The Liar's Paradox

This is a paradox that, as I, a logic purist (If such a label exists anywhere) have read, troubles me, for it seems to strike a jab at one of the three most foundational laws of logic, which is that of noncontradiction (A or not A, or formally, A v ~A). If such a principle of logic was not necessarily true, or else, in every instance of a possible world, it does not hold in at least one of them, then utterly absurd things about reality could be true, such as that it is possible that I am writing this post and that I am not writing this post at the same time. Indeed, it would, if the law of noncontradiction is not necessarily true, be possible for me to be both fat and not fat, for the sun to also not be the sun, and for me to know English (Necessary for me to write this very sentence) and to not know English (And so I shouldn’t be able to write this sentence), both simultaneously with one another, not merely as a sequential change of state from, say, Not A into A (For example, for me to go from not knowing English to knowing English). This seems completely wrong, and it should be. But then here comes the liar’s paradox, stated thus:

This sentence is false.

Interesting, for if it is the case that it is true that the sentence is false, then it is false despite it being true. And if it is false that the sentence is false, then the sentence must be true, despite it being false, which is an outright logical contradiction, for this means, then, that the sentence is, depending on what truth value you assign to it, both true and false at the exact same time. What should follow from this, then, is that certain propositions, such as ā€œThis sentence is falseā€, do not follow the law of noncontradiction, and so cannot simply be either true or not true, and if this is the case that some propositions don’t follow the law of noncontradiction, then it follows that the law of noncontradiction does not hold in every possible world, and if the law of noncontradiction does not hold in every possible world, or else is contingently true, then it is possible, in some world, for me to know English and to also not know English. Yikes.

But something seems wrong with this Paradox, and it can be unearthed if we dissect what ā€œThis sentence is falseā€ actually means. And we can do it with this simple question: What is ā€œThis sentenceā€? Is ā€œthis sentenceā€ the full proposition as stated, ā€œThis sentence is false,ā€ which can be expressed as A, or else is it ā€œThis sentenceā€ by itself, expressed as A, and then the ā€œis falseā€ expresses a negation of ā€œThis sentenceā€ (A), which would look like ~A? If it is the former, then what would be true is that if it is true that ā€œThis sentence is falseā€, or it is true that A, then all that would happen is that it turns out that A is asserted to be a true proposition, and that the only problem, then, would be to characterize what A actually means, or else it would be up to us to characterize what the contents of A have, or what their predicates are, to determine if the assertion that A is true is either a true assertion or a false one. For example, if I assert that it is true that I dropped my ice cream, then it would be up to others to determine whether or not my assertion of the truth of me dropping ice cream, or else the truth of me declaring A, is a true assertion or not a true assertion, on the basis of the existing evidence in support of the assertion, and this would not be a contradictory, nor an unreasonable, way to characterize ā€œThis sentenceā€. Now if the latter possibility is true, which is that ā€œThis sentenceā€ means the first part of the sentence, ā€œThis sentence,ā€ is the proposition, or A, and that the last part of the sentence, ā€œis falseā€ is a negation of the first part of the sentence, or A, then ā€œthis sentenceā€ can be expressed as Not A, or formally, ~A. So if we defined ā€œthis sentenceā€ this way, then if we assert that ā€œthis sentence is falseā€ is true, then we would be saying, instead, that it is true that Not A, or it is true that ~A. Therefore, if defined this way, then this would be analogous to me asserting that it is true that I didn’t drop my ice cream, or it is true that Not A, or ~A, and this would not be a contradictory way to define ā€œthis sentenceā€.

What I think follows from this argument, then, is that the liar’s paradox, which is ā€œThis sentence is false,ā€ is not a true instance of a proposition that can violate the law of noncontradiction, but instead is an instance of a proposition that, depending on how specifically it is defined, can lead to instances where the truth value assigned to the specific definition can be preserved or negated. Therefore, since definitions aren’t fixed, but the objects which propositions can be assigned to are fixed, then the paradox seems, from this, to be a larger problem in linguistics than it does for logic itself.

