Fixing Content and Function in Neurobiological Systems:

Keywords: adaptationism, Daniel C. Dennett, electric fish, electroreception, evolution, evolutionary function, indeterminism, mental content, neuroethology, sensory modality, underdetermination

Introduction

In his 1987 book, The Intentional Stance, Daniel Dennett defends a position concerning attributions of function (in biology) and content (in psychology) according to which neither is determinate. Content and function attributions are the result of a process of “retrospective radical interpretation” (283). In psychology, we begin our explanations of behavior, according to Dennett, by assuming the rationality of the agent and then attributing content (i.e., intentional states such as beliefs, desires, intentions) to the agent according to a strategy he calls the “intentional stance”. Similarly, in biology, we begin our explanations for the existence of behaviors and structures by assuming the optimality of natural selection and then attributing functions these structures and behaviors in the process commonly known as “adaptationism”. Dennett writes, “The problems of interpretation in psychology [determining content] and the problems of interpretation in biology [determining evolutionary function] are the same problems, engendering the same prospects–and false hopes–of solution, the same confusions, the same criticisms and arguments” (1987: 277, emphasis in original).1

Dennett argues that, in biology, “We take on optimality assumptions not because we naively think that evolution has made this the best of all possible worlds, but because we must be interpreters, if we are to make any progress at all, and interpretation requires the invocation of optimality” (278-279). Dennett offers a similar analysis of the rationality assumption in psychology. He writes,

We cannot begin to make sense of functional attributions until we abandon the idea that there has to be one, determinate, right answer to the question: What is it for? And if there is no deeper fact that could settle that question, there can be no deeper fact to settle its twin: What does it mean?

(319)

Dennett nicely sums up his own position: “It is not just that I can't tell, and they can't tell; there is nothing to tell” (312). Content and function are indeterminate because there simply is no “fact of the matter” in either case. For Dennett then, attributions of content and function are intimately related, and for both he draws the same indeterminist conclusion.

In this paper, I begin by accepting Dennett's claim for the close relationship between content and function attributions. Instead, I challenge his characterization of these attributions as indeterminate and unfixable. I will argue that, contrary to his analysis, attributions of content and function are determinate, at least insofar as it is relevant to scientific attributions of both. There are facts of the matter, I claim, and we do make such determinations. To support my argument, I will consider the case of the twentieth century discovery of electroreception. Electroreception is relevant because this non-human sensory modality–the ability to perceive the environment via electricity–was only discovered this century, and yet the existence of this modality commands a strong consensus among present-day scientists. That is to say, there is today a strong scientific consensus that certain organisms possess anatomical structures which have the function to process electrical information about the organisms' environments. (It is notable that other proposed non-human modalities, such as vertebrate magnetoreception, lack a similarly strong consensus.)

Electroreception is a sensory modality and thus provides an added advantage, vis-a-vis Dennett's claimed relationship between content and function, in that it cuts across the distinction between content and function. Attributing electroreception to particular animals simultaneously involves hypotheses about both content and function. More specifically, it proposes that these animals possess structures, undergo processes, and exhibit behavior that have the function of bringing about a particular kind of contentful connection between the world and the organism. A claimed discovery of a new modality in certain fish and other organisms is a claim about the evolutionary function of particular structures those animals possess. It is the function of these structures to act as sensory organs to process electrical information, just as it is the evolutionary function of hearts to circulate blood and the function of birds' wings to allow flying.

The electroreceptive hypothesis is also a claim about the kinds of sensory content that such organisms possess. Sensory systems are, by their very nature, perceptual content production and manipulation systems. To claim that sharks, say, are genuinely electroreceptive is to claim that these organisms are capable of representing a particular kind of environmental information, i.e., electrical information.2 Nonetheless, given Dennett's arguments for a connection between content and function attributions, then if I can undermine either the argument for content indeterminacy or the argument for function indeterminacy, the remaining claim will as a consequence meet the same fate. That is to say, since Dennett argues that attributions of function and content are indeterminate for the same reasons, then (by his lights) undermining one set of arguments should simultaneously undermine the other.

In their own attempts to undermine Dennett's arguments, some (Kitcher and Kitcher 1988; Amundson 1988) have pointed out a curious mismatch between his conclusions and day-to-day scientific practice. Far from being mired in a morass of indeterminism, biologists seem to fix the evolutionary function (or functions) of biological structures, processes, and behavior all the time. Biologists take as their ken, such questions as: Why do male guppies have spots? Why do fish and birds flock? What does this structure do in this organism? In answering these questions, biologists often make reference to the concept of evolutionary function; guppies have spots, say, because male guppies with spots (of a particular type) have historically had a reproductive advantage over those which did not.3 The actual story is no doubt a great deal more complicated than this simple statement, but as shorthand for the long-winded, causal-historical description, we say that male guppies have spots in order to attract females; the function of the spots is to attract females.

Dennett offers three counter-arguments to such claims. First, function and content determinists generally fail to tell stories in enough detail to refute the thesis of indeterminacy. Dennett (1988b) points out that, “The guppy example is supposed to exhibit a case in which careful biological research yields determinate (but complex) attributions of function. Why then do [functional determinists] refrain from concluding their tale by telling us exactly what the function (or functions) of those guppy spots is and is not?” (540). Dennett is here setting a high standard. Philosophical “toy examples” will not do; neither will hand-waving. I accept Dennett's challenge and, in Section II, I provide a highly detailed account of the scientific history of the discovery of electroreception. This discovery was the result of an interdisciplinary investigation–an endeavor best characterized under the rubric “neuroethology”, the biological study of the neural basis of animal behavior.

Second, Dennett (1988b) notes that it is no argument against indeterminacy arguments that some attributions are, “obviously wrong, and it is no argument in favor of functional determinacy that some functional attributions are obviously right (the eagle's wing is obviously for flying, and the eye for seeing)” (540, emphasis in original). Setting aside worries about this claim's cogency (see Note 5, below), let me stress that the possession of electroreception is far from obvious. Indeed, it is so non-obvious that even though humans have been aware of the existence of electric fish for at least three thousand years, it is only during the last century that the possibility of electroreception has been imagined.

Third, Dennett (1988b) gives reason to believe that a longer and more detailed story–even one packed with causal details–would still fail to answer his philosophical objections: “There is causation to be taken into account at all times by the psychologist, and by the biologist. Neither, however, has succeeded in telling a watertight causal story that licenses functional, or referential, interpretations” (541, my emphasis). According to Dennett, no matter how much causal detail is amassed, there will always be room for multiple interpretations. To address this worry, a careful analysis of the components of the neuroethological case for electroreception will be developed in Section III. This will demonstrate how various kinds of evidence are necessary for such attributions, but also how each, on its own, is insufficient. The point of this analysis is to show how specific kinds of scientific detail both constrain the space of possible explanations and offer support for specific interpretations. Then, in Section IV, I argue that such a story is as “watertight” as it needs to be in order to fix content and function.

The metaphysics and epistemology of content and function

What are content and function, such that they are either determinate or not? In this paper, I endorse the standard etiological accounts of content and function. The primary contemporary proponent of the etiological account of both content and function is Ruth Millikan. In two influential books, Millikan (1984, 1993) has argued for a causal-historical approach to psychology and biology–an approach she calls the theory of “Proper Function”. For Millikan (1993), content and function are best understood in terms of evolutionary history: “A proper function of ..... an organ or behavior is, roughly, a function that its ancestors have performed that has helped account for its own existence” (14). To use one of her favorite examples, what makes it the function of a heart to circulate blood is that it is in virtue of its blood-pumping properties (and not, say, its weight or color) that organisms which have possessed hearts have differentially successfully reproduced in competition with organisms whose hearts did not possess the same blood pumping properties. According to Millikan, function is a matter of an organism's evolutionary history and is not due to current properties or propensities.4_

A similar story can be told about content. Dretske (1981, 1988, 1995) and others have defended a causal-historical account of content. On this account, a particular physical or neural state has a particular content if that state has the appropriate causal connection to the appropriate state or states of the world. In many respects, this account of content parallels the etiological account of function. On the causal account of content, a state has the content it has because of its place within a particular causal history. According to Dretske, content is a matter of an organism's causal history and is not due to current properties or propensities.

