Do Explanation Formats in Chemistry Depend on Agent Causality?

Rom Harré


`Causality’ is the name for the relation between cause and effect. However, there is no agreement as to the nature of this relation. Is causality the production or generation of effects by material agents, the Aristotelian view; or is it no more than a regular concomitance between pairs of events of similar types leading to an expectation of the occurrence of a consequent event, as Hume argued? Or some more up to date version of the Humean notion of causality as a relation between events? The philosophy of chemistry has inherited this problem `space’. Should we follow the ideas of Aristotle or those of Hume? One would have thought that by the beginning of the third millennium the issue would have been settled. Recent attempts by followers of the Aristotelian plan have revived the concept of causal power. These attempts have included analyses of the logical form of attributions of causal powers (Cheng, 1978 and Hiddleston, 2005). However, it is well to remind ourselves of the origins of the traditional alternative to causal powers, the regularity analysis pioneered by David Hume (1739). Instead of causality as a process in which active agents exercise their causal powers, the regularity analysis is based on the presumption that causes are events. Can the use of causal concepts in chemistry be accounted for in the Humean way as referring to regular sequences of events, or does it need to include the role of causal agents?

Recent discussions of the concept of causality (for example Cheng, 1997, Hitchcock, 2001 and Hiddleston, 2005) seem to be based on two interrelated conceptual contrasts.

  1. Causes as ephemeral events, echoing Hume (1737); and causes as persisting agents (echoing Leibniz (1714), though it seems unknowingly.
  2. Causality as correlation (passivity) and causality as the exercise of causal power (activity).

One would have thought that there was simple one to one mapping between these pairs of concepts. Thus we would have correlated events and persisting powerful particulars, causal agents. Unfortunately, despite a superficial clarity, all the arguments of all three authors cover over a chaotic muddle of rival ontologies.

In addition, Cheng assumes that a third contrast between observables and non-observables maps neatly on to both 1 and 2 above. Since she uses the contrast between an empirical regularity and a theoretical explanation to locate the place for causal powers in her analysis, the psychological experiments of Michotte (1963) displaying observable exercises of causal powers are left out of her study of how people come to ascribe causality to sequences of events on the basis of the experience of causal activity. Nevertheless, there is something right about the thought that causal powers are unobservables, while exercises of power are routinely observable.

The difficulties of assessing the contributions these authors have made to this ancient debate are exacerbated by the use of algebra to express their insights. A close study of the papers by Hitchcock (2001) and Hiddleston (2005) shows that the equations with which causal relations are expressed do not contribute anything that is not already visible in the vernacular presentation of examples, mostly from medicine. Algebra can display the general forms of chunks of discourse, such as syllogisms. Its great value lies in its `computability’, the way that formal operations can bring to light new structures, as in Aristotle’s syllogistic Frege’s begriffschrift. The algebra used by the authors cited has no algorithmic value, and an examination of the formalizing moves shows that the level of abstraction is achieved at the cost of concealing a chaotic muddle of different ontological categories presumed by the various examples so encoded. I will demonstrate this below.

The Grammar of Event Causality

Before I turn to discuss some recent accounts of causality let us briefly return to Hume’s ingenious and powerfully reductive analysis and one of its modern descendents, John Mackie’s reduction of the concept of causality to the satisfaction of what he calls the INUS conditions on the antecedents of an effect. According to Hume (1739: Sect XIV) the causal relation as it is thought to hold between a pair of events, one the cause and the subsequent event the effect, includes three components. There must be regular concomitance in contiguity and succession between the types of events of which the cause event and the effect event are instances. Though Hume uses the word `object’ in the earliest formulations of his analysis, it seems clear that by this word he means events rather than things. According to Hume the elements of experience are impressions and ideas are their mental shadows. As well as regular contiguity and succession the idea of causality includes the idea of necessity, that is the idea that the effect event must follow the cause event, ceteris paribus. After pointing out that natural necessity is not the same as logical necessity, Hume offers his ingenious reduction of necessity to regularity. The meaning of an idea derives from a corresponding impression, so to understand what the idea of necessity means in the context of natural science we must locate the impression that corresponds to it. According to Hume it is the psychological state of expectation brought about by repeated experience of the correlations of types of events. `Necessity’ as a feature of the flow of events disappears in favour of regularity in contiguity and succession.

There must also be a second order regularity between first order regularities of contiguity and succession and the psychological state of expectation, since the former, according to Hume, causes the latter.

Two and a half centuries later Mackie (1974: 62) offers something similar, the INUS conditions on which he bases his analysis of causality in The Cement of the Universe (1974: 36). He argues that an event is to be taken as causal only in certain material circumstances, a field of relevant conditions.

