1. Introduction

William James developed an evolutionary objection to epiphenomenalism that is still regularly discussed today.Footnote ¹ Perhaps because the classic passage where he lays out the objection is so pithy, epiphenomenalists have offered responses that do not, I will argue, grasp its full depth. I begin with a brief history of James’s own encounters with epiphenomenalism. Then I make use of more recent theoretical tools from evolutionary biology to unpack his worry.

The most important source for James’s early views on phenomenal consciousness generally is a Mind essay entitled “Are We Automata?” (James Reference James1879).Footnote ² That essay’s attack on epiphenomenalism was directed at a host of authors who had advocated the latter view, including William Clifford and Herbert Spencer.Footnote ³ But chief among James’s targets was T. H. Huxley, whose 1874 Belfast address was a high-water mark for epiphenomenalism. That position had emerged as an attractive account of consciousness, particularly for a new generation of physiologists (like Huxley) who were attempting to extend the strictly mechanistic conception of a reflex arc to cover even voluntary action (Klein Reference Klein2018).

Although James seems to have been the first to use the term “epiphenomenon” in its philosophical sense (Reference James, Burkhardt, Bowers and Skrupskelis1890/1981, 133, 139, 1186; see Robinson Reference Robinson and Zalta2015), neither he nor his opponents used the word “epiphenomenalism” in published work, typically employing the phrase “automaton theory” instead. For convenience I will use this latter phrase interchangeably with our more current “epiphenomenalism.”

Hodgson characterized such a view as asserting that “states of consciousness are not produced by previous states of consciousness, but both are produced by the action of the brain” (Reference Hodgson1865, 278). Huxley would offer a famous metaphor: like the relationship between the sound of a steam whistle and an engine that forces steam through said whistle, conscious states are always products of bodily states, but conscious states never in turn make a causal difference to bodily states, according to epiphenomenalism (Reference Huxley1874/1894, 240). This view portrays humans as automata in the sense that all behavior is understood to be a mechanistic product of prior bodily states, with no causal intervention from consciousness. Consciousness thereby becomes “a simple passenger in the voyage of life,” as James would put it, something “allowed to remain on board, but not to touch the helm or handle the rigging” (Reference James1879, 1).

2. Against Epiphenomenalism

It is useful to distinguish two related objections James raised, one domain general and the other domain specific. The domain-general objection targets the notion that any conscious states could have been shaped by selection if epiphenomenalism were true. The domain-specific objection targets the notion that phenomenal pleasures and pains associated with “fundamental vital processes” could have been shaped by selection. The latter issue is now the more widely discussed.

The domain-general objection is that automaton theory depicts consciousness as a “mere supernumerary” in the sense that it could make no causal difference to the animal’s behavior. But then consciousness “would be useless” from an evolutionary standpoint, in that it could not itself have contributed to the animal’s reproductive success and so could not have evolved via natural selection (James Reference James1879, 3; also see Reference James, Burkhardt, Bowers and Skrupskelis1890/1981, 142). I take this part of James’s criticism to be uncontroversial. As we will see, the prevailing epiphenomenalist response today (which is to contend that consciousness evolved through evolutionary processes other than selection) does not try to resist this initial move.

James’s domain-specific objection targets epiphenomenalism’s prospects for making evolutionary sense of a narrower band of our phenomenally conscious experiences—that is, of a particular subset of our phenomenal pleasures and pains. James wrote that there is a “set of facts which seem explicable on the supposition that consciousness has causal efficacy” (and not explicable by epiphenomenalism; Reference James, Burkhardt, Bowers and Skrupskelis1890/1981, 146). The facts in question concern the link we typically find between what phenomenally feels good or bad and what physically benefits or harms us. I quote this widely discussed, classic passage (CP) at length:

Consider a subset of our phenomenal pleasures and pains—namely, those that natively go along with “fundamental vital processes.” Examples include unpleasant feelings associated with being burned, suffocated, or physically injured and pleasurable feelings associated with eating, drinking, and resting when needed. Epiphenomenalists accept that these life-essential, phenomenal pleasures and pains (LEPPPs) are effects of the beneficial and harmful brain states (BHBSs) with which they are natively associated.Footnote ⁴ But epiphenomenalists (like “Mr. [Herbert] Spencer”) cannot consistently say that they are adapted effects, or in other words proper etiological functions, of those BHBSs.Footnote ⁵ In other words, if epiphenomenalism were true, there can have been no selection pressure on any BHBSs to produce their associated LEPPPs. This is because if epiphenomenalism were true, LEPPPs could have no “efficacy” and so (recall James’s uncontroversial argument) could have made no difference to reproductive success. But LEPPPs have three features that do suggest that they were shaped by selection: our LEPPPs are (a) natively patterned (they have a characteristic “distribution”),Footnote ⁶ (b) those patterns are systematically linked with antecedent BHBSs (this is the “well-known fact”), and (c) the patterns are “universal” among humans. James concludes that epiphenomenalism cannot make sense of how our natively patterned LEPPPs (with their systematic connections to antecedent BHBSs) could have evolved.

