1. Introduction

In defending a theoretical definition of health, Christopher Boorse asserts that disease recognition is “a matter of natural science, not evaluative decision” (Reference Boorse1977, 543). The account that he offers is theoretical in a second, and related, sense. As Boorse has repeatedly claimed, often in response to objection, the account is not primarily about medical practice. It is not meant to guide decisions concerning whether to treat, or what course of treatment should be attempted.

Nevertheless, it does have some relationship to medical practice. For example, Boorse claims that medicine is characterized by a commitment to fight disease (Reference Regan, Humber and Almeder1997, 25), and one advantage of an account such as his is that “all therapeutic programs converge on one goal” (Reference Boorse1977, 571), namely, health. Boorse asserts, “Clearly, almost all traditional medical therapy aims at promoting health” (Reference Boorse, van DeVeer and Regan1987, 382). Moreover, it seems important that the account have some such relationship; otherwise, we might start to question what it is that a theoretical account of health helps us to understand. If it has no link to what doctors actually do, and if it doesn’t help us understand anything about medicine as it is actually practiced, we might wonder whether the account is nothing more than an empty abstraction, of little interest or importance.

My aim here is to evaluate Boorse’s theoretical account of health, not by directly examining the role of values in medicine, but by asking whether medicine really does aim at health so defined. If we accept that medicine takes health as its aim, we should expect much of medical practice to be intelligible as health seeking under our theoretical understanding of that term. At times, other practical considerations may trump the central aim, but, in general, what medicine does should be intelligible as health seeking, if that is indeed the goal on which therapeutic programs converge.

I argue here that Boorse’s account faces particular difficulties in this regard. In section 2 I present the basic problem that his account faces. In short, in order to be health seeking (given his account), medical interventions must seek to restore parts and processes to normal functional ability. But, in fact, relatively few medical interventions do this. Thus, most of what doctors do cannot be understood in terms of restoring health or removing disease. Section 3 then summarizes two possible responses to this basic problem and responds to one. First, one might suggest that the argument here is wrongly focused on medical practice. Given that Boorse offers a theoretical account of health, the objection here misses the mark. I argue that we have good grounds for regarding the objection as a theoretical problem, not a practical one. It is fundamentally about the intelligibility of medicine as health seeking, not the basis for making particular treatment decisions.

Section 4 takes up the bulk of this paper and responds to the second defense of Boorse’s position. That defense suggests that all that is required to avoid the basic objection is to show that his account can support a system for making comparative judgments concerning degrees of health. I argue that his understanding of disease is going to face significant, perhaps insurmountable, challenges supporting such a system. In large part, this is because it faces challenges grounding a clear understanding of how to individuate pathologies in the way that would be required to make such comparisons.

Finally, section 5 argues that there is a more promising route to solving the basic problem presented here. It begins with Daniel Hausman’s recently proposed revision of Boorse’s approach, but it emphasizes that the efficiency of higher-order functional systems must be able to trump lower-order failures. In this way, the solution involves giving up on seeing local functional failures as central to understanding pathology. Instead, higher-order functional failures, failures at the level of the whole organism, become conceptually basic. Thus, I argue that if we begin with something like Boorse’s account of health and we seek to make medicine intelligible as health seeking, we are ultimately forced to give up on one of the core features of the account. Our understanding of pathology is going to be tied to assessing probability of success achieving the higher-order functional goals of the organism, not to local functional failures.

2. The Basic Problem

Health, as Boorse defines it, is normal functional ability. This is understood as “the readiness of each internal part to perform its normal functions on typical occasions with at least typical efficiency” (Boorse Reference Boorse1977, 555). Thus, a part is functioning normally when it makes “a statistically typical contribution . . . to . . . individual survival and reproduction” (555). Further, that contribution must be typical for “an age group of a sex of a species” in order to rule out normal (nonpathological) forms of variation in organisms (555). Such groupings constitute the relevant contrast classes grounding judgments of statistically normal functioning.

Health is then defined as the absence of any disease (Boorse Reference Boorse1977, 567). Thus, it does not matter whether an individual regards his or her condition as undesirable (or desirable), nor does it matter whether the condition is, in fact, a good (or bad) thing to have. All that matters for the question whether pathology is present is if some underlying biological part or process is falling short of normal functional ability. The overall goals of reproduction and survival help us to identify the proper way to decompose the organism into functional parts. These goals are selected only because they are most central to physiology, and Boorse argues that those functions discussed by physiology are the only ones that appear to be relevant to health (Reference Boorse1977, 556).

