elder vas.pdf.pagespeed.ce.yrcnhpsu 7

8/10/2019 Elder Vas.pdf.Pagespeed.ce.YrcNHpsU 7

1/15

RE-EXAMINING BHASKARS THREE ONTOLOGICALDOMAINS: THE LESSONS FROM EMERGENCE

Dave [email protected]

Dave Elder-VassSchool of Politics and SociologyBirkbeck CollegeUniversity of LondonMalet StreetLondon WC1E 7HX

Paper for presentation at IACR Conference, Cambridge, August 2004


2/15

D. Elder-Vass 2

RE-EXAMINING BHASKARS THREE ONTOLOGICALDOMAINS: THE LESSONS FROM EMERGENCE

Introduction

Although Roy Bhaskars ontology in A Realist Theory of Science (Bhaskar1978) is explicitly stratified, he makes relatively little attempt to examine the basis ofstratification (Collier 1994, p. 130), or its implications for his three domains of theempirical, the actual, and the real 1, 2 . This paper attempts to remedy that deficiency, byinvestigating the nature of a stratified reality based on emergence, and consideringhow this impacts our understanding of experiences, events, entities, and causes. Ihasten to emphasise that my objective here is primarily to refine Bhaskars argument,and to repackage it in a form that provides greater clarity, rather than to undermine orcontest its essential content. My argument in no way conflicts, for example, with the

need for a multi-domain depth ontology, or with the argument that reality is stratified.What the paper does seek to do, on the other hand, is to add some depth to the

characterisation of experiences, events, and entities, thus clarifying the distinctionsbetween ontological domains. Ultimately this will lead me to propose an alternativeway of conceiving Bhaskars domain of the real. While many simpler causal accountswork well without them, the modifications proposed in this paper are of particularsignificance when we turn to the argument against reductionism.

The paper will begin by introducing the key terms: Bhaskars three domainsand the concept of emergence. It will then move on in turn to discuss the implicationsof emergent stratification for events, entities, causes, and experiences. Finally it willbring together the threads of the argument to re-evaluate Bhaskars three domains.

Bhaskars three domains

In A Realist Theory of Science , Bhaskar argues from the intelligibility ofexperimental activity that there is an ontological distinction between scientific lawsand patterns of events (Bhaskar 1978, p. 12). Such laws depend upon the existence ofnatural mechanisms, and it is only if we make the assumption of the realindependence of such mechanisms from the events they generate that we are justifiedin assuming that they endure and go on acting in their normal way outside theexperimentally closed conditions that enable us to empirically identify them (p. 13).Similarly, events must occur independently of the experiences in which they areapprehended. Structures and mechanisms then are real and distinct from the patternsof events that they generate; just as events are real and distinct form the experiences inwhich they are apprehended. Mechanisms, events and experiences thus constitutethree overlapping domains of reality, viz. the domains of the real , the actual , and theempirical (p. 56). The relationship between these domains is summarised in a table,reproduced below as Figure 1.

1 I assume the reader has a certain degree of familiarity with some of Bhaskars concepts, notably thetransitive/intransitive distinction.2 There seems to be some ambiguity in the use of stratification in the critical realist literature; I take itto mean the division of the world into emergent explanatory levels, not the division of ontology intodomains.


3/15

D. Elder-Vass 3

Domain of Real

Domain of Actual

Domain of Empirical

Mechanisms x

Events x x Experiences x x x

Figure 1 Bhaskars three domains: populating entities (Bhaskar 1978, p. 13)

Bhaskar clearly intends the domain of the empirical to be a subset of thedomain of the actual, which in turn is a subset of the domain of the real (Bhaskar1978, Note to Table 1, p. 56; Bhaskar 1993, p. 207). We can represent these inclusionrelations in a Venn diagram (see Figure 2).

Figure 2 Bhaskars three domains: inclusion relations

Emergence and its basis

The second element of Bhaskars ontology with which this paper will engageis the stratification of the intransitive world into levels the atomic, the molecular, thebiological, and the like. This stratification depends upon the phenomenon ofemergence, which is most simply described as the relationship which makes itpossible for a whole to be more than the sum of its parts. Bhaskar himself definesemergence as the relationship between two terms such that one diachronically, orperhaps synchronically, arises out of the other, but is capable of reacting back on thefirst and is in any event causally and taxonomically irreducible to it, as society is tonature or mind to matter (Bhaskar 1994, p. 73).