On both of these accounts, what it is to have a particular function or content is to have a particular causal history. This account is fine as far as it goes. It is limited, however, to the theoretical or metaphysical nature of content and function. Such characterizations tell us what it is to be (or to have) a function or content. The central issue here has a distinctly more epistemic flavor. My central question is: Whatever content and function are, can we ever know them? Can we ever discover, in Dennett's phrase, “some fully determinate truth about what things mean, or what their functions really, truly are” (1988b: 540)? Borrowing a phrase from Peirce, how do we fix content and function (if content and function are the sorts of things that can be fixed)?

To date, the philosophical work on content and function has mainly focussed on the metaphysical issue, and less attention has been given to the epistemic standards of evidence required to make content and function claims, supposing that some version of the etiological theory is correct. Notice that Dennett's claim about indeterminacy involves both the metaphysics and epistemology of content and function. He claims that content and function are epistemically unfixable (“we can't tell”). However, the reason for this epistemic situation is his claim that content and function are metaphysically indeterminate (“there is nothing to tell”). For Dennett, the metaphysics and epistemology go hand-in-hand. It is the nature of content and function that they are unfixable.5

My strategy is to argue backwards through this relationship. Regardless of the ultimate metaphysical nature of content and function, the case of the discovery of electroreception demonstrates that they are indeed fixable. As the Kitchers, Amundson, and others have pointed out, biologists fix function and psychologists fix content all the time. That content and function are fixable does not tell us everything about the metaphysical or theoretical nature of content and function, but it does take out much of the epistemic sting said to derive from their (alleged) indeterminacy.

Nonetheless, there is still the question of how content and function are fixed. I argue that content and function are fixed by collecting a variety of cross-constraining and mutually supporting evidence. To make an attribution of content or function, one needs concurring evidence from behavioral analysis, anatomical analysis, physiology, ethology, evolutionary biology and adaptationist reasoning. By itself, not one of these sources of evidence is sufficient to explain the strong consensus on behalf of electroreception, although each provides a necessary piece of the evidentiary story. This collective, neuroethological approach to content and function fixation for which I argue will, with luck and diligence, license the appropriate attributions.

In Section II, I review the history of the discovery of electroreception, tracing the history of the science that led to the discovery of this non-human sensory modality. In Section III, I analyze the evidentiary contributions and limitations of the scientific components that make up neuroethology. The point of this section is to show how these components complement one another; allowing the collection to license content and function claims not licensed by any single component. This then allows me to return, in Section IV, to Dennett's arguments on behalf of indeterminacy using an example that meets his three desiderata: (1) I present a thorough and detailed story of the attribution of electroreception, (2) the attribution of electroreception is not an “obvious” attribution, and (3) such neuroethological claims are sufficiently “watertight”.

Humans have long been familiar with the phenomenon of animal electricity.6 The electric catfish of the Nile (Malapterurus electricus) is featured in Egyptian murals and statuary. Hippocrates, Plato, and Aristotle were familiar with the stunning charge of Torpedo ocellata, a Mediterranean species of ray from which the modern words “torpor” and “torpid” derive their meanings.7 Even prior to the development of the modern theory of electricity in the eighteenth century, it was recognized that there was something unusual about certain kinds of catfish, rays, and eels. Soon after the discovery of ways to detect and measure electricity, the true nature of these bioelectric animals was revealed.

The electric ray, the electric catfish, and the South American electric eel (Electrophorus electricus) possess “electric organs” that allow them to produce powerful electrical discharges capable of paralyzing prey or discouraging would-be predators. This discovery led to new mysteries, particularly the discovery of what were called “pseudo-electric” fish. These fresh-water fish from both Africa and the New World were so-called because their anatomy includes structures of highly regular morphology; a morphology strikingly similar to that of the electric organs of the electric eel and the electric ray. However, pre-nineteenth century technology could detect no electrical discharge from these “pseudoelectric organs”. That these organs do indeed generate a genuine, albeit weak, electrical discharge was discovered by Robin in 1865, and Babuchin in 1877. (See Moller and Fritzsch 1993.) Notwithstanding, their discharge was clearly too weak to subserve the kind of offensive and defensive function possessed by electric rays, eels, and catfish (see Textbox 1).

{Textbox 1 about here. Textboxes unavailable for web version.}

The possible function of such a weak electric organ was utterly mysterious and this, in turn, presented a problem to Charles Darwin. In the fourth and subsequent editions of his Origin of Species, Darwin notes that the existence of pseudoelectric fish pose an important challenge to his theory of evolution by natural selection. Darwin (1897) writes,

Although we must be extremely cautious in concluding that any organ could not have been produced by successive, small, transitional gradations, yet undoubtedly serious cases of difficulty occur.

[…]

The electric organs of fishes offer [a] case of special difficulty; for it is impossible to conceive by what steps these wondrous organs have been produced. But this is not surprising, for we do not even know of what use they are. In the Gymnotus [sic8] and Torpedo they no doubt serve as powerful means of defence, and perhaps for securing prey; yet in the Ray [Raja clavata], as observed by Matteucci, an analogous organ in the tail manifests but little electricity, even when the animal is greatly irritated; so little, that it can hardly be of any use for the above purposes. (234)

The existence of pseudoelectric organs posed a significant problem for the nineteenth century evolutionary biologist. The first step in giving an evolutionary account of a trait is to form a hypothesis as to its possible adaptive function. Nineteenth century biologists–as Darwin attests–could not begin to imagine the function of pseudoelectric organs. I will refer to this as “Darwin's Problem”. Darwin's own solution was to bide his time until more evidence could be collected and an answer found:

[We cannot currently go far] in the way of explanation; but as we know so little about the uses of these organs, and as we know nothing about the habits and structure of the progenitors of the existing electric fishes, it would be extremely bold to maintain that no serviceable transitions are possible by which these organs might have been gradually developed.

(235)

According to one line of reasoning, pseudoelectric fish organs no longer have a function. That is to say, these structures are vestigial (Du Bois-Reymond 1884; Rosenberg 1928; Dalhgren 1910). (See Moller 1995: 26.) On this account, the pseudoelectric organs are best thought of as atrophied versions of the powerful electric organs of electric eel or electric ray (which have a more obvious, offensive or defensive function). Of course, the gradualist problem about the evolution of strong electric organs in the first place would still remain.

By the turn of the century, some began to suspect that, in addition to the ability to generate electricity, some animals could detect the presence of electricity in their environment. In 1891, Fritsch “observed that mormyrids [the African family of pseudoelectric fish], when `surprised' by a pair of recording electrodes, avoided the metal with great agility.....” (Moller and Fritzsch 1993). The electrosensory hypothesis gained credibility in 1917 when Parker and van Heusen discovered that the non-bioelectrogenic catfish, Ictalurus (Amiurus) nebulosus, could detect galvanic and direct currents. They did so through a series of experiments in which they presented blindfolded catfish with glass, wooden, and metal rods. The fish were able to detect the presence of metallic rods at a distance, but only reacted to the glass and wooden rods when they touched the surface of the fish. They went on to show that the type of behavior (to flee from or to approach and “nibble” at the rod) elicited by the presentation of rods could be modulated by changing the length of rod exposed to the water. Parker and van Heusen correlated the amount of exposed metal with the amount of galvanic current produced by the rods, and then reproduced the behavioral results using direct electrical currents presented via electrodes placed in the aquaria with the catfish.

To what do Parker and van Heusen attribute this observation (that their catfish behave differently depending on the amount of electrical stimulation in their environment)? Notably, they do not posit an electroreceptive modality. Rather they propose that electric detection is mediated by the gustatory system, more specifically, by the taste buds. Their reasoning was that, (1) electrical stimulation elicits feeding responses and these behaviors are typically mediated by the gustatory system, (2) the head of the catfish is the most sensitive to stimulation, and most taste buds are found on the head,9 and (3) “This assumption is completely in line with what has long been known of human taste organs for these are easily stimulated by direct currents of very low energy value” (419). (Think of the distinctive sensation elicited by touching a 9-volt battery to the tongue, and 9 volts is far above the human threshold of sensitivity.) However, they admit that such evidence is inconclusive, for “[a]side from these general indications, however, we have no grounds for any determination as to the exact sense organ concerned” (418). They conclude, in the terminology of sensory physiology, that these catfish have only the capacity to electrodetect and do not possess a true electroreceptive modality (see Textbox 2).