There is a crucial but suppressed premise here: if a–c cannot be explained by appeal to selection, then their evolution cannot be explained at all. It is exactly this premise that today’s epiphenomenalists effectively deny.

What is now the standard response was first articulated by Jackson (Reference Jackson1982). He suggests that phenomenal consciousness could be a spandrel (in the sense of Gould and Lewontin [Reference Gould and Lewontin1979]), in which case the evolutionary objection to epiphenomenalism would be overcome. In my terminology, Jackson is suggesting that our BHBSs could have evolved to produce our native LEPPP patterns through processes other than selection. This response has been developed more recently by Robinson (Reference Robinson2007, Reference Robinson2014) and Robinson, Maley, and Piccinini (Reference Robinson, Maley and Piccinini2015).

Jackson is correct that selection is not the only cause of evolutionary change. Common examples of nonselective evolutionary factors include phyletic inertia, genetic drift, linked genes, pleiotropy, and developmental constraints. The latter three can produce so-called by-products, or traits that have proliferated because they happen to have been genetically or developmentally linked to another trait that itself has been selected. So when James claims that epiphenomenalism makes the evolution of our native LEPPP patterns inexplicable, did he simply overlook the possibility that those patterns could have been evolutionary by-products rather than adaptations?

The answer is an emphatic no. Elsewhere in the Principles James repeatedly explained other mental traits as evolutionary by-products. He also identified clear and broadly acceptable criteria for distinguishing between by-products and adaptations. When one reads James’s domain-specific objection to epiphenomenalism in light of his criteria for distinguishing by-products from adaptations, a much more powerful objection emerges than has been appreciated in the literature, namely, that our BHBSs have phenomenal effects that have clear adaptive hallmarks, according to widely accepted standards in biology, and so contra epiphenomenalism, those effects are highly likely to have evolved through direct selection.

So what features of traits are adaptive hallmarks for James? His criteria for identifying likely adaptations are on especially clear display in his explanation of the evolution of the morbid fear of heights, which he takes to be a by-product.

James offers two reasons for thinking acrophobia is likely a by-product (or what he often calls an “incidental peculiarity”).Footnote ⁷ First, “individuals differ enormously” with respect to the trait. Rather than approaching ubiquity in the human population, as he thinks one would expect if the trait were a product of selection, the fear of heights “varies so much from one person to another” that it is likely to be a by-product. Second, James suggests that the trait’s “effects” lack any clear “uses” that would outweigh its costs. If natural selection had designed our psychological response to heights for climbing utility, one would expect that response to be “level head[edness],” perhaps dampened by a twinge of “timidity,” not the near paralysis we find in cases of acrophobia. In short, James expects traits that are products of selection to be ubiquitous and to be well designed, perhaps in the sense of producing effects whose uses outweigh costs.

These criteria are actually reflected in CP, albeit obliquely. Recall that I already identified three features of LEPPPs that CP identifies as evidence of adaptation. What I called feature c is just the “universality” of a trait in a population. And James might plausibly take features a and b (LEPPPs are natively organized into patterns and systematically linked to antecedent BHBSs) as amounting to the kind of natural design that only evolves through selection.

But what is it, specifically, about features a and b that might suggest evolution by selection? And in any case, how does James’s argument fare with respect to more recent biological practice?

One way we now establish that some trait is a likely adaptation (particularly when we cannot yet identify a specific etiological function) is to appeal to a set of features one expects to find in a trait that has evolved through selection. The standard list comes from Williams (Reference Williams1966, 10), who suggests looking at a trait’s “precision, economy, [and] efficiency.” That list is sometimes extended, with Symons (Reference Symons1990, 429) adding “complexity, and constancy with which effects are achieved” and Andrews, Gangestad, and Matthews (Reference Andrews, Gangestad and Matthews2002, 496) adding “specificity, proficiency … reliability of development, [and] complexity of design,” to give two examples. I will call these features “adaptive hallmarks.” James was effectively highlighting two adaptive hallmarks that are still recognized today as evidence of adaptation: complexity and precision.

