Ordinary verbs perform extreme selective compression of multimediated living process. They articulate only a narrow fraction of any occurrence and thereby render recursive coordination possible. Symbolic overreach arises when the stability introduced by that compression is later forgotten and treated as a property of the process itself. The clearest instance is the shift from verbal predication to nominal construal. A fluid process becomes a portable noun such as knowledge or life. That noun then appears as an independent entity or philosophical problem. The resulting amnesia distinguishes symbolic overreach from ordinary reification and explains why even process oriented philosophies repeatedly return to static terms. Much of 20th century philosophy, including Wittgenstein and Heidegger’s, fell prey to the symbolic overreach of verbs turned into nouns.
I. Introduction: Why Verbs Are Interesting
Verbs might seem, at first glance, an unlikely place to look for anything new. Living Value Theory has spent a great deal of effort on nouns: on classifications, categories, sensory terms, and the ways recursive life sediments into stable symbolic ground. Verbs have mostly waited in the background, treated as the grammatical category naturally suited to a theory built on process rather than substance. If reality consists not primarily of stable things but of recursively unfolding living process, then verbs, which name doing rather than being, might seem to come closer to that ontology than nouns ever could.
The intuition has an obvious appeal. Adam, in Genesis, can name thousands of animals one after another, each requiring its own word, while the act he performs each time remains a single process: to name. A person can walk across innumerable roads, through innumerable landscapes, in innumerable weathers, and the verb that covers all of it, to walk, never needs to multiply the way the nouns for terrain, footwear, and destination do. Hundreds of thousands of entities invite their own lexical differentiation, while a comparatively small set of recurrent processes appears able to organize an enormous range of what those entities do.
This article resists the tempting conclusion that follows from that observation, that verbs simply correspond to processes while nouns correspond to things. Nouns regularly denote events, an arrival, a collapse, a departure, and verbs regularly denote comparatively stable states, to remain, to belong, to know. Grammatical category and ontological category are not the same architecture, a point linguistic typology has established in its own right (Hopper and Thompson 1984), and mapping one onto the other produces exactly the kind of premature classification a theory of recursive mediation should distrust.
The question worth asking instead is what particular symbolic work verbs perform when they articulate a process at all. Pursuing that question across the sections that follow leads to nine related discoveries. Verbs turn out to be extreme process compressions, articulating only a fraction of what any actual occurrence contains. The mediational bandwidth they amplify appears remarkably narrow on inspection, usually one mediation and rarely more than two, though this pattern is treated throughout as a hypothesis to be tested rather than a demonstrated result. They generally do not specify at what recursive level an event was lived, even as they differ considerably in recursive valency, the range of recursive statuses that can occupy their participant positions. Their participant structures can flatten the difference between a nonrecursive and an interrecursive occupant of the same grammatical role, disguising a change in kind as a change in degree. The strictly universal core of the verbal lexicon may be far smaller than intuition suggests, a question this article distinguishes carefully from the separate question of how many nouns a language needs for its entities, and the differences between languages are better described as differing compression profiles than as differences in richness. Nominal construal, the operation that gives a process a stabilized noun to be handled by, shows with unusual clarity how the stability introduced by symbolic articulation of a process can be mistaken for a property of the process itself, and philosophy turns out to repeat this mistake often enough that it is worth treating as one of its most characteristic habits.
The aim of what follows is not to produce another taxonomy of verbs alongside the many linguistics already offers. It is to use verbs as an unusually clear site for watching multisymbolization act on a process world, compressing what it touches, and then, when the compression is forgotten, mistaking the compressed symbol for the world it was compressed from.
II. What Verbs Actually Classify
Linguistics has long resisted the idea that there exists a simple inventory of natural actions which languages merely go on to label. Different traditions classify verbs along quite different dimensions: agency (Davidson 1963; Anscombe 1957), event structure and temporal profile (Vendler 1957; Moens and Steedman 1988), causation, the number and kind of participants a verb allows (Levin 1993), manner and result (Rappaport Hovav and Levin 2010; Beavers and Koontz-Garboden 2012), and the sheer granularity at which an event is cut up for naming (Zacks, Tversky, and Iyer 2001). None of these dimensions delivers a single correct partition of process into verb-sized units, because there is no such partition waiting to be discovered independently of the classifying activity itself. The present article brackets the first of these dimensions almost entirely. Whether an occurrence is something an agent does or something that merely happens to them is a substantial question in its own right, and nothing in the argument that follows depends on resolving it, since the compression this article is concerned with applies equally to intentional and non-intentional predicates alike.
A simple case makes the point. To run can name an open-ended activity, as in he ran, with no implied endpoint, what Vendler's classification calls an activity (Vendler 1957). The same verb changes its event structure once a goal or a distance is added: he ran a mile, he ran to the gate, shifting the predicate toward what Vendler calls an accomplishment and illustrating the more general point, developed at length by Pustejovsky, that event type is a property of the predicate in its clause rather than of the bare verb (Pustejovsky 1991). The physical episode being described has not changed; what changes is the profile the construction imposes on it. The same holds for events that can be segmented at several levels at once. Moving a hand, holding a knife, chopping vegetables, making soup, cooking dinner, and preparing a meal can all describe the same stretch of activity, sliced at different grains, an instance of the hierarchical structure independently documented in experimental research on how people segment and describe ongoing activity (Zacks, Tversky, and Iyer 2001), and none of them is the uniquely correct verb for what happened.
This suggests a distinction worth stating explicitly: the occurrence itself is one thing, and the verb-mediated profile of the occurrence is another. A verb does not retrieve an event that was already packaged by the world and waiting to be picked up. It makes one particular processual articulation of that event salient, at the expense of every other articulation the same stretch of process would equally have supported.
