Dissolving the paradox of ineffability

1. Introduction

Sentences like $⌜\delta$ is indescribable $⌝$ appear to state that the speaker is unable to do what they just did. This apparent contradiction is known as the paradox of ineffability. The paradox has historical origins in both the East and the West. Kuiji, a Buddhist in the Yogācāra tradition writing in the 7th century, was committed to all concrete particulars being indescribable. He posed the paradox as a question: “[A]ll real things being unspeakable, wouldn’t it be inappropriate as well to speak the word”unspeakable”?” (Ho 2015, 72). Augustine of Hippo, a Christian theologian writing in the 4th century, wrestled with the paradox in connection to God’s ineffability:

God is unspeakable. But what I have spoken would not have been spoken if it were unspeakable. For this reason, God should not even be called unspeakable because, even when this word is spoken, something is spoken. There is a kind of conflict between words here: if what cannot be spoken is unspeakable, then it is not unspeakable because it can actually be said to be unspeakable. It is better to evade this conflict silently than to quell it disputatiously. (Augustine 2008, §1.13)

The paradox surfaces whenever something is regarded as indescribable. Unlike other ancient paradoxes like This sentence is false, which is paradoxical but never sincerely said, $⌜\delta$ is indescribable $⌝$ is paradoxical and sincerely said.¹ Indeed, it is paradoxical because it is sincerely said.

In the history of philosophy, topics including God, particulars, sensory experience, the Dao, nothingness, and artwork are regularly characterized as ineffable. As Moore (1992, 427) puts it, “Ineffability $\dots$ has had one of the longest and most distinguished of philosophical histories.” The Corpus of Contemporary American English also contains many examples of speakers describing everyday things as indescribable. Examples include sex, emotion, wine, film, pain, the passing of time, beauty, clothing, and even large quantities (Davies 2008). So the paradox of ineffability is topic neutral. Whatever stands-in for $\delta$, we have it.

To solve the paradox is to explain how $⌜\delta$ is indescribable $⌝$ can be true in a context. It is widely thought that a solution is a pre-requisite for explicating the coherence of ineffability because it is how one answers the self-refuting objection. If $⌜\delta$ is indescribable $⌝$ cannot be true, nothing can be claimed to be ineffable without yielding a contradiction. As such, everything is describable. This objection is found regularly in the philosophy of religion (Alston 1956; Yandell 1975; Plantinga 2000). It has ancient origins too. For example, Ānandavardhana, a 9th century Indian literary critic, objected to the view that the value of poetry was its ineffability by arguing that predicating ineffability of poetry was self-refuting.

These special [virtues] of words and meanings can be explained and have been explained in many ways. To imagine that there is some ineffable virtue over and above these is to admit that one’s power of analysis has ceased. For ineffability, in the sense of being beyond the range of all words, cannot be predicated of anything, because in the last analysis a thing can be designated by the word “ineffable.” (Ānandavardhana 1990, §347a)

Like the paradox, the self-refutation objection is topic neutral. So any putative solution would need to be neutral too.

This essay provides what I call a dissolution of the paradox of ineffability as opposed to a solution. I argue that $⌜\delta$ is indescribable $⌝$ is always false in a context for any $\delta$ whatsoever (§3-§4). However, I further argue that a solution is not a pre-requisite for explicating the coherence of ineffability (§5). The self-refuting objection can be answered without finding a way for $⌜\delta$ is indescribable $⌝$ to be true. A middleground is thereby charted: contra advocates of ineffability, $⌜\delta$ is indescribable $⌝$ is always false in every context, but, contra critics of ineffability, this generalization does not show that ineffability is incoherent. Both of these conclusions follow, I argue, from paying close attention to how predicates like indescribable are gradable adjectives that express agentive modality. That such terms express agentive modality will help explain why $⌜\delta$ is indescribable $⌝$ has truth-conditions that cannot be satisfied in a context. That they are gradable will help explain why ineffability can be explicated without solving the paradox. Accordingly, this essay begins with the natural language semantics of such adjectives (§2). It ends by discussing how the dissolution of the paradox relates to contemporary theories of ineffability developed in the philosophy of religion by Jacobs (2015) and Keller (2018).

