Troubles with Trivialism

According to the trivialist, everything is true. But why would anyone believe that? It turns out that trivialism emerges naturally from a certain inconsistency view of language, and it has significant benefits that need to be acknowledged. But trivialism also encounters some troubles along the way. After discussing them, I sketch a couple of alternatives that can preserve the benefits of trivialism without the corresponding costs.


Introduction
Trivialism is the view according to which everything is true. You're reading these words right now-that is true. But the trivialist would insist that it's also true that you are not reading them. After all, everything is true! Trivialism isn't the sort of philosophical view one would arrive at easily. In fact, until very recently, it's a view that hardly anyone would even take very seriously. An incredulous stare would seem enough to dispose of it.
But, of course, incredulous stares are never enough. Once trivialism is on the table, it's not obvious how one can argue against it. The trivialist would agree with all the points that were raised in criticism of the view, just to take them back in the very same sentence-assuming that ''taking back'' is the right description of what someone who believes that everything is true is doing.
In this paper, I examine some of the motivations for trivialism, how it emerges rather naturally in the context of the development of a particular inconsistency view of language, and some troubles that the view faces. I conclude by offering some alternatives.

I. Inconsistent languages, trivialism and dialetheism
Let me start by making two preliminary remarks. (1) We can grant-and we should grant-that English (among other natural languages) is inconsistent. Sentences such as the Liar can be easily and naturally expressed in this language, and the inferences leading from the Liar to a contradiction are perfectly valid. This establishes the inconsistency of English (and other natural languages in which the same steps can be derived).
It is taken by most parties in this debate about the inconsistency of natural languages that there is nothing wrong with the notion of a language being inconsistent. I share this view. There is an extremely idealized notion of language-the notion of a formal language-that derives from mathematical logic, according to which language and logic are, at least in principle, separate notions. A language is taken here mainly as a list of symbols, some syntactic rules of construction of expressions, and some rough and ready rules of interpretation of these expressions. A logic, in turn, is basically a system of derivation of expressions. Given that a formal language is detached from a logic, a language cannot be inconsistent, since it doesn't presuppose a logic. On this conception of (formal) language, it is ultimately a category mistake to state that a language is consistent (or inconsistent). Languages, similarly to tables and hammers, are not the kind of thing that can be consistent or not. Without some logic or other, no consequence can follow from a language, and thus no contradiction follows either. Moreover, languages, as opposed to theories, are simply a list of wellformed formulas, and do not express anything. Languages need to be used in certain ways for them to express something-in particular, an inconsistency.
But clearly, languages, and especially natural languages, can be inconsistent. I just mention two reasons here: (a) We do have the resources to express inconsistent state of affairs in natural languages, and we often end up with such states when we model various aspects of our cognitive tools expressed in natural languages. Obvious examples are things such as belief systems, data sets, fictional discourse, and many theories in the history of science and mathematics (such as the early formulation of the calculus, Bohr's atomic model, and Frege's original reconstruction of arithmetic). Moreover, (b) we do derive results, such as the inconsistency of the Liar sentence, in perfectly acceptable steps in English (among other natural languages). So, on both counts, it's perfectly natural to defend the claim that natural languages are inconsistent. 1 (2) My second preliminary remark is that trivialism should never be confused with a view that is quite different from it: dialetheism. According to the dialetheist, some contradictions (that is, statements of the form: A and not-A) are true (see Priest (2006a) and (2006b)). Of course, if we assume classical logic, and if the language we use has a truth predicate satisfying the usual disquotational conditions and is not limited in its expressive power, the two views are ultimately equivalent. On the one hand, since dialetheism is committed to at least one contradiction, and given that a contradiction entails everything in classical logic (due to the so-called principle of explosion), every sentence in the language follows from that contradiction. With the truth predicate, the statement to the effect that ''everything is true'' follows. Thus, trivialism holds. 2 On the other hand, if trivialism is true, then everything is true, and, in particular, some contradictions are true. Hence, dialetheism is also true.
