Deniz Saraç | Middle East Technical University


In this paper, I attempt to analyze Ed Feser’s Aristotelian Argument from his ‘Five Proofs of the Existence of God’ (2017). More precisely, I argue that his argument is unsound on the grounds of one of the premises, namely the tenth premise, including a false dilemma. For this purpose, in the first section of this paper I investigate his notion of hierarchical causation, especially regarding whether his examples reflect a unitary notion, in order to construct the most defensible characterization of hierarchical causation. In the second section, I try to argue that there is nothing in this characterization that prevents one from asserting the existence of symmetrical instances of hierarchical causation, which will mean that the premise 10 includes a false dilemma. In the third section I will consider and respond to some possible objections.

Key Words: Hierarchical causation, metaphysics, change, existence, god, relations, Feser.

SECTION I: Hierarchical Causation

In his ‘Five Proofs of the Existence of God’ (2017) Feser provides what he calls an Aristotelian argument for the existence of God. The argument comprises two stages. In the first stage Feser argues that there must be a purely actual actualizer, and in the second stage that this purely actual actualizer must be God. In this paper, I will argue that the argument is unsound due to a false premise in the first stage of the argument[1]. Although Feser first presents abovementioned argument informally, he also graciously includes a formal presentation. Here is the formal presentation of the first stage of the argument (2017, pp. 35-36).

  1. Change is a real feature of the world.
  2. But change is the actualization of a potential.
  3. So, the actualization of potential is a real feature of the world.
  4. No potential can be actualized unless something already actual actualizes it (the principle of causality).
  5. So, any change is caused by something already actual.
  6. The occurrence of any change C presupposes some thing or substance S which changes.
  7. The existence of S at any given moment itself presupposes the concurrent actualization of S’s potential for existence.
  8.  So, any substance S has at any moment some actualizer A of its existence.
  9. A’s own existence at the moment it actualizes S itself presupposes either (a) the concurrent actualization of its own potential for existence or (b) A’s being purely actual.
  10. If A’s existence at the moment it actualizes S presupposes the concurrent actualization of its own potential for existence, then there exists a regress of concurrent actualizers that is either infinite or terminates in a purely actual actualizer.
  1. But such a regress of concurrent actualizers would constitute a hierarchical causal series, and such a series cannot regress infinitely.
  2. So, either A itself is a purely actual actualizer or there is a purely actual actualizer which terminates the regress that begins with the actualization of A.
  3. So, the occurrence of C and thus the existence of S at any given moment presupposes the existence of a purely actual actualizer.
  4. So, there is a purely actual actualizer.

In the informal statement, Feser makes a distinction among two types of causal series: the linear and the hierarchical (2017, p. 21). The linear series is extended in time and the relations within a linear causal series correspond roughly to what we intuitively consider to be causation. The elements of a hierarchical series, that is, instances of hierarchical causation (as present in the premise 11 of the argument above) are such that they can take place in a single moment in time.

This notion of hierarchical causation is crucial for Feser’s argument. Earlier, historical first cause arguments were undermined by the possibility of infinite series, because they employed linear causation. But, Feser argues, this is not a possibility for hierarchical causal series, which makes hierarchical causation crucial for his argument and worthy of further examination. In characterizing his notion of a hierarchical series, Feser provides us with two main examples. His first one is a cup of coffee being held up by a desk, which is in turn held up by the floor, and so on until the series reaches the Earth (2017, p. 21). The Earth is the first cause in this case, as every intermediary is said to derive its causal power (of holding up, I presume) from the one that is hierarchically prior to it and eventually from the Earth; while the Earth doesn’t have to be so held up itself[2]. The intermediary steps, on the other hand, are characterized as instrumental, which means that they derive their causal power from something else[3].

