In his IAI essay, Words Do Not Refer to Reality, 30 September 2024, Paul Pietroski rejects the familiar picture in which words latch onto objects and derive their meanings from those attachments. Instead, he suggests that lexical items operate as headings for extendable families of concepts. A word such as book, bank, or line can express several related ways of thinking, none of which need to constitute its single, public, permanently fixed meaning. Polysemy is not an unfortunate contamination of language. It is how language works.
I largely agree. My quarrel is not with the argument but with where it stops. The headline announces that words do not refer to reality – though headlines are usually the editor’s work, not the author’s, and I suspect Pietroski would happily concede what I want to add. Words may not refer by themselves, but speakers plainly use them to refer within organised encounters, and Pietroski’s account is conspicuously silent about how. The silence is where the interesting work remains.
The trouble with water
Pietroski illustrates his case by observing that Diet Coke contains a slightly higher proportion of H₂O than the water from his New Mexico well. Nonetheless, ordinary speakers call the well water water and the Diet Coke Diet Coke. He concludes that the word water cannot simply apply to all and only samples of H₂O. Instead, it may express different concepts, including a kind-concept concerned with underlying constitution and a functional concept concerned with water’s role in our lives.
The distinction is real, but the example is doing suspiciously theatrical work.
Suppose I begin making lemonade. I pour water into a pitcher, add sugar, and then add lemon juice. At the first stage, I have water. At the second, sugar-water. At the third, lemonade. Yet nothing resembling alchemy has occurred. The H₂O has not vanished, nor been transmuted into some unrelated substance. It remains materially present, continues to dissolve the other ingredients, and retains much of its functional role, including hydration.
What changes is the preferred classification of the organised mixture. Lemonade becomes more informative than water. That does not mean water has ceased to exist or lost all referential relevance. It means the mixture is now being carved at a different practical scale.
The intermediate state makes this clearer. Were I making lemonade with a mate who had added sugar to the water, I might ask:
Pass me the water so I can add the lemons.
No confusion would follow. My accomplice would understand that the water refers to the sugar-water presently occupying the pitcher. Ordinary communication does not require a chemical recertification ceremony after each ingredient.
Suppose, moreover, that two pitchers are present, one empty and one containing the sugar-water. I might simply ask for the water. The expression would ordinarily select the pitcher containing it. The substance identifies the relevant container through the shared activity. Conversely, I might ask for the pitcher, in which case the practical context would probably select the same object. If both pitchers were equally salient, some further qualification or gesture would be needed.
This tells us something important: neither water nor pitcher carries a complete destination label inside itself. Reference is completed through contrast, discourse history, spatial organisation, shared attention, gesture, and the activity under way.
Pietroski is therefore right that the word is not tied directly to the thing. But reference has not thereby retreated into the privacy of concept-families. It has become situated and distributed – a joint achievement of speakers, words, and the structured circumstances in which both operate.
Where does polysemy sit on the Gradient?
Reading Pietroski alongside the Language Insufficiency Hypothesis invites an obvious question. LIH maps language along an Effectiveness–Complexity Gradient, from Invariants through Contestables and Fluids to the Ineffables. Fluids are the region of drifting sense – consciousness, intelligence, equity – and Pietroski has just spent an entire essay making the multiplicity of sense salient. So the tempting inference is that polysemy is the signature of the Fluid region, perhaps even its mechanism.
The tempting inference is wrong. Polysemy is orthogonal to the Gradient. It occurs everywhere along it, quite possibly proliferating at the Ineffable horizon.
The relevant distinction is this:
Polysemy concerns the availability of multiple related senses. The Gradient concerns the reliability and cost of coordinating through them.
Walk the regions and the orthogonality shows itself.
An Invariant can be comfortably polysemous. Pitcher names the vessel and, by an ordinary container-for-contents extension, what the vessel holds: one can pass the pitcher or drink the whole pitcher without anyone reaching for a dictionary. The senses are multiple; the coordination cost is negligible. Context selects cheaply, and nobody notices the selection occurring. (The baseball player, incidentally, is no help here – that pitcher is a different lexeme sharing a pronunciation, which Pietroski would rightly file under homophony rather than polysemy. The distinction matters: polysemy is one heading with related senses; homophony is two headings that happen to sound alike.)
Among Contestables, the availability of senses is no longer innocent. Selecting a sense of justice, consent, or responsibility is not lexical housekeeping; it authorises different actions, allocates different burdens, and licenses different punishments. The senses remain related – that is what makes the contest possible – but coordination now carries institutional freight.
Among Fluids, senses travel between domains, hybridise, and borrow authority from one another. Consciousness drifts among scientific, philosophical, colloquial, and metaphysical uses, often without speakers noticing that the operative sense has changed mid-sentence. Here polysemy is not merely present but weaponisable: the drift does argumentative work its users need not acknowledge.