To be frank, writing this post kind of melted my brain, and I’ve not looked into other solutions to the problem very deeply other than the problem itself, so if something seemed to be of question in this post (This is my first post), then say something.

1 Like

What does it mean for a sentence to be true or false?

ā€œIt is rainingā€ is true if it is raining and false if it is not raining. Can we explain what it means for ā€œthis sentence is falseā€ to be true or false in a similar way, i.e. without using the words ā€œtrueā€ or ā€œfalseā€? If not, I’d question its very coherence; it’s gobbledygook that only appears to be saying something.

2 Likes

That is basically the point being made. Unless we are allowed to press into what ā€œThis sentenceā€ even means, then ā€œThis sentenceā€ could, in truth, be referring to three different possible things at the same time:

  1. It could be referring to a proposition outside of itself, such as if I said, ā€œI played dominos,ā€ and the sentence would be saying, in effect, that ā€œThis sentence is falseā€, or else saying that I did not play dominos, and this is, if defined this way, not incoherent if this is the case.
  2. ā€œThis sentenceā€ literally refers to ā€œThis sentence is falseā€, because it is a full sentence, and because the liars paradox, taken as is, doesn’t generally refer to anything else outside of itself. But then if this is true, then ā€œThis sentence is falseā€ would be a whole proposition, and so would be formally expressed as A. So if someone were to say that ā€œThis sentence is falseā€ is true, then they would merely be saying that A is true, which is not a contradiction.
  3. ā€œThis sentenceā€ could refer to the first half of the sentence, ā€œThis sentenceā€, and so that would be the full proposition, or A, and then the second half, ā€œis falseā€, would be an informal expression of a negation, or a Not, or a ~, and so, put together in this way, ā€œthis sentenceā€ would, in formal expression, be referring to say Not A, or ~A, and so defined this way, then to say that ā€œthis sentenceā€ is true, would be to say that ~A is true, and if defined this way, this would not be incoherent.

What this is getting at, then, is that the liar’s paradox works only if you permit ā€œThis sentenceā€ to remain ambiguously defined, because if done so, then ā€œThis sentenceā€ is equally capable of saying A, or Not A, which is why the contradiction seems to emerge if you allow it. It is a problem regarding semantics and word meaning, more than some issue with logic itself, and this, in a sense, demonstrates why equivocation is such a serious fallacy in Philosophy, because undefined words can cause serious problems in preserving an argument’s truth value, if not clarified up front. Like that time I spent 6 posts trying to debate the semantics of ā€œGodā€ with you, I hope this comes to show why semantics is so important.

The point I was suggesting is that saying that ā€œthis sentence is falseā€ is true is like saying that ā€œbish bash boshā€ is true.

What do ā€œthis sentence is falseā€ and ā€œbish bash boshā€ even mean? If we can’t explain this then we don’t even reach the point where we have to consider whether or not them being true is a contradiction.

I’m going to adopt a convention where (1), (2), and so forth serve as names to the sentence which follows them.

(1) – (1) is false.

That’s pretty clear, no?

I’m not sure I’d go this far. I’d rather say that the liar’s sentence shows a paradox about self-reference, and points to logics which drop the PNC in favor of some other method of stopping explosion, which is what you’re referring to here.

Explosion is undesirable because it leaves us with little to evaluate after.

But if you can contain explosion then perhaps you could have a class of sentences called dialethia. These are the sentences which are both true and false.

Now how this stops your ice cream example is we could claim that ā€œI dropped my ice creamā€ is bounded by the circumstances we’re asserting which, in this case, includes the LNC.

But dialethia are the sorts of sentences which clearly do not.

A self-reference to what?