To deal with the problem of specifying the link between causes and effects he introduces the quite undifferentiated idea of `the world runs on’ (1974: 55). Causal statements are generally taken to support or even entail corresponding counterfactuals. If it is true that the freezing of water causes it to expand, then it should also be true of some sample of liquid water that if were to be (had been) frozen then it would (would have) expanded. Such inferences seem to require that the possibilities in the situation are taken seriously, perhaps in planning some action. Mackie offers no rationale for delimiting the domain of possibilities upon which the truth conditions of a relevant counterfactual depend. `The world runs on’ is clearly inadequate.

In defining his version of causality as a feature of the conditions under which an event occurs, he writes that a state or event A is the cause or some part of the cause, if it is `an insufficient but non-redundant part of an unnecessary but sufficient condition (Mackie, 1974: 62). This yields the neat acronym `INUS conditions’. There may be several sets of conditions sufficient to produce an effect, so no one of them is necessary, that is without which the effect would not occur. However, among these conditions are some which are more relevant to a causal story than others. These are candidate causes. How the relative relevance of conditions is established in the absence of attention to causal mechanisms and natural agents is left mysterious.

For Mackie the ontological categories of causes must be found in what could count as a relevant condition for the coming to be of an effect. This is a vague specification since the conditions we advert to science and everyday life include such things as standing states, events, material agents, substances, and anything else natural science might favour. What determines the relevance of conditions could hardly be empirically observed correlations since without a rationale for the relevance of some event to the production of the effect it would not conform to the INUS requirements for identifying causes.

One could read Hume, and perhaps Mackie as well, as outlining the conditions under which we are entitled to surmise that a causal relation exists between an initial state of a system and a subsequent state rather than analyzing the meaning of the causal relation. This would leave room for a distinction between the criteria for picking out a certain pair of events as causally related and analyzing the content of the concept of causality. Eliding the content of a concept with its truth conditions is the defining mark of positivism. For all sorts of reasons we want to avoid being driven to adopt that stance. Some of the recent writing cited in this paper seems to fall back into verificationism.

Despite the fact that it is quite difficult to find Humean explanations in everyday discourse, many philosophers have taken event causality as paradigmatic of the concept of causality and some still do. Instead of saying that it was water that washed away the foundations, that it was aspirin that cured the headache, that is was the brick that broke the window, and so on, we transpose these explanations into the `event’ form. Then we have something like this: the impact of the water, the taking of the aspirin and the striking of the window brought about the effect event or state. However behind the advice one might give to someone complaining of a head ache, `Take an aspirin’, it is surely not a belief that the taking of the pill will cure the head ache. It is the aspirin, the stuff, acetylsalicylic acid, working away on the nervous system that does the trick

Why does any of this matter? However artificial the paradigm of event causality may seem when imposed on everyday discourse, it serves to support the denial of agentive powers to people. There are many philosophers who think that the discoveries of neurophysiology and genetics have rendered the idea of human agency redundant. The defence of the role of agent causality in explanations in recent years has been concerned very largely with the metaphysics of human action, with consequences for moral philosophy and philosophical psychology (O’Connor, 2002: 45). O’Connor, comments that Thomas Reid was sceptical of `the very idea of event causation’. Reid seems to have held that agent causation was given in human experience, and was extended from that origin to inanimate beings. Agent causation in the human case is the root idea.

The argument of this paper is rather the reverse. En passant, I wish to throw doubt on the claim made by some of those who would deny radical agency to people that physics does not make use of the concept of causal agency in explanations. Here is a spirited version of this claim:

How does an agent cause an effect without there being an event (in the agent, presumably) that is the cause of that effect (and itself the effect of an earlier cause, and so forth)? Agent causality is a frankly mysterious doctrine, positing something unparalleled by anything we discover in the causal processes of chemical reactions, nuclear fission and fusion, magnetic attraction, hurricanes, volcanoes, or such biological processes as metabolism, growth, immune reactions, and photosynthesis (Dennett, 2003: 100).

Much as I admire Dennett’s approach to cognitive psychology, this passage betrays a woeful misunderstanding of the ontology of chemistry and physics. What is supposed to be the event internal to a South magnetic pole that brings it about that it attracts a North pole? And what might the internal events be that bring it about that an electron displays a unit negative charge? How could there be photosynthesis without photons?

