How Arguments Work: Claims, Reasons, and Deduction, Induction and Abduction

Start here"All toasters are made of gold; everything made of gold is a time-travel device; so all toasters are time-travel devices" is a valid argument. An invalid argument can still end in a true conclusion. So what does a good argument need?

An argument, in the logic sense, is not a fight. It is a claim plus the reasons offered for it. This page shows how to find the two in everyday talk and how to tell whether the reasons work. It also covers the three ways of reasoning that logicians name: deduction, induction and abduction.

011. What an argument is

In logic, an argument is a set of reasons offered for a conclusion, not a quarrel. The word “argument” can mean a dispute or a fight, or it can be used in a more technical sense. In philosophy, an argument is a set of statements, called premises, offered as grounds for another statement, the conclusion.

So every argument has two parts: a claim you are asked to accept, and the reasons offered for it. Each part is something that can be true or false. On this page, “claim” and “conclusion” mean the same thing, and “reason” means the same as premise.

For example, “The last bus leaves at eleven, and it’s already quarter past, so we’ll need a taxi.” The claim is “we’ll need a taxi”. The reasons are the bus time and the clock.

Explanations look similar but are not usually counted as arguments. If someone asks why a metal expanded and you answer that it was heated, and all metals expand when heated, your reasons help them understand why it expanded rather than give them grounds to believe that it did.

Finding the claim and its reasons

Signal words help, but they are clues, not proof. Typical premise indicators include “because”, “since”, “for” and “as”; typical conclusion indicators include “therefore”, “thus”, “hence” and “so”. The same encyclopedia warns that these words do not always work this way.

In practice, ask two questions. What am I being asked to believe? What is offered as the reason? The answer to the first is the conclusion, even when it comes first in the sentence. For example, in “We should move the meeting online, since the trains are on strike”, the claim comes first and “since” marks the reason.

The reason nobody says out loud

An argument can leave a reason unspoken. Logicians call an argument with at least one part left unstated an enthymeme.

For example, “She’s a nurse, so she’ll know what to do” rests on an unspoken reason: that nurses know what to do in this situation. Our reading: saying the hidden reason out loud often shows where the real disagreement is. Filling in that gap fairly is what the principle of charity and steelmanning are about.

022. Valid and sound: two different questions

Valid asks whether the reasons would force the conclusion; sound asks whether they are also true. A deductive argument is valid if its form makes it impossible for the premises to be true and the conclusion false. It is sound if it is valid and all its premises are actually true.

So validity is about the shape of the reasoning; soundness adds the facts. A valid argument can still end in a false conclusion when one of its premises is false.

The encyclopedia’s example: all toasters are made of gold, everything made of gold is a time-travel device, so all toasters are time-travel devices. It is valid, because if the premises were true the conclusion would have to be, but it is not sound.

It works the other way too. Arguments that are invalid, or valid but unsound, can still have true conclusions. A weak argument does not make its conclusion false. It only means this argument does not show it. Our cheat sheet calls this the fallacy fallacy.

A quick test for a valid shape

Keep the shape, swap the words: true reasons with a false conclusion mean the shape is invalid. On the view most logicians hold, validity is decided by the argument’s form: words such as “all”, “and”, “not” and “some” that remain when the topic words are taken out. Aristotle showed forms were invalid by building counterexamples; in modern logic, a single instance of a form with true premises and a false conclusion is all it takes.

One of the encyclopedia’s examples: all basketballs are round; the Earth is round; therefore the Earth is a basketball. Its shape is “All A are F; X is F; so X is an A”. For example, the same shape turns up in “Good leaders are confident; he’s confident; so he’s a good leader”. The basketball version shows that the shape alone does not get you there.

Why a believable ending feels like a good argument

People tend to accept arguments whose conclusions they already believe, valid or not. Researchers call this belief bias. In one classic example, “All flowers have petals. All roses have petals. Therefore, all roses are flowers” is invalid, yet many people judge it valid because the conclusion fits what they know. Syllogisms with believable conclusions are accepted more often than syllogisms of exactly the same form with unbelievable ones. Researchers have proposed several competing theories of why this happens.