And the point I would be making here is that, because the contents of the liar’s sentence is ambiguous, there are three equally plausible definitions as to what the liar’s sentence self-refers to, specifically, and that each one is capable of agreeing with, or negating, the truth value assigned to it. This is not a problem with logic, specifically, but with language itself, as language cannot perfectly map out the rules of logic, due to certain words and phrases being multidefinitional, and the liar’s sentence is one sentence which is not only multidefinitional, but a multidefinitional sentence where the different possible definitions are the opposite of themselves, or A and ~A, depending on how it is defined. This is an issue with semantics, not just the rules of logic, such as the law of noncontradiction.

True and false simultaneously? Words, and their possible definitions, are merely a possible way which a word can be validly defined, not one which is actually the case at the same time as other definitions. This is why Wittgenstein’s concept of ā€œword gamesā€ is very important, because the context which surrounds a specific word or phrase can actualize one possible definition over another, because such a definition is the one that applies more appropriately to the rules of the specific ā€œgameā€ being played, such as the word, ā€œWater!ā€, meaning a command to pour water on a POW who is being waterboarded if someone is carrying out a torture session, over being a signal that someone found potable water on a desert island, and needed to inform his friends; one use of the word, ā€œWater!ā€, is correct over the other one. And because the liar’s sentence has no context, and because it is multidefinitional, then this is why it is apparent that it is at the same time true and false, because there is no correct definition of the Liar’s sentence which can be actualized by context, unless you decide to probe into what the liar’s sentence means, independently of context.

I ain’t no linguist, nor why language possesses this property, but I don’t see it truly proving the law on noncontradiction to be incorrect, because language maps onto logic, and not the other way around, so if the liar’s paradox exposes a problem regarding language, it doesn’t therefore follow that it exposes a problem regarding logic, even though language is capable of replicating logic, correctly, very much of the time. I wonder what you say of the matter yourself, since you seem to know about the liar’s paradox quite well yourself.

It seems to me, it’s almost always interpreted like in 2. But you haven’t solved the contradiction.

Right, so A is ā€œThis sentence is falseā€ and ā€œthis sentenceā€ refers to the whole proposition, i.e. A so A is ā€œA is falseā€. If you say A is true then ā€œA is falseā€ is itself false. But A is ā€œA is falseā€ so A is false. You have the contradiction.

To itself.

Where you say:

I’m saying it’s not so ambiguous as all that.

ā€œThis sentenceā€, all by itself, cannot be true or false because it is not a sentence. And since it is sentences which are truth-apt, rather than nouns, ā€œThis sentence is falseā€ refers to the sentence being uttered, written, heard, or read.

Yes.

In analogue consider Russell’s Paradox which has a similar trick going on in that the set refers to itself.

I think that’s something the liar’s points out: there is something funny about our reasoning when we include self reference and are consistent.

The context then, to answer:

The context is philosophy so I figure it’s fairly apparent that we’re talking about some philosophical problems related to the liar’s. One of those is whether or not there is one true logic, which I don’t think there is so the liar’s is useful for pointing that out (we make up the rules of logic). The other is self-reference.

Even if we reduce the problem to one of semantics I could just stipulate that I’m speaking in the context of logic and philosophy so, in this case, the sentence ā€œThis sentence is falseā€ refers to itself by stipulation.

Nice instance of cherry picking BTW. If you read the argument just below this quote, you would see why this is relevant to the liar’s paradox, because just below this quote, I say:

So, ā€œthis sentenceā€, if defined as 2, as you said it is often done so, is defined this way, then it would simply become a basic proposition, A, not A is false (As the ā€œis falseā€ is a part of the full proposition here), and so when someone would say that ā€œthis sentenceā€ is true, they would only say that A is true, and so is not a contradiction. Please observe surrounding context in further criticisms, rather than direct ā€œgotchaā€ responses my way.

I may have misunderstood what you meant but I did not cherry picked anything. Sure, if someone were to say that ā€œThis sentence is falseā€ is true, then they would be saying A is true. I agreed with all of that, and just continued the analysis.

