1. Introduction

Talk of mechanisms is quite common in organic, inorganic, and physical chemistry as well as biochemistry. Chemists take mechanisms to be one important way of explaining why reactions give the products they do, proceed at the speed they do, and lead to different products in different reaction conditions. My primary aim in this paper is to examine the usefulness of the mechanism approach (see Craver and Darden [Reference Craver and Darden2005] for an overview) in explicating the explanatory use of mechanisms by organic chemists.Footnote ¹ I address the following questions:

1. 1) What does explanation by means of mechanism in organic chemistry look like?

2. 2) Why do chemists consider mechanisms to be explanatory? and

3. 3) What is the nature and status of the explanation achieved?

To answer the first question, I outline a fragment of the theory of organic reaction mechanisms.Footnote ² I use this to answer the subsequent questions. At least in large parts of organic chemistry, chemists consider a mechanism to be explanatory because it accounts for the “dynamic process of bringing about” (Tabery Reference Tabery2004, 10) a change from reactants to products. The accounting is causal and most closely resembles the interventionist account of causal explanationFootnote ³ (Woodward Reference Woodward2003; Woodward and Hitchcock Reference Woodward and Hitchcock2003) for two reasons: the mechanisms are constructed via a series of interventions; and, due to a variety of explanatory challenges, succeed only in showing “how some possibilities depend on others” (Woodward Reference Woodward2003, 375). In important ways, then, the explanations offered by chemists are contingent.

My analysis in this article is limited in at least two important respects. First, the conception of mechanism I elaborate here is tied to a period lasting from the 1930s until about the 1970s or so. Explanation by mechanisms was rare before the 1930s. At the end of the article in order to support the further point about contingency, I comment briefly on developments after the 1970s. Second, except for a few parenthetical comments about thermodynamics, I ignore thermodynamic and quantum mechanical explanations, which are widespread in chemistry and do not involve a step by step accounting of entities and their productive interactions. Discussions of these issues are subjects for other papers.

2. Organic Reaction Mechanisms

For organic chemists, “a reaction mechanism is a specification, by means of a sequence of elementary chemical steps, of the detailed process by which a chemical change occurs” (Lowry and Richardson Reference Lowry and Richardson1981, 174; cf. Streitweiser and Heathcock Reference Streitweiser and Heathcock1981, 57). The elementary reactions are conceived of as simple substitution, elimination, or addition reactions. To keep matters simple, I focus the discussion below primarily on substitution reactions, in which an atom or a group bonded to a carbon atom is displaced by some other atom or group. A simple example of such is the reaction

where I stands for iodine, Cl for chlorine, R for a generalized carbon-containing substructure, and the minus for a negative charge. This is a substitution reaction because the iodine is substituting itself for the chlorine in the parent organic compound.

Chemists do not initially begin with a mechanism. Mechanisms are inferences based on observational data. Chemists begin with data about the speed of the reaction in different solvents, at different temperatures and at different concentrations of reagents. Collectively, this information is referred to as the reaction kinetics. In the kinetic method,Footnote ⁴ chemists follow how the speed with which the concentration of one chemical appears or disappears is affected by changing the concentrations of other reactants. The result is a rate law. As limiting cases, one observes either a second-order rate law, $\mathrm{Rate}\,=k_{2}[ A] [ B] ,$ where the rate is proportional to two species, or a first-order law, $\mathrm{Rate}\,=k_{1}[ A] ,$ where the rate is proportional to only one species.

