1. Introduction.

Ellis (Reference Ellis2001, Reference Ellis2002), in his recent book Scientific Essentialism, and its companion The Philosophy of Nature: A Guide to the New Essentialism argues that certain essential properties of objects in the world not only determine the nature of these objects but also how they will behave in any situation. This project is an ambitious entry into the metaphysical discussions of natural kinds and the laws of nature. Ellis believes that our current scientific understandings of physics and chemistry suggest that our ontology should be built of irreducible properties, necessary connections and a modernized version of Aristotelian essences. He argues that scientific realists have every reason to regard inanimate objects in the physical world not as passive receptors of “the forces of nature” but instead as active agents. “Things behave as they do,” Ellis suggests, “not because of any external constraints that force them to, but because this is how they are intrinsically disposed to behave in the circumstances” (Ellis Reference Ellis2001, 3).

In this paper I will critique Ellis's essentialism from the perspective of the philosophy of chemistry. I believe such a project is warranted because one of the characteristic parts of this metaphysical project is its reliance on examples from chemistry. Historically, philosophers of science tend to focus on physics as the paradigmatic science, and more recently philosophy of biology has come into vogue. But Ellis not only uses chemical elements and molecules as paradigmatic examples but expresses his hope that one of the main philosophical implications of his project would be “the development of a realistic logic of chemistry … that did justice to the complexities of the subject matter, and which chemists themselves would find useful” (Ellis Reference Ellis2002, 139). I will argue that our current knowledge of chemistry in fact does not lend itself to essentialist interpretations and that this seriously undercuts Ellis's project.

2. A Brief Introduction to Ellis's Essentialism.

In the 1950s students of modal logic, including Kripke and Putnam, revived the essentialist debate that had been largely dead since the discrediting of Aristotelian natural kinds by Locke and Hume. Among the most important new defenders of the existence of essential properties and the reality of natural laws is Armstrong (Reference Armstrong1978, Reference Armstrong1983). He argues that “behind the set of ‘surface’ properties there is a set of more deeply hidden properties: the real essence of the natural kind. The laws which a thing obeys are in a sense part of its essence” (Armstrong Reference Armstrong1978, 66). Note that these essential properties are not individuating properties—objects that have the same essential properties are members of the same natural kind. This idea is the basis of Armstrong's defense of universals. Armstrong and Ellis can both be described as essentialists about natural kinds. Where Ellis diverges from Armstrong is in his description of natural laws (see Ellis Reference Ellis2005 for a helpful discussion). Armstrong argues that the laws of nature are best understood as interactions between universals. Such interactions are contingent upon circumstances but necessary if the right circumstances are in place.Footnote ¹ Armstrong is an essentialist about natural kinds but a nonessentialist about the laws of nature. For Ellis, however, the laws of nature are more than epistemological conveniences—they are ontologically necessary, for reasons that will become clear in the next section. It is at this point that most metaphysical critiques of his work are applied.Footnote ² I believe, however, that it is at the level of the essential properties, not the natural laws that they purportedly cause, that a critique of Ellis's work should begin.

Ellis wants to jettison what he calls a passivist view of nature in which “inanimate things are capable of acting only as directed—depending, for example on how they are pushed or pulled around by God, or by the forces of nature” (Ellis Reference Ellis2002, 1). But this is a new essentialism, and Ellis must strike a balance between distancing himself from both the Newtonian mechanist and the Aristotelian. One of the things which distinguishes new essentialism from Aristotle's, Ellis suggests, is “a much stricter conception of natural kinds … any two members of the same natural kind must be identical in all essential respects … essentiality implies necessity in the strict sense of metaphysical necessity” (Ellis Reference Ellis2002, 14). This concept of metaphysical necessity has two key properties:

The second half of this description is particularly important. Essential properties on this definition are those properties that are discovered by scientific investigation to be the necessary and sufficient conditions by which objects are grouped into natural kinds. Science, not logic or linguistic analysis, is the final arbiter on this issue.

Ellis defines three types of natural kinds, the first quite a traditional definition and the second two distinctive to his project:

• • Substantive: divisions of physical things, some more fundamental than others, that can be subsumed into broader categories arranged in a hierarchical structure, e.g., an electron is a natural kind, but so is an atomic particle and the latter includes the former as one of its types.

• • Dynamic: processes such as diffraction, ionization, and vaporization are natural kinds within this ontology.

• • Property: natural property kinds include both particular (spherical, mass m, charge q) and the broader kinds(shape, mass, charge).Footnote ³

One final distinction is key to both Ellis's project and my critique of it. In his third chapter, Ellis devotes considerable space to explicating a compound distinction: between intrinsic and extrinsic properties and essential and accidental properties.