Call this a process specification. A verb identifies some recurrent organization of change, activity, state, causation, relation, or trajectory that can be found across otherwise heterogeneous occurrences. To stumble is already an extraordinary abstraction in this sense. No two stumbles share identical bodily dynamics, the angle of the ankle, the distribution of weight, the speed of recovery all vary from case to case, and yet all of them can be gathered under a single verb as instances of stumbling. To hammer abstracts across an equally enormous range of material variation, from the weight and balance of the tool to the hardness of the surface being struck. To name abstracts across radically different kinds of entities and circumstances, covering the naming of a child, a ship, a hurricane, and a hypothesis with the same verb.
Existing lexical resources pursue a related but distinct project at a different level of granularity, and the boundary is worth marking clearly. FrameNet organizes verb senses around the structured situations they evoke (Baker, Fillmore, and Lowe 1998), VerbNet links thematic roles and syntactic frames to broad classes of verbs (Kipper-Schuler 2005), and WordNet organizes verbs primarily by sense and synonymy (Miller 1995). None of these resources competes with the argument developed here. A process specification in the sense intended above is compatible with any of several classificatory schemes built on top of it. What follows is concerned with what compression a verb achieves in articulating process at all, not with which lexical family, frame, or synonym set a given verb sense belongs to once that compression has already occurred.
This is precisely why verbs matter for a theory built on process. Their importance lies not in mirroring process faithfully, as though a verb were a small, accurate replica of what it describes, but in showing, with unusual clarity, how multisymbolization makes an otherwise unmanageable process world recursively tractable. A verb takes a configuration of change that would resist any exhaustive description and turns it into something that can be uttered, repeated, negated, combined with other verbs, and applied to a new case without having to be reinvented each time.
This yields the first caution the rest of the article will return to repeatedly. A verb is a stabilized symbolic means for articulating process, and not the process itself. A theory of recursive mediation should resist a temptation that might be called verbism just as strongly as it resists the substantialism that treats nouns as though they simply named stable things waiting to be discovered. What it gains, once both temptations are set aside, is a clearer view of what verbs actually do: they compress, and the compression is the achievement worth examining, not an approximation to be measured against some more accurate description that ordinary language merely failed to produce.
One distinction is worth making explicit before the argument proceeds further, since later sections depend on it directly. Using a verb in ordinary discourse is itself a symbolic articulation at L3 or above, since any utterance of a verb is already multisymbolic activity. This does not imply that the process being articulated itself occurred at L3. A verb may articulate an L1 bodily remediation that required no separate registration at all, an L2 felt disturbance that had not yet been put into words, an L3 articulation performed for the first time in the exchange in which it is uttered, an L4 institutional operation drawing on an established category, or an L5 repositioning of such a category. The level of the articulation and the level of the articulated process must therefore be kept separate. The corollary is worth adding immediately: a verb sitting in a dictionary is not itself performing L3. It is a sedimented symbolic trace, a poietic deposit like any other, available for future articulation whenever a living occurrence calls it back into use.
III. Ordinary Language as Extreme Recursive Compression
Consider an ordinary utterance: he is walking. Ask what the sentence preserves of the occurrence it describes, and what it lets go.
The utterance does not specify his speed, his gait, the terrain underfoot, its gradient, what he is wearing on his feet, the temperature of the air, what fills his visual field, the sounds around him, the proprioceptive sense he has of his own movement, the muscular effort involved, where he is going, who if anyone is with him, the social situation he is walking through, where his attention is directed, what he remembers or expects, why he is walking at all, his mood, whether the walking is routine or unusual, whether it hurts, whether he knows he is being observed, or whether he is consciously monitoring his own gait as he goes. The actual occurrence, whatever it is, will contain many of these simultaneously, layered into a single lived stretch of embodied activity, and the sentence contains almost none of them while remaining, for most ordinary purposes, a completely adequate thing to have said.
This shows why the common complaint that language fails to capture reality is only half right. Ordinary language is not attempting a total description that then falls short but performing a relevance-sensitive compression that was never aiming at totality in the first place. Ordinary language compresses the multimediated fluidity of living process down to the distinctions relevant for present recursive coordination, and no further. If the question at hand is where is Peter, the answer he is walking home may contain exactly what the coordination requires, and supplying the precise pressure distribution under his left foot as he walks would not make the utterance more accurate in any sense that mattered to the coordination it was serving. It would only interfere with it, adding noise where the exchange needed signal. The dimensions a compressed utterance leaves out are not simply discarded at random. They are dimensions that have not, for the coordination currently under way, become recursively relevant, which is why compression is better understood as recursive relevance enacted symbolically than as reduction for its own sake.
The same asymmetry appears in embodiment itself. A living body does not see, and then separately smell, and then separately hear, and then separately feel, in the serial way a sentence must lay these dimensions out one after another, since embodied experience is multisensorially integrated at the moment it occurs. Language can approximate the content of that integration only by stringing its components together in sequence: I smelled smoke, felt heat, heard crackling, and saw flames. Even a sentence built from four such clauses, one for each sense, remains four serial articulations rather than one integrated event, and their accumulation does not recreate the simultaneity that made the original occurrence what it was. L3 gains its remarkable articulatory precision by sacrificing exactly this simultaneity, trading the integration that belongs to lived process for the seriality that symbolic mediation requires.
This claim should not be overstated into a claim about the capacities of symbolic mediation in general. Sign languages exploit simultaneous manual and non-manual channels to encode several grammatical and semantic relations at once. Prosody, gesture, and facial expression accompany speech concurrently rather than in strict sequence. Diagrams, maps, and written layouts can present several relations at a glance rather than one after another. None of this restores the simultaneity of the living occurrence itself. What these cases show is that symbolic mediation is not mathematically incapable of simultaneity, only that even the most multimodal articulation preserves a minute fraction of the simultaneous bandwidth present in the occurrence it articulates. The crucial claim is a relative one. Symbolic resolution remains vastly below process resolution, however many channels a given articulation manages to coordinate at once, and it ordinarily achieves tractability through selective decomposition rather than through any strict impossibility of doing otherwise.
The same trade-off extends beyond the senses to all five mediations at once. A living occurrence can be embodied, social, material, spatially situated, and symbolically laden all at the same time, each mediation fully present and none of them waiting its turn. Symbolic articulation can move across these dimensions, naming first one and then another, and even the richest multimodal act, gesture accompanying speech accompanying a shared material setting, holds nothing like the fullness with which the occurrence itself keeps all five open together.