Given this understanding of Boorse’s account of health, the basic problem it faces can be stated simply. If therapeutic interventions generally aim at health so defined, then they must seek to restore normal functional ability to the part or process that is falling short. However, this is frequently simply not the case. Many examples of therapeutic intervention seek to replace, rather than restore, the functioning of the normal part or process and thus cannot be understood as aiming at health so defined. An insulin pump would not restore the normal functional processes responsible for insulin production. An artificial heart, even if developed to the point of equaling the efficiency of a natural human heart, would not restore normal function to the heart. Many orthopedic surgeries replace the normal part with either an artificial alternative (e.g., an artificial hip) or graft (replacing a torn ligament with a tendon, as in ulnar collateral ligament reconstruction—“Tommy John” surgery). Organ transplants do not restore normal function to the normal part or process. One’s own liver, for example, does not contribute to survival and reproduction after the transplant is completed, even if the surgery is successful. Similarly, many drug therapies—to replace missing compounds, supplement in cases of underproduction, or inhibit in cases of overproduction—aim not to restore the normal means of production (or regulation), but rather to replace that role (or some part of that role) by other means (i.e., drug therapy). In all of these cases, the underlying dysfunction, the disease in Boorse’s sense, remains or is made worse.

In short, if health is to be found in normal functional ability of the relevant parts and processes, a great deal of what medicine actually does would seem to be irrelevant to restoring health. The same disease would remain both before and after treatment. Moreover, the goal of such therapies is explicitly not the restoration of normal functional ability to the normal part or process. The goal, on this understanding, is not to restore health. Understanding what doctors are doing would seem to require some other conceptual foundation.

3. Health Comparisons in Theory and Practice

This conclusion will, of course, seem much too quick. For example, it is quite plausible to claim that someone with a pacemaker does continue to fall short of perfect health. The underlying failure on the part of the relevant biological mechanism remains. We do not need to claim that health is fully restored (or even that full restoration of health is the aim of treatment). It is enough to say that health would be improved (that disease would be reduced). Thus, for example, when Boorse identifies medical treatment in its “core sense” as health seeking, he immediately notes that it is “the removal, mitigation, or prevention of pathology” (Reference Boorse, van DeVeer and Regan1987, 382–83). Further, he notes that while cure of an infectious disease is an example of removal, “examples of mitigation (substitution of mild for severe pathology) include most other medical treatments and all surgery” (383). He thus acknowledges that the majority of medicine cannot be understood in terms of restoration of normal functional ability to the normal part or process. It cannot be understood as restoring health in this sense. Rather, it has to be understood as mitigating disease (improving health)—as he puts it, trading one form of pathology for a less severe form of pathology.

On this account, understanding medicine as health promoting is dependent on our ability to make comparative judgments concerning degrees of health. Given Boorse’s understanding of health, these comparisons will depend on some measure of the relative severity of disease in comparison cases. The question, then, is whether an account such as Boorse’s can support a system for making such comparisons. If it can, then there is still the chance that the account can render core medical activities intelligible as health seeking.

Before directly addressing this question, however, it is important to address a possible objection to the way this problem has been identified and presented. As noted above, Boorse has consistently drawn a distinction between his theoretical account of health and medical practice. Thus, one might object that the concern advanced here does not touch the core account of disease as Boorse conceives it. The understanding of health that he offers is “the pathologist’s concept, whereby many pathological conditions are wholly local and trivial” (Boorse Reference Almeder and Gifford2011, 57). It is not the clinician, or clinical decision making, that he aims to understand, so one might argue that the kinds of clinical examples mentioned here are irrelevant to the account he is offering. On Boorse’s understanding, we don’t turn to the pathologist’s understanding of disease alone in clinical cases; we turn to that account of disease plus any number of additional factors that give clinical relevance to the disease (Reference Regan, Humber and Almeder1997, 100). When assessing clinical interventions, it is these other factors that are of particular interest, not the underlying, theoretical conception of disease.