What is A? A is ā€œthis sentence is falseā€ but we can continue further and say A is ā€œA is falseā€ as we know ā€˜this sentence’ refers to the whole proposition, i.e. A. That’s what I am saying.

Presumably, if our domain includes all sentences, it includes ā€œThis sentence is falseā€.

If meaning is understood extensionally, it is unclear what the extension of either ā€œThis sentence is trueā€ or ā€œThis sentence is falseā€ would be - either might be true or false.

One approach is simply to say that not all sentences are either true or false. This is to deny the excluded middle, giving us a non-classical logic.

Another approach is not to assign truth values in the language in question, but to keep them in a metalanguage, a language about the target language. This is the approach adopted by Tarski and much of model theory, avoiding the Liar by rendering ā€œThis sentence is falseā€ inadmissible, since a language cannot contain its own truth predicate.

Another approach, from Kripke, is to permit a language to contain a truth predicate, but to assign truth values in a strict sequence. Here’s an example:

Suppose we restrict the language to being about a group of people, Adam, Bob and Carol… an so on, and their respective nationalities, English, French… etcetera. We can construct any number of sentences from these: ā€œAdam is English", ā€œBob is Englishā€, ā€œAdam and Bob are frenchā€ā€¦

We start by assigning ā€œtrueā€ and ā€œfalseā€ to statements of our language on the basis of an extensional interpretation. We assign ā€œtrueā€ or ā€œfalseā€ to these as appropriate; so in our interpretation, ā€œAdam is Englishā€ is true, and ā€œAdam is Frenchā€ is false, and so on.

Notice that so far any sentence that contains the term ā€œtrueā€ is not yet assigned a truth value. So ā€œā€˜Adam is English’ is trueā€ is at this stage neither truth nor false.

We then start to permit sentences that contain ā€œtrueā€ or ā€œfalseā€ to be assigned values, but under strict conditions. So:

  • If ā€œAdam is Englishā€ is true, then we allow that ā€œā€˜Adam is English’ is trueā€ is also true.

  • If ā€œAdam is Frenchā€ is false, then we allow that ā€œ'Adam is Frenchā€ā€™ is false" is true.

And so on. Generally, if p is true, then ā€œp is trueā€ is true, and ā€˜ā€œp is trueā€ is true’ is true, and so on; if p is false, then ā€œp is falseā€ is true, and ā€˜ā€œp is falseā€ is true’ is true, and so on.

But notice that in this construction, we never get to assigning a truth value to the sentence ā€œthis sentence is falseā€. So it remains without a truth value. The Liar is thus not considered inadmissible so much as un-modelled.

Yet another approach is to accept that the truth value flip-flops between true and false, and build a logic that incorporates this as a process. On this account we might begin step one by assigning ā€œtrueā€ to ā€œThis sentence is falseā€. Since if ā€œThis sentence is falseā€ is true, the sentence is false, our second step is to assign ā€falseā€ to ā€œThis sentence is falseā€; and again the consequence is that the sentence is true, so in step three we again assign ā€œtrueā€ to ā€œThis sentence is falseā€; and so on. This may not be a solution so much as an explication, and interestingly has been formalised by Hans Herzberger and Anil Gupta.

Compare an alternative, ā€œThis sentence is trueā€. If we begin by assigning ā€œtrueā€ to ā€œThis sentence is trueā€, our conclusion is that the sentence is true, and the process of revision ceases. And if we begin by assigning ā€œfalseā€ to ā€œThis sentence is trueā€, then our conclusion is that ā€œThe sentence is trueā€, is false. Here there is a problem, which encourages the revision of assigning ā€œtrueā€ to ā€œThis sentence is trueā€. What the revision process shows that assigning ā€œtrueā€ to ā€œThis sentence is trueā€ results in stability, but assigning ā€œfalseā€ results in a need for revision.

So, four differing approaches: adopting a non-classical logic; adopting a Tarskian hierarchy; adopting Kripke’s assignment of truth values in stages; and adopting a revision theory of truth. What I’d like to point out is that the choice here is not forced— we can choose whichever of these approaches suit our purposes.