I should point out that I am operating with a slightly different conception ofemergence from that which seems to be adopted by Bhaskar and some other criticalrealists. I argue that emergence is a property of entities, and that the concept ofemergence is inherently compositional, in the sense that higher-level entities always

Empirical

Actual

Real


4/15

D. Elder-Vass 4

emerge from collections of lower-level entities that are their components or parts 3.Although sometimes critical realists seem to adopt a compositional definition ofemergence (see, for example, (Collier 1998, p. 264)) at other times this is denied (e.g.(Bhaskar 1978, p. 169; Collier 1994, p. 116)). While there is not space to go into theargument more deeply here, it is worth enumerating two advantages of the

compositional approach. First, it maintains a clear relationship between emergenceand its primary theoretical usage: to maintain the tenability of a stratified view ofreality in the face of reductionist arguments, which assert that explanations in terms ofhigher level entities and their properties can be eliminated in favour of explanations interms of lower level entities. And second, it maintains a degree of simplicity andhence clarity to the concept of emergence that makes it easier for us to understand itstheoretical role.

But how is emergence possible? If we accept that emergent wholes are morethan the sum of their parts, but also that emergence allows nothing to be present in thewhole that is not present in the parts, then where does that more than come from?

I argue that it comes from the organisation of the parts, from the maintenanceof a stable set of relations between the parts that constitute them into a particular kindof whole (Emmeche, et al. 1997, p. 106; Lloyd Morgan 1923, p. 64). Thus it is thefact that a higher-level entity is composed of a particular stable organisation oflower-level entities that gives it the possibility of exerting causal influence in its ownright. In other words, it is the set of relations between the lower-level entities thatmakes them more than the sum of their parts. Only when this particular kind of partsis present in this particular set of relations to each other does the higher level entityexist, and only when this particular kind of parts is present in this particular set ofrelations to each other do they have the causal impacts that are characteristic of thehigher-level entity. As Archer puts it, Emergent properties are therefore relational:they are not contained in the elements themselves, but could not exist apart fromthem (Archer 1982, p. 475). Note that a higher-level entity is only emergent when it

just so happens that, when a set of lower level entities is so organised as to create it,the resulting entity has a consistent causal impact that is not a simple summation ofthe impacts of the its components. Now, the particular causal influences that anyparticular entity type may exert, and the way in which the presence of its parts in therequired relations produce these higher level effects, are a matter for the particularscience of the case we cannot go any further at the philosophical level in explainingwhy particular cases of emergence work.

We can go further, however, in identifying another general pre-requisite foremergence. As the existence of the whole is inseparable from the continuing presence

of the required parts in the required arrangement, then emergence itself depends uponthe set of causes that maintain a set of such parts in just such an arrangement. Bhaskarhas commented on the dual aspect of emergence as a synchronic and diachronicrelation; but the diachronic aspect of his account seems to relate to the originalcreation of the new level of reality. While this original creation is clearly necessary,the maintenance of the particular entities that constitute that new level is equallyimportant. There is not only a set of causes that brings the entity about, but also afurther (possibly overlapping) set that maintains its continuing existence what I willcall, after Buckley, its morphostatic causes (Archer 1982, n8, p. 480). It is these

3 Bhaskar sometimes uses higher and lower in the opposite sense to this, but I shall maintain theusage that lower entities are components of higher entities for the sake of consistency with mostother work on emergence.


5/15

D. Elder-Vass 5

causes that are responsible for the stability of the organisation that constitutes thehigher level entity from the composing entities.

The role of these morphostatic explanations of continuity of structure iscritical to emergence. Any number of implausible combinations of lower-level entitiesmay be brought about by a vast range of morphogenetic causes over the course of

time, but it is only those combinations that have continuity of structure that persist. Atany time, it is possible that a more powerful morphogenetic cause may overcome themorphostatic causes for any given entity. At this point, the emergence of the higherlevel entity is dissolved. It is the ability of morphostatic causes to resist such effectsthat sustains the existence of higher-level entities and hence any emergent propertiesthey may have. But this continuing existence is always contingent on the outcome ofthe ongoing tension between these different types of cause 4.

I therefore propose that we can describe emergence as the outcome of aprocess by which a set of morphostatic causes, which may be both internal andexternal, sustain a set of lower level entities in relationships that constitute them into astably organised higher level entity that can as a result exercise powers that are notpossessed by its component entities either in isolation or in an unstructuredaggregation.

With this account of emergence, we can now turn to examining the nature ofthe elements that inhabit Bhaskars three domains in a stratified world.

Experiences, events, and entities

Let me begin with experiences. There are two key factors that influence theshape of our experiences.