{Textbox 2 about here}

A consensus on the ability of certain animals to electrorecept did not come until after the publication of several landmark papers by Hans W. Lissmann and his colleagues in the 1950s. The first, Lissmann's 1951 Science paper, confirmed that the so-called pseudoelectric fish were genuinely, albeit weakly, bioelectric. He measured weak, but highly regular fields generated by these fish and showed that such signals were very similar within species, but differed between species. Lissmann's discovery contributed to the renaming of pseudoelectric fish, which have since been known as “weakly electric” fish.

However, Darwin's Problem remained: Exactly what is the function of these weakly electric organs? Lissmann (1958) acknowledges Darwin's Problem: “The inadequacy of functional and evolutionary theories of electric organs in fish has been apparent for a long time” (156). He explicitly addresses this inadequacy: “In the absence of any existing, coherent theories about the evolution of electric organs, and about the function of weak electric organs, the speculative picture presented here may fill a gap” (186). Lissmann proposed his answer in two papers published in 1958. The first paper, entitled “On the function and evolution of electric organs in fish”, is a detailed exposition of an idea first suggested in 1947 by C. W. Coates of the New York Aquarium. Coates had speculated that the discharges of weakly electric fish are used to locate objects in their environment. That is to say, the electric organs of weakly electric fish subserve an electrosensory function.

Lissmann rejects the vestigial account. This account claims that strong electric organs evolved first and that they later degenerated into weak, pseudo-electric organs. Instead, he concludes that, “the easiest explanation for the evolution of strong electric organs would appear to start from ..... muscular action potentials, and proceed via weak electric organs used for orientation, to the powerful offensive and defensive organs [of electric eels, say]” (188). He supports this conclusion by presenting a series of discharges recorded from a variety of weakly electric species. Across every species investigated, these electric organs produced precise, stereotyped discharges. Lissmann reports on comparative ethological (in the field) and behavioral (in the laboratory) studies of representative species of seven genera of the African family, Mormyridae, and on behavioral studies of several members of the South American gymnotids.

Lissmann's conclusion from his work is that, contrary to the vestigial account, “the weak electric fish discussed here represent already very advanced, highly specialized forms” (186). According to Lissmann, far from being vestigial and without function, these electric organs play an important role in the lives of these animals: weakly electric fish use the discharges of their organs to actively locate objects in and navigate through their often dark and murky environments. He demonstrates the role of their electrosensory abilities with a series of experiments. Here is a description of a characteristic set of observations:

..... some preliminary experiments were carried out in Africa. A cloth partition was fixed in an aquarium dividing it into two equal halves. This partition consisted of a wooden frame over which the cloth was stretched on both sides, so that the two layers of cloth were about 2 cm apart. This frame was fitted into the aquarium by means of plasticine, and could be expected to be transparent to electrical but not visual stimuli. One Gnathonemus senegalensis [an African species of weakly electric fish] was introduced into one compartment of this tank and allowed to settle down for 2 days. After this period a second fish of the same species was carefully introduced into the second compartment. Both fish were resting motionless on the bottom. Recording electrodes introduced into the tank showed that both fish were discharging at a fairly low and regular rate. Whenever one of the two specimens was gently touched with a glass rod its discharge rate went up abruptly and the fish in the other compartment usually followed suit. When one fish was removed, a similar movement of the glass rod in the empty compartment remained without effect. These two specimens were left in the tank overnight. After darkness the light of a dim torch showed them both swimming up and down on opposite sides of the partition, obviously taking note of each other's presence and discharging with higher frequencies. Bursts of discharges from both fish coincided when they came close together, but no correlation in the timing of the individual pulses was noted. This observation, though not conclusive, does suggest that the electrical discharges may play a social role in the life of the Mormyridae. (169-170)

Lissmann does not pursue that final speculation on the social role of electrical discharge in these papers. (Although see Lissmann 1961.) However, later work has shown him to be largely correct. Lissmann instead focuses on the possible role of electrical discharges in perceptual and orientating behavior. He does this by:

(1) describing the nature of the electrical field generated around the fish, particularly focussing on the changes (at the surface of the fish) to that field as a result of objects of differing conductances being brought near the fish;

(2) noting the existence of pores distributed across the skin of these fish, and that “these pores lead through canals filled with a jelly-like substance to a variety of sense organs termed `glandular sense organs' or `mormyromasts'“ (181). These mormyromasts seem appropriately distributed across the body to participate in electroreception and are innervated by lateral line ganglia (which typically innervate sensory systems);

(3) noting the ecological conditions under which the fish live (often dark and turbid waters);

(4) noting natural behaviors (in some electric fish) such as being nocturnal and exhibiting a marked and unusual tendency to swim backwards and to explore novel environments tail-first; and finally,

(5) pointing out the unique swimming behavior of certain species of electric fish, who move primarily by undulating a highly specialized anal (or, in some species, dorsal) fin that runs most of the length of the body. This would tend to affect the generated electrical field less than swimming by undulating the whole body or by the use of large lateral fins as most bony fishes do. That so many species of electric fish in both major groups (one from Africa, the other from South and Central America) show a similar fin morphology is evidence of convergent evolution, according to Lissmann (Figure 1).

{Figure 1 about here. Figures unavailable for web version.}

Lissmann does more than offer a merely plausible account of the evolution of electric organs and the adaptive significance of an electric modality. The linchpin of Lissmann's account is his detailed proposal about how electric fish perceive electricity: that, “it can be imagined that such a fish, living in a private, electric world of its own, receives a variety of information through sense organs distributed over the surface of its body which may be likened to an `electro-receptive retina'“ (186).

Having presented an account of its possible adaptive significance, what Lissmann next presents is a hypothesis concerning the mechanism of electroreception. He does this in his second paper of 1958, written with K. E. Machin, entitled “The mechanism of object location in Gymnarchus niloticus and similar fish”. They first demonstrate a distinction between the presumably large number of fish which can electrodetect (or which can passively electrorecept) from those that exhibit genuine, active electroreception. They note that the galvanic currents used by Parker and van Heusen in their catfish experiments are several orders of magnitude larger than the sensitivity threshold of Gymnarchus: “It is clear that the sensitivity of Gymnarchus is of an entirely different order of magnitude to that of the other fish [previously investigated. These include minnow, carp, goldfish, catfish, and stickleback.]” (452). This is crucial because Lissmann and Machin's proposed mechanism of electrolocation requires detecting very small electrical changes; changes much smaller than the comparatively large changes detected by Parker and van Heusen's catfish (see Textbox 3).

{Textbox 3 about here.}

They next go back over the ground originally trod by Parker and van Heusen and demonstrate that their fish are actually detecting electrical changes in the environment, and are not making discriminations based on some other environmental properties. I have not the space here to rehearse these elegant experiments. I will only note the authors' conclusion: “..... [I]t has been shown that Gymnarchus can distinguish between geometrically identical objects if they have different electrical conductivities, and cannot distinguish between objects which, although geometrically identical and with similar electrical effects, have different internal arrangements” (454).

Having shown that Gymnarchus can perceive objects on the basis of their electrical properties alone, Lissmann and Machin then ask what kind of mechanism might carry out such a perceptual feat? They begin by assuming that the mormyromast-type structures found in the skin of Gymnarchus are electroreceptors (455). They next ask how sensitive these “electrical receptors” would have to be in order to carry out the process of electroreception they observed in their behavioral studies? Their answer is based on both physical and mathematical models of an electric fish. The physical model involved an artificially produced electric field mimicking that produced by weakly electric fish. They placed recording electrodes in the water at points corresponding to the surface of a fish, and then recorded changes to electrical properties as objects were passed through the artificial field. This physical model of the kinds of perturbations demonstrably detectable by the fish was supplemented by a mathematical model of the effect of a cylindrical object, of known conductivity, on a dipole electric field in water of known conductivity. They conclude from their models that, depending on the electrical properties of the receptors (specifically, the resistivity of the jelly-filled canals of the mormyromasts), the fish are probably performing both temporal and spatial integration of the potential or its second derivative (see Figure 2). It is important to note here that Lissmann and Machin not only suggest that electric fish are capable of electrolocating, they also outline with a good degree of precision, how that perceptual ability might be carried out in physical, neurophysiological, and computational terms.

