1. Introduction
Many philosophers have argued that diseases are biological dysfunctions (Boorse Reference Boorse1977; Wakefield Reference Wakefield1992). Dysfunction-talk also features prominently in the relevant sciences and clinical fields. However, a simple dysfunction-based theory struggles to make sense of some seemingly non-pathological instances of dysfunction. For example, does a single dead cell in an otherwise healthy body suffice for disease?
Some theorists have used trivial dysfunctions of this sort to argue that ‘disease’ is an anthropocentric or inherently evaluative notion that cannot be analysed in purely biological terms (Cooper Reference Cooper2020; Wakefield Reference Wakefield2014). This paper advances instead a previously unexplored thesis: diseases are biological dysfunctions, but of a special kind. On the theory I advance diseases are “domino dysfunctions;” dysfunctions in traits which, when they fail, cause other traits to fail too – like chains of toppling dominoes.
I begin by clarifying my method and approach to the question. I then define biological dysfunction, before introducing a set of problem cases from the literature which complicate a theory of disease as, simply, dysfunction. I then provide an account of environmental mismatches, before applying this distinction to the body’s internal environment, coining the term “somatic mismatch.” On this basis, I define “domino dysfunctions” as biological dysfunctions that cause dysfunctions and/or mismatches in other traits, and distinguish them from “trivial dysfunctions;” that is, dysfunctions which fail to have adverse biological “knock-on” effects of this kind. I then show how the Domino Theory can make sense of the problem cases which initially motivated us to refine the dysfunction theory: they aren’t diseases, because they aren’t domino dysfunctions.
2. Approach
In this section, I clarify my methodology and highlight a few assumptions which underpin my view. The disease debate in philosophy is traditionally understood as a debate over the correct conceptual analysis of disease; in other words, over the criteria which people have in mind when they use terms like “disease.” The theory I advance here, however, is not intended as a conceptual analysis, but instead as a descriptive theory of the structure underpinning a natural kind (more akin to what Millikan calls a “theoretical definition”; Millikan Reference Millikan1989, Reference Millikan2017). Consistent with this, I assume that disease is a natural kind.Footnote 1
It will also be helpful to say a few words about how I will be using the term “disease” in this paper. In the philosophical debate, contributors have often ended up employing distinct terminology (disease, disorder, malady, pathology, etc.) to talk about what is ostensibly the same phenomenon. I have previously argued that philosophers’ divisions over terminology, despite their insistence that they’re all “talking about the same thing,” itself motivates thinking of disease as a natural kind (Fagerberg Reference Fagerberg2023). In what follows, I assume that all diseased, disordered, and legitimately medical conditions form a kind, and I will use the terms “disease,” “disease state,” or “pathological state” to refer to this kind.
Before moving on, I also need to make explicit two theoretical assumptions which I am unable to defend within the scope of this article. First, as noted, there seems good prima facie reason to think that pathology has something to do with biological dysfunction, and this is also my starting point (see, e.g., Wakefield Reference Wakefield1992; cf. Cooper Reference Cooper2002). My second assumption is that the selected effects theory and its related conceptual apparatus provides the best foundation for thinking clearly about biological function and dysfunction (see Neander Reference Neander1991; cf. Boorse Reference Boorse1976). I lean particularly on the distinction between proximal functions and distal effects, and the Neanderian notion of a “response function” (Neander Reference Neander1995; Reference Neander2017).
While I realize that these assumptions are to various extents controversial, I hope the sceptical reader will nonetheless be willing to grant them for the sake of argument, even if merely to see where they lead. The fact that these assumptions do support a coherent and plausible theory of disease as a natural kind arguably lends some further credence to them.
3. A dysfunction account
In order to understand what dysfunctions are, we must first get clear on what functions are. According to the selected effects theory, functions are those effects which caused their underlying trait to be selected.
A function (F) of a trait (T) in an organism (O) is an effect that items in T’s lineage yielded which increased the inclusive fitness of O’s ancestors, in recent evolutionary history, such that T was naturally selected (Neander Reference Neander1991; Neander and Rosenberg Reference Neander and Rosenberg2012; Godfrey-Smith Reference Godfrey-Smith1994).
So, for example, it is a function of my heart to pump, because that is what items of the lineage “human heart” did in the past which conferred upon the organisms which possessed these hearts such fitness advantage that the genotype which codes for the phenotype “human heart” was naturally selected.Footnote 2
Having settled on a theory of biological function, we are now closer to being able to define biological dysfunction. At first glance, it seems obvious: if a function is a selected effect of some trait T, then surely a dysfunction is simply the failure of a selected effect of T? Unfortunately, it isn’t quite so simple. Technically, hearts did many things which caused them to be selected. For example, hearts pumped, and hearts helped to circulate blood around the body. Which of these is the heart’s function? The selected effects theory appears to support more than one conclusion. As Neander writes: “The answer seems to be both, for both were done, both were adaptive, and both caused the underlying genotype to be selected.” (Reference Neander1995, 114)
Now, consider that blood circulation may fail through no fault of the heart – as in the case of an arterial clot. In this case, the heart is failing to do one of the things it did in the past which caused its selection (for blood is not being circulated), and yet the heart itself is not dysfunctional. How do we know which selected effect needs to fail for there to be a biological dysfunction of the heart in particular?
An answer to this question was originally proposed by Neander, and has recently been expanded and championed by Garson (Neander Reference Neander1995; Garson Reference Garson2019; Fagerberg and Garson forthcoming). The short answer is that a trait only dysfunctions when it cannot yield that effect which is most proximal or specific to this particular trait.
