Soundness is Neither Necessary nor Sufficient for Goodness

In this (very) short article, I am going to try to explain what makes an argument valid (comparing two views), what makes an argument sound (again comparing two corresponding views) and then I aim to distinguish ‘good’ arguments from either of these. I will attempt to obviate why validity, on either interpretation, will be a necessary but insufficient condition for soundness (on that respective interpretation). It will turn out that soundness (on either interpretation) will not be a sufficient or a necessary condition for goodness, and that validity (on either interpretation) will be a necessary but insufficient condition of goodness. It will also be shown that goodness is neither a necessary nor sufficient condition of soundness. This article can, perhaps, serve as a useful prolegomenon to introductory deductive logic, though its distinctions are themselves somewhat unorthodox and reach beyond the scope of formal logic.


One definition of validity which is relatively common, easily found in most introductory textbooks on deductive logic, is the following:

An argument is valid if and only if it is not logically possible for the premises to be true and the conclusion false.

Let’s explore the dynamics of this definition. It would mean that an argument of the following sort would be considered valid:

  1. All men are human
  2. Socrates is a man
  3. Therefore, Socrates is human.

Clearly, in this argument, it is not logically possible for the premises to (both) be true, and for the conclusion to be false. The same can be said of the following argument:

  1. Bob loves Carroll
  2. If Bob loves Carroll then Carroll loves Joe
  3. Therefore, Carroll loves Joe

This is pretty obviously logically valid. So too, though, is the following argument:

  1. All women are purple.
  2. Socrates is a woman.
  3. Therefore, Socrates is purple.

The reason this argument is valid is that it is not possible for the premises to (both) be true, and for the conclusion to be false. Perhaps the premises and conclusion are all, in fact, false, but in any possible world in which the premises were true, the conclusion would be true. Thus, validity is not concerned with truth so much as truth-preservation. The concern is to ensure that one cannot, in a ‘valid’ argument, move from true premises to a false conclusion. Take the following example as well:

  1. All women are purple.
  2. If all women are purple, then evolution is true.
  3. Therefore, evolution is true.

In this argument, we have a conclusion which is (I presume) true in fact, while the premises are all false. However, the argument is clearly valid as well, since it is not logically possible that the premises be true and the conclusion false. Remember that validity requires nothing more than that it is not possible for both (i) the premises to be true, and (ii) the conclusion to be false.

The difficulty with this account of validity arises when we are confronted with examples of the following variety:

  1. All men are animals.
  2. If all men are animals then Tyrannosaurus Rex makes a good pet.
  3. Therefore, 3+4=7

This argument is logically valid, since it is not logically possible for the premises to be true while the conclusion is false (mostly because it isn’t possible for the conclusion to be false, and it is ‘possible’ for the premises to be true). Such an argument, however, doesn’t have any dialectical appeal. Consider also:

  1. I once drew a square-circle,
  2. If I once drew a square-circle, then I am a married bachelor,
  3. Therefore, I once drew the impossible.

This argument can be tricky; in order to find out whether it is valid we have to ask whether it is possible for both (i) the premises to be true, and (ii) the conclusion false. As it turns out, it is not possible for the premises to be true and the conclusion false, precisely because it is not possible for the premises to be true. Thus, formally speaking, it is a logically valid argument.


The definition of a sound argument is pretty straightforward: an argument is sound if and only if it is (i) logically valid, and (ii) all of its premises are true. For example,

  1. Socrates was mortal.
  2. Everything that was mortal, was once alive.
  3. Therefore, Socrates was once alive.

In this argument, we find that it is not logically possible for the premises to be true while the conclusion is false, and in addition, we find that both premises are clearly true. Thus, we have a sound argument on our hands. Any argument which is logically valid is sound just in case all of its premises are true. Thus, for example, the following argument is sound:

  1. A tautology is a tautology.
  2. 6-2=4
  3. The sentence ‘is this a question‘ expresses a question.

This exemplifies the problem with the formal definitions of validity and soundness. It shows that one can construct sound and vacuous arguments by simply ensuring that the premises and conclusions are all necessary truths, or at least that the conclusions are necessary truths while the premises are true. In the interest of more off-the-cuff examples, take for instance:

  1. I once wrote this sentence.
  2. If I once wrote this sentence, then I have written at least one sentence.
  3. Therefore, 3+4=7

