Resilience as a Unifying Concept

Abstract

In sustainability research and elsewhere, the notion of resilience is attracting growing interest and causing heated debate. Those focusing on resilience often emphasize its potential to bridge, integrate, and unify disciplines. This article attempts to evaluate these claims. Resilience is investigated as it appears in several fields, including materials science, psychology, ecology, and sustainability science. It is argued that two different concepts of resilience are in play: one local, the other global. The former refers to the ability to return to some reference state after a disturbance, the latter the maintenance of some property during a disturbance. An implication of this analysis is that the various uses of the resilience concept are more closely related than has been previously been suggested. Furthermore, it is argued that there is a preference towards using highly abstract versions of the concept. This explains the apparent context insensitivity of the concept, but presents a problem for those hoping to establish a research programme based on it.

1. Introduction

The notion of resilience is being used increasingly in a number of scientific and non-scientific contexts: archaeology, national defence, ecology, psychology, and sustainability science, to mention just a few (see Rutter 1993; Almedom and Glandon 2007; Bonanno et al. 2007; Walker and Cooper 2011; Parker and Hackett 2012; Weiberg 2012). The precise meaning being attached to this notion, as well as its usefulness have, however, been fiercely disputed (Davidson 2010; Hornborg 2013). In the literature, and especially in work on sustainability, resilience often appears in contexts where discipline bridging and integration are central themes. It is an essential part of the framework proposed by, for example, Lance H. Gunderson, Crawford S. Holling, and others in the influential collection, Panarchy (Gunderson and Holling 2002). The authors seek to develop a general theory of change and express concerns that ‘approaches’ in which resilience has no role are partial in the sense that they ‘are too simple and lack an integrative framework that bridges disciplines and scales’ (Holling, Gunderson, and Ludwig 2002, 8). They go on to claim that one way to put in place ‘robust foundations for sustainable decision-making’ is through the ‘search for integrative theories that combine disciplinary strengths while filling disciplinary gaps’ (Holling, Gunderson, and Ludwig 2002, 8). Others are more explicit. Charles Perrings, for instance, notes that the concept has broad appeal for both natural and social sciences:

While the notion of system resilience has its roots in ecology, it is concerned with something that is common to any stochastic evolutionary system—the effect of the stability domain structure on the system's dynamics. These matters are currently attracting attention in a variety of different fields within economics, and in a variety of different disciplines. (Perrings 1998, 511)

More recently, in another influential volume, Gunderson and Lowell Pritchard treat resilience as a ‘unifying concept in both ecological and social systems’ (Gunderson and Pritchard 2002, xxi). For sustainability scientists the concept has particular appeal because it promises to connect social and natural science disciplines—an integrative achievement that some see as a prerequisite to the solution of the problems of sustainability (Kates et al. 2001; Kates 2011).

In this article, the focus is on two interrelated questions. One concerns how the concept of resilience can be used to connect disciplines, and the other concerns what might come of this. Ultimately, the interest here is in whether, for instance, the concept of resilience provides a sound basis for interdisciplinary research. In approaching this issue, however, a cross-disciplinary analysis of various resilience concepts is provided.

The article is structured as follows. Section 2 provides an overview of a range of definitions of resilience together with some background to the concept. In section 3, two conceptual cores that most forms of resilience revolve around are presented: local and global resilience. It is then argued that it is primarily just one of these, global resilience, that is associated with claims of integration and discipline bridging. Section 4 contains arguments for the view that the apparently wide applicability of resilience can only partly be explained by treating it either as a metaphor or a boundary object (two suggestions that have been made in the literature). An overlooked and important feature of the concept of resilience is instead that it is highly abstract. Context insensitivity of the sort accomplished through abstraction comes at a price, since the unification that ensues is often relatively weak.

Finally, section 5 contains a discussion of unification in which the abstract character of resilience is very much in the foreground. It is argued that resilience, in either of the senses presented, is multiply realizable, and that this feature has consequences for how resilience should be studied empirically.

2. The Many Concepts of Resilience

According to the Web of Science, the number of published works with the term ‘resilience’ in their titles has been growing steadily since the early 1980s. In 1993, 60 papers were published; in 2013, close to 800. In part this increase reflects uptake of the concept in various fields and disciplines, including the environmental sciences and ecology, sociology, anthropology, history, polymer science, urology, urban studies, materials science, and so on. The 10 publications with ‘resilience’ in their titles that appeared three decades ago, in 1983, were strongly focused on either ecology and biology or materials science, with one exception (Morrison 1983). More recently the term has found its way into the reports and working papers of highly influential institutions such as the World Bank (see e.g. World Bank 2012) and the IPCC (see e.g. IPCC 2007, 4.6.1; IPCC 2013).

Looking further back, ‘resilience’ figures in the titles of scientific publications as early as the second decade of the twentieth century. Materials science, especially textile research, is prominent there.¹ The general idea is that resilience is the ability of something to return to some reference state (a particular shape, for instance) after a disturbance of some sort. It might, for example, be the ability of a yarn to return to its previous length after being stretched with a weight (Hoffman 1948, 141–142).

From the 1970s until quite recently two fields adopting resilience stand out: psychology and ecology. In both disciplines during this period the concept of resilience was the subject of protracted discussion. In psychology, and particularly child and adolescent psychology, resilience came to replace the problematic concept of invulnerability.² Actually psychologists have used the term in several senses, including the following three:

1. Meeting developmental goals in spite of adversity.³

2. Sustained competence under stress.⁴

3. The ability to recover following trauma.⁵

The general aim has often—although not always—been to find so-called protective factors, or protective mechanisms, which account for the resilience of individuals (see Rutter 1987).

