Suppose you are to answer the following two questions:
(1) Does the sentence “God exists” express a proposition?
(2) If so, then is that proposition true or false?
If you say no to the first question, then you may be classified as a noncognitivist with regard to God-talk. If you say yes to it, thereby allowing that the given sentence does express a proposition, then you are a cognitivist with regard to God-talk. (Let us henceforth abbreviate these expressions, simply using the terms “cognitivist” and “noncognitivist”.) All theists, atheists, and agnostics are cognitivists, so the second question applies to them: is the proposition that God exists true or false? You are a theist if and only if you say that the proposition is true or probably true, you are an atheist if and only if you say that it is false or probably false, and you are an agnostic if and only if you understand what the proposition is, but resist giving either answer, and support your resistance by saying, “The evidence is insufficient” (or words to that effect).
One virtue of this way of characterizing the three groups of cognitivists is that it captures the way the terms are commonly used in ordinary language, and, in particular, it makes the groups mutually exclusive. No one can consistently be both a theist and an agnostic, or both an atheist and an agnostic. Some other ways of drawing the distinction fail to capture that important feature. For example, if the term “agnostic” were defined as anyone who claims that there is insufficient evidence to know whether or not God exists, then it would be possible for a person to be both a theist and an agnostic. He could be what is called a fideist and say, “I realize that the evidence is insufficient, but I believe in God anyway.” (Incidentally, I am here taking the expression “believe in God” simply as shorthand for “believe that God exists.”) Alternatively, such an “agnostic” could be a “fideistic atheist” and say, “Though the evidence is insufficient, I deny God’s existence anyway.” That would make it possible for someone to be both an atheist and an agnostic. This result is a drawback to such a definition of “agnostic,” for it conflicts with the way the term is commonly used.
Another definition of “agnostic” which is at variance with common usage is that of Thomas Huxley, who first coined the word in 1869. He said the following:
Agnosticism is not a creed but a method, the essence of which lies in vigorous application of a single principle. Positively the principle may be expressed as: in matters of intellect, follow your reason as far as it can carry you without other considerations. And negatively, in matters of the intellect, do not pretend the conclusions are certain that are not demonstrated or demonstrable. It is wrong for a man to say he is certain of the objective truth of a proposition unless he can produce evidence which logically justifies that certainty. That is what Agnosticism asserts; and, in my opinion, it is all that is essential to Agnosticism. … The application of the principle results in the denial of, or the suspension of judgment concerning, a number of propositions respecting which our contemporary ecclesiastical “gnostics” profess entire certainty.
Thus, Huxley would call anyone an “agnostic” who abides by the principle that propositions should be believed exactly in accord with the available evidence. However, he does not make the application of that principle to God’s existence a part of the definition of the term “agnostic.” Thus, people who claim to be able to prove God’s existence or nonexistence could still be “agnostics” by Huxley’s definition, for they may indeed be abiding by the given principle. Since the definition allows one to be both an agnostic and a theist or atheist, it fails to capture that ordinary-language concept that is the focus of this investigation.
An agnostic could also be an atheist if the term “atheist” were defined more broadly, for example, as anyone who lacks a belief in God, or who classifies the proposition that God exists as anything other than true. Such a definition is recommended by George H. Smith in his book Atheism: the Case Against God. Other writers who support the definition are cited in Michael Martin’s book Atheism: A Philosophical Justification. According to this usage, people would be “atheists” even if they answer the question whether it is true that God exists with “no one knows.” This is a departure from the most common use of the word “atheist” in ordinary language, which is in itself an important reason to avoid it. Another reason is that infants and fetuses have no belief in God, yet it would be perverse to say that they are all atheists.
Sometimes the use of the term “atheism” to mean “lack of theistic belief” is supported by an appeal to etymology. For example, Martin, in the book mentioned above, says the following:
In Greek a’ means without’ or not’ and theos’ means god.’ From this standpoint an atheist would simply be someone without a belief in God, not necessarily someone who believes that God does not exist. According to its Greek roots, then, atheism is a negative view, characterized by the absence of belief in God.
This argument is rather unsatisfactory for at least two reasons. First, it is not completely clear that the correct translation of the Greek prefix “a” is “without.” It might also mean “no,” in which case “a-the-ism” could be translated as “no-god-ism,” or “the view that there is no god.” Note that there is no “ism” in Greek. Second, even if the etymology of the word “atheism” did indicate that it once meant “without belief in God,” that is still not a good guide to current usage. It is quite common for words to acquire new meanings over time. It seems far more important what people mean by a word today than what it once meant long ago.