From the dependencies discovered in the rate laws, chemists make inferences about the mechanism that would produce the experimentally observed rate law. They appeal to a set of elementary steps set out originally by Ingold (Reference Ingold1953).Footnote ⁵ In this version of theory, the elementary substitution reactions are designated as S_N2 and S_N1 reactions. “S” stands for “substitution,” the subscript N for “nucleophilic,” and the integers for the number of molecules that undergo bonding changes in the intermediate phase of the reaction. An example of an S_N2 reaction is the same one mentioned before, namely,

It is a substitution reaction. It is nucleophilic because the negatively charged iodine is substituting in virtue of its affinity for a positively charged center in the carbon substructure. A wide variety of evidence indicates that it is bimolecular, the primary pieces being that the rate law is second order and all such reactions involve a stereochemical inversion at the central carbon atom (imagine your umbrella inverting in the wind). These and other pieces of evidence are interpreted as two species undergoing bonding changes simultaneously in what is known as the transition state. The transition state is defined as the intermediate state lying between the reactants and the products. It is usually envisaged as a molecular complex that exists momentarily on the path from reactants to products. In the case of an S_N2 reaction, the molecular complex is a conglomeration of the two reactants.

In an S_N1 mechanism, the parent carbon compound first dissociates into R⁺ and an anion. Subsequently, the reactant anion joins with the R⁺ cation. This is still a nucleophilic substitution, but the step determining the overall speed of the reaction involves only the parent carbon compound, since no reaction occurs unless the cation is present. As a result, it is given the “1” designation, since the rate depends in the first instance only on the supply of R⁺. The rate law for such a reaction is first order (subject to complicating factors to be discussed later).

There is no general mechanism for a given reaction. Every reaction can be made to display at least two and very often more than two mechanisms. For instance, we can force a reaction occurring via the S_N2 mechanism to change over to the S_N1 mechanism. Take a series of substitutions, by different nucleophiles, in the same alkyl compound under similar conditions as to concentration, solvent, and temperature. The following ions: OH⁻, OPh⁻, HCO₃⁻, OAc⁻, Br⁻, Cl⁻, are arranged in order from strongest to weakest affinity for the carbon atom in the alkyl compound. The rate law for the reaction of the hydroxyl (OH⁻) shows that it reacts by the S_N2 mechanism. As one goes down the series, the rate of the reaction loses its dependence on the presence of the ion somewhere in between the phenoxide and carbonate. The mechanism has switched over to the unimolecular process. Under other conditions, the mechanism will switch partially or completely to an elimination, in which part of the parent compound is eliminated as a new, independent compound.

Four factors affect which series of steps the reaction will take: the solvent, the nucleophile, the leaving groups, and the substrate structure. With suitable knowledge, insight, and skill, a chemist can both explain and predict the course of a reaction by invoking these four factors. Ingold, the creator of this theory, hoped that this theory would be a priori. While the theory has never achieved this status, it did become part and parcel of the language of organic chemists by the early 1940s, only some 10 years after it was introduced. With some elaborations and refinements, it has remained.

3. Explanation and the Theory of Organic Reaction Mechanisms

Why do chemists consider mechanisms to be explanatory? Reaction mechanisms are one important way to reveal the “dynamic process of bringing about” (Tabery Reference Tabery2004, 10) chemical change.

Tabery (Reference Tabery2004) has examined two prominent analyses of mechanism and found them wanting as accounts of explanation. On Glennan's (Reference Glennan1996) account of mechanisms as state transformations, once the “property change is induced and the state transformation is accounted for, the cause is known and so the philosopher's job is essentially finished” (Tabery Reference Tabery2004, 10). For all we know, the transformation could have been effected in any number of ways. Thus, while Glennan's account is epistemically adequate, it does not suffice as an account of explanation. Likewise, on Machamer, Darden, and Craver's (Reference Machamer, Darden and Craver2000) account of mechanisms as organized entities and activities, it seems as if merely listing the entities, activities, and the spatial or temporal organization would suffice for an explanation. Again, while this may be epistemically adequate, it is not sufficient to say we have an explanation. For example, a listing of “dominoes,” “falling,” and “within one inch of each other in a linear array” does not tell us how the falling is productive. We need to be told how the falling is occurring and why that particular manner of falling can keep the mechanism going. The abstract list does not tell us.