This produces a fundamental asymmetry worth stating on its own terms, even once the multimodal qualification above is taken into account. Living process is simultaneous and multimediated at a bandwidth no symbolic system, single-channel or multimodal, comes close to matching, and no amount of elaboration closes this gap, because elaboration only ever adds further selective articulations rather than converting resolution into the integration the occurrence itself possessed. A thousand-page description of a single walk remains, however long, an ordered symbolic traversal of aspects of that walk, taken up one at a time and set down in sequence, and it never becomes the integrative mode of the original occurrence, no matter how many further pages are added.
The trade-off can be put simply: loss of process bandwidth produces a gain in recursive tractability. This is one of the conditions of ordinary language being usable at all rather than an accidental limitation that better language could someday overcome. A symbolic system that tried to preserve the full multimediated simultaneity of the occurrences it named would not be a more accurate language but would fail to be usable as language at all, since nothing recursively tractable would be left of what it produced.
IV. Mediational Bandwidth: How Much of an Event Can a Verb Carry?
The compression argument can now be translated directly into the five mediations Living Value Theory distinguishes: multisensorial embodiment, being-with, multiversal dwelling, multimaterial forming, and multisymbolism. Shorthand forms are used from here on, embodiment, dwelling, multimateriality, and multisymbolization for the corresponding activity of symbolic deposition, while multisymbolism remains the name of the mediation itself. Doing so requires separating two things that are easy to run together, the mediational profile of an event and the mediational profile of the verb used to describe it.
An event can recruit all five mediations strongly at once. The verb that names it ordinarily does not. Take cook. An actual episode of cooking can involve embodiment in the full sense, taste, smell, heat on the skin, the proprioceptive sense of stirring and lifting, being-with in the form of cooking for or alongside others, multimateriality in the transformation of ingredients through tools and heat, dwelling in the specific organization of a kitchen and a domestic space, and multisymbolism in the recipes, quantities, culinary categories, and ritual prescriptions that shape what counts as cooking correctly in a given household. The verb cook carries almost none of this configuration into the sentence that uses it, its default lexical profile centred on multimateriality, the transformation of ingredients, with embodiment sufficiently prominent in most uses to count as a second amplified mediation, and the remaining three mediations left almost entirely unmarked.
A narrower verb makes the contrast sharper still. Hammer strongly profiles multimateriality alone. Not even a body is required, a lot of machines can do the hammering. A verb like table, in a use such as table the proposal, is instructive for a different reason, since a verb historically derived from a material object can undergo symbolic extension until its current mediational profile, centred on multisymbolism and institutional procedure, has almost nothing left in common with the multimaterial object it once denoted. Add to this that for the longest time of our existence as Homo sapiens, no one used a table.
Further cases confirm the pattern. Walk amplifies embodiment, and sometimes dwelling where terrain is at stake. Teach amplifies being-with together with multisymbolism. Write amplifies multimateriality together with multisymbolism. Visit amplifies being-with together with dwelling. Travel amplifies embodiment together with dwelling. Promise amplifies being-with together with multisymbolism. In each of these cases, at most two mediations appear to be strongly carried by the verb itself, however many more mediations the actual event they describe might involve. The list is small and deliberately illustrative rather than a systematic sample, and the pattern it suggests should be read accordingly.
The term amplified needs a definition of its own at this point, since without one a critic could keep adding mediations to any verb's profile indefinitely. A mediation counts as amplified by a verb sense when removing it would substantially alter the verb's conventional semantic profile, as distinct from a mediation that is merely required for the event to occur but not itself profiled by the verb. A body is necessary for any ordinary act of hammering, since something has to swing the tool, but this embodied requirement is enabling rather than amplified: change the swing and the description of what happened changes, but what hammer conventionally profiles, the transformation worked on the material, does not. Cook, by contrast, loses a substantial part of its conventional sense once the sensory and gustatory dimension of embodiment is subtracted from it, which is why embodiment counts as a second amplified mediation for cook and only an enabling condition for hammer.
This suggests mediational bandwidth as a name for the number of mediations a verb amplifies simultaneously in a given use, on the working definition of amplification just given. The strong hypothesis worth testing, rather than assuming, is that most ordinary verbs amplify only one mediation strongly, that two-mediation verbs are common but secondary, and that verbs amplifying more than two mediations at once are unusual rather than the norm. Nothing in this hypothesis should be built into the theory by definition, and nothing in the handful of examples considered above should be mistaken for a demonstrated generalization about the verbal lexicon as a whole. It is an empirical claim about the lexicon that a systematic survey across a large sample of verb senses could confirm or overturn, and it is offered here in exactly that spirit.
A further distinction sharpens the point. A verb carries a lexical default profile, the mediational amplification it carries in its most conventional use, but any particular enactment of that verb can depart from the default considerably. Cooking with a child, cooking professionally under commercial pressure, cooking ritually as part of a religious observance, and cooking while simultaneously teaching someone else to cook are all instances of the same verb, and each produces an event whose total mediational structure differs from the others in ways the verb itself does nothing to signal. The lexical profile of cook stays constant across all four cases. The mediational reality of the event does not.
The narrow mediational bandwidth of a verb is not a complaint about the poverty of verbs but evidence about how symbolic articulation works: it says something about the compression a verb performs, not about how narrow the living process it names actually is. A verb that amplifies only multimateriality is not thereby announcing that the event it describes was mediationally thin but doing what verbs do everywhere, selecting one or two dimensions of a richer configuration and making those dimensions recursively available, while leaving the rest of the configuration to be carried, if it is carried at all, by context, by the surrounding sentence, or by nothing.
V. Verbs Do Not Tell Us How Recursively an Event Is Lived
Turn now from mediational bandwidth to recursivity, and the same pattern of narrow specification reappears in a different register.