The most proximal effect of trait T is that effect which occupies the lowest level of a functional analysis wherein T is still visible as an unanalyzed part (Neander Reference Neander1995; see also Cummins Reference Cummins1975; Griffiths Reference Griffiths1993). Put differently, the most proximal function is the effect in virtue of which all the other, more distal, effects of T are carried out (Garson Reference Garson2019). The heart pumps, the heart contributes to blood circulation, the heart helps supply oxygen to the tissues of the body, including the brain, and thus enables normal cognitive function. However, one of these effects is more specific to the heart than the others. The heart cannot circulate blood without the contribution of the rest of the cardiovascular system, nor can it ensure normal cognitive function without the lungs and, of course, the brain itself. Moreover, the heart’s contributions to these more distal effects are all consequences of its ability to pump. We could perhaps appeal to selected effects at even lower levels – such as the individual contributions of the ventricles – but at that point in the analysis the heart itself would no longer appear as an unanalyzed component. On this basis, we can say that the proximal function of the heart is to pump (see figure 1).
Figure 1. Functions are hierarchically organised.
We can now see that these other, more distal, selected effects of the heart are, in fact, proximal functions of larger systems which include the heart among their constituent parts. As we move up the hierarchy of our functional analysis, the selected effects in question (blood circulation, providing oxygen to tissues) involve larger and larger biological systems, of which the heart makes up a smaller and smaller part. As Neander put it, the most proximal function is that thing which the trait can do “more or less on its own” (Neander Reference Neander1995).
As Neander and Garson argue, T is only dysfunctional in a scenario in which T cannot perform its most proximal function. If T can perform its most proximal function, but is unable to yield one of the more distal selected effects associated with the trait, then T is strictly speaking performing its proper proximal function just fine and therefore cannot count as being dysfunctional. However, T may instead be “mismatched.” (I will return to and refine my approach to environmental mismatches in section 5.)
We are now able formulate a simple theory of pathology as biological dysfunction:
A function (F) of a trait (T) in an organism (O) is an effect that items in T’s lineage yielded which increased the inclusive fitness of O’s ancestors, in recent evolutionary history, such that T was naturally selected.
A dysfunction is the failure or inability of some trait T to perform its most proximal function.
A disease is a dysfunction.
My view is that this simple theory puts us on the right track. All diseases are dysfunctions, and most dysfunctions are also diseases. However, due to the intricate dependence relations which obtain between evolved traits within the body, paradigm diseases are not just dysfunctions; they are dysfunctions which cause problems for other traits. In other words, diseases are domino dysfunctions.
Before outlining the Domino Theory in full in section 5, I first introduce a set of “problem cases” which illustrate the deficiencies of a simple dysfunction theory, and which motivate distinguishing domino dysfunctions from dysfunctions simpliciter.
4. The problem cases
I have said that, unlike most previous definitions of disease, the Domino Theory is not intended as a conceptual analysis. However, this does not make the traditional disease debate entirely irrelevant to our discussion here. Our thoughts, inferences and, indeed, philosophical debates may be shaped by the existence of real kinds and properties in our world, even if we lack explicit awareness of this structure (Fagerberg Reference Fagerberg2023). As such, classic problem cases in the traditional disease debate may indicate something important about the nature of the kind which we seek to understand. In what follows, I shall consider a cluster of cases which I believe provide a clue as to the distinction between diseases and non-pathological conditions.
As noted, there seems to be a strong relation between dysfunction and disease. Many instances of biological dysfunction are also paradigm diseases, for example liver failure, stroke, hypothyroidism. However, some plausible cases of dysfunction appear to be non-pathological or compatible with health, and as such challenge the simple dysfunction theory outlined in the previous section. Consider the following:
A single dysfunctional kidney: A kidney’s function is to filter blood and remove waste from the body, because that is the effect which led to its natural selection. As such, should some particular kidney fail to perform this function, that would constitute a dysfunction of that kidney (Wakefield Reference Wakefield1992). And yet one can live a healthy life with only one kidney.
One dead cell: A dead or dysfunctional cell is not functioning as it should.Footnote 3 However, we are all currently operating with dead cells within our bodies (Boorse Reference Boorse1977; cf. Nordenfelt Reference Nordenfelt1995). And yet it does not seem that we are all, thus, diseased.
Prima facie fitness-reducing sexual orientations: Another possible category of dysfunction in the absence of disease is sexual orientations that would appear, on the face of things, to be fitness-reducing, such as exclusive same-sex attraction or asexuality (Boorse Reference Boorse2014). Assuming that these are (or could be) dysfunctions, should they count as pathological?Footnote 4
Contraception: We often choose to disable reproductive function for pragmatic reasons. We do not wish to conceive, and therefore choose to use contraceptives or undergo sterilization. In such cases, the reproductive system cannot perform its function (see, e.g., Kingma Reference Kingma2014). However, these would not appear to be cases of us inflicting disease upon ourselves.
Wakefield and Cooper use cases such as these to motivate the necessity of attributions of harm for judgments about disease (Cooper Reference Cooper2020; Wakefield Reference Wakefield2014). Boorse prefers to bite the bullet and count such cases as pathological – albeit divorced from any negative evaluation (Boorse Reference Boorse2014). As I shall argue, these cases strike us as non-intuitive cases of disease because they are of a different biological kind to paradigm disease-states; they are not domino dysfunctions, but instead (at most) “trivial dysfunctions.”
5. The Domino Theory
Domino dysfunctions are biological dysfunctions in traits which are situated within the rest of the body in such a way that, when they fail, they disrupt the normal functioning of other traits which depend on them. More precisely, the view I defend is that diseases are biological dysfunctions which either cause other traits of the organism’s body to become somatically mismatched, or which cause them to become dysfunctional, or both.
As now becomes clear, the viability of the Domino Theory will depend on a well-articulated account of dysfunctions, mismatches, and the relationships between these concepts. In section 3, I defined dysfunction. In section 5.1, I give an account of environmental mismatches, distinguishing two types of mismatch via the types of evolved dependence relations which give rise to them: “fuel failures” and “signal failures.” In section 5.2, I then generalize this account to the dependence relations which obtain between evolved traits within the body, coining the term “somatic mismatch.” This finally allows me to outline the Domino Theory in full, and to distinguish domino dysfunctions from trivial dysfunctions.
