Frege's Foundations of Arithmetic: The Logicist Program
Gottlob Frege's 1884 pursuit of the logicist program, arguing that arithmetic truths are analytic and can be derived from purely logical principles.
PhilosophyAll chapters Reading tools Notes & comments
What numbers are › One Collection, Several AnswersFull chapter · Sections in this chapter1 / 8 ↓
One Collection, Several Answers
Suppose you have just spread a deck of cards on the table. How many are there? You could reply “one deck,” “four suits,” or “fifty-two cards.” Your answer depends on what you are counting. Someone who picked up a single card while you were out would still leave one deck and four suits. But “fifty-two cards” would become an error. A number claim has a condition of correctness, even when the objects admit several different ways of being counted.
In The Foundations of Arithmetic, §46, Gottlob Frege asks what goes on in statements of this sort. His examples range from counting divisions of a whole to saying that Venus has zero moons. The latter case helps isolate the problem. There is no group of moons to which you can attach a property. The statement tells you that nothing falls under the concept moon of Venus. The contribution of zero can be understood by asking what the whole sentence says about that concept.
This is the beginning of a larger problem in the 1884 book. A number statement can tell us something about a concept, but an ordinary equation like “the number of cards in the first deck equals the number of cards in the second” seems to name the same object twice. Frege wants an account that makes room for both observations, and then lets us see what sort of reasoning could establish the laws of arithmetic.
Where an Arithmetic Proof Starts
Frege opens by asking for the grounds of familiar arithmetic truths. We may know the answer to a simple sum long before we can say what justifies it. A child starts with counters, learns some examples, and eventually answers without needing to move anything around. An account of that learning is useful. It leaves a further question about what would make the child's answer correct, no matter how the child arrives at it.
In §3, Frege proposes answering this question by tracing a proof to its ultimate premises. If the proof needs only general logical laws and admissible definitions, he calls the result analytic. A result is synthetic if its proof needs truths about a special domain beyond those logical laws and definitions. Whether a truth is a priori is a separate question. This asks whether its proof can proceed from general laws without needing particular facts. The distinctions are about what warrants the result. They are independent of how early in life someone learns it or whether its proof is easy to invent.
This way of drawing the distinction lets Frege retain part of Immanuel Kant's question while contesting Kant's answer. Arithmetic can yield consequences we could spend years trying to discover. That discovery can be new and difficult even if the proof ultimately turns on general logical laws. In §5, Frege asks whether 135,664 + 37,863 = 173,527 is immediately apparent. It takes work to establish that sum. He argues that this counts against treating numerical equations as unprovable. The fact that the result needs to be found does not settle which premises are needed to establish it. His aim is to examine the grounds of a proof closely enough to tell whether it uses any distinctively arithmetical insight.
John Stuart Mill raises a different issue. Counting things in one place can make a sum look like the result of an observation about how things combine. In §§7–10, Frege presses this idea with the case of the number 777,864. What particular physical observation would belong to the definition of that number? Observations may answer whether the number applies to a particular population, and whether particular collections join or separate under particular circumstances. The arithmetical proof is supposed to explain how a result holds independently of any such collection. Frege's example asks the empiricist to show where an observation would enter that proof.
Concepts, Objects, and Context
Three principles at the end of the Introduction guide the inquiry. Frege distinguishes the logical and objective from the psychological and subjective. He asks about the meaning of a word in the context of a sentence. He also keeps concepts distinct from objects. A deck can fall under the concept complete deck. The deck is a particular thing and can fall under several concepts. The concept supplies a condition that can be met by other objects as well.
The context principle gives Frege an alternative to asking what picture the word “seven” brings to mind. Suppose two people read “there are seven envelopes in the drawer.” One pictures the envelopes in a row. The other sees them in a stack. They understand the same assertion well enough to disagree over whether it is true. If three of the envelopes are removed, their mental images need not change for the assertion to become false. Starting with what the whole sentence says helps explain how speakers can share a claim despite picturing it differently.