The language of ‘resilience’ in ecology should be understood in the context of the debate over the stability-diversity thesis, i.e. the idea that diversity and stability are positively covariant. This thesis was famously defended by influential ecologists such as Robert MacArthur (1955) and Charles Elton (1958), and it was received opinion in ecology throughout the 1950s and 1960s.⁶ Within the terminological framework that then emerged ‘resilience’ is a common term. It appears alongside terms like ‘resistance’, ‘persistence’, ‘constancy’, ‘hysteresis’, and ‘elasticity’. Volker Grimm and Christian Wissel (1997) list 70 different stability related terms and a staggering 163 definitions. The semantic diversity here is considerable. One can nevertheless discern two concepts of resilience of special significance. First, in a very common interpretation the term ‘resilience’ denotes the ‘returning to the reference state (or dynamic) after a temporary disturbance’ (Grimm and Wissel 1997, 325). This is how Stuart Pimm (1984) used the term in his influential paper on stability. This notion being pressed into service here is sometimes referred to by other terms. Kirstin Schrader-Frachette and Earl McCoy (1993, 33) label it ‘dynamic balance’ and Holling calls it ‘stability’ (Holling 1973), and then in later work ‘engineering resilience’ (Holling 1996; Holling and Gunderson 2002). In this article, the term ‘local resilience’ is used to indicate this ability of a system to return to a reference state (see section 3.1).

The other significant concept of resilience is often traced back to Holling (1973)—which, incidentally, is also the most cited paper with ‘resilience’ in the title by a wide margin.⁷ Holling makes a distinction between stability and resilience, regarding the latter as follows:

Resilience determines the persistence of relationships within a system and is a measure of the ability of these systems to absorb changes of state variables, driving variables, and parameters, and still persist. In this definition resilience is the property of the system and persistence or probability of extinction is the result. (Holling 1973, 17)

On this conception resilience is not about returning to a reference state, but rather a kind of buffer within a system. It is a margin that allows the system to retain certain structural features when perturbed. For Holling resilience is a magnitude—ideally one that may be measured—that tells us how large a disturbance a system can withstand without being pushed on to a trajectory where the system will, through its own dynamics, become extinct. This notion of resilience is common in the work of ecologists, although it is not always associated with the term ‘resilience’. Gordon Orians defines inertia in a similar way as the ‘ability of a system to resist external perturbations’ (Orians 1975, 141). Schrader-Frachette and McCoy (1993, 33ff) prefer the term ‘persistence’, and Grimm and Wissel (1997) associates Holling's resilience with the term ‘domain of attraction’.

2.1. Resilience Crossing Boundaries

Holling's resilience (as it were) is of particular interest in view of its impact outside ecology. First, it has drawn the interest of researchers in other fields, such as archaeology (Weiberg 2012), sociology (Davidson 2010), community psychology (Norris et al. 2008), and social medicine (Almedom and Glandon 2007). Second, promoters of the concept, and most prominently Holling himself, have been arguing explicitly that it should be applied in more fields. The idea is that resilience can capture a dynamic property not just of ecosystems, but others as well, including the social-ecological, the economic, and the social (Gunderson and Holling 2002; Adger et al. 2005; Folke et al. 2010).

Resilience also appeals to those interested in sustainability and sustainable development. It offers an opportunity to connect, and perhaps coordinate, information from different disciplines, which is seen as instrumental in effective sustainability science (Kates et al. 2001; Jerneck et al. 2011). It serves an explanatory purpose by accounting for interactions and events as well as providing a way of analysing actual systems for their resilience (thus in effect becoming a kind of diagnostic tool). It also offers neutral grounding for otherwise normatively charged concepts such as sustainability.⁸

Although there is always a strong link to Holling's resilience, variation in the way different authors use the concept is considerable, at least at a first approximation. Fridolin Simon Brand and Kurt Jax (2007) list 13 definitions, dividing these into four main groups on basis of their normative content (see Table 1). One such definition runs as follows:

Table 1 Thirteen definitions of resilience according to Brand and Jax (2007)

	Definition	GR	Specification
1	Measure of the persistence of systems and of their ability to absorb change and disturbance and still maintain the same relationships between populations or state variables	S	System
		D	Unspecified
		I	Structure
2	The magnitude of disturbance that can be absorbed before the system changes its structure by changing the variables and processes that control behaviour	S	System
		D	Unspecified
		I	Structure
3	The capacity of a system to experience shocks while retaining essentially the same function, structure, feedbacks, and therefore identity	S	System
		D	Unspecified
		I	Structure, function, feedbacks
4	Quantitative property that changes throughout ecosystem dynamics and occurs on each level of an ecosystem's hierarchy	S	?
		D	?
		I	?
5	Resilience of what to what?	S	?
		D	?
		I	?
6	The ability of the system to maintain its identity in the face of internal change and external shocks and disturbances	S	System
		D	Unspecified
		I	Identity
7	The ability of groups or communities to cope with external stresses and disturbances as a result of social, political, and environmental change	S	Groups, communities
		D	Social, political, environmental change
		I	Unspecified
8	Transition probability between states as a function of the consumption and production activities of decision makers	S	?
		D	?
		I	?
9	The ability of the system to withstand either market or environmental shocks without loosing the capacity to allocate resources efficiently	S	Unspecified
		D	Market and environmental shock
		I	Capacity to allocate resources efficiently
10	The underlying capacity of an ecosystem to maintain desired ecosystem services in the face of a fluctuating environment and human use	S	Ecosystems
		D	Environmental and human use
		I	Ecosystem services
11	The capacity of a social-ecological systems to absorb recurrent disturbances ( … ) so as to retain essential structures, processes and feedbacks	S	Social-ecological systems
		D	Unspecified
		I	Essential structures, process, and feedbacks
12	Flexibility over the long term	S	?
		D	?
		I	Flexibility
13	Maintenance of natural capital in the long run	S	?
		D	?
		I	Natural capital

Notes: GR denotes global resilience; S, D, and I signify specifications of the system, the type of disturbance, and the property maintained, respectively. A question mark means that the place-holder is not only unspecified but not even present in the definition.