Let us look more closely at the first of these. Complexity of design is still widely thought to constitute evidence of evolution by selection (Symons Reference Symons1990, 429; Andrews et al. Reference Andrews, Gangestad and Matthews2002, 496). In fact, that complexity is a hallmark of adaptation is a point of basic biology one can find in respected textbooks on evolution (e.g., Futuyma Reference Futuyma2005, 250).

Futuyma offers the pecten as an illustrative example. This is a highly complex structure in birds that protrudes in front of the retina and is supplied with an extensive network of blood vessels (for an illustration, see Futuyma [Reference Futuyma2005], 262). Because of its complexity, biologists have long believed that the trait is an adaptation, even before they had a good explanation of what the specific, adaptive function of this structure might actually be. Futuyma writes: “Even if we cannot immediately guess the function of a feature, we often suspect it has an adaptive function if it is complex, for complexity cannot evolve except by natural selection” (261–62). I will now argue that feature a (LEPPPs are natively organized into patterns) amounts to the kind of complexity that biologists still think can only evolve via selection.Footnote ⁸ We can get an initial sense of this phenomenal complexity by consulting Grant Allen, whose account of “the distribution of our feelings” of pleasure and pain we saw James approvingly citing in CP.

Allen documented and classified a rich diversity of pleasures and pains. His taxonomy of pains, for example, finds a place for discomfort associated with “the amputation of a limb, the excision of an ulcer, … the removal of a scalp,” “wounds, cuts, pricks, and scratches,” “burning off a finger, having the feet frozen so that the joints drop off, destroying the skin and muscles with a corrosive acid,” “par[ing] or break[ing] the nails below the quick, … pull[ing] open a sore, … hav[ing] the face or lips chapped,” cases in which “portions of the body waste away in eating sores, such as abscesses, cancers, ulcers, whitlows, &c,” “corns, bunions, bedsores, and lacerations,” “intestinal pain,” “the passage of renal calculi, gall stones, or clotted catamenial discharges,” “sprains, cramps, and spasms,” rubbing “salt or pepper” on “a wound or burns,” attempts to “to tear off the nails, to flay alive, to pull out the hair, to draw a tooth,” “mustard and cayenne pepper in excess,” “very loud sounds,” “fatigue after muscular exertion; mental weariness; inanition from want of food; faintness from anaemia, loss of blood, sleeplessness, or over-exertion; weakness from fever or other depressing disease; nervous debility; and those undefinable organic feelings which result from general ill-health,” among others (Allen Reference Allen1877, 6–9, 11–15). His taxonomy of pains gives order to a surprisingly diverse set of phenomenal states, in short, and similarly with his taxonomy of pleasures (which can be found at Allen [Reference Allen1877], 20–26).

We should count an apparent pattern in our phenomenal states as adaptively complex (i.e., as bearing the adaptive hallmark of complexity) if and only if the pattern satisfies two criteria. First, the pattern must involve a variety of phenomenal state types. Allen amply documented such variety in the case of our native LEPPPs, as I illustrated in the prior paragraph. Second, the pattern must be stable in the sense that relations between the patterned state tokens must be isomorphic not only when other tokens of the same types recur in one person but also when tokens of the relevant types occur in different persons, regardless of cultural context. For example, it cannot just be that a severe burn happens to feel worse than a hangnail here today but that my preferences might change tomorrow or might have been different had I been raised in a different culture.

We must admit that patterns among pleasures and pains that are not life essential do not count as adaptively complex precisely because those patterns do not meet this second criterion (i.e., they are not appropriately stable). For instance, music appreciation or the love of drunkenness can vary dramatically from person to person (some but not all prefer music to silence, some but not all prefer harmonious to cacophonous music, etc.).Footnote ⁹ So the strongest version of James’s argument against epiphenomenalism must confine its scope to life-essential pleasures and pains.