Take she stumbled. Nothing in the verb tells us whether her body recovered immediately without any further disturbance, whether the stumble produced an extended felt misalignment that lingered afterward, whether she consciously wondered what had just happened, whether she went on to describe the stumble to someone else, whether the event was later classified, in a clinical or institutional setting, as evidence of impairment or illness, or whether she subsequently contested that classification. The verb remains equally compatible with every one of these outcomes, because it specifies the process and says almost nothing about the recursive configuration in which that process was embedded.
This yields a distinction worth stating on its own terms: process specification is not the same as recursive specification. A language can discriminate walk, stride, stagger, stumble, trip, limp, creep, sprint, and totter with impressive processual precision, capturing fine differences in gait, speed, and control, while saying almost nothing about how recursively available any of these actions was to the person performing it, at what level, if any, it registered, or what further recursive activity, if any, it triggered.
The principle that follows is a strong one. Recursive level is ordinarily a property of the occurring configuration as a whole, not a property of the verb chosen to describe it. A small number of verbs appear at first to be exceptions, but on closer inspection they establish only lower bounds rather than fixing an exact level. To notice requires some minimal recursive registration, since nothing can be noticed without at least L2 salience. To wonder requires more than that, some articulable uncertainty. To ask requires overt symbolic articulation, placing it at L3 at minimum. Verbs such as to explain, to justify, to promise, to classify, to diagnose, to certify, to legislate, to theorize, and to problematize presuppose increasingly elaborate symbolic and institutional conditions the further along this list one moves. Even these verbs resist assignment to a single fixed level, however, since to ask can occur as a casual, unremarkable request between friends or as a formally structured act within a legal proceeding, and to classify can occur informally, as when someone privately sorts their own experience into rough categories, or through the full apparatus of an L4 bureaucracy equipped with forms, criteria, and appeals procedures. The right way to speak of such verbs is therefore in terms of a recursive-level range and a minimum recursive requirement, not a fixed assignment to L1 through L5.
A further concept becomes necessary once verbs are examined for what they permit rather than for what they specify: recursive valency. Standard linguistic valency asks what kinds and how many participants a verb allows to occupy its argument positions. Recursive valency asks a related but distinct question, what recursive statuses can occupy those positions while the verb retains its central sense. Some verbs are recursively indifferent to this question. One can touch a stone, touch oneself, or touch another person, and the verb does not change its sense across the three cases. One can wash a car, wash oneself, or wash a child with the same indifference. Verbs of this kind tolerate a nonrecursive, a selfrecursive, and an interrecursive object with no discernible shift in central sense.
Other verbs are considerably more selective. To persuade, to deceive, to forgive, to congratulate, to negotiate, and to promise all strongly require an interrecursively responsive participant on at least one side of the verb, since none of them makes sense applied to a stone or a storm. At the opposite extreme, to rust, to evaporate, and to crystallize characteristically select nonrecursive material participants in their central senses, rather than applying cleanly only to such participants. Biological tissue can crystallize, organic material can rust, and living substances can evaporate, so the restriction is a strong tendency rather than an absolute one, and forcing a straightforwardly interrecursive participant into their argument structure still produces either metaphor or a markedly non-central sense rather than an ordinary literal description. Recursive valency, provisionally defined, is the range of recursive statuses that can occupy the participant positions a verb licenses while the verb preserves its central sense.
This concept in turn makes visible a phenomenon worth naming separately: recursive flattening. Grammar frequently assigns the same participant role to occupants of radically different recursive status, using the identical construction for both. Remove rubble sits alongside remove protesters. Process raw material sits alongside process refugees. Deploy equipment sits alongside deploy troops. Manage assets sits alongside manage people. Lose money sits alongside lose patients. In each pair, the grammatical architecture of the sentence remains almost unchanged while the recursive status of what occupies the object position shifts from nonrecursive to interrecursive, and the sentence does nothing to mark that this has happened.
Verbs can classify what happens without registering the recursive status of those to whom it happens, allowing the same construction to describe the disposal of debris and the disposal of people without any grammatical signal that anything different has taken place. Institutional discourse can exploit this grammatical indifference, allowing recursively responsive persons to occupy syntactic positions structurally indistinguishable from those occupied by materials, resources, or objects. This is an affordance the grammar makes available rather than something institutions strictly depend on, since the same institutions could in principle mark the difference explicitly and, in some registers, do.
The four dimensions distinguished across the preceding sections can now be drawn together. Every verb sense can provisionally be examined along at least four largely independent axes: its process specification, the recurrent organization of change or activity it identifies; its mediational bandwidth, the number of mediations it amplifies; its recursive-level range, the span of recursive levels compatible with its use; and its recursive valency, the range of recursive statuses its participant positions tolerate. None of the four can be inferred straightforwardly from any of the others. A narrowly multimaterial verb such as hammer can still have a wide recursive-level range, since hammering can be entirely unreflective or the deliberate, monitored centrepiece of a craft demonstration. An interrecursive verb such as negotiate can have a comparatively narrow recursive valency even though the event it names is mediationally rich. A highly articulated L3 or L4 description can refer back to a process that occurred, for the person living it, at L1. The point of distinguishing the four axes is not to classify verbs once and for all but to prevent these different dimensions of process articulation from collapsing into one another, a collapse the sections above have shown to be easy to make and consequential once made.
VI. How Many Verbs Might Really Be Universal?
Return now to the question that motivated the inquiry from the start. Verbs might be expected to show greater cross-linguistic convergence than nouns, on the reasoning that recurrent processes cut across enormous numbers of differentiated entities, so that a comparatively small store of process categories could in principle organize the coordination of every human community regardless of what particular things its members happen to encounter.
Genesis 2 offers the cleanest possible image of the asymmetry this reasoning points toward. Adam is presented with an expanding world of differentiated entities, one animal after another, each requiring its own name, while the process he performs each time remains a single verb: to name. The same asymmetry recurs across an unlimited range of further cases. To eat applies across innumerable different foods. To carry applies across innumerable different objects. To walk applies across innumerable different terrains. To sleep applies across innumerable different sleeping arrangements. To name applies across innumerable different nameable entities. Entity differentiation promotes lexical proliferation, while process abstraction permits lexical recurrence across whatever entities the process happens to be applied to.
