1 Introduction

The machine metaphor certainly provides insights, but these come at the price of overlooking much of what biology is... A metaphor far more to my liking is this. Imagine a child playing in a woodland stream, poking a stick into an eddy in the flowing current, thereby disrupting it. But the eddy quickly reforms... Organisms are resilient patterns in a turbulent flow—patterns in an energy flow. A simple flow metaphor, of course, fails to capture much of what the organism is. None of our representations of organism capture it in its entirety.—Carl Woese (2004, p.176), in A New Biology for a New Century

This century has seen the rise of two new philosophies of biology: New Mechanism and Process Ontology. I refer to these as mechanism and processualism. These views are distinguished by their ontological commitments and by how they provide scientific explanations. This essay argues that the ontological commitments of these philosophies are in conflict and shows how to achieve a form of reconciliation between them.Footnote 1 It is addressed to ontologists of either sort who take issue with the other, and to those who aim for reconciliation between them without commitment to either. I first argue against two common forms of pluralism about these ontologies (§ 2), then present another option (§ 3). This option is to adopt an ontology that combines elements of both processualism and mechanism while localizing them to different durations surrounding phenomena of interest. On this view, some parts of the world are mechanisms, others are processes, and neither is the most fundamental or global. I argue that this is an overall better way of capturing the processual and mechanistic features of different parts of the world, organisms especially.

Mechanism began this century with Machamer et al.’s (2000) exploration of a new mechanism (though some of the key ideas pre-existed in Glennan, 1996 and were followed by Bechtel & Abrahamsen, 2005; Craver & Darden, 2013; Glennan, 2017; inter alia, see Craver & Tabery, 2019). Mechanist views descend in part from the mechanical philosophy of Descartes. While largely dispensing with metaphors or analogies to organisms as machines (c.f. Glennan, 2002, 2017; see Krickel, 2018, p.33), they retain commitment to a dualist ontology of entities and activities. Mechanists argue that their ontology provides the correct basis for explanations in bioscience. A decade later saw the development of contemporary biological processualism, developed from the process ontology of the 20th century largely by Nicholas Rescher, Johanna Seibt, Dan Nicholson, Anne Sophie Meincke, and John Dupré and collaborators, (Rescher, 1996; Seibt, 2009; Dupré, 2013, 2020; Meincke & Dupré, 2020 and collaborators therein, Nicholson and Dupré 2018 and collaborators therein). Processualism ultimately descended from accounts of flux or change in Heraclitus and Nietzsche or, more recently, Whitehead. Processualists likewise argue for “process ontology as the most appropriate ontological framework for research on organisms in science and philosophy” (Meincke, 2020). Many arguments have been put forth in support of processualism, usually as an alternative to a substance or thing based metaphysics (e.g. Rescher, 1996; Dupré, 2013), or to a point or particle based metaphysics (Bickhard, 2011). Although also largely ridding itself of reliance on flow metaphors, processualism retains a commitment to a monist ontology of processes.

Both of these ontologies are contenders for our scientific ontology and they are in tension with one another. Both ontologies have proven fruitful for understanding certain biological phenomena, enabling distinct though complementary forms of understanding. A large part of biology is concerned with the elaboration of mechanisms explaining phenomena of interest (e.g. evolutionary mechanisms, such as the mechanism of hybrid dysgenesis, and molecular mechanisms, such as the mechanism of glycolysis to explain respiration, or the mechanism of sodium-potassium channel activation to explain axon-dendrite depolarization). An equally large part concerns processes (e.g. evolutionary processes, such as the process of speciation or lineage formation, as well as metabolic processes, explaining catabolism or analbolims generally, or specific processes such as concentration dependent feedback). Talk of mechanisms and processes, indeed sometimes about the same event (e.g. gluconeogenesis) is ubiquitous, and arguably pervades every area of the life sciences. I take this as sufficient reason to aim for a reconcilation between them.Footnote 2

The conflict between these ontologies is in part as Woese highlights above. The metaphors and representations commonly associated with these ontologies—machines and flows or vortices—do provide valued insights across science, but neither alone seems to be appropriate to or fully capture every phenomenon we are interested in. Moreover, for some phenomena one or the other seems markedly inappropriate. Critics of the machine metaphor, e.g., have argued that it is widely over-applied, if not generally indefensible for understanding organisms (Nicholson, 2013, 2018, 2019). However, the conflict between these ontologies is not just about their associated metaphors.

Some conflict is about their respective empirical or descriptive adequacy. Krickel (2018, ch.2, p.34) has argued that one form of mechanismFootnote 3 is not descriptively adequate for the explanatory practices of the life sciences. Drawing support from the mechanistic criticism of Nicholson (2012) and the process philosophy of Dupré, Krickel ([ibid]) argues that, when using mechanistic language, scientists are not usually referring to mechanisms qua objects with stably arranged parts. Similarly, Godfrey-Smith (2016); Skillings (2015) have argued that mechanism is unable to capture some of the empirical features of key molecular biological events (e.g., stochasticity in protein synthesis), and Dupré and Nicholson (2018) have claimed that mechanism is unable to account for various metabolic turnover events. On the other hand, Austin (2016) has argued that processualism is unable to handle some cases of modular dynamics from developmental biology, which he believes can be provided for by mechanism. Indeed, critical engagement between these ontologies is often coeval with their development. Machamer et al. (2000) include criticism of processualism within their presentation of mechanism (e.g. p.5); Dupré and Nicholson (2018) include criticism of mechanism within their presentation of processualism (e.g. p.29, also in Nicholson, 2012; Dupré, 2013). There is significant, and more than merely verbal disagreement between these views.

Since both have been successful, but both in conflict, if we understand scientific phenomena solely through one of these metaphysical lenses then we may miss out on features captured, or captured better, by the other. As with all cases of theory choice, when faced with two conflicting ontologies a Quinean attitude is both ideal and initially plausible—“the obvious council is tolerance and an experimental spirit” (Quine, 1948, p.38). In service of understanding the heterogeneous phenomena of biology and in recognition of the mutual contributions of such understanding by mechanism and processualism, it seems worthwhile to give up on choosing one over the other of these ontologies absolutely and to adopt some kind of pluralist ontological position. However, there are significant obstacles to any straightforward attempt to adopt a pluralism about mechanism and processualism.