And when it comes to LEPPPs, Allen documents patterns that are remarkably stable. To use two of his examples, virtually all persons would find the pain of passing a “gall stone” to be more severe than that of having the “lips chapped.” He also makes his point by asking readers to “remember the sharpness of the smart which is produced by pulling out a hair or tearing a small piece of skin off the back of the fingers, and then contrast its volume with that which is experienced on breaking a limb or undergoing a severe surgical operation” (Allen Reference Allen1877, 13). Breaking a limb feels worse than pulling out a hair to virtually everyone, in any time or culture. These patterns in the “volume” of our life-essential phenomenal pains appear to be stable in the relevant sense, and the same presumably applies to life-essential pleasures. Of course, we have already noted the impressive variety of phenomenal state types involved in such patterns. So, just as the complexity of the pecten suggests evolution by selection, there is excellent reason for thinking that the complex “distribution” of our LEPPPs suggests evolution by selection as well—a point epiphenomenalists cannot accept (per James’s domain-general argument).

What is it about feature b (LEPPPs are systematically linked to antecedent BHBSs) that suggests evolution by selection? The adaptive hallmark in play, here, is what Williams calls “precision.” I will give an uncontroversial example of adaptive precision before returning to the case of LEPPPs.

Consider the Australian Stylidium bicolor, or triggerplant. The sexual structures are housed in a column that normally rests below the surface of the flower (for an illustration, see Futuyma [Reference Futuyma2005], 321). The petals are shaped in such a way that when a pollinator lands, its weight “triggers” the column to swing up and tag the insect with its pollen-covered tip. The precise delivery of the pollen depends on the column having exactly the right length and weight and the petals having exactly the right shapes, surface areas, thicknesses, and so on. The only evolutionary mechanism that can plausibly explain this sort of precision is selection (Williams Reference Williams1966, 10; Symons Reference Symons1990, 429; Andrews et al. Reference Andrews, Gangestad and Matthews2002, 496).

So, the triggerplant is shaped in such a way that it precisely delivers pollen to visiting insects. Can BHBSs be understood as producing LEPPP patterns with similar precision? Again, we do well to consult Allen, who did not just document complexity in our LEPPP patterns. He also identified a remarkable “concinnity,” as he aptly called it, between BHBSs and the LEPPPs they produce. He wrote: “the consciousness of Pain or Discomfort bears somewhat the same relation to other conscious states as the physical fact which underlies it bears to other conditions of the system” (Allen Reference Allen1877, 20). Of two pains, the one that is relatively more phenomenally distasteful is also very likely to have been caused by a bodily state that is relatively more threatening to the physical organism, according to Allen (also see 17–18). I suggest that James had precisely this concinnity in mind when he cited in CP the “well-known fact that pleasures are generally associated with beneficial, pains with detrimental, experiences,” for Allen himself had put the concinnity claim in remarkably similar language (29).

Importantly, Allen pointed out that we could identify this concinnity “without entering into the ultimate question of the connexion between physical and psychical states” (Reference Allen1877, 19–20). He meant that one can accept concinnity without begging Huxley’s question: whether conscious states are causally efficacious. For consider a causal claim that neither interactionists nor epiphenomenalists dispute: that BHBSs reliably bring about their characteristic LEPPPs. If Allen’s concinnity claim is correct (and I think it is), then epiphenomenalists and interactionists alike should both accept that there is a kind of quantitative precision with which BHBSs bring about their associated LEPPPs. In particular, a brain state that registers relatively more (or less) harm to the organism (e.g., a brain state that registers a broken limb as compared with a state that registers a yanked-out hair strand) will tend to produce pain states that are more (or less) phenomenally repugnant.

Epiphenomenalists and interactionists alike must also recognize a qualitative precision with which BHBSs bring about their associated LEPPPs. We experience the distinctive qualitative discomfort of a sprain if we try to walk on an injured ankle and a distinctive urge to drink water when dehydrated; we feel the distinctive urge for sexual gratification when we have been aroused and the distinctive craving for salt when sodium levels in our blood drop; we experience distinctive bodily relief when we overcome constipation and distinctive chills when we have a raging fever; and on and on.

One could perhaps believe that some pattern of neural flickering happens to produce some one phenomenal property as an accidental by-product. But it defies the standards of evolutionary biology to assert without argument that nonselective forces might have produced the spectacular precision—both quantitative and qualitative—with which our BHBSs bring about our LEPPPs.

To recap, I take James’s attack on epiphenomenalism to work like this. BHBSs produce a complex effect (i.e., stable patterns among a variety of LEPPP states) and to produce that effect with both quantitative precision and qualitative precision. This multifaceted complexity and precision strongly suggests evolution by selection. But if epiphenomenalism were true, these effects could have made no causal difference in the life of any organism, and so the effects could not possibly have been selected. Hence, epiphenomenalism is empirically implausible.Footnote ¹⁰