A more satisfactory analysis of explanation by means of a mechanism can be recovered by invoking the notion of “dynamic production” (Tabery Reference Tabery2004, 12). Dynamicity removes the dangers inherent in talk of mere state transformations and in mere, bare activities because it supplies an account of how the state transformation is brought about and it requires talk of activities as productive in particular interactions. “The concept of productivity … becomes specified by reference to the particular property changes in the particular mechanism” (Tabery Reference Tabery2004, 11). In the process of specification, explanation arises because an account of how the property changes can be and are changes is supplied.Footnote ⁶

Tabery's analysis makes sense of the explanatory practice of organic chemists. Talking only of reactants transforming into products without an account of how the transformation occurs is taken by chemists to be the paradigm of a nonexplanation. It is equivalent to a student scribbling, “And then magic occurs!” on the examination. Likewise, simply saying that electrons have moved about is not taken to be explanatory. Electron movement is assumed, at least in this version of the theory, so saying that electrons are moving doesn’t explain in this context. We have to be told where and why the electrons are moving in order for the mechanism to be explanatory. Thus, the explanation is achieved by showing how the electronic activity manifests itself in the change from reactants to products: “the new bond is formed by co-ordination, and the old one broken by heterolysis.Footnote ⁷ … There is an electron transfer from the substituting agent Y to the centre of substitution in Alk, and from this centre to the expelled group X; so that, in consequence of the substitution, Y becomes formally one electronic unit more positive, and X one unit more negative” (Ingold Reference Ingold1953, 308). Note here the language of productivity, namely, the transfer of electrons in specific situations and how the transfer changes the reactants. Note also the explicit causal language, namely, “in consequence of the substitution.” While there is appeal to electron transfer, it is electron transfer from and to particular agents that is taken to explain the change; not every alkyl center can be a center of substitution for in every circumstance. In short, the explanation is given by showing how the movement of the electrons interacts with the molecules to dynamically produce the reaction products.

Chemists’ mechanistic explanations are apparently different than biologists’ in two ways: they are thoroughly theoretical, and they involve an appeal to powers or capacities. While many biological mechanisms do not involve theory (Bechtel Reference Bechtel2006, 2), mechanistic explanation in organic chemistry is theoretical in at least two senses: it is a theory, and it is underwritten by a theory. It is a theory in the sense that the mechanisms are general for all reactions, and the mechanisms explain the reaction phenomena. Further, in concert with considerations about the entities and their properties, the mechanisms support a number of generalizations about reaction phenomena. Finally, those other considerations—which are Ingold's four factors—support generalizations about when and where we should expect the mechanisms.

The theory of mechanisms is underwritten by various parts of the electronic theory of chemistry. To simplify greatly, the electronic theory of valence—at least in its 1920s and 1930s version—postulated that molecules like to have complete octets of electrons. This theory tells us why the entities are ‘wanting’ to add or eliminate electrons. Further, the property of nucleophilicity is underwritten by the Lewis acid and base concept. A Lewis acid is any entity that can act as an electron pair acceptor; a base acts as an electron pair donor. This is part of the electronic theory but is independent of the claim about octets; thus it is a separate theoretical underpinning.

The second unique aspect of mechanistic explanation in organic chemistry—and one that is crucial for explanation—is the reference to a variable power or capacity of the reagents. If a chemist says the reaction is a bimolecular substitution reaction, she is leaving out a crucial piece of information regarding how the change is occurring. A reference to the capacity must be included. This capacity is not simply the activity. The activity here is electron movement. The capacity tells us how the activity is manifesting itself, that is, how the activity is being productive in this context. Further, whether the electron movement is nucleus seeking or electron seeking “is not a simple and invariant property [of a reagent] but depends on the specific reagents, conditions, and solvent” (Streitweiser and Heathcock Reference Streitweiser and Heathcock1981, 169). That is, how the activity manifests itself must be specified by reference to the particular conditions that obtain in the reacting system.