To see what has gone wrong here reflect on the conditions under which a heavy body falls in a gravitational field. True, an event occurs – the removal of a support – prior to the beginning of the descent, a process not an event. But it would be bizarre to cite the removal of the support as the cause of the body’s descent. The cause of the body’s downward acceleration is the potential energy of the gravitational field. Try removing the support from under a cannon ball in inter-stellar space. Nothing happens! Why? No gravitational field! Fearing we might want to bring back spirits, souls and other dubious beings, some philosophers slip back into the old positivist move of eliminating them by eliminating everything agentive from our ontologies by eliminating all that is unobservable. Of course causal powers are unobservable, just as their physical species, such as charges, field potentials and the like are unobservable.

It does not follow that a defence of the fundamental and indispensable role of agentive causality as it appears in the physical sciences provides any ready made answer to the question of human agency. It does follow that the alleged support from the way those sciences have developed for dispensing with the foundational role of powerful particulars in philosophical psychology and moral philosophy is chimerical. Explanatory patterns in the natural sciences are based on some version of agentive causality. The concept of `agency’ in the natural sciences, especially physics and chemistry, includes the root ideas of `spontaneity’ and `activity’. But agency in the human domain also includes the root idea of `free choice’ which is foreign to the ontology of the physics and chemistry.

From Hume’s writings of 1739 through Mackie’s proposal of INUS conditions in 1974 we can find the same inadequacies, when we try to match up their analyses against the intuitions we extract from the conceptual systems in use during that period in the natural sciences. There are two areas of inadequacy:

  1. The ontological status of causes and effects is not well studied, eventuating in a ragbag of very different kinds of beings offered as causes and as effects, at least things, substances, conditions and events!
  2. The link between cause and effect is not well studied either, with suggestions ranging from Hume’s psychological account to positivistic regularities of a variety of kinds.

Ehring’s (1997) study of causation does make use of the idea of an intervening mechanism which, triggered by a stimulus of a certain type, regularly brings about a subsequent type of event, the relation between the events meeting Hume’s criteria for causality. However, he makes no place for causal agents in his account so the efficacy of such mechanisms is left unaccounted for.

The Grammar of Agent Causality

I will try to show that a pair of events meeting the Hume/Mackie conditions can be certified as causally connected only if a plausible account of the production of the effect event, citing a causal agent, can be supplied. After thirty years it is pleasant to find a new generation of allies such as Cheng (1997) and Hiddlestone (2005) actively engaged in the project of the restoration of natural agency as a root concept for the sciences and as a key component in the causal insights of lay people. Unfortunately, both seem to slip back into a verificationist way of fixing the meaning of a key concept, as we shall see.

The ontology of the natural sciences, as they are actually practised, rests on the concept of a natural agent, that is a material being endowed with certain causal powers. I propose to demonstrate the plausibility of this intuition with examples from chemistry as my exemplary science. As remarked above, the generic concept of `power’ seems to involve two root ideas, `spontaneity’ and `efficacy’. The field potentials of an electric charge do not require anything to bring them into being. Furthermore they are displayed when constraints on motion are removed or absent.

My argument will unfold in two phases. In the first I offer an account of natural agency as it appears in the way we treat a wide variety of non-human entities or stuffs as causally efficacious. The second phase presents an analysis of causal regresses in chemistry, in modern times reaching through the level of chemical agents into physics.

E. H. Madden and I argued that attributions of event causality made sense only on the presumption of an underlying agent causality (Harré and Madden, 1975). There were material beings with powers to bring about various kinds of changes, events and new states of affairs. Our argument was based on the analysis of the concept of causality as it appeared in everyday thinking, but also the discourse of the physical sciences, though rarely expressed by the word `cause’ in that context.

Causal efficacy is ascribed to a powerful particular in the language of dispositions. `If such and such conditions are fulfilled, then such and such an effect will (is likely to) occur’. A dispositional ascription is a truth function of two empirical concepts, one representing the conditions and the other the effect. However, to answer the question `Why does this or that entity or stuff display such and such a disposition?’ we need to invoke causal powers. Causal powers are theoretical properties manifested but not displayed in observable phenomena.. `Causal powers’ are attributes of material entities, such as a plasmodium or a magnet, an engine or a force field. Some powerful particulars are observable, some are not.

The causal powers of natural agents usually persist in time. There are two patterns of persistence. The fact that this stuff in the bottle has the power to alleviate pain is true now, but lasts only until the analgesic is metabolized. The fact that this elementary charge has the power to attract other like elementary charges is true now and it persists through a great many interactions in which this material entity plays a part.

Causal powers are individuated in terms of their effects. For example, `Aspirin has the power to reduce pain’, that is it is an analgesic. Codeine too has the power to reduce pain, that is they display the same dispositional property, just as blood and stop signs are both red. Powers are attributes in just the way that `red’ is an attribute of flags and nose. However, the differentia of powerful particulars, the agents that have such powers, include the chemical structure of the molecules of each type of particular and the active route by which each brings about its effect.