Below (§ 2) I argue that two prominent varieties of pluralism are untenable for this pair of ontologies. First, I argue against varieties of pluralism that attempt to unify both views or hold the commitments of both simultaneously, what I call unified pluralism. I argue against this by showing that the commitments of these ontologies are inconsistent with one another (§ 2.1). Mechanism is a dualist ontology, involving commitment to disjoint sets of entities and activities, and processualism a monist ontology involving commitment to processes alone, so their union would need to be both.Footnote 4 No matter how fruitful that may be, an ontology cannot be both monist and dualist.Footnote 5 Second, I argue against varieties of pluralism that allow the free adoption of each view separately within a community, what Garson (2020), in another context, refers to as within-discipline pluralism. I argue against this sort of pluralism by showing that ontologists of each sort are in possession of arguments—here called covering arguments—that have the intended effect of putting their own ontology in a better position, as either the most fundamental or globally applicable ontology (§ 2.2). This puts any discipline with such pluralism in an unstable and unhelpful position: a reconciliation by stalemate. This is good reason to abandon these varieties of pluralism and look elsewhere for a way to reconcile these ontologies; Quinean tolerance will need to be obtained by other means.

In the final section (§ 3) I describe and argue for another sort of reconciliation. Another variety of pluralism, perhaps, but one with a distinctly ontological flavour. This reconciliation is to adopt a local ontology.Footnote 6 For a pair of ontologies, a local ontology consists of two central claims,

  1. 1.

    To each location there corresponds one or the other ontology, exclusively.

  2. 2.

    Neither ontology is most fundamental nor most global.

I elaborate on both of these claims below. On a local ontology for processualism and mechanism, (1) each location is either processual or mechanist, exclusively, and (2) neither ontology is held most fundamental or globally applicable. This involves rejecting both absolute monism and dualism and adopting a locally variable patchwork of ontologies instead. Moreover, I show that (2) can be held in a way that is maximally permissive of our choices for different structural features organizing these ontologies. I describe a local approach to these ontologies, sketch how it should be structurally constrained, and discuss its use using examples from bioscience. I recommend a local ontology to anyone who believes that biology must carry on in its current heterogeneous and partial ontological position, but who finds the consequences of adopting pluralism unacceptable.

Location is here intended in a broad sense, to include spatiotemporal regions, timescales and durations. I do not mean to exclude locations in an even broader sense, perhaps including contexts or domains of inquiry. However, context and inquiry, while sometimes ordered in a way similar to locations, include additional connotations related to the particular inquiry or relationships among the inquirers involved. To say that to each context (e.g. of inquiry) there corresponds an ontology, is just to say that different inquirers happen to hold different ontologies. That is plainly true, so prescribing it is of dubious value. At best, a view with this consequence is local epistemology, not local ontology, and at worst it is disciplinary pluralism. Pluralisms of this form can be problematic, for reasons explained in § 2, and do not resolve the problem of conflict between these ontologies. On the other hand, I take a local (and disunified) epistemology as a given.Footnote 7 A local ontology is intended as a way of reconciling some of the ontological conflicts that arise once the heterogeneity in our scientific inquiry is taken seriously.

I take this account to fit well with other local approaches to resolving theoretical disagreements. In the specific context of mechanism and processualism, location is already emphasized by proponents of both ontologies.Footnote 8 Moreover, a local approach preexists in the disordered ontology of Dupré (1993), the dappled ontology of Cartwright (1999), and the local epistemology of Longino and Lennon (1997). Local approaches have also been applied to resolve conflicts in logic (Goldblatt, 1984, Ch. 14; Simons, 2006), category theory (Bell, 1986), algebra (see Bergman, 2011, ch.8), physics (Jozsa, 1979, in Fourman 1979), and in the metaphysics of causation (Brunet, 2021). What these approaches share is a scepticism about universal or absolute criteria of theory choice and a shift of focus towards theories that account explicitly for local variation. The aim of this essay is to argue that an appreciation of local variation in processual and mechanistic features of the world is our best way to achieve tolerance and benefit from both.

2 Problems with pluralism

In this section I argue against two forms of pluralism about processualism and mechanism. These are (§ 2.1) what I will call unified pluralism: the view that one can assent to both processualism and mechanism together, as a unified theory. My argument against this form of pluralism is that it is inconsistent. Next (§ 2.2) I consider within-discipline pluralism: the view that a discipline can accommodate both processualism and mechanism by allowing individuals working within that discipline to assent to either of these views separately. So far as I can tell, this view is consistent. My argument against within-discipline pluralism is that it is unstable and unhelpful. Within the philosophy of science at least, this disciplinary plurality is already the status quo. Allowing these two flowers to bloom in the same garden has generated conflicts about which ontology is the more fundamental, rather than fostering synergies that allow philosophers of science to benefit from both ontologies. Of course, argument against these two specific forms of pluralism is not exhaustive of pluralist options; I examine these options in detail because they are common and initially plausible. In the following section (§ 3) I offer another way forward, arguing that we are better served by a local ontology.

2.1 Unified pluralism: processualism and mechanism are mutually inconsistent

I define a unified pluralism as follows:

Definition 1

A unified pluralism about ontologies T1 and T2 is the ontology committed to the union of T1 and T2, that is, the ontology T which is committed to everything in T1 and to everything in T2.

This section argues that unified pluralism about processualism and mechanism is inconsistent. This inconsistency is due to their different claims about the number of ontological primitives. Mechanism and processualism differ fundamentally in the number of ontological primitives admitted: processualism is a monism about processes and mechanism is a dualism about entities and activities. Since a (strict) monism and dualism are plainly contradictory ontological positions—there cannot be exactly one primitive ontological category and exactly two primitive ontological categories—it suffices to show that processualism and mechanism are indeed committed to monism and dualism, respectively. I begin by explaining why they adopt different primitives. I take this to show that we cannot be pluralists by adopting both ontologies simultaneously as individuals. Someone who claims to be both a mechanist and a processualist is contradicting themself.

Consider mechanism first. Machamer et al. (2000) state explicitly that their view is dualist, and claims that dualism is a fundamental part of mechanism can also be found in Glennan (2017, p.19-21). Mechanism is dualist in part because it is an attempt—not unlike this essay—to unite two ontologies within one. For Machamer et al. (2000) at least, mechanism is intended to unite processualism and substantivalism.

Both activities and entities must be included in an adequate ontic account of mechanisms. Our analysis of the concept of mechanism is explicitly dualist. We are attempting to capture the healthy philosophical intuitions underlying both substantivalist and process ontologies —MDC (2000, p.4)

Machamer et al. (2000) rightly note that there is a conflict between substantivalism and processualism. Much more contentiously, they claim that mechanism resolves this conflict. Their resolution involves a dualist ontology having two primitive notions, with entities evidently meant to capture but differ from substances and activities meant to capture but differ from processes. Moreover, mechanists identify specific problems with monism about entities (substantivalism) and monism about activities (process ontology).