A number statement assigns something to a concept. This does not mean the number itself is a concept. In §§55–60, Frege explores why numbers behave like objects. Compare “there is one survivor” and “the number of survivors is one.” In the second sentence, one occurs alongside another expression for what Frege takes to be that very number. The statement identifies an object. It also tells us how many things fall under survivor. Frege needs to explain why these two aspects can go together. One cannot simply point to the object one in the room.
The distinction also matters when a count changes. “The number of people in this room” may designate a different number in five minutes. In §46 Frege compares a condition that changes with the time to one whose reference time has been fixed. The concept person in this room at noon on August 3 does not acquire new instances when someone comes in at half past twelve. The number stays the same. This is compatible with there being something objective to discover about how many people satisfy the present-tense condition.
In §§52–54, Frege distinguishes a characteristic that belongs to the definition of a concept from a property of the concept itself. If a pencil falls under red pencil, it has to be red. But the concept red pencils on this desk can have exactly one instance, even though a pencil itself cannot have exactly one instance in this sense. Adding a second red pencil makes the statement about how many there are false. The first pencil remains red. The change affects a property of the concept in the situation, without altering a characteristic its instances must have. This explains how a numerical statement can be about a concept even when its truth depends on objects you can see and count.
Comparing Counts Before Counting
Suppose every cup in a cupboard is sitting on a saucer, no saucer holds more than one cup, and no saucer is left over. We can tell that there are as many cups as saucers without being able to say how many there are. Conversely, if one cup has been taken away and its saucer remains, the original pairing has stopped showing that the two counts are equal. In that finite cupboard there is now no way to pair every remaining cup with every saucer on the same terms.
Frege calls two concepts equinumerous when their instances can be put in a one-to-one correspondence. His discussion in §§62–63 and §§70–73 makes this precise using a relation. Each object falling under the first concept is related to exactly one falling under the second, and conversely. The cup example happens to supply a spatial relation we can inspect. A pairing could also follow a rule, as when we take the disjoint pairs (1, 2), (3, 4), and so on up to (39, 40), and match each pair with its odd-numbered seat. Nothing in the relation requires us to discover the count of each collection first.
We can now give a criterion for an equation between two numbers. The number belonging to F equals the number belonging to G exactly when F and G are equinumerous. This is now commonly called Hume's Principle. The criterion explains how statements about which objects match can settle an identity between numbers. The proposed condition also has useful features. Any concept is equinumerous with itself under the identity relation. A matching can be reversed. Two successive matchings can be composed. These facts correspond to familiar properties of identity.
The pairing test gives a second way to check that two counts agree. Give two librarians two copies of the same inventory. They can cross off every item on each list against its corresponding item on the other list, or they can count the lines and compare their answers. Either procedure can establish equality of the counts. If they disagree, finding the missed line or duplicate entry can tell them why. The first route makes good sense before either librarian has used a numeral to state how long the list is. That independence is the reason Frege can use correspondence to explain something about number instead of taking number as already explained.
When the identity criterion is given as an equivalence, its two directions matter. Equal numbers imply the existence of a suitable matching. In a small collection we might build the matching simply by arranging the objects in rows. Going the other way, a matching guarantees equality no matter how we have arranged the objects. The count is preserved even if, instead of lining up items by name, we line them up by date of purchase. The criterion singles out the existence of a matching and leaves the particular matching open.
The Julius Caesar Problem
Frege uses another case to test the method in §§64–67. We call two straight lines parallel when they have the same direction. Going in the other direction, we might try to explain an identity between directions by saying that it holds exactly when the corresponding lines are parallel. Given two lines, this can tell us whether their directions agree. It does not directly answer whether the direction of one line is the same object as something introduced by a different description.