By resilience, we mean the capacity of linked social-ecological systems to absorb recurrent disturbances such as hurricanes or floods so as to retain essential structures, processes, and feedbacks. Resilience reflects the degree to which a complex adaptive system is capable of self-organisation (versus lack of organisation or organisation forced by external factors) and the degree to which the system can build capacity for learning and adaptation. (Adger et al. 2005, 1036)

Similar formulations can be found in Carl Folke et al.:

Resilience, for social-ecological systems, is related to (i) the magnitude of shock that the system can absorb and remain within a given state; (ii) the degree to which the system is capable of self-organisation; and (iii) the degree to which the system can build capacity for learning and adaptation. (Folke et al. 2002, 438)

In the same vein, Brian Walker et al. (2004, 1) understand resilience as ‘the capacity of a system to absorb disturbance and reorganize while undergoing change so as to still retain essentially the same function, structure, identity, and feedbacks’.

Before we proceed, it is important to note that what is sometimes called ‘resilience thinking’ or ‘resilience theory’ (Folke et al. 2010; Strunz 2012) will be left aside. These concepts capture something much broader: ‘a resource management approach and a view of the world that is not necessarily tied to scientific discourse and academic institutions’ (Strunz 2012, 113f). No attempt will be made to analyse this idea here, as we are interested in it only insofar as it concerns the specific concept of resilience.

3. Abstraction and the Conceptual Core(s)

The abstract nature of the concept of resilience plays an important role in the argument presented here. It is thus important to distinguish abstraction from other ways in which a concept may be said to be unspecific or imprecise. Especially important here are vagueness and ambiguity. A vague concept is often thought of as a concept with borderline cases—that is to say, a concept whose extension cannot be precisely determined. Consider the concept of a copse. A copse is a small group of trees that is not a forest. We have many examples of things that are clearly copses and we have many examples of things that are clearly forests, and thus not copses. However plenty of groups of trees lie somewhere in between and are hard to classify. This kind of vagueness can be stipulated away—in order, for instance, to make forest management easier. An arbitrary, sharp boundary (29 trees is a copse, 30 is a forest) can be imposed. Such a stipulation, however, does not mean that we now have discovered the real boundary between copses and forests, or learnt something about these concepts. Vagueness in this sense can be problematic and unwanted in science, although it is standard practice within many disciplines to make concepts more precise than vernacular usage would suggest.

Turning to ambiguity, this is a property of terminology. An ambiguous term, in other words, can be used to signal more than one kind of thing. Thus the term ‘bat’ can either pick out the blind, flying mammal that uses sonar for navigation, or the thing commonly used in baseball to hit the ball across the pitch. So two concepts are associated with the term ‘bat’, neither of which is vague, or particularly abstract. In this case disambiguation is not challenging. Other cases prove more difficult. Sebastian Strunz (2012) has argued that resilience is one of them, and that its disambiguation is challenging.

3.1. Determinates and Determinables

William Ernest Johnson (1921) introduced the distinction between determinates and determinables to highlight the logical structure of expressions such as ‘red is a colour’. The distinction operates as follows: the concept red is a determinate of the determinable coloured, and scarlet is a determinate with respect to the determinable red. Determinables have no ‘pure’ instances—nothing can be coloured without being some specific colour (and hue of that colour). Abstract concepts tend to have wider extensions than more specific ones. In fact, Johnson thought of concrete things in the world as absolutely determinate (see Sanford 2013). We might expect the concept red to be applicable to more things in the world than the concept scarlet. Obviously, this is not necessarily so, since it might be the case that all red objects are scarlet. It is, however, impossible for the concept red to have a smaller extension than the concept scarlet.

Strunz (2012) has argued that ‘resilience’ is indeed ambiguous, but that this is not necessarily a bad thing. I will return to this discussion in sections 4.1, 4.2, and 5. If ‘resilience’ is indeed ambiguous, there can be no common core to all resilience concepts—here Strunz appeals to the Wittgensteinian notion of a family resemblance. I shall not reject Strunz's claims. I wish instead to explore, and attempt to formulate, a conceptual core that is shared by many, though perhaps not all, concepts of resilience. I will argue that there is a tendency among those who use the term ‘resilience’ to prefer more abstract interpretations of its meaning, and that this means that the concept of resilience being deployed is quite close to the conceptual core we identify.

On this account abstraction is a form of ambiguity, albeit structured in a specific way. Consider again R. M. Hoffman (1948) and resilience in textile research. Hoffman himself notes the ambiguity in the notion of resilience he uses when describing retracting yarns: one might be interested in the speed of return, how close to the original shape the yarn becomes after a disturbance, or how much stretching it can cope with (Hoffman 1948, 141f).

3.2. Two Definitional Schemas

If we employ Johnson's conception of abstraction, then, in our search for the conceptual core of resilience we should be searching for a determinable common to many determinates; that is to say, some general, or abstract, conception of resilience.

In section 2, I presented a range of definitions of the concept, which might be taken to suggest that no such core could possibly be found. Here, however, two resilience concepts commonly seen in many disciplines are identified. As mentioned in the introduction, I shall call these local and global resilience. This division roughly traces the conceptual line, in Schrader-Frachette and McCoy (1993), between dynamic balance and persistence. It is also shadows Holling's (1996) distinction between engineering and ecological resilience.

Local resilience, thus named because it concerns a single domain of attraction of some system, is characterized as follows:

Local resilience: the ability of S to return to state E after disturbance D.