Of course, the dialetheist realizes all too well that this equivalence holds, and to avoid it, he or she adopts a paraconsistent logic. Given that explosion fails in paraconsistent logics, the argument from dialetheism to trivialism is blocked. But trivialism could emerge from dialetheism in a different way. Suppose that the commitment to the truth of some contradictions leads to the commitment to the truth of all contradictions. As a result, if all contradictions are true, and if the language has a truth predicate satisfying the disquotational conditions, we get, once again, trivialism-even if the underlying logic is paraconsistent. 3 It's not surprising then that the dialetheist explicitly avoids the slide from some contradictions being true to all of them being true (see Priest (2006b)).
Given its implausibility, trivialism doesn't seem to be the sort of view that one willingly decides to defend. Rather, it's a view that unexpectedly emerges given other commitments. Typically, the view arises from the combination of a commitment to classical logic and to the premises of a valid argument leading to a contradiction. Paradoxes are the trivialist's paradise.

II. Trivialist maneuvers
But why would anyone endorse trivialism? According to Jody Azzouni, trivialism emerges from natural language, given that sentences such as the Liar (''This sentence is not true'') can be expressed, the logic is classical, and for familiar reasons, the Liar sentence entails a contradiction (Azzouni (2006) and (2007)). The combination of these assumptions leads to trivialism. The trivialist has a significant benefit on his or her side: the view preserves perfectly well significant features of our practice, and in particular, of ordinary (non-philosophical) responses to paradox.
But trivialism also faces significant costs. Our phenomenology-the way we experience the world and the way we use natural languages-doesn't square well with trivialism. It simply doesn't seem to be the case that everything is true. While you're reading this sentence, it clearly doesn't seem to you that you're also not reading it. The trivialist is aware of the issue, and offers as an explanation the fact that we are often unaware of how our language functions. But even though this fact is correct, to invoke it in this context is to suppose that we live with a massive mistake akin to the Troubles with Trivialism 657 supposition that we could be brains in vats. This contradicts so blatantly our experience (our phenomenology) that a much stronger motivation is needed to suppose that this is what is going on. (I'll return to this point below.) The trivialist insists that we live with an inconsistent language, given that English and other natural languages are inconsistent. But how can the trivialist explain why is it that it simply doesn't seem as though English (among other natural languages) is trivial? In other words, why is it that not every sentence of English seems to be true? What is the reason why we seem to be able to distinguish so clearly true sentences from false ones (except perhaps in cases of vagueness and semantically defective sentences)?
The trivialist recognizes that this is indeed a problem and offers some solutions. Azzouni (2007) insists that speakers are unaware of the semantic theory for their own languages, just as they are unaware of the deep structure of their languages. And so they are unaware of the inconsistencyand hence, given the use of classical logic, of the triviality-of the language they speak.
Even though speakers are indeed unaware of the semantic theory of their languages, this lack of knowledge is not sufficient to explain why natural languages do not seem to be inconsistent-let alone trivial-to speakers. What needs to be explained, in particular, are two crucial aspects of our practice given that natural languages are taken to be trivial, as the trivialist insists: (1) Why is it that speakers do not derive every sentence in the language they speak? (2) What are the mechanisms speakers use to derive only some results and not everything despite the triviality of their languages?
Clearly, the fact that speakers are unaware of the semantic theory for their languages-and hence, that these languages are inconsistent-doesn't account for these features. The speakers' unawareness of the inconsistency of the languages they use doesn't explain why speakers refrain from deriving everything in these languages. After all, even if the inconsistency is pointed out to them, speakers still refuse to derive every sentence from a contradiction. And it doesn't matter whether the speakers are ordinary people or specialists in some area of knowledge. We find the same attitude in either case. For instance, unbeknownst to him, Frege was dealing with an inconsistent system in his logicist reconstruction of arithmetic. Despite this fact, he didn't derive everything from his system, even after Russell pointed out to him that the system Frege crafted was inconsistent. Of course, given that the underlying logic that Frege was using was classical (in fact, it was a particular second-order logic), as soon as Frege realized that his system was inconsistent, he knew that it was trivial, and hence worthless as a tool to distinguish true claims from false ones. 4 The trivialist could insist that Frege's case actually supports trivialism. After all, Frege managed to do exactly what the trivialist predicted anyone in Frege's predicament would do: when dealing with an inconsistent system, Frege didn't derive everything from it. Somehow, he derived the relevant conclusions-although the irrelevant ones, including the negation of every result Frege obtained, were also consequences from Frege's trivial system. The fact that Frege managed to get the correct results is remarkable.