Another argument sometimes put forward is that we should ascertain what the word “atheist” means by taking a poll among atheists. But that is an unclear suggestion. How are we to decide who is an atheist (and thus to be polled) prior to ascertaining what the word “atheist” means? Let us assume that the poll is to be taken among all those native speakers of English who are not theists. It is still not clear what the result of such a poll would be. I have never seen any statistical result presented on the matter. My conclusion here is that no good case has ever been made for using the word “atheist” in the sense of “one who is without belief in God.”
In this essay, I shall use the term “atheist” in its (more common) narrow sense. Martin draws a distinction between “negative atheists,” who are without any belief in God, and “positive atheists,” who deny God’s existence.) Applying that distinction, it could be said that I (and most people) use the term “atheist” in the sense of “positive atheist.” It should be noted that all positive atheists are automatically negative atheists, which may sound somewhat peculiar when those expressions are used.
In place of the expression “negative atheist,” I shall use the term “nontheist.” That seems to be a better term (than “atheist”) for capturing the more general concept of “one who is without belief in God,” for several reasons:
(1) Almost everyone who employs the term “nontheist” already uses it in the given way.
(2) As indicated in dictionaries, most native speakers of English use the term “atheist” for the more definite concept of “one who denies that God exists.” It is desirable that we abide by common usage and it is foolish (and probably futile) to try to reform people’s usage of terms.
(3) It would be more natural to call infants and fetuses “nontheists” than to call them “atheists.”
(4) It is desirable to have a system in which the familiar three classes, theists, atheists, and agnostics, are mutually exclusive, and that would not be possible if the term “atheist” were instead used for the more general concept.
Although the usual way of defining “theist,” “atheist,” and “agnostic” (in terms of the proposition that God exists) is commendable, it is also incomplete, for it yields no classification for someone who says, “It depends on how you define the word God’: for some definitions the proposition is true, for others it is false, and for still others I just don’t know.”
Let us consider an alternate way of formulating the definitions. Since the word “God” has many different meanings, it is possible for the sentence “God exists” to express many different propositions. What we need to do is to focus on each proposition separately. Subscripts could be used for the different senses of “God.” Thus we have the proposition that God1 exists, the proposition that God2 exists, the proposition that God exists, etc., with each different sense of “God” suitably defined. For each different sense of the term “God,” there will be theists, atheists, and agnostics relative to that concept of God. A person might be a theist relative to one concept of God but an atheist or agnostic relative to a different one. If the question is raised whether God1 exists, then theists relative to that concept (of God) are people who answer that question affirmatively. Atheists relative to that concept are people who answer the question negatively. And agnostics relative to that concept are people who understand the question but who avoid committing themselves to a “yes” or “no” answer to it by maintaining that the evidence either way is insufficient.
Now suppose someone says of the sentence “God1 exists,” where “God1” has been previously defined in some way, that it is a cognitively meaningless sentence and does not express any proposition at all. Should such a person be called an “atheist” or an “agnostic”? In The Encyclopedia of Philosophy he is classified as an atheist. From one point of view, that seems reasonable, since such a person is obviously very far from being a believer in God1. On the other hand, he is not giving a negative answer to the question whether God1 exists. He is not saying of the proposition that God1 exists that it is a false proposition. So, from that perspective, he seems to fit more into the category of an agnostic.
Of course, the person in question is not taking a noncommittal stance relative to the proposition that God1 exists, for he does not even concede that there is any such proposition. So the “agnostic” label doesn’t fit either. A. J. Ayer, a philosopher who held this type of position, rejected both labels “atheist” and “agnostic.” And that is the terminology that I prefer. I call such a person a “noncognitivist with regard to God-talk” (or just “noncognitivist” for short) and place him into a fourth category. The correctness of this terminology cannot be settled by appeal to ordinary language, for the type of position being considered is practically unknown among the lay public. It is almost exclusively held by those analytical philosophers known as logical positivists. Rudolf Carnap employed it with respect to the word “God” in his essay “The Elimination of Metaphysics Through Logical Analysis of Language.” A somewhat similar outlook had been expressed about seventy years earlier by Charles Bradlaugh in his essay “A Plea for Atheism,” however the view was there referred to as “atheism.” Anyway, there is no basis for assessing how it would probably be classified by ordinary people. What is called for is simply some sort of linguistic stipulation.