This definition can be made more specific by considering S to be a piece of yarn retracting after having had a weight attached to it, a child returning to normal behaviour following the loss of a close relative, or a population of fish regaining its former number after having been overfished. As explained in section 2, concepts that relate in this way to local resilience are common and are found in many fields. Interestingly, even with specifications of S, E, and D this notion can still be operationalized in various ways. For example, one may be interested in the speed of return (what Holling calls stability), or the difference between the states S is in after D and E.

As already noted, not all concepts of resilience are captured by the notion of local resilience. In fact some are defined explicitly by contrast with local resilience (as is the case in Holling 1973). What here has been called global resilience is not about the returning of a system to some reference state, but the absorption by a system of a disturbance. The choice of ‘global’ is motivated by the fact that it is usually considered a property of a system as a whole, rather than a feature pertaining only to a system as it is close to some domain of attraction. Global resilience involves keeping some property of a system fixed as the system is disturbed—in other words, the maintenance (not recoverability) of some property during disturbance. Sustained competence under stress, and the sort of resilience Neil Adger et al. (2005, 1036) describe as the retaining of ‘essential structures, processes, and feedbacks’ when a system is disturbed, both fit this characterization of resilience. They do not satisfy the requirements of local resilience.

A first approximation of global resilience is as follows:

Global Resilience (1): the ability of S to absorb disturbances and maintain I.

S is often a part of definitions and characterizations and should be understood as a preliminary delimitation. As can be seen in the examples that have been provided references to specific classes of system are fairly common. I here denotes a property of the system that is maintained during the disturbance. It is ordinarily associated with the identity or persistence of the system, and thus it should, ideally, come with some explicit criteria governing when either of those relations can be said to hold. That, however, is quite rarely the case, and both content and degree of specificity varies considerably. In the examples offered here (Table 1) it is structural features, such as relationships between state variables and parameters, that recur, but other varieties are conceivable, including the upholding of a particular function, for example.

Stephen R. Carpenter and Gunderson (2001) have pointed out that global resilience is not an intrinsic property of systems, but a relation between a system and a particular (type of) disturbance: a given system S can be resilient in different degrees to different disturbances. An ecosystem might be resilient with respect to forest fires but not with respect to a ruinous pest. Thus, in our definitional schema the disturbances need to be taken into account:

Global Resilience (2): the ability of S to absorb disturbance D and maintain I.

D here represent some disturbance, or class of disturbances. This schema fits well with many of the available definitions, as shall be seen in the next section, although one should note that there is a general issue here about how disturbances are to be individuated. Individuation may be based upon the way disturbances affect the system in question, or it may be achieved via some other criterion of individuation. If one opts for the second alternative, it may be the case that this definition captures several, relevantly different forms of resilience. There will be cases where a single disturbance affects a single system in several ways, and some systems are complex in the sense that they have many dimensions that can be subject perturbation. Real social-ecological systems, like those that exist in coastal areas (Adger et al. 2005), are good examples. Prima facie, the disturbances such systems are typically assessed for—floods, rising sea level, and so on—are disruptive of these systems in any number of different dimensions. In all probability, the system will be resilient in different ways, and to varying degrees, in these different dimensions. Generally it seems preferable to think of D in terms of how it affects the system under scrutiny, but here I will stick to this schema as it is more representative of the definitions one actually find.⁹

Global resilience, like local resilience, can be operationalized in several ways even if S, D, and I are fully specified. Moreover, I do not suggest that these definitional schemas need to be precisely specified under all circumstances. Sometimes it is appropriate and preferable to use more abstract formulations. Rather this is a way of analysing resilience that appears to capture a structure present in most deployments of the notion of resilience in sustainability research, and common also within, for instance, psychology. Our suggestion is that these schemas express abstract notions that can be made more specific by filling in S, D, and I. The degree of specificity needed depends on the context, but more specific versions relate to more abstract ones as determinates to a common determinable.¹⁰

3.3. Conceptual Pluralism Revisited

Two questions arise. The first is: how representative are these definitional schemas of the actual usage of concepts of resilience? This question has already been answered in part, but in order to provide a fuller answer, I use Brand and Jax's (2007) list of 13 definitions as a comparative base (see Table 1). The second question has to do with general preferences as to the degree of abstraction: How abstract are the resilience concepts that are preferred? More abstract concepts are less sensitive to specific contexts, but also less informative about underlying structures and mechanisms, and this has implications vis-à-vis the strength of the unification achieved (if this is indeed the goal).

Table 1 shows how the 13 definitions of resilience collected by Brand and Jax (2007) fit with our definitional schema of global resilience. The reason local resilience has not been included is that none of them fit with that schema. Five of the 13 definitions do not fit global resilience either. Two can be disregarded immediately: no. 5 is not a definition at all and no. 4 gives only a very broad characterization of features of the property.

In the remaining three definitions that do not fit global resilience, i.e. nos. 8, 12, and 13, the assignments of S, D, and I are partial with respect to our definitional schema. Definitions 12 and 13 give identity criteria—a system is resilient if either flexibility or natural capital is maintained over time—but are otherwise incomplete. None of these three is, as it stands, inconsistent with global resilience (although in this regard I harbour reservations about no. 8, about which I am not sure).

As already pointed out in the previous section, local resilience fits well with resilience concepts of the sort used in materials science, psychology, and ecology. Within psychology especially, many forms of resilience are determinates of the determinable global resilience. Sustained competence under stress was already mentioned, but the definition also fits with the idea of resilience being the ability to follow developmental trajectories in spite of trauma.

Moving to the second question about the degree of abstraction displayed by resilience concepts in use, it is notable that only one definition (no. 10) specifies all three variables, and then not in particularly precise terms. Should one conclude, then, that there is a preference for abstract notions of resilience? Perhaps. But taking the contexts in which these definitions occur into account may well reveal them to be more specific. For example, 1, which is due to Holling (1973), appears in an ecological context, and thus an argument could be mounted that it should be understood as pertaining to ecosystems only. On the other hand, one might also look at practice to see if there are implicit preferences. I think there is this kind of preference for abstraction. The following two examples are perhaps not conclusive evidence, but they provide support of a kind for this suspicion.