2. The paradox

To build up to the paradox, let’s start with morphemes, the smallest building blocks of natural language. From most verbs $𝒱$, an adjective can be formed by combining $𝒱$ with the modal suffix -ible/-able.² The result is an adjective expressing that an agent is able to do the action denoted by $𝒱$. Examples include readable, singable, and drinkable.

These adjectives have a number of interesting semantic properties. Of central importance to the paradox is that they are agentive modal adjectives. These adjectives express that an agent can/is able to $𝒱$. To add precision to our discussion, I assume the semantics for agentive modals owed to Mandelkern et al. (2017). On their account, agentive can in $⌜$S can $\varphi ⌝$ expresses that an action $\varphi$ is a practically available action to S such that, in the closest worlds where S tries to $\varphi$, they $\varphi$. As such, agentive modals express a hypothetical guarantee. An agent may not try to $\varphi$. But if they can $\varphi$, they will $\varphi$ in the closest worlds where they try. Applying this semantics to the suffix -ible/-able, agentive modal adjectives express that that the action denoted by the relevant verb $𝒱$ is hypothetically guaranteed. For example, The book is readable expresses that reading is practically available to the relevant agent such that if that agent tried to read the book, they would succeed at reading in the closest worlds.

But which agent? Constructions like $⌜$S can $\varphi ⌝$ have the agent as the grammatical subject. But $⌜\delta$ is $𝒱$able $⌝$ does not. Instead, modal agentive adjectives are perspective-sensitive.³ The adjectives are interpreted with respect to the perspective of an individual or group who possesses the ability to $𝒱$. This perspective can be made explicit with a prepositional phrase such as to S in $⌜\delta$ is $𝒱$able to S $⌝$, but the perspective is often implicit. When implicit, the perspective often defaults to that of the speaker. So the agent who can $𝒱$ is the speaker who uttered the sentence $⌜\delta$ is $𝒱$able $⌝$. But sometimes the perspective receives a generic interpretation. For example, The sidewalk is walkable is interpreted as walkable for a normal agent as opposed to walkable for the speaker.

We now have the requisite pieces to extend the formal semantics of agentive modal auxiliaries from Mandelkern et al. (2017) to agentive modal adjectives. To start, let ${𝒜}_{{𝑆}_{𝑐},𝑐,𝑤}$ represent a set of actions that are practically available in context c and world w to S ${}_{𝑐}$, or the speaker of context c. Next, let f ${}_{𝑐}$(A,w) be a selection function provided by the context c that delivers the world in which S ${}_{𝑐}$ tries to perform an action A in world w. Then $⟦\delta$ is $𝒱$able ${⟧}^{𝑐,𝑤}$ = 1 iff $\exists$A $\in$ ${𝒜}_{{𝑆}_{𝑐},𝑐,𝑤}$: $⟦𝒱$(S, $\delta$) ${⟧}^{𝑐,{𝑓}_{𝑐}(A,𝑤)}$ = 1.

The semantics adopted from Mandelkern et al. (2017) appeals to actions that are practically available. They do not define this notion but do provide a rough-and-ready gloss. An action $\varphi$ is practically available to S just in case S could reasonably plan to $\varphi$ in order to achieve a relevant goal. Some readers may not be satisfied with this gloss. However, it comes with the territory. Semantic explanations of agentive modality at some point introduce a notion to explain how some actions are accessible or not to the agent. For example, Maier (2015) appeals to actions that are options, or actions that figure in choice situations. Likewise, Lewis (2020), in a posthumously published outline, toyed with agentive modality being explained in terms of the absence of obstacles, or something that “wouldn’t go away if things were just a little different.” I take all of these notions to be explicating the same intuition that an action an agent is able to do is an action that is open to an agent. Any will do.