In sum, organic chemists take mechanisms to be explanations. They do so because mechanisms provide a clear sense of why those reactants produced those products. That mechanisms do so is tied to theoretical considerations on a number of levels and to the capacities of molecules to display particular productive relations under certain conditions but not others.

4. The Nature of the Explanation

What kind of explanation is achieved by the appeal to mechanisms in chemistry? While there are many deductions made in the course of explaining a reaction, the explanation of the phenomenon as a whole is not deductive. Despite Ingold's hope, there are no general laws about the behaviors of organic reaction systems. The mechanisms are general, but one does not and cannot explain why those particular reaction products appear simply by citing the presence of a particular type of mechanism. Other factors affect how the mechanism plays out in that specific instance. And while the theory of reaction mechanisms does provide a great deal of unification among various types of reactions, it is not the unification itself that is seen as explanatory. Rather, the unification follows as a result of the mechanistic explanation. That leaves the causal account of explanation, and explanation in organic chemistry is clearly centered in this conception. As indicated in the earlier quotation from Ingold, nucleophiles cause the displacement of substituents. In particular, mechanistic explanation in organic chemistry looks like Woodward's ‘intervention’ conception of explanation (Woodward Reference Woodward2003; Woodward and Hitchcock Reference Woodward2003). Here “the central idea is that successful explanation has to do with the exhibition of patterns of counterfactual dependence describing how the system whose behavior we wish to explain would change under various conditions” (Woodward and Hitchcock Reference Woodward2003, 2). The generalizations that figure in such explanations are invariant in the sense that they will support ‘intervention’ counterfactuals, that is, counterfactuals that would hold when certain values of certain variables are changed. But they are not invariant in any more comprehensive sense, that is, there is no claim that the invariances will obtain for objects and conditions other than those described.

In addition, Woodward has hinted that this interventionist or manipulationist account involves a different attitude toward contingencies in explanations. I take him to be thinking along the following lines: both the nomic and unificationist accounts propose that explanatory activity reduces contingencies. If I say that the law of gravity explains why my pen will drop when I release it, I am usually entitled to think that the dropping is more than simply one possibility among many or that phenomena other than dropping will occur under new sets of conditions. “The manipulationist [or interventionist] account involves the apparently different idea that explanation is a matter of showing how some possibilities depend on others; it involves tracing some contingencies to others, but needn’t involve anything that looks much like a reduction in contingency” (Woodward Reference Woodward2003, 375). In other words, we understand why something has happened under those conditions, but whether the explanation will hold under new conditions cannot be taken for granted. We haven’t done the interventions under the new conditions yet, so we just don’t know if the explanation will be portable.

Mechanistic explanation in classical organic chemistry is interventionist and contingent in this sense. That the explanation is interventionist is fairly easy to substantiate. Even when equipped with the general theory of organic reaction mechanisms, chemists do not presume that the theory will provide all of the detailed causal happenings in a reaction or a series of reactions. These have to be explored with the mechanism in mind. Generalizations are built on interventions. Chemists are famous (or infamous) for writing long series of papers with titles such as “Mechanism of Elimination Reactions. Part XIII.” This might look like naive empiricism. But if one thinks in terms of interventions, one can see that it isn’t. Typically, each paper in the series takes up the next analogue in the series (e.g., elimination from tert-butyl compounds after a study of n-butyl compounds) or examines the kinetics in changed solvents or with different charged species.Footnote ⁸ Chemists are trying to see what series of counterfactual generalizations can be supported as one intervenes in the system. Each paper is designed to test or expand the generalization. As long as further chemicals and conditions are tied to interventions that have been studied or are similar enough to those that have been, the generalizations are taken to be explanatory.

Such an interventionist approach makes sense because the four factors form a causal network. Many influences compete in determining how a reaction will proceed. Teasing them apart in a given instance does not guarantee that they can be teased apart in the same way in the next instance.