5.1. Environmental mismatches
Although environmental mismatches are not necessarily diseases in and of themselves, as we shall see, giving an account of environmental mismatches will allow me to define its analogue within the body – the somatic mismatch – which in turn will help me define domino dysfunction and, thus, disease.
While accounts of environmental or evolutionary mismatches which bear some relation to the view I shall propose already exist in the philosophical literature (Cofnas Reference Cofnas2016; Morris Reference Morris2020; Matthewson and Griffiths Reference Matthewson and Griffiths2017; Bourrat and Griffiths Reference Bourrat and Griffiths2021), none is explicitly premised upon the distinction between proximal functions and distal effects. As such, they are unable to distinguish types of mismatches in terms of the trait’s ability to perform its proximal function, as I do below. Obscuring the distinction between functions (which are proximal) and more distal selected effects renders functions technically indeterminate and so inevitably yields some imprecision (Fagerberg and Garson forthcoming; cf. Matthewson and Griffiths Reference Matthewson and Griffiths2017). Because mismatches play a key role in defining domino dysfunctions, and thus disease, it will be important in what follows that they are precisely defined. Existing views were not quite up to this task.
In general, environmental mismatches occur where the functioning of a biological trait T is impacted or interrupted by some factor in T’s actual environment differing from T’s selective environment – that is, the environment in which traits of T’s type evolved. A little reflection on this phenomenon reveals to us that environmental mismatches are possible because biological traits are adapted to and depend on features of the organism’s external environment. Because traits depend on their environment, abnormalities in a trait’s actual environment (relative to its selective environment) can prevent a trait from functioning ‘as it should’. And yet the trait is not necessarily dysfunctional. In some cases, the trait is functioning precisely as it was designed to function by natural selection – and yet, it is operating maladaptively due to a mismatched environment. In what follows, I distinguish two types of environmental mismatch on this basis: (1) “fuel failures” and (2) “signal failures.”
Recall that we defined dysfunction as the failure of T to perform its proximal function. The first type of environmental mismatch occurs when T depends on some feature or aspect of the environment for the performance of its proximal function – i.e., to avoid failing to perform its function or becoming dysfunctional. Let’s call environmental mismatches of this type “fuel failures”:
(1) Fuel failure: Trait T depends on some feature C of its selective environment for the performance of its proximal function. Because T’s actual environment differs from its selective environment in respect of C, T cannot perform its proximal function.
By “fuel” I just mean some environmental condition without which T cannot perform its proximal function. Let us consider an example. It is a proximal function of my lungs, along with some muscle groups, to enable breathing. However, my lungs cannot perform this function in an environment without breathable air. As such, they are suffering fuel failure or, put differently, a type 1 environmental mismatch.
The second type of mismatch is due to some abnormality in a “signal” in T’s actual environment which it is T’s proximal function to respond to. Signal failures are distinguished from fuel failures in that they are compatible with trait T performing its function precisely as designed, and yet functioning strangely, even maladaptively. This phenomenon reflects not the dependence of traits on their environments for the performance of their proximal functions, but rather their dependence on environmental cues for their proximal functions, whatever they are, to actually contribute to more distal selected effects.
(2) Signal failure: Trait T has as its proximal function F to yield some outcome or effect in response to or in proportion to some input S in its selective environment. Because T’s actual environment differs from T’s selective environment in respect of S, T is caused to perform its proximal function F in such a way that F fails to contribute to more distal selected effects of T.
Signal failures are only possible where some environmental dependence is built into trait T’s most proximal function – in Neanderian terms, where T’s most proximal function is a “response function” (Neander Reference Neander2017). Response functions are functions where trait T has as its proximal function to yield some outcome in response to or in proportion to some input from the environment.
Let us consider a plausible example. Imagine a young gosling that happens to imprint upon a fox due to hatching into an abnormal environment (see also Wakefield Reference Wakefield2017; Garson Reference Garson2021). The gosling’s imprinting mechanism is performing its proximal function just fine – that is, the mechanism is causing the gosling to become attached to “the first large, suitably moving object” (Garson 2021, 342) it meets upon hatching. However, the imprinting mechanism is not causing the gosling to form an attachment specifically to a its mother – a more distal selected effect which the imprinting mechanism normally contributes to in its selective environment. Indeed, being emotionally attached to an unsympathetic fox likely yields none of the mechanism’s usual beneficial effects.
In this case, the imprinting mechanism is in a state of “signal failure” or type 2 environmental mismatch. Due to the gosling’s actual environment differing from its selective environment (the environment in which the mother is present) in respect of some input on which the imprinting mechanism depends, imprinting fails to contribute to the more distal, fitness-boosting effects which it had in its selective environment.
Type 1 and 2 environmental mismatches – or fuel failure and signal failure – are similar in that they reflect how evolved traits depend on features of their selective environments in order to function “as they should.” The difference between them is that they reflect different types of evolved dependence relations. A trait can depend on its environment for “fuel” – some precondition for the performance of its proximal function – or for “signaling” – some cue or input which T has as its proximal function to respond to. In the latter case, the responsiveness to environmental cue S is built into what it is to perform T’s proximal function F correctly, and thus, T behaving in a strange and seemingly fitness-compromising manner is compatible with T performing its proximal function precisely as designed. In fuel failure, on the other hand, T is prohibited from performing its proximal function by an inappropriate environment.
Let us now consider some cases that would not count as environmental mismatches on the view I’ve detailed in the above. Firstly, if, while minding my own business in an environment approximating the one in which my species evolved, I suddenly suffer a blood clot in my leg that prohibits the passage of blood through my vein, that would be a straightforward endogenous dysfunction – not an environmental mismatch of any kind. Dysfunctions aren’t necessarily mismatches, just as mismatches aren’t necessarily dysfunctions. However, as we shall see, mismatches are often causes of dysfunction.
