Crítica, Revista Hispanoamericana de Filosofía, Volume 10, number 30, December 1978
Tipos de causalidad
Ernesto Sosa
Brown University

Abstract: To elucidate the concept of causality was a matter of special importance for empiricists such as Hume, because for them causality is not, at least in a strict sense, perceived. If the causal relationship is not a temporal relationship between two phenomenons, what kind of a relationship is it? For a causal relationship to exist there must be, according to Hume, a universal connection between type A antecedent phenomenons and type B subsequent phenomenons, so that whenever A occurs, B must also occur. From here are originated the “nomological” explanations of causality, which accept the requisite of universality, but furthermore require the existence of a natural law according to which B must occur whenever A occurs.
I. All nomological explanations of causality accept (T): An occurrence, or state of things, P, is a causal factor of another one, Q, only if there are such actual conditions, C, and such a law of nature, L, that if P, C, and L exist together, Q must also exist; but without L, the existence of P and C does not imply the existence of Q.
Furthermore, it is assumed (i) that natural laws, or their expression, do not refer to any particular being, i.e., they are natural laws, and (ii) that laws of nature, as opposed to those of logic, are not necessary truths.
Let us propose the principle (L): If a type τ wood-plank comes into relationship R with a type κ wooden box, at a given time and place, then a table begins to exist at that time and place.
Such a generation is, to be sure, a paradigm of causality. However, what law of nature, or better still, what semi-law or principle could play the role there required by the nomological explanations of causality?
Assuming C — plank T belongs to type τ, and box K to type к — it is possible to explain, without violating thesis (T), the fact that the existence of the table springs from the union of T and K. Let Q be the table coming into being at time t and place l, and let P be the state of things in which T enters into relationship R with K at time t and place l. If P, C, and L exist together, this implies Q.
In this way a victory, however partial, is won for the nomological method. But there is still the problem of understanding how the union of plank T and box K causes this table — table M — to exist. The problem is not raised by the fact that the effect involves a particular entity, table M; it is raised because the effect is the generation of something.
The answer to the problem lies in relinquishing the idea that contingent, purely general principles underlie every instance of causality — which is a central idea for the nomological model. This shall allow us to widen our perspective in order to include causality types, such as generation, that do not rest upon those principles.
II. E. Sosa considers then a more general argument, whose purpose is to prove the nomological model to be incompatible with the coming into being of anything for whatever cause, and not merely with a coming into being caused by the required components’ entering into the required relationship.
If mereological essentialism is true, the nomological model may help to explain at least one type of generation. But the fact that the nomological model helps to explain how some causes result sometimes in the generation of some entity does not imply that generation offers no difficulties for the nomological model. There may be some other types of generation which such a model is unable to explain.
III. The following are instances in which it is unclear how the causal relationship might be explained by means of the nomological model:

(d1) A thing such as table M exists at moment to because at to plank T rests upon box K;

(d2) Plank T rests upon box K at to because it is then that it comes to rest there;

(d3) Plank T comes to rest upon box K at to because the carpenter puts it there at that moment.

What are, then, the purely general contingent principles with which the above statements share the relationship required by the nomological model?
IV. The above argument can be partly rounded in the following manner: If there is now a composite A, then there are now entities C1 and C2, and a relationship R, such that (i) if C1 is now into relationship R with C2, A necessarily exists now, and (ii) C1 might now, possibly, enter or have entered into relationship R with C2. But, as a result of this, A would now come or have come into being. So, if there is now a composite, there is a possible instance of causality (i.e., the generation of that composite) which cannot be understood by means of the nomological model. In such a way, therefore, composites can have no true existence for the nomological model; they can only be, at most, shadows or figments.
V. The common feature in all forms of causality appears to be the necessary implication. If this is true, we can interpret nomological causality as a relationship which binds together a state of things P that includes a causal law, and a result Q. Under this interpretation, the nomological cause as a whole does imply necessarily a result.
Assuming the necessary implication as an essential factor in every type of causality, then whenever something, Q, is a result or a consequence of something, P, it is a necessary truth that, if P exists or occurs, Q must also exit or occur: P necessarily implies Q. However, P can necessarily imply Q without Q being a result or a consequence of P. To find out the special factors which distinguish a causal from a non-causal necessary implication is a task for analysis.
To sum up, nomological causality is a relationship between a result or consequence and that to which it owes its being, and the same thing holds for material, supervenient, and inclusive causality. The generated is a result or a consequence of the generating; an apple has a color as a result or consequence of its being red, and a periscope operates as it does because it is a part of a submarine that operates as it does. These are instances of results or consequences. The use of the same terminology for all these cases is justified by the fact that they are all instances of necessary implication.
If this suggestion is right, even nomological causality is essentially a relationship based upon necessary implication. For, strictly speaking, true causes are not those commonly cited, but much larger complexes which may also include some causal laws. And these total, strict causes do necessarily imply their results or consequences.

Sebastián Lamoyi

| PDF en Spanish (432 Kb)