This asymmetry needs to be handled carefully, because it actually supports two different claims that are easy to run together. The first is a claim about lexical proliferation within a single language: a language that must name a thousand differentiated entities does not thereby need a thousand differentiated process verbs to describe what is done with, to, or by them. Adam's animals demonstrate this asymmetry cleanly, and nothing about it depends on any particular language's history. The second, quite different claim concerns the cross-linguistic universality of particular verb meanings, whether the same recurrent process is lexicalized in roughly the same way in every language regardless of how many process verbs any one language happens to have. The Genesis asymmetry predicts the first, a smaller recurrent process lexicon relative to entity differentiation. It does not by itself predict the second, and whether a smaller lexicon of this kind also turns out to be a more cross-linguistically stable one is a separate empirical question that the rest of this section takes up on its own terms.
The universal question needs sharpening before it can be answered, however, because at least three distinct claims travel under the single label universal verb, and they could in principle have very different answers. The first and strictest sense would be a dedicated lexical verb with roughly the same semantic range in every language, a single word doing the same job everywhere. The second, looser sense would be a conventional predicate, not necessarily a single morpheme, capable of expressing approximately the same process in every language, however that expression happens to be built. The third and loosest sense would simply be a recurrent human process that every language must somehow be capable of articulating, by whatever lexical or grammatical means it has available. These three inventories could differ dramatically in size. The first, strictest sense may turn out to include only a small handful of verbs. The second may perhaps run to a few dozen. The third could plausibly include several hundred recurring processes drawn from the shared mesocosmic conditions of embodied, social, and material human life.
A provisional list of candidates helps make the scale of the question concrete, offered here without prematurely declaring any of them universal in the strict sense: do, happen, move, live, die, eat, drink, sleep, wake, dream, see, hear, feel, say, think, know, want, come, go, walk, sit, stand, fall, give, take, hold, carry, touch, hit, cut, burn, wash, cry, and laugh. Even the most apparently irresistible candidates on this list, come and go among them, or the verbs of basic perception, turn out on closer typological inspection to be lexicalized quite differently across language families, sometimes split across several distinct verbs where English uses one, sometimes collapsed into a single verb where English insists on a distinction.
The important prediction that follows is therefore not that every language possesses some twenty verbs closely resembling their English glosses but a different and more interesting claim, that human languages may share a highly constrained mesocosmic process space, the recurrent activities, states, and events that embodied social life across the species reliably generates, while differing considerably in exactly where within that shared space each language chooses to draw its lexical boundaries. Two languages can be organizing the identical underlying process space and still look, from the outside, as though they classify action in entirely unrelated ways.
This question has in fact already been pursued at length under a different name. The Natural Semantic Metalanguage programme associated with Anna Wierzbicka and Cliff Goddard proposes a small set of universal semantic primes, terms argued to be indefinable and out of which more complex meanings are built, and its inventory includes clearly process-like items among them: DO, HAPPEN, MOVE, LIVE, DIE, THINK, KNOW, WANT, SEE, HEAR, and SAY (Wierzbicka 1996; Goddard and Wierzbicka, eds., 2002). The overlap with the candidate list above is close enough to be worth stating plainly. The important difference is that NSM does not require every prime to correspond to a dedicated, freestanding verb in every language. A prime can be lexicalized as a bound morpheme, a grammatical construction, or a meaning bundled together with other primes, provided the meaning itself is arguably expressible in the language somehow. The minimal verbal compression of the human mesocosm proposed here should therefore be read as a new interpretation of a problem the semantic-primes tradition has already been investigating for decades, using independent methods and independent evidence, rather than as an unnoticed reinvention of it. Where the present framework adds something is in asking why a residue of this kind should exist at all, tying it to the compression argument developed over the preceding sections rather than treating it as a brute semantic fact to be catalogued and left unexplained.
An empirical project follows directly from this reframing, worth stating even though it lies outside what this article can carry out. Begin with some fifty apparently universal process concepts, drawn from candidate lists of the kind offered above, and then attempt, through broad typological comparison across unrelated language families, in the spirit of the cross-linguistic testing already applied to motion-event categories (Bohnemeyer et al. 2007), to break each candidate in turn, looking specifically for languages that fail to treat the candidate process as a single lexical or grammatical unit. Whatever survives sustained attempts at destruction in this way would be considerably more theoretically interesting than any list generated from English intuitions alone, because its universality would have been earned rather than assumed. The residue that survived such testing might reasonably be called the minimal verbal compression of the human mesocosm, a small set of process categories that recur not because every language happened independently to invent the same convenient labels, but because the underlying mesocosmic conditions of human life leave certain recurrent processes too pervasive for any language to do without some means of articulating them.
VII. Linguistic Diversity as Differences Between Compression Profiles
The preceding argument now needs to confront linguistic relativity directly, since the claim that languages converge on a shared process space might seem to sit uneasily beside the well-documented fact that languages differ enormously in what their grammar obliges speakers to mark.
Temporal marking supplies a familiar example. Some languages grammatically require speakers to mark distinctions of tense or aspect that other languages leave to context, to optional particles, or to adverbs supplied only when the speaker judges them necessary. Considered purely from within linguistics, comparing one language's grammar directly against another's, these differences can look enormous: one language may possess an elaborate tense morphology marking several distinctions obligatorily on every finite verb, while another manages without obligatory tense marking of any kind.
The comparison changes considerably once the denominator changes. Rather than comparing Language A against Language B directly, compare each language against the temporal texture of the lived process itself, the actual temporality of an occurrence rather than any other language's grammar. That temporal texture includes biological rhythms, felt duration, anticipation, retention, memory, expectation, uncertainty about what is to come, several overlapping social temporalities running at once, material trajectories of growth, decay, and repair, and a continuously changing recursive orientation toward past and future that shifts with mood, context, and company. Measured against this bandwidth, even the most elaborate tense system available in any human language remains radically reductive, capturing only a thin selection of everything temporally present in the occurrence it marks.