• • The intrinsic/extrinsic property distinction: Intrinsic properties are roughly the same as Locke's primary qualities, “the properties of the objects themselves” as Ellis puts it. Extrinsic properties may interfere with our perception of the object but they should not figure into any definition of natural kinds. For example, position is considered an extrinsic property while atomic number and molecular weight are intrinsic.

• • The essential/accidental distinction is required to describe an important difference between the fundamental particles of physics and atoms and molecules at the chemical level. “For the simplest natural kinds, the intrinsic properties are all essential properties. For natural kinds at the atomic, molecular and higher levels, the intrinsic properties and structures always include some states that are accidental” (Ellis Reference Ellis2001, 55). For example, a chlorine atom whose electrons are in an excited state is as much chlorine as one whose electrons are in their ‘normal’ orbitals. This is an intrinsic accidental property.

Ellis uses this distinction to make it clear that while intrinsic properties can be both essential and accidental properties, no extrinsic properties are essential.

In summary, this new essentialism is an ontology in which the laws of nature and the properties that define natural kinds are necessarily connected. There is a hierarchy of natural kinds of objects whose members must share certain essential properties and may or may not share certain accidental properties. It is the job of science to observe the behavior of objects and thus discover what the properties of objects are and to distinguish the accidental from the essential. Within this ontology, natural laws are descriptions of how objects are intrinsically disposed to behave.

3. Critique of Scientific Essentialism from Chemistry.

In this paper I suggest that Ellis's examples of essential natures from chemistry do not conform to our best scientific understandings of the chemical world. Ellis thinks his new essentialism ought to appeal to chemists, and on a surface level it does. Chemists have what Ellis calls the most obvious examples of natural kinds—the elements—and of the utility and predictive power of categorization into natural kinds—the periodic table. The laws of periodic regularity provide a clear example of how intrinsic properties of elements affect their outward behavior. Probing deeper, however, it becomes clear that chemical elements and molecules are not good candidates for natural kinds in Ellis's sense. His paradigm examples of metaphysical necessity break down under scrutiny. Many of the properties Ellis suggests are essential—shape, atomic weight, atomic number, electron affinity—have serious problems. Whether any given property should be considered accidental or essential, and whether only internal properties are essential seems to vary depending upon which natural kind is under discussion.

First, consider the concept of metaphysical necessity. Ellis defines a metaphysical necessity as a state of affairs that could not be otherwise than it is and must be discovered by scientific investigation. As an illustration of metaphysical necessity, Ellis quotes Putnam's example of a twin earth on which there exists a substance with all of H₂O's properties, which is also called water, but is actually structure XYZ. Putnam uses this example to make a point about modal relationships in possible worlds, and it is not my purpose here to critique his theory of modality. More interesting for my purposes is the widespread use of Putnam's example, which makes it clear that most philosophers are under the impression that water is nothing more than a bunch of H₂O molecules. This is simply not the case. An individual molecule of H₂O doesn't have any of the observable properties we associate with water. A glass of water, pure as water can be, is better understood as containing H₂O, OH^–, H₃O⁺ and other related but less common ions, and even this is a vast oversimplification (if we could get truly pure water, which we cannot). Our current best understanding of the electron transfers that give water the properties we observe is a statistical average of ever changing interactions so complex as to be quite literally unthinkable. Indeed, the problem is “not that we are unsure which (distribution of types of) microstructure is the correct one. The point is that there is no one correct microstructure, because the microstructure depends as much on the context and functions just as another nominal essence would” (van Brakel, Reference van Brakel2000b, 80–81).

This is why chemists use the ‘mixture of ions’ model to describe water's macroscopic behavior. The only thing we can say about a glass of water that is not, strictly speaking, an error is that the average ratio of atoms in the glass is 2 H: 1 O and that it has the macroscopic properties of water. If there are other kinds of atoms in the glass, or if the ratio is other than $2\thinspace{:}\thinspace 1$ , then we do not have pure water. If the ratio is $2\thinspace{:}\thinspace 1$ but it does not have the macroscopic properties (pH, boiling point, etc.) of water then we have not water but a mixture of hydrogen and oxygen molecules in their elemental form. Chemical analysis and experiment deals with water not as an individuated molecule but in macroscopic quantities. All the typical observable properties of water—its pH, its density, its boiling and freezing points, its utility as a solvent, are dependent not upon its atomic ratio but the interactions between the dissociated ions.Footnote ⁴ Philosophers of chemistry have been arguing this point for at least 25 years; the fact that Ellis can blithely use this example in his earliest chapters suggests that they either aren't being read or haven't been understood (see especially Primas Reference Primas1983).