To see how Lissmann's predictions about the physiology of electroreception were born out, we need to turn our attention to a second stream of research. While work on pseudoelectric fish and catfish can be characterized as a function in search of an organ, work on the ampulla of Lorenzini was an organ in search of a function. Found in the skin of sharks and rays, the ampulla of Lorenzini has a striking structure. It consists of a long canal leading to swelling (or ampulla), all filled with jelly (Figure 3). Until this century, it was generally accepted that the ampulla was responsible for some form of mechanoreception.

Physiological studies (Sand 1938; Hensel 1955) demonstrated that ampullae are extremely sensitive to changes in temperature, a finding which suggests that perhaps the ampullae are thermoreceptors. However, this hypothesis leaves some unanswered questions. For example, what is the purpose of the long canals? Murray (1960) and Loewenstein (1960) showed that, in vitro at least, these cells exhibit a response to large mechanical distortions. However, this evidence offered only dubious support to a mechanoreceptive hypothesis because the degree of mechanical distortion required to elicit ampullary responses is much greater than that typically experienced by sharks and rays.

In 1962, both Murray and Loewenstein and Ishiko published results indicating that the ampullae of Lorenzini are highly sensitive to chemical changes, specifically small changes in salinity, thereby suggesting a possible chemoreceptive function. However, strong conclusions could not be drawn from such physiological research unless it could also be shown that such changes in salinity actually played a role in the lives of the organisms and that the organisms indeed made behavioral distinctions on the basis of such stimuli. Except for a few species that wander in and out of river mouths, such ethological evidence was not forthcoming any more than it was for the mechanoreceptive hypothesis.

{Figure 2 about here}

These two streams of research–a function in search of an organ and an organ in search of a function–came together in the 1960s. T. H. Bullock and colleagues (Bullock, Hagiwara, Kusano, and Negishi 1961), recording from the lateral line nerve of the electric fishes Gymnotus and Hypopomus, measured differential neurophysiological activity when the fish were stimulated by passing a conductive or insulating object nearby. Significant neural activity was not observed when the fish were stimulated by ethologically-plausible levels of mechanosensory stimulation (brushing the skin with a brush or with weak water currents). They conclude that they have demonstrated the existence of “true electroreceptors” (1427). Part of the basis of this claim is the work that had already been done by Lissmann, Machin and their predecessors–work that demonstrated the ethological, behavioral, and historical cogency of the sensation and discrimination of small changes in the naturally available electric organ discharge current path. As the history of this entire episode plays out, Bullock's physiological work was one of the last pieces of an intricate puzzle, whose solution resulted in a scientific consensus that a new sensory modality, electroreception, had been discovered; there exist organisms that possess anatomical structures (mormyromasts in weakly electric fish, ampullae of Lorenzini in sharks, skates, and rays) with the function of mediating electrical information about their world.10

In following years, electroreception has become a mainstay of neuroethological research. The number of species for which electroreception has been claimed has grown to include salamanders, sharks, skates, rays (Kalmijn 1982, 1987), catfish, sturgeons, paddlefish, lungfish, lampreys, and even the platypus (Scheich, et al 1986). In 1986, scientists at the Smithsonian Institution claimed to have discovered that electroreception even occurs in a species of placental mammal, the Star-nosed mole (Gould, et al 1993), but this claim has recently been called into question (Schlegel and Richard 1992; Catania 1994).11

{Figure 3 about here.}

Research on weakly electric fish has continued. Much of the neural circuitry underlying electroreception and electrically-mediated behavior–from receptor cells to central nervous system nuclei to motor cells–has been uncovered (Heiligenberg 1991). In terms of known circuitry, the electric fish electrosensory/electromotor system is one of the best understood vertebrate systems in contemporary neurobiology. Beyond its use in electrolocation, bioelectricity has been shown to subserve important communication functions, allowing electric fish to easily identify and communicate with conspecifics, even in dark and turbid environs (Hopkins 1977). Fish use their electrical signals as part of complex electrical “mating dances” in much the same way that courting birds “dance” (Hagedorn and Heiligenberg 1985). The ability to send and receive electrical signals provides these animals with an essentially private channel of communication in their crowded, tropical environs.

III

The final stages of the history described above indicate the strength of the consensus in favor of electroreception. Scientific effort has shifted away from trying to show whether electroreception exists to trying to discover exactly which species do and do not possess it and to elucidating further the neurobiological mechanisms that subserve this modality. All this, with a function whose possibility was not even imagined a century ago. What has led to this strong consensus? I claim that it is due to the depth and breadth of evidence brought in its favor. Electroreception has been fixed through a process in which behavioral, ethological, physiological, computational, and evolutionary evidence have all been marshalled in a mutually supporting fashion. The result of this interdisciplinary endeavor is a hypothesis that unifies findings from all of these areas. Lissmann is given primary credit for having discovered electroreception because he had a hand in more areas than any other single researcher. Each area contributes answers to different aspects of the issue, and together they tell the strongest story we can practically expect to have.

The name given to this combination of mutually constraining areas of research is “neuroethology”–a name that traditionally captures a wide variety of research priorities in biology, from neurobiology and ethology (as the name suggests) to computation and evolutionary biology.12 These different areas make a nicely complementary suite of priorities. We need the kind of evidence encompassed by neuroethology to license claims about content and function. The strong consensus on behalf of electroreception is not due to some singular piece of evidence that is both necessary and sufficient. For example, it might be claimed that it is the evolutionary reasoning alone which licenses the claim for electroreception and all the rest (physiology, anatomy, ethology, etc.) is just icing on the cake. To show that this is not the case and that no one area alone is sufficient, let us look at these areas in more detail and determine what they can and cannot do for hypothetical content and function attributions.

Anatomical and Behavioral Analysis

Anatomy and behavior are almost always good places to start, because more often than not they provide that which requires explanation. Having identified a biological trait, organ, capacity, etc., we can begin an investigation by asking two basic questions: (1) Of what is it composed (anatomy)? (2) What does it do (behavior)? The answers to these questions set the initial limits on function/content hypotheses. But research in the other areas may send us back to take another look at anatomy and behavior to see if we overlooked something.

This situation is well illustrated in the case of electroreception. Prior to the birth of modern science, it was well known that certain fish (the strongly electric fish) do something unusual: they induce a numbing feeling in those who would harass them. Anatomical investigation revealed that these organisms have in common an elaborate and intricate organ. Additional research revealed that other animals (the pseudoelectric fish) also share this anatomy, but do not share the striking behavioral anomaly. This discrepancy between anatomy and behavior set the basis for Darwin's Problem. What is the function of pseudoelectric organs, if not to produce the type of behavior associated with strongly electric fish?

Behavioral analysis provided the next piece of the puzzle, when Parker and van Heusen demonstrated that their catfish exhibited an unexpected behavioral capacity for distinguishing objects differing only in their electrical properties. However, lacking firm evidence as to the anatomical basis of this perceptual skill, they cautiously concluded that their catfish were capable of electrodetection. That is to say, the analysis of behavior alone is insufficient to support the stronger claim of electroreception. Looking at behavior alone, it is impossible to distinguish between detection via a previously identified modality and reception via some new modality. Lissmann replicated their work in his weakly electric fish as the first piece of evidence for the hypothesis that such organisms possess an electroreceptive function.

Physiology

Anatomy and behavior go hand-in-hand with physiology. In a sense, physiology is a science situated between these other two areas. Where the behavioral analysis described above primarily focussed on the overt behavior of the organism as a whole, physiology explores the behavior of its components as identified by the anatomist–behavior that ultimately underlies the overt behavior exhibited at the organismic level.