Chemists are quite upfront about the contingent nature of the explanations built on the basis of the interventions. I believe this is because they are fully aware that they are dealing with a causal network and because they face a number of other challenges when constructing their mechanistic explanations. These challenges include: (1) the evidence for a mechanism is nearly always indirect; (2) the mechanism of a reaction is closely tied to the conditions that obtain; (3) mechanisms often compete. Taken together, these three lead to an explanation that suffices to explain the course of a given reaction. However, they also lead to an explanation that shows only how some possibilities depend on others.

The kinetics of a reaction is by no means a completely reliable indicator of the mechanism. There are a number of circumstances that might lead to first order kinetics even if the reaction is bimolecular. For example, if one of the reactants is in constant excess, which will be the case when the Y is a main constituent of the solvent, the concentration of Y will not vary and thus the rate equation will appear first order. Likewise, a reaction proceeding according to the S_N1 mechanism can have a rate law that makes it look like an S_N2 mechanism. This will be the case when the back reaction, the reformation of the reactants from the products, is significant. In addition, many intermediate situations arise, so that the rate equation often has no integral order and is not constant for the same reactants across different reacting conditions.

As in other experimental sciences (cf. Craver Reference Craver2002), much work is devoted to distinguishing the ‘true’ mechanism from all of the ‘false friends’ provided by the kinetic studies. Despite this, chemists are quite aware that the kinetic evidence may not provide us with the true rate dependencies. “No uniform one-to-one correlation between mechanism and reaction order is to be expected. … In judging the significance of reaction order, all the circumstances must be taken into account” (Ingold Reference Ingold1953, 314). In short, experimental evidence—kinetic as well as other kinds (cf. Lowry and Richardson Reference Lowry and Richardson1981, 175)—always leaves other mechanistic pathways as possible interpretations of the course of a reaction. If more than one fit can be found, chemists are usually happy to say that any one of the possibilities could be the mechanism.

Even if a chemist could get direct evidence of the operation of the mechanism under the given conditions, there are no general mechanisms for a given reaction. That is, in different conditions the same starting materials proceed along different paths to different sets of products. I illustrated this point earlier by discussing the series of substitutions by different nucleophiles to the same alkyl compound. R. B. Woodward, one of the synthetic chemists par excellence of the twentieth century, gave nice voice to this when he claimed, “There are no general reactions” (cf. Berson Reference Berson1999, 77). Which mechanism will be operative depends on the relative causal importance of four factors in the network.

To further complicate matters and make even more pathways available to the reactants, all four factors are kinetic, that is, they affect the rates of formation of the products. However, reactions are also under thermodynamic control. If there is significant equilibration among reactant and product molecules, such that the product molecules, once formed, have time to revert to the reactant molecules while the system is under observation, the product ratio will reflect thermodynamic rather than kinetic considerations. “Kinetic control might pertain under one set of conditions and thermodynamic under another” (Lowry and Richardson Reference Lowry and Richardson1981, 181–182). While the assumption of kinetic control can be checked, subject to certain limitations, the important point remains that different pathways will activated under different reaction conditions.

Finally, even under the exact same conditions, many reactions follow two or more paths at the same time. That is, mechanisms often compete. The standard example here is competition between S_N2 and E2 reactions. Both mechanisms involve a Lewis base, that is, the nucleophile, attacking a substrate, causing another Lewis base to depart from the substrate. If the energy differences in the transition state complexes are small, the paths are equally available. Another common example is shown in Figure 1. The first of the three products is produced via the S_N1 mechanism, the latter two by E2 mechanisms. Even though an unwanted product can be suppressed or eliminated with appropriate reaction conditions (cf. Lowry and Richardson 1981, 554; Streitweiser and Heathcock Reference Streitweiser and Heathcock1981, 171–172), the phenomenon of multiple, simultaneous mechanisms remains basic.

Figure 1.

