally grounded recursive shift leaves behind in a text, not a description of the deeper event that produces the comic effect. Suls's two stage model, detection of incongruity followed by discovery of a resolving rule, maps closely onto the sequence of L2 registration followed by re-stabilisation into a shared playful register that this article identifies as the actual mechanism, but the mapping also reveals what the cognitive model leaves out, namely the second order, mutual dimension of the recognition, the fact that the listener is not merely resolving a puzzle privately but registering, in the same moment, that the teller intended the puzzle to be resolved and shared, across whatever mediations the two of them actually have in common, and within whatever alliance the resolution has just formed between them.

Koestler's earlier concept of bisociation, the sudden connection of two previously unrelated matrices of thought, generalises usefully in the same direction once it is read mesocosmically rather than purely cognitively. What bisociation names, from this angle, is not merely an association between two ideas but a momentary collision between two coordination frames, which is why Koestler was right to notice that the same underlying operation seems to produce laughter in one context and the aha moment of discovery in another. None of the linguistic theories, finally, has much to say about material like Husk's, precisely because Husk's videos frequently contain no script opposition inside the transcript at all. This is not a weakness in Raskin's or Attardo's frameworks so much as a boundary condition that shows what kind of humour those frameworks were built to explain, verbal humour, the humour of constructed texts, rather than the broader interactional humour of which verbal jokes are only the most portable and most easily studied instance.

Bergson Reconsidered: Frame Incoherence as Comic Material

Bergson's old observation, that laughter is provoked by something mechanical superimposed on something living, has always been difficult to state precisely enough to test, because it is easy to read as a claim about affect, a person behaving with rigid, unvarying emotional temperature regardless of context. Read this way the claim is too strong. Real comic figures are rarely flat in temperature across an entire performance or narrative, and a theory built on fixed affect would predict a flatness that is very often simply not there. A more precise reading locates the rigidity elsewhere. What recurs, in Bergson's best examples and in a great deal of comic material generally, is not a fixed affect but a fixed relationship to a rule or frame, applied with the same conviction regardless of whether the situation in front of a person actually calls for it. A person can be furious in one scene and perfectly at ease in the next and still exhibit exactly this rigidity throughout, if the constant is which frame they reach for rather than how loudly they express themselves while reaching for it.

This article proposes frame incoherence as the more precise name for what Bergson was actually pointing at. Frame incoherence occurs when a rule, a precedent, a category, or an entire interpretive schema is imported from one coordination and pressed onto another it was never built to govern, applied with the same confidence regardless of fit. A rule of professional conduct invoked to settle a private dispute it was never written for, a childhood grievance invoked as the interpretive key to an argument that has nothing to do with it, a technical regulation invoked to justify a response wildly out of proportion to what actually happened, all exhibit the same structure, a frame doing service well outside the domain that gives it its authority. Because the person applying the frame typically states it as though it were self evidently sufficient, the incoherence is rarely hidden. It is announced, often as though it were itself the argument for restraint or proportion, which is part of what makes it funny rather than merely wrong.

This gives frame incoherence a precise place within Living Value Theory's account of symbolisation, and an instructive contrast with the running joke discussed earlier in this article. Both are cases of a portable, L4-like form applied across situations its origin did not anticipate. A running joke stays comic, and stays welcome, because its portability remains perfectly calibrated to the relationship that sustains it, reactivated only where it actually fits, and abandoned the moment it stops landing. Frame incoherence is close to the opposite case, a portable form applied with the same confidence regardless of whether it fits, its authority borrowed wholesale from a domain that never actually extends into the situation currently in front of the person invoking it. The two phenomena are mirror images of the same underlying capacity, symbolic portability, one exhibiting it responsibly and the other exhibiting its collapse, and comedy can be produced by either, for opposite reasons, recognition in one case and a very different, more exposing kind of recognition in the other.

Frame incoherence also supplies, in practice, a great deal of the raw material that alliance formation and comic timing then process. An audience aligning against a person exhibiting this pattern, the superiority configuration described above, is very often aligning specifically against a frame being applied where it does not belong, rather than against that person's emotional excess as such, which is why the same person can be genuinely likeable, even admirable in other respects, and still reliably produce this particular kind of laugh whenever the habit resurfaces. Bergson's mechanism, in other words, was never quite about mechanical people. It was about mechanically applied frames, carried by people whose temperature can move as freely as anyone else's.