This argument is both logically valid, and sound, and yet it appears to be a very bad argument. Nobody who didn’t already accept the conclusion could be led by it to accept the conclusion. It is a bad argument, even for those of us who accept the conclusion; if this argument were submitted as our reason for believing the conclusion then our mathematical belief that 3+4=7 would literally be unjustified (a necessary self-evident truth in which we believe can, of course, be unjustified). What all this illustrates is, first, that the formal definitions of validity and soundness are concerned only with truth preservation, and not with the persuasive force of an argument at all. As philosophers who specialize in the study of modal logic often make a distinction between ‘broad’ and ‘narrow’ logical possibility (eg. a square-circle is broadly logically impossible, but narrowly logically possible since there isn’t any purely formal way to obviate a contradiction between the predicates ‘square’ and ‘circle’), so too, perhaps, should we make a distinction between broad and narrow validity & soundness. What we have looked at so far would be the purely formal or ‘narrow’ accounts of validity and soundness. Maybe a ‘broad’ view of validity (which I will henceforth write as ‘validity*’) would be something like: an argument is valid* if and only if i) it is not possible for the premises to be true while the conclusion is false, and ii) the conclusion meaningfully follows from the truth of the premises. This definition of validity* says everything the former one did, with the addition that the premises and conclusion have to be semantically related (i.e., meaningfully related; they have to have something to do with one another). We can correspondingly say that an argument is sound* just in case it is valid* and its premises are true.

Now, validity* and soundness* are not appropriate distinctions in an introductory course on deductive logic, and so are somewhat philosophically unorthodox. However, they are rather useful outside of that narrow context, and in the context of doing philosophy. In philosophy, we don’t just want sound arguments, we want sound* arguments!


Speaking of what philosophers want, there is another issue I wish to examine, which is what makes an argument ‘good’ by philosophical standards. It turns out, I will argue, that neither soundness nor soundness* are necessary or sufficient conditions of ‘goodness’.

I submit that the goodness of an argument consists in two things: i) that the argument is logically valid*, ii) that the accumulated uncertainty of the premises to the argument’s intended audience sets a reasonably high lower bound on the probability of the conclusion. This second criterion is specially crafted to avoid the common mistakes which have, in the past, been made even by some relatively good philosophers like William Lane Craig; namely, the mistake of thinking that premises in a valid argument need be merely each more plausible than their respective negations for the conclusion to follow forcefully. Indeed, the (probability of the) premises of an argument merely set a lower bound on the probability of the conclusion.[1] If that lower bound on the probability of the conclusion is less than or equal to 0.5 then the argument is not compelling. Whether an argument is persuasive or not to some subject is going to depend on their appraisal of the premises, of course, but a good argument will consist of premises which are not merely more plausible than not, but also highly plausible – plausible enough, at least, that the conclusion will also seem highly plausible. This definition obviously subjectivizes ‘goodness,’ making it dependent upon an audience’s appraisal, but that shouldn’t bother us very much because plausibility has to figure into the goodness of an argument in some way, and ‘plausibility’ is already a term of epistemic appraisal.

Consider the following two arguments, both of which are valid and at least one of which is sound. First, the modal ontological argument, which we can roughly reconstruct as:

  1. God possibly exists (i.e., God exists in at least one logically possible world).
  2. If God exists in one logically possible world then God exists in all logically possible worlds.
  3. If God exists in all logically possible worlds then God exists.
  4. Therefore, God exists.

This argument is sound just in case the conclusion is true. However, that doesn’t make it a very good argument in my sense. Indeed, consider its parody:

  1. God possibly does not exist (i.e., there is at least one logically possible world in which God does not exist).
  2. If there is at least one logically possible world in which God does not exist, then there is no logically possible world in which God exists.
  3. If God exists in no logically possible worlds then God does not exist.
  4. Therefore, God does not exist.

At least one of these two arguments is valid, valid*, sound and sound*, but it is arguable that neither of them are good. Goodness, then, consists in more than just soundness*. So, given the way I’ve just outlined things, we can imagine any number of arguments which are good without being sound, sound without being good, valid* without being sound, sound* without being good, but none which are good without being valid*. The goodness of an argument, it seems, is largely in the eye of the beholder; the goodness of a valid* argument is entirely in the eye of the beholder.