Moreover, it is important to note that water is by no means unique in this regard. If a molecule is defined as “the simplest structural unit that has the characteristic physical and chemical properties of a compound” as it is by Webster's dictionary, then no dissociating liquid, metal, salt, or electrolyte fits the definition. Van Brakel argues that:

All this is meant to establish that if only science can establish what the metaphysically necessary properties of water are, we know that “being H₂O” is not one of these properties. We must ask again, what the real essence of water is. Just what properties and structures are metaphysically necessary? In seeking possibilities we must consider Ellis's distinction between intrinsic and extrinsic properties and essential and accidental properties. A metaphysically necessary property must be neither extrinsic nor accidental. “Essential properties distinctive of a natural kind,” Ellis reminds us “must be independent of the histories, locations and surroundings of its members” (Reference Ellis2002, 51). To clarify this point, Ellis notes that where uranium was mined has nothing to do with the essential properties that make it uranium. But the case of water makes this distinction less obvious. In order to behave like water, the H₂O molecules and ions must be in contact. Compare this to the uranium mining example—here it seems that where the molecule is found (in contact with other molecules and ions of oxygen and hydrogen and not, for example, isolated in the interstices of a crystal structure) is an extrinsic quality. But it is nonetheless essential to for it to have the properties of water.

There are also serious difficulties raised by the essential/accidental distinction. Recall Ellis's example of recently mined uranium. He notes that if the electrons of the uranium were in an ‘accidentally’ excited state, this would clearly not impinge upon its essential status as an element. Where the essential/accidental distinction becomes less easy to draw, however, is with intrinsic qualities like the number of neutrons in a nucleus. Uranium-235 and uranium-238 are, in one sense, both ‘essentially’ uranium, as they share the same atomic number. But the extra neutrons give the atoms quite different properties—one is intensely radioactive and the other is stable. It is unclear in this case whether having a particular number of neutrons constitutes an accidental or an essential intrinsic property. This is a serious problem, not just for radioactive elements. Ellis argues that “the chemical elements and compounds constitute the most readily accessible system of natural kinds of substances, their properties are mostly their essential properties, and the processes they undergo in chemical interactions are all natural kinds of processes that display the essential properties of the substances involved” (Reference Ellis2002, 139). Atomic and molecular weights would seem to be among the most obvious candidates for essential properties. They affect the properties of the macroscopic substance and were the original means by which the elements in the periodic table were ordered.Footnote ⁵

So what are the essential properties of the natural kind, H₂O? Molecular weight is a problematic choice, because there are multiple weights for this compound, due to the existence of isotopes of both hydrogen and oxygen. Molecular weights of anything from 18 to 24 are possible given various combinations of the isotopes. It is possible that we could define each isotope of an element as a separate natural kind, but chemists, have chosen not to do so.Footnote ⁶ Oxygen atoms with weights of 16, 17, and 18 are referred to as the same element. It could be argued that this is just a linguistic convention favored by chemists; in which case isotopes do constitute distinct natural kinds. On this picture the element oxygen is the natural kind of which the ¹⁸O isotope is a subset. Molecular weight would thus be an essential property of the natural kind ‘isotope’, but not of the natural kind ‘element’. In favor of this view we can consider the isotopes of hydrogen. Because the addition of a neutron more than doubles the weight of the nucleus, this “heavy hydrogen” has distinctly different properties than the more common hydrogen atom and even has its own name, deuterium and its own chemical symbol, D. The boiling point of D₂O differs from H₂O by four degrees Celsius.Footnote ⁷ But D does not have its own place in the periodic table. Chemists have very good reasons for the convention of referring to isotopes of an element as the same natural kind—with the exception of hydrogen, isotopes of elements have identical chemical properties. They form the same compounds, exhibit the same electrochemical properties, have the same pH in solution and so on. It is notoriously difficult to isolate the isotopes of an element using chemical methods.Footnote ⁸ The regularity and predictive power of the periodic table depends not on atomic weight but on atomic number, meaning the number of protons in the nucleus of an atom. The essence of the definition of the element oxygen is a nucleus containing eight protons.

My point here is that if we are observing natural laws in order to understand what the essential properties of objects are and thus sort objects into natural kinds then we encounter a serious difficulty with the hierarchy of natural kinds. If the essential properties of an isotope are different from the essential properties of an element then what are the essential properties of a molecule? H₂O and D₂O are made of essentially the same elements but not of the same isotopes. It would seem that which properties chemists will deem essential to explain an object's behavior depends upon what behavior they happen to be observing. If you are interested in solubility, H₂O and D₂O are the same molecule. But if you are measuring boiling points it is clear that the essential property is actually atomic weight and you are dealing with two quite different substances.