To take another step closer to the paradox, consider agentive modal adjectives morphologically derived from verbs of saying such as say, speak, and describe. The result is adjectives like sayable, speakable, and describable. To take us all the way there, incorporate negation with the prefix in-/un- and we have an ineffability predicate:

$\begin{array}{c}\hfill 𝑖𝑛-\hfill \\ \hfill 𝑢𝑛-\hfill \end{array}$ + $\begin{array}{c}\hfill 𝑑𝑒𝑠𝑐𝑟𝑖𝑏𝑒\hfill \\ \hfill 𝑠𝑝𝑒𝑎𝑘\hfill \\ \hfill 𝑠𝑎𝑦\hfill \end{array}$ + $\begin{array}{c}\hfill -𝑖𝑏𝑙𝑒\hfill \\ \hfill -𝑎𝑏𝑙𝑒\hfill \end{array}$

Like any negated agentive modal adjective, ineffability predicates are true when the action denoted by the verb $𝒱$ is not practically available to the relevant agent. Since the verb is a verb of saying, an ineffability predicate is true when saying or describing $\delta$, the grammatical subject in $⌜\delta$ is indescribable $⌝$, is not practically available to the relevant agent.

In what follows, I will continue to focus discussion on the adjective indescribable. I ignore ineffable because it is a loan word from Latin ineffabilis acquired either directly or by way of French (Harper Accessed 2022). As a result, the verb of saying from which it is derived—effari—is not itself in the English vocabulary. (Though this does not stop many from making effing jokes.) It is thereby difficult to elicit reliable linguistic intuitions about the meaning of ineffable. I will also not discuss unspeakable extensively because it is sometimes interpreted as a deontic modal adjective as opposed to an agentive modal adjective. For example, What he did was unspeakable is interpreted as what happened being too terrible or evil to permissibly be spoken about.

The paradox can now be unpacked. By default, the perspective anchoring the ineffability predicate in $⌜\delta$ is indescribable $⌝$ is the speaker. So $⌜\delta$ is indescribable $⌝$ entails that describing $\delta$ is not practically available to the speaker. But insofar as the speaker just performed an act of description with the predicate indescribable, describing is practically available to them. So using $⌜\delta$ is indescribable $⌝$ appears to be a performative contradiction. What is done by asserting $⌜\delta$ is indescribable $⌝$ is what its truth-conditions require to be practically unavailable.

3. Dissolution

It is a performative contradiction that cannot be avoided. A comparison with other performative sentences like $⌜$I (hereby) say that p $⌝$—where p is any proposition whatsoever—is instructive. The semantics of these sentences is such that their utterance is self-verifying.⁴ To utter them is to create the event of saying that constitutes their truthmaker. That self-verification results from the statement being, in part, presently about the speaker. It disappears—along with the ability to insert the adverb hereby—when the tense or subject changes. Neither $⌜$I said p $⌝$ nor $⌜$Thanasis says that p $⌝$ is performatively self-verifying like $⌜$I (hereby) say that p $⌝$ is self-verifying.

Instead of being self-verifying, $⌜\delta$ is indescribable $⌝$ is self-refuting. To utter it is to bring about the event of description that constitutes the truthmaker for its negation. That self-refutation similarly results from the statement being, in part, presently about the speaker. The contradiction disappears when the sentence is not presently anchored to the speaker. There is nothing amiss about $⌜\delta$ was once indescribable (to me) $⌝$. This sentence entails that describing $\delta$ was not practically available at a prior time. But there is no conflict between an act being unavailable in the past and that act now being available to the speaker. Likewise, there is nothing problematic about $⌜\delta$ is indescribable to Thanasis $⌝$. This sentence merely entails that describing $\delta$ is not practically available to Thanasis. What goes for whoever is talked about need not go for the speaker.