[1] See:


Notes on a Transcendental Argument from Logic

Nearly ever since I was first exposed to transcendental argumentation through listening to that famous debate between Greg Bahnsen and Gordon Stein,1 I have retained the intuition that there is an interesting potential argument from the fact that there are necessary propositions (necessary, that is, simpliciter) to the conclusion that there is a necessary mind. While the analysis of what it means to be a necessary mind will fall short of the God of perfect being theology or classical theism, it will still provide a being which so resembles God that it significantly undermines atheism. This being may not have all the superlative attributes, but it will be a metaphysically necessary immaterial spaceless timeless being with an intellect (and whatever that entails), et hoc omnes intelligunt Deum. However, to avoid the charge of using St. Thomas’ famous phrase in order to paper-over the chasm between my conclusion and full-blown theism, I will state the conclusion more modestly in terms of which the good old reverend Bayes would approve. Enjoy;

1) There are laws of logic.
2) Logical laws are identical to necessary propositions (exempli gratia [P v ~P])
3) Therefore, there are necessary propositions.
4) Propositions are not real entities which exist mind-independently, but are mind-dependent (i.e., there is no proposition for which there is not at least one subvenient mind).
5) A necessary truth is a truth which obtains in all logically possible worlds.
6) Necessary truths are either grounded in at least one contingent mind, or at least one incontingent mind.
7) There are logically possible worlds without any contingent minds.
8) Therefore, there must be at least one necessary mind.
9) If there is at least one necessary mind then it is a being with intellect (plausibly knowing all necessary truths), which is immaterial (spaceless, timeless) in nature.
10) The conditional probability of theism is, ceteris paribus, greater than the conditional probability of not-theism on the condition that there is at least one metaphysically necessary immaterial being with intellect. 
11) Therefore, theism is probably true, 
ceteris paribus.

There are plenty of points at which one could still object to this argument, but it seems to me that most objections are philosophically more costly than the conclusion. One might also just accept the conclusion but deny that, in fact, things really are equal (i.e., cetera non sunt pariba) in this case. For instance, the objector could insist that there are no propositions which are ‘necessary’ in the sense required here (that is, necessary simpliciter – not a merely model-dependent necessity). They might also insist, for some odd reason, that there are not possible worlds without contingent minds, or that those worlds are possible in a merely model-dependent way while other possible worlds are possible simpliciter. That would be pretty wild. Another might argue that the existence of a metaphysically necessary immaterial mind doesn’t raise the conditional probability of theism at all (maybe because the probability of theism is ‘0’ – or because it is ‘1’). Somebody could, of course, deny the major premise, that there are laws of logic. Somebody may also insist that laws of logic are not identical to the propositions which express them (though that seems to reify them so much as to put the objector, for other reasons, in the near occasion of belief in theism anyway). Alternatively one may think that each premise on its own seems more plausibly true than false, but that the collection of them together seems to have a upper-bounded probability of lower than or equal to 0.5, and that would be a principled way to object.

Edit*: it occurs to me that there’s no way of which I’m aware to really set an upper-bound on the probability of a conclusion. What the objector could say, then, is either that the conclusion just seems to be no more likely than 0.5 (notwithstanding the plausibility of the individual premises), or that the premises collectively set a lower-bounded probability on the conclusion of less than or equal to 0.5, in which case the argument fails to be compelling.

To be fair, this argument of mine very likely draws significantly from the influence of James N. Anderson and Greg Welty,2 whose argument seems, to me, much better than what often passes for responsible argument among presuppositionalists (among whom, I should take a moment to clarify, I adamantly do not count myself).

1 For those interested, you can find the audio of the debate, and the transcript (because the audio is really not great) at the following two links: and

2 James N. Anderson and Greg Welty, “The Lord of Non-Contradiction: An Argument for God from Logic” Philosophia Christi 13:2 (2011).

An Amended Minimal Principle of Contradiction

The law of non-contradiction seems self-evidently true, but it has its opponents (or, at least, opponents of its being necessary (de dicto) simpliciter). W.V.O. Quine is perhaps the most well known philosopher to call the principle into question by calling analyticity itself into question in his famous essay “Two Dogmas of Empiricism,” and suggesting that, if we’re to be thoroughgoing empiricists, we ought to adopt a principle of universal revisability (that is to say, we adopt a principle according to which absolutely any of our beliefs, however indubitable to us, should be regarded as revisable in principle, including the principle of revisability). Quine imagined that our beliefs were networked together like parts of a web in that we have beliefs to which we aren’t strongly committed, which we imagine as near the periphery of the web, which are much less costly to change than the beliefs to which we are most strongly committed, which we imagine as near the center of that web. Changing parts of the web nearer to the periphery does less to change the overall structure of the network than changing beliefs at the center of the web. Evolution has, in operating upon our cognitive faculties, selected for our tendency towards epistemic conservatism.