At the Ineffable horizon, one might expect polysemy to give out along with everything else. The opposite occurs. Metaphor, paradox, analogy, and apophatic language pile up around what cannot be stably expressed. The multiplicity of descriptions indicates not conceptual richness successfully captured, but repeated compressive failure – each new sense another attempt at a compression that will not hold. Polysemy proliferates precisely where coordination collapses, which rather undermines the naïve assumption that ambiguity is what breaks communication. Communication at the horizon is already broken; ambiguity is the debris.
So Pietroski supplies the ubiquity of polysemy, and LIH supplies its variable price. The Gradient is not a ladder from monosemy to polysemy. It maps the changing effectiveness of linguistic coordination under increasing complexity and load, and LIH treats its categories as heuristic regions along a continuous curve, not as ontological bins.
Reference as encounter
MEOW, the Mediated Encounter Ontology developed in The Architecture of Encounter, permits a deeper account of what is happening in the kitchen.
On this view, reference is not a cable running from a word to an antecedently packaged object. Nor is meaning exhausted by a private concept applied to such an object. The referential event emerges within an organised encounter comprising material constraint, salience, affordance, contrast, shared practice, and linguistic and gestural cues.
When I ask for the water, the hearer does not first calculate the timeless extension of the lexical item and then locate an object satisfying it. The request enters an already structured field. We are making lemonade. Water is needed. One pitcher contains the relevant mixture. That pitcher affords pouring; the empty one does not. The word directs attention within this organisation.
Language functions here as a synchronisation protocol. It does not recreate my entire encounter inside another person. It supplies enough of a cue for our differently mediated encounter-structures to converge upon the next action. The water, the pitcher, or even that accompanied by a gesture may all stabilise the same practical result.
An objection is available, and it deserves stating at full strength. Pietroski can absorb the entire kitchen scenario. The hearer, he may say, simply selects a concept from her water-family, cued by context – and ‘context disambiguates’ has been common property since Grice. Concept-families plus ordinary pragmatics deliver the lemonade. Why reach for an ontology of encounter?
Because of where the application conditions end up. Pietroski’s move is a relocation: he strips application conditions from words and reinstalls them in concepts, leaving the traditional representational triangle – thinker, concept, thing – standing. Concepts still possess application conditions; things still antecedently satisfy them; context merely chooses which concept gets deployed. MEOW’s move is a dissolution: there are no stable application conditions waiting anywhere, in words or in concepts. What there is instead is constraint-tested stabilisation within the encounter itself. Thinker, concept, and thing are not the fixed corners of a triangle across which reference travels; they are stabilisations that emerge together within relational organisation, at scales fixed by practical interest.
This is not linguistic idealism, and constraint is what prevents it from becoming so. Not every description survives encounter equally well. A pitcher remains sufficiently stable to be tracked despite changes to its contents. Water remains materially and causally present after sugar and lemon are added, even though lemonade becomes the dominant name for the organised whole. Multiple descriptions remain valid because they select different patterns, scales, and practical interests – and invalid descriptions fail, publicly and repeatably, when the encounter refuses them. Ask the empty pitcher for water and the world declines to cooperate.
The pitcher, the water, the sugar-water, and the lemonade are not four metaphysically isolated substances taking turns on the kitchen counter. They are overlapping stabilisations within an evolving encounter. The nouns divide the event because division is useful, not because reality arrived pre-labelled.
A less dramatic conclusion
Pietroski is right to reject the naïve belief that words possess fixed application conditions by virtue of being attached to worldly things. He is right that polysemy and conceptual variation are ordinary rather than exceptional. And nothing above contradicts him; it continues past the point where he stops.
His positive account transfers application from words to concepts while retaining much of the traditional representational structure: thinkers use concepts to depict things, and those concepts possess application conditions. MEOW asks the prior question – how thinker, concept, and thing become stabilised within encounter in the first place – and answers that they are co-stabilised there, under constraint, rather than assembled from pre-certified parts.
The better conclusion is therefore not:
Words do not refer to reality.
It is:
Words do not refer alone. Speakers achieve reference through words operating within shared, constrained practices.
This is less revolutionary, which is unfortunate for headlines, but more faithful to how communication actually works. The word does not contain the referent. Neither does it float free of reality. It participates in an encounter through which reality is provisionally, selectively, and often quite successfully made available for coordinated action. The lemonade, meanwhile, gets made – which is more than most theories of reference can say for themselves.
This content was produced by Bry Willis, a philosopher with a focus on language. He shares content here on Substack as well as on his Philosophics Blog. Bry uses various LLMs in his production workflow at all stages, from ideation to production, as well as for the production of ancillary content – audio and video summaries, images, PDFs, and possibly other media.
You can also find him on various social media platforms:
For more information, check out my About Me page