First, as Bhaskar tells us, Experiences, and the facts they ground, are socialproducts (Bhaskar 1978, p. 57). Experiences are social products because ourexperiences are not simply a set of sense-data, but rather the result of our applicationof a socially-influenced conceptual framework to the interpretation of that sense-data.Our eyes may detect a pattern of colours; but what we experience is seeing a set ofmeaningful objects behaving in meaningful ways. It is in this interpolation of ourconceptual frameworks between sense data and experience that experiences becomesocial products. Hence experiences are no longer purely the outcome of the eventsthey might appear to reflect, but rather a product of the combination of those eventswith our prior knowledge.

Second, our experiences, despite being interpreted, are constructed on thebasis of our sense-perceptions. Those sense perceptions are inevitably limited to

impressions of those segments of reality that we are capable of perceiving with thesenses we possess, as augmented by any artificial tools that are available.Now, the combined effect of this process of interpretation and our restricted

perceptual (and perhaps cognitive) abilities is that we generally perceive reality asflat in the sense that our experiences are interpreted as impressions of entities at asingle level of stratification. When we perceive the human being, we do notsimultaneously and inseparably perceive the organs, the cells, the molecules that makethem up. If we perceive the cells of a living tissue under a microscope, we do notsimultaneously and inseparably perceive the organ or the organism to which theybelong.

4 There is a useful role in the explanation of morphostasis for concepts like negative feedback fromcybernetics and strange attractors from complexity theory.


6/15

D. Elder-Vass 6

Thus, our experiences are already, through the process of abstraction that isinherent to perception, and as a result of the limited slice of reality to which oursenses give us access, single-levelled abstractions from what is in actuality aninherently multi-levelled occurrence.

Now, because of the nature of our experiences, our everyday (empirical)

concept of an 'event', which we take to be the naturally-given subject of anyexplanation in science, is itself an abstraction from reality. Thus, when we say, forexample, 'the pen fell on the floor', we are already, in framing our reportage of theevent, making an assumption about which abstraction from what was happening in amulti-level stream of interconnected happenings is the one that is relevant andrequires explanation. We could have looked at the same happenings and chosen toexplain the behaviour of the molecules or atoms involved, or the writing process orthe world historical events or the social history of which the falling of the pen formeda part. But in selecting out one of these happenings from the rest as the thing to beexplained, we have already created the illusion that this is an event that can be givenan explanation in its own right, independently of its component events and of thelarger events of which it forms a part. Let me label events conceived in this single-levelled way as level-abstracted events .

In seeing events as level-abstracted, we implicitly frame the question of howthey are caused in a way that demands explanations of a particular form - in a waythat pushes us into thinking in terms only of a particular stratum or level oforganisation. But any causal account of a level-abstracted event forms only part of alarger picture. A more complete explanation can always be provided by re-integratingthe level-abstracted event into the larger stratified picture of which it forms a part, andrelating the explanation of this 'event' to the explanations of the other level-abstractedevents in which it is inextricably implicated, either as subset or superset.

Now, I suggest that the way to make sense of causal explanations of individualevents in this context, where an event is defined as the behaviour of a given entity at agiven time, is to allow that in reality every event taken as an individual instanceinescapably includes the behaviour of the composing lower level entities as well(Lloyd Morgan 1923, p. 15). This allows us to ignore yet higher-level events of whichthis event is a part, since it is when we turn to the causal explanation of those higher-level events that we will deal with the implications of this particular event being asubset of them. Given that any given event may be a subset of an indefinitely largerange of other higher-level events, this would seem to be the only practical way ofdealing with the higher-level implications. I shall label events, conceived as includingthe simultaneous lower-level events which are its parts, as downwardly-inclusive

events. Hence, in explaining a downwardly-inclusive event, we recognise, forexample, that when a pen drops, it is inseparably part of this individual event that thecomponents composing the pen remain in a set of relationships through which theyconstitute the pen, and behave in whatever ways are required for the pen to drop. Thisis the inevitable consequence of the set of morphostatic causes whose operation mustbe present for the pen to exist as such through the entire course of this event. Thus, thevarious material parts of the pen go through a series of events that forms part of thehigher event, the molecules that compose those parts go through another series thatalso forms part of the higher events, and so on through the atoms, subatomic particles,and so forth. Given that we do not have fully adequate understandings of the lower

end of this spectrum, we must accept that only partial descriptions and hence only


7/15

D. Elder-Vass 7

partial explanations are possible of the lower-level set of events that composes thehigher level event.