This suggests a distinction between relative linguistic resolution and absolute process resolution. Languages can differ considerably on the first measure, in how finely they carve up temporal or other distinctions relative to one another, while remaining remarkably similar on the second, in how much of the underlying process bandwidth any of them actually preserves. Introduce, alongside this distinction, the idea of a compression profile: languages differ not so much in how much of reality they capture overall as in which distinctions they preferentially preserve while compressing everything else. One language may invest heavily in evidentiality, marking obligatorily how a speaker knows what they are claiming. Another may invest in social status, direction of motion, kinship relations, distinctions of smell, exact number, fine temporal relations, or precise spatial orientation. Each of these represents a genuine investment of grammatical resources. None of this need be treated as zero-sum, however, as though every language operated on a fixed budget of grammatical precision that had to be divided among competing domains. A language can obligatorily mark evidentiality and also maintain rich spatial or kinship distinctions, since obligatory marking in one domain does not draw down some fixed allowance available to the language as a whole. What differs across languages is better described as the distribution of obligatoriness, conventionality, accessibility, and lexical resolution across domains, not the total quantity of precision any language can afford.
Motion verbs supply a well-documented case of this same kind of profile variation, developed independently of the temporal domain just discussed. Languages differ systematically in whether path of motion is characteristically encoded in the verb itself or in a satellite element accompanying it, with manner correspondingly encoded in whichever slot the verb does not occupy (Talmy 1985, 2000). English characteristically packages manner into the verb and path into a satellite, as in run out or climb down, while languages of other families characteristically package path into the verb and leave manner to be added separately or left unexpressed. Later typological work has complicated this binary considerably, showing that many languages mix strategies or rely on serial-verb constructions that fit neither pattern cleanly. The point that matters here is not which pattern any given language prefers but that the underlying motion event does not change across these languages, only which of its components each language's verbal system happens to compress into the verb slot, a lexicalized instance of exactly the compression-profile variation this section has been describing in the temporal domain.
This reframes what linguistic relativity is actually a claim about. Linguistic differences remain consequential, because they shape which distinctions are cheap to express, habitual, obligatory, immediately salient, or easily recoverable from memory for speakers of a given language. What they should not automatically be read as showing is that speakers of different languages inhabit fundamentally different ontologies, since the differences at stake concern how each language distributes obligatoriness and resolution across domains, not how much of the underlying process world any of them managed to capture in total.
Max Frisch's observation about Swiss German is a useful illustration, provided his explanation and his observation are kept apart. Frisch noted that Swiss German could encode directional distinctions in perfectly ordinary expressions where High German left the same distinctions unspecified, and he interpreted this specificity through Switzerland's agrarian and topographic history, farmers working spatially demanding terrain who needed to track direction with unusual precision. Whether that historical explanation is correct is a separate empirical question, since farmers exist outside Switzerland, High German has long been spoken across intensely agricultural landscapes of its own, and directional systems can develop for reasons that have nothing to do with terrain at all. What Frisch's observation nevertheless suggests, independently of whether his own explanation of it is right, is the testable possibility that long-term mesocosmic stakes of a particular kind can leave differential mediational signatures in a language's lexical and grammatical resolution. Read this way, Swiss German preserves unusually high resolution at particular points associated with dwelling and multimaterial practical coordination, a hypothesis about mediational lexical investment worth testing on its own terms rather than a settled fact about Swiss thoroughness.
The scale correction still has to be applied even to this hypothesis, however, and would need to be applied whether or not Frisch's own historical explanation survives testing. A slightly finer spatial distinction, however useful within its domain, still captures almost nothing of the total embodied, material, and spatial process occurring when an object falls, moves, or is carried from one place to another. The formulation this suggests is worth stating directly: Swiss German may preserve more of dwelling at particular points while still preserving almost none of dwelling overall. Linguistic anthropology may, in this respect, have periodically exaggerated the differences among languages' compression profiles, because the comparison it usually performs sets one language's symbolic representation against another's rather than setting either against the bandwidth of what was compressed in the first place.
VIII. From Predication to Nominal Construal: The Cleanest Case of Symbolic Overreach
The article now reaches its conceptual centre. Consider a familiar set of pairings between a verbal and a nominal way of construing the same process: to relate and relation, to perceive and perception, to intend and intention, to know and knowledge, to live and life, to die and death, to value and value.
These pairings are not all of the same linguistic kind, and the difference matters enough to state before going further. Some are overt derivational nominalizations, a well-studied grammatical process in which a nominal form is built regularly from a verbal root (Comrie and Thompson 1985), as with relate to relation, perceive to perception, and intend to intention. Others are historically related without one member being synchronically derived from the other in any straightforward sense. Life is not a morphological nominalization of live in the way relation is of relate, and death stands in a comparably loose relation to die, however closely their histories are entangled. Value has its own complicated history as both noun and verb, with neither obviously prior to the other. What all of these pairings share, whatever their grammatical history, is not derivation but nominal construal: a process that could be predicated of a subject, something lives, something dies, something is valued, can also be handled through a stabilized noun that names it, life, death, value, and the two options coexist in the language regardless of which grammatical route produced the noun in the first place.
This distinction matters because the argument that follows depends on it and not on etymology. The claim is not that life is historically derived from live, a claim a linguist could reasonably dispute. The claim is that living, however that predicate came to have a corresponding noun, can be recursively handled through the stabilized symbolic object life once that noun is available in the language, and that derivational nominalization is only the most visible grammatical instance of a considerably broader symbolic operation.
Trace what happens across this operation carefully. An already compressed verbal articulation isolates a recurrent process profile, in the way the preceding sections have described at length. Nominal construal then takes that articulation a further step, making it available as a comparatively stable symbolic object rather than a fleeting predication attached to a particular subject on a particular occasion. Once nominally construed, the process can serve as the grammatical subject of further propositions in its own right: meaning is, knowledge requires, value consists in, identity is. None of this is initially mistaken, since nominal construal is one of multisymbolization's most powerful operations, allowing a process, once articulated, to become recursively portable, comparable across cases, classifiable within a wider system, teachable to those who did not witness the original occurrence, disputable in argument, and eventually sedimented into the institutions that depend on such stable handles to function at all.