The other main example of hierarchical series provided by Feser concerns the very existence of the cup of coffee. Feser describes this series as such: 

The potential of the coffee to exist here and now is actualized, in part, by the existence of the water, which in turn exists only because a certain potential of the atoms is being actualized, where these atoms themselves exist only because a certain potential of the subatomic particles is being actualized. This is a hierarchical series—one which, as we have seen, must have a first member. (2017, p. 26)

In their Has Oppy Done Away With The Aristotelian Proof, McNabb and Devito characterize Feser’s distinction as being a distinction “between two types of series of causes that result in change: a linear series of causes and a hierarchical series of causes.” (2020, p. 728). I disagree with this description, on the grounds that the latter type does not “result in change” at all. Feser’s argument starts with the observation that change is a real feature of the world, because his analysis of change as actualization of potential is useful to him for showing that there are real cases of potential actualization in the world. But to say that every instance of change is an instance of potential actualization is not to say that every instance of potential actualization is an instance of change. We cannot infer from the facts that “For each x, if x is an instance of change, then it’s an instance of potential actualization” and “x is an instance of potential actualization”, the conclusion that “x is an instance of change”; as it would amount to affirming the consequent. Jumping from the observation that all change is potential actualization to the conclusion that the instances of change and potential actualization must perfectly overlap is erroneous.  We can see this point by noticing that the paradigmatic examples provided by Feser for the latter kind of potential actualization, as we have seen above, are obviously not instances of change: there is nothing changing in them. I think this is pretty apparent, but if one wishes an argument for this point (that these examples don’t involve change), consider the fact that these relations (such as the cup of coffee’s being held up by the table) are taking place in a single instance in time. Any acceptable characterization of the notion of change should mention the gain or loss of an attribute. To say of something that it gained (or lost) an attribute is to say that it now has (or lacks) an attribute it lacked (or had) before. Thus, I take it that change requires duration and cannot take place in a single instance, and consequently, these examples are not instances of change. Thus, it is wrong to characterize this latter relation as involving change. 

But then we have to confront an initial problem about the argument. If this kind of potential actualization is not a type of change, what use is it to start the argument by appealing to the observation that there are instances of change in the world? After all, this second kind of potential actualization might have no instances. An analogy might clear up my point. Suppose I start by ostensively proving that there are humans in the world. Suppose that I then explain that humans are a kind of animal. If I then go on to talk about another species of the kind animal, my initial ostensive proof has no bearing on the existence of this latter species belonging to the same kind. Similarly, change might exist, it might be a type of potential actualization; nonetheless, it might be the case that no other type of potential actualization exists. However, we can try to charitably investigate whether the kind of relation assumed in the latter characterization of potential actualization, that is, hierarchical causation, also exists. Hence, we should examine the information and examples provided to see if it corresponds to a real relation in the world, and even whether there is a unitary notion of hierarchical causation in the examples[4]. If we are successful in constructing a unitary description of this relation using clearly understandable terms, which also plausibly corresponds to a feature of reality, I take it that this will not be a vital objection to the argument.

We can now turn to the question of whether there really is a single, unitary notion of hierarchical causation present in these examples. Firstly, let us examine how the two main examples differ. The main difference seems to lie in the fact that one is about the property of entities of being held up by something else (or the holding up relation between pairs of entities) and the other about the very existence of these entities. This difference may be more important than Feser assumes it to be, as it is controversial whether existence is something similar to all the other, more mundane properties; and indeed, whether it can be said to be a property in the first place. For instance, it can be doubted whether the existence of something is a property of it, since it seems that for anything to have any property at all presupposes that it, in some way or another, exists (Nelson 2020). Thus, these two examples may not be so analogous as to pertain to a unitary kind of relation, viz. hierarchical causation. 

I am personally of the opinion that it is certainly possible existence turns out to be best understood as a property. However, even if it turns out that existence is best understood as a property (and so, is analogous to the property of being held up in relevant respects), I believe this should at least be argued for, since it is pretty much a minority view to assert that existence is a property (Nelson 2020, para. 6)[5].

Nonetheless, I think this need not be a crucial problem for Feser. The proponent of his argument can claim that the former example is provided simply for illustrative purposes and just abandon it. After all, it seems that what really matters for the argument, what is necessary for the conclusion it’s trying to arrive at, is the sort of hierarchical causation as present in the second example.