Contemporary physics is grounded in charges as powerful particulars and their dispositions as associated fields of spatio-temporally distributed potentials. In so far as contemporary chemistry is grounded in contemporary physics, it too rests on a metaphysics of charges and fields. Nancy Cartwright has defended a similar metaphysics for physics based on the concept of `natural capacities’ (Cartwright, 1987).

According to Purvis and Cranefield (1995: 4) `There are two basic aspects of a causal agent model:

  1. The nature or structure of the agent itself, and

  2. The architecture of the system in which the agent operates.’

The first basic aspect covers two very different cases. A simple agent may possess the power to act in a certain way, as its defining or one of its defining attributes. The question by virtue of what internal structure does an electron have a unit charge has no place in physics. An electron is a simple agent. A complex agent may possess its powers as emergent properties of a more basic structure. Here is an example of the use of the concept `causal agent’ from medicine. According to the CDC Medical Dictionary (2005) the `Causal Agents [of malaria are] blood parasites of the genus Plasmodium.’ Their power to cause malaria is an emergent property of their anatomical/physiological structure which plays an essential role in the causation of the symptoms in the appropriate circumstances.

To what sort of material beings can powers be ascribed? The following taxonomic tree sums up my example-driven intuitions.

Causal Agent




The powers severally ascribed to such beings include the power to repel positively charged bodies, the power to cause the symptoms of malaria, the power to etch metal and the power to do mechanical work.

Proposing events for the ontology of causality renders the link between causal laws and counterfactuals mysterious. Events are ephemeral and so cannot serve to support counterfactuals, since these refer to possibilities rather than actualities. Causal powers endure whether exercised or not are so are well fitted to be an ontological basis for causality that makes the link between causality and the link with counterfactuals intelligible. The emphasis that Purvis and Cranefield put on the structure of the context in relation to the possibility of a power being exercised ties in with the requirement that a power be possessed by a powerful particular whether exercised or not.

An intriguing way of introducing possibility into the analysis of causation, that fits well with analyses that give priority to agency has been proposed by Belnap (2005). However, though he adds branching space-time to the repertoire of analytical tools nevertheless the ontology of causality on which the analysis rests is still Humean. His causae causantes are events. The causative events are picked out from all antecedent ephemeral states of the universe by Mackie’s INUS conditions. His account inherits the same problem of relevance. However, there is a more fundamental difficulty in that is very hard to see how an event could be a genuine originating condition, unless it were to appear spontaneously in the flux of time.

The endurance requirement on causal powers leads to the quest for an account of how a powerful particular can properly be said to have a power when not exercising it. In chemistry this quest leads to a hierarchy of structural hypotheses, which are taken to be the ontological basis for the powers of complex beings as emergent properties. Acidity is an emergent property of molecules which yield H+ ions in solution. These ions continue to exist whether or not the acidity of hydrochloric acid is being exercised on anything, for example zinc.

To return to the authors briefly introduced above. Getting behind the misleading façade of `directed acyclic graphs’ and the related `structural equations’ device (Cheng, 1997; Hitchcock, 2001). Hiddleston (2005) has made an interesting attempt at a comprehensive analysis of the concept of a causal power using these devices. It will prove instructive to test out his analyses of the main features of the concept as intuition reveals then in its use in chemistry, past and present. Chemistry is replete with causal power concepts, such as `elective affinity’, `valency’ and so on, and so an ideal test context for this analytical tool and the results of using it.

To make this discussion somewhat more critical, it also poses the question as to how far the analyses to be discussed depend on independent and unexamined intuitions of the authors. This might lead to a wider discussion of the point of using algebra in philosophy. If the algebra turns out to be a device for the perspicuous representation of intuitions then we must look through the algebra to examine the intuitions which lie behind it.

After a somewhat breathless encomium on the importance of the algebraic representation of some aspects of causality concepts Hiddleston (2005) proceeds to a critical examination of Lewis’s claim that the (or at least some) concept of causality is transitive. Thus, if A causes B and B causes C then A causes C. Clearly the Lewis claim assumes an event ontology for causality. Rightly Hitchcock centres his discussion around counterfactuals. Why should be believe the associated counterfactuals?