Now consider processualism. It is a central tenet of processualism that all entities can be understood as dynamic processes. Dupré’s (2013) view of biological processualism is explicit on this point.

[I]n fact the entities we treat as things are typically very specifically stable, which is to say stabilized, stages in processes. The constant flow of material through a cell or an organism, a flow that includes many other elements that we are inclined to think of themselves as things... seems much better understood in terms of a hierarchy of processes than of things. —Dupré (2013, p.30)

Monism about process is not a feature unique to biological processualism. Perhaps the most direct statement of monism is in Rescher’s (1996) introduction to process metaphysics. This appears as the first of four competing contentions about metaphysics—the one that “represents process philosophy... in its stronger (Heraclitean)... version” (ibid. p.3). The contention is this: “Process has primacy over things. Substance is subordinate to process: Things are simply constellations of processes.” (ibid p.2). Rescher’s contention is a strongly monist reductive claim about the relationship between processualism and substance or thing-based ontologies. Seibt (2012 Sec.5) likewise claims that, despite other differences, all process philosophers share the view that “being is dynamic or that there are no concrete static entities”. With processes and without entities, or with entities being somehow reducible to processes, processualism is monist. Indeed, this monism about processes—or at very least a rejection of a dualism about substances and processes—is one of the key features of processualism that its advocates point to, locating this as far back as Heraclitus’ panta rei. I remain agnostic about the status of such a reduction, but maintain that it is part of the processual view.

I take this to show that processualism and mechanism have different ontological primitives and a different number of them. There cannot be exactly two ontological primitives (mechanism) if there is only one ontological primitive (processualism), so the attempt to hold a unified pluralism about these these ontologies is inconsistent. Perhaps both ontologies (or some best reconstruction of them) indeed require more primitives. E.g., perhaps they require relations. The contradiction remains nonetheless; it is a conflict between a different, but still unequal, number of primitives.

I respond to an objection before turning to another form of pluralism below. Ontologists of either sort may object that a unified pluralism about their ontology and the other is not inconsistent, the number of primitives is not really different, since the primitives of the other ontology may somehow be defined within their own. Indeed, ontologists of both sorts also support their ontology by claims about how the primitives of the other relate, are subordinate, or reduce to their own.Footnote 9 For example, Glennan (2017) argues that processes are temporally extended mechanisms. Whether we accept any of these relationships then affects whether we find anything problematic about adopting both of these ontologies together. I respond to this objection by showing that if we hold the reductive claims of both ontologies, as a unified pluralism would suggest we may, then this is also contradictory. In particular, we end up with a monism that directly contradicts the explicit dualism of mechanism.

Fig. 1
figure 1

Mechanist view of processes as activities

The figure (Fig. 1) above is meant to capture that Machamer et al. (2000) support their view by reductive claims that equate processes with activities while denying that entities can be reduced to processes. Specifically, they claim,

Substantivalists confine their attention to entities and properties, believing that it is possible to reduce talk of activities to talk of properties and their transitions... In contrast to substantivalists, process ontologists reify activities and attempt to reduce entities to processes. While process ontology does acknowledge the importance of active processes by taking them as fundamental ontological units, its program for entity reduction is problematic at best.—MDC (2000, p.5)

It is clear that they must adopt a view something like this, or else have mechanism collapse into processualism, or some other monism. The commitment to dualism in mechanism is not idle. In the mechanist ontology, both entities and activities are thought to be ineliminable parts of mechanisms, with neither reducible to the other. Entities alone are thought not ontologically adequate, as we can see from Machamer et al.’s (2000) criticism of substantivalism. Likewise, the mechanist ontology does not allow for activities alone; there are no free-floating activities in mechanisms. For instance, Glennan (2017, p.22) claims that “[t]here cannot be activities without entities.” An ontology consisting of only activities or only entities would not be dualist, and would accomplish entity reduction.Footnote 10 Moreover, they deny that entities are sorts of processes, as we can see from their criticism of the processualist program for entity reduction, which they attribute to process ontology. This leaves an ontology which is still essentially dualist, while attempting to incorporate processes as a sort of activity.

For their part, process ontologists following Rescher (1996) and Dupré (2013) do approach mechanism (and substantivalism) in a way that might be called entity reduction. As above, Rescher insists that substances are essentially just aggregates or “constellations” of processes. Dupré takes this further by describing entities as a particular sort of process: stabilized processes (as in Fig. 2). If we put both of these views together—the mechanist view that activities are sorts of processes, and the processualist view that entities are sorts of processes—then we end up again with a monism about processes (Fig. 3). This might be acceptable to some processualists interested in reducing mechanism to processes, but it is not a viable form of unified pluralism. That both ontologies together with their inter-ontological reductive claims end up leading to a monism directly contradicts the explicit dualism of mechanism, so this option is also inconsistent.

This is reason enough to believe that dualism and monism are non-optional aspects of the views of mechanists and processualists. If this is correct, then we cannot consistently adopt both ontologies as a package. Pluralism would have to be achieved by other means. Moreover, this also reaffirms that there is a still a conflict in our 21st century biological ontology. New mechanism cannot “capture the healthy philosophical intuitions underlying both substantivalist and process ontologies” and remain ontologically adequate, since its union with processualism is not consistent.Footnote 11 The following section considers another form of pluralism that might be hoped to both overcome this inconsistency and allow us to benefit from both ontologies.

Fig. 2
figure 2

Processulist view of entities as stabilized processes

2.2 Within-discipline pluralism: processualism and mechanism are antithetical

Since pluralism in the form of unification is an inconsistent option, it is reasonable to aim next for a sort of pluralism falling short of unification. The next most plausible option is to adopt a “within-discipline pluralism”. Garson (2020) introduced the concept of within-discipline pluralism about function concepts as the view that, “seeks out and emphasizes the plurality of function concepts within any branch of biology.” This form of pluralism is intended as an alternative to between-discipline pluralism about function concepts, i.e. the view that different function concepts are appropriate to different disciplines of biology. This view is far more permissive of theoretical heterogeneity among members of a discipline. It is helpful to borrow this pluralism concept while leaving behind its association with function concepts and considering instead a within-discipline pluralism about ontologies. With that purpose in mind, I define this sort of pluralism as follows:

Definition 2

A within-discipline pluralism about ontologies T1 and T2 allows that, for different individuals \(i_{1}\) and \(i_{2}\) within a discipline, \(i_{1}\) adopts T1 and \(i_{2}\) adopts T2.