For numerical expressions, Frege makes the point sharply in §56. An account of “the number of Fs” should let us tell whether Julius Caesar is identical to that number. The obvious answer is that he is not. Yet the equivalence between equal numbers and a one-to-one pairing supplies an answer only if we already have two expressions of the form “the number of this concept.” The equivalence itself says nothing about whether one of the objects named might be Caesar. Simply excluding anything introduced under a different description would go too fast. Frege needs room for a single object to be introduced in different ways. That is why an identity statement can be worth finding out.
“The number of cups equals the number of saucers” asks whether two number expressions designate the same object. A matching can answer it. “The number of cups is Julius Caesar” also purports to express an identity, but this time the matching rule gives us no second concept to use. The issue is how a proposed definition lets us distinguish its objects from others.
A Definition of Number and a Starting Point
In §68, Frege defines the number belonging to a concept F as the extension of the concept equinumerous with F. An extension is understood here as the object associated with a concept through what falls under it. In this case, what falls under the concept of being equinumerous with F is itself a concept. Related concepts share the same number because each is equinumerous with the same concepts. This gives the equality of numbers a particular proposed object to be about, beyond the pairing rule that tells us which numbers are identical. It also creates an obligation to explain which extensions there are and when they are identical.
The construction starts without presupposing the answer to “what is three?” In §74, Frege defines zero using the concept not identical with itself. No object falls under it, and Frege can make that claim using the laws of identity, without an empirical inspection. The number zero is assigned to this concept. It is also assigned to any other empty concept. An empty matching between the two concepts satisfies the condition on correspondences because there are no objects left unmatched on either side.
The contrast between zero and one is especially revealing. The concept identical with zero has zero as its sole instance, so it is a concept to which the number one applies. Zero is playing different roles in these steps. First it is the number attached to an empty concept. Then it is an object under the concept identical with zero. Treating zero as an object makes the second step possible. Frege uses it in §§76–83 as part of an account of how a number follows another. Roughly, if one concept results from another by leaving out exactly one specified instance, the numbers belonging to the two stand in the appropriate successor relation. The full extension of arithmetic is a further task.
Which Commitments Do the Work?
Frege presents much of the argument in prose in 1884. The later Basic Laws of Arithmetic attempts to carry out the formal derivation. Its Basic Law V governs the identity of courses of values, including extensions associated with concepts. Frege's 1902 reply to Russell acknowledges that the contradiction Russell found bears on that law. The question about extensions in §68 is therefore consequential. A move from a criterion for equality to a theory of objects has to be evaluated by what that theory permits. A logic that supplies these objects too freely can license contradictory claims.
The pairing criterion itself can be understood without accepting Basic Law V. When cups and saucers match, we can understand why they have the same count without specifying numbers as the extensions Frege proposes. The converse question is more demanding. Even if a theory starts with this equivalence, it needs to show what follows about numbers and whether those commitments qualify as purely logical. Distinguishing the criterion from the proposed definition lets us see why the first can be studied after difficulties have emerged for the second.
There is a difference too between using arithmetic to answer a question and using a proof to learn what arithmetic depends on. Establishing the count of people in the room uses information about actual people and a time. Establishing the identity conditions of the numbers used in the answer has a different aim. Frege's inquiry begins with an ordinary number statement and makes that difference explicit. The analysis of counting still explains something even if one remains unconvinced by the proposed foundations.
Reading Frege
The 1884 German text preserves the section numbers. J. L. Austin's English translation is titled The Foundations of Arithmetic. Start with the Introduction, §3 on grounds of proof, and §46 on counting under a concept. §§55–68 give the path from number as an object through equinumerosity to an explicit definition. §§74–83 let you try out those commitments by following the treatment of zero, one, and successor.
Join the discussionAsk a question or share what you learned.
Keep exploring
Reference
Citation details
Cite or share this article.

Comments
Add to the conversation
A question, a correction, or an example from your own work.
No comments yet. Be the first to share your thoughts!