My concern here, however, is not that Boorse can’t offer an understanding of such interventions from a practical point of view. Instead, it is that the interventions identified are not intelligible within medicine from a theoretical point of view. Conceptually, we cannot understand them in terms of health and disease alone. The only way to make most medical interventions intelligible is to turn to any of a number of auxiliary concepts. Thus, in wide swaths of medicine, the theoretical understanding of disease is irrelevant to our understanding of what physicians do and what their goals are. It isn’t just irrelevant to questions of practical decision making (e.g., should I treat); it is irrelevant to understanding what physicians are doing at all. If we regard health and disease as the central concerns of medicine, then failing to make interventions intelligible in terms of health and disease is a serious problem. To put the point in Boorse’s own terms, the worry is that on his account, nearly everything that medicine does would fall outside the “core sense” of medical therapy. Almost all of medicine would be “peripheral medical treatment” akin to elective plastic surgery (Boorse Reference Boorse, van DeVeer and Regan1987, 383).

Thus, the concern here is a theoretical one. Medicine is primarily (though not exclusively) a health-seeking endeavor. Any understanding of what health is must preserve this fact about medicine. The concern isn’t whether therapy should be attempted in a given case. It is whether swaths of medicine are intelligible as health seeking and thus intelligible in terms of the central goal of medicine. Failing to make a range of paradigmatic medical interventions intelligible as health seeking severely limits the ability of any theory of health and disease to aid us in understanding medicine.

4. Counting Dysfunctions

Can Boorse’s definition of health and disease support a system for making health comparisons? Given his account, closeness to the relevant normal functional state would seem to be the pertinent measure. If we could assess changes in the efficiency of the relevant parts and processes, we could track changes in the severity of a pathology. If the efficiency of the normal part or process that is falling short were to be improved, the person would be brought closer to health as Boorse defines it.

This basic approach, however, will not work for many examples identified earlier. In many cases, the intervention in question brings no improvement in the efficiency of the normal part or process. Implanting an artificial pacemaker does not restore the normal means of regulating the heart; it replaces a defective system with an alternative means of carrying out the same functional goal. As with this case, many examples mentioned earlier involve some other, often artificial, process wholly replacing the normal part or process. It would seem that any movement is away from, not toward, health. Replacing a diseased heart with an artificial heart would result in a reduction of the (admittedly limited) contribution that the normal, natural heart makes to the survival and reproduction of the organism. Replacement would not improve the efficiency of the normal part. The individual would seem to be, in Boorse’s sense of the term, less healthy after the operation than before it, even if the operation were entirely successful.

This suggests that instead of focusing on a single part or process, we should look to the interconnected nature of functional systems within the body. One part or process failing to perform its immediate function will likely result in other parts and processes falling short of typical efficiency in carrying out their tasks. An intervention could then stop the cascade of problems that results by replacing the functional role of the problematic part or process. Thus, while pathology remains (there are still parts and processes not playing their typical roles with at least typical efficiency), the individual is still brought closer to health because other parts and processes are returned to normal efficiency. By counting the number of parts and processes falling short before and after an intervention, one could argue that health is improved based on a reduction in that number.

In this way, a general account of health comparisons could be tied to the number of pathologies present in different cases. One could compare degrees of health either for a single person over time or across individuals. If one could assess the typical number of further impacts caused by a particular pathology, then one could offer general comparisons between types of disease. One could argue that disease of type X tends to be worse (all things being equal, from a theoretical perspective) than disease of type Y, because X tends to cause more disruptions to further systems within the body, or perhaps has the potential to cause more such disruptions.Footnote ¹

Obviously, the key to making such an account work is providing clear guidance on how to count. That system of counting needs to sustain consistent comparisons across time, individuals, and types of pathologies. In order to count, we must be able to identify and individuate pathologies. We must be able to tell when we should count disruptions as parts of a single pathology and when we should count multiple, different pathologies.

To illustrate complexities here, it will be useful to consider a simplified case. We can take as an example a very basic genetic disease. A single mutation to the normal sequence results in the creation of a protein that lacks the structural properties needed to carry out its ordinary functional roles. Assume further that the condition is recessive, requiring the inheritance of two copies of the mutated sequence for the further dysfunctions to manifest. The inheritance of a single copy of the normal sequence results in adequate production of the required protein for all further function to be carried out. So long as the protein produced does, in fact, normally play some role in systems that impact reproduction and survival, we have a basis for saying that pathology is present given Boorse’s definition.