If we wish to understand the role of emergence in individual events, and therelations between causes at different emergent levels, then, the correct account ofindividual events is inherently stratified. We need to recognise that the events which

populate Bhaskars domain of the actual are downwardly-inclusive and multi-levelled. This clearly corresponds to Bhaskars conception of the actual as thatdomain of reality in which a vast range of particular causes interact to cause events.And on my account the actual includes not only events that are unobserved by virtueof the absence of an observer, but also those levels of downwardly-inclusive eventsthat are unobserved by virtue of operating below (or above) the perceived levels ofreality.

Now, events involve the behaviour of things, or entities. Like events, we areaccustomed to perceiving entities in level-abstracted terms. But downwardly-inclusiveevents involve the behaviour of entities that are also defined in downwardly-inclusiveinstead of level-abstracted terms. A downwardly-inclusively defined pen includes itsmaterial components, its molecules, its atoms, and so on, and when we give ainclusive casual account of the dropping of the pen, we will be giving an account thatpresumes that the falling of the molecules, atoms, etc, is inherently part of that event,since these are inherently part of the entity that has been dropped. In this inclusiveontology, then, it is not only events, but also entities that must be treated as existing atmultiple levels all the way down. Both events and entities can be imagined now aspyramids, consisting of a single level-abstracted event or entity at the top, all of itscomponents at the next level down, all of the components of those components at thenext level down, and so on, until the base of the pyramid is lost in the mists of ourlimited understanding of quantum and sub-quantum science.

Where do entities fit in Bhaskars three domains? If events are constituted bythe behaviour of entities, and if (downwardly-inclusive) events belong in the domainof the actual, then it would seem clear that (downwardly-inclusive) entities must alsobelong in the actual. Indeed, at one point Bhaskar indicates that the domain ofactualities may be extended to include things as well as events (Bhaskar 1978, p.32). However, Bhaskar also identifies causation with relatively enduring structuresand mechanisms that are nothing other than the ways of acting of things (Bhaskar1978, p. 14); or to put it in other words, the generative mechanisms of nature exist asthe causal powers of things (Bhaskar 1978, p.50) (Lawson 1997, p. 21). Now forBhaskar, generative mechanisms exist in the domain of the real, but not in the actual,and this seems to imply that the same is true of entities.

One way to resolve this apparent contradiction is to suggest that entities existin these two domains in two different forms. Within the domain of the actual, entitiesexist as individual instances of things. But within the domain of the real they alsoexist as generalisations (transitively), or as types (intransitively). This is the form inwhich entities appear in our statements of causal law. Causal laws are generalisationsover many cases as well as an abstraction from multiple levels. The whole logic ofscientific experiment depends on the generalisability of causes if experimentationidentifies causal mechanisms, it is the belief in their enduring quality that rendersthem real rather than actual, and not only their enduring quality as ways of acting ofparticular things, but as ways of acting of whole classes of thing. Those entities thatare found only in the domain of the real are types of entity, natural kinds in fact

(Fodor 1974), whereas entities in the domain of the actual are specific instances ofthose things.


8/15

D. Elder-Vass 8

Furthermore, actual entities are inherently and inclusively multi-levelled, butwhen we label them in empirical experience, and also when we employ them in causalstatements, we typically abstract from most of those levels. Fleetwood seems to implysomething similar when he argues that when one writes that a mechanism has atendency to x , one is, in reality, referring to the ensemble of structures, powers, and

relations: it is, strictly speaking, the ensemble that has a tendency to x. Onceunderstood, however, there is no harm in shortening the phrase by omitting referenceto structures, powers and relations (Fleetwood 2001, p. 211). We can translate thisinto the language of emergence by equating ensembles with higher-level entitieswhose components are lower-level entities and the relations between them.Fleetwoods argument thus translates into the claim that we can allow level-abstractedentities to stand for downwardly-inclusive entities in causal explanations. Thus,entities in the real are both level-abstractions-from and generalisations-over entitiesin the actual.

Real causes and actual causation

Bhaskars account of cause in A Realist Theory of Science is focussed on therole of causal mechanisms, which, as we have seen, he identifies as part of the domainof the real (Bhaskar 1978, p. 13). As we have seen, these mechanisms exist as thecausal powers of things (Bhaskar 1978, p.50). In such situations, we can, asFleetwood suggests, work successfully with an abstracted ontology that ignores thefact that each entity or thing is composed of a variety of levels of lower entities, andsimply sees it as existing at a specific level of organisation. The composition of theentities we seek to explain (or use as causal factors) is simply one of the many thingsthat we abstract from in formulating our laws. It therefore seems in the resultinggeneralisations that the entities which cause and whose behaviour is caused arefree-floating level-abstracted entities that are autonomous of their component parts,and that can be treated in those causal accounts as if they had no component parts atall.