Given the difficulty of identifying metaphysical necessity based on structural and dynamic properties, we might focus on natural property kinds, such as shape, mass and charge. After all, this discussion of the nucleus of the atom is somewhat misleading. Chemical reactions are, by definition, reactions between the electrons of elements and compounds. Interactions at the nuclear level are formally the domain of physics. The chemical properties of an atom are the result of the nucleus’ relative affinity for its electrons and their placement in the purely theoretical spaces called orbitals. If we return, once again, the H₂O molecule, we find that many of the macroscopic properties of water are at least partially attributed to the ‘bent’ shape of H₂O, which helps create the phenomenon of hydrogen bonding. Ellis believes that substances have an essential shape to which they always want to return, although circumstances may distort the shape we observe. He provides the example of a rubber band. A stretched rubber band has a shape but this is not the ‘intrinsic shape’ of the band; “its actual shape is best considered as being made up of two distinct parts: an underlying intrinsic shape, and a superimposed distortion” (Ellis Reference Ellis2002, 52). However, in his article on the distinction between primary and secondary qualities in chemistry, Ramsey (Reference Ramsey, Bhushan and Rosenfeld2000) argues that while the shape of molecules is explanatory of physical properties, it is not an essential property. The best explanation we have for chemical properties is that molecules are in constant, internal motion—the bonds twist, lengthen, contract and bend—and the degree to which they do this depends upon the context they are in. Ramsey concludes that “philosopher and scientist cannot and should not rely uncritically on the claim that microstructure is an essential property of substances … a particular shape does not ‘belong’ essentially to matter as a ‘basic’ physical feature. Whether a particular ‘structure’ can be imputed to a sample is a contingent matter” (Reference Ramsey, Bhushan and Rosenfeld2000, 122).

In light of Ramsey's work the question for Ellis is: what basis do we have for designating one particular shape as ‘intrinsic’ while all others are merely ‘superimposed distortions?’ It is possible to isolate an H₂O molecule within a crystal lattice structure and supercool it, with the result that its bonds twist, contract and bend very little. If this is the intrinsic shape of the H₂O molecule then it is hard to understand how the laws of nature which build up the macroscopic properties of water from the intrinsic essential properties can arise from it—because this is not the shape that the H₂O molecule has in contexts where it is neither isolated nor supercooled. This is a distinctly unnatural state for an H₂O molecule on earth. Suppose we choose instead to analyze the shape of any single H₂O molecule at 20°C, just one molecule, at one instant of time, in a bowl full of liquid water.Footnote ⁹ But then we have the problem that most of the molecules in the bowl are actually ions and all of them have bonds in motion. Either selection seems entirely arbitrary. One option is to say that the object has “a shape” but not “a particular shape” but this cannot be what Ellis means, because that would make shape explanatorily uninteresting,Footnote ¹⁰ and Ellis expressly selects shape as his paradigm example of a “causally intrinsic” property. A possible response is that for us to choose an intrinsic shape is arbitrary but this does not mean such a shape doesn't exist, only that science cannot determine precisely what it is. The problem with this suggestion is that Ellis defines metaphysically necessary properties as those that science can discover. If science in principle cannot know what the intrinsic shape of water is, then shape cannot function as a metaphysically necessary property within scientific essentialism.

When we examine Ellis's defense of intrinsic qualities it seems that Ellis is embracing an even more controversial stance; he moves from a scientific model to a metaphysical necessity. In defending the concept of causal intrinsicality, Ellis argues that it “is not only required for an understanding of essentialism, but it is also one that has a fundamental role in scientific theory.” He gives two examples of this “model-theoretic” type of scientific explanation.

It is somewhat odd that Ellis uses an example from the Newtonian mechanist world view, which is the very position he's trying to oppose, as an example of the kind of scientific theory best preserved by essentialism. Be that as it may, it is far from clear that Newtonian motion is in any way ‘natural’ since it leaves out all of the natural variables involved in actual motion. When our experience contradicts mathematical idealizations like Newton's ideal of projectile motion, one option is to attribute the ‘error’ to moderating effects—friction, the medium, and so on. The other, just as viable possibility, is that nature is simply not in conformity with ideal constructs. On this view, the picture of a perfectly spherically shaped earth, like a picture of a bent H₂O molecule, is an inaccurate statement we knowingly make in order to simplify our calculations. Statistically, the average of a vast number of H₂O molecules in various shapes seem to behave as if they were bent to a certain angle with bonds of a certain length, twist, and so on. But it is far from obvious that a mathematical idealization of a far more complex actual behavior should qualify as our basis for belief in a metaphysically necessary property. I would suggest that the ‘bent molecule’ picture, like the round earth and Newtonian projectile motion, is better viewed as a scientific model—a convenient fiction of scientific theory—than an essential property.