Budworms. In their review of an example that is familiar from the literature—spruce budworms in eastern Canada originally described in Ludwig, Jones, and Holling (1978)—Donald Ludwig, Brian Walker, and Crawford Holling (1997) note that there are many ways in which resilience applies within their model. Essentially the budworm interacts with a number of species of tree in eastern Canada and is prone to periodic outbreaks that can be described by differential equations.

Consider the diagrams in Figure 1. R and Q are composite parameters that depend on the relationships between a number of lower-order parameters in the model—parameters that specify carrying capacity without predation, upper limits for predation, and so on. These lower-order parameters are actually variables in the model, albeit slowly changing ones, and depending on their relationship the model comes to rest at two, three, or four points. One unstable equilibrium occurs when budworm density is nil. For a range of relationships between R and Q a low, stable equilibrium B ₋ appears towards which the system will tend. As can be seen in Figure 1, for a fixed value of Q it will be the case that above certain values of R ≥ R ₂ the stable B ₋ and unstable B _c, first coincide, and then disappear altogether. That leaves the system with only one stable equilibrium, here designated B ₊. Thus, an outbreak occurs and the system remains within this higher equilibrium until (again for the same fixed value of Q) R < R ₁.

Figure 1 Four different values of R given a fixed value of Q. Different relationships between R and Q generate systems with different numbers of, and values for, those equilibria.

Ludwig, Walker, and Holling (1997) acknowledge that we may focus on the resilience of some lower—and, for the forest industry, preferable—equilibrium, and thus view the abrupt disappearance of lower equilibriums as loss of resilience. This construal is not, perhaps, straightforward, but certainly possible.¹¹ On the other hand, the budworm naturally interacts with its surrounding environment. The parameter α that determines when budworm density becomes saturated, for instance, is proportional to foliage density and will change with a growing forest. This affects R, effectively creating an oscillating system in which irregular, but recurrent outbreaks of budworm control other species. Where budworm outbreaks serve to regulate the balance over long periods of time, the balance between balsam fir, spruce and birch is especially important (Holling 1973, 13f). Ludwig, Walker, and Holling (1997) argue that such oscillators can be very important in biological and ecological systems, as ‘they continue to oscillate more or less with the same frequency and amplitude under a wide variety of disturbances’ (Ludwig, Walker, and Holling 1997, 9). The oscillator is, then, a constitutive part of what realizes the resilience of a system, and dampening it—as in this case could, and did, happen through the suppression of budworms—actually renders the larger system less resilient.

The mere fact that Ludwig, Walker, and Holling elaborate these different potential applications of the concept of resilience indicates that they are in fact endorsing an abstract notion of resilience, one that is not sensitive to the particular system it is applied to. The decision to pick one interpretation over another is then made on other grounds.

Migration and mental health. Psychologists have often focused on the outcomes of resilience highlighting ‘the maintenance of functionality’ in the individual (Olsson et al. 2003, 2). Indicators here may include good mental health, functional capacity, and social competence. There is an issue, however, about how these outcomes are to be evaluated. As Craig Olsson et al. point out, ‘emotional well-being’ poses particularly ‘perplexing’ challenges:

It is tempting to define adolescent resilience solely in terms of maintenance of emotional well-being in the face of adversity. However, it may be unrealistic to believe that young people can quickly resolve the emotional ramifications of serious threat to personal values (e.g. illness, death of a loved one). Distressing emotion must in some way act as an index of adversity. (Olsson et al. 2003, 3)

The issue is that it is not clear whether we should regard the emotional dimension as that which is being disturbed (absence of emotional well-being is a sign of impact) or as an indicator of the degree of resilience (as the absence or presence of resilience). In fact, as Rutter (1993) once indicated, some have described extreme emotional responses, such as clinical depression, as the ‘price of resilience’ (Rutter 1993, 627). On this view, depression is essentially an adaptation to adverse conditions and hence a sign of resilience rather than its lack.

An analogous problem appears in sustainability science. Consider again the example of migration and mobility as indicators of social resilience (see Adger 2000). Displacement can be a sign that a community is collapsing, the idea being that had the system been more resilient the displacement would not have occurred. Migration thus becomes ‘an indicator of the breakdown of social resilience’ (Adger 2000, 357). On the other hand, mobility might be considered as a preservation strategy and an adaptation to an external factor of some kind, in which case it looks more like an indicator of resilience than a deficiency in it.¹²

Uncertainty over whether this or that feature should be interpreted as a sign of resilience or a symptom of its lack flows from the implicit, or explicit, use of a highly abstract concept of resilience. The problem can be ameliorated by stipulation, but that solution is susceptible to the suspicion that it is arbitrary, or ad hoc. The point I wish to make is that the appearance of the problem itself is enough to demonstrate that there is a preference, among the authors here considered, for a concept of resilience that is in fact highly abstract.

4. Resilience Bridging Disciplines

To recap, then, two highly abstract resilience concepts have been identified. They have been labelled local and global. These concepts can be used to structure many of the more specific resilience concepts in actual use, as the latter tend to relate to the former as determinates relate to a determinable. It has also been established, on the basis of the two examples above, that in at least some of the literature there is a preference for rather abstract concepts of resilience, and it has been noted that abstraction promises to link disparate cases of resilience in different disciplines. Now we shall move to discuss two other forms of conceptual connection: boundary objects and metaphor.

4.1. Resilience as a Boundary Object

Brand and Jax (2007) have proposed that resilience works as a kind of boundary object. Three features are essential to such objects (Star and Griesemer 1989; Collins, Evans, and Gorman 2007):

Flexibility: Adaptation to local needs.