Such a level-abstracted conception of cause is perfectly usable and indeedpositively desirable in the process of formulating theoretical laws. It also works quitereliably in many practical applications, both everyday and scientific, when level-abstracted causation often seems to reflect what is going on well enough to provide uswith reliable explanations and hence expectations. However, as Bhaskar himselfrecognises in more recent work, it is quite inappropriate for the discussion of what ishappening over multiple levels when we turn to causation at the level of individual

instances:

unlike theoretical explanation in at least many of the natural sciences, viz. fromexplanatory significant structures to their higher-order structural explanation,applied explanation of concrete singulars, like changes in a particular structuratum,are a much messier affair. In a dialectical pluriverse an event e at a level L is aslikely to be (multiply) explained by elements at the same and lower-order levels inaddition to higher-order (deeper) ones, and/or even laterally, diagonally,tangentially by elements not locatable in the categorical or generic order at all(Bhaskar 1993, p. 133).

Let me now use an example to show why level-abstracted causal accounts areinadequate to the causal explanation of individual events over multiple levels.Consider an event in which one set of entities and properties B is considered to have


9/15

D. Elder-Vass 9

caused a subsequent set B* (perhaps the same entities with an altered set ofproperties). Let us further assume that there is some level-abstracted causalmechanism L that, when applied to B, leads us to expect or predict that it will lead to

B* . It is tempting to argue here that Ba in combination with L has caused Ba*, wherethe subscript a indicates that we are dealing with an abstracted definition at the

highest level, and as long as we do not intend to dig into the other strata of thisparticular event, this may well be a viable strategy.

But once we seek to account for B* i (where the subscript i indicates that weare dealing with an inclusive definition of B* , which includes all of its components,and their components all the way down), or to examine the relationship between Ba*and its lower level components (e.g. to discuss whether the set of components C a andC a* of B and B* had exercised a determinative influence on them), this strategybecomes untenable. The causation of B* as a multi-levelled individual instance canonly be understood if we see it as the causation of B i* by B i. In other words, it isimpossible to explain fully the causation of the event except as the outcome of acausal interaction between the whole pyramids, and not just the single points at thetop 5.

We can see why this is a useful way to look at causation if we consider theproblem posed to level-abstracted accounts by multiple realizability, i.e. in caseswhere the higher-level outcome is consistent with a variety of different lower-levelconfigurations. Such accounts are underdetermined, in that they can provide anaccount of how Ba caused Ba*, but not an account of how the other components of B i*were caused, hence leaving Ba* floating unsecured without any confidence in how itscomponents could have been brought to a state consistent with it . The inclusiveaccount, by contrast, resolves this underdetermination since the whole range of statesof all the component entities and sub-entities of B i is available to contribute to thecausation of the component entities and sub-entities of B i*. (This is closely related toBhaskars concept of multiple determination see below).

Of course, each of the interactions at the lower levels can also be considered asinclusive events in their own right, so the higher-level event is at least partiallycomposed of a whole series of smaller pyramidal events. Now a reductionist mightperhaps argue that the inclusive account suffers from the opposite problem: it mayseem to be overdetermined 6, if we believe that the higher level entities are no morethan the sum of their parts, and lower level explanations are available for thebehaviour of each of those parts. In this case, it would seem that causes at the higherlevel are redundant to the explanation of B i*, since the lower level causes do all thecausing that is needed to produce it.

But there are a number of problems with this reductionist claim. First, themeanings of some of the terms that describe the events to be explained may beincoherent at a lower level. Many animals can see things, for example, but theconcept of seeing is meaningless when reduced to the behaviour of organs or cells. Itis hard to see how lower level explanations of a concept that is meaningless at thelower level can be complete. Second, there will generally be features of B i* that arecontingent on the relations between its components, and not just their separate

5 This is a sub-case of the determination of events in the actual by a mix of many causes; and also acase of what Bhaskar calls multiple determination, which will be discussed below.6 I use overdetermined here to indicate a logically impossible case i.e. where the set of causallyeffective factors exceeds those required to explain the set of outcomes, with the result that they appearto mandate a set of outcomes that may be inconsistent with each other.


10/15

D. Elder-Vass 10

presences summed; and it is the addition of these relations as an ongoing feature thatdistinguishes the higher level entity from the mere collection of lower level ones. Acausal account of the lower level entities will not explain the higher level entity unlesswe go beyond that causal account to explain the set of relations between them too, andwhen we do this we have reintroduced the higher level entity into the explanation.