On this account, instead of individuals adopting a unified pluralism, i.e. accepting both ontologies at once, we would allow different researchers to adopt different ontologies. In effect, our community would be split into a mechanist group and a processual group—we would have an ontologically heterogeneous Biology, as a discipline. In this section I argue against this option. My argument against this sort of pluralism is that it fails to give individuals a stable criterion to decide which ontology to adopt in any case, so is unhelpful in every case.Footnote 12

Fig. 3
figure 3

Putting both mechanistic and processual reductive claims together collapses to processualism

I assume that for any within-discipline pluralism we require a criterion to decide when one or the other view should be applied. We need such a criterion in this case because the application of both ontologies to the same case is, as argued above, inconsistent (§ 2.1). However, in the case of mechanism and processualism, we should also assume that we cannot make this choice by applying both ontologies and seeing which enjoys empirical adequacy. If one and not the other turned out to be empirically inadequate, then that would indeed settle the matter entirely: the adequate ontology would be preferable and we could set aside any worries about pluralism and tolerance. It is acceptable to be intolerant of empirically inadequate ontologies. However, I do not think we have any reason to assume that either ontology is overall empirically inadequate; both ontologies have been very successful at providing descriptions of phenomena from across the sciences, and both seem to be able to provide some description of any case.

If empirical adequacy cannot provide a criterion to choose between these ontologies, then how should we proceed as members of a pluralist discipline? Even if both ontologies are empirically adequate, might they still differ, in some cases, in relative goodness of fit? That is, might we adopt a within discipline pluralism that varies by individual according as that individual works in an area where one ontology is more empirically adequate? Supposing that these ontologies do differ in how well they fit given cases of interest, then one or the other ontology might also be an overall better fit in the following sense: one ontology may be best for several cases and may be able to cover any case where the other is judged best. By covering a case I mean the following.

Definition 3

An ontology T1 covers another ontology T2 in case X provided T1 is the best fit for another case or series of cases Y, and that Y in some sense includes X.

The idea here is of ‘covering’ by analogy with a cover of a set in topology—a collection of (covering) subsets the union of which contains the covered set.Footnote 13 Of course, there is some necessary ambiguity about what it takes to best fit a case, and what senses of inclusion between cases are admitted. However, we will shortly see examples of covering, which should help to fill in these ambiguities.

When one sort of ontologist argues that their ontology can cover a case accounted for by the other ontology, I will call this a covering argument. Indeed, ontologists of both sorts often provide covering arguments in support of their view. Below I argue that covering arguments and counterarguments are available to ontologists of both sorts, so do not provide a stable way of deciding on which ontology to apply in any given case. I take this to show that even if both ontologies often differ in their goodness of fit for given cases (indeed, even if they always differ), this still does not provide a stable way of choosing which ontological framework to apply in a within-disciplinary pluralist context—there will always be a covering argument available which speaks against your choice of ontology, whatever the choice.Footnote 14

It will first help to give some general description of covering arguments. Covering arguments have two flavours (Fig. 4): one in which a given first case is covered by a larger or more inclusive case containing the first as a sub-case, and another where the first case is covered by a set of cases the union of which contains the first. We will see examples of both in discussions about processualism and mechanism below. For an example of the first sort, it might be the case that a mechanistic ontology best fits the case of the events of gluconeogenesis (X)—that is, there might be a mechanism that best fits gluconeogenesis. Processualism might nonetheless cover gluconeogenesis if it turns out that the events of metabolism (Y) overall are best understood as a process, since metabolism includes gluconeogenesis. Which of these two flavours of covering argument we ascribe to a particular case indeed turns on how we choose to individuate cases. E.g., if we prefer to individuate metabolism as a series of overlapping metabolic processes, such as sugar metabolism (Y1), protein metabolism (Y2), and fatty acid metabolism (Y3) then we might instead see the covering of gluconeogenesis by metabolism as an argument of the second flavour. I separate them because I think it is helpful to recognize both as covering arguments.

If we adopt an ontology in some case only to find that the other covers it, then it is reasonable to switch to the covering ontology and the more inclusive case. Consider again the example of processualism covering mechanism via a covering argument that metabolism is best seen as processual. If we accept this covering argument, it would be reasonable to switch to processualism and embed our analysis of gluconeogenesis as a sub-case of metabolism. However, similarly to choosing ontologies on the basis of empirical adequacy, we have reason to believe that both mechanism and processualism can cover the other in any case. In the case of mechanism and processualism, this happens because ontologists of both sorts have reason to claim that their ontology can cover more inclusive phenomena that occur over different durations—any practitioner within a discipline where both ontologies are freely available will find herself marching on metaphysical Penrose stairs, perpetually ascending but getting nowhere. This point will occur again in the following section (§ 4) where I suggest a way of handling it that is better than within-discipline pluralism.Footnote 15

Though not in name, covering arguments occur frequently in discussions of mechanism and processualism. I begin by discussing the case of processualism. Part of the processual critique of mechanism is that the entities appealed to in mechanistic explanations of some phenomenon—in biology especially—often do not persist over the duration of the phenomenon Dupré (2017); Nicholson and Dupré (2018). Moreover, in many cases it seems that there is always a longer duration over which some entity does not persist. Entities have origins and destructions, so take any duration overlapping one or both of these. Moreover, processualists take this as a criticism of mechanism because they believe—rightly, so far as I can tell—that a given mechanism for a phenomenon requires persisting entities, while processes do not. If we can always choose some such duration over which entities do not persist, and we take this to imply that a processual ontology best describes the (origin or destruction) events involved, then processualism arguably covers mechanism.

Consider an enzyme catalysis. The processualist covering argument in this case is based on the fact that enzymes sometimes do not persist over the course of the reactions they catalyze. If an entity does not persist over a duration and a mechanism that refers to that entity is given to explain a phenomenon over that duration, then the mechanism is arguably not the best fit in that case. However, since processualism does not require persisting stable entities, it seems to fit better.