The reason I call the view “noncognitivism with regard to God-talk” rather than “noncognitivism with regard to religious language” is that the sentence “God exists” occurs in contexts other than religious language. One such context is the field of metaphysics. The existence of God is a standard topic in metaphysics, and there need be no reference within that context to religion or to religious discourse. Thus, the noncognitivist is rejecting the cognitive meaningfulness of various sentences that contain the word “God,” whether those sentences occur within religious discourse or not. Consider now the sentence “God1 exists,” where some definition has been previously given for the subject term “God1.” Relative to that sentence, we may put forward the following definitions:
A noncognitivist is someone who declares that the sentence does not express any proposition at all.
A theist is someone who allows that the sentence expresses a proposition and who classifies the proposition as true or probably true.
An atheist is someone who allows that the sentence expresses a proposition and who classifies the proposition as false or probably false.
An agnostic is someone who allows that the sentence expresses a proposition, and who grants that he/she knows what that proposition is, but who is noncommittal about its truth or falsity on the grounds of insufficient evidence.
It is to be borne in mind here that each category is relative to a particular interpretation of a sentence of the type “God exists,” A person may be in one category relative to one interpretation (or one sense of the word “God”), but in a different category relative to another interpretation.
To illustrate these distinctions, consider the following four responses to a request for a definition of the term “God”:
God1 = the universe itself (all that exists). [Or, alternatively, God1 = love.]
God2 = the powerful being who created the universe.
God3 = the omnipotent creator of the universe whose highest goal regarding humans is that they believe that he has a son who died for them so that they might obtain salvation.
God4 = ? (No definition is possible; the word is indefinable.)
Now suppose there were a philosopher who examined these four responses. When asked the question “does God exist?” he might very well respond as follows:
In the case of God1, yes, God definitely exists, for it is obvious that the universe [or love] exists. In the case of God2, I understand the question but have no answer to it since the evidence is insufficient. In the case of God3, there is good evidence that such a being does not exist, for most humans do not believe in his son, etc., yet, if such a being were to exist, then probably he would have done things to cause them to have the given belief. And in the case of God4, I do not understand the question. Since no definition of “God4” has been given, the sentence “God4 exists” expresses no proposition whatever.
Given this response, we should say of such a philosopher that he is a theist relative to God1, an agnostic relative to God2, an atheist relative to God3, and a noncognitivist relative to God4. I would say that these answers to the four “does God exist?” questions are reasonable, though they are not necessarily the correct (or “best”) answers.
It should be noted that the term “theist” is here being taken in a broad sense, one which includes what are often referred to as “deists” and “pantheists.” In a narrower sense, a theist only affirms the existence of a certain type of deity (a personal deity who rules the universe). The distinction between different types of theist (or different senses of the term “theist”) is outside the scope of the present essay.
A question might be raised here. Suppose “God5” is defined as “a transcendent spirit capable of thought, feeling, and action,” and suppose there is a man who says of the sentence “God5 exists” that it is cognitively meaningless and expresses no proposition. When he says that, does he mean to speak just for himself or does he mean to speak for everyone? In other words, is he merely claiming that he himself does not understand the sentence, whereas there may be others who do? Or is he, instead, claiming that no one understands it? We can initially refer to these two types of noncognitivist as the “subjective noncognitivist” and the “objective noncognitivist.” When the man claims not to understand the idea of a transcendent spirit capable of thought, feeling, and action, we need to ask him: Do you mean just that you personally can’t grasp that idea (at the present time), allowing the possibility that others may understand it (or that you yourself may come to understand it in the future)? Or, alternatively, are you claiming that the idea is inherently unintelligible, so that no one can possibly grasp it or think it? These are two essentially different types of claim, and so we have here a fundamental distinction between two essentially different types of noncognitivist.
The objective noncognitivist is making a bolder claim, one which is in the public arena and in need of support. His view could be refuted by showing that the term “God5” can be understood sufficiently to allow the sentence “God5 exists” to express a proposition. There could be philosophical debate about such an issue. In the case of the subjective noncognitivist, presumably he is not making a claim in the public arena. He only says that he himself fails to grasp the concept, allowing that possibly others might (or perhaps that he himself might come to grasp it later). Of course, if he were to have very strong doubts about such a possibility, then his position would come close to that of the objective one. In expressing such doubts, he would be making a statement in the public arena.
A similar distinction could be drawn in the case of agnosticism. Subjective agnostics would be people who simply make autobiographical reports regarding their own situation. Each of them says, for example, “I don’t have enough evidence one way or the other to give an answer to the question whether God5 exists.” In contrast, objective agnostics would be making a statement in the public arena. It is the statement that there just does not exist sufficient evidence to warrant an answer to the given question, and if anyone at all were to answer the question with a yes or no, then that would be a mistake, and perhaps irrational.