Third, even the separate presences of those component entities in a particular situationis often difficult to account for except as the consequence of their being part of thehigher level entity concerned. Why do we find this particular collection of lower levelentities or events present in the first place, and not some other, perhaps random,collection? Why, because it is precisely this collection that constitutes the higher levelentity and is held together by its morphostatic causes.

The causation of individual instances of actual events thus operates across thewhole pyramid of entities and sub-entities involved, not at a single level of it.Individual actual events, in all their multi-levelled glory, are the products of acombination of a variety of causal mechanisms operating on the prior state of the setof entities involved in the event. This individual instance causation (which is ofcourse interlinked with other individual instances of causation) occurs within thedomain of the actual.

It is not just the process of causation, however, that belongs in the actual. As Ihave argued above, individual entities, or things, belong within the actual. But asBhaskar insists, causal mechanisms are just the ways of operating of things (Bhaskar1978, p. 50). If those things are to be found within the actual, and also the eventscaused by the operation of their powers, then doesnt this suggest that the actualprocess of operation of their powers belongs there too?

The best answer would seem to be that individual instantiations of the causalmechanisms associated with particular entities do indeed belong in the actual, but that,like entities, the generalisation of those mechanisms into a type, a transfactual causalmechanism that will operate whenever certain circumstances are instantiated, belongsonly in the domain of the real.

These causal mechanisms, both in their real and their actual forms, take alevel-abstracted form; but they can only be applied in the actual in combination with amultiplicity of causal mechanisms from other levels of the ontological strata. Thusreal causation can be expressed in a level-abstracted form, while actual causationalways occurs in the context of downwardly-inclusive events. Transitively, we isolateeach real causal mechanism in thought, before we can understand how they areinstantiated and recombined to cause particular events in the actual (cf. Sayersaccount of the concrete (Sayer 1998, p. 123)). Intransitively, real level-abstracted

transfactual causal mechanisms will be instantiated into the actual whenever thecorresponding entities exist, but the causation of any actual (multi-layered) event willbe the result of the combined effects of a number of such mechanisms operating at thedifferent levels included in the event.

Multiple determination

Bhaskar himself addresses this question of the contribution of causes operatingat different levels through a concept which he calls dual control, multiple control,or multiple determination. In considering actual natural and social events, he argues,we must accept that different causal mechanisms and the interactions between them

account for different aspects of the events concerned, and that no single lawdetermines the whole result:


11/15

D. Elder-Vass 11

Laws leave the field of the ordinary phenomena of life at least partially open... Tosay that laws situate limits but do not dictate what happens within them does notmean that it is not possible to completely explain what happens within them. Thequestion how is constraint without determination possible is equivalent to thequestion how can a thing, event or process be controlled by several different kinds

of principle at once? To completely account for an event would be to describe allthe different principles involved in its generation. A complete explanation in thissense is clearly a limit concept. In an historical explanation of an event, forexample, we are not normally interested in (or capable of giving an account of) itsphysical structure (Bhaskar 1978, pp. 110-111).

Bhaskar makes the link to stratification explicit in a more recent work:Emergence makes possible the important phenomena of dual and multiple control (Bhaskar 1994, p. 75). It is precisely because the [actual] ordinary phenomena of theworld are inherently multi-layered, that we need to bring to bear different [real]single-layered causal mechanisms to explain different aspects of them. Thus

explanation at each level, in the area of autonomy left by the incompleteexplanations at other levels, requires a putatively independent science of that level(Bhaskar 1978, p. 114). And it is in combining all these level-specific explanations ofthe different levels of a particular multi-layered event that we completely account foran event. Although, of course, because we do not have viable sciences of every level,we can only produce incomplete subsets of the complete multi-layered account,which is why such a complete account can be seen only as a limit concept.

To put this in my terms: in decomposing the behaviour of a downwardly-inclusive entity across its ontological levels, it is the organisation that appears at eachlevel, the set of relations between the relevant lower-level entities, that is the 'extra'piece of explanatory information that appears at that level; and this is what makes the

'multiple determination' approach viable. We attribute a portion of the causalinfluence on a particular event to the level of organisation at the topmost level, aportion to the level at the next level down, and so on. This allows us to constructcausal accounts of multi-levelled single instance causation in which all the levels ofthe prior situation can have an appropriate influence on the various levels of theoutcome. In this model, any insistence on explanatory priority for any particularlevel becomes nothing more than a metaphysical prejudice.

