Overview

This supplement provides a brief overview of some interesting concepts and facts that fall under the broad heading of “metalogic” or “metatheory of logic”. The material here might show up under the guise of a bonus question on an exam, but it is not material I otherwise intend to hold you responsible for. Students interested in an accessible overview of some of these topics might look to David Papineau, Philosophical Devices (Oxford University Press, 2012), Part IV.

Logic vs. Metatheory

We have been doing logic. We’ve learned how to symbolize sentences in logic, how to check for tautological validity using truth tables, and how to construct derivations.

Metatheory involves stepping back and thinking about the logic we have been using: what are its properties? does it work properly? how does it compare to other logical systems?

What is a Logic?

A logic is a formal language together with a system of deduction and a semantics.

A formal language consists of a set of symbols (in our case, the sentence letters, the parentheses, and the five connectives), together with a syntax—that is, rules that determine which combinations of those symbols are well-formed sentences. (The Parsing module was designed to test your understanding of the syntax of our formal language.)

A system of deduction is a set of inference rules (in our case, R, DN, MP, MT, S, ADJ, MTP, ADD, CB and BC), a set of methods of derivation (our case, DD, CD, and ID). When you construct derivations in the Derivation module, you are using our system of deduction.

A semantics is a theory of meaning for a given language. There are various ways of providing a theory of meaning. We have done so by assuming that each sentence letter can either be T or F, and then associating each connective with its characteristic truth function, e.g.,

Other Logics

Other logics either have a different language, a different system of deduction, or a different semantics.

Some logics are different from ours only in trivial ways. For example, our choice of sentence letters was arbitrary: a trivial variant of our logic would use lowercase letters as sentence letters instead. Or we might have used different symbols for our connectives.

Some logics differ from our in less trivial ways. Our logic is a sentential logic: the basic atoms, or logical units, are sentence letters. Predicate logic, which we will study next, takes the basic atoms to be predicates, like ‘is sitting’, and singular terms, like ‘Socrates’. Term logic, which we looked at briefly in the introduction, takes the basic logical units to be common nouns, like ‘Cars’ and ‘Horses’.

Our logic is a truth-functional sentential logic: the connectives in our language are all truth-functional. That is why, if you know the truth values of the atomic sentence letters in a complex sentence, you can calculate the truth value of the complex sentence. But there are sentential connectives in English that are not truth-functional: ‘because’ is one, as is the non-material conditional.

Two unary connectives that are not truth functional are ‘possibly’ and ‘necessarily’. Consider:

• ‘2+2=4’ is true and ‘Necessarily 2+2=4’ is true.
• ‘There are cows’ is true but ‘Necessarily there are cows’ is false.

Likewise,

• ‘2+2=5’ is false, and ‘possibly 2+2=5’ is false.
• ‘There are purple cows’ is false, but ‘Possibly there are purple cows’ is true.

So ‘possibly’ and ‘necessarily’ are not truth-functional connectives. The logic of ‘possibly’ and ‘necessarily’ is called modal logic. ‘◻’ is used to represent ‘necessarily’, and ‘◇’ is used to represent ‘possibly’.

Truth-Functional Completeness

Confining our attention to truth-functional sentential logics, we might consider logics that have more connectives than ours: for example, a logic that includes a connective for exclusive ‘or’, or a connective for ‘neither…nor…’

Here is an interesting fact about our logic: it is truth-functionally complete. That means that every truth-function is expressible in our language. For example, we can express the truth-function that goes along with an exclusive ‘or’ using the sentence ‘(P ∨ Q) ∧  ∼ (P ∧ Q)’, read in English as ‘P or Q, but not both’. And we can express the truth-function that goes along with ‘neither…nor…’ using ‘ ∼ (P ∨ Q)’.

So adding more connectives to our language would not add to its expressive power. What about removing connectives?

We already know that every conditional is equivalent to some disjunction, and vice versa. This means that our language would remain truth-functionally complete even if we got rid of one or other of these two connectives.

As a matter of fact, the language of Chapter 1, containing only ‘→’ and ‘∼’, was truth-functionally complete. Likewise, a language containing only ‘∨’ and ‘∼’ would be truth-functionally complete, as would a language containing only ‘∧’ and ‘∼’.