It turns out that Frege's overall reconstruction of arithmetic can be preserved almost completely; the only exception is Frege's derivation of Hume's principle from Basic Law V, which was responsible for the inconsistency of Frege's system (see Boolos (1998)). To explain Frege's impressive achievement simply in terms of his lack of knowledge of the inconsistency of his system cannot be right. Being aware or not of the inconsistency of the system we work with (including the semantic theory for our language) doesn't play any role in our ability to derive results from such a system. After all, the same results are derived in either case. Even when people are aware of the fact that the system they work with is inconsistent, they still fail to derive everything from that system. People don't derive everything from the Liar sentence, despite acknowledging that the Liar is a perfectly meaningful sentence and the steps leading to the contradiction from that sentence are perfectly valid. So, awareness of the inconsistency or lack thereof doesn't change our attitude toward the inferences we draw. And it didn't change Frege's attitude either. After Russell's letter to him, Frege knew his system was trivial and so he tried (unsuccessfully, as it turns out) to provide a consistent successor. Frege operated remarkably well with an inconsistent system. Some other explanation is needed: the one offered by the trivialist is not enough.
The trivialist is invoking a massive form of mistake to account for the fact that natural languages don't seem to be trivial, given that the folk would be systematically mistaken in thinking that not everything is true. To have an account based on this sort of widespread oversight requires a very good explanation as to why such a global error happens, and, in particular, why no one thinks (except for the trivialist) that we could be so massively mistaken when we don't take everything to be true. Any error at that scale demands a powerful, but also plausible, explanation. And in the end, the price that the trivialist has to pay for being able to accommodate ordinary folk's responses to paradoxes is to leave the folk entirely in the dark regarding their own understanding of language.
It seems perfectly acceptable to argue that ordinary people are blind to the deep syntactic structures of the language they speak, since this move doesn't locate folk into any situation of massive mistake, but simply acknowledges a reasonable form of ignorance on their part. It's not that people have entirely false beliefs about the syntactic structures of their language; for the most part, they simply lack belief about the issue altogether. And people are very comfortable in acknowledging that ignorance when the issue is raised to them. The situation, however, is very different with the trivialist's claim that people are blind to the fact that everything is true in the language they Troubles with Trivialism 659 speak-particularly given their explicit denial that this is the case! A massive mistake at that scale demands some form of explanation.
Is such a gargantuan form of error even possible? The trivialist's scenario of a massive mistake resembles, of course, global forms of skepticism, which are invoked to challenge the belief that we know anything about the world. But there is a significant difference here. Instead of arguing, as the traditional skeptic is taken to do, that everything we believe may turn out to be false, the trivialist insists that everything is true-even though it doesn't seem that way to the rest of us.
There are, of course, many responses to global skepticism. One of them, in particular, is worth considering in this context. This is the response offered by Donald Davidson, who invokes a ''coherentist'' account of knowledge to undermine the global skeptic's challenge (Davidson (1989)). Although I don't think that Davidson's response to skepticism ultimately succeeds (see Bueno (2005)), perhaps a version of Davidson's argument can be used to dispute the reasonableness of the trivialist's approach.
Davidson first insists on the veridical nature of our beliefs: most of our beliefs are true. And by this, he doesn't mean that everything is true, but just that our beliefs are-and not necessarily all of them, only most. As Davidson puts it: The agent has only to reflect on what a belief is to appreciate that most of his basic beliefs are true, and among his beliefs, those most securely held and that cohere with the main body of his beliefs are the most apt to be true. (Davidson (1989), p. 153) When Davidson claims that most of our beliefs are true, he is not asserting that the beliefs in question are also false-although, given a fallibilist picture, each belief can be false. In Davidson's own words: I think the independence of belief and truth requires only that each of our beliefs may be false. But of course a coherence theory cannot allow that all of them can be wrong. (Davidson (1989), p. 140) The global skeptic then faces an awkward predicament, since it's not reasonable to ask for additional justification that the beliefs an agent has are true.