4. Why Physics Alone Isn't the Answer—Problems of Reductionist Essentialism.

I have argued above that while chemical elements and molecules and processes provide Ellis's primary examples for natural kinds, most of the obvious candidates for essential properties of these natural kinds have significant problems. One answer to my critique might be to deny the chemical properties are essential, and constrain essentialism to sub-atomic physical properties. After all, Ellis argues that “for the simplest natural kinds, the intrinsic properties are all essential properties. For natural kinds at the atomic, molecular and higher levels, the intrinsic properties and structures always include some states that are accidental” (Ellis Reference Ellis2002, 55). Several difficulties I identified in chemistry center around the problems of distinguishing accidental from essential properties. At the subatomic level, however, Ellis believes this problem is eliminated. Perhaps all that is needed to ground an essentialist ontology is to reduce chemistry to physics.

The problem with this argument is that it is not as easy to reduce chemical properties to quantum or statistical mechanics as physicists have tended to suggest.Footnote ¹¹ Certainly such reduction is currently impossible given the sheer size of the calculations involved.Footnote ¹² But there remains considerable debate as to whether chemistry is even in principle reducible to physics. Primas argues that if by reduction we mean “higher level theory together with its interpretation can be deduced from the basic theory, then chemical kinetics cannot be derived from quantum mechanics, the theory of heat has not been reduced to statistical mechanics, temperature is not a molecular concept, the model of hybridization of orbitals is not predicted by quantum theory … and so on” (Primas Reference Primas and Agazzi1991, 163). The literature on the reduction of chemistry to physics is vast and inconclusive.Footnote ¹³ For my purposes, however, the most important point is the problems this raises for essentialism. Ellis admits that the two key problems for essentialism are:

I have two objections to this: first, most “working scientists” are much less interested in the essential natures of things than whether or not their models of how nature behaves yield the results they desire. The identification of natural kinds is much more likely to be a philosophical project than a scientific one. Second, and more importantly, these “kinds of doubts and concerns” sometimes do seem to be irresolvable, for reasons that are less metaphysical than epistemological. Whether we define H₂O and D₂O as members of the same natural kind depends upon what properties of these molecules we wish to investigate. Organic chemists, examining the interactions between the billions of molecules in an aqueous solution, will define the essential properties of water as a natural kind quite differently from a physical chemist studying individual H₂O molecules in crystalline structures, and both will differ from a physicist studying which types of quarks make up the proton. Because different fields focus on different properties of the natural world, debate over whether isotopes or elements are the more fundamental natural kind began in the early years of the 20th century and has not been decided today. The most fundamental ordering of chemistry—the periodic table—is not derivable from quantum theory (see Scerri Reference Scerri, Bhushan and Rosenfeld2000). Mendeleev's periodic law is a marvelous tool for ordering the chemical elements based on atomic number, but it does not accurately predict atomic weights. Neither can it suggest which is the more essential property. Again, it depends upon what we want to know. Ellis has turned the two most difficult questions in the formation of his ontology over to science but it is not clear that scientists are particularly interested in the question and, where they are interested, that they are capable of providing the answers.

5. Conclusion.

There are elements of this new essentialism that make it a very tempting theory for philosophers of chemistry. Ellis takes scientific research as the font information about the structure of the natural world. He offers a simple account of natural laws from within causal realist metaphysics. Chemists are comfortable with the essentialist idea that the internal properties of a molecule are necessarily connected to the law like behavior it exhibits. But the key words here are “law like.” The “laws of chemistry” are approximations based on models the chemists know to be oversimplified.Footnote ¹⁴ Scientific essentialism relies on scientists to identify natural kinds and the essential properties of natural kinds, and a sophisticated examination of chemical properties shows just how difficult this is to do. Many of these same critiques apply to Armstrong's description of universals, because he and Ellis share key assumptions about the existence of essential properties.

After more than one hundred years of debate, it remains unclear whether different isotopes of the “same element” belong to the same natural kind. In practice, however, this ontological difficulty has had little impact on scientific research. In some situations, such as in ordering the periodic table, it is convenient to place U²³⁵ and U²³⁸ together. In other contexts, such as harnessing radioactivity, it is not. I suggest that in the end debating which properties of a chemical element are essential seems a less valuable project for science than making note of which properties are of interest in which context.