The work of Bullock and colleagues is an excellent example of the role of physiological evidence. They discovered cells in the peripheral nervous system whose physiological responses tracked electrical properties in the environment. In essence, Bullock and colleagues replicated the experiments of Parker and van Heusen, but recorded the neural activity directly instead of simply observing overt behavior.

Physiology offers powerful tools for determining what activity in sensory cells represents. It might seem initially plausible to say that a true sensory cell is one which responds best to some particular type of sensory stimulus. For example, a true electroreceptor would be one which responds best to electrical stimuli. We would have to say “responds best” on this account, because nerve cells will typically respond to many different kinds of stimuli. Press the outside of your eyeball with your thumb and you will elicit a visual percept. However, we do not normally think of the eye as a pressure receptor. A battery touched to the tongue causes a quite distinctive flavor sensation. However, it seems absurd to say that humans are electroreceptive. To avoid such absurdities, someone wishing to wield physiological evidence alone to determine content must be able to show that a given type of stimulus is best at eliciting a physiological response from the nerve cells of the structure in question.

However, this will not work. First, as Bullock (1974) has noted, there is no metric for comparing intensity of stimuli across modalities. Say a given cell responds to intensity x of auditory stimulation. However, it responds equally well to intensity y of chemical stimulation. Based on intensities in qualitatively distinct domains alone, one cannot decide to which stimuli the cell responds best. The problem here is that we lack a cross-modal metric. Based on physiological data alone, we simply have no way of evaluating the intensity of stimuli of different modalities, because we have no way of comparing different quantities of different qualities. For this reason, physiology alone is helpless to distinguish between the various functional hypotheses attributed to the ampulla of Lorenzini: mechanoreception, thermoreception, chemoreception, or electroreception.

A second limitation of physiological evidence is that one of the best stimuli for eliciting responses in sensory cells (actually, to be accurate, in all neurons) is a properly inserted stimulating electrode. Neuronal response covaries very nicely indeed with the amount of current injected into a neuron by an electrode. Chances are, this covariation is much stronger than that recorded with any other stimuli. But it seems absurd to call every neuron known to humanity a “stimulating electrode detector!”

However, these problems are insurmountable only if we consider physiology and nothing else. These problems indicate only that physiological evidence needs to be buttressed by other kinds of evidence. For example, one might respond to the stimulating electrode problem by noting that “electrode detecting” is an unlikely candidate for neuronal function (and “stimulating electrode” is an unlikely candidate for neural representational content) because animals do not typically encounter stimulating electrodes in the natural environments in which they have evolved. Since the etiological/causal theory of content and function defines them in terms of the selection history of the organism, a consideration of normal or natural environments is called for. Such a consideration is the purview of ethology, to which we now turn.

Ethology

Behavioral analysis is typically carried out within the well-controlled confines of the laboratory. However, that which makes the lab attractive (greater control over the stimulus environment of the organism) is also a potential weakness. This is because we are generally more interested in what organisms do naturally in the wild than what they can be coaxed to do in the lab. Physiological investigation also has related problems, as just noted. Because of these problems, ethology offers an important constraint on content/function theorizing. Ethology is the study of animal behavior within natural settings. Ethologists study naturally behaving organisms and attempts to discover to what sorts of stimuli they are normally exposed, as well as to what sorts of stimuli they naturally respond. Where physiology and behavioral analysis tell us what organisms are capable of doing, ethology tells us what they actually do.

The story of electroreception is shot through with ethological considerations. Lissmann collected data in the field, observing the behavior of wild electric fish. Also, in proposing that electric fish were active electrolocators he proposed that weakly electric fish are perceiving electrical signals in their environment that they themselves produce. That is, he simultaneously proposed a sensory function and the ethological plausibility of the signals detected by that sensory system. Electrical fields are indeed in the fishes natural environments because they themselves naturally generate these stimuli. This ethological evidence in favor of electroreception stood in stark contrast to the poverty of ethological evidence in favor of the mechano-, chemo-, and thermoreception hypotheses.

Ethology can be helpful in conjunction with other approaches, but it is not without its limitations. First, the fact that the natural environments of organisms are simply packed with potential stimuli is problematic. In any given situation, there is a virtually infinite set of possible stimuli to which an organism might be attending. The natural world is full of electromagnetic radiation, vibrations, chemicals, etc., and most organisms generally pay attention to all and only those stimuli which it can profitably exploit. Determining what stimulus or set of stimuli is responsible for eliciting a given behavior requires controlled experiments with known stimuli.

As if the plurality of stimuli were not bad enough, the ethologist is also faced with the same problem on the other side of the coin: not only are natural environments just packed with potential stimuli, but naturally behaving organisms do lots of things. Animals rarely perform only a single behavior at a time. A variety of interpretations can be placed on a given behavior. Distinguishing between intentional and accidental actions can also be difficult. Did that fish rise six inches by the action of its swim-bladder or was it moved by an upwelling current? Careful experimentation can help narrow the hypotheses down, but in doing so one encounters the Ethologist's Dilemma: the more control an experimenter exercises over the stimuli and behavior of an organism, the less natural and biologically normal such stimulation and behavior becomes. Thus, the experimental ethologist must be careful not to undermine the very naturalness that motivates this approach in the first place.

Again, one should not conclude from such ambiguities and difficulties that ethology is pointless or irrelevant. As with physiology, ethology cannot by expected to answer content and function questions unaided. For example, knowing the physiological response ranges of putative electroreceptor sensory cells constrains behavioral interpretations–an animal cannot respond to stimuli its sensory system is incapable of detecting. Also, as with physiology, ethology provides important evidence that may support some hypotheses about content and function and cast doubt on others. In this way, knowledge of the life of sharks, skates, and rays gives unequal support to the different hypotheses concerning the ampullae of Lorenzini. It is unlikely that the ampullae are mechanoreceptors, if the severe degree of mechanical distortion required to elicit response is much greater than that ever experienced by these organisms.

Evolutionary analysis and adaptationism

Finally, consider the arena of historical evidence–evidence of a species' evolutionary history. Ethology is helpful for determining content and function because, as a general rule, organisms are well adapted to their environment. Why? Because organisms are the product of evolution by natural selection, a process that systematically eliminates organisms which are less well adapted in favor of those which are better suited to their environment. Given the theoretical characterization of content and function in Section I, none of this should be surprising. On the accounts presumed here, what it is to be a particular function or content is to have a particular causal history. Therefore, it is no surprise that evidence of causal history is relevant to such attributions. Lissmann, for example, presents such evidence, by noting similarities in fin morphology and behavior of presumably widely divergent species of fish and suggesting that these are the result of convergent evolution. So, why not cut out the ethological middleman and answer questions of content and function determination directly on the basis of historical evidence alone?

Such an approach is not without its proponents. Presumably, Millikan would argue that to answer the question–”Are the ampullae of Lorenzini thermo-, chemo-, mechano-, or electroreceptors?”–one need only determine to what use the organism's ancestors put the structure. If sharks and rays have historically used these structures to gather temperature information, then the ampullae are properly thought of as thermoreceptors. If, instead, they used these structures to detect the electrical properties of their environments, then they are properly thought of as electroreceptors.

Aye, but there's the rub! The very thing to which Millikan calls upon us to consider–the selection history of the ancestors of a currently extant species–is the very evidence to which we do not have access in the vast majority of cases. Generally, all we ever have are extant species. We do not have direct access to selection histories or the past competitors of current designs, those designs that failed in the struggle of natural selection. What we do have is a severely restricted window on evolution via paleontology, genetics, comparative biology and cladistic analysis.

Curiously, Millikan herself barely suggests what actual methods ought to be used to gather the historical evidence her theory requires. (Indeed, her own favorite examples–hearts, magnetosomes, etc.–are without the very kind of evidence her account requires. We simply do not know much at all about the evolutionary history of hearts and magnetosomes.) Instead, she falls back on the only option open to one who argues for the importance for evolutionary history, but who has little access to the actual data of evolution: adaptationism. Without access to selection histories, the historian has little choice but to use plausibility as the only test of hypotheses about content and function.13

Adaptationism is exactly what Dennett would have us pursue. He is quite correct about the linchpin role adaptationist reasoning plays in functional theorizing. The electroreception case offers us an obvious example of its importance. Darwin's Problem posed the difficulty: What is the function of pseudoelectric organs? To Lissmann's lasting credit, he offered an answer: electric organs have an electroreceptive function.