The question ''How do I know my beliefs are generally true?'' thus answers itself, simply because beliefs are by nature generally true. Rephrased or expanded, the question becomes, ''How can I tell whether my beliefs, which are by nature generally true, are generally true?'' (Davidson (1989), p. 153) As a result, the skeptic's objection is just beside the point.
What is Davidson's anti-skeptical strategy then? The idea is that the skeptic cannot consistently claim that all of our beliefs can be false. Such a claim would be incoherent, given the veridical nature of beliefs. Although Davidson notes that each belief individually can be false, it's not consistent to maintain, as the skeptic does, that all beliefs can be false. After all, beliefs are, by nature, generally true.
Whether Davidson's argument works against skepticism or not, a similar argument can be used to resist the tenability of trivialism. If beliefs are, by nature, generally true-although individually each of them can be falsethen a system of beliefs cannot have all of their components false. This, however, doesn't mean that everything is true, since a system of beliefs, most of which are true, is not one in which the beliefs in question are false as well.
It might be said that, as an argument against trivialism, this move begs the question, since it assumes the veridical nature of our beliefs, and excludes that such beliefs are also false. But for the trivialist everything is trueincluding the claim that everything is false. So, for the trivialist, everything is both true and false.
I think this response is ultimately correct. Similarly to what happens with global skepticism, once the skeptical scenario is in place, it's unclear that we can effectively refute the skeptic. For similar reasons, it's not clear to me how we could refute trivialism. What we can do, and what the argument just offered is supposed to suggest, is that trivialism is not a plausible, tenable position, since it requires from each belief that it's taken to be true and false, contrary to the nature of beliefs.
At this point, the trivialist may suggest an additional sort of explanation (see Azzouni (2007)): the regimentation move. Although a natural language is indeed inconsistent (in fact, trivial), part of the language can be regimented in a consistent language whose underlying logic is classical. In this way, given that we understand well how the regimented language works, we also understand well the workings of the natural language in questionor, at least, of its consistent part.
The trouble with that move is that the trivialist is creating a parallel discourse via regimentation, and it's not clear that the regimentation actually helps in explaining how the natural language under consideration actually works. After all, given that the latter is inconsistent, there are parts of that language that are in fact eliminated in the regimentation, and thus are not accounted for in the regimented language. But the behavior of the inconsistent part of the language is precisely what we need to understand. Simply eliminating the inconsistency of the language won't help here.
Moreover, the regimented language, being consistent, is expressively poorer than the ordinary language we speak. This cost is typically associated with consistent approaches to paradoxes (see Priest (2006a)). The fact that an inconsistent view ultimately has this cost is unexpected. Ultimately, the Troubles with Trivialism 661 trivialist view ends up both endorsing inconsistencies (in the natural language) and incompleteness regarding expressive power (in the regimented language). For a view that presupposes classical logic, we have costs on both fronts.

III. Alternatives to trivialism
There are many alternatives to trivialism. Here I'll sketch two of them, and argue that they offer the benefits of trivialism without the cost. Both strategies have the naturalness of trivialism and its capacity to take natural languages literally. 5 Despite this, the strategies don't have the main cost faced by the trivialist proposal, since they don't take everything to be true.
The first alternative is the compartmentalization strategy. Following the trivialist suggestion, suppose that when we use a natural language, e.g. English, the logic we adopt is classical. In this case, given certain perfectly meaningful and acceptable English sentences, such as the Liar, we immediately realize that a contradiction follows. To avoid the contradiction, we should reason carefully, and prefix each inference in such a way that contradictory premises are never used together. In this way, we can keep classical logic, but we need to specify explicitly the context in which we apply each inference rule. We also need to prevent ever using inconsistent premises simultaneously; otherwise given the use of classical logic, we would obtain triviality.