The connection to poiesis is direct. Recursive life continually depends on the production of relatively nonrecursive grounds, documents, objects, habits, classifications, and names that make further recursive coordination possible without having to be reconstructed from nothing each time. A noun performs an analogous operation symbolically. It manufactures a symbolic rock out of a process, giving what was fluid and occasion-specific a stability the underlying process never possessed on its own.
The overreach becomes visible once the stabilization that symbolization introduced is retrospectively forgotten, and its stability comes to be treated as though it belonged to the process itself rather than to the symbolic operation performed on it. The sequence runs in one direction: a fluid process undergoes verbal compression, the resulting verbal articulation receives a nominal construal, the construal produces a symbolic object, that object becomes recursively stabilized through repeated use, the stabilized object comes to feel like an entity in its own right, and the entity is then treated as an ontological problem requiring philosophical solution. Symbolic overreach occurs when properties generated by symbolic mediation are retrospectively attributed to the process that was symbolically mediated in the first place.
This diagnosis runs deeper than the more familiar concept of reification with which it might initially be confused. Reification names the eventual appearance of an abstraction or a relation as though it were a thing, describing the endpoint of the process. Symbolic overreach names the entire sequence of operations through which that endpoint is reached, verbal compression followed by nominal construal followed by forgetting that either operation occurred, identifying where in the sequence the mistake actually enters rather than only naming the mistake once it has already arrived.
The governing formula for the whole article can now be stated in the form it has been building toward throughout. Compression is the achievement. Forgetting that compression occurred is symbolic overreach. Verbal compression on its own is not a defect, since the preceding sections have argued at length that compression of exactly this kind is what makes ordinary language usable at all, and nominal construal on its own is not a defect either, since portable symbolic handles are what allow a process, once articulated, to travel beyond the occasion of its first use, whether the handle arrived through regular derivation or through an independently lexicalized noun standing beside its verb. What produces the overreach is neither operation in isolation but the subsequent amnesia about having performed them, the point at which a symbol's hard-won stability is mistaken for a property the underlying process always had.
IX. Philosophy After the Verb: Why the Problem Keeps Returning
The final section widens the argument and draws it to a close. Philosophy, on the reading developed here, repeatedly turns verbs into nouns and then treats the resulting nouns as problems requiring an ontology of their own. Meaning, knowledge, being, value, consciousness, perception, intention, relation, action, and identity all name philosophical problems built from exactly the nominal construals the previous section traced. Philosophical thought needs stabilizations of this kind, because sustained recursive analysis is impossible if everything under discussion remains perpetually in flux, with no stable term available to hold a distinction in place across an argument's several steps. Philosophy therefore has an unusually strong structural incentive toward nominal construal, stronger than that of most other forms of discourse, and this explains both its distinctive power and its characteristic vulnerability at once.
The strongest version of this claim is not that philosophy has simply failed to notice verbs, or that some philosophers naively favour nouns while others more wisely favour process. It is that even philosophers who explicitly and self-consciously resist substantialization remain drawn back through the nominal bottleneck, because the bottleneck is built into what sustained symbolic analysis at L4 and L5 requires and not into any particular philosopher's carelessness. Heidegger and Wittgenstein are worth examining together for exactly this reason. Both built philosophies organized specifically around resisting the reduction of lived activity to static objects, and both, on close inspection, illustrate how difficult that resistance is to sustain rather than how easy it is to achieve.
Heidegger provides a striking illustration of the problem from within a philosophy explicitly designed to guard against it. He is acutely concerned that inherited philosophical concepts turn lived phenomena into static objects, and his notion of formal indication is meant to keep concepts open, directional, and enactable rather than closed and inert. Yet the difficulty he is responding to may arise in part because philosophical inquiry has already selected its objects of analysis in nominalized form before any formal indication can be applied to them: Being, Time, Death, Care, World, and Guilt all enter the analysis as nouns from the outset. Heidegger nominalizes, diagnoses the resulting stasis with unusual acuity, and then develops elaborate procedures for making the resulting noun move again. His occasional deliberate verbalizations of otherwise nominal terms become especially revealing under this reading, since they show philosophy having to reintroduce happening into concepts from which its own prior symbolic procedures had already removed it.
Wittgenstein appears at first to be considerably more resistant to the same difficulty, since his later philosophy emphasizes use, practice, action, rule-following, and ostension rather than static essences or fixed meanings. The residual nominal construal nonetheless persists, visible the moment it is looked for directly. Consider language-game rather than playing language: even the concept built to emphasize activity has itself become a noun, carrying the process of playing inside a term that itself sits still. Use becomes use, a noun standing in for using. Rule-following becomes a rule, a noun standing in for the ongoing activity of following it. Forms of activity sediment into conceptual nouns even within a philosophy organized specifically around resisting exactly that sedimentation. Wittgenstein is therefore not a target this argument needs to defend itself against. He is evidence for how deep the phenomenon runs, since if the nominal bottleneck reappears even inside a philosophy built to dissolve it, no philosophy organized around nouns from the outset stood any real chance of avoiding it.
Pointing offers a particularly clear miniature case, worth developing in some detail. Many philosophical treatments of ostension ask what a person points at, and how the act of ostension is supposed to establish reference between a word and an object. Begin instead from the verb rather than the noun that referential treatments of ostension quietly presuppose. Someone points. Someone else attends to the gesture. Their attention is redirected by it, from wherever it had been, toward whatever the gesture indicates. The pointer monitors whether the intended uptake has actually occurred, adjusting the gesture if it has not. The recipient either follows the gesture successfully or fails to, and either outcome can itself become the occasion for correction, repetition, or renewed attempt. Described this way, pointing is an interrecursive unfolding, a small negotiated exchange between two recursively responsive participants, well before it is the static bridge between a word and an object that a referential treatment of ostension takes it to be. A referential treatment of this kind has already nominalized the interaction into ostension before beginning its analysis, and everything that follows inherits the stillness that nominal construal introduced. This is not a criticism that lands equally on every philosopher who has written about pointing. Wittgenstein's own discussion of ostensive definition repeatedly insists that pointing means nothing apart from the training, uptake, and surrounding practice that make a gesture intelligible as pointing to this rather than that, which is close to the interrecursive unfolding just described. The criticism lands on treatments that take ostension as a static bridge to be analysed rather than as a practice to be described, a distinction Wittgenstein himself was already drawing attention to.