It is worth noting that this conception of multiple determination is alsorequired by the transcendental argument from the nature of experimental science. Themost obvious causal regularity in experimental situations is the causal impact that theintervention of the experimenter has on the results of the experiment. Clearly there isa sense in which the experimenter causes the results of the experiment. It is onlywhen we have a concept like multiple determination that allows differentmechanisms at different levels to contribute to the determination of a multi-layeredevent that there is room for any other sort of cause to operate in experimentalconditions as well as the causal input of the experimenter. Since experimental scienceworks on the assumption that such other causes are in fact at work in experimentalsituations it also assumes that multiple determination is a feature of the world.

Muti-levelled causation of the actual, then, is an unavoidable feature ofBhaskars ontology.


12/15

D. Elder-Vass 12

The consequences for Bhaskars three domains

Let me now pull together and round out the implications of the foregoing forthe three domains of Bhaskars ontology.

There is relatively little to be said here about the empirical domain. In the

context of stratification, it is important to recognise that our experiences take the formof single-levelled abstractions from a multi-levelled reality. This is the result of thecombination of two factors: the inherent limitations of our perceptual tools, and theinterpretive habits that are integrated into the very process of perception. Now, it is ofcourse true that this form of perception is enormously effective in practical situations,or it would not have been favoured by biological evolution. And this effectiveness inturn derives from the fact that in many practical situations we can afford to ignorelevels of stratification other than those we are in the habit of perceiving. Level-abstracted perception and indeed level-abstracted approaches to the causation ofeveryday behaviour work well for normal human purposes.

But science seeks to go beyond this type of understanding of the world we livein, and in delving into other layers of our stratified world it reveals that there is moreto events than meets the eye. An event at any given level is inseparably also a seriesof events at lower levels (and may be a part of other events at higher levels). Thetheory of emergence enables us to see that if we want to explain a multi-levelled eventthen there will be a whole series of causal mechanisms involved, all operatingsimultaneously at multiple levels. If we wish to understand the relations betweencauses at different emergent levels we need to re-integrate these partial explanationswith the other levels that are inseparably part of the same event, in what I have calleda downwardly inclusive causal account. These multi-levelled individual events, theparticular multi-levelled entities whose behaviour they involve, and the multiplecausal mechanisms invoked, when they operate in practice as the powers of theseparticular entities all belong, I argue, within Bhaskars domain of the actual. Theeffect of my argument here, then, is to populate that domain with a variety of elementsthat we do not find in Bhaskars original conceptualisation of it.

The most radical consequences of my argument, however, are those for thedomain of the real. Here, too, I have added at least one new type of element to thepopulation of the domain, this time the concept of entity-types or natural kinds. But inthe process of examining the roles of entities and causal mechanisms a moresignificant point has become clear: that the new entities and mechanisms we find inthe domain of the real are of a radically different character from those we find in thedomain of the actual. On the one hand, actual entities and their powers are

downwardly inclusive individual instances; on the other, the kinds of entities andcausal mechanisms that are introduced when we come to the domain of the real arelevel-abstracted transfactual generalisations or types.

But in Bhaskars conceptualisation of the real, it includes everything in theactual, and also all those events and entities that could possibly happen or exist buthave not done so the unrealised. It seems, then, that he uses the domain of the real toachieve at least two different things (a) the introduction of negativity, in the shapeof unrealised possibilities, i.e. events that have not actually occurred, and (b) theintroduction of transfactuality, in the shape of causal mechanisms, and by implicationentity types. It is far from clear what the value is of combining (b) in the same domainas the actual, since they seem to be real in a very different sense; and I am tempted to

suggest that we could achieve greater clarity by using separate domains to achieve (a)and (b). We could, for example, call these a domain of the possible including both


13/15

D. Elder-Vass 13

the actual and the unrealised; and a domain of the transfactual, including causalmechanisms and entity types, but not the actual or the unrealised 7.

In this model, both the actual and the possible are populated with multi-levelled events and entities, and with causal mechanisms in the shape of the particularpowers of particular entities, while the transfactual is populated only by level-

abstracted entity types and generic causal mechanisms. Some might be tempted toargue that this transfactual domain is populated entirely by transitive entities theabstractions and laws we create to categorise and describe the entities and causalmechanisms that surround us. But this would be an error. Those entity types and thecausal mechanisms inherent in their powers continue to be instantiated independentlyof our knowledge of them. Entity types and causal mechanisms represent a real fact that there are certain possible configurations of lower-level entities that are stable andthat do have these powers whenever they are instantiated.

The relationships between these and Bhaskars domains are illustrated inFigures 3 and 4 below. Here, the empirical remains a subset of the actual, but theactual now becomes a subset of the possible, and the transfactual, although it isinstantiated by elements in the possible and the actual, does not overlap with either ofthem, since it is populated with types while the other domains are populated byinstances. Now in a sense this would not make a substantive difference to Bhaskarsontological arguments, since his domain of the real is simply the summation (or

joined set) of my domains of the possible and the transfactual, but I suggest it wouldlead to improved clarity in our thinking about the domain of the real.