The compartmentalization strategy resembles in various ways Azzouni's own proposal (see Azzouni (2006)). 6 First, similarly to Azzouni's move, the underlying logic of the compartmentalization strategy is classical. Moreover, Azzouni recognizes that careful ordinary people do tend to compartmentalize their reasoning strategies when they face inconsistent contexts. Instead of deriving everything, they separate the assumptions that together would yield a contradiction, and are particularly wary of not employing such assumptions simultaneously.
The compartmentalization strategy (or, as we can also call it, the strategy of tagging premises) consists in indicating explicitly the assumptions involved in each derivation, so that a context is determined, and in each context, inconsistent premises are never used in conjunction nor are selfcontradictory sentences (such as the Liar) ever invoked. In this way, even though the language employed is inconsistent, no derivations are allowed from inconsistent premises. Every sanctioned inference is classical, and in this respect, the underlying logic is classical as well.
However, since nothing is allowed to follow from a contradiction, it might be argued that the underlying logic is actually paraconsistent. After all, no longer can we infer an arbitrary statement B from a contradiction of the form A and not-A. The inference that sanctions that conclusion-namely, explosion-is, of course, valid in classical logic, and as is well known, explosion's invalidity is a distinctive feature of paraconsistent logic. This suggests that the logic turns out to be paraconsistent in the end. Now, this response, however plausible, is not entirely cogent. The compartmentalization strategy offers a mechanism for speakers to use classical inferences in inconsistent contexts without triviality. In this sense, the strategy provides a pragmatics for inference making. However, it doesn't offer a semantics for the language in use. From a semantic point of view, the logic would indeed be paraconsistent, given that in effect one fails to derive everything from a contradiction. For this reason, the compartmentalization strategy turns out to be different from Azzouni's proposal, which is firmly committed to classical logic and hence, given the presence of inconsistencies in the language, to trivialism.
The second alternative to trivialism is the paraconsistency strategy. This involves the full-fledged use of paraconsistent logic in every context, whether inconsistent or not, including the contexts in which we provide the semantics for our own language. To obtain verbal agreement with the trivialist's insistence that the underlying logic is classical, this strategy highlights the fact that, in consistent contexts, exactly the same inferences are sanctioned by classical and paraconsistent logics. So, there is no disagreement between the classical and the paraconsistent logicians in such contexts. As a result, there is also no disagreement between the trivialist and the paraconsistent theoretician regarding the use of classical logic in consistent contexts-exactly the same inferences are sanctioned. The disagreement emerges from the trivialist's claim that everything is true: this is not something that the paraconsistent theoretician-and pretty much everyone else for that matter-is prepared to grant.
Note that the paraconsistent strategy does not assume dialetheism, the claim that there are true contradictions (see Priest (2006a)). We can obtain all the benefits of the use of a paraconsistent logic without the costs of having to believe something that clearly hardly anyone is inclined to believe (namely, true contradictions). After all, dialetheism is neither required nor guaranteed by the use of a paraconsistent logic. One can just be agnostic about the existence of true contradictions, and simply use a paraconsistent logic as a device of inference making, without having any commitment to the truth of any contradiction that may emerge.