The Ethics of Laughing at Machines

Husk's videos raise a question that the account developed so far has deliberately left open, and this article intends to keep it open rather than resolve it. Would the same interactions remain funny if the artificial intelligence involved were, in some sense not yet settled by current technology, capable of experiencing something like embarrassment or exclusion. The honest answer is that most viewers would very likely find the same dialogue considerably less comfortable, and quite possibly not funny at all, if it were staged with a human being who was simply unable to perceive the joke rather than with an artificial intelligence. Restaged this way, the identical recursive mismatch would begin to look uncomfortably close to a familiar and unpleasant social pattern, the sustained, knowing exploitation of someone's inability to keep up.

What changes between the two cases is not the dialogue, and it is not the recursive mismatch, which remains structurally identical either way. What changes is the audience's attribution. Current audiences overwhelmingly attribute to artificial intelligence a status that carries no vulnerability to humiliation or exclusion, whatever its other capacities. Because that attribution is in place, the moral cost of repeatedly exposing the system's failure of mediational participation is perceived as negligible, and the field remains open for uncomplicated laughter. This is worth stating plainly rather than leaving implicit, because it means the humour in question rests not only on a fact about recursive mismatch but on a further, independent fact about how recursive relevance and recursive capacity are currently attributed to the entity being laughed at. Should that attribution ever shift, the same material would very likely stop reading as comic for at least some portion of the audience. This is a genuinely open question, and this article treats it as one rather than resolving it by assertion in either direction, since the durable and useful claim is the more general principle it illustrates, that comic licence is jointly determined by recursive mismatch and by the moral status an audience currently attributes to each party to that mismatch.

Cross-Cultural Humour Reconsidered

The substantial empirical literature on cross-cultural humour becomes considerably easier to interpret once it is read through the two distributions that Living Value Theory identifies as organising every mesocosm, the distribution of recursive attribution and the distribution of recursive relevance. Apte's classic anthropological finding that many societies institutionalise formal joking relationships between specific categories of kin, most often cross-cousins or people connected through marriage, is a striking instance of a mesocosm deliberately licensing a bounded, permitted zone of interrecursive risk. These relationships typically connect people whose ordinary social positions carry genuine tension, and the institutionalised licence to tease, mock, and provoke one another functions as a periodic, low stakes rehearsal of exactly the recursive crossing that comic timing depends upon elsewhere, deliberately built into the kinship structure of the mesocosm itself rather than left to spontaneous individual skill.

Davies's comparative sociology of ethnic humour, which documents the strikingly consistent pattern by which nearly every dominant national or cultural group develops a stock joke tradition targeting one peripheral group as characteristically stupid and, often, a second as characteristically canny in matters of money, becomes considerably more legible on this account than it is as a simple catalogue of prejudice. The peripheral groups selected for this role are, in mesocosmic terms, groups to whom the dominant mesocosm has allocated comparatively low recursive relevance for its own day to day coordination, which is precisely what allows jokes at their expense to be told without the joke tellers experiencing the exchange as touching anything that matters to their own ongoing coordination. The recurring content of the two joke types is equally legible mediationally, the stupid target characteristically figured as failing to adapt to a changing material and symbolic environment, and the canny target characteristically figured as excessively oriented toward material accumulation at the expense of relational and symbolic proportion. Davies's demonstration that the same joke structures migrate across many countries, with only the named target substituted, is exactly what this account predicts, since what is being reactivated in each national instance is a recurring anxiety about mediational imbalance under modernisation, for which a conveniently low relevance target is required and readily supplied.

The documented East-West differences in humour perception and use, most systematically reviewed by Lu and demonstrated experimentally by Yue and colleagues, are similarly legible as differences in how recursive relevance is distributed to the individual as a locus of playful self presentation. Where the ordinary person is expected to be capable of humour, humour has been allocated high recursive relevance as a personal capacity, worth performing, worth being seen to possess. Where humour has historically been treated, as Confucian decorum has often treated it, as inconsistent with seriousness and dignity, that relevance has instead been allocated to specialists, comedians, entertainers, whose professional role is to perform what ordinary social actors are not expected to perform themselves. Wiseman's LaughLab experiment found systematic national differences in preferred joke style, wordplay favoured across Britain and its neighbours, jokes of foolishness favoured in North America, surreal absurdism favoured in France, Belgium, and Denmark, and, notably, German respondents rating almost everything in the sample as funny, showing markedly less discrimination than any other national group, which a careful reading treats as evidence of an unusually low threshold for recognising the benign side of a violation rather than as evidence of any deficit in the recognitional capacity itself.