In sum, I believe chemists are offering contingent mechanistic explanations built on interventions. Which mechanism predominates under a given set of reaction conditions can be more or less determined, but there is no general set of conditions we can lay down to explain the overall course of the reaction. As one leading chemist puts it, “All conceivable mechanisms seem to be available in a closely balanced array, and which of them dominates seems to be a delicate function of substrate structure and reactions conditions” (Berson Reference Berson1999, 102–103).Footnote ⁹ Such a situation is typical of causal networks (Thagard Reference Thagard1999, 113–114). Because of the interaction among the factors, the generalizations they support are only weak counterfactuals. The generalizations are riddled with (usually implicit) ceteris paribus clauses and are subject to any number of exceptions. But there are generalizations nonetheless.

Ingold hoped his scheme would be deductive, allowing chemists to escape the contingencies (cf. Hughes and Ingold Reference Hughes and Ingold1935; Ingold Reference Ingold1964). However, his hope has never been fully realized. Even as early as 1940, Hammett (Reference Hammett1940, 184) noted, the assumption that “like substances react similarly and that similar changes in structure produce similar changes in reactivity … requires so great an exercise of judgment, offers so wide an opportunity for the wisdom that comes only with experience and for the genius that seems almost intuition that there is some justice to the compliment or gibe, whichever it be, that this is an art not science.” There are many very specific trends in reactivity that can be captured by an understanding of the four factors, but one can get only trends, not precise understandings. Often this is because the four factors interact in surprising ways in the causal network. In short, the four factors provide only a rough and ready heuristic for prediction.

One might think that with new technologies this contingency would be eliminated. It has not. The advent of lasers and molecular beams, first in physical chemistry and increasingly in organic and inorganic chemistry, has led to a shift in the definition of mechanism but not to eliminating the tracing out of contingencies. While lasers and molecular beams allow greater specificity in the choice of reaction conditions, the increased specificity was accompanied by a decreased range of applicability of the definition of a mechanism. Lasers and beams allow one to react molecules together under very precise velocities and angles. For a chosen set of impact parameters, one can describe in great detail the mechanism by which the product is formed. However, if the velocities or angles are changed, an entirely new mechanism might be necessary. If you were to press practitioners of molecular beam studies as to what the mechanism of the reaction might be outside the carefully controlled conditions of the laboratory, they would probably offer an answer consisting of many parallel threads, each thread corresponding to one of the mechanisms observed in the laboratory.Footnote ¹⁰ Mechanisms are no longer a single specification of the process of chemical change; rather, they are summations of many individual processes, all of which have been discovered via interventions. The notion of a multiplicity of mechanisms for and thus contingency in the explanation of a single chemical reaction has remained even with the change to the new technology.

5. Conclusion

I have argued that, in order to consider a mechanistic explanation adequate or good, organic chemists require entities to be in productive, dynamic relationships. The claim of productivity is underwritten by appeal to a number of elements of classical chemical theory. Moreover, productivity for the organic chemist makes essential reference to a variable power or capacity of the entities. Productivity is discovered by a series of interventions. The mechanistic theory of organic chemistry is causal and appears most like Woodward's intervention version of that theory of explanation. The theory supports a number of generalizations tied to experimentally determined conditions. Change the condition, and the generalization may not hold. In addition, the nature of the evidence and the fact that reactions often precede simultaneously along many different reaction paths leads to the conclusion that the explanation given by means of the mechanism is, in important respects, contingent.

I take it as not insignificant that Woodward's hint about possibilities depending on other possibilities is substantiated in the explanatory practice of organic chemists. In addition to potentially helping philosophers sort out which of the competing models of explanation are more appropriate to a given situation, the example of organic chemists shows that the hint has legs. Contingency and explanation need not be at odds.

This examination of the practice of organic chemists allows us to characterize more closely what a mechanistic explanation looks like and what it needs to look like in order for this group of scientists to consider it to be explanatory. Further, it clarifies the nature of the explanation that is achieved. Whether either of these points can be extended to mechanistic explanations in other parts of chemistry and in other sciences is an open question.