In this way, we can also accommodate the overall impression that the logic that is used in most contexts is classical. Other non-classical logics tend to be used only in very special situations. For example, if we need to reason about incomplete but consistent situations, we can use some constructive logic. If we need to reason about some situations where non-distributive conditions are involved, some quantum logic would be adequate. But in most contexts, where we reason with consistent, complete, and distributive situations, the use of classical logic is perfectly adequate. As a result, the Troubles with Trivialism 663 paraconsistentist is able to accommodate the widespread use of classical logic without triviality. This is not the place, of course, to articulate this version of paraconsistency further (for some details, see, e.g., da Costa, Krause, and Bueno (2007)). The point is just to indicate that there are alternatives to trivialism that preserve the benefits of the view without its costs. We can acknowledge, with the trivialist, that natural languages are indeed inconsistent, and so is the semantic theory for such languages (see Patterson (2007)). Moreover, there is no need for sacrificing the expressive power of these languages to make them consistent, which inevitably happens with consistent approaches to paradoxes (see Priest (2006a)). But, with a paraconsistent logic in place, we need not pay the price of accepting that everything is true. Now, how should the issue of the underlying logic of a natural language be addressed? The issue is delicate. The paraconsistent theoretician can grant the trivialist that, to the extent that there is an underlying logic in natural languages, it seems to be classical-at least in most contexts. But the paraconsistentist will also highlight the fact that speakers often tend to find explosion an invalid inference, given that the arbitrary statement in the conclusion of explosion has nothing to do with the contradiction in the premise. In other words, there is no relevance relation between premise and conclusion. Of course, the speakers' sensitivity to relevance considerations such as this-precisely at an inconsistent context-is an indication that the underlying logic need not be classical. If we use a paraconsistent logic, however, we can understand perfectly well such a response. In fact, this response is exactly what we would expect.
This doesn't establish, of course, that the underlying logic of a natural language is paraconsistent. As should be obvious already, it's not clear to me that there is only one logic in any particular context. Given finitely many pieces of reasoning, there will always be distinct logics that sanction these pieces of reasoning as valid, but which differ in the assessment of other pieces of reasoning (see Bueno (2002)). To illustrate this point, consider the case of paraconsistent logic. As is well known, there are several nonequivalent paraconsistent logics (for a survey, see da Costa, Krause, and Bueno (2007)). And in consistent situations, exactly the same inferences are sanctioned by classical logic and all such paraconsistent logics. However, in inconsistent contexts, significant differences emerge. Classical logic is trivialized by any contradiction; that is, if a contradiction is added to classical logic, the resulting system is trivial. But no paraconsistent logic is so easily trivialized. Consider the family of paraconsistent logics known as C n -logics, 1(n(v (da Costa (1974)). This is a sequence of progressively weaker logics, such that each logic in the sequence is trivialized by a certain form of contradiction; but contradictions of that form fail to trivialize the logics that come later in the sequence. The latter logics, however, are trivialized by contradictions of another form that, in turn, don't trivialize the logics that come still later in the sequence. 7 Until we get to an extremely weak logic, C v , which is not finitely trivialized. As a result, each paraconsitent logic in the sequence is importantly different from the others. The point here is to indicate that there is a range of alternatives to consider, since there is a family of (in fact, infinitely many) paraconsistent logics that would be adequate to the task of helping us understand the inferences made in inconsistent contexts.
As a result, it's not clear that there is only one logic of natural language. But the paraconsistent theoretician who acknowledges the plurality of logics would be able to accommodate this fact, including the appearance that, in most consistent contexts, the logic used is classical. We thus have alternatives to trivialism that don't force us to believe that everything is true, while still acknowledging the inconsistency of natural languages.

Conclusion
As we saw, as part of an inconsistency view of language, trivialism has some significant benefits. The view accommodates perfectly well the response that folk have to paradoxes such as the Liar (namely, they acknowledge that there is something strange going on, but ultimately ignore the paradoxes, leaving others to grapple with them). It also accommodates well the alleged fact that the underlying logic of natural language is classical. But trivialism also faces some significant costs. First, it demands a commitment to everything being true in natural language (something not so easy to swallow). Second, the explanation of why it just doesn't seem that way (that is, why it doesn't seem that everything is indeed true) is not very plausible, given that it introduces a massive form of mistake. Third, to avoid these troubles, a regimentation of natural language is offered, which assumes classical logic, and is thus expressively poorer than the natural inconsistent language.
As an alternative, I offered some possibilities that were meant to preserve the benefits of trivialism without the costs. Both alternatives that were suggested involved, in one way or another, considering logics other than classical. But because it's not clear that we can ever settle the issue of which logic is actually being used in a given context (given the plurality of logics compatible with the inferences in that context), this move seems well motivated. In the end, in all consistent contexts, classical logic will work just fine-exactly as the trivialist would tell us, but without everything being true. 8

1.
The inconsistency of natural languages is, of course, a point recognized by Tarski (see Tarski (1936)).