However, it must always be kept in mind that adaptationist reasoning is but one method among many, and just as with the areas discussed above, adaptationist reasoning encounters important limitations when trying single-handedly to answer questions about function. For example, how can we even begin the process of content and function attribution when we do not know what possible contents and functions there are? Electroreception went undiscovered for millennia in part because it was not realized that electrical content was a possibility. Without first having sufficient knowledge of the life and physiology of an organism, it is devilishly difficult to generate plausible hypotheses of content and function. Darwin himself recognizes this when he identifies his ignorance of the lives of pseudoelectric fish as the reason for his inability to determine the function of their pseudoelectric organs. Adaptationism does not work well where we are forced to expand the space of possible functions and contents, as opposed to simply attributing those which we already know to be possible.

Limitations on imagination and knowledge are not a problem for adaptationism alone–they can plague any approach. A more serious and unique problem for adaptationism is raised in the context of vestigial organs. What if the system under investigation is vestigial, once having a function but no longer? The adaptationist's first step is to assume that it does have a function. (Dennett's intentional stance has similar problems dealing with cognitive agents precisely when they are being irrational or willfully ignorant.) In such cases as these, adaptationism is a non-starter.

A final problem for adaptationism arises when we consider the distinction between detection and reception, as exhibited in the case of Parker and van Heusen's work on the sensory skills of catfish. In the case of their catfish, Parker and van Heusen felt that they had strong warrant for claiming electrodetection in catfish, but not for the stronger claim of electroreception. They were reticent because, while they had discovered robust behavioral evidence that electrical stimuli could be detected, they lacked any neural mechanism whereby such stimuli might be processed. They also lacked ethological evidence that such electrical stimuli were naturally occurring in catfish environments. They could only imagine that the finely tuned taste receptors known to be scattered over the head and body were, as human taste receptors are capable of, responding to electrical changes. This, together with the further evidence that feeding behavior could be so elicited, pointed to electrodetection (via the gustatory system) over electroreception. Lissmann was able to hypothesize genuine electroreception in the case of weakly electric fish, because he could sketch a dedicated neurobiological mechanism and present ethological evidence, in addition to behavioral evidence.14

However, it is not clear whether adaptationism alone allows us to maintain such a distinction between detection and sensation. Where sensory physiologists posit two distinct kinds of content (on the basis of physiological differences) to explain often indistinguishable behavior, adaptationists who rely primarily, if not entirely, on behavior cannot make such a distinction. This should come as no surprise, since such philosophically well-motivated frameworks often do not provide for the fine-grained analysis required by practicing scientists.

The example of the discovery of electroreception shows us how evidence in other areas constrains and helps us in our adaptationist reasoning. Lissmann deserves credit for discovering electroreception not because he correctly posited the function of this sensory system but rather because he presented this adaptationist reasoning together with a great deal of supporting evidence not derived from adaptationist reasoning. It was the combination of his modeling work, his anatomical speculation, his ethological observations, and his behavioral analyses, together with his identification of an evolutionary function that licensed his claim that electroreception existed.

We are now in a position to return to the debate with which I opened this paper: How is it possible to fix content and function? I have described in detail the cautious, meticulous ploddings of biologists over many decades, and how their work resulted in a strongly supported claim: that weakly electric fish possess anatomical structures, undergo physiological processes, and exhibit behaviors that have as their function the perception of electrical stimuli in their environment.

I also want to argue that the case presented here successfully meets Dennett's three desiderata for potential counter-arguments. First, it certainly meets his requirement of thoroughness. The electroreception hypothesis is supported by a great deal of work, as described above, in a variety of different scientific fields. The evidence for this hypothesis presents a great deal of causal detail. The electrosensory hypothesis successfully unifies findings in all of these fields, and to date, no other hypothesis has been put forward with anywhere near the strength of evidence as this functional attribution. If one still wishes to object on the grounds of thoroughness, I will only note that my own account more or less ends in the 1960s. For the most part, the account given here does not describe the great amount of work conducted since then–work which adds still further support for this particular functional attribution.

Furthermore, it is the nature of the electroreception case that it meets Dennett's second desideratum concerning the non-obviousness of the attribution. Electroreception is too bizarre a sensory modality (relative to those possessed by humans) ever to be considered obvious. It took millennia for the electroreception hypothesis to arise. It took decades to garner sufficient evidence on its behalf.

The real crux of Dennett's objection to content and function determinism lies with his third requirement: that I tell a “watertight” story–a story that leaves no room for alternative interpretations of the evidence presented. Recall that Dennett (1987) wishes us to abandon the idea that, “there has to be one, determinate, right answer” (319) to questions about content and function–a claim that he implies function/content determinists must hold. Note, however, that Dennett is embedding three different criteria in his claim here. The first, that there must be one answer, is clearly a red-herring. No determinist worth her salt would restrict herself to the claim that, in all situations, there is but a single function or content to be attributed. Real examples are more complicated than this.

The second and third criteria–determinacy and correctness–are more involved. Dennett is claiming that determinists believe that we can identify functions and contents and have good reason to believe in the appropriateness of our identifications. To begin with, I admit that, on one reading, Dennett is correct. The evidence on behalf of electroreception cannot logically rule out all other possible interpretations. But, to do so would be an impossible task for standard reasons of empirical underdetermination. Quine, Goodman, and others have argued that any claim is underdetermined by empirical evidence, because any hypothesis sufficiently general to be of interest “goes beyond” available evidence (cf. Goodman's New Riddle of Induction). Furthermore, we can never be absolutely certain of the truth and accuracy of our empirical observations. For instance, we may be the unwitting victims of a skewed sample, or our instruments may have consistently malfunctioned at the crucial moments. However, this lack of logical certainty is part and parcel of scientific theorizing. Scientists must guard against it, but even in the best cases, logically speaking, nothing can be done to silence the extreme skeptic.

Therefore, the underdetermination and uncertainty of empirical claims point to two reasons why we might agree with Dennett that content and function attributions are neither determinate nor fixable. However, Dennett's initial position in The Intentional Stance is that there is something special about content and function claims. This difference sets them aside from other empirical claims scientists tend to make. Indeterminism is supposed to result from the interpretive move that must be made when attributing content and function. This hermeneutic move requires that we “go beyond” the available evidence. However, we just noted that underdetermination also requires that we go beyond the available evidence. There does not seem to be any relevant difference here.

My response to Dennett is to point out that in the case of content and function, there exist constraints on our interpretation. It is not the case that one is free to posit just any function or content. The electroreception hypothesis, for example, meets the constraints set by evidence uncovered in each of the various areas of neuroethology. This is why the breadth of evidence in favor of electroreception is compelling. Furthermore, this is nothing special to content and function claims, this is a general virtue of empirical hypotheses. As Hempel (1966) observes, “For the confirmation of a hypothesis depends not only on the quantity of the favorable evidence available, but also on its variety: the greater the variety, the stronger the resulting support” (34). However, the story presented here goes beyond Hempel's virtue of evidentiary breadth. The evidence in favor of electroreception is more than just broad; it is mutually constraining and mutually supporting. For example, ethological evidence supports some interpretations of physiological data while casting doubt on others.

We are left with the conclusion that attributions of content and function are empirical claims like any others, subject to the vagaries of uncertainty and underdetermination, but that with enough evidence of the right kind, such attributions are ultimately as fixable as any other empirical claims. It then comes as no surprise that we speak of the “discovery” of electroreception, just as we talk of the “discovery” of planets around other stars or the “discovery” of a new strain of tuberculosis. At the same time, legitimate concerns about underdetermination force us to conclude that the evidence in favor of any such claim can be compelling but can never force a conclusion. However, by requiring a “watertight” account in order to license function/content attributions Dennett asks more of content and function claims than other empirical claims. The painstaking and cautious investigation of neuroethologists during this century has produced a varied collection of evidence, all of which points to the claim that they have discovered a new sensory modality and that they have fixed the electroreceptive function of the structures, processes, and behaviors that subserve that modality.