So, let us inspect whether we can find a single unitary notion of hierarchical causation within the second example. It will be helpful to keep in mind that hierarchical causation is the relation invoked in answering the question, in Feser’s words, regarding “what makes it true that the coffee exists here and now, and at any particular moment that it exists? What keeps it in existence?” (2017, p. 26). The first step in the second example is the relation between the coffee and the water that makes up the bulk of it.  This is a somewhat odd example, due to its being what one might call a part-whole relation. Are mereological relations (at least sometimes) instances of hierarchical causation? A metaphysical picture in which wholes depend on their parts in order to exist would seem to accommodate for such a reading of hierarchical causation. But the most charitable option would be to stay neutral on this question and move on. The next step in this example is the relation between the body of water and the atoms it is made out of. According to Feser, the potential of distinct quantities of hydrogen and oxygen of existing as water is actualized, and this is the cause of the sustained existence of water. Similarly, these atoms’ potential to exist is actualized by yet another cause, that is the potential of certain subatomic particles to exist in the way they do, and so on. Let’s focus our characterization only on these last two steps in order to be able to give a unitary account of the relation in question; since, as we have seen, the other examples and the previous steps in this example may hinder our efforts to provide a unitary characterization as they possibly involve quite different relations. Then, perhaps we can affirm the reality of hierarchical causation.

Firstly, hierarchical causation certainly seems to be transitive. The subatomic particles can be understood to be the cause of the body of water, even if they’re not the very next member in the series. Actually, the whole point of the argument relies on the fact that hierarchical causation is transitive; if this was not so, the first cause would only be the cause of the very next entity in the causal series, whereas Feser wants it to function to be the ultimate cause of all existence, including the cup of coffee. Secondly, regarding reflexivity, according to Feser, the very characterization of the members of the series as instrumental amounts to saying that they derive their causal power from some other entity (2017, pp. 65-66). It seems prima facie reasonable to think that this rules out their being self-caused, making the relation irreflexive. However, I will try to challenge this point in the third section, so let’s keep it out of our characterization for now. Thirdly, regarding symmetricity, we don’t have much information to infer from the two steps we’re focusing on and Feser’s own discussion on the topic. The two steps can be argued to be asymmetric, as there seems to be some sort of hierarchy between the entities in question: the subatomic particles are hierarchically prior to the atoms, and the atoms hierarchically prior to the body of water. But this can be challenged, why not suppose the bigger parts cause the smaller ones? More importantly, even if we accept that these two instances of the relation are asymmetric, this doesn’t prove that the relation is asymmetric as well. It only shows that the relation in question is non-symmetric (that is, it can have instances that are not symmetric). Would we be justified in going one step further and claiming it is asymmetric as well? I think not, and will argue for it in the next section. Finally, the function of this relation is to be involved in answering the question regarding the sustained existence of entities (as opposed to, say, their generation).

So let us try to provide a characterization of hierarchical dependence that doesn’t employ any mysterious metaphysical terms. Thus, hierarchical causation can be narrowly characterized as a transitive relation in which members depend on other members for their sustained existence, and don’t derive their causal power to keep other members in existence directly from themselves but from some other thing. As we have seen, the existence of change is not sufficient reason to believe that this relation exists as well. So, how should we answer the question regarding whether it is real[6]? I will now attempt to show that even if we accept that it exists, it cannot be used to justify Feser’s conclusion. Whereas earlier first cause arguments that employed linear causation were undermined by the possibility of infinite causal series; this one is undermined by the possibility of causal loops. This will be the subject of the next section.

SECTION II: Possibility of Symmetricity

Now I can try to show what I claim to be the false dilemma is in the tenth premise of Feser’s argument, which states: “If A’s existence at the moment it actualizes S presupposes the concurrent actualization of its own potential for existence, then there exists a regress of concurrent actualizers that is either infinite or terminates in a purely actual actualizer” (2017, p. 35).

