Although the fallacy of composition is standardly classified as an informal fallacy, I see it is a formal fallacy, one rooted in logical form. Let W be any sort of whole (whether set, mereological sum, aggregate, etc.) Suppose each of the proper parts (if any) of W has some property P (or, for the nominalistically inclined, satisfies some predicate F). Does it follow that W has P or satisfies F? No it doesn't. To think otherwise is to commit the fallacy of composition: it is to argue in accordance with the following invalid schema:
1. Each member of W is F
Therefore
2. W is F.
To show that an argument form is invalid, it suffices to present an argment of that form having true premises and a false conclusion. (This is because valid inference is truth-preserving: it cannot take one from true premises to a false conclusion. But it doesn't follow that invalid inference is falsehood-preserving: there are valid arguments with false premises and a true conclusion. Exercise for the reader: give examples.) Here is a counterexample that shows the invalidity of the above pattern: Each word in a given sentence is meaningful; ergo, the sentence is meaningful. (Let the sentence be 'Quadruplicity drinks procrastination.') Since the premise is true and the conclusion false, the argument pattern is invalid. So every argument of that form is invalid, even in the case in which the premises and conclusion are both true.
Why then is Composition standardly grouped with the informal fallacies? Petitio principii is a clear example of an informal fallacy. If I argue p, therefore p, I move in a circle of embarrassingly short diameter. But the inference is valid. (Bear in mind that 'valid' is a terminus technicus.) And if p is true, the argument is sound. Nevertheless, any argument of this form is probatively worthless: it it does not prove, but presupposes, its conclusion. Since this defect is not formal, we call it informal!
So there are clear examples of informal fallacies. But what about Equivocation? It is usually classed with the informal fallacies. Consider the syllogistic form Barbara (AAA-1):
All M are P
All S are M
All S are P.
Suppose there is an equivocation on the middle term 'M.' Although this is an informal defect (in that it has not to do with logical syntax, but with semantics) it translates into a formal defect, the dreaded quaternio terminorum or four-term fallacy, which is of course a formal fallacy: no syllogism with more than three terms is valid. (A syllogism by definition is a deductive argument having exactly two premises and exactly three terms.)
It can be shown that every equivocation on a key term in an argument induces a formal defect. So the standard classification of Equivocation as an informal fallacy cannot be taken too seriously. By contrast, Petitio Principii is seriously informal in its probative defectiveness.
I say that Composition is like Equivocation: it is a formal fallacy in informal disguise. (And the same goes for Division, which is roughly Composition in reverse.) So I disagree with the author of a logic book who writes:
. . . the fallacy of composition is indeed an informal fallacy. It
cannot be discovered by a mere inspection of the form of an
argument , that is, by the mere observation that an attribute is
being transferred from parts onto the whole. . . . The critic must
be certain that, given the situation, the transference of this
particular attribute is not allowed. . . .
So the fallacy of composition is not always a fallacy, but only when it is a fallacy? That is the silliness that the author seems to be espousing. He is saying in effect the following: if you transfer an attribute from parts to whole, that is fallacious except in those cases in which it is not fallacious, i.e. those cases in which the transfer can legitimately be made.
But then what is the point of isolating a typical error in reasoning called Composition? What is the point of this label? Why not just say: there are many different part-whole relationships, and it is only be close acquaintance with the actual subject-matter that one can tell whether the attribute transfer is legitimate?
Logic is formal: it abstracts from subject-matter. So mistakes in logic are also formal. A mistake that is typical (recurrent) and sufficiently seductive to warrant a label is called a fallacy. To say or imply that the fallaciousness of a fallacy depends on the particular subject-matter of the argument is to abandon logic and embrace confusion.
Example. Every brick in this pile weighs more than five lbs; ergo, the pile weighs more than five lbs. This is an example of the fallacy of composition despite the fact that it is nomologically impossible that the pile not weigh more than five lbs.
Another example. Every being in the universe is contingent; ergo, the universe is contingent. This too is the fallacy of composition. And this despite the fact that it is metaphysically impossible that a universe all of whose members are contingent be necessary.
Recent Comments