A great variety of items of very different ontological categories figures in his examples. In one he wobbles between `taking some pills’, an event, and `the pills’, causal agents, in setting out his example of an active route. It is the pills that prevent the pregnancy not the taking of them. That is no more than setting up the conditions under which the pills will actively do their biochemical work. Still expressing himself in event terms Hiddleston further develops the excellent idea of an active route between cause-event and effect-event. But when it comes to illustrative examples he abandons an event ontology for an agent ontology in two striking cases. In the case of the sharpshooter and the apprentice the problem of associated counterfactuals is resolved by introducing the moving bullet as the basis of an `active route’ between the cause-event and the effect-event, or in this case state, the being dead of the victim. Similarly in the case of someone who ducks out of the way of a falling boulder, but might not have done, both counterfactuals are supported by the boulder as persisting causal agent, which status it has whether or not the hiker ducked. The moving boulder is there as an active agent in both scenarios. The hiker is dead, and we say had he ducked he would have survived. The hiker is alive, and we say had he not ducked he would be dead. Of all the goings on in the vicinity of the hiker, why is the boulder so important? It is the powerful particular.

If no reductive analysis of `causal power’ into something else seems likely to be successful then the presumption that there are causal powers of active material agents underlying those sequences of phenomena we identify as causal chains must be reckoned a Wittgensteinian `hinge’ on which much of science turns, particularly chemistry and physics.

A Hybrid Structure

Cheng’s (1997) studies seem to suggest that there is no radical disparity between event-causality and agent-causality in actual cases of people reasoning causally. They tend to identify a correlation as causal if they think that there has been a causal power at work. According to scientific realism every reliable observable causal relation between events and conditions antecedent to an event taken as their effect, must be explicable by reference to a generally unobservable generative processes. Cheng summarizes her project as follows:

I propose that causal power is to covariation [among instances of event-types] as the kinetic theory of gases is to Boyle’s law. When ordinary folks induce the causes of events, they innately act like scientists in that they postulate unobservable theoretical entities (in this case the intuitive notion of causal power) that they use to interpret and explain their observable models (in this case their intuitive covariational model). That is, people do not simply treat covariation as equivalent to causal relations; rather, they interpret their observations of covariation as manifestations of the operation of unobservable causal powers, with the tacit goal of estimating the magnitude of these powers (Cheng, 1997: 369).1

Cheng’s thesis needs to be `bulked out’ to include the concept of `causal agent’ or `powerful particular’. People surely do not just postulate causal powers as unobservable entities, but material beings with causal powers. So, in addition to satisfying the Hume/Mackie requirements that an observable pattern of events must satisfy to be a candidate causal relation, there must be the presumption of a material, spatiotemporally continuous link between cause-event and effect-event. This is the role for the causal agent or powerful particular. For events to count as causes they must activate agents with causal powers. This is the principle on which the identification of certain covariations in event-types as picking out causal chains is based, whether in physics, chemistry and biology or in everyday life. This is what is missing in Cheng’s account.

How can one explain this omission? Recent revisions of Hume’s account of covariation between types of events abandons the implausible associationist theory that Hume himself proposed, for the idea that to take an event as a cause is to think that its occurrence increases the probability that a certain kind of event will subsequently occur2. This move leaves the tendentious presumption of `event causality’ intact. Cheng’s use of the probability reducing concept imports event causality into her treatment throughout. This leaves the concept of `causal power’ adrift with no ontological anchors.

Hiddleston’s recent article (2005) looks promising. Fortunately he sets out his insights both in the form of examples as well as in the idle rhetoric of algebra. In critically examining Hitchcock’s analysis he emphasises the role of an `active route’ presumption from cause to effect, but both are events. His concept of causal power is, correctly, ontologically located among properties. One would have thought that the next question would be `properties of what?’ In setting out his examples, he sometimes introduces material entities as powerful particulars, such as boulders, basket balls and pills, but in stories in which they are mixed up with an ontology of events. Yet, the introduction of these beings into the story can sustain counterfactuals only if they continue in existence, whatever else does or does not happen. For example, in discussing several examples in which powerful particulars are in action, the analysis slips back to events and worse probabilities.

In discussing an example of Cheng’s, Hiddleston (2005: 38 – 9) interprets her conclusion that it `attributes to smoking a “generative causal power”’. This derives from Cheng’s proposal that the generative power of an (unspecified) C is the probability that C produces E in the relevant circumstances. Now this is, to say the least, weird. The example is replete with causal agents, material beings with causal powers, such as asbestos. If we want to understand the generative power of asbestos to bring about cancer, we had better spend time on tracing the biochemical pathway, rather than doing statistical analyses of death rates. The generative power cannot be the probability, unless we have slipped back into neo-verificationism. A probability assessment might be a hint that some biochemistry would be in order in this case. A numerical measure never caused cancer in anyone.