The false dilemma is this: there is a third option regarding the series’ regress. It might terminate in a purely actual actualizer, or regress infinitely, or it might also form a causal loop. This third option, which I will articulate more thoroughly below, is important because Feser uses the premise 10 in conjunction with the premise 11, which states that the infinite regress is impossible, to arrive at the conclusion that there must be a purely actual actualizer in which the series terminates (by modus tollendo ponens).

This unconsidered third option claims that an entity can be present in its own hierarchical “family tree”, so to speak. This might seem unintuitive at first, mainly because our ordinary conception of causation (what Feser calls ‘linear causation’) does not seem to permit such a situation. However, I will argue there is nothing in Feser’s hierarchical series that excludes this possibility. If I succeed, this will mean that the premise 10 is false on the grounds of presenting a false dilemma.

            I don’t mean to argue that the world is really this way. To show that the argument is unsound and that the tenth premise presents a false dilemma, I only need to prove that this last option is a metaphysical possibility. From our final characterization of the hierarchical causation relation, this seems to be the case. I will try to provide some examples to motivate this claim. But first, I will try to clarify my position by analogy.

What I mean by the third possibility can be best understood as the analogue of the coherentist theory of epistemic justification from epistemology. Regarding the structure of epistemic justification, three proposed theories are infinitism, foundationalism, and coherentism. The infinitist claims that the justificatory structure of our beliefs forms an infinite regress, each belief being justified by some other belief, which is in turn justified by yet another belief, and so on for an infinite number of beliefs. The foundationalist claims all of our beliefs ultimately rest on a set of noninferential beliefs that are nonetheless able to justify the superstructure beliefs. So, the regress present in infinitism is avoided due to the existence of a foundational layer. This position is analogous to Feser’s own position regarding his hierarchical series. Finally, the coherentist claims that a belief is justified if it coheres with a set of beliefs. In other words, the members of a set of beliefs are justified by the other beliefs in their epistemic vicinity, forming what can be called a web-like structure. A naive form of coherentism where the chain of justification creates loops (i.e, A justifies B, which justifies C, which justifies A) can be analogous to the option I am going to present.

There has also been recent work in spelling out similar possibilities in metaphysical series (Barnes 2018; Thompson 2018). Here I will argue that such a ‘coherentist’ series may be possible in Feser’s hierarchical causation. I will present two examples analogous to the two main examples given by Feser. In these, I will consider what is perhaps the simplest way such a causal loop can take place, namely symmetrical causation, although this is not to exclude the possibility of much more complicated loops. I hope that if I am successful in persuading the reader that these instances of symmetrical causation are metaphysically possible instances of Feser’s hierarchical causation, she will also be convinced that they represent simple instances of the broader thesis that a coherentist picture of the mentioned series is also possible. Since, if we grant that a case in which A causes B and B causes A is possible, there is no reason we shouldn’t extend this to situations where A causes B, B causes C and C causes A, etc. I also find it possible that these relations form even more complicated web-like structures.

The first example is this. Consider two celestial bodies and their orbits, say, the Earth and the Moon. Since each has a certain mass, each of them is the cause of the other’s staying in that very orbit. If one has the lingering feeling that the Earth is somehow hierarchically superior in this case simply because it is more massive, one can also consider the case of binary stars in which the two objects attracting each other can be of similar masses. This example is analogous to the hierarchical causation present in Feser’s first example, in which the cup of coffee is supported by the table and the table by the floor and so on until eventually all turn out to be supported by the Earth. But unlike Feser’s example, it doesn’t terminate in a single entity that causes the other entities to have a certain property without needing to be caused to have that property itself. Rather, the entities are the cause of each other’s having a certain property.

The second example is this. For Feser, the death of a previously living being is an example of substantial change (2017, p. 17). That is to say, since living beings are genuine entities –substances—in virtue of their being alive, their deaths amount to a change in the ontological landscape of the world: some individual substance goes out of existence when something dies.Hence, keeping something alive is causing a substance’s sustained existence. Such concurrent causation of sustained existence is supposed to be what we mean by hierarchical causation. Now consider a surgical operation in which two otherwise healthy people are attached to one another such that person A loses her heart and depends on person B for her circulation, and person B loses his kidneys and depends on person A for his urination. Both urination and blood circulation are vital processes for human beings, one cannot survive without them. In our case, each person keeps living because the other person does. They keep each other alive and hence are the concurrent actualizers for each other’s potential to keep existing. I think this is an instance of symmetrical hierarchical causation and shows that the relation in question is non-symmetric rather than asymmetric.