Fig. 4
figure 4

Two kinds of covering argument. (left) X is covered by a case Y which is more inclusive, and (right) X is covered by a series of cases Y1-Y3, the union of which is more inclusive

Fig. 5
figure 5

A form of covering argument where a mechanism (solid arrows) occurring during some duration is covered by a process occurring over a longer duration, containing a process of production and/or decay of an entity (dashed arrows)

We can state this sort of convering argument generally, at the level of abstract characterizations of mechanisms with covering processes. Consider an abstract mechanism where a component A is converted into B by an enzyme E, perhaps an enzyme catalysis of a single step of glycolysis. Supposing E persists over some duration, then the activity of E on A can explain the production of B. Both process philosopher and mechanist should agree on this, so constrained. However, we can assume that there are longer durations over which E does not persist, is not stable, degrades, or is otherwise not stably acting on A. This constitutes a covering argument of the first sort discussed above, where the processualist provides some covering duration over which entity E is produced or degraded, as in the following Fig. 5. This (and Fig. 6) are intended to depict, alongside the usual abstract arrow depictions of reaction events, which events cover which others (via bracketing).

Fig. 6
figure 6

A form of covering counterargument where a series of mechanisms, including a production mechanism \(E^{\circ } \xrightarrow {X} E\) and degradation mechanism \(E \xrightarrow {Y} E'\), are proposed to cover a process of enzyme production and degradation, itself covering a mechanism of enzyme catalysis

Over the longer duration, the processualist can claim that the activity of E on A cannot correctly explain the production of B. Indeed, during those periods when E is not present or stably acting on A, a mechanistic representation which indicates that it is present and acting on A is incorrect.Footnote 16 To processualists, these consequences of considering longer durations are taken to support the idea that the “entities” appealed to in mechanisms are essentially processes (see Nicholson & Dupré, 2018, p.15-22 on metabolic turnover), thus reinforcing their monism. Dupré and Nicholson (2018, pg.15) say that metabolic turnover is “one of the strongest motivations for adopting a process ontology” and that “from a metabolic perspective, it is simply a matter of fact that, in an organism, everything flows” (ibid, p.17).

However, this form of covering argument does not uniquely favour processualism. There is a symmetry in the availability of covering arguments; mechanists can supply covering counterarguments. They can do this by proposing mechanisms over longer durations to explain phenomena that fit best within processualism over a shorter duration. Consider an enzyme catalyzed reaction again. Plausibly, processualists are correct to point out that over timescales including enzyme degradation processes, that a mechanism which treats an enzyme as an entity fails to capture something important about the catalysis (i.e., it would get the concentration-dependent rates wrong). Indeed, in the case of glycolysis in common yeast, Daran-Lapujade et al. (2007) found that metabolic flux—the enzymatically controlled change in concentration of glycolytic intermediates—is controlled predominantly by regulation of production and degradation of glycolytic enzymes. However, that has not prevented a mechanistic account of enzyme degradation.

In the framework of covering arguments, once a processualist has provided a case X covering a mechanism and including enzymatic turnover processes, a mechanist can provide another series of cases Y1-3, where Y1 is a mechanism of protein synthesis, Y2 is the initially covered mechanism, and Y3 is a mechanism of protein degradation. Moreover, the union of Y1-3 includes the case X identified, so covers it. Biochemists do this regularly. In the case of yeast again, many such mechanisms are reviewed in Zattas and Hochstrasser (2015). This constitutes a covering argument of the second sort discussed above, where a mechanist provides some covering durations Y1-3, as pictured in Fig. 6.

Given a case for which a process ontology seems the best fit (e.g., the sum-total of metabolic processes in a given organism), a mechanist can respond by arguing that, in each of a series of shorter durations the union of which includes the case at hand (e.g. overlapping sections of each metabolic event), that a mechanism is best, so that mechanism covers processualism after all. This is the sort of argument provided when mechanists respond to a non-mechanistic (processual) description of some phenomena by reporting on underlying or lower-level mechanisms, e.g., Craver’s (2013) definition of multi-level mechanisms. Nonetheless, symmetrically, a processualist has the same form of argument available to them again. A processualist can respond to a mechanistic description of some case by considering their own series of cases including it and providing a fitting process for each—perhaps by stepping down again from molecular biology to series involving cases from more fundamental physical phenomena, for which plausible processual accounts have been given (see references in Seibt, 2012).Footnote 17 I see no in principle reason that this chain of covering arguments and covering counterarguments should end.

Taken together, that these forms of argument are available to ontologists of both sorts makes the within-discipline approach to pluralism unstable. Each time a mechanism is proposed, a processualist may counter with a covering process, and vice versa. This puts us in a position where it is justifiable to continually switch between ontological frameworks with no apparent point of settling on one or the other. Switching between ontological frameworks for different cases is not problematic in itself; in the following section I argue that this is precisely what we should be willing to do. However, in the context of within-discipline pluralism, this instability presents a problem for anyone wishing to report on any single case within either ontological framework. If they choose one (as a within-discipline pluralism would suggest they freely might), they are open to criticism that they have not chosen the best framework on the basis of a covering argument. As above, choosing both is not an option. Someone may defend their choice of framework by provision of a covering counterargument, but then so may their opponent. A committed within-discipline pluralist may of course bite the bullet of instability. However, in the final section below I argue that there is an overall better and stable option available.

It is worthwhile to emphasize two points before continuing with a different sort of resolution in the final section. First, this is not intended as an argument against either ontology, and I do not expect covering arguments to convert any ontologist who is firmly committed. Indeed, no ontologist should be converted by a covering argument, so long as they are perpetually willing to provide their own covering counterarguments. This section is addressed rather to the reader interested in adopting a pluralist approach but unsure of how to do so. Second, in this within-discipline pluralist context, faced with parallel covering arguments, both mechanists and processualists can respond to one another with the same form of argument. Both ontologists can claim that there is some longue durée over which their ontology is favourable, but their opponents can also always claim the existence of some plus longue durée. I do not take this to be a sign of any defect or weakness in either or both ontologies. Rather this is a problematic consequence of having two sufficiently powerful ontologies to choose from; the incompatible choices are too good to comfortably coexist. Put another way, the within-discipline pluralist is in an ontological version of the fable of The Fox and the Cat, this puts the pluralist in the position of the fox, having many ontologies to choose from and no way to settle on a choice when pressed about a particular case.Footnote 18 This antithetical situation unfortunately leaves us with two powerful ontologies but without a pluralistic way to both consistently and stably benefit from the powers of both. In the following section I argue for another way forward.