Another way to illustrate the performative contradiction is to show that a logical contradiction can be derived from the supposition that $⌜\delta$ is indescribable $⌝$ can be truthfully asserted. We need just two additional premises. Here is the first premise:

1. (I)

   If $⌜\delta$ is indescribable $⌝$ is truthfully asserted by a speaker, then describing $\delta$ is a practically available action to the speaker.

Asserting is a complex action that consists in acts of referring and acts of predicating that correspond to the subject/predicate structure of a sentence (Austin 1962; Searle 1969). As such, a speaker cannot assert $⌜\delta$ is indescribable $⌝$ without predicating indescribable of $\delta$. So if a speaker asserts p, they have described $\delta$. If they have described $\delta$, description is practically available to them. The second premise is this:

1. (II)

   If describing $\delta$ is a practically available action to the speaker, then $⌜\delta$ is indescribable $⌝$ is false.

This premise results straightforwardly from the natural language semantics. Since the truth-conditions for $⌜\delta$ is indescribable $⌝$ require that the act denoted by the verb describe not be a practically available action, the sentence is false if describing is practically available to the speaker. To get a contradiction, we just need to suppose that $⌜\delta$ is indescribable $⌝$ can be truthfully asserted and modus ponens does the rest.

1. (I)

   If $⌜\delta$ is indescribable $⌝$ is truthfully asserted by a speaker, then describing $\delta$ is a practically available action to the speaker.

2. (II)

   If describing $\delta$ is a practically available action to the speaker, then $⌜\delta$ is indescribable $⌝$ is false.

3. (III)

   $⌜\delta$ is indescribable $⌝$ is truthfully asserted by a speaker.

4. (IV)

   Therefore, $⌜\delta$ is indescribable $⌝$ is false.

(III) and (IV) are contradictory. You cannot truthfully assert what is false. Since (I) is motivated by banal observations about speech acts, and (II) just elaborates the truth-conditions for agentive modal adjectives, (III) is what ought to be jettisoned of the three premises. $⌜\delta$ is indescribable $⌝$ cannot be truthfully asserted because it is always false.

4. Failed solutions

Once we appreciate that ineffability predicates are agentive modal adjectives, it is straightforward to see that an assertion of $⌜\delta$ is indescribable $⌝$ is a performative contradiction. However, solutions to the paradox have been developed, especially in the philosophy of religion. To complete the case that the performative contradiction is unavoidable, five solutions are discussed. Three have been proposed, and two can be envisioned. None succeed in showing that $⌜\delta$ is indescribable $⌝$ can be true.

5. Dissolution continued

We should accept that $⌜\delta$ is indescribable $⌝$ is always false. This follows from the natural language semantics of indescribable (§3), and is further confirmed after the survey of various solutions that we just completed (§4). But what does this tell us about ineffability in general?

The answer given is usually bleak. At the essay’s outset, we saw how Ānandavardhana understood this result for the ineffability of poetry. Here is how Plantinga (2000, 5-6) similarly answers a question about whether the ineffability of God is coherent:

[T]he answer seems to be no; one who makes the claim seems to set up a certain subject for predication—God—and then declare that our concepts do not apply to this being. But if this is so, then, presumably, at least one of our concepts—being such that our concepts don’t apply to it—does apply to this being. Either those who attempt to make this claim succeed in making an assertion or not. If they don’t succeed, we have nothing to consider; if they do, however, they appear to be predicating a property of a being they have referred to[.]

As such, accepting that $⌜\delta$ is indescribable $⌝$ is always false seems to be tantamount to conceding that the self-refuting objection against ineffability is successful. I suspect this is why many have been motivated to solve the paradox. Hewitt (2020, 19), for example, suggests that the absence of a solution “fatally undermine[s]” the view that God is ineffable.

In considering the self-refuting objection, we must distinguish two questions. The semantic question concerns how to understand the natural language semantics of sentences like $⌜\delta$ is indescribable $⌝$. The explication question concerns whether ineffability can be explicated without internal contradictions that reveal incoherence (van Elswyk 2025).