For the moment, we can focus on just the local, limited context of protein production. Even in this limited context we have a number of parts and processes to consider. For parts, we have not just the relevant DNA sequence and protein but also the mRNA and tRNA (and other molecules) involved in the process of protein synthesis. For processes, we can not only consider the whole process of protein synthesis but also break that process down and consider transcription and translation as separate processes (or even further, consider different steps in those processes as themselves separate processes).

Which of these parts and processes, given our hypothetical, fit Boorse’s definition of pathology, and how should we count them? Consider first the relevant segment of DNA. In the species design, the DNA segment that corresponds to our hypothetical mutated sequence makes a contribution to reproduction and survival through its role in the production of the functional version of the relevant protein.Footnote ² Since the mutated sequence is not playing this functional role (at the very least in the case where two copies of this sequence have been inherited), it would seem to count as pathological.

Accepting this, how then should we count instances of the sequence in order to make comparisons? Do we count every instance of this sequence, or do we only count when the production of this particular protein is normally called for? Here, it is important to remember precisely how Boorse defines health. He contends, “Health in a member of the reference class is normal functional ability: the readiness of each internal part to perform all its normal functions on typical occasions with at least typical efficiency” (Reference Boorse1977, 562). Disease, then, is “a type of internal state that impairs health” (562). More directly, “A condition of a part or process in an organism is pathological when the ability of the part or process to perform one or more of its species-typical biological functions falls below some central range of the statistical distribution for that ability in corresponding parts or processes in members of an appropriate reference class of the species” (Boorse Reference Boorse, van DeVeer and Regan1987, 370). Thus, any impairment of the ability of a part or process to carry out its functions when typically called upon counts as pathological. It is important to focus on the readiness, or ability, of parts and processes to perform functions, because as Boorse points out, “biological functions are usually performed on appropriate occasions, not continuously” (Reference Boorse1977, 562). Thus, even if a hemophiliac is never injured and clotting is never required, hemophilia remains pathological (562).

If we follow this same line of reasoning, we can conclude that we should count the sequence as pathological at all times, even when it is not being called upon. But what should we say about the sequence in other tissues, in parts of the body where this sequence will never be called upon to produce this particular protein? Are these further examples of the same part not being ready to carry out the needed functions when called upon? Medical usage would seem to suggest that they are. We can speak about the specific gene for beta globin, for example, and identify its location (near the end of the short arm of chromosome 11) even in DNA taken from cells not associated with production of blood cells. We can identify specific mutations to the normal sequence and demonstrate the role they play in giving rise to molecules that cannot carry out their normal functional roles with typical efficiency. This allows us to undertake genetic screening for conditions such as sickle cell disease or beta thalassemia.Footnote ³ It is not hard to find examples of medical professionals speaking about the gene for sickle cell disease; in those contexts, it seems that they refer not just to the sequence when it is found in the specific context where it results in the production of the problematic form of beta globin. The relevant mutated sequence found at this location on chromosome 11 is regarded as pathological, as associated with the specific condition sickle cell disease, wherever in the body it is found.

This may not, however, be as conclusive as it first seems. Finding the mutated sequence anywhere in the body would indicate that the sequence is also present in the specific cells where it is normally called upon. Thus, even when drawn from any other context, the sequence is a sign that pathology is present elsewhere. Shorthand references to genes for conditions may not indicate the judgment that the sequence constitutes pathology in all cases. It may simply indicate this general fact that the presence of the sequence anywhere means that pathology is present in those specific contexts where this mutation results in lowered functional efficiency.

Nevertheless, it is hard to see how to resolve the question whether to count every instance of the sequence in the body or just count in those specific cases where the sequence might be called upon to produce the problematic version of the relevant protein. In favor of counting just in specific cell types, one could emphasize the contextual nature of many genetic functions. Specific parts of chromosomes can play multiple functional roles and be a part of multiple different genes, depending on the context in which they are found. As the organism develops, some genes may only be involved in protein synthesis in specific tissues at specific times. Thus, we can say that a copy of a particular sequence, in particular tissues, will not play any role in the production of a particular protein. As such, in those contexts in the species design, it does not make a causal contribution to the functional goals associated with that protein. There is some reason, then, to suggest that on Boorse’s account, it does not have the relevant function in those contexts. Thus, it cannot be falling short of contributing with typical efficiency in those contexts and should not be counted.