Of course, the operation is not ethically permissible and perhaps the medical technicalities (like the amount of blood being too much for a single heart to pump) may be important. However, at most, the medical technicalities will provide a physical impossibility. That is, the impossibility of the operation being successful in a world governed by our physical laws. But it is important to remember that we are looking for metaphysical possibility as the matter is a metaphysical one. Thus, this isn’t sufficient reason to reject these examples and the following conclusion that symmetrical hierarchical causation is possible[7].

The examples I have provided for symmetrical hierarchical causation are in line with our characterization of hierarchical causation from the last section. The members of the relation depend on other members for their sustained existence, and don’t derive their causal power to keep other members in existence directly from themselves but from one of the other members in the series. Thus, these examples illustrate a third alternative to the two provided by Feser in the premise 10, which in turn makes said premise erroneous as it implies the two alternatives to be exhaustive.

SECTION III: Two Objections Answered

The causa sui objection:

One can perhaps object to my argument by stating that since hierarchical causation is a transitive relation, by admitting that there can be symmetrical cases of it, we must accept that these cases are reflexive as well. In other words, if A hierarchically causes B and B hierarchically causes A, by the transitive property of hierarchical causation we can infer that A hierarchically causes A. But this seems like a causa sui fallacy, nothing can be said to be the cause of itself. However, I think this objection fails. The notion of causation at hand is very different from the notion of causation in which the causa sui fallacy is a legitimate concern. The reason why nothing can be said to be the cause of itself because it would imply that the thing would exist temporally prior to itself. This is what makes self-causation impossible. But in hierarchical causation, the relation can take place in a single moment, it need not be linearly extended in time. Thus, the reflexivity (and self-causation) is not problematic.

The objection from definition; or, alternatively, the “proof is in the name” objection:

One can also object that the notion of causal series in question must have a first member due to its definition. After all, even its name is “hierarchical causation”. It must have a first member if it is to satisfy this hierarchical quality expected of it. However, I think this objection fails as well. It would be question-begging to assert that hierarchical series must have a first member because they are defined to be such. Feser himself considers the notion of question begging as a possible objection to his argument. He says: 

It is also sometimes objected that the argument for a first member of a hierarchical series begs the question, insofar as characterizing other causes as instrumental itself presupposes that there is such a first member. But there is no begging of the question. To characterize something as an instrumental cause is merely to say that it derives its causal power from something else. There is nothing in that characterization that presupposes that a series of such causes cannot regress to infinity or that there must be some cause which has underived causal power. Even the skeptic can perfectly well understand the idea that a stick cannot move the stone under its own power, whether or not he goes on to agree that a regress of such moved movers must terminate in a first member (2017, p. 65).

I agree with him on the point that our characterization of the hierarchical series and the members it comprises do not presuppose the existence of a first member on the condition that by calling the members ‘instrumental’ we only mean that they derive their causal power from something else. But, as we have shown, this quality can be satisfied by what we have called coherentist causal series. This means that we have to actually investigate whether alleged instances of hierarchical series in the real world are such that they terminate in a first member. Thus, to argue that coherentist series aren’t permissible on the grounds that hierarchical series needing to have a first member by definition would be fallaciously extending the definition so as to include a quality that this type of causation doesn’t necessarily have. Hopefully, we can leave the business of defining entities into existence to the realm of ontological arguments.

SECTION IV: Conclusion

In this paper, we started by examining the first stage of Feser’s Aristotelian argument for the existence of God and noticed that the jump it makes from the reality of change to the reality of a special kind of potential actualization (hierarchical causation) is unwarranted. We tried to charitably reconstruct the account of hierarchical causation provided by Feser and assumed that it too corresponds to a feature of reality. We noticed that in this account there is nothing that prevents one from having instances of causal loops, and tried to motivate this point by providing some examples of symmetrical causation. If these examples and the argumentation so far has been successful, this will mean that the premise 10 is false because it fails to acknowledge the possibility of coherentist causal series. This, in turn, would mean that the argument is to be rejected (or at least modified) on the grounds of its being unsound.