This demonstrates how readily philosophy gravitates from an activity toward the stabilized entities that result from, or merely participate in, that activity, mistaking the entity for the more basic phenomenon when in fact the entity was produced by compressing the phenomenon in the first place. The qualification that must be added immediately, before the argument overreaches in its own right, is that the answer is emphatically not to replace a noun ontology with a verb ontology, on the assumption that verbs finally give unmediated access to process where nouns had failed. Verbs also compress, as the whole of this article has argued. To walk is not walking itself, in exactly the sense that knowledge is not knowing itself. Becoming, the philosophical term most explicitly built to resist stasis, can itself become one of philosophy's most thoroughly reified nouns the moment it is treated as naming a stable underlying reality rather than as a further compressed articulation of process, subject to precisely the same overreach as any noun it was meant to replace.
What follows instead is a general discipline worth stating directly. Whenever an apparently stable concept is encountered, ask what process was compressed to make the resulting symbol recursively available in the first place. What was happening. Through which mediations. With which participants occupying which recursive statuses. At what recursive level or levels. Over what temporal trajectory. What was discarded in the course of the compression, and what became stabilized in its place. Which properties belong to the process being described, and which belong instead to the symbolic articulation that was built to describe it.
The modest title with which this article began has turned out not to deliver a theory of a single grammatical category after all. Verbs have instead provided an unusually clear experimental site for observing a much larger feature of human multisymbolization at work. The process world is too rich to be articulated at anything approaching its own bandwidth, and human language survives, and indeed flourishes, through extreme selective compression rather than despite it. The resulting symbols are powerful precisely because they are radically poorer than what they articulate, and their poverty is what makes them portable, repeatable, combinable, and recursively available across occasions their original occurrence never anticipated. Three propositions can carry the whole of this argument forward from here. Living processes exceed their verbal articulation. Ordinary language works because it does not attempt to preserve that excess. Symbolic overreach begins at the point where the success of a compression persuades its users that nothing important was ever compressed in the first place. The world does not resemble our words because our words describe it so well. The extraordinary achievement of language is that life can coordinate itself through symbols that preserve so little of the processes from which they were made.
Selected References
• Anscombe, G. E. M. 1957. Intention. Oxford: Blackwell.
• Baker, Collin F., Charles J. Fillmore, and John B. Lowe. 1998. "The Berkeley FrameNet Project." Proceedings of COLING-ACL 1998: 86–90.
• Beavers, John, and Andrew Koontz-Garboden. 2012. "Manner and Result in the Roots of Verbal Meaning." Linguistic Inquiry 43(3): 331–369.
• Bohnemeyer, Jürgen, Nicholas J. Enfield, James Essegbey, Iraide Ibarretxe-Antuñano, Sotaro Kita, Friederike Lüpke, and Felix Ameka. 2007. "Principles of Event Segmentation in Language: The Case of Motion Events." Language 83(3): 495–532.
• Comrie, Bernard, and Sandra A. Thompson. 1985. "Lexical Nominalization." In Language Typology and Syntactic Description, vol. 3, edited by Timothy Shopen. Cambridge: Cambridge University Press.
• Davidson, Donald. 1963. "Actions, Reasons, and Causes." Journal of Philosophy 60(23): 685–700.
• Goddard, Cliff, and Anna Wierzbicka, eds. 2002. Meaning and Universal Grammar: Theory and Empirical Findings. Amsterdam: John Benjamins.
• Hopper, Paul J., and Sandra A. Thompson. 1984. "The Discourse Basis for Lexical Categories in Universal Grammar." Language 60(4): 703–752.
• Kipper-Schuler, Karin. 2005. "VerbNet: A Broad-Coverage, Comprehensive Verb Lexicon." PhD diss., University of Pennsylvania.
• Levin, Beth. 1993. English Verb Classes and Alternations: A Preliminary Investigation. Chicago: University of Chicago Press.
• Miller, George A. 1995. "WordNet: A Lexical Database for English." Communications of the ACM 38(11): 39–41.
• Moens, Marc, and Mark Steedman. 1988. "Temporal Ontology and Temporal Reference." Computational Linguistics 14(2): 15–28.
• Pustejovsky, James. 1991. "The Syntax of Event Structure." Cognition 41(1–3): 47–81.
• Rappaport Hovav, Malka, and Beth Levin. 2010. "Reflections on Manner/Result Complementarity." In Lexical Semantics, Syntax, and Event Structure, edited by Malka Rappaport Hovav, Edit Doron, and Ivy Sichel. Oxford: Oxford University Press.
• Talmy, Leonard. 1985. "Lexicalization Patterns: Semantic Structure in Lexical Forms." In Language Typology and Syntactic Description, vol. 3, edited by Timothy Shopen. Cambridge: Cambridge University Press.
• Talmy, Leonard. 2000. Toward a Cognitive Semantics, Vol. 2: Typology and Process in Concept Structuring. Cambridge, MA: MIT Press.
• Vendler, Zeno. 1957. "Verbs and Times." The Philosophical Review 66(2): 143–160.
• Wierzbicka, Anna. 1996. Semantics: Primes and Universals. Oxford: Oxford University Press.
• Zacks, Jeffrey M., Barbara Tversky, and Gowri Iyer. 2001. "Perceiving, Remembering, and Communicating Structure in Events." Journal of Experimental Psychology: General 130(1): 29–58.