In this way, this case is not analogous to the hemophiliac. For the never-injured hemophiliac, nothing in principle about the functioning of the organism prevents the kinds of injuries that would call upon the functions of the clotting factor. In this sense, the organism is simply lucky to never get injured. If we imagine the possibility of manipulation of the genetic sequence in specific cells, we can imagine a case where gene therapy could be carried out in just those contexts where the sequence in question would be called upon to create the relevant protein (and perhaps the germ line as well). In such a case, where the normal gene is present in every circumstance in which the protein might be produced (and there is no possibility of passing the gene on to a future generation), it would seem that no function is not being carried out in the organism in question. Moreover, it would seem that no part would fail to function with typical efficiency when called upon.

The examples of the sequence in other contexts can’t be called upon in the way required, because they are not in the relevant context. The body simply does not call for production of this protein in those contexts. It isn’t just that the circumstances happen to not arise; the biology of the situation is such that they will not arise. We can say that the mutated sequence in question has no role to play, and no potential role to play, in all those wider contexts, and the sequence that could be called upon (assuming our genetic manipulation) is capable of fulfilling its functional role. Thus, for example, having a sequence associated with sickle cell present in a skin cell would not constitute pathology.

At the same time, the very possibility of genetic manipulations suggests that we have good reason to continue to regard the sequence as pathological. We would not want to put this mutated sequence where it might be called upon to produce the relevant protein precisely because it is incapable of playing a role in producing a functional version of that protein. In that sense, the mutated sequence is clearly not ready to play all of the functional roles that might be expected of it within the species design. So, if in the abstract we would want to recognize this fact, should we not recognize it in every instance where it is found in a cell in the body? In this way, the sequence is a pathological sequence regardless of where it is found.

Resolving this question would, in part, require greater clarity concerning what constitutes a medically relevant part type. Since our hypothetical sequence plays no role in producing proteins in many contexts within the body, we could say that it doesn’t constitute a medically relevant part in those contexts. The relevant type here isn’t the sequence found at location X on chromosome Y; it is the sequence potentially triggered to initiate in the production of protein Z. In other contexts, the sequence at that location makes no causal contribution to survival and reproduction and hence has no function relevant to medicine (assuming, again, that the sequence plays no other functional roles in those other contexts).

The alternative would regard the sequence at that location as a single part type regardless of the context in which it is found. Its medically relevant functions would be those tasks it is called upon to carry out in specific contexts that causally contribute to reproduction and survival. In our hypothetical case, the only relevant function is the creation of a specific protein. If the sequence in question has a defect that renders it incapable of producing a functional copy of that protein, then that mutated sequence is pathological wherever it happens to be found.

To put the point another way, it isn’t immediately clear how we should characterize the “corresponding parts or processes in members of the appropriate reference class of the species” (Boorse Reference Boorse, van DeVeer and Regan1987, 370). Is the corresponding part here the sequence found at a specific location on a chromosome, or is it the specific sequence called upon in those functional systems that produce the relevant protein? There are competing visions of relevant medical part types and no clear way to resolve the question.

Moreover, the difference in these two understandings for counting is potentially huge. If we must count every instance, we will have to count instances equal to twice the number of cells within the body (focusing on the case where two copies of the gene are inherited). If we are just counting instances where the sequence may actually be called upon to produce the protein in question, that number is going to be significantly smaller depending on when and where the protein is normally produced. If comparing counts is our measure of disease severity, there is potentially a very large difference in our results.

Further complications appear if we shift our attention to the process of protein synthesis. First, much as we can question what is the relevant part type, we can question how we should conceive of each process. In our hypothetical example, we can think about protein synthesis as a single process, carried out constantly throughout the body. That process makes its contributions to survival and reproduction through making the functional products required to carry out many further goals within the organism. If one of those products is unable to carry out its required task with typical efficiency, then the process of protein synthesis as a whole is also not contributing with typical efficiency. The genetic mutation results in the process of protein synthesis being unable to function with typical efficiency.