Secondly, trait T does not count as mismatched when a distal selected effect of T fails to be performed, but for reasons that are unrelated to T. Consider our gosling again. Suppose everything went as planned and the gosling imprinted on its mother. In this case, the imprinting mechanism is contributing as normal to the more distal effect of creating a fitness-boosting relationship with the mother goose. Let’s also stipulate that the gosling is not in its selective environment (say it lives on a farm run by humans), and so his mother is removed from him at an early stage of his development. In this case, again, the unfortunate gosling does not enjoy a fitness-boosting attachment to its mother. However, this outcome has nothing to do with the imprinting mechanism. The imprinting mechanism contributed as usual to the more distal selected effects to which the imprinting mechanism normally contributes – however, something else went wrong which interrupted the casual chain.
I have said that environmental mismatches are not necessarily dysfunctions. However, being in a mismatched environment can be a cause of dysfunction, even of independent dysfunctions which endure even after the organism is relocated to an appropriate environment - in other words, dysfunctions which do not depend on the mismatched environment for the failure of function to occur. For example, a fish out of water is in a mismatched environment. Being out of water will quickly cause the functioning of the fish’s brain to break down due to lack of oxygen (fuel failure). Now the fish is both currently mismatched relative to its actual environment, and the fact of it being mismatched is currently causing a failure of function. Over time, the tissues of the fish’s brain will become damaged by prolonged oxygen deprivation. At this stage, if we reintroduce the fish into water, it will no longer be in a mismatched environment, but the dysfunction – via damage to its neural tissues – will remain (figure 2). In this sense, the mismatched environment has caused an independent dysfunction.
Figure 2. Mismatches can cause dysfunction.
We can see, then, that a mismatched environment can occur without dysfunction, can be a current cause of current dysfunction, and can be a past cause of current dysfunction.
5.2. Somatic mismatches
With this understanding of mismatches in hand, we can now turn to the dependence relations which obtain between evolved traits within the body. Traits do not rely solely on preconditions and cues in their external “ecological” environment. Traits are also adapted to and dependent upon other traits within the organism. For example, the brain relies on the heart to pump properly, so that the cardiovascular system can supply oxygen to the brain which, in turn, enables cognition. Once we recognize these intricate dependence relations between evolved traits within organisms, we can see that there is an internal “somatic” analogue to the classic environmental mismatch. When the somatic environment within the body is disrupted by some trait becoming dysfunctional, dependent traits are often affected too.
These dependent traits are, on my view, “mismatched” to their somatic environment. Just as there are two ways in which a trait T can be adapted to its external environment, there are two ways in which a trait T1 can be adapted to another trait T2 in its somatic environment, and thus two types of somatic mismatch.
(1) Fuel failure: Trait T1 depends on some other trait T2 for the performance of its proximal function. Because T2 is dysfunctional, T1 cannot perform its proximal function.
Suppose there is a dysfunction in the heart which is causing a severely reduced heart rate (bradycardia). This reduction in heart rate in turn causes insufficient oxygen supply to the brain which, in turn, causes the brain to fail to perform some of its proximal cognitive functions – perhaps the individual simply loses consciousness and passes out. Now the brain is unable to yield one or more of its proximal functions, because it lacks one of the environmental predispositions (oxygen) for the performance of those functions. As such, the brain is experiencing somatic fuel failure or a type 1 somatic mismatch.
We turn now to signal failure, or type 2 somatic mismatch:
(2) Signal failure: Trait T1 has as its proximal function F to yield some outcome or effect in response to or in proportion to some input from T2. Because T2 is dysfunctional, T1 is caused to perform its proximal function F in such a way that F fails to contribute to more distal selected effects of T1.
As an example of somatic signal failure, consider the dependence relation which obtains between the heart and the vagus nerve. One of the functions of the vagus nerve is to regulate the heart rate, and one of the functions of the heart is to regulate its rate in response to input from the vagus nerve (a response function). As such, if there is a dysfunction of the vagus nerve – if, for example, it is overactive – this can lead to an abnormally low heart rate which, in turn, may cause other problems.
If part of the function of the heart is to regulate its rate in response to input from the vagus nerve, then the heart is not dysfunctional when it lowers its rate accordingly. Nevertheless, if given the wrong input from the vagus nerve, the heart may perform its proximal function in such a way that it fails to contribute to more distal effects of the heart (such as supplying oxygen to cells). In this case, the heart is a victim of somatic signal failure or, in other words, somatic mismatch type 2.
5.3. Secondary dysfunctions
As we have seen, mismatches can also cause new dysfunctions. For example, insufficient oxygen supply to the brain as a result of heart failure can make it impossible for the brain to perform its proximal functions, and over time may cause such damage that the dysfunction remains even when the trait is returned to its selective environment (in this case, when the brain is resupplied with oxygen). In such cases, the mismatch (type 1) is causing a secondary dysfunction.
Similarly, type 2 mismatches can over time put such stress on mismatched traits that new dysfunctions arise. For example, elevated levels of the hormone thyroxine, as a consequence of hyperthyroidism, can stimulate the heart to pump rapidly and irregularly which, over time, may lead to heart failure. Thus, traits being somatically mismatched (type 2) can cause them to become dysfunctional over time.
Due to the hierarchical organization of the body, whereby more specific functions (e.g., the heart’s pumping) contribute causally to increasingly general selected effects (e.g., blood circulation) of larger and more complex systems (e.g., the cardiovascular system), there is also another way in which a dysfunction in some trait T2 can cause a dysfunction in a dependent trait T1. If T2 is a part of T1, on which T1 relies for the performance of T1’s proximal function, then a dysfunction in T2 could conceivably “bring down” the whole larger system T1. For example, if the heart (a part of the cardiovascular system) cannot perform its proximal function of pumping, then the cardiovascular system cannot circulate blood around the body. If so, a dysfunction in the heart is causing a dysfunction of the cardiovascular system.
For present purposes, we shall call dysfunctions such as those described above “secondary dysfunctions.” Secondary dysfunctions are dysfunctions which are caused by some original “primary” dysfunction – either via a somatic mismatch, or via the failure of a part causing the failure of a larger trait.