Challenge: find a sentence, using only ‘∼’ and ‘∨’, that expresses the same truth function as ‘P ∧ Q’.

In fact, it is possible to have a truth-functionally complete language that contains only a single connective. The Sheffer stroke has the following truth table:

This connective is sometimes called ‘NAND’ in programming contexts. Looking at the truth table, do you see why it might give given that name?

Challenge: using just ‘↑’, find a sentence that is equivalent to ‘ ∼ P’.

Of course, if our language only had ‘↑’, we’d need different rules and methods of derivation. What should they be?

Classical and Non-classical Logics

Our logic is a classical logic. That means it has several important features. For example, in our logic, each of the following is a theorem:

The Law of Excluded Middle:

∴ P ∨  ∼ P

The Law of Non-Contradiction:

∴  ∼ (P ∧  ∼ P)

The Principle of Explosion

∴ P ∧  ∼ P → Q

Also, in our logic, you can always add additional irrelevant premises to an argument, without affecting its derivability (derivability is “monotonic”). That is, for any list of sentences, Γ, and any sentences ◻ and ○,

Monotonicity:

If there is a derivation from Γ to ◻, then there is a derivation from Γ . ○ to ◻

Each of these properties might be thought problematic: the Law of Excluded Middle, if you think that some sentences lack truth values, or that there are truth values other than T and F; the Law of Non-Contradiction and the Principle of Explosion, if you think that some contradictions are true; and Monotonicity, if you think the premises of an argument must be relevant to the conclusion. This has led to the development of a wide variety of non-classical logics.

You might be a logical monist, and think of classical logics and non-classical logics as competitors for the title of the one correct logic. Or you might be a logical pluralist, and think that there is no such thing as the one correct logic. The issues here quickly get complicated in fascinating ways. For more, see the Stanford Encyclopedia of Philosophy article on logical pluralism.

For a decent overview of the issues dividing classical and non-classical logics, see the Wikipedia articles on Classical Logic and Non-classical Logic. For a more advanced and rigorous overview, see several articles in the Stanford Encyclopedia of Philosophy, including:

• Classical Logic
• Intuitionistic Logic
• Linear Logic
• Paraconsistent Logic
• Relevance Logic

Soundness and Completeness

Our logic includes both a system of deduction—a set of rules together with methods of derivation—and a truth-functional semantics. It also has two important features, where Γ is any set of symbolic sentences and ◻ is any symbolic sentence:

Soundness

If there is a derivation from Γ to ◻, then Γ∴ ◻ is tautologically valid.

Completeness

If Γ∴ ◻ is tautologically valid, then there is a derivation from Γ to ◻.

The first of these, Soundness, says that our derivations never go wrong: you cannot construct a derivation from true premises to a false conclusion. This would not be true if we introduced an invalid rule—say, a rule that allowed you to infer P from P ∨ Q.

The second, Completeness, says that every tautologically valid argument can be derived. This would not be true if we had not introduced enough rules and methods, or if our rules and methods were not powerful enough.

Some of our rules and methods are redundant: we could get rid of them, and our system would remain complete. For example, every derivation completed using DD could instead be completed using ID (how?) DN is also redundant: given  ∼  ∼ Q, you can derive both Q and  ∼  ∼  ∼  ∼ Q using nothing but ID and R (can you do this?) MT is similarly a rule we could do without.

This suggests that we might “purify” our system by removing DD, DN, MT, and any other redundant basic rules and methods (we could always reintroduce them as derived rules or methods). For an example of a system that is similar to ours but more pure in this sense, see Bergmann, Moor, and Nelson’s The Logic Book.

A standard course in Metatheory of Sentential Logic would culminate in proofs of soundness and completeness. You can find an accessible sketch of such proofs in Papineau’s Philosophical Devices, and you can find more rigorous proofs in any of several logic texts, including The Logic Book (but note that the sentential logics presented in both Philosophical Devices and The Logic Book differ from ours both in their choice of symbols and, more importantly, in their choice of rules and derivation methods).