However, similar to our discussion of the DNA sequence, we can ask whether we should count protein synthesis as a single process that isn’t contributing, or whether we should count only those particular instances of the process when it fails to produce the functional protein that would be produced in the species design. Does protein synthesis in general count as a particular process type, or, instead, should we identify the production of each particular protein type as a separate process? Or should we perhaps consider and count both the general process and the specific protein production processes? Here we have the added complication that the process of producing this particular protein is an instance of the more general process type, protein synthesis. The general process isn’t contributing normally precisely because the production of this particular protein is failing. In that sense, it seems difficult to count the particular process without also counting the general process. The particular process failing to have the ability to make its normal functional contribution means that the general process will also fall short.

We could block that linkage if we could argue that, in some sense, the general process doesn’t constitute a genuine, medically relevant process type. In this case, the argument might be that the general process itself isn’t, in some sense, a real process, and that it doesn’t have any existence in the functional design of the organism other than as an abstraction from the many particular protein production processes that actually do play causal roles in the organism. Here, though, we might worry that the same line of reasoning could be used to suggest that there isn’t a general process for producing any particular protein type either. There are just the particular processes that go on in each cell; that is the only reality to the process type picked out by reference to production of a particular protein.

Each time we reason in this way, however, we seem to lose some general knowledge about processes within the body. There are general features of protein synthesis that we understand and can discuss. It is because those general roles are carried out that particular proteins get made. Carrying out those roles successfully is crucial to the survival and reproduction of the organism. If a general problem with the process of protein synthesis prevented part of the process from being executed in every instance, that would constitute a catastrophic pathology that would be incompatible with life.

Thus, it seems that we need to count the process of protein synthesis in general, because it is a process that contributes to survival and reproduction. If we do include it, however, we encounter a different problem. When we note that the general process of protein synthesis is not making its contribution to reproduction and survival, this is because the process that creates the specific protein in our hypothetical pathology is falling short. This failure, in turn, is due to the problematic mutation to the normal DNA sequence. Thus, while the protein synthesis process is not contributing with typical efficiency, it is also not the case that there is some inherent problem with that process in general. That process is failing only because a component that serves as input into one instance of the process type is defective. There is nothing abnormal about the general way the organism is processing DNA sequences to produce proteins. That process is functioning as it should. Even in the context of the production of our particular example protein, the process of reading and processing the sequence to produce the (defective) protein is happening as it normally should. The process is not the problem. The DNA sequence is. Thus, if we count the process as pathological, our understanding of pathology fails to account for an important distinction between a part or process that is intrinsically pathological and one that only behaves as such owing to problems elsewhere in the organism.

If we simply try to answer the question of when we can say that pathology is present in the organism, then there is no problem with the account. The general process not contributing with typical efficiency does, at least in this case, mean that there is pathology present in the organism. If, however, we aim to identify those parts and processes that are themselves pathological, failing to account for this distinction does constitute a problem. In our example, the DNA sequence and protein are pathological in a way that the process of protein synthesis is not.

Revising Boorse’s definition of pathology to try to exclude this kind of example would create further problems. If we refuse to count the process of protein synthesis in making health comparisons, we run the risk of undermining the basis for our system of comparing cases. The process of protein synthesis isn’t making its normal contribution because of the underlying genetic pathology. If we don’t count it, because we do not regard it as a truly independent pathology (we don’t regard the process as, itself, intrinsically pathological), then to be consistent we should do the same in all similar cases. This means, however, that only functional systems that are damaged in some ongoing way by the effect of pathology elsewhere will make it into our count. If restoring normal function to one part of the system results in the other parts of the system returning to normal functional ability, then we can say that those parts and processes continued to have the ability to function normally in normal circumstances. The problem was not that they were pathological; it was only that the circumstances in which they were asked to function were abnormal owing to pathology elsewhere. By not counting them, however, we cut off the cascade of problems (and restorations) that was supposed to allow us to make medical interventions again intelligible as health seeking. Those further failures that are restored by our intervention would, on that basis, be ruled out of our count used to assess level of health. They would be ruled out precisely on the basis of their restoration.

If we do count the process and yet refuse to call it pathological, we have to admit that our assessment of health is something other than the absence of pathology. When we count parts and processes to make judgments of relative degrees of health, we count things that are not themselves pathologies. They constitute signs that pathology is present (or effects of pathologies that are present), but they are not themselves pathologies. That means that the elimination of these other sorts of things is equivalent to improving health (lowering the number that we are using to count and compare). Health, then, would not just be the absence of disease.