5.4. The Domino Theory in full
Due to the intricate dependence relations which obtain between traits within the bodies of complex organisms, there are a number of ways in which a dysfunction of some trait T2 can cause problems for another trait T1. We have considered these dependence relations in depth above, but let’s briefly recap.
Firstly, a dysfunction in T2 could cause T1 to become somatically mismatched in two overarching ways. It could be that T2 has as its function to provide some precondition or “fuel” which T1 needs for the performance of T1’s proximal function, as in the case of fuel failure or type 1 somatic mismatch. Alternatively, it could be that T1’s proximal function is a “response function” which depends on some cue or input from T2, as in the case of signal failure or type 2 somatic mismatch.
Secondly, a dysfunction of T2 could, over time, cause an independent secondary dysfunction of T1 as a consequence of being either type 1 or type 2 mismatched. Alternatively, it could be that T1 is a larger system that depends on a part T2 for the performance of T1’s proximal function.
My claim is that diseases or “domino dysfunctions” are dysfunctions of traits on which many other traits depend in the ways outlined above, and which therefore cause widespread problems for other functional traits and systems of the body when they fail. Finally, we are able to sum up the Domino Theory in full:
A dysfunction is the failure or inability of some trait T to perform its most proximal function.
A somatic mismatch occurs where a trait T1 is in state of fuel failure or signal failure relative to some dysfunctional trait T2.
Fuel failure: Trait T1 depends on some other trait T2 for the performance of its proximal function. Because T2 is dysfunctional, T1 cannot perform its proximal function.
Signal failure: Trait T1 has as its proximal function F to yield some outcome or effect in response to or in proportion to some input from T2. Because T2 is dysfunctional, T1 is caused to perform its proximal function F in such a way that F fails to contribute to more distal selected effects of T1.
Domino dysfunctions are dysfunctions which either cause other traits of the organism’s body to become somatically mismatched, or which cause other traits of the organism’s body to become dysfunctional, or both.
Diseases are domino dysfunctions.
We are now in a position to draw a principled distinction between two kinds of biological dysfunction: (i) pathological or domino dysfunctions, and (ii) non-pathological or “trivial dysfunctions”. Because our bodies are characterized by a high degree of functional integration and interdependence, most dysfunctions are not just dysfunctions – they are also causes of failures of normal functioning in other traits. However, not all biological dysfunctions have such widespread knock-on effects. Some traits relate to other traits in such a way that, when they fail, there are no, very few, or only very local consequences for other traits. These dysfunctions are on my account “trivial,” and do not constitute diseases.
Trivial dysfunctions are dysfunctions which cause no, or few, other traits of the organism’s body to become somatically mismatched or dysfunctional.
In the following section, I shall consider trivial dysfunctions – and when they might occur – in more depth.
6. Types of trivial dysfunction
Having distinguished trivial dysfunctions from domino dysfunctions, let us return to the problem cases which initially motivated a revision of the simple dysfunction theory. Can the Domino Theory explain why these cases are anomalous? I will argue that the problem cases are all plausibly instances of trivial, rather than domino, dysfunction, and that trivial dysfunctions fall into three broad sub-types. While explaining how the problem cases fit this scheme, I will also extend these principles to consider other cases which have received less attention in the literature, but which nonetheless are plausible examples of trivial, non-pathological dysfunction.
6.1. Spare traits
Sometimes the dysfunction of some trait T fails to have any adverse knock-on effects on other functional traits simply because the trait is “spare.” That is, although the body depends on traits of this type (for example, the body certainly needs cells), the body does not depend crucially on any given token of this type (it does not depend on any given cell). Thus, when a spare trait fails, this need not have any adverse effects.
It is no accident that the body contains spare traits. Natural selection favors robustness, and thus often selects for “spare” parts and fallbacks that can step in and ensure that the overall functioning of the body continues intact even if a part is damaged. As Plutynski puts it: “Multicellular organisms like us have distinctive functional organization and a high degree of … redundancy. These features are adaptive; redundancy enables functional parts to take over when one such part fails” (Plutynski Reference Plutynski2018, 165–6). Because of redundancy, some dysfunctions have few, or only very local, knock-on effects in the form of somatic mismatches and secondary dysfunctions. This is why a biological dysfunction in a single cell, or even a single kidney, is compatible with overall health.
6.2. Isolated traits
I have said above that, because of the functional interdependency of traits, the failure of one very often leads to the failure of others. However, not all traits are equally tightly integrated into the functional architecture of the body. More specifically, some traits do not have, or have very few, dependent traits. These traits are, as I shall term it, “functionally isolated.” When isolated traits fail, precisely because other traits are not reliant on them, the ramifications also tend to be local and isolated.
Let us consider an example: the peacock’s tail. The peacock’s tail was sexually selected, and its function as such is to be aesthetically pleasing to peahens. However, it is not the case that many other biological traits depend on the tail performing its function of looking good for their function. For example, it is not the case that the peacock’s vital organs depend on the tail for their normal operation, nor that the tail is needed for motor control – indeed, the most impressive tails are often so large as to be a hindrance to the peacock’s free movement. Thus, if the tail were absent, or otherwise failed to perform its proximal function (that is, to be attractive), this would likely not lead to many adverse knock-on effects.
A similar story might be told about sexually selected traits in our own species. For example, while females of most other mammalian species remain flat-chested when not breast feeding, some have hypothesized that the “perennially enlarged breasts” of human females are the result of sexual selection (for a discussion, see Pawłowski and Żelaźniewicz Reference Pawłowski and Żelaźniewicz2021). Supposing this hypothesis is correct, if a human female fails to have perennially enlarged breasts, this would be a dysfunction, because her breasts would not be doing one of the things they were naturally selected for (aesthetic signaling), and yet this is unlikely to affect other traits. Sexually selected traits have a function, and can as such dysfunction, but when they do, it has little effect on the rest of the body. They are as such functionally isolated.