Domain ofTransfactual

Domain ofPossible

Domain of Actual

Domain of Empirical

Mechanisms Types Instances Instances

Entities Types Instances Instances Events x x

Experiences x x x

Figure 3 An alternative set of domains: populating entities

7 Although the transfactual does include both actualized and unactualized entity-types and causalmechanisms. Unactualized entity-types (and their powers) are those that could possibly emerge buthave not actually done so hitherto.


14/15

D. Elder-Vass 14

Figure 4 An alternative set of domains: inclusion relations

This improvement has most immediate value when we apply depth ontology inthe context of stratification, as I have tried to do in this paper, and particularly whenwe seek to refute reductionist responses to stratification. When we seek to understandthe relations between causal mechanisms operating in/at different strata but allcontributing to the causation of a single event, we need to be able to distinguish thelevel-abstracted entity-types and causal mechanisms of the transfactual from thedownwardly-inclusive entities and causes operating in the actual. It is this distinctionthat is brought into focus by separating the transfactual from the possible.

Conclusion

I believe that many of the difficulties of existing approaches to emergence andreduction stem from the inappropriate application of a level-abstracted ontology. Ihave sought to demonstrate that Bhaskars depth ontology, as enhanced and, I hope,clarified in this paper, offers part of the solution to this problem. Conversely, I havealso sought to show that the careful study of emergence can enrich Bhaskarsontology.

On the one hand, depth ontologys division of actual events from realmechanisms is an essential prerequisite for the understanding of causation in a multi-levelled world. On the other, when we need to take full account of the stratification ofthe world, it seems that we need to see actual events and entities as inherently multi-levelled or downwardly inclusive, while we need to see real, or at least transfactualentity-types and causal mechanisms (as well as empirical experiences) as level-abstracted. It is thus only by combining causal mechanisms from a number ofdifferent levels that we can provide an adequate causal account of a downwardly-inclusive event. This seems consistent with what Bhaskar anticipates in his account ofmultiple determination.

Having said this, it seems to me that there is still much more work to be doneon this question of the relationships between and among levels in causal accounts, aquestion which is fundamental to understanding emergence and causation. Havingunderstood that emergence is important, we need to examine in more detail just how itworks, following in a tangential sense Bhaskars own advice: When a stratum of

reality has been adequately described the next step consists in the discovery of themechanisms responsible for behaviour at that level (Bhaskar 1978, p. 169). The

Empirical

Actual

Possible

Transfactual


15/15

D. Elder-Vass 15

outcome of such an exercise may well lead us to conclude that the higher stratum hadnot been adequately described in the first place. Hence examination of themechanisms of emergence may alter our perception of the nature of emergence itselfand its ontological role, as I hope I have already demonstrated in this paper.

References

Archer, M. 1982 'Morphogenesis versus Structuration: On Combining Structure andAction', British Journal of Sociology 33(4): 455-83.Bhaskar, R. 1978 A realist theory of science , 2nd Edition, Hassocks: Harvester Pressetc. 1993 Dialectic : the pulse of freedom , London ; New York: Verso. 1994 Plato etc : the problems of philosophy and their resolution , London: Verso.Collier, A. 1994 Critical Realism , London: Verso. 1998 'Stratified explanation and Marx's conception of history', in M. Archer, R.Bhaskar, A. Collier, T. Lawson and A. Norrie (eds) Critical Realism: Essential

Readings , London: Routledge.Emmeche, C., Koppe, S. and Stjernfelt, F. 1997 'Explaining Emergence: Towardsan Ontology of Levels', Journal for General Philosophy of Science 28(1): 83-119.Fleetwood, S. 2001 'Causal Laws, Functional Relations and Tendencies', Review ofPolitical Economy 13(2): 201-220.Fodor, J. A. 1974 'Special Sciences (Or: The Disunity of Science as a WorkingHypothesis)', Synthese 28: 97-115.Lawson, T. 1997 Economics and reality , London ; New York: Routledge.Lloyd Morgan, C. 1923 Emergent evolution : the Gifford lectures, delivered in theUniversity of St. Andrews in the year 1922 , London: Williams and Norgate.Sayer, A. 1998 'Abstraction: A realist interpretation', in M. Archer, R. Bhaskar, A.Collier, T. Lawson and A. Norrie (eds) Critical Realism: Essential Readings , London:Routledge.

elder vas.pdf.pagespeed.ce.yrcnhpsu 7

Documents