A widespread assumption in the existing literature made by advocates and critics of ineffability alike is this: if an answer to the semantic question finds that $⌜\delta$ is indescribable $⌝$ is always false, then the answer to the explication question is no. However, this assumption is false. We can see why it is false by revisiting how the falsity of $⌜\delta$ is indescribable $⌝$ results from a performative contradiction and that indescribable is a gradable adjective.

Ineffability predicates result in self-refuting truth-conditions only when anchored to the speaker in the present tense. This is why asserting $⌜\delta$ is indescribable $⌝$ is a performative contradiction. As a result, there is nothing incoherent about predicating indescribability non-performatively. As mentioned earlier, there is nothing self-refuting about $⌜\delta$ is indescribable to Thanasis $⌝$, which anchors the perspective to someone else. Nor is there anything amiss with $⌜\delta$ was indescribable $⌝$ or even $⌜\delta$ will be indescribable $⌝$, which avoid the present tense. The reason why is that the speaker’s present act of describing is not relevant to what is describable for someone else or what is describable for the speaker at different times. So there is nothing like premise (I) in my earlier argument to allow us to reason to a contradiction. The assumption that the unavoidable falsity of $⌜\delta$ is indescribable $⌝$ tells against the coherence of ever predicating ineffability is therefore a mistake. it is a mistake that results from considering only ineffability predicates in performative constructions.

In fact, predicating indescribability non-performatively happens frequently in cross-cultural research into the language of sensory experience. Languages starkly differ in their vocabularies associated with particular senses. For example, Yélı̂ Dnye (Papua New Guinea) lacks terms for any colors other than black, white, and red (Levinson 2000). In contrast, English has more terms than that just for blue such as periwinkle, ultramarine, cobalt, and azure. On the flip side, English lacks a dedicated smell lexicon. We can use general evaluative adjectives like good or bad, or we can comment on the source of the smell such as floral or rotten. After that, our descriptive powers quickly run out. But languages such as Cha’palaa (Ecuador) have 15 dedicated terms that are even incorporated into the language’s grammar as nominal classifiers (Floyd et al. 2018). These fascinating differences make it perfectly intelligible to predicate indescribability of other speakers and to generally talk about the differential ineffability of sensory experience for other speakers (Levinson and Majid 2014).

What these considerations show is that ineffability can be explicated coherently. There is nothing about the concept of ineffability that is confused or internally inconsistent. This is true even though instances of $⌜\delta$ is indescribable $⌝$ are always false. The difference-maker is that assertions made in such cross-cultural research are non-performative. They are not about what is practically available to the speaker conducting the research. Nor do they have any immediate consequences for the speaker. So what remains to be discussed is whether ineffability can be explicated from a first-person perspective, or in a manner that does have consequences for the speaker.

That $⌜\delta$ is indescribable $⌝$ is always false does carry the consequence that $⌜\delta$ is describable $⌝$ is always true for any $\delta$. Such a consequence might seem significant. It seems to bring everything within the reach of description. It does—but only in the most minimal possible way.

As a gradable adjective that is absolute, describable has a closed scale. The threshold for being true of an object is set to the minimum endpoint on the scale. For $⌜\delta$ is describable $⌝$ to be true is for $\delta$ to be at the lowest degree of describability on the scale. The lowest point is very low. Just as $⌜\delta$ is indescribable $⌝$ is performatively self-refuting, $⌜\delta$ is describable $⌝$ is performatively self-verifying. In asserting it, the speaker engages in an act of description with describable. The sentence cannot therefore fail to be true when uttered because describing is a practically available action in that moment. The minimum endpoint, in other words, is describing with describable.

As a result, maintaining that $⌜\delta$ is indescribable $⌝$ is always false only brings everything within reach of the predicate describable. Whether $\delta$ is mostly or entirely describable is a separate issue. The truth-conditions of $⌜\delta$ is mostly describable $⌝$, for example, are markedly different. The adverb mostly moves $\delta$ higher on the describability scale past the threshold set at the minimum endpoint. For that sentence to be true requires much more than one description. In this way, the self-refuting objection establishes very little. It does not show that other descriptions are practically available.