How does this apply to the problem cases we discussed in section 4? Well, consider asexuality. Suppose a neural mechanism was selected for opposite-sex attraction, and that this mechanism is dysfunctional; that is, it is failing to cause the individual to be attracted to members of the opposite sex. Even so, it is hard to see which other dependent traits would be affected. Other functions of the brain (perception, mood, etc.) operate normally, other systems of the body such as the endocrine system or the digestive system, are not affected at all. In this sense, asexuality – whether a biological dysfunction or not – fails to qualify as a domino dysfunction, and thus is not pathological.
At this point, some may wonder whether the reproductive system of, say, an asexual woman wouldn’t nonetheless count as being somatically mismatched in the event that her asexuality causes her to fail to become pregnant.Footnote 5 Isn’t her sexual orientation then causing a failure of function, for it is preventing her reproductive system from fulfilling its biological function of gestating fetuses? I don’t think this follows. The proximal function of the female reproductive system is not to “gestate fetuses”, regardless of context or circumstance. Rather, the female reproductive system has a complex response function, more along the lines of “develop a fetus when an egg is fertilized” (see Fagerberg & Garson, forthcoming). If so, then the failure to develop a fetus in the absence of fertilization is not a failure of a proximal function and, hence, not a fuel failure.
But does the female reproductive system not perhaps have as its function to respond to input from the mechanism for opposite-sex attraction, such that the reproductive system could instead be in a state of signal failure when opposite sex attraction isn’t occurring? I don’t think so. I agree that the female reproductive system has as its proximal function to gestate fetuses in response to other processes occurring, but in response to fertilization, not opposite-sex attraction. It is not as if the female reproductive system has as its function to spontaneously develop a fetus at the mere sight of an attractive male. As such, there is no signal failure either. If the condition in question causes no secondary dysfunction, no fuel failure, and no signal failure, then it does not count as a domino dysfunction.
Functional isolation can arguably also account for why we sometimes choose to disrupt reproductive function without any sense that we are causing pathology. If I were to disable some part of my cardiovascular system, many other traits would soon be affected. Similarly, if endocrine function or liver function were disrupted, the effects would be far-reaching. However, if the reproductive system is disabled (say, a man chooses to undergo voluntary sterilization by having his vasa deferentia severed) very little happens elsewhere in the body. This is why voluntary sterilization and contraception differ in principle from having a vital organ such as the liver removed. Traits in the rest of the body simply do not rely on reproductive traits for their normal operation.
We have seen how the functional isolation of certain traits and systems can make sense of some of the problem cases cited in section 4. It is also possible that this idea has novel implications for some conditions we have not yet considered. If contraception and voluntary sterilization are not diseases on the grounds that the rest of the body does not depend on reproductive function, then it would seem that involuntary infertility is not a disease either. However, I suggest in section 7.2 that infertility (voluntary or otherwise) may in fact best be counted among a number of interesting borderline cases.
6.3. Traits that can be compensated for
I now consider a third and final type of trivial dysfunction: dysfunctions in traits that can be compensated for in some way should they fail. That is, even when the trait in question is failing to function, other traits are able to compensate adequately for this loss of function such that dependent traits remain unimpacted.
For example, in some cases of injury to a brain region which was selected for the performance of some function, other areas are able to reconfigure themselves through neuroplasticity in such a way as to compensate for the loss of function in the injured area. This does not change the fact that there is neurological dysfunction in the damaged region of the brain – it is no more able to perform its function than before – but the compensation renders the dysfunction trivial by guarding against any impacts on other traits.
For example, Liu et al. (Reference Liu, Adrian Nestor, Vida, Pyles, Ying Yang, Freud and Behrmann2018) describe the remarkable case of U.D., a six-year-old boy who had his entire right occipital lobe and parts of his temporal lobe removed as a treatment for epilepsy. These regions are implicated in the recognition of objects and faces, among other capacities, and damage to the occipito-temporal lobe is associated with acquired prosopagnosia (Biegler Reference Biegler2018; Corrow et al. Reference Corrow, Dalrymple and Barton2016). The researchers tracked U.D.’s recovery and found that, by the age of ten years and ten months, he had regained almost all cognitive and perceptual functions, including normal capacities for facial recognition. Imaging data indicated that regions in the left hemisphere had been recruited to compensate for the missing regions (Biegler Reference Biegler2018).
If we suppose that face recognition is a selected effect of (some part of) the missing occipital and temporal areas, then it appears there is a clear brain dysfunction in this case. These regions of U.D.’s right hemisphere are in no position to perform face recognition – in fact, these regions are entirely missing. However, because other regions have reconfigured themselves through plasticity to perform this function instead, this failure will not have any further adverse effects on other dependent functions (such as, for example, social functions of the brain which might depend on facial recognition).
Similarly, it may be that biological dysfunctions which are compensated for by highly effective prosthetics fail to count as diseases. Consider a case of valvular dysfunction that is sufficiently compensated for by an artificial heart valve. The dysfunctional heart valve is of course still failing to perform its function, but because this failure of function is sufficiently compensated for by an artificial valve, this will not have the sorts of domino effects which a dysfunctional heart valve would normally have.
7. Implications
Having outlined the Domino Theory, and the types of trivial, non-pathological dysfunction implied by it, I now consider two implications of this theory. Firstly, on the Domino Theory the distinction between disease and non-disease is not strict, but rather exists on a continuum. Secondly, the Domino Theory appears to give us the resources to distinguish between diseases and certain types of risk factors.
7.1. A continuum of pathology
I have said that domino dysfunctions cause widespread dysfunctions and/or mismatches in other traits, while trivial dysfunctions cause none, or very few. If this is so, then the distinction between disease and non-disease is not strict, but rather exists on a continuum. This is not a problem per se, so long as we can account for why some conditions are more paradigm disease states than others. In what follows, I explore this implication and propose some distinctions, before suggesting that the degree and extent to which a dysfunction has knock-on effects tracks something like “clinical significance.”
