Ad Hominem
A reader's summary of the best-known argumentative fallacy — answering the person instead of the claim — covering Locke's older and milder meaning of the term, Schopenhauer's stratagem of getting personal, the three textbook forms, Paul Graham's hierarchy of disagreement, and the cases where looking at the speaker is not a fallacy at all.
The term at a glance
“Ad hominem” is Latin for “to the man.” In everyday use it has become a general word for a personal attack, and in logic it names the fallacy of rejecting a claim by criticising whoever made it. It is the usual first entry in lists of informal fallacies and the label most often thrown at a stranger in an online argument — frequently, as discussed below, without checking whether it fits.
Origin
The Latin phrase belongs to the old logic of disputation, where it described arguments aimed at a particular opponent rather than at the matter in general. Locke gave it a place in English philosophy in 1689, and Richard Whately's Elements of Logic (1826) kept the meaning of an argument built from the opponent's own admissions. The modern sense — insult in place of reasoning — settled over the nineteenth and twentieth centuries as textbooks collected it into their lists of fallacies. In 1970 Charles Hamblin's Fallacies criticised that textbook tradition for treating such labels as self-evident, and the argument-theory literature that followed, notably Walton's, has been more careful about when the label applies.
History and context
Rhetoric has always known that audiences judge speakers as well as speeches; Aristotle counted the speaker's character as one of the three modes of persuasion. What the fallacy tradition adds is a discipline for keeping the two apart when the question is whether a claim is true. Schopenhauer's notebook of stratagems shows how old the temptation is, and the same move appears in the related patterns catalogued elsewhere on this site: poisoning the well delivers the attack before the person has spoken, whataboutism is a tu quoque scaled up to politics, and Godwin's Law describes where an unrestrained thread tends to end up. The internet did not create the fallacy; it supplied an audience, a profile history and a reply button within reach of every participant.
Main ideas
The Latin means 'to the person'
Argumentum ad hominem shifts the target of a reply from what was said to who said it. The structure is simple: a claim is made, the speaker is attacked, and the attack is offered as if it settled the claim. The fallacy lies in the last step — a speaker's flaws, however real, are a separate question from whether the claim is true.
Locke used the term for something milder
In Book IV of the Essay Concerning Human Understanding (1689), John Locke listed ad hominem among four 'sorts of arguments' and meant pressing a man with the consequences of his own principles or admissions. That older sense — 'you said so yourself' — is closer to what logicians now call an internal critique, and it is why some modern authors resist treating every use of the term as a fallacy.
Schopenhauer filed the abusive form under last resort
Arthur Schopenhauer's The Art of Being Right (Eristische Dialektik, written around 1830 and published after his death) lists thirty-eight stratagems for winning a dispute regardless of truth. The last, number 38, is to become personal, insulting and rude once the argument is lost — the stratagem he says works because it is always available and always costs the attacker nothing.
Textbooks usually sort it into three forms
The abusive form attacks the speaker's character ('you are a fool, so you are wrong'). The circumstantial form points to the speaker's situation or interests ('you would say that'). The tu quoque — Latin for 'you too' — answers a criticism by noting that the critic does the same thing. Douglas Walton's 1998 book Ad Hominem Arguments is the standard modern treatment of how each form can be weak in one context and reasonable in another.
Pointing at interests is sometimes just evidence
At the 1963 trial of Stephen Ward during the Profumo affair, Mandy Rice-Davies was told that a peer she had named denied any involvement; her reply, 'Well he would, wouldn't he?', has become the canonical example. A person with a stake in the outcome is a less reliable witness on it, and saying so is an argument about the weight of testimony, not a refusal to engage. The error begins only when the stake is treated as proof that the claim is false.
Graham ranked disagreement from name-calling to refutation
Paul Graham's 2008 essay How to Disagree arranged responses on a seven-step pyramid: name-calling at the bottom, then ad hominem, responding to tone, contradiction, counterargument, refutation, and at the top refuting the central point. The ladder is useful because it shows that most online replies sit on the lowest rungs and that moving up costs effort rather than intelligence.
Online, the person is the only thing on screen
A comment thread presents a handle, an avatar and a few lines of text. Attacking the handle is faster than reading the argument, and a profile page offers a ready supply of material — past posts, follower counts, affiliations. The reply can feel responsive while saying nothing about the claim, and a pile-on of such replies makes the original point effectively unanswerable by anyone but its author.
Critique
- Credibility is a legitimate question. Courts allow a witness to be impeached for a record of dishonesty (in the US federal rules, Rule 608 deals with character for truthfulness), and a reader is right to ask who funded a study. When the claim rests on the speaker's word alone, the speaker's reliability is part of the evidence. The fallacy applies when the claim can be checked independently and the attack is used to avoid checking it.
- The accusation is itself overused. Calling any reply that mentions the speaker an “ad hominem” can become a way of dodging a fair challenge to one's credentials, motives or consistency. A short test: would the criticism still stand if the claim were true? If the answer is yes, it was probably not meant to refute the claim.
- Tu quoque can be a point about consistency. That a critic does what they condemn does not make the condemned act acceptable, but it can fairly show that the critic does not believe their own standard, or that the standard is being applied selectively. What it cannot do is answer the criticism itself.
Impact
The working lesson is to separate two questions that an argument constantly blurs: is this true? and should I trust this person? Both are fair, but only the first decides the claim, and the second only decides how much checking it needs. It pairs naturally with steelmanning, which asks for the strongest version of the claim before any verdict on its author, and with tone policing, the mirror error of attacking how something was said rather than what. On a platform where the cheapest reply is to mock the person, answering the point is the rarer and more persuasive move.