The same line of argument appears in Hiddleston’s (2005: 42) analysis of `preventative power’. Roughly a preventative power `is the probability that P prevents generative cause C from producing E’. What then is P? It might be platform on which a cannon ball rests – it might be an antibody in the bloodstream, it might be a mosquito net in Panama, and so on. Each of these entities has the relevant causal powers by virtue of their material nature. Each answers the relevant `Why?’ question, such as `Why doesn’t the cannon ball fall?’ P is certainly not a probability! I go to the pharmacy. `I need something to stop my flea bites itching’. `Here you are’ says the pharmacist, `a nice probability.’ `But’ I remonstrate, `I do not want an abstract mathematical “object”. I want some Betnovate!’

Further analysis of Cheng’s and Hiddleston’s papers offers no further help. At best they have come up with an empirical criterion for the hypothesis that a causal power is at work. Neither has addressed the question of the status of the beings of which causal powers are properties. Neither has offered any kind of analysis of the concept of `causal power’, other than what we can glean en passant from examples.

Causality in Chemistry

In all the test cases to be discussed in what follows the concept of causality appears in the hybrid form. It is expressed in terms of hypotheses about the unobservable links between initial states of a mix of material substances of certain kinds, the reaction among which leads reliably to new substances which have come into being in the course of the reaction.

Newton’s `Force’ Hypothesis

Newton was reluctant to accept the idea of simple inter-atomic linkages. Of what stuff would these be made? After several false starts his account of chemical bonding and so of chemical processes in the Opticks, Query 31, expresses a view similar to that generally current today.

Have not the small Particles of Bodies certain Powers, Virtues, or Forces, by which they act at a distance … on one another for producing a great Part of the Phaenomena of Nature? For it’s well known, that Bodies act one upon another by the Attractions of Gravity, Magnetism, and Electricity; and these Instances shew the Tenor and Course of Nature, and make it not improbable that there may be more attractive Powers then these. …How these Attractions may be performed, I do not here consider. .. I use [the word `attraction’] here to signify only in general any Force by which bodies tend towards one another whatsoever be the Cause’ (Newton, 1730 [1952]: 376).

Since forces are continuously existing active beings, their causes must also be continuously existing beings. By mid nineteenth century the forces themselves had come to be regarded as the relevant causal agents.

Nineteenth Century Chemistry and the Electrical Theory

The spontaneity component of the content of the concept of `chemical agency’ is displayed in reactions.

Certain bodies, when placed in contact, exhibit a proneness to combine with one another, or to undergo decomposition, while others may be most intimately mixed without change. The actual phenomena of combination suggest the idea of peculiar attachments and aversions subsisting between different bodies. … A specific attraction between different kinds of matter must be admitted as the cause of combination, and this attraction nay be conveniently distinguished as chemical affinity (Graham 1850: 217).

Fownes (1856: 214) introduces the causal powers terminology as follows:

The term chemical affinity, or chemical attraction, has been invented to describe that particular power or force, in virtue of which, union, often of a very intimate and permanent nature, takes place between two or more bodies, in such a way as to give rise to a new substance, having for the most part, properties completely in discordance with those of its components.

Though this passage was written in mid-century, already by 1807 Humphrey Davy had made the connection between chemical affinity and electricity.

In the present state of our knowledge it would be useless to speculate on the remote causes of electrical energy. Its relation to chemical affinity is, however, sufficiently evident. May it not be identical with it, and an essential property of matter? (Davy, 1807: 39).

The final step in building something like the near-modern conception of chemical reaction came with the idea of `order of affinity’.

The affinity of bodies appears to be of different degrees of intensity. Lead, for instance, has certainly a greater affinity than silver for oxygen. … But the order of affinity is often more strikingly exhibited in the decomposition of a compound by another body (Graham, 1850: 223).

Here Graham completes the analysis of the concept of chemical attraction in terms of causal powers, adding `efficacy’ to `spontaneity’. However, we are still within the Newtonian framework in which `forces’ do duty as the relevant powerful particulars. What sort of entities are these?

From Forces to Fields

By mid nineteenth century the ontology of physics was changing from an essentially Newtonian atomism with inter-atomic forces towards charges and their fields. The concept of `field’ in more or less its modern form had been introduced by Gilbert (1600) as the foundations of a working ontology for explaining magnetic phenomena. The orienting power of the `orbis virtutis’ accounted for the way a compass needle behaved in the vicinity of a magnet, rather than the interaction of polar forces. Faraday’s elastic lines of force model brought the concept to life again. However, chemists were slow to adopt field concepts. Only in mid twentieth century do they displace force concepts altogether (Harré, 2004).