Firstly, it may be relevant to draw a distinction between those dysfunctions that cause dysfunctions in traits of which they are not part, including independent dysfunctions that remain in the absence of a mismatch, and those dysfunctions that do not cause additional dysfunctions of this kind. For example, heart failure will quickly cause other vital organs to fail, and eventually lead to the death of the organism. Similarly, paradigm cases of serious disease – malignant cancers, degenerative diseases like Huntington’s, or other types of organ failure – would, if left untreated, likely have many adverse effects on traits in the rest of the body, including secondary dysfunctions in a wide variety of traits and systems.
Contrast these conditions with dysfunctions which cause the larger system of which the trait is a part to fail (in virtue of that system’s reliance on this constituent trait) but do not cause dysfunctions in any other traits outside of this system. Perhaps infertility is in this category; if I were to sever my fallopian tubes this would prevent my reproductive system as a whole from performing its function, but only in virtue of the dysfunction in my fallopian tubes. Moreover, no traits or systems outside of the reproductive system would seemingly be affected. This also suggests another dimension along which domino dysfunctions may vary: domino effects may be more or less widespread within the systems of the body, and transcend levels of biological organization to a greater or lesser degree.
Similarly, we might think dysfunctions which cause mismatches, but do not cause enduring, independent secondary dysfunctions, are less paradigm disease states. Perhaps some of what we normally think of as paradigm disabilities are in this category. For example, my systems for motor control are certainly adapted to me being sighted. As such, if I lost my sight, it is plausible that these systems might be caused to perform their functions in ways that compromise more distal selected effects (thus instantiating a type 2 mismatch). However, my blindness is not going to damage the internal constitution of dependent traits, thus causing further independent dysfunctions, and thus would not be degenerative in this sense.
As such, it seems the Domino Theory gives us the resources to distinguish between serious degenerative conditions, on the one hand, and “stable” conditions on the other, exemplified perhaps by paradigm disabilities such as deafness or lacking a limb. This distinction goes some way to explain why a social model of disability might be thought appropriate for deafness, but inappropriate for, say, Huntington’s. Deafness does not cause widespread and enduring dysfunction in other systems, and thus exists at the opposite end of the spectrum to Huntington’s.
My suggestion, illustrated in figure 3, is that the more extensive adverse effects (in the form of dysfunctions, mismatches, and, in particular, dysfunctions in systems of which the affected trait is not part) are caused by a particular biological dysfunction, the more disease-like it is. Conditions that are perhaps associated only with mismatches, or only with higher-level dysfunction due to a faulty part, are less disordered.
Figure 3. A continuum of pathology.
It is interesting to observe, at this stage, that the degree of domino effects appears to track something like “clinical significance.” The more dysfunctions and mismatches are associated with a condition, and the more widespread these effects are, the more clinically significant, serious, or incompatible with health the condition seems to be. Conditions that are associated with few adverse knock-on effects tend to be controversial cases of disease, or the types of conditions which are sometimes argued to be compatible with health. For example, one might be considered healthy while lacking a limb, but few would argue that the absence of a vital organ is similarly compatible with health. My view is that this is because the failure of a vital organ, unlike the absence of a limb, is almost guaranteed to cause a large number of dysfunctions and mismatches in a wide variety of dependent traits, and as such is a paradigm domino dysfunction.
Drawing a strict line anywhere along this continuum is neither possible nor desirable. The important point is that biological dysfunctions are not created equal. Liver failure is different in principle to contraception, or the death of a single cell, and the Domino Theory gives us the resources to draw this distinction on objective biological grounds – even if this distinction is not strict.
7.2. Diseases versus risk factors
Secondly, I suggest that the Domino Theory can help us make sense of the distinction between diseases and risk factors. Sometimes, trivial dysfunctions, which do not count as pathological, may nonetheless count as risk factors for disease in that they increase the likelihood of future domino dysfunctions.Footnote 6
I noted in section 6.1 that natural selection favors spare parts and fallbacks. Precisely because they are spare and as such not crucially relied upon, dysfunctions in spare parts don’t count as pathological. However, damage to a spare part may nevertheless cause the system to become less resilient against future damage. The more spare parts and fallback mechanisms are lost to damage and dysfunction, the less robust the system becomes. In other words, dysfunctions of spare traits may not be pathological in and of themselves, but they will put the organism at greater risk of pathology.
For example, with two intact kidneys, either kidney is spare. Thus, dysfunction of either one of these is compatible with health. However, if one of the kidneys is removed or compromised, the remaining kidney becomes essential, and no longer spare. Whereas the renal system could previously afford to lose a kidney, it now relies crucially on the one kidney it has left. As such, even if the individual is not suffering a domino dysfunction, they are now at increased risk of domino dysfunction due to a loss of robustness. The loss of a single kidney is in this sense a risk factor for disease.
Similarly, it seems plausible that, in certain circumstances, trivial dysfunctions can over time cause biological changes that do not count as domino effects per se – i.e., they aren’t fuel failures or signal failures – but which nevertheless affect the organism in ways which increase the risk of future pathology. For example, suppose a woman uses a copper coil and therefore (by choice) never becomes pregnant. This would not amount to a (paradigm) disease state. However, never being pregnant could have consequences which, although they aren’t domino effects, increase the risk of breast cancer (Husby et al. Reference Husby, Wohlfahrt, Øyen and Melbye2018). In this sense too, trivial dysfunctions may be risk factors – even if they aren’t diseases.
Perhaps this notion of risk factor does not capture all of the ways in which risk factor is used in medicine – more work is required to situate this distinction within existing theories and concepts employed in the medical literature. However, the Domino Theory does neatly capture a distinction between full-blown pathological conditions (domino dysfunctions) and at least some types of risk factor for disease.