How sure: Strong found across three decades of syllogism studies; why it happens is still argued.

For example, before you agree with an argument, run the same steps toward a conclusion you dislike, and see whether they still feel convincing.

033. Three ways to reason

Deduction aims at a conclusion that must follow, induction at one that is likely, abduction at the best explanation. Philosophers normally count three major types of inference: deduction, induction and abduction. In deduction, true premises guarantee a true conclusion; in induction and abduction they do not. In those two, the conclusion goes beyond what is contained in the premises, which philosophers call ampliative.

	Deduction	Induction	Abduction
What it does	Applies a rule to a case	Generalises from cases seen	Forms a hypothesis, or picks the best explanation
What it can give	A conclusion that must follow, if the premises are true	A likely conclusion	A best guess, open to a better one
How it is judged	Valid or invalid; sound or unsound	Strong or weak; cogent or not	How well it explains, against the rivals

Deduction: the reasons force the conclusion

Deduction can give certainty, but only about what follows from your premises. Aristotle defined a deduction as speech in which, “certain things having been supposed, something different from those supposed results of necessity because of their being so”. His Greek word was sullogismos; in modern English, “syllogism” means an argument of one very specific form.

The textbook example: Socrates is a man; all men are mortal; therefore, Socrates is mortal. A three-line argument of this shape is a classic syllogism. Charles Sanders Peirce called deduction “merely the application of general rules to particular cases”.

For example, every ticket in this sale is non-refundable, and this ticket is from the sale, so it can’t be refunded. If both reasons are true, the conclusion is too. Deduction never checks its own starting points, so the real work is checking the premises.

Induction: from what you’ve seen to what you haven’t

Induction makes a conclusion likely, never certain. Inferences from the observed to the unobserved, or to general laws, are known as inductive inferences. Aristotle described induction as argument from the particular to the universal.

An encyclopedia example: most Greeks eat olives, Socrates is a Greek, so Socrates eats olives, which is likely but not guaranteed. Inductive arguments are judged strong or weak rather than valid or invalid, and a strong one with true premises may be called cogent.

David Hume set out the classic problem with induction in 1739, and gave a shorter version in 1748. He wrote that all inferences from experience suppose, “as their foundation, that the future will resemble the past”. Hume did not deny that we reason this way; he asked what justifies it. So induction is how we learn from experience, and it always leaves room for a surprise.

For example, the last ten buses on this route ran late at rush hour, so tomorrow’s probably will too. Our reading: more cases, and fairer ones, make that stronger. Too few cases is the hasty generalisation in our fallacies cheat sheet.

Abduction: the best explanation you can find

In its common modern sense, abduction picks the explanation that best fits the facts, and stays open to a better one. In that sense it is also called inference to the best explanation. An encyclopedia example: you know two friends had a terrible row, then hear they were seen jogging together, and you conclude they have made up. The conclusion does not follow logically; what leads you to it is that it would best explain what you heard.

The term “abduction” was coined by the American philosopher Charles Sanders Peirce. Around 1865 he began to hold that probable reasoning came in two distinct kinds: induction, and a second kind he called by several names, hypothesis, retroduction and abduction. He gave it a short pattern: “The surprising fact, C, is observed. But if A were true, C would be a matter of course. Hence, there is reason to suspect that A is true”. For Peirce, abduction was the step of forming a hypothesis worth testing. In its most common modern sense, inference to the best explanation, it is about judging which hypothesis to accept. Peirce’s writings on abduction never refer to a best explanation. So the word has two related uses: Peirce’s, for coming up with a guess, and today’s, for choosing the best of the guesses you have. The phrase “inference to the best explanation” is the title of a 1965 paper by the philosopher Gilbert Harman.

Astronomers John Couch Adams and Urbain Leverrier explained why Uranus’s orbit departed from predictions by proposing an undiscovered planet, and not much later, Neptune was found.

What it gives is a best guess that new facts can undo. If you learn that the two joggers are former business partners with money matters to discuss, “they made up” may no longer be the best explanation. The best explanation you thought of is only the best of those you considered, which may be what the philosopher Bas van Fraassen called “the best of a bad lot”.