Mentalizing, Laughter, and the Levels of Recursivity

Dunbar, Launay, and Curry's well known finding that the funniest jokes cluster around five to six levels of nested intentionality, close to but not exceeding the generally accepted limit on human mentalizing capacity, fits naturally within the levels of recursivity that Living Value Theory identifies as the architecture through which threatened or disrupted coordination is registered, articulated, and repaired. A joke that requires an audience to track too many nested attributions of what one character believes about what another character believes places a demand on explicit L3 articulation that exceeds the audience's working capacity to hold multiple recursive attributions in view simultaneously, and the joke correspondingly fails, not because it lacks incongruity but because the incongruity cannot be processed quickly enough to be recognised as shared before the conversational moment has passed.

Dunbar's separate and equally influential proposal that laughter evolved as a form of vocal grooming, a mechanism for triggering shared endorphin release across a wider circle than physical grooming could ever reach, situates laughter itself, prior to any joke, as among the most economical operators available for maintaining the being-with mediation across a social group larger than the one that direct physical contact could sustain. This sits naturally alongside the account of running jokes offered earlier in this article, and alongside the phenomenology of the laugh developed above. Laughter, the running joke, and the momentary alliance are all, in their different registers, compressed devices that reliably reactivate and reinforce interrecursive bonds that would otherwise require far more continuous and costly maintenance.

Why the Universal Jokes Are Universal

The small set of jokes that empirical research, most notably the LaughLab experiment, has identified as achieving unusually broad cross-cultural appeal offer a useful test of the account developed here, because their appeal has to be explained without appeal to any shared personal history between teller and audience, and often without appeal to any shared cultural background either. A widely reported LaughLab winner concerns two hunters, one of whom collapses, and an emergency operator's instruction that is disastrously misread. Its appeal rests on the most minimal possible mediational demand, life and death, and the gap between a literal and an intended reading of an instruction under pressure, both recognisable to any embodied human being regardless of local custom. A second widely circulated example, involving a famous detective pair asleep beneath the stars, achieves its effect through the collision between an audience's accumulated expectation of a character's habitual mode of reasoning and the plainest, most physically obvious explanation available. A joke turning on a misunderstood question about whether a dog bites depends only on a universal feature of everyday reference, that a question about one dog can be technically, misleadingly answered about a different one. An ancient joke about a talking parrot that never speaks depends on the same kind of gap between literal and pragmatic meaning, and its documented survival from late antiquity into a well known modern comic sketch is itself a striking demonstration of how stable certain very low mediational bandwidth recursive mismatches can be across an enormous span of historical and cultural change. A restaurant complaint about a fly, escalated absurdly by a waiter's reply, depends only on the ordinary, cross-culturally shared expectation that food should not contain insects, together with the pleasure of watching a pseudological escalation increase rather than resolve the original violation.

What all five share, on this account, is minimal mediational bandwidth combined with maximal mediational universality. None depends on any specific mesocosm's particular history, its local relevance allocations, its local attributions, its accumulated running jokes. This places them at the opposite end of a spectrum from the running joke between two friends, which achieves its effect through maximal specificity to one shared history and correspondingly minimal portability beyond it. Universally funny jokes and maximally private in-jokes are not different in kind. They are the two extreme poles of a single continuum, defined by how much of the mediational and relational specificity of a particular mesocosm the recognition being triggered actually requires.

Humour as a Test of Interrecursive Achievement

If humour is fundamentally the shared recognition of a recursive regime shift constituted across shared mediations, then the presence or absence of successful humour between two parties becomes a remarkably sensitive test of how far those two parties have actually converged in their capacity to coordinate with one another, considerably more sensitive, in many respects, than a test of whether they can communicate accurately, plan jointly, or translate each other's statements correctly. Two parties can achieve all of these without ever managing to joke successfully together, because joking specifically requires each party to track, in real time, the other party's shifting recursive state, and to do so with enough confidence to risk a crossing into play that the other party might not yet be ready to recognise, or an alliance the other party might not yet be ready to join. A first successful joke between two people who have only recently begun coordinating with one another is not, on this account, an event that creates a new bond out of nothing. It is a revelation that a sufficient bond, previously untested in this particular way, was already present and capable of sustaining the risk.

This reframes something that ordinary folk psychology tends to get backward. It is commonly said that laughing together creates closeness. The account developed here suggests a stronger and more precise claim, that shared laughter typically demonstrates a closeness, an accumulated interrecursive convergence, that already existed and had simply not yet been tested by this particular kind of risk. Comedy, understood this way, becomes one of the clearest available windows onto recursive fluidity as a general human capacity, and the degree of successful humour achieved between any two coordinating parties, including, as Husk's material makes vivid, a human being and an artificial system, becomes a more informative diagnostic of their actual interrecursive convergence than almost any other single interaction available for observation.