There is much more that could be said here, even about our simplified, hypothetical case. A truly full examination of the issues raised, however, is not possible in the space available here. What can be said, based on the arguments to this point, is that a system for counting local pathologies consistent with Boorse’s definition will encounter significant hurdles in identifying which part and process types are correct for comparison with the species design. Such an account will also need to resolve how to count consistently without cutting off the cascade required to resolve the problem that the account was meant to solve, or obscuring the difference between systems that are actually pathological and those that are merely behaving as such owing to the effects of pathology elsewhere. The arguments here give reasons to be skeptical that defensible solutions to these difficulties can be found.

5. Global versus Local Definitions of Health

As noted earlier, Boorse intends his account to capture the pathologist’s understanding of disease. This focus gives local part dysfunctions conceptual priority. The species design tells us what contributions local parts should make, and those parts can then be assessed independently of facts about how other parts and processes in the organism are behaving. As we have seen, however, we cannot use just the local functional achievement of a particular part or process to assess degrees of health. We need to turn to broader systemic considerations. Counting local pathologies is, then, a way to try to account for these systemic effects without having to give up the local focus of the account. As we have seen, this approach forces us to systematically examine parts and processes throughout the organism and determine whether each should be counted. In many cases, it is not clear what parts and processes we should consider and count. It is worth asking, then, whether there are alternatives to this focus on local pathology.

One author who has taken steps in that direction, within a general approach similar to that of Boorse, is Daniel Hausman. Hausman argues, “what defines greater health is greater functional efficiency—that is, enhancement of the probability of survival and reproduction or of the probability that systems within organisms achieve their goals” (Reference Hausman2012, 540). Notice that this has shifted attention away from just the definition of a state as pathological to consider degrees of health. Thus, the definition he is offering should be well placed to resolve the central problem identified here. Notice also that the assessment of health is no longer entirely local. The probability of overall success, of reproduction and survival, is directly relevant to the assessment of degrees of health.

I say that the account is not entirely local because Hausman also clearly tries to maintain an understanding of local pathology. That is the importance of the final clause of the definition above. That creates an ambiguity in the account. If probability of reproduction and survival is to serve as the measure of degree of health, then a purely local failure of a system to carry out its goals would not lead us to count an organism as less healthy than one without that local failure. Without an impact on probability of achievement with respect to these goals, there is no difference in health. If purely local failures do, in fact, affect our assessment of level of health, then achievement with respect to reproduction and survival is not what defines level of health. If both together are to be assessed in making the comparisons, we need some way to combine them into a coherent single measure.

We need to consider, then, whether Hausman gives us a basis for deciding between these alternatives. He does assert, for example, that “functional efficiency within the organism is usually of more interest and importance” than efficiency within a particular system (Reference Hausman2012, 535). There is good reason, however, to take an even stronger position than this. Consider, for example, the way that Hausman deploys his account to try to resolve a lingering objection to Boorse’s basic approach. Authors such as Cooper (Reference Cooper2002) and Kingma (Reference Kingma2007) have suggested that Boorse cannot independently justify the selection of the relevant reference class for identifying pathology. For example, in cases such as different blood types or eye colors, Boorse asserts that the trait can be handled disjunctively. The relevant species design includes having type A blood, or type B blood, or type O blood, or type AB blood (Boorse Reference Boorse1977, 558). The only apparent reason to allow such disjunctions within the design, however, is a prior belief that such variation in functional design is not pathological.

Hausman, however, believes he can explain why such differences are not pathological. He points out that while “only a small percentage of human beings have blood type AB,” it is the case that “type AB blood is still a normal variant, not a pathology, because the functional efficiency of type AB blood is no worse than the functional efficiency of more common blood types” (Reference Hausman2012, 539). This answer, however, depends on accepting that the achievement of a higher-order systemic goal can trump lower-order failures. This is because a part in one system can be a goal in another. We can thus often shift our level of analysis and find a local failure to function efficiently even when there is no effect on the probability of achieving a higher-order goal. This is easy to see in the case of eye color. Blue eyes result from the failure to produce sufficient melanin to color the eye brown. Thus, in persons with blue eyes, the systems responsible for producing melanin in the eye are not functioning efficiently in comparison to the normal case. That failure, however, does not count as pathological because having blue eyes does not significantly reduce the overall functional efficiency of the eye or organism. If higher-order functional efficiency cannot trump in this way, we lose the explanation for why blue eyes should be counted as a healthy variant rather than as pathological.