3 Reconciliation using a local ontology

How pluralism is understood—whether, for instance, it affirms radical ontological or epistemological heterogeneity or merely the diversity of mechanisms in nature—varies from thinker to thinker and topic to topic.—Kellert, Longino and Waters (2006, p.vii)

The view advocated in this section is, in a slogan: Mechanisms here, processes there, and the two never mix, like oil and water. Whether a view like this seems unpalatable turns importantly on one’s view of the goals of fundamental metaphysics. If metaphysics is supposed to provide a homogeneous collection of ontological primitives, then the view advocated in this section will taste bad. To a monist metaphysician of either sort considered here, it cannot really be that there are both mechanisms and processes in the world. Or rather, they might accept that there are both, but only insofar as one can, ultimately, be analysed into the other. For instance, a mechanist might insist that there are processes, though analyse these as a sort of mechanism.Footnote 19 The view intended here is not like this. It is intended as a reconciliation that affirms a radical ontological heterogeneity. There is not only a diversity of mechanisms and processes in nature, there is also a diversity in whether parts of nature are mechanistic or processual.

Something important is captured by the pluralistic approach discussed in the section above: that mechanistic and processual ontologies are ultimately to be applied to individual cases, to specific phenomena, and thus should be relativized to these phenomena somehow. This section argues for a view about why this is so. Instead of a pluralist ontology of either sort discussed above, I suggest something similar: local ontology. By a local ontology I mean the following general sort of metatheoretical situation. Given a set of equally good though conflicting ontologies, a local ontology for these is itself an ontology, constrained by two additional claims (§ 1). First, that it varies by location which of these ontologies holds, and second that none is presumed to hold most fundamentally or globally. In the case at hand, I will discuss locations in terms of a duration surrounding a given phenomenon.

There is no reason in principle that a local ontology should be only about two ontologies. Perhaps a triad of mechanism, processualism and trope theory would be better. Indeed, some advocates of all three of those ontologies have proposed that their their primitives should be individuated by location (see, respectively, Glennan, 2017; Dupré, 2013; Schaffer, 2001). An n-ary ontology is just as conceivable. Though for the present discussion I consider only a local ontology for the pair of mechanism and processualism.

A local ontology for mechanism and processualism is a view that claims that the truth of the ontological claims of mechanism and processualism vary by chosen duration and phenomena; over some durations the world machinates and over others it flows. Moreover, in a rephrasing of Rescher’s (1996) first contention about metaphysics (§ 2), a local ontology also advances the view that, over varying durations, neither process nor mechanism has primacy over the other, and neither is subordinate to the other. We cannot hold processualism and mechanism as a package, but we can still view things as patchwork constellation of processes and mechanisms. This section will describe this view and argue that it is a better way to benefit from the ontologies of mechanism and processualism.

I draw the inspiration for this ontological view from topos and sheaf theory (see Bell, 1986; Lovering, 2010). When faced with incompatible theoretical constructs—e.g., set theories where the continuum hypothesis holds and those where it does not—it makes sense to adopt a strategy of localizing our judgements about these constructs. We can localize our analysis to a model or, further, to its submodels. The same is true when dealing with incompatible properties. It is often difficult to assess (or uninteresting) whether some property—e.g., the continuity of a function—holds globally over a space. The more tractable, and interesting question is whether some property holds at all, and if so where it holds. In these kinds of cases, local analyses serve in the manner of a divide and conquer strategy: incompatible alternatives are assigned separately to different locations, so incompatibilities do not arise anywhere (nor globally, in particular). I suggest a similar approach in the service of our incompatible scientific ontologies, where our ontological analyses are localized to durations.

By a duration I mean a concrete, particular, bounded, uninterrupted span of time during which some phenomenon occurs. We should not expect any particularly well-behaved association between non-durations and ontologies. Here is a series of examples and non-examples. “A span of five minutes” is a span of time, but is not a duration because (among much else) it is not concrete; “The span of five minutes from 12:00-12:05” is not a duration because it is not particular; “From 1:00 pm on today’s date and indefinitely onwards” is not a duration because not bounded; “All of today except 12:00” is not a duration since it is interrupted; “from 1:00-2:00 on today’s date” is not a duration because no occurring phenomena is given. However, “the party from 1:00-2:00 on today’s date” is a duration; “the span of time from the synthesis to the degradation of some (specific) molecule” is also a duration; your life from birth up to the present is a duration; so, presumably, is your life from the present to your death. From the point of view of metaphysics concerned with everyday science, it is irrelevant whether there is an empty duration or a universal one. I take the terms ‘spatiotemporal location’, ‘timescale’ and ‘duration’ as nearly synonymous and use ‘location’ and ‘duration’ in the remainder.

Though of similar use, I take a local ontology to be distinct from theoretical pluralism, i.e., views about how ontologies as theories are best applied (as in § 2-3). Theoretical pluralisms are views about theories; local ontology is a view about what there is. It is an ontology where mechanism or processualism are two modes or ways of being that a given duration of a phenomenon can have. On this account, given a duration and a phenomenon, that duration is either a processual duration or a mechanistic one. This might be difficult to decide in some cases and easier in others.Footnote 20 Nonetheless it enjoys theoretical virtues not had by either pluralism considered above. Firstly, it provides an ontology that is consistent, or at least is not inconsistent on grounds of an internal contradiction between monism and dualism (§ 2)—of course, there might be some other hidden inconsistency to discover. Secondly, it provides a stable way of drawing understanding by framing phenomena within both mechanistic and processual modes. At least, it provides a way that is not unstable in the way that within-discipline pluralism is (§ 3), since here covering arguments have no force (more below).

We may think of a local ontology as entailing an assignment of ontologies to durations. We should of course like to know how to practically go about assigning ontologies to durations. However, as discussed above, the plausible assumption that both ontologies are empirically adequate places a strong epistemic limitation on our ability to do so. We cannot, for instance, assign ontologies to durations by investigating to see which best fits or explains the contained phenomena since, as argued in § 3, both can claim to do this. Instead we require some other principle to partition durations between ontologies. For the purpose of this essay I remain agnostic about how precisely to do this (though discuss some positive and negative implications of a local approach to ontology choice below). This is not a problem in any way unique to a local ontological approach. The problem of deciding when to apply one member of a pair of empirically equivalent scientific theories is a common one. The problem of deciding where to apply one member of a pair of equally empirically adequate metaphysical theories is similar.