Barnes, E. (2018). “Symmetric Dependence,” in Reality and Its Structure: Essays in Fundamentality, R. Bliss & G. Priest (Eds.), Oxford: Oxford University Press, pp. 50–69.

Berto, F., Plebani, M. (2015). Ontology and Metaontology A Contemporary Guide, Bloomsbury Academic.

Feser, E. (2017). Five Proofs of the Existence of God, San Francisco: Ignatius Press.

McNabb, T., and Devito, M. (2020). “Has Oppy Done Away With The Aristotelian Proof?,” The Heythrop Journal, 61: 723–731.

Nelson, M., “Existence”, The Stanford Encyclopedia of Philosophy (Summer 2020 Edition), Edward N. Zalta (ed.), URL =<>.

Oppy, G. (2019). “On stage one of Feser’s ‘Aristotelian proof’,” Religious Studies, 1–12.

Thompson, N. (2018). “Metaphysical Interdependence, Epistemic Coherentism, and Holistic Explanation,” in Reality and Its Structure: Essays in Fundamentality, R. Bliss & G. Priest (Eds.),Oxford: Oxford University Press, pp. 107–125.

[1] In my opinion, other objections to either stage of the argument are available, and one can see Oppy’s On stage one of Feser’s ‘Aristotelian proof’ (2019) for a notable example. However, I will limit my discussion to a specific objection to a specific premise.

[2] One might wish to say that this picture provided by Feser regarding the relation between the Earth and objects standing on it is lacking since it doesn’t employ the notion of gravitation; however, I think this would be missing the point. The purpose of this example, I believe, is to show that a relation as mundane as holding up is characterized by hierarchical causation.

[3] Another example provided by Feser concerns the holding up of a lamp by the chain, the chain by the fixture on the ceiling, and so on (2017, p. 21). This example is very similar to the one previously mentioned and gives no new information regarding the nature of hierarchical series, so I will not talk about it.

[4]  There is a discussion regarding the objections to the reality of the hierarchical series in Feser’s book (2017, pp. 60-66), but most of my discussion will be unrelated to the arguments and counterarguments mentioned there.

[5] More precisely, it is against the commonly accepted analysis of existence that is endorsed by Russell and Frege, and is reflected in the Quinean orthodoxy of today’s ontological landscape. For further discussion, see Berto & Plebani 2015; Nelson 2020.

[6] A great discussion that argues that such a causal relation is unnecessary, that things don’t need an actualizer for their potential to exist after they begin existing can be found in the aforementioned Oppy paper (2019) and a response to Oppy in the aforementioned paper by McNabb and Devito (2020).

[7] One might also wish to respond by saying that even if the examples show that there can be symmetrical cases of hierarchical causation, such as in the surgical operation example, this does not exclude the possibility of hierarchical series that are of the form described by Feser. For example, even if we accept that persons A and B cause each other’s sustained existence, each of them also owes their sustained existence to the atoms that they are constituted out of, and the atoms owe their sustained existence to the subatomic particles, and so on. But, as I have stated above, the examples aren’t intended to show that each hierarchical series has to be terminated in this circular fashion. I only propose the examples to hopefully persuade the reader that they describe possible cases that are in line with our best characterization of hierarchical causation and have a circular character. It is sufficient to show that this is a metaphysical possibility in order to undermine the argument which presents only two possibilities and proceeds by refuting one of them. I find it interesting to note that although this response is unable to illustrate a fault in my example as it is irrelevant to the intentions behind the example, it may lead to questions regarding the possibility of causal overdetermination in hierarchical series (both Person B as an individual and Person A’s own atoms hierarchically cause Person A’s sustained existence). Inspecting the consequences of permitting overdetermination in hierarchical causation is, however, beyond the scope of this paper.