Rigidity: Maintenance of identity across sites.

Communication: Mediate communication between different cultures that could not otherwise communicate.

The notion of boundary objects is a descriptive one introduced to show how otherwise incommensurable cultures can communicate in spite of their differences. Establishing a boundary object is hence not preferable to, say, sharing a language. It should be thought of as a way of establishing communication when other options are not available.

This is how these requirements, as they apply to concepts, will be interpreted here. A flexible concept is ambiguous with respect to different ‘cultures’, i.e. different disciplines. Perhaps with meanings close to one another, what Strunz (2012) call polysemy or vagueness depending on whether these different meanings are easily separable or not. The precise nature, or degree, of this ambiguity can vary, but clearly ambiguity alone is not enough. Two disciplines may use identical ambiguous concepts: conceptual joints must coincide with disciplinary boundaries.

At the rigidity end of the spectrum there is every reason to be pluralists, and to insist that what, precisely, maintains identity across disciplines can vary. As long as there is some degree of flexibility, and communication in fact obtains, the concept is a boundary object. An extreme case would involve sharing only a label. One possibility would be to construe less radical cases in which the concepts involved are structured in exactly the way that has been suggested above as examples of different disciplines embracing different determinates of the same determinable. An obvious worry about this proposal is that it would threaten to dilute the notion of a boundary object unacceptably. Moreover, I suspect it would be hard to find cases in which the conceptual connection is one where a conceptual core is shared and no other modes of communication are available to the parties involved.

Indeed the resilience example is a case in point. A number of resilience concepts are spread among different disciplines. Mostly, however, no communication has ensued with resilience as a basis. If one brings to bear the more specific concepts of local global resilience one finds more signs of exchange, especially in respect of the latter. Interdisciplinary endeavours like the Stockholm Resilience Centre, which build on Holling's concept of resilience, appear to promote communication across disciplines successfully, but they do so by pressing a much narrower, and more fixed, resilience concept into service. Interestingly, considerable variability in the concept occurs frequently within disciplines. In both ecology and psychology several more specific varieties of both global and local resilience are visible. In such contexts, however, resilience cannot be a boundary object as there is no boundary to begin with.

There are problems with the notion of boundary objects when applied to concepts that present a challenge here, especially concerning the communication criteria. How do we verify that communication is maintained through the concept in question and how can one can one determine that there are no other venues of exchange available (anyway, how is that to be interpreted exactly)? Falsification, on the other hand, is somewhat easier. It is clear that merely sharing some term is not sufficient. Two disciplines can share a label by accident, perhaps without members of the respective disciplines even noticing. To describe such a situation as one where a boundary object is present makes the notion of a boundary object both trifling and somewhat arbitrary. So, it is obvious, that the concept of resilience has not been a boundary object between materials science and psychology although it may be argued that there is both some degree of flexibility and some degree of rigidity.

4.2. Resilience as a Metaphor

Resilience has also been treated as a metaphor. Thus it has been said that resilience ‘has multiple levels of meaning: as a metaphor related to sustainability, as a property of dynamic models, and as a measurable quantity that can be assessed in field studies of SES’ (Carpenter et al. 2001, 765). A metaphor involves seeing something as something else—the brain as a computer, or natural selection as a kind of artificial selection (Sloep 1997). A recurring theme in the literature on the role of metaphors in science associates the value of metaphors to their heuristic ability to generate new and interesting hypotheses to be investigated (Black 1962; Hesse 1966; Kittay 1987; Kellert 2008). Importantly, the metaphorical relation is one that holds between systems (entities, structures) and not between a concept and that to which it is applied. The resilience concept is indeed part of a metaphorical transfer but is itself not a metaphor; instead the relevant metaphor—in the context in which Carpenter et al. are writing—is the seeing of complex social entities (social systems on their terminology) as ecosystems.

The question whether the ecosystem–social system metaphor is suitable depends on its fruitfulness, and this in turn depends on both quantitative and qualitative features of the similarities that obtain (or are made to obtain) between the systems in question (together with a range of other factors). Proper assessment of the metaphor is beyond the scope of this article. Instead I wish to raise two concerns. First, this metaphor can only partly explain the wide applicability of the resilience concept. Psychologists developed their concept of resilience independently of ecologists, and this metaphor is nowhere to be found in that literature. The proliferation of the concept is considerably broader than the proliferation of the ecosystem–social system metaphor. Second, questions can be raised about whether this metaphor is indeed interdisciplinary in the collaborative sense preferred here. Among some ecologists, particularly those with a strong interest in sustainability like Holling himself, the metaphor is popular, but elsewhere it seems to have attracted little attention.¹³

5. Unity and Multiple Realizability

It is quite clear that the ecosystem–social system metaphor is, at least for many scientists interested in sustainability, what carries the concept of resilience from one domain of inquiry to another. With the proviso that one must be sensitive to the limits of metaphors in general, and this metaphor in particular, this might well prove fruitful. However, the notion of metaphor alone fails to explain the widespread use of the concept of resilience. Psychologists, for example, appear to swear by no such metaphor.

The thesis advanced here is that the wide applicability of the resilience concept is underwritten by its highly abstract character. The implications of this are important. Some of the criticisms levelled against the concept focus on what appears to be the suitability of the underlying metaphor. Hornborg, for instance, argues that the framework tends to ‘mask the power relations … that to a large extent determine how humans utilise the surface of the Earth’ (Hornborg 2013, 116). Not all uses of the concept, however, build on this metaphor. The idea of resilience as a boundary object appears to be limited in similar ways. The criteria required for something to qualify as a boundary object are rarely met. Note that it is not denied that they are sometimes met, but rather that this accounts for the way in which the concept of resilience has been deployed in general.