However, this displacement affects only the most recondite research. In practice, most chemists in most situations continue to make use of a theory of valency based on electric polarities that Berzelius would have recognised. This is elementary chemical theory, but instructive in the way it reveals the underlying ontology of the practice of chemistry, even into the third millennium. Discussions with one of the research chemists in my college have shown me how widely the electron model is used in the everyday work of chemists3

Covalency, for example in the union of two chlorine atoms to form the Cl2 molecule, is explained as the result of the sharing of two electrons, one electron from each atom completing the 8 shell of the other and so enobling both4.

According to Friend (1915: 170), theories postulating attractive and repulsant forces in explaining molecular structures and the reactions which give rise to them ought to be considered as the ground level of chemistry. Those theories which set out to explain the `actual causes of the chemical affinity postulated by the chemist’ are properly the province of physics. Even in 2005 something like this is still true, in that the majority of chemists are most likely to go no further than the electron theory leaving the formulation of the wave functions for multiple electronic configurations to the members of another `club’.

The situation is actually more interesting ontologically than this division of labour would suggest (Brown: 1975). To the ontology of charges and their fields that is realised in the electron theories of valency chemists added a third concept with ontological overtones, namely `energy’. For example ionisation energy is the energy required to abstract an electron against the attraction of the nucleus, which is complemented by the energy released when an electron is taken up by an atom or ion.

Forces have become fields and fields have been substantialised into energy. The final step along these lines is the sort of thinking that underlies the work of John Polanyi (Harré, 2005).

By mid nineteenth century the ontology of physics was changing from an essentially Newtonian atomism and inter-atomic forces towards charges and their fields. The concept of `field’ in more or less its modern form had been introduced by Gilbert (1600) as the foundations of a working ontology for explaining magnetic phenomena. The orientation of the `orbis virtutis’ accounted for the way a compass needle behaved in the vicinity of a magnet, rather than the interaction of polar forces. Faraday’s elastic lines of force model brought the concept to life again. However, chemists were slow to adopt field concepts. Only in mind twentieth century do they displace force concepts altogether (Harré, 2004).

Suppose we explain the existence of NaCl molecules by the Berzelian hypothesis that there is an electrostatic attraction between Na+ ions and Cl-. Why should this set up be stable? The solution is due to Albrecht Kessol (1853 – 1927). By losing one electron to the chlorine atom the sodium atom displays a net positive charge (a cation), and the outer shell of the remaining electrons matches that of the noble gases in having 8 electrons. By acquiring an electron the outer shell of the chlorine also emulates the electron configuration of a noble gas, and moreover is now negatively charged (an anion). So we have an account of the electrostatic forces and of the stability of ions within the same picture. This is `ionic or electro-valency’.

How would the Hume/Mackie way interpreting causality fare in chemistry? Take the case of a simple double replacement reaction. In symbols:

2NaCl + Pb(NO3)2 → 2NaNO3 + PbCl2

all in aqueous solution.

By considering electron transfer among the ions a new pair of +/- charged entities comes into being. The key is the transfer of powerful particulars which endow the ions with the Berzelian charge structure that accounts for their stability. Electron transfer leads to new products either because one of them is insoluble and precipitates, or the new products are more stable than the reactants were.

Similarly there are many cases of proton transfer, in which a hydrogen ion is relocated. Consider the reaction between ammonia and hydrochloric acid yielding ammonium chloride. The ion equation is something like this:

NH3 + H3O + Cl- → NH4 + Cl- + H2O

By deleting the Cl- on either side of the equation we get a net ion reaction, which reflects the transfer of H+ , a proton, from the complex water ion to the ammonium radical (Atkins and Beran (1989).

These transfers of electrons and protons as powerful particulars endow the ions with the necessary electrostatic charges to sustain Berzelian bonding. Covalency is more complicated, but the principle of referring bonding `forces’ to elementary charges is made use of there to.

To return to Hume’s criteria and Mackie’s INUS conditions; could their style of causal concepts be applied in the cases I have described, which are no more than routine inorganic chemistry? Well … the test tube must actually contain ammonia and hydrochloric acid brought together in solution. They could perhaps have been poured from separate vessels into a common container. The contiguity and succession conditions are clearly met since reaction occurs almost immediately and the reaction products are readily identifiable afterwards. Since this reaction always occurs, ceteris paribus, we might just admit the possibility that experience of this would lead to an expectation that it would happen again, Humean `necessity’, though most chemists have seen it only once or twice. An event, the mixing as the cause of the process that leads to the reaction products, would be a candidate cause. These are more or less Mackie’s INUS conditions, though we might just squeeze the reagents themselves into the story via the relevance requirement. Given the implausibility of Hume’s psychological analysis of causal necessity would a new version of the regularity theory in terms of probability make sense here? Do chemists mean by talking of reagents and products, their favoured causal concepts, that adding ammonia to hydrochloric acid would increase the probability that ammonium chloride would be formed? Of course not. But given the causal agents involved, they would, no doubt, conclude that the formation of ammonium chloride was very likely, but in a jokey kind of voice!