This is going to be particularly important in making therapeutic interventions intelligible as health seeking. The kinds of interventions discussed above often left in place, or made worse, a local failure to function efficiently. What interventions sought to accomplish were improvements in functional efficiency at more general levels of functional organization. Thus, in order to make such interventions intelligible as health seeking, the probability of success at achieving those higher-order functional goals must be the relevant measure, not the local failure.

If we take this route, then, and revise Hausman’s definition to reflect the priority of higher-order functional systems, we can say that greater health is greater probability of functional efficiency with respect to the overall functional goals of the relevant whole system. The relevant overall goals are the highest available goals for the relevant whole, often the organism, and thus will not be trumped at a further level of analysis. Notice that this proposal does not mention survival and reproduction specifically. In this context, it is far from clear that survival and reproduction are the only relevant goals for assessing health. Moving to this kind of approach is going to require a much wider range of ultimate functional goals. Those goals are now directly relevant to assessing levels of health. They are not just used to identify the relevant functional design. Clearly, a great deal more needs to be said about the identification of the relevant goals (a discussion that may ultimately call into question whether the account is still one that is free from evaluative judgments). That discussion, however, would go well beyond what can be accomplished in the space here. For the moment, it will be enough to say that this looks like the more promising route to a defensible theory of health that maintains our understanding of medicine as primarily aimed at health preservation.

To further support that conclusion, we can show how this approach would avoid the problems encountered by counting pathologies. Since we have shifted the focus of the account from counting local pathologies, we no longer need to identify and classify nearly every relevant part and process in the body to arrive at a basis for comparison. We instead look to the probability of achieving overall goals of the organism.Footnote ⁴ Thus, in our example, we are not forced to classify the DNA sequence as generally pathological or not. We can say that a DNA sequence is pathological for an organism if, when called upon, its actions cause functional failures that affect the overall organism’s chance of functioning efficiency. If this sequence generally has this effect, we can say that the sequence is generally pathological for organisms of the relevant type. We can also say that the sequence, in contexts where it is not called upon, does not count as pathological for an individual, even though we generally consider the sequence pathological because of the effects it has in those contexts where it is called upon for organisms of this type. Thus, on this approach, the puzzles discussed above do not arise. We can accept different classifications in different contexts because being pathological depends on further effects on the functioning of the organism.

At the same time, we need not regard the failure of the general process of protein synthesis as itself pathological, even though it is not making its normal contribution. The failure of the general process is not the cause of the more general failures of the organism to function efficiently. A defect (apart from the mutated DNA sequence) in the specific process of protein synthesis that resulted in the defective protein is also not the cause of the further failure. The causes of the further failures are the mutated sequence and the defective protein. Failing to classify the process as defective does not threaten to undermine the identification of further parts and processes that are restored by therapy because we do not need to identify particular local pathologies that are restored. It is enough to say that the probability of achieving some functional goals of the organism as a whole is improved.

Finally, it is worth asking whether this is giving up on an understanding of purely local pathologies. In one sense, yes it is. Pathology is assessed in terms of the effect on the probability of achievement of the ultimate functional goals of the whole in question. In that sense, assessments of health are always global. We can, however, still discuss a pathological liver cell. In such cases, the relevant whole just is the liver cell. The ultimate goals of the liver cell are determined by its being a liver cell and having certain functional roles to play in the body, but that doesn’t prevent us from assessing whether a liver cell is capable of carrying out its particular overall functional goals. Whether that liver cell affects the health of an organism, however, depends on examining the effect of that cell on that organism. Whether pathological cells of this type generally should be regarded as affecting the health of individuals of a certain type depends on what general effects, all things being equal, such cell types have on the probability of achieving the relevant overall functional goals of the organisms of that type. The pathologist’s local understanding of disease thus becomes a special case of a more general account, not the foundation for a general theory of health.

Again, it is worth mentioning that there are still questions about such an approach that would need to be resolved, particularly concerning the identification of the relevant overall functional goals of the particular wholes. Thus, there is further work to be done in developing and defending a final account. The arguments here, however, suggest that this is a more plausible route to providing a theory of health that can render medicine intelligible as primarily aimed at maintaining, improving, and restoring health.