Analogy: What Makes One Good or Bad

6. The modesty of the conclusion. The less the conclusion claims relative to the premises, the stronger the argument. Concluding that B is probably liable is far better supported than concluding that B is certainly liable to the same extent and on the same terms.

Putting the tests in order

An examination answer that lists six tests as equals has missed the point. They are not equal.

Relevance is the master test. Numbers and variety matter because they make relevance more likely to be real rather than accidental; they have no independent value. An analogy with two relevant resemblances beats one with fifty irrelevant ones, always.

Modesty is free. It costs nothing to weaken the conclusion and it improves every analogy.

Disanalogies are the opponent's tool and the arguer's blind spot. Anyone constructing an analogy will have thought about the resemblances and will not have looked for the differences, which is why the differences are where an argument by analogy is attacked.

A worked example: one analogy, six tests

The situation. A court has held that a shopkeeper who left a loose floor tile unrepaired for a week is liable to a customer injured by it. A different case now arises: a bank has left a loose floor tile unrepaired for a week and a customer has been injured.

The analogy. The two occupiers resemble each other in inviting the public onto premises for the purposes of their business, in controlling the premises, in the nature of the hazard and in the period for which it was left. The shopkeeper was liable, so the bank is probably liable.

Now run the tests.

Number. One prior case, which is the minimum. A line of decisions covering shops, restaurants and cinemas would be much stronger.

Variety. Nil, since there is only one instance. If the line existed and covered shops, restaurants and cinemas, the variety would be high and would show that the liability does not depend on being a shop.

Number of resemblances. Four, listed above.

Relevance. Strong, and this is the whole argument. Inviting the public, controlling the premises, the hazard, the delay: every one of them bears on why an occupier is liable at all. Add that both premises had white walls and the number of resemblances rises to five and the strength of the argument does not move at all.

Disanalogies. Look for them deliberately. The bank's floor may be cleaned by a contractor rather than by staff; the customer may have been in a part of the premises not open to customers; the tile may have been loosened that morning by another customer. The first is probably irrelevant, because control is what matters and not who does the cleaning. The second and third are highly relevant, and either would defeat the analogy.

Modesty. "The bank is probably liable" follows. "The bank is liable in the same amount" does not, because nothing in the analogy concerns quantum.

The result. A reasonably strong analogy, resting almost entirely on the relevance of four resemblances, and vulnerable to exactly two facts. That is a usable answer, and it is what the six tests are for.

Bad analogies and how they go wrong

The false analogy is the standard fallacy of this form: an argument whose resemblances are numerous but irrelevant. "This agreement, like the one in that case, was made on stamp paper of the same value, in the same city, before the same sub-registrar, so it must be enforceable in the same way." Nothing in the list bears on enforceability.

The analogy stretched too far takes a comparison that holds for one purpose and carries it into another. A company is like a person for the purposes of owning property and suing; it is not like a person for the purposes of imprisonment.

The analogy that ignores the crucial difference is the commonest in practice, because the arguer is looking at resemblances by construction.

The analogy used as proof rather than as support. An analogy can never establish its conclusion with certainty, and an argument that treats it as if it had has mistaken an induction for a deduction.

Distinctions that carry marks

What this does not mean

More resemblances is not automatically better. Adding irrelevant resemblances adds nothing, and an answer that says "the more points of similarity the stronger the analogy" without the qualification about relevance has stated the common error.

An analogy is not a metaphor. A metaphor illustrates; an analogy argues. Calling a company a person may be either, and only one of them supports a conclusion.

A weak analogy is not invalid. The word does not apply to inductive arguments at all, as chapter 90 established.

Quick revision

Form: A and B resemble each other in p, q, r; A has s; therefore B probably has s.

Six tests: number of instances; variety of instances; number of resemblances; relevance of the resemblances; number and importance of disanalogies; modesty of the conclusion.

Relevance is the master test. One relevant resemblance beats a hundred irrelevant ones.

Modesty is free and improves every analogy.

Disanalogies are where an analogy is attacked, because the arguer will not have looked for them.

False analogy: many resemblances, none relevant. The standard fallacy of this form.

It moves particular to particular and never states a general rule, which is why one prior case is enough to argue from.

Test yourself

1. Define argument by analogy and give its form.

An argument by analogy infers, from the resemblance of two or more things in certain respects, that they resemble one another in a further respect. Its form is: A and B resemble each other in respects p, q and r; A has the further property s; therefore B probably has s. Its conclusion is probable only, and it moves from one particular to another without asserting any general rule.

2. State the characteristics of a good analogy.

A greater number of instances; greater variety among them, so that the resemblance relied on is not accidental; a greater number of respects in which the things resemble each other; relevance of those resemblances to the property inferred; few and unimportant disanalogies; and a modest conclusion that claims no more than the premises support.

3. Which of the tests is decisive, and why?

Relevance. A resemblance counts only in so far as it is connected with the property being inferred, so a single relevant resemblance outweighs any number of irrelevant ones. Two documents may be alike in paper, typeface, length and place of execution without any of it bearing on enforceability, while one resemblance in the capacity of the parties bears on it completely. The other tests matter because they make a relevance claim more credible, not on their own account.

4. What is a false analogy?

An argument by analogy in which the resemblances relied on, however numerous, have no connection with the property inferred. It is the characteristic fallacy of this form and it is persuasive precisely because the list of similarities can be made long. The answer to it is not to deny the similarities, which are usually admitted, but to show that none of them bears on the conclusion.

5. Why are disanalogies the natural place to attack an argument by analogy?

Because whoever constructed the argument was looking for resemblances and will not have searched for differences. The tests for a difference are the same as for a resemblance: what matters is whether it bears on the property inferred. A difference in some feature unconnected with the conclusion is harmless, while a single relevant difference may destroy the argument by itself.

6. Why does weakening the conclusion strengthen an analogy?

Because the strength of an inductive argument is the relation between what the premises support and what the conclusion claims. Reducing the claim leaves the same support carrying a lighter load. Concluding that a party is probably liable is far better supported than concluding that he is certainly liable in a stated amount, and since the modest conclusion is usually all that is needed, the improvement costs nothing.