This, he thinks, is why we don’t mind changing our peripheral beliefs (for instance, beliefs about whether there is milk in the fridge or whether a certain economic plan would better conduce to long-term increases in GDP than a competing plan) but we stubbornly hold onto our beliefs about things like mathematics, logic, and even some basic intuitive metaphysical principles (like Parmenides’ ex nihilo nihil fit). Nevertheless, indubitability notwithstanding, if all our knowledge is empirical in principle, then everything we believe is subject to revision, according to Quine. He boldly states:

… no statement is immune to revision. Revision even of the logical law of the excluded middle has been proposed as a means of simplifying quantum mechanics; and what difference is there in principle between such a shift and the shift whereby Kepler superseded Ptolemy, or Einstein Newton, or Darwin Aristotle?1

This statement is far from short-sighted on Quine’s part. Those who defend his view have suggested that even the law of non-contradiction should be regarded as revisable, especially in light of paraconsistent systems of logic in which the law of non-contradiction is neither axiomatic, nor derivable as a theorem operating within those systems. This is why Chalmers calls attention to the fact that many regard Quine’s essay “as the most important critique of the notion of the a priori, with the potential to undermine the whole program of conceptual analysis.”2 In one fell swoop Quine undermined not only Carnap’s logical positivism, but analyticity itself, and with it a host of philosophical dogmas ranging from the classical theory of concepts to almost every foundationalist epistemological system. The force and scope of his argument was breathtaking, and it continues to plague and perplex philosophers today.

More surprising still is the fact that Quine isn’t alone in thinking that every belief is revisable. Indeed, there is a significant faction of philosophers committed to naturalism and naturalized epistemology, but who think that a fully naturalized epistemology will render all knowledge empirical, and, therefore, subject to revision in principle. Michael Devitt, for instance, defines naturalism epistemologically (rather than metaphysically):

“It is overwhelmingly plausible that some knowledge is empirical, justified by experience. The attractive thesis of naturalism is that all knowledge is; there is only one way of knowing”3

Philosophical attractiveness, I suppose, is in the eye of the beholder. It should be noted, in passing, that metaphysical naturalism and epistemological naturalism are not identical. Metaphysical naturalism does not entail epistemological naturalism, and neither does epistemological naturalism entail metaphysical naturalism. I have argued elsewhere that there may not even be a coherent way to define naturalism, but at least some idea of a naturalized metaphysic can be intuitively extrapolated from science; there is, though, no intuitive way to extrapolate a naturalized epistemology from science. As Putnam puts it:

“The fact that the naturalized epistemologist is trying to reconstruct what he can of an enterprise that few philosophers of any persuasion regard as unflawed is perhaps the explanation of the fact that the naturalistic tendency in epistemology expresses itself in so many incompatible and mutually divergent ways, while the naturalistic tendency in metaphysics appears to be, and regards itself as, a unified movement.”4

Another note in passing; strictly speaking Devitt’s statement could simply entail that we do not ‘know’ any analytic truths (perhaps given some qualified conditions on knowledge), rather than that there are no analytic truths, or even that there are no knowable analytic truths. Quine, I think, is more radical insofar as he seems to suggest that there are no analytic truths at all, and at least suggests that none are possibly known. Devitt’s statement, on the other hand, would be correct even if it just contingently happened to be the case that not a single person satisfied the sufficient conditions for knowing any analytic truth.

Hilary Putnam, unfortunately writing shortly after W.V.O. Quine passed away, provided a principle which is allegedly a priori, and which, it seems, even Quine could not have regarded as revisable. Calling this the minimal principle of contradiction, he states it as:

Not every statement is both true and false”5

Putnam himself thought that this principle establishes that there is at least one incorrigible a priori truth which is believed, if at all, infallibly. Putnam shares in his own intellectual autobiography that he had objected to himself, in his notes, as follows:

“I think it is right to say that, within our present conceptual scheme, the minimal principle of contradiction is so basic that it cannot significantly be ‘explained’ at all. But that does not make it an ‘absolutely a priori truth’ in the sense of an absolutely unrevisable truth. Mathematical intuitionism, for example, represents one proposal for revising the minimal principle of contradiction: not by saying that it is false, but by denying the applicability of the classical concepts of truth and falsity at all. Of course, then there would be a new ‘minimal principle of contradiction’: for example, ‘no statement is both proved and disproved’ (where ‘proof’ is taken to be a concept which does not presuppose the classical notion of truth by the intuitionists); but this is not the minimal principle of contradiction. Every statement is subject to revision; but not in every way.”6

He writes, shortly after recounting this, that he had objected to his own objection by suggesting that “if the classical notions of truth and falsity do not have to be given up, then not every statement is both true and false.”7 This, then, had, he thought, to be absolutely unrevisable.