The suggestion here is that abstraction has a larger role in explaining the broad applicability of the concept of resilience than has generally been admitted. It allows resilience to appear in highly differentiated subject matters because highly abstract concepts are comparatively context insensitive. However, abstraction cuts both ways, and this impressive level of mobility comes at the cost of weak interdisciplinary links. A shared resilience concept will, in general, provide little or no incentive for further integration. This fact might also explain why there has not been more intellectual exchange despite the similarity of the concepts of resilience in different disciplines.

One feature of highly abstract concepts is that they tend to be multiply realizable.¹⁴ This insight can be fleshed out in various ways. It can be expressed somewhat crudely, however, as follows: some phenomena are not systematically realized at the microscopic level. To borrow an example from Jerry A. Fodor, many general and interesting things that can be said about the economic phenomenon of monetary exchange are irreducible to, for instance, physics, in this sense:¹⁵

banal considerations suggest that a description [in the vocabulary of physics] which covers all such events must be wildly disjunctive. Some monetary exchanges involve strings of wampum. Some involve dollar bills. And some involve signing one's name to a check. (Fodor 1974, 103)

The notion of multiple realizability has been exploited in arguments for both the autonomy (Fodor 1974) and, curiously, for the disunity (Kim 1992) of the special sciences. The direction of the argument depends on the purpose and intent of the study: sometimes we are interested in the realizers, and sometimes we wish to leave them out.

It is clear that both local and global resilience are context-insensitive concepts in the sense that they are multiply realizable; any number of different mechanisms, or underlying structures, can be responsible for a system's resilience in either of these senses. One realizer of global resilience in organisms is redundancy. Many phenotypes will develop even if the relevant genes are knocked out. Other genes compensate for the loss (Mitchell 2009). Similar mechanisms can be observed at higher levels of organization as well—for example, where functions associated with damaged structures of the brain are taken over by other structures (Richardson 2009). This is one way in which global resilience can be realized within a particular system, but there are many others. At a more concrete level, the individual mechanisms and structures responsible for realizing resilience in individual systems come in many forms.

Broadening the perspective somewhat, suppose we explain the fall of the Western Roman Empire by saying that it was a consequence of her resilience to barbarian invasions being eroded? A number of different facts about the Empire may be said to have something to do with her resilience: maintenance of a large and sophisticated military force, effective political institutions, peaceful succession, a reliable system of collecting taxes, and so on. The resilience of the influenza virus to the countermeasures taken by its hosts, on the other hand, would probably be due to genetic, epigenetic, and structural factors: ‘good’ genes, of course, a small and easily manipulable genome, short cycles of reproduction, vast numbers of individuals in each generation, et hoc genus omne. Ecosystems, on the other hand, are resilient to external damagers such as famine, disease, and hurricanes, largely thanks to the specific relationships between, and numbers of, interacting species, as well as complex multi-state cycles. The mechanisms involved in making a child resilient to psychological trauma are different again: heritable factors, previous history of mental health, and the social network in which the child is situated, are all likely to play a role. At any other than the most abstract level—that is, the level of global resilience itself—these mechanisms and structures appear to be just the kind of ‘wildly disjunctive’ sets that Fodor mentions.

If our interest lies with underlying mechanisms, the observation that resilience is multiply realizable will matter for the unificationist. Mechanisms and structures that differ so radically from one another must be studied using different methods, tools, and representation. For example, modelling practices tend to be relative to specific subject matters (Dupré 1996). A study of resilience that takes its realizers into account is a pluralist project in many respects.

Are these mechanisms the locus of study then? The answer to this question depends on several factors whose analysis is largely beyond the scope of this article, but our strong suspicion is that the mechanisms are important, especially within emerging fields such as sustainability science where the debate over the concept of resilience is most active. This suspicion is anchored both in the high degree of abstraction in the resilience concept itself, since this renders a study of resilience in the absence of its realizers a strictly formal exercise, and in the empirical and concrete goals of sustainability scientists, i.e. the objective of providing diagnostics, predictions, and prescriptions relating to actual social-ecological systems.

6. Conclusions

It has been argued that most resilience concepts can be traced back to either of two general forms of resilience: local and global. Specific concepts of resilience in a range of disciplines, including ecology, psychology, materials science, and sustainability science, relate to these more abstract concepts as determinates to determinables. Furthermore, I have argued that there is a tendency, among those who make explanatory use of resilience, to implicitly or explicitly prefer abstract understandings of the concept. This has bearing on questions about the way concepts of resilience connect disparate disciplines. Several authors, including Brand and Jax (2007) and Strunz (2012), have noted that the concept of resilience is heterogeneous, and have argued that this heterogeneity may, in various ways, be helpful in connecting different disciplines. It is suggested that abstraction is, in the resilience case, an overlooked and important further way of connecting disciplines.

This suggestion is potentially important for those hoping that resilience is a concept that may unify and connect the natural and social sciences. The concepts of resilience that have been discussed pick out diverse kinds of mechanism, and to the extent that mechanisms are the locus of inquiry, a general study of resilience, global or local, needs to be a methodologically and theoretically pluralist venture. A research programme with empirical components founded on the resilience concept alone, in either of the suggested senses, can be expected to be highly diversified in many respects.

Lastly, it is my view that the current research programme dedicated to locating and describing mechanisms and structures that give rise to resilience is underdeveloped. Instead it seems that the lion's share of empirical research is devoted to finding correlates for resilience. These often appear to be little more than heuristics selected on the basis of persistence, and not particularly reliable heuristics at that. This unreliability, I suspect, flows from the fact that these indicators rarely appear capable of discerning between the ability to absorb an impact, on the one hand, and the capacity to avoid an impact, on the other. Some systems persist as a consequence of being resilient, others because they are not disturbed. If we consider these as ‘strategies’, we can see that they may work equally well. Thus it is widely contended that wealth is an indicator of resilience, but surely it is also an indicator of avoidance. This matters in comparative terms. For the wealthy, both avoidance and resilience are possible strategies, and, being wealthy, they can choose either as it suits them. For the poor, however, avoidance may not be option. This might indicate that poverty, rather than wealth, is an indicator of resilience, albeit not perhaps the kind of resilience accompanied by persistence. A mixed strategy may well be more successful, on the whole. In many cases, no doubt, mechanistic or structural explanations of resilience would help us to separate the two forms of strategy.