8. Counter examples and disease-talk
I have said that some commonly recognized pathological conditions do not count as diseases on the Domino Theory. As such, is the Domino Theory just too revisionist?Footnote 7 In what follows, I respond to two versions of this objection: the first pertains to the Domino Theory’s alleged inability to explain how the term “disease” is ordinarily used; the second pertains to the role that disease-talk plays in medical discourse.
The first version of the objection states that the Domino Theory violates normal medical usage of terms like disease. For example, the Domino Theory implies that even life-long unwanted infertility, which (let us suppose) is due to an injury and causes emotional distress to the individual, does not count as a paradigm case of disease – because any domino effects are local. And yet, many would say that such infertility is clearly diseased. As such, the Domino Theory provides a poor explanation for medical usage, and thus fails to live up to a decisive desideratum: that an adequate account of disease should be able to make sense of our normal application of the term within medicine.
My first response is to point out that this desideratum is decisive only for a certain brand of conceptual analysis – which the Domino Theory is not. The Domino Theory is a descriptive theory of an underlying natural kind, and there is no reason to assume that the boundaries of the kind must conform precisely to our intuitions as to when we would, and would not, be inclined to apply the term. As such, the Domino Theory is not beholden to usage at every twist and turn.
Some may be inclined to respond that, even so, if the Domino Theory is so revisionary as to have completely veered off topic, or “changed subject,” then it is a poor theory of the kind which it purports to describe (see, e.g., Haslanger Reference Haslanger2020). However, the Domino Theory is not wildly revisionist. In fact, it concurs with normal usage in almost all cases, and where it does depart from such usage, a little reflection on the case usually reveals that the condition was always in some sense anomalous. For example, it is true that unwanted infertility is normally considered a disease. However, if we were to classify infertility per se as disease, then that would yield an unintuitive result for contraception and sterilization. This suggests, to me, that reproductive dysfunction is an unusual case. There are not many other functional traits which people voluntarily disable, and where doing so is socially sanctioned and considered compatible with health.
The second version of the objection states that the Domino Theory compromises some important pragmatic functions of disease-talk which we would want to preserve. For example, on the Domino Theory, whether a condition counts as a disease can depend on facts external to the dysfunctional trait, such as whether compensatory mechanisms are in place to guard against domino effects. However, sometimes a clinician or researcher might simply want to say: “This brain is missing large sections of the occipital lobe, and so is clearly pathological.” Perhaps disease-talk of this kind – which lacks regard for what is going on in the rest of the organism – is also quite useful in certain contexts. Does the Domino Theory then not just violate ordinary usage, but in fact impede some important pragmatic functions of disease-talk?
In response, I will make two somewhat related points. Firstly, there is a usage of pathological (exemplified by the use case considered above) which is ambiguous between abnormal, dysfunctional, and diseased. However, there are good reasons to think that these ideas are not co-extensive. A trait can be structurally abnormal while performing its function impeccably, such as with a foot with two fused toes. Moreover, as argued in this paper, dysfunction can seemingly occur in the absence of disease, such as in the case of a single dead cell in an otherwise healthy body. As such, although we certainly want a philosophical theory that is able to say something informative about cases such as that of U.D., I believe that this folk usage of “pathological” is unhelpful in that it is imprecise and open to misinterpretation.
The question from here becomes: does the Domino Theory give us the conceptual resources to discuss cases such as that of U.D. in useful and precise language? Of course, the Domino Theory gives us the conceptual resources to do much more than just point out domino dysfunctions. For example, the judgment that the trait in question is dysfunctional – that the occipital lobe itself cannot perform its function – does not depend on what’s going on in the rest of the body, and so can be inferred from facts about the trait itself. Thus, the Domino Theory is no barrier to describing such cases adequately; we can agree that the trait is dysfunctional without any need to assess the extent of the adverse effects.
However, the conceptual framework suggested by Domino Theory also prompts new questions which challenge us to be even more precise in our characterisation of the case. If the trait is dysfunctional, what sort of dysfunction is it? If it is a domino dysfunction, what kind of effects does it have? If it is a trivial dysfunction, could it nevertheless be a risk factor? In this way, rather than obstructing disease-talk, the Domino Theory in fact has the potential to improve medical discourse by prompting us to be more precise and to the point. Instead of simply saying that the trait is “pathological” (which is potentially ambiguous), the Domino Theory invites more nuanced disease-talk, where things going awry within a complexly interdependent organism can be expressed in more biologically precise terms.
9. Conclusion
I have argued that diseases are domino dysfunctions – that is, dysfunctions that cause adverse knock-on effects in other, dependent traits. Domino dysfunctions can be distinguished on principled grounds from non-pathological or “trivial” dysfunctions. Previous attempts at defining pathology in terms of dysfunction overlooked the degree of functional integration within organisms, and thus failed to recognize that most biological dysfunctions are not singular, local phenomena, but rather causes of biological disruptions in dependent traits and systems elsewhere in the body.
In order to spell out the Domino Theory in full, I first defined dysfunction and distinguished dysfunctions from environmental mismatches. I then applied my account of environmental mismatches to the relationships which obtain between traits within the organism’s body, distinguishing two forms of “somatic mismatch”: fuel failures and signal failures. Domino dysfunctions, I argued, are dysfunctions which cause other traits to become either somatically mismatched, or dysfunctional, or both. I then considered three clusters of “trivial dysfunctions;” dysfunctions which fail to constitute domino dysfunction and thus disease. I finally considered two implications of my view, before responding to the objection that the Domino Theory is too revisionist.
Acknowledgments
I’m especially grateful to David Papineau, Justin Garson, Nick Shea, Tim Lewens, Marta Conti Lorenzo and Simon Lord for their feedback on earlier iterations of this research. I’d also like to thank participants at the King’s College London Start of the Year Conference, the London Mind Group, and the “New Work on the Concepts of Health and Disease” Sowerby Colloquium Series for their insightful questions and comments. Finally, I’d like to thank three anonymous reviewers to this journal for their helpful input.
Open access