This minimal principle of contradiction, or some version of it, has seemed, to me, nearly indubitable, and this despite my sincerest philosophical efforts. However, as I was reflecting more deeply upon it recently I realized that it is possible to enunciate an even weaker or more minimalist (that is to say, all things being equal, more indubitable) principle. As a propaedeutic note, I observe that not everyone is agreed upon what the fundamental truth-bearers are (whether propositions, tokens, tokenings, etc.), so one’s statement, ideally, shouldn’t tacitly presuppose any particular view. Putnam’s statement seems non-committal, but I think it is possible to read some relevance into his use of the word ‘statement’ such that the skeptic may quizzaciously opine that the principle isn’t beyond contention after all. In what follows, I will use the term ‘proposition*’ to refer to any truth-bearing element in a system.

Consider that there are fuzzy logics, systems in which bivalence is denied. A fuzzy logic, briefly, is just a system in which propositions are not regarded (necessarily) as straightforwardly true or false, but as what we might think of as ‘true’ to some degree. For instance, what is the degree to which Michael is bald? How many hairs, precisely, does Michael have to have left in order to be considered one hair away from being bald? Well, it seems like for predicates like ‘bald’ there is some ambiguity about their necessary conditions. Fuzzy logic is intended to deal with that fuzziness by allowing us to assign values in a way best illustrated by example: “Michael is 0.78 bald.” That is, it is 0.78 true that Michael is bald (something like 78% true). Obviously we can always ask the fuzzy logician whether her fuzzy statement is 1.0 true (and here she either admits that fuzzy logic is embedded in something like a more conventional bivalent logic, or she winds up stuck with infinite regresses of the partiality of truths), but I digress. Let’s accept, counter-possibly, that fuzzy logics provide a viable way to deny bivalence, and thus allow us to give a principled rejection of Putnam’s principle.

Even so, I think we can amend the principle to make it stronger. Here is my proposal for an amended principle of minimal contradiction:

“Not every single proposition* has every truth value.”

I think that this is as bedrock an analytic statement as one can hope to come by. It is indubitable, incorrigible, indubitably incorrigible, and it holds true across all possible systems/logics/languages. It seems, therefore, as though it is proof-positive of analyticity in an impressively strong sense; namely, in the sense that necessity is not always model-dependent. At least one proposition* is true across all possible systems, so that it is necessary in a stronger sense than something’s merely being necessary as regarded from within some logic or system of analysis.


As a post-script, here are some principles I was thinking about as a result of the above lines of thought. First, consider the principle:

At least one proposition* has at least one truth-value.

To deny this is to deny oneself a system altogether. No logic, however esoteric or unconventional or counter-intuitive, can get off the ground without this presupposition.

Consider another one:

For any proposition* P, if we know/assume only about P that it is a proposition*, then P more probably than not has at least one truth-value.

I’m not certain about this last principle, but it does seem intuitive. The way to deny it, I suppose, would be to suggest that even if most propositions* were without truth-values, one could identify a sub-class of propositions with an extremely high probability of having a truth-value, and that will allow one to operate on an alternative assumption.

[Note: some of the following footnotes may be wrong and in need of fixing. Unfortunately I would need several of my books, currently in Oxford with a friend, to adequately check each reference. I usually try to be careful with my references, but here I make special note of my inability to do due diligence.]

1 W.V.O. Quine, “Two Dogmas of Empiricism,” in The Philosophical Review, Vol. 60, No.1 (Jan., 1951), 40.

2 David J. Chalmers, “Revisability and Conceptual Change in “Two Dogmas of Empiricism”.” The Journal of Philosophy 108, no. 8 (2011): 387.

3 Louise Antony, “A Naturalized Approach to the A Priori,Epistemology: An Anthology. Second Edition, Edited by Ernest Sosa, Jaegwon Kim, Jeremy Fantl and Matthew McGrath. (Oxford: Blackwell publishing, 2000), 1.

4 Hilary Putnam, “Why Reason can’t be Naturalized,” Epistemology: An Anthology. Second Edition, Edited by Ernest Sosa, Jaegwon Kim, Jeremy Fantl and Matthew McGrath. (Oxford: Blackwell publishing, 2000), 314.

5 Hilary Putnam, “There is at least one a priori Truth” Epistemology: An Anthology. Second Edition, Edited by Ernest Sosa, Jaegwon Kim, Jeremy Fantl and Matthew McGrath. (Blackwell: 2000): 585-594.

6 Auxier, Randall E., Douglas R. Anderson, and Lewis Edwin Hahn, eds. The Philosophy of Hilary Putnam. Vol. 34. (Open Court, 2015): 71.

7 Auxier, Randall E., Douglas R. Anderson, and Lewis Edwin Hahn, eds. The Philosophy of Hilary Putnam. Vol. 34. (Open Court, 2015): 71.