I want to suggest that it is Dennett's distinctly unnatural sense of “adaptationism” which generates indeterminism, not the actual practice of science.15 If biologists stopped their investigations after presenting adaptively plausible stories, Dennett's criticisms would be correct, but they do not. I am not alone in noting the unnaturalness of Dennett's adaptationism. Amundson (1988) accuses Dennett of failing to appreciate the fact that practicing biologists go beyond the adaptationist reasoning that Dennett advocates. They take into account the causal mechanisms of natural selection that brings the attributed functions about. “If adaptationists are conscientious”, Amundson notes, “the dose of `goodness' [that Dennett claims is necessary to functional theorizing] will be heavily constrained by the facts of natural history. Adaptationism is not only a functional theory, but also an evolutionary theory” (506). Dennett's version of adaptationism (which deemphasizes the underlying causal mechanisms of natural selection) is of a piece with Ryle's logical behaviorism (which deemphasizes the causal mechanisms of thought and action). As such, Amundson interprets Dennett as advocating “logical adaptationism”.

Dennett (1988b) objects that Amundson has misunderstood him: “..... I don't `deny the relevance' of causal theories; I just deny their sufficiency” (541). But what would satisfy Dennett here? Avowals to the contrary aside, Dennett very often sounds as if he is advocating logical adaptationism. While it is true that he rarely explicitly denies the relevance of non-behavioral, non-adaptationist evidence, he often simply fails to consider it. If he is not guilty of a sin of commission, then he surely is guilty of a sin of omission.

One interpretational problem with Dennett is that it is difficult to see exactly where he stands on the status of adaptationism. On the one hand, he introduces this scientific strategy as a stopgap measure, a “sound interim way of speaking” (237) until we have a firmer grasp of other pools of evidence, such as that provided by the neurosciences and comparative biology. He cites with approval Oster and Wilson's claim that the adaptationist strategy should be considered “provisional guides to future empirical research and not as the key to deeper laws of nature” (ibid.) Dennett endorses this claim and asserts that, “Exactly the same can be said about the strategy of adopting the intentional stance in cognitive ethology” (265). This invites in the other kinds of evidence I have been discussing here. (Although we may well wonder why we should bother if, at the end of the day, there is no fact of the matter?)

However, when push comes to shove, this ecumenicalism goes out the window. Space is short here, so let me take just one example, involving a very Quinean thought experiment.16 We are asked to imagine a tribe that uses a word, “glug”, for an invisible, explosive gas found in their local swamp, a gas that we know to be methane. We are then to imagine them being presented with acetylene. If they use the word “glug” to refer to acetylene, do we take them to be using the term correctly? If glug really means “methane” then they would be wrong, but if “glug” means “gaseous hydrocarbon” then they would be quite correct. But how do we tell what they really mean? Dennett's response is telling:

If, as seems likely, no answer can be wrung from exploitation of the intentional stance in their case, I would claim (along with Quine and the others on my side) that the meaning of their belief is simply indeterminate in this regard. It is not just that I can't tell, and they can't tell; there is nothing to tell. (312)

This is exactly the sort of move that leads one to conclude that Dennett is a logical behaviorist (and, when the same move occurs in evolutionary contexts, a logical adaptationist). When confronted with an example in which behavioral analysis (by means of the intentional stance) seems to make no headway, what is Dennett's reaction? Does he treat the intentional stance as a stopgap measure, one among many possible avenues of investigation? No. The intentional stance cannot answer the question, so he concludes that the question must not have an answer. But this seems to throw in the towel before the fight has even begun. Furthermore, it undercuts his claimed naturalism.

Perhaps it should come as no surprise that two of the late twentieth century's most famous indeterminists (Quine and Dennett) are also two of its most famous behaviorists (although Dennett routinely denies cleaving so closely to the position of his former teacher, Gilbert Ryle). Dennett cannot have it both ways. He cannot say, on the one hand, that adaptationism and the intentional stance are not the only games in town–that they are but two sources of evidence among many–and, on the other, that if adaptationism and the intentional stance fail to produce fruitful hypotheses then no other avenue need be explored because there must be no answer. This makes no more sense than a physiologist claiming that there is no fact of the matter as to whether all neurons are “stimulating electrode detectors” merely because physiological evidence fails to rule out such an interpretation, or for a behavioral scientist to claim that there is no distinction to be made between detection and reception because behavioral analysis reasoning fails to provide evidence for such a distinction.

Ultimately, Dennett and those who would support his indeterminacy thesis are faced with a dilemma. On the one hand, they can honestly advocate logical adaptationism as an approach to the issues of content and function attribution. Such an account indeed has indeterminism as a consequence. However, the example of adaptationism's role in neuroethology indicates that logical adaptationism is a poor caricature of adaptationism as it is practiced in contemporary biology. On the other hand, they can renounce logical adaptationism and embrace a version of adaptationism more in line with what neuroethology calls for and which biologists practice. However, the neuroethological approach to content and function attribution is not, in practice, subject to indeterminism (although, like every other area of science, it is subject to worries about uncertainty and underdetermination). At least, that is what seems to be the most obvious conclusion from the detailed account of electroreception described in the history of this discovery. Therefore, Dennett is faced with a dilemma: either embrace an indeterminist, logical adaptationism that bears little resemblance to adaptationism as it is practiced in biology, or adopt a more accurate portrayal of adaptationism from which indeterminism does not follow.

Several decades ago, C. P. Snow (1959) bemoaned the emergence of two cultures, one scientific, the other humanist. Naturalized philosophy, of the kind represented here, attempts to bridge Snow's gap between biology (on the scientific side) and philosophy (on the humanist side). On the surface, the debate over the degree of determinism available for content and function attributions seems to be a debate between these two opposing camps. Dennett is here accused of being too “philosophical” and insufficiently “scientific” when arriving at the conclusion that content and function attributions are essentially different from other types of scientific claims. To his credit, Dennett denies that he fails to appreciate the contribution science has to make.

The details of this debate aside, there does seem to be a dearth of detailed considerations of scientific reasonings within the philosophical literature–even the philosophical literature dealing directly with science. Content and function attribution are among the things that scientists do on a daily basis and it would behoove those of us who wish to draw conclusions about this process to pay closer attention to how scientists actually do it. The distinction here is not unlike that between laboratory behavioral analysis and ethology. Philosophical analysis has, in the past, taken place within the well-controlled boundaries of toy examples and thought experiments. While such an approach has its benefits, we should not forget that we can supplement our investigations by looking at how science is actually done. I suggest that we not forget about the ethology of actual, practicing scientists. The present paper is offered in this ethological spirit. By looking in detail at a particular episode in the history of biology, we may learn something about content and function attribution. At the very least, it should give us a new arena within which to test our philosophical intuitions.17

The discovery of electroreception offers a particularly nice example for the issues at hand. In a relatively short period of time, a strong consensus was built and maintained on behalf of a particular functional hypothesis: that weakly electric fish (and many other animals) possess a unique, non-human perceptual capacity called “electroreception”. As it involves a sensory system, the electroreception proposal cuts across both the content and function attribution issues–it simultaneously makes a claim for a particular evolutionary function and a claim about the perceptual content possessed by an organism with this evolutionary function.

How was electroreception in weakly electric fish discovered? The discovery was made by the combined effort of several different strands of scientific research, including neurophysiology, anatomical & behavioral analysis, ethology, and evolutionary & adaptationist reasoning. While none of these approaches can, on its own, marshall sufficiently strong evidence for functional or content hypotheses, their combined, mutually constraining use–within the multidisciplinary framework of neuroethology–can and has. In doing so, the discovery of this unique and interesting sensory modality offers a case study in function/content attribution that calls out for continued detailed philosophical consideration. This paper is, I hope, only the beginning.