The way I have been construing agnosticism, it is the objective view that I have in mind. And similarly for noncognitivism. When I speak of that position, it should be understood that it is the objective form to which I mean to refer. Thus, given a specific definition of “God,” agnostics are people who claim that no one has sufficient evidence to warrant acceptance of either theism or atheism. And noncognitivists are people who claim that no one understands the sentence “God exists” in a way that would allow it to express a proposition.
One objection to my definition of “atheism” is that it seems to allow no way for anyone to simply proclaim, “I am an atheist” irrespective of the definition given for “God.” It seems to force one to always listen to a definition of “God” before saying, “I am an atheist relative to that definition.” A better way of defining “atheism,” it is claimed, is “denial of the existence of all gods.” I have two replies to this objection. First, to define “atheism” as the denial of the existence of all gods is unsatisfactory, because there may be gods that clearly do exist. Some people say, “God is the universe” or “God is love.” The reasonable response here should be to grant that God does exist when defined that way. Consider also primitive tribes who worshiped huge statues as gods. We should be able to say, “Their god was a huge statue.” But if we say that, then we need to grant that their god did indeed exist. Second, discussions of the existence of God almost always occur within linguistic contexts in which a certain particular concept of God is understood. So one could legitimately proclaim “I am an atheist” in such a setting without first agreeing on a definition of “God,” for the simple reason that some definition of “God” is already being assumed. For example, if one is discussing the topic with Christians, then it might be assumed that it is the Christian concept of God that is at issue. There would then be no problem in simply proclaiming, “I am an atheist,” for it would be understood within that particular context that what is meant is “I am an atheist with regard to the God of Christianity.”
As indicated previously, agnostics allow that “God exists” expresses a proposition, but they regard the evidence as insufficient to warrant committing oneself one way or the other about the truth value of the proposition. They are not even willing to say that the proposition is probably true or that it is probably false. If a person were to say that it is probably true, then I would call him a “theist,” even if he is not willing to go so far as to say that he “believes in God.” He may just say that the evidence presently available favors the proposition that God exists (where the term “God” is given some particular definition). I would call him a “theist” so long as he leans to that side. Similarly, I would call a person an “atheist” even if he is not willing to say that he believes in God’s nonexistence. He may just say that the available evidence favors the proposition that God does not exist (given some particular definition of “God”). I would say that’s enough to classify the person as an “atheist,” at least relative to the given definition of “God.” We could draw a distinction here between “weak theists” and “strong theists,” and between “weak atheists” and “strong atheists,” depending on how strongly they proclaim their view. But in contrast to these positions, the agnostic does not claim that the available evidence favors either side, even to the slightest degree.
Let us now look at the concepts of atheism and agnosticism more closely to see whether any further distinctions may be drawn within them. Among atheists, I would distinguish four different types, as follows:
Disproof atheists, who claim that there is good objective evidence for God’s nonexistence, and if there is any good objective evidence at all for God’s existence, it is significantly outweighed by the evidence for God’s nonexistence.
Methodological atheists, who claim that there is no good objective evidence either for God’s existence or for God’s nonexistence, but there is a certain methodological principle which places the burden of proof upon theists, and since they fail to meet that principle, the only rational position to take is that of atheism. (Some methodological atheists formulate the principle by saying that the burden of proof is always on any person making an existence claim, since, from a logical point of view, existence claims are only capable of proof, not disproof. No one has ever proven the nonexistence of Santa Claus, or elves, or unicorns, or anything else, simply because the very logic of an unrestricted existential proposition prohibits its disproof. It is impossible to go all over the universe and show that, for example, there are no elves anywhere. For this reason, rational methodology calls for us to deny the existence of all those things which have never been shown to exist. That is why we all regard it rational to deny the existence of Santa Claus, elves, unicorns, etc. And since God is in that same category, having never been shown to exist, it follows that rational methodology calls for us to deny the existence of God.)
Mystical atheists, who claim that there is no rational support for theism or atheism, but who believe in God’s nonexistence on the basis of a private, subjective, mystical experience.
Faith atheists, who accept atheism, not on the basis of any grounds or experiences, but simply on the basis of “faith,” regarding it to be a “properly basic belief.” They say that one could be rational in accepting a noetic system that has atheism as its foundational presupposition, since there is no good objective evidence for God’s existence.