4 Likes

What you may have captured here is the observation that the paradox does not lie in the syntax of the logical system adopted, but in the semantics, in the interpretation. However, in formal logic, definitions are fixed by being explicitly stipulated. Above, I’ve given a few examples of how a formal system can accommodate assigning truth values in a consistent fashion.

There is a distinction to be made between propositions, which are sentences which carry a truth value such as true or false, and other types of sentences, such as ā€œWhat time is it?ā€, which cannot be declared as either true or false, because such sentences fail to make a truth claim. What doesn’t seem to follow from this, though, is that if it is the case that not all sentences carry a truth value, but the reason why is because not all sentences make truth claims, then the mere existence of non-propositional sentences does not seem like a sufficient ground to reject the law of excluded middle, and so I don’t see, personally, how granting that not all sentences can be true or false justifies that the law of excluded middle can be reasonably abolished. Unless you meant by this that not all propositions can be true or false, which would be a very different story.

Yes, but that is the problem with the liar’s paradox, and where my critique is really trying to hone in on right now, is that the liar’s sentence seems to do a very poor job at characterizing itself, even though, with all of the responses and feedback I’ve received thus far, there seem to be four equally valid ways to characterize the liar’s sentence, either as a negation of some other proposition, a simple proposition, a negated simple proposition, or else the liar’s sentence is not a proposition at all, but an empty sentence, and so cannot coherently be assigned any truth value at all without contradiction. This argument I’ve been setting forth has been more of an internal critique of the problem itself than it is an argument where I grant that the liar’s paradox holds, and then see what follows from it, and its effect on how logic is to be conducted in the future, as you seem to have done so in your responses. Categorically, you seem to approach the problem in a different way than I do.

Indeed, at least in natural languages; less so in most formal languages, where a sentence usually equates to a statement. There’s of course a long tail of discussion here.

Another approach is simply to stipulate that a proposition is a sentence with a truth value, or if you prefer, what is said by a sentence that has a truth value. If we adopt that approach, then ā€œThis sentence is falseā€, if it does not accept a truth value, is not a proposition. A good reason for not talking of propositions.

I was following Davidson out of habit, using ā€œsentenceā€ rather than reifying propositions as entities. I’ll continue to do so, and trust you to understand the implications.

There are logical systems that make use of the excluded middle, and others that do not. I don’t see a reason to prefer this system or that one, except for what is useful in the context of our conversation. So I’ve offered four different systems that each avoid the issues presented by the liar. Take your pick, or come up with an alternative.

At least in the last three approaches, the liar can be characterised formally.

For Tarski, the Liar requires a predicate having a truth value within its own language, which simply cannot happen given the way Tarski defines truth in terms of satisfaction.

For Kripke, the Liar remains ungrounded, never being assigned a truth value.

For Hans Herzberger and Anil Gupta, the Liar never stabilises over revisions. This last preserves classical logic while explaining the instability quite directly, which is why it is worth considering.

Tarski will not admit the Liar. Kripke admits it by refusing to evaluate it. Revision admits the liar and sets out the problem. That’s not far at all from what you have been saying, as I’ve understood.

The liar paradox, just as Russell’s paradox, just shows our false assumptions or false premises about logic, when we generalize something (like a set) to contain everything and do not take into account negative self-reference (which, in my view, is the most awesome rule in logic that we still dismiss and do not understand even it’s extremely crucial for logic to be consistent).

Put it short, it is about what happens in logic with negative self-reference.

Remember that the original Liar paradox was the Epimenides paradox, where Epimenides, a Cretan, says ā€œAll Cretans are liarsā€.

OK, if Epimenides would have been an Athenian, there’s no problem. But now as he belongs to a group of Cretans, he is, as a Cretan, doing a self reference. Ok, still no problem. But when talking about ā€œliarsā€, there’s the negative self-reference. As a Cretan, then he is a liar.