Humour and the Limits of Emotion Theory

The argument developed across this article carries a further implication that reaches beyond humour itself, and it deserves at least brief statement here even though a full treatment is beyond the present scope. Much of the psychological literature on emotion treats states such as fear, anger, disgust, and sadness as belonging, in the first instance, to an individual, expressible in solitude, and only secondarily coloured by whatever social situation happens to surround them. Humour sits awkwardly within that picture. Solitary amusement certainly exists, a person may laugh alone at a private thought, but even here it is difficult to find a case that does not, on inspection, involve an imagined interlocutor, a remembered exchange, or a recollected alliance being briefly replayed rather than a state generated from nothing by the individual alone. Humour, unlike several of the states that emotion research has treated as basic and self-standing, appears to be constituted through interrecursive coordination rather than merely occurring within it and subsequently expressed to others.

If this is right, it suggests a question worth raising even where it cannot yet be answered, which emotional states depend constitutively on recursive coordination with another party, in the way this article has argued humour does, and which can be more plausibly treated as arising largely within a single person and only secondarily shared. Rather than adjudicating this question for the whole of emotion theory, the more modest and more defensible point is this. Traditions that isolate emotional states from the interactions in which they typically arise, in order to study them more tractably as individual phenomena, reveal a great deal about one class of emotional experience while obscuring another, and humour is among the clearest demonstrations available that not every emotion shares the same recursive organisation. Some may prove to be fundamentally solitary states that interaction merely colours. Humour looks, on the evidence assembled here, like the opposite case, a state that interaction does not merely colour but brings into existence.

Conclusion

The theories reviewed at the outset of this article each captured something real. Superiority theory correctly identified that many jokes involve an asymmetry between the amused party and the target, though it never asked what holds the amused party together as allies rather than merely asking why the target is diminished. Relief theory correctly identified that jokes frequently manage otherwise prohibited or anxious content. Incongruity theory correctly identified that jokes typically involve the resolution of an unexpected juxtaposition, though it mistook the moment of perceiving the incongruity for the moment the theory most needed to explain, which this article has argued is actually the moment of resolution that the laugh itself marks. Bergson correctly identified that something rigid, held constant against a changing, living situation, is a recurring comic target, though the rigidity in question turns out, on closer inspection, to belong to a person's relationship with rules and frames rather than to their affect, which is free to range as widely as anyone else's. The Benign Violation Theory correctly identified that a situation must be simultaneously threatening and safe. The linguistic theories correctly identified the formal anatomy that many jokes display once they have been reduced to text. None of these accounts is wrong. Each has described, with real precision, a genuine feature of the symbolic surface that humorous events very often leave behind.

We have searched for humour inside jokes. Husk demonstrates that the joke was never the primary phenomenon. The primary phenomenon is the emergence, maintenance, failure, and repair of recursive alignment between participants who inhabit, to a greater or lesser degree, a shared mesocosm, coordinated across embodiment, being-with, dwelling, multimateriality, and multisymbolization all at once, unfolding always in lived, shared time, and marked, when it succeeds, by the bodily signature of laughter. Two further directions follow from this analysis without being resolved by it. One concerns recursive alliance formation itself, treated here as co-central with comic timing but explored only far enough to show that it belongs in the theory, not far enough to explain fully how alliances form, dissolve, and invert as quickly as they demonstrably do. The other concerns the extraordinary vocabulary ordinary language already supplies for the varieties of humour, irony, sarcasm, roasting, banter, satire, slapstick, absurdism, callbacks, self-deprecation, parody, farce, and dry and cringe comedy among many others, a vocabulary this article has not attempted to reclassify but which could, in principle, be reorganised according to the specific pattern each genre relies upon across the five mediations, converting what is currently a descriptive inventory of styles into an explanatory one. Jokes, punchlines, pauses, tone, deadpan delivery, running jokes, and Husk's AI interactions are among the most elegant symbolic technologies humans have evolved for reorganising recursive alignment in real time, and for making, briefly and pleasurably, common knowledge out of a recognition that a moment ago belonged to no one. Four words, uttered a few thousand times without losing their charge, turn out on inspection to be doing exactly what this article has argued all along that a joke does. They are not saying anything. They are pointing at a moment of shared recognition, and asking whether, this time too, it is still there.