Acknowledgements

I want to thank Johannes Persson, Lennart Olsson, and James McAllister as well as two anonymous reviewers for many helpful comments and suggestions in developing this manuscript. Furthermore I am in a debt of gratitude to Paul Robinson, Brian Hepburn, and the research seminars at the centres of excellence LUCID at Lund University and TINT at Helsinki University, as well as the seminar at the Centre for Science Studies at Aarhus University, where earlier version of this paper was presented. This work was supported by the Linnaeus programme LUCID (Lund University Centre of Excellence for Integration of Social and Natural Dimensions of Sustainability, www.lucid.lu.se), FORMAS 2008-1718. Moreover this research benefited greatly from discussions within the research project, ‘Measuring, Assessing and Profiling (MAP) Human Resilience’, funded by the Rockefeller Foundation (contract no. 2012 RLC 304).

Notes

[1] Other fields are represented as well, such as medicine and economics, but to a diminished extent in comparison. All in all, Web of Science finds only 51 papers published before 1970.

[2] See Rutter (1993) for a discussion of the merits of resilience with respect to invulnerability. See also Rutter (1985), Garmezy (1991), and Olsson et al. (2003).

[3] Fogany et al. consider resilience to be ‘normal development under difficult conditions’ (Fogany et al. quoted in Daniels 2008, 60)

[4] This variety of resilience is used by Bonanno et al. (2007): they use the term to denote the ‘capacity to maintain healthy, symptom-free function … following PTEs’ (181).

[5] See for instance Dyer and McGuinness (1996). They endorse a version of the concept closer to the usages common in material science: resilience as the ‘process whereby people bounce back from adversity and go on with their lives’ (Dyer and McGuinness 1996, 277).

[6] Around 1970 critical voices were raised—among them Holling, as discussed below—and by the end of the 1980s the thesis had been largely discredited. For discussion of this debate see Schrader-Frachette and McCoy (1993); Redfearn and Pimm (2000); Justus (2008); deLaplante and Picasso (2011). For a detailed overview of the various stability concepts in ecology, Grimm and Wissel (1997) is an excellent resource, while Hansson and Helgesson (2003) have developed a context-independent and parsimonious account of the concept of stability.

[7] Holling's (1973) paper, published in an ecological journal, has to date had over 2000 citations. Until 1998 it had between 10 and 30 citations a year, but since then the number of citations has grown almost exponentially, with almost 230 in 2013 alone.

[8] The relationship between resilience and sustainability is contested, although the connection is often made. See for example Common and Perrings (1992); Lélé (1998); Perrings (2006); and Derissen, Quaas, and Baumgärtner (2011).

[9] If one does prefer to consider the ‘effect side’ of disturbances, it is important that D and I concern different parameters, or properties, of the system. In cases where they are identical the system cannot be perturbed without changing and thus, trivially, is not resilient at all.

[10] Are there senses of resilience that might have been missed? As I have spelled them out here, global and local resilience are roughly equivalent to two of the three senses of stability that Hansson and Helgesson identify. In their terminology, robustness denotes the ‘tendency of a system to remain unchanged, or nearly unchanged, when exposed to perturbations’; they reserve the term ‘resilience’ for the ‘tendency of a system to recover or return to (or close to) its original state after a perturbation’ (Hansson and Helgesson 2003, 222). The third sense, which they call constancy, is non-dynamic; it describes the historical property of not changing. I have found no uses of the term ‘resilience’ that correspond to constancy, which was only to be expected. The word ‘resilience’ has roots in the Latin resilire, which means to jump or to bound back. In other words, dynamics appear to be at the heart of the concept. Since stability is a more abstract concept, I believe that the two senses of resilience presented above capture the vast majority of resilience concepts.

[11] For values of R some distance from the critical R₂ the system has a buffer of sorts and may therefore absorb disturbances to a number of variables and parameters without much in the way of consequence. The closer R is to R₂, however, the more probable it becomes that an otherwise minor disturbance might push the system beyond this boundary, causing an outbreak.

[12] Adger seeks to resolve this problem by drawing up a distinction between migration that is circular and seasonal, and migration that is not. The former, he argues, is a ‘strategy for risk spreading at the household level’ and may thus be an important constituent in resilience (Adger 2000, 357). This is unconvincing in my view. Disturbances that are highly regular are usually considered parts of the system itself. Given this, resilience concerns the ability of a system to absorb irregular (though sometimes recurring) stress. Unless migration is removed from consideration by stipulation, no conceptual barrier prevents us from viewing it as a sign of resilience.

[13] This may well change, but currently one social science journal dominates, according to Web of Science, and that is Ecology and Society. This should come as a surprise to no one, as this journal is the main voice of the Resilience Alliance. Mainstream social science has not taken notice of this, however. Among the 10 highest-ranked journals in economics, social science, and political science, respectively, I found no hits on searches for the terms ‘resilience’ and ‘ecology’ and only six hits between the 30 of the terms ‘resilience’ and ‘system’.

[14] Multiple realizability is an idea originating within the philosophy of science and the philosophy of mind. Putnam (1967) introduced the notion in the course of presenting an argument to rebut reductionist claims. See also Fodor (1974).

[15] The empirical validity of this thesis has come into question, especially as it pertains to mental kinds (Bechtel and Mundale 1999; Richardson 2009). This does not matter in the case of abstract notions, however.