So the outcome is totally similar like with Russell’s Paradox with idea of there being an ā€œset of all setsā€, and then using negative self-reference and getting a set, that does not contain itself. So we have a paradox, because there are sets and we can imagine all of them… but then that isn’t a set, but a very paradoxical set!

Before you comment anything, here me out, here is the moment where many people take go down the futile and useless rabbit hole of discussing liars can tell truths or not. This is about simple rules of logic, not a World hugging debate about liars or just what truth of false just means. And let’s leave also Russell’s paradox and naive set theory aside here.

Let’s look first how powerful negative self-reference is with an example that doesn’t end up with a Paradox. I’ll give an example.

Statement 1: People can write in the Philosophy Forums any finite response they want in PF. (Only afterwards the administrator can erase nonsensical or comments that are against the rules.)

The above statement seems obvious at first. But it isn’t true, for many reasons, like we have only a finite time that we are live etc. Yet it is also logically for the logical limitation that negative self-reference gives us. Why?

Because using the negative self-reference rule, we can also say that:

Statement 2: People cannot write a response in the Philosophy Forum which they do not write.

Now, are there responses that people don’t write here? Yes. You cannot write something that you don’t write. I cannot either. @Banno cannot write something that he doesn’t write. But obviously you can something that @Banno doesn’t write, and also you and me can copy him. Is this a limitation on what we can write? Surely not, but in some situations can be.

Now it really sounds incredibly simply (or stupid) to say:

Statement 3: People can write in the Philosophy Forum anything, except what they never will write.

But that still is true!

So why do we have the Paradox or Antinomy, as Kant put it? I think it is because we absolutely hate limitations and see something like the paradox as an obstacle to solve. Just remember how devastated Russell along everybody else was after the re-emergence of the paradox. It isn’t something to be solved, but simply something to be understood.

There are things you might have written, but didn’t, and never will. so ā€œPeople can write in the Philosophy Forum anything, except what they never will writeā€ looks in this sense false. One might have written what one didn’t write. We need here to be clear as to whether we are using a existential quantification, as you are, or a modal quantification, as I am. I think the instability here comes from modality, not truth.

The Liar is not dependent on indexicality, as ā€œI am a Cretanā€; nor modality, as in that we might have written something other than we do write; nor self-reference. All good points, since each of these can be used with without contradiction or paradox.

There are versions of the Liar that do not rely on self-reference. The obvious one is

  • (A) Sentence B is false
  • (B) Sentence A is true

Yalbo’s paradox is even more fun, since no sentence ever ultimately refers to itself, nor is it clear that there is some overall circularity.

It’s worth here again drawing parallels to my discussion of Aesthetics. There I proffered the argument that whatever definition of art one might offer, some artist will produce an item of art that undermines that definition. It may be that whatever general solution or analysis of paradoxes is offered, some logician will produce a paradox that bypasses that analysis.

At the least, no simple diagnosis of paradoxes is apparent.

I think here ā€œwhat people writeā€ would refer to everything people do write in the entirety of their lives. That is a large yet still finite amount of writing and doesn’t take into account the temporal aspect of this (which you think about when thinking about what they might but didn’t). Obviously what people write is something we couldn’t even list as long as the people are interacting with the world and writing.

To refer to modality more like the question if liars can also say something true every once and a while… Hence Russell’s paradox is more easier to understand as there isn’t the problem of the complex nature of physical or real life examples, which can be understood in many ways.

Yet both sentences reference each other, right?

Yep. The existential quantification - there is indeed stuff that I will not actually write. The modal quantification remains - there is stuff that I will not actually write, but which I might have written. I’m just pointing to the need to seperate out the two, for clarity.

Yep. Not so in the Yalbo case.

Huh.

There’s always more to learn that you didn’t know. Thanks for the reference.

1 Like

Interesting, have to attempt to understand that one.

1 Like