Experimental studies have shown that when people can think of an explanation for a possible event, they tend to overestimate how likely it is. One study found that, in some situations, people tend to greatly overrate the probability of simpler explanations. The encyclopedia adds that these studies were not directly about abduction, but suggest that leaning on explanations may not always help.

How sure: Unsettled most philosophers agree abduction is common; its exact form and how far to trust it are still argued.

For example, the kitchen floor is wet and the dog’s bowl is tipped over, so the dog probably knocked it. Before settling, ask what else would explain it, and what you would expect to see if you were wrong. Simplicity is one of the qualities often used to judge the best explanation, alongside generality and fit with well-established theories. For where the “simplest is best” rule comes from, see Occam’s razor.

Peirce’s bag of beans: the three side by side

Peirce showed all three as the same three statements, put in a different order. In an 1878 article, “Deduction, Induction, and Hypothesis”, Peirce illustrated the three with bags of beans. The three statements are a rule (all the beans from this bag are white), a case (these beans are from this bag) and a result (these beans are white).

• Deduction: from the rule and the case, conclude the result.
• Induction: from the case and the result, conclude the rule. Peirce called induction “the inference of the rule from the case and result”.
• Hypothesis, his name for abduction in this article: from the rule and the result, conclude the case. He finds a handful of white beans on a table and a bag holding only white beans, and guesses that the handful came from that bag.

He gave everyday cases too. Landing at a seaport in a Turkish province, he met a man on horseback under a canopy held by four horsemen, and inferred he was the governor. He added that “as a general rule, hypothesis is a weak kind of argument”.

044. How the three work together

In Peirce’s account, you guess by abduction, work out what should follow by deduction, and test by induction. In his later work, Peirce treated abduction, deduction and induction as three phases of scientific method: making a conjecture, inferring what should follow, and testing. If the predicted results do not appear, the method loops back to abduction for a new hypothesis.

For example, your phone won’t charge. You guess the cable is faulty. If so, a different cable should work. You try another cable a few times. If the phone still won’t charge, you go back and guess again.

The line between the types is less tidy than it first looks. One encyclopedia article finds the many proposals for separating deductive from inductive arguments in “a state of largely unacknowledged chaos”. It suggests asking of any argument whether it proves its conclusion, only makes it probable, or does neither.

055. Easy mix-ups

Each of these is easy to slip into; each has a one-line fix.

• Mixing up “valid” and “true”. In logic, valid only means the conclusion would follow if the reasons were true.
• Thinking a flawed argument means a false conclusion. One cannot reject a conclusion simply because an argument for it is flawed.
• Taking “general to particular” as the test for deduction, and “particular to general” for induction. One encyclopedia article calls this rule “much too crude”. An argument made only of general statements, such as “All As are Bs; all Bs are Cs; so all As are Cs”, would count as neither.
• Trusting an argument because its conclusion is believable. That is belief bias, above.
• Taking the best explanation you can think of as true. It is only the best of the ones you considered.
• Treating an explanation as an argument. An explanation helps you understand why something is so; an argument offers reasons to believe that it is so.

06Quick answers

07Use this

08Quick quiz

How we checked this · Last checked November 2026

• Definitions come from the Internet Encyclopedia of Philosophy (Argument; Deductive and Inductive Arguments; Validity and Soundness) and the Stanford Encyclopedia of Philosophy (Aristotle's Logic; Abduction and its supplement on Peirce; Charles Sanders Peirce; The Problem of Induction), each read on the day.
• We read Peirce's 1878 article in the Wikisource transcription and Hume's Enquiry in the Project Gutenberg text.
• For belief bias we read the 2018 paper's full text and abstract; we did not open the older studies it builds on.
• Peirce's short pattern for abduction is quoted from the Stanford Encyclopedia, which gives it no date, so we give none.
• The experiments on explanations are as reported in the Stanford Encyclopedia; we did not open those studies.
• Everyday examples that start "For example" are ours and are not data.

How we check every article →

09Sources