In all the test cases that have been discussed the concept of causality appears in the hybrid form, that is expressed in terms of hypotheses as to the unobservable links between the initial and final states of some mix of material substances, when there seems to be a good correlation between them.. However, these links do not conform to Cheng’s conception of a causal power as a probability.

In analyzing a case of the application of dynamicist metaphysics, the prime ontological problem is to locate the powerful particulars, the sources of activity. These are the `uncaused causes’. Charges satisfy the requirements for having this status in chemistry in the case of the electrovalent bond and the chemistry of ions perfectly. To the question `Why has this ion a net negative charge?’ the answer is found in the distribution of elementary charges. But the question `Why has this electron a unit negative charge?’ there is no answer. We have reached bedrock and `my spade is turned’.

Where Does All This Lead?

The discussion so far has been focussed on electrovalent compounds and the activity concepts needed to account both for formation reactions and for the stability of their products. However, one presumption was left on the sidelines, so to speak. As well as the final mystery of electron and proton charges, there is the puzzle of the stability of the `noble octet’, the eightfold `shell’ of electrons on the integrity of which the ionic story depends. The answer to that question is to be found in the properties of the wave function for that configuration, and hence in quantum mechanics. It will be something to do with minimum energy conditions.

However, most of chemistry is not based on electrovalency and the electrostatic model, but on covalency. In that configuration each constituent atom shares an electron with those from the other constituents, in the simplest cases to create a noble octet. So whatever is the binding force in covalency it is not electrostatic. At this point we move fully into quantum mechanics, and the theory of molecular orbitals. Since the underlying ontology is based on energy, a new metaphysical project opens out – the study of the energy concept. That takes us away from the philosophy of chemistry into the conceptual underpinnings of physics.



Belnap, N. (2005) `A theory of causation: causae causantes (originating causes) as Inus conditions in branching space-times’ British Journal for the Philosophy of Science. 56 221 – 253.

Brown, G. I. (1973) A New Guide to Modern Valency Theory. London: Longman.

Cartwright, N. (1987) Natural Capacities and their Measurement. Oxford: Oxford University Press.

Cheng, P. (1997) `From covariation to causation: a causal power theory’ Psychological Review, 104, 367 – 405.

Davy, H. (1807) `On some Chemical Agencies of Electricity’ Phil. Trans. Roy. Soc. 97 1 – 56.

Dennett, D. (2003) Freedom Evolves. London: Allen Lane.

Ehring, D. (1997) Causation and Persistence. New York and Oxford: Oxford University Press.

Friend, J. N. (1915) The Theory of Valency. London: Longmans Green.

Gilbert, W. (1600) De Magnete. London.

Harré, R. and Madden, E. H. (1975) Causal Powers. Oxford: Blackwell.

Harré, R. (2005) `The metaphysics of chemical reactions’. Foundations of Chemistry.

Hitchcock, C. (2001) `The intransivity of casusation revealed in equations and graphs’ Journal of Philosophy 98 273 – 299.

Hiddleston, E. (2005) `Causal powers’. British Journal for the Philosophy of Science. 56 27 – 59.

Hume, D. (1739) A Treatise of Human Nature. London: John Noon, Book One.

Leibniz, G. W. (1714) The Monodology. London: Everyman-Dent.

Mackie, J. R. (1974) The Cement of the Universe. Oxford: Clarendon Press.

Michotte, A. (1963) The Perception of Causality. London: Methuen.

O.Connor, T. (2002) Persons and Causes. Oxford: Oxford University Press.

Purvis, M. K. & Cranefield, S. J. S. (1995) `Causal agent modelling’ Unpublished paper, Victoria University, Wellington, New Zealand.



1 Admirable though Cheng’s approach is there are some serious shortcomings, in particular in her understanding of the Kantian proposal of schematisms (Wittgenstein’s `hinges’). Close study of her paper shows that it deals with the conditions under which the attribution of a causal power is reasonable, but has nothing to say on what causal powers are!

2 Surely the thought that an event is a cause if it increases the probability of an expected effect is a consequence of believing that the event is a cause, and not an analysis of the causal concept itself (unless, of course, one were under the spell of neo-positivism!)

3 I am grateful to Gareth McGuire for instruction in these matters.

4 The expression `noble octet’ for the stable electron configuration derives from the determining role this structure has on the properties of the `noble’ gases, argon, neon etc.