It is an interesting question which, if any, of the positions represented here is true or correct or the best one to hold, but I take that to be outside the scope of the present essay. I turn now to an examination of types of agnosticism. Among the types which may be distinguished are the following:
Unknowability agnostics, who say that not only is there no evidence relevant to the issue of God’s existence but that there could be none. They may view the issue as so abstract and removed from human experience that it is impossible that we should have any evidence regarding it, one way or the other, so the matter is essentially unknowable. (Note that people with this outlook could easily move over to the position of noncognitivism. They are making the claim that the matter of God’s existence is essentially unknowable and that there could not possibly be evidence regarding it, one way or the other. If they were to conclude that the sentence “God exists” therefore does not express any proposition at all, then they would be noncognitivists. But if, instead, they allow that there is a proposition expressed but it is simply one that we cannot ever legitimately claim to be true or false, then they are to be classified as “unknowability agnostics.”
Zero-data agnostics, who allow that it is possible to know whether or not God exists, but no one does in fact know it for the simple reason that there is zero data regarding the matter. Although no one at present has any relevant data, it is possible that some may come along in the future.
Data-vs.-data agnostics, who concede that there is evidence relevant to the issue of God’s existence, but who have judged it to be perfectly balanced. That is, the evidence presently available in support of God’s existence is exactly matched by evidence presently available against God’s existence. They could be called “knife-edge agnostics,” since they view the positive evidence for God’s existence as so perfectly balanced by negative evidence that it is as if they were balanced on the edge of a knife without falling to one side or the other.
Data-vs.-principle agnostics, who are another type of “knife-edge agnostics.” They agree with the methodological atheists, described above, who claim that there is a certain methodological principle that places the burden of proof upon the theists. The question is raised as to how much evidence is needed for the existence of something in order for it to satisfy the burden of proof required by the given principle. Presumably there is some minimum amount such that any lesser amount would still call for us to deny the existence of the thing. Take, for example, Bigfoot or the Loch Ness Monster. It may be argued that there is some slight evidence in favor of the existence of such entities but it is less than the amount needed to satisfy the burden-of-proof requirement. Thus, the rational stance to take with respect to them, at least at the present time, is that of denying their existence. If, in the future, additional evidence is found favoring the existence of the entity, and that evidence is of sufficient quantity to satisfy the burden-of-proof requirement, then at that time the rational stance to take would be to believe that the entity exists. Suppose, now, that additional evidence is found for, say, the existence of Bigfoot. But the evidence is not quite sufficient to satisfy the burden-of-proof requirement. It is almost enough. In fact, it is as close to satisfying the requirement without actually satisfying it as it is possible to get. One might say, in a case like this, that the evidence for the existence of Bigfoot is balanced on a knife-edge. It is not enough to warrant belief, but it is also too much evidence to warrant denial of Bigfoot’s existence. In such a case, it might be argued, the rational stance to take for anyone aware of the evidence is that of withholding judgment on the matter. Well, this is the situation with data-vs.-principle agnostics. They are not claiming that there are two opposing bodies of evidence, one favoring God’s existence and the other favoring God’s nonexistence. Rather, there is only a body of evidence that favors God’s existence. However, the evidence, at least at the present time, is not sufficient to satisfy the burden-of-proof requirement imposed upon theism, though it is right at the threshold of satisfying it. It is not yet enough to warrant belief that God exists, but it is also too much to warrant denying God’s existence. Hence, the rational stance for anyone to take regarding God’s existence is that of agnosticism. (Thus, both types of “knife-edge agnostic” claim that there is definite evidence in favor of God’s existence, but that it is exactly balanced by something else. One type, data-vs.-data agnostics, say that the something else is definite evidence in favor of God’s nonexistence. The other type, data-vs.-principle agnostics, say that the something else is a principle of rational methodology which places a certain burden of proof upon the theist.)
It must be remembered that all of these positions are relative to a given definition of “God.” Someone may be one type of agnostic relative to one definition but a different type of agnostic (or a theist or an atheist or a noncognitivist) relative to another definition. I leave the reader with these questions: Are there definitions of “God” with regard to which you are an atheist or agnostic on the matter of God’s existence? If so, can you locate yourself on the above lists with regard to each of those definitions? And are there definitions of “God” with regard to which you are a theist and perhaps even ones with regard to which you are a noncognitivist? If you are a rational person, then I think you should be willing to say yes to all of these questions.
 Rudolf Carnap, “The Elimination of Metaphysics Through Logical Analysis of Language,” originally published (in German) in Erkenntnis, vol. 2 (1932), and translated by Arthur Pap for the anthology Logical Positivism, edited by A. J. Ayer (The Free Press, 1959), p. 66.