The assignment of ontologies to locations should have a structure that allows us to have a picture of the world that is pieced together, location by location. If we assume we are restricting our attention to a given phenomenon (or spatial location), then we may picture this arrangement, as in the figures below (Figs. 7-8), by containing ontologies over durations on a timeline. Minimally, we should assume at least that every duration is assigned an ontology and that no duration is assigned both ontologies (on pain of inconsistency).Footnote 21

To see what sort of overall structural features a local ontology must have to avoid the problems associated with covering arguments, we can consider again how mechanists and processualists can use covering arguments (§ 3) to argue in favour of their ontology globally. Must a local ontology prescribe anything about which ontology holds “globally”, over “all” durations or as we indefinitely extend our durations outward in both directions? For example, is this ontology just a concession that, although mechanisms might be successful over some small local durations, globally we are admitting an ontology of radical flux in the long term?Footnote 22 No. Nor must it presume that the mechanistic mode holds globally. To see this, consider the following two principles that a processualist and mechanist might adopt in attempt to favour their own ontology globally.

  1. 1.

    Global Processes: For every phenomenon over a duration where a mechanist ontology is appropriate, there is a longer duration over which a process ontology is required.

  2. 2.

    Global Mechanisms: For every phenomenon over a duration where a process ontology is appropriate, there is a longer duration over which a mechanist ontology is required.

I take it that processualists adopt the former and reject the latter, and vice versa for mechanists. However, within a local ontology, both principles are mutually consistent provided we do not assume the existence of a most global duration, i.e. longest timescale. If there is no global duration, at which we would be obliged to stop considering longer durations, then there is no reason that oscillation between ontologies need cease. This situation, where different ontologies may be appropriately nested within one another, is roughly captured in Fig. 7.

Fig. 7
figure 7

Nested ontologies

What other local arrangements of ontologies should be allowed, and should any be disallowed? We at least have reason to allow (1) alternation between ontologies over a given duration, (2) overlap between ontologies on the same segment of a duration (Fig. 8), and (3) non-terminating descending chains (also Fig. 7). For an instance of 1, we might alternate between a mechanistic and processual ontology, living in a mechanistic duration before our entities degrade and until new entities are formed, then shift to a processual one during intermittent periods of flux, then return to a mechanistic ontology when stability re-emerges. For an instance of 2, we might consider some phenomena, such as the degradation of a given tissue, which is an unstable or destabilized process over a substantial duration. Nonetheless, when considering another duration overlapping this process, we might successfully individuate certain molecular biological mechanisms that include entities and activities that persist over this duration.

We should allow 3, non-terminating descending chains, since terminating descending chains are used in covering arguments. If all series of containing sub-durations terminated with some particular ontology—say processualism—then that ontology would have a strong claim to being the most fundamental. It is easy to provide examples of short alternating chains, although knowledge of events at smaller scales or short durations drops off with orders of magnitude. For a short chain, consider any duration overlapping the decay process of an entity, e.g., the death of an organism. We may then consider a shorter duration not overlapping this decay process at the tail-end, and we may assume for example that this duration is mechanistic. Nonetheless, we might consider some shorter-lived phenomena, such as the decay-process of a given cell within the focal organism, and again consider a duration just overlapping this. Again, we are free to consider a shorter duration, not overlapping this tail-end decay process, and we might then again find ourselves in a mechanistic ontology. If we want to block this kind of reductionist argument from settling on a given fundamental ontology, then we should allow that such chains do not terminate.

Fig. 8
figure 8

Alternation and overlap

We should allow ourselves at least sufficient freedom in proposed local ontologies to block the force of covering arguments. However, I do not have any strong criteria or intuitions about what sort of structural features are necessary for a local ontology generally, and it is not in the Quinean spirit of tolerance to bar structure to ontologies outright. Nonetheless, there are sorts of local structure that would trivialize the localization of ontologies to durations, and other structures that are prima facie unnecessarily complex or fanciful.

For an example of a trivializing structure, it would not be much of a reconciliation if we assumed that there is some most inclusive or highly inclusive duration relative to which one ontology held, past which the same ontology always held, and the other did not. That would just be monism (as opposed to pluralism) from some duration and up. Similarly, if we assumed that each duration assigned to one ontology could always be broken up into a series of contiguous parts, each of which was assigned the other ontology, we would essentially be excluding the former piecemeal. Local structures like this are local in name only. For an example of unnecessary structure, since my purpose in providing a local ontology is to allow both ontologies to coexist within biological science, it would be unfortunate for the structure of this ontology to be too fine for empirical investigation—such as by having oscillations between ontologies over non-differentiable durations, or only over Planck time. Perhaps oscillations between ontological frameworks over Planck seconds would be useful for some pairs of incompatible ontologies, but this structure is too fine to capture processual and mechanistic features of organisms.

A local ontology does contrast starkly with monism about either mechanism or processualism, but I do not think it is too great a departure from the perspective of either. I recommend this solution in part because the means of adopting it already exist, in embryo, within both ontologies. Both mechanists and processualists accept that a fundamental part of scientific explanations of a phenomenon is identifying its timescale or duration. Consider Machamer et al. (2000) and Dupré (2017),

Entities often must be appropriately located, structured, and oriented, and the activities in which they engage must have a temporal order, rate, and duration (p.3) ... Activities, likewise, may be identified and individuated by their spatiotemporal location. They also may be individuated by their rate, duration, types of entities and types of properties that engage in them (p.5). —MDC (2000)

[T]he constituents of a biological mechanism are themselves dynamic and more or less transient entities. Mechanistic explanations will be successful only to the extent that the constituents identified are sufficiently stable on the timescale of the phenomenon under investigation. —Dupré (2017, p.3)

Mechanists, for their part, already accept that mechanisms and their constituents are individuated relative to durations. I am simply insisting that they can only be individuated relative to a given duration—since they may not even exist relative to others. Moreover, Glennan (2017, p.41) notes that “we will get different and incompatible decompositions of mechanistic processes... depending on what phenomenon we are seeking to explain”. I agree and would add only that the incompatibility can run ontologically deeper: we can get different and incompatible ontologies entirely, depending on the duration over which we choose to explain phenomena. On the processual side, I take Dupré (2017) to be providing a local concession to mechanism, that mechanistic explanations can be successful only over some durations. I would only add that, on a local ontology, the same is true of processualism.

Mechanism and processualism, as scientific ontologies, were intended in part to capture the explanatory activities of science, and these activities are local to the duration surrounding particular phenomena of investigation. Glennan (2017, p.33) highlights the local explanatory character of mechanistic explanations, that they explain causation in particular contexts over particular durations. The view argued for here is just slightly stronger. It is not just that applying an ontology requires attention to the “timescale of the phenomenon under investigation” or that processes and mechanisms may be “individuated by their spatiotemporal location”; rather, location contingently determines the correct ontology.