Acknowledgements

This paper is a modified chapter of my Ph.D. dissertation: Keeley (1997). I have had extensive help on this paper. It makes sense to thank people in groups: (1) members of my lab and the electric fish community: Walter Heiligenberg, Ted Bullock, John Spiro, Calvin Wong, Jim Murray, Jim Prechtl, and Gerhard von der Emde; (2) philosophers: Kenna Barrett, Laura Reider, Laura Perini, Jessica Pfeifer, Adrian Cussins, Fred Dretske, and the members of the Experimental Philosophy Lab; and (3) members of my dissertation committee: Sandy Mitchell, Pat Churchland, Paul Churchland, Jim Moore, and Marty Sereno. Earlier versions of this paper were presented at the University of Birmingham and at fellows meetings of the McDonnell-Pew Center for Cognitive Neuroscience. This work was supported by grants from the National Institute of Mental Health (NRSA #1 F31 MH10676-01), the McDonnell-Pew Center for Cognitive Neuroscience, and the University of California, San Diego Department of Philosophy.

Notes

1. In equating these two domains of attribution, Kitcher and Kitcher (1988) observe that on Dennett's account, “Darwin emerges as the spiritual grandfather of Quine's famous indeterminacy argument” (517).

2. My characterization of content points to issues that may appear to be a long way from typical philosophical concerns about content, e. g., misrepresentation, linguistic translation, etc. Nevertheless, inasmuch as an organism's sensory manifold partially determines the intentional relationship of that organism to its phenomenal world, the individuation of sensory modalities involves content. However, I am not arguing that the case of electroreception will enlighten every discussion about content.

3. Kitcher and Kitcher (1988) raise the guppy example.

4. Millikan intends her own account to extend to issues of mental content as well, hence the title of her first book, Language, Thought, and Other Biological Categories. In this way, she agrees with Dennett on the close relationship between issues of content and function.

5. Except, apparently, in the worrisome case of “obvious” attributions. Dennett seems to accept that we do fix function in the case of birds' wings and eagles' eyes. However, it is unclear what is fixed in these cases, since it is the nature of function (on Dennett's account) that it is indeterminate. In any case, one wonders why, if it is OK to fix (indeterminate) function in these cases, we can't fix (indeterminate) function in all cases?

6. The following discussion is not intended, and should not be taken, as a complete and fully accurate depiction of this scientific history. Instead, I plan to present a rational reconstruction of the work that was carried out. Many good histories have been written; including several by major participants. I point interested parties to those that were of the greatest help to me: Bullock 1974; Bullock and Szabo 1986; Wu 1984; Moller and Fritzsch 1993; and Moller 1995.

7. Hippocrates discusses the dietary benefits of the Torpedo in On Regimen. Plato describes Socrates' torpid effect on his audience in the Meno (80b). Aristotle discusses the powerful discharge of the Torpedo in Book IX of the Historia Animalium.

8. This is most likely a misidentification of the electric eel, Electrophorus electricus.

9. In fish and other aquatic vertebrates, taste buds are found not only within the mouth, but also on practically the entire surface of the body. In catfish, they are densely distributed on the head.

10. Ironically, Murray (1960, 1962) had noted that the ampullae are responsive to electrical currents, but favored the hypothesis that this was simply part of a mechanism for detecting salinity (there is a direct connection between salinity and the electrical conductivity of water). In other words, Murray supposed that the ampullae allowed rays to use the conductive properties of water to extract information about its salinity (as opposed to using the chemical properties to extract information about the electrical properties of the surrounding water).

11. To tie up a few loose ends, I should note that, as it turns out, Parker and van Heusen's catfish were passively electroreceptive, after all. In addition to taste buds, further investigation revealed that they possess an ampullary electroreceptor system. The unique ampullary structure of catfish, shark, skate, and ray electroreceptor cells gives these organisms the high degree of sensitivity required by passive electroreception.

12. In 1992, I joined the electric fish neuroethology laboratory headed by the late Walter Heiligenberg. The most striking element of this lab was the variety of research interests represented by the graduate and post-doctoral students gathered there. In residence was a neuroanatomist (mapping forebrain structures), a neurophysiologist (investigating cellular properties of pacemaker nucleus neurons), a biophysicist (exploring how the pacemaker nucleus could generate such a highly precise rhythmic output), a developmental biologist (studying the ontogeny of structure and behavior), an evolutionary molecular biologist (performing a cladistic analysis of mitochondrial DNA to determine the phylogenetic relationships between the many species of electric fish) and finally, a philosopher turned behavioral scientist. What unified the disparate interests of the lab was Heiligenberg's wide-ranging curiosity and the object of study: electroreception in weakly electric fish.

13. To my knowledge, Millikan herself has never explicitly dealt with adaptationism and anti-adaptationist objections, even though there seems to be little else she might be suggesting as a research methodology.

14. For those who doubt that the discovery of electroreception is an issue of content (rather than simply being an issue of function), note that the distinction between detection and reception is a distinction in the kinds of contents used by an organism. The primary difference between an organism possessing electroreception and one which can only electrodetect via another modality is a difference between the perceptual contents accessible by each.

15. “Unnatural” in the sense that Dennett's “adaptationism” differs from the scientific use of “adaptationism”. In its workaday sense, biologists who call themselves “adaptationists” are typically naturalists who would not give up access to the kinds of evidence discussed above without a fight. However, the kind of adaptationism I claim Dennett practices does bear a striking resemblance to the “adaptationism” that is the target of Gould and Lewontin's (1979) criticisms.

16. This particular example concerns content indeterminism. I remind the reader that Dennett repeatedly claims that the logics of content and function attribution are the same. The two issues are intimately intertwined: “After all these years we are still just coming to terms with this unsettling implication of Darwin's destruction of the Argument from Design: there is no ultimate User's Manual in which the real functions, and the real meanings, of biological artifacts are officially represented. . . .You can't have realism about meanings without realism about functions.” (321, emphasis in original).

17. For a project very closely related in spirit to mine, see Akins (1993) on philosophy of mind and the neuroethology of bats.

References

Akins, K.: 1993, `What Is It Like to be Boring and Myopic?' in B. Dahlbom (ed.). Dennett and His Critics: Demystifying Mind, Blackwell, Oxford, 124-160.

Amundson, R.: 1988, `Logical Adaptationism', Behavioral and Brain Sciences, 11, 505-506.

Bullock, T. H.: 1974, `An Essay on the Discovery of Sensory Receptors and the Assignment of Their Functions Together with an Introduction to Electroreceptors', in A. Fessard (ed.), Handbook of Sensory Physiology, (Vol. 3), Springer-Verlag, Berlin, 1-12. Reprinted as Bullock (1993, 171-182).

Bullock, T. H.: 1993, How Do Brains Work? Papers of a Comparative Neurophysiologist, BirkhŸuser, Boston.

Bullock, T. H., Hagiwara, S., Kusano, K. and Negishi, K.: 1961, `Evidence for a Category of Electroreceptors in the Lateral Line of Gymnotid Fishes', Science, 134, 1426-1427. Reprinted as Bullock (1993, 140).

Bullock, T. H. and Heiligenberg, W. (eds.): 1986, Electroreception, John Wiley & Sons, New York.

Bullock, T. H. and Szabo, T.: 1986, `Introduction', in Bullock and Heiligenberg (1986, 1-12).

Catania, K. C.: 1994, The Sensory Biology of the Star-nosed Mole. Unpublished Ph.D. thesis, University of California, San Diego.

Coates, C. W.: 1947, `The Kick of an Electric Eel', The Atlantic, 180, 75-79.

Dalhgren, U.: 1910, `The Origin of the Electric Tissues in Fishes', The American Naturalist, 44, 193-202.

Darwin, C.: 1897, On the Origin of Species, (6th ed.), D. Appleton and Company, New York.

Dennett, D. C.: 1987, The Intentional Stance, M.I.T. Press (A Bradford Book), Cambridge, Mass.

Dennett, D. C.: 1988a, Pr›cis of The Intentional Stance. Behavioral and Brain Sciences, 11, 495-505.

Dennett, D. C.: 1988b, `Author's Response: Science, Philosophy, and Interpretation', Behavioral and Brain Sciences, 11, 535-546.

Dretske, F.: 1981, Knowledge and the Flow of Information, M.I.T. Press (A Bradford Book), Cambridge, Mass.

Dretske, F.: 1988, Explaining Behavior: Reasons in a World of Causes, M.I.T. Press (A Bradford Book), Cambridge, