A good logic puzzle gives you enough information to reach a conclusion, but not necessarily in the form you would prefer. The challenge is to track implications precisely. Words such as “if,” “only,” “every,” and “at least” do real mathematical work. Replace one casually and you may solve a different problem. The first discipline of deduction is therefore simple: make the rules explicit before trying to be clever.
Which cards must you turn over?
Four cards lie on a table, showing A, D, 4, and 7. Every card has a letter on one side and a number on the other. You want to test this rule: “If a card has a vowel on one side, then it has an even number on the other.” Which cards must you turn over to check whether the rule holds for these four cards? Choose the smallest sufficient set.
This is a version of the selection task associated with psychologist Peter Wason. Its difficulty comes less from arithmetic than from our tendency to seek examples that agree with a rule. A proper test also looks for the one kind of example that could break it.
Choose your inspection
Hint: look for a violation
The rule fails only when a vowel is paired with an odd number. It says nothing about what must be behind a consonant.
Reveal the full explanation
Turn over A to check that its number is even. Turn over 7 to check that its letter is not a vowel. D cannot violate the rule, regardless of its number. A 4 can have either a vowel or a consonant behind it, because the rule does not say that every even number must have a vowel.
In symbols, “if P, then Q” rules out P together with not-Q. It does not assert “if Q, then P.” That reversed claim is the converse, and it needs its own justification. The contrapositive, “if not-Q, then not-P,” is logically equivalent to the original statement. Keeping converse and contrapositive separate prevents a surprisingly large family of mistakes.
Truth-tellers, liars, and useful assumptions
Suppose two inhabitants, Ada and Ben, each belong to one of two types. Truth-tellers make only true statements; liars make only false statements. Ada says, “We are both liars.” What can you conclude? Treat the sentence as a complete statement about these two people, and assume the type rules apply without exceptions.
If Ada were a truth-teller, her statement would be true, making her a liar. That contradiction eliminates the truth-teller case. Ada is therefore a liar, and her statement is false. Since she is already a liar, the claim that both are liars can be false only if Ben is a truth-teller. No guess about personalities, motives, or conversational politeness is relevant.
Notice that this conclusion depends on the artificial rules. In ordinary life, a person who lies once is not someone whose every possible sentence is false. Logic puzzles build simplified worlds so that implications become exact. Raymond Smullyan's What Is the Name of This Book? is a celebrated exploration of this style of reasoning, including much more intricate self-reference.
Knowing is not always knowing together
Some of the hardest reasoning problems are about information rather than truth values. A fact can be true, known by everyone, and still not be common knowledge. Common knowledge means everyone knows it, everyone knows everyone knows it, and that pattern continues through every finite level. Public announcements can therefore matter even when they appear to add no new factual content.
Consider a simplified two-person hat puzzle. Both people can see the other's hat but not their own. Hats are either red or white. Both hats are actually red. The rules, each person's perfect reasoning ability, and the public answering rounds are common knowledge. An announcer publicly says, “At least one hat is red.” In the first round, each person is asked whether they know their own color; both say no.
In the next round, each can infer that their own hat is red. If one person's hat were white, the other would have seen white and used the announcement to identify their own hat as red immediately. The first-round uncertainty rules that out. The announcement and the public responses create a shared reasoning structure. Timing and public access are essential; remove them and the conclusion need not follow.
A practical deduction checklist
Translate a complicated sentence into a short rule. Ask what would make it false. Separate cases that cannot both occur. When possible, construct a small table of candidate worlds and cross out only those directly contradicted by the evidence. If two worlds remain, you have not yet proved a unique answer, however appealing one of them feels.
George Pólya's How to Solve It offers a broadly useful problem-solving attitude: understand the problem, devise a plan, carry it out, and look back. In logic, the looking-back step includes checking whether you accidentally assumed the conclusion. A valid deduction should survive a skeptical reader who accepts the stated rules but shares none of your intuitions.
That is the deeper reward of a good reasoning problem. You are not training yourself to distrust every first thought. You are learning to ask what evidence would distinguish a plausible answer from a necessary one.