I do not have any strong intuitions about whether the correct ontology is caused, grounded, or necessitated somehow by locations. What I have in mind is more contingent: just that different ontologies happen to hold over different durations. This raises an epistemological problem of determining, for a fixed location or duration, which ontology is correct. In making these choices, I think we should lean heavily on Quinean tolerance—we should be much more tolerant than would be appropriate if either mechanism or processualism were globally true. Ontology choice, like theory choice, is underdetermined by our evidence. However, we do not need to fully capitulate in the face of such a choice; there are negative and positive things that a local ontology entails about choosing the ontology given a location and duration. The remainder of this section discusses these.

On the negative side, a local ontology entails that, in all likelihood, a monist ontologist who recommends the same ontology universally, for every case at hand, is bound to be wrong at least some of the time. Indeed, monist ontologists are not in much, if any, better position when it comes to determining an ontology for any particular case; they recommend the same ontology in every case, so succeed in determining a specific ontology for a particular case in the same way that a constant function succeeds in determining an output for a particular input. Also on the negative side, a monist who universally uses covering arguments to negotiate in favour of their preferred ontology is mistaken about the significance of those arguments. If there are arguments that generally favour one over the other ontology, covering arguments are not among them, since covering counterarguments lead to stalemate. However, on the positive side, a local ontology tells us that our choice of ontology for a given duration and location need not always be constrained by our other choices for other cases. On a local ontology, e.g., there is no contradiction in saying that glycolysis is a mechanism while metabolism is a process. So, even fixing one set of ontology choices does not necessarily fix a choice for a new case at hand. This added freedom of choice, together with the avoidance of the problems with pluralism, is the primary benefit of adopting a local ontology.

Before concluding I return to the problems with pluralism identified in § 2, and how a local ontology avoids them. First, a local account resolves the contradiction (§ 2.1) that emerges from an attempt to straightforwardly unite the claims of mechanism and processualism. It does so because it is a rejection of both dualism and monism. Not absolutely, but over different durations, we may adopt an ontology consisting of entities, activities, and processes. Different medium-sized dry goods happen to exist in different locations. In advocating a local ontology I am claiming that the same is true of metaphysically primitive goods as well. Globally, this gives us an ontology with three primitives (processes, entities and activities), though one that is locally (exclusively) either monist or dualist. No contradiction of the form “There is exactly one primitive and there are exactly two primitives” erupts from a local ontology.

Of course, if we attempt to hold, universally, all the mechanistic and processual claims over all durations, i.e., we hold explicit dualism and monism over all durations, then we end up back with the contradiction we started with. That is a good reason not to hold those universal claims. As such, I understand, this relativization may be dissatisfying to both dualist and monist. However, resolving a contradiction between two of our major scientific ontologies is a significant payoff for rejecting these fixed ontological assumptions. If we want to be tolerant towards both mechanism and processualism or, more ambitiously, to unite them within a common framework, then we require some way of overcoming the inconsistency in their primitive notions. The reconciliation offered here at least overcomes this inconsistency.

How does this resolve the instability problem identified in § 2.2? It blocks the conclusion that the (absolute) number of primitives can be decided by moving to a longer duration. Within a local ontology for mechanism and processualism, it is no victory for a processual ontology nor loss for mechanistic ontology that the former can cover the latter over some durations. At least, it is no further loss, above and beyond the loss that comes with acceptance of a much more strongly local constraint on the explanatory purchase of the mechanistic ontology. That a given part of some phenomenon can be an entity with activities over some duration but a process in flux over a longer duration is precisely what a local ontology entails. Similarly, that unstable processes might nonetheless be constituents of (stable) entities over the longer term is no loss for processualism. Again, at least no more a loss than accepting a local ontological constraint on the explanatory purchase of processualism. There is no global metaphysical payoff to marshalling evidence to support covering arguments, so covering arguments justify no instability.

4 Conclusion

Though the problem at hand is framed at the level of scientific ontology, the best independent motivation for adopting this resolution is epistemic. Science is a local epistemic activity, so our scientific ontology should also be a local ontological activity. Science generates knowledge by activities that are necessarily conducted with phenomena at particular places, durations and timescales, so the most we may know about which ontological framework obtains is localized similarly. Moreover, a good explanation for why our analyses seem to capture different phenomena over different durations—Woese’s conundrum in his New Biology—is that those durations are ontologically different.

Assuming that ontology varies locally comes with the benefit of allowing us a freer and consistent interplay of mechanistic and processual notions. However, it comes at the cost of rejecting the fundamentality of both mechanism and processualism. It also entails its own peculiar epistemic limitations. As Longino and Lennon (1997) noted about the position of feminist epistemology within any local epistemological background, from a local perspective, those working only under single ontology must accept that “inquiry conducted under its auspices will not reveal a total, but an incomplete, picture of reality” (Longino & Lennon, 1997, p.33). The most discomforting consequence of adopting a local ontology is that it explains why none of our representations of some phenomena seem to capture them in their entirety. Perhaps this is especially true for our incomplete picture of organisms; they are mechanistically and processually heterogeneous.

However, as a happy corollary, since our ontology is relativized to particular durations of phenomena, anyone wanting to confine their inquiry to some particular phenomenon no longer needs to worry about objections (such as the covering arguments examined in § 2.2) to their chosen ontology that are lodged in terms of other phenomena over other durations. If we find a molecular phenomenon for which we believe a mechanism is ontologically adequate, we need not worry about the turnover rates of other phenomena, of other potential non-mechanisms. Likewise, an organism-level biological process ontologist need not worry about the mechanistic nature of the molecular realm any more than they worry about cosmological mechanisms. Researchers whose work is confined to particular locations can also locally confine their choice of ontology. Woese need not worry about absolute metaphysical theory choice when leaving the eddies in his stream and returning to mechanistic benchwork.

I think this way of joining these philosophies preserves the largest part of their ontological commitments while blocking the problems that erupt from holding a pluralism about them. If so, then it should be at least more acceptable to ontologists of either suit than abandoning their own view or accepting the other. The difficult part for either sort of ontologist is not deciding whether a local ontology allows them to benefit from some aspects of the ontology of the other—it is easy to see that it can do this—but whether this benefit is worth the price of abandoning the universality and fundamentality of their view. I think it is, but I have no argument that the same benefits cannot be had more cheaply elsewhere, and in a way that allowed to each their own fundamental ontology. I have argued here that a local ontology is one way to reconcile the differences between mechanism and processualism, and that the payoff is great enough to choose this approach over pluralism.