It also does not show that the no answer to the explication question is forced when explicating from a first-person perspective. As long as one’s explication can accommodate the minimum describability that makes $⌜\delta$ is indescribable $⌝$ always false and $⌜\delta$ is describable $⌝$ always true, the explication will be coherent just as it is in cross-cultural research into the language of sensory experience. Of note is that many contemporary attempts at explication can. I will illustrate with two examples from the philosophy of religion.

The first example involves views according to which some class of predicates apply to an object and others do not. An example is Keller (2018). She explicates ineffability as the view that only predicates ascribing extrinsic or negative properties apply to God. Predicates ascribing intrinsic properties do not. Strictly speaking, the falsity of $⌜$God is indescribable $⌝$ is compatible with Keller’s approach. Let’s unpack why. Our sentence is false because describing is a practically available action. But this is compatible with Keller’s view if description is accomplished with predicates that ascribe extrinsic or negative properties. And, in this case, it is. Describing with describable is what is practically available. However, the property being describable fits the bill as an extrinsic property. It is a relational property between an object and a speaker that does not reveal anything qualitative about the object. Accordingly, Keller’s explication of ineffability is not fatally undermined.

Next, let’s entertain views according to which some sentences about an object can be true but not true in a special way. Our example here is Jacobs (2015). He draws a distinction between truth and fundamental truth. A sentence is fundamentally true when it carves reality at its joints. He explicates ineffability as the view that no sentences about God are fundamentally true. At most, sentences can be non-fundamentally true because there is no joint-carving when it comes to God. Again, strictly speaking, the falsity of $⌜$God is indescribable $⌝$ is compatible with Jacobs’s view. Its falsity does not mean that there are fundamental truths. We have explored that its falsity means that $⌜$God is describable $⌝$ is true. But Jacobs does not need to be committed to this being fundamentally true. Nor is it a good candidate for being a joint-carving truth. Jacobs’ explication of ineffability is also not fatally undermined.

This survey of views of ineffability is far from exhaustive. But the compatibility of these explications with $⌜\delta$ is indescribable $⌝$ always being false cements that a no answer is not forced to the explication question. Admittedly, it is hard to avoid saying some self-refuting instance of $⌜\delta$ is indescribable $⌝$ when developing these approaches. But such slips do not reveal a deep incoherence with ineffability. Instead, they reveal the hazard of speaking about the limits of a language with that very language. Taking a step back to theorize about the nature of description or truth more generally, as Keller’s and Jacob’s approaches do, enables an explication that accommodates minimal describability. As a result, neither requires a denial of (I), (II), (III), or (IV).

6. Conclusion

This essay has not followed the advice in §1 from Augustine of Hippo. My approach to quelling the paradox of ineffability has been disputatious. His diagnosis was that the paradox is a “conflict between words.” My diagnosis is that the paradox is a conflict between words and action. Seeing it as such follows from taking the semantics of ineffability predicates like indescribable seriously. That $⌜\delta$ is indescribable $⌝$ is always false and $⌜\delta$ is describable $⌝$ always true is the result of an agentive modal adjective being used in a performative sentence. But the fact that ineffability predicates are gradable is why $⌜\delta$ is describable $⌝$ always being true is nearly trivial. Its truth does not secure that $\delta$ can be described with something other than describable.

A question to end with is why speakers then and now assert $⌜\delta$ is indescribable $⌝$ if it is always false. Though some speakers may think it is not a performative contradiction, or that ineffability can be coherently explicated in some manner, as I have suggested, the natural answer for others is that such statements serve a pragmatic purpose such as revealing piety or personal affection (Stace 1960; O’Meara 2000; Lebens 2017). Whatever the purpose of such assertions, there is much we do with language beyond express truths. It should not come as a surprise that, in grappling with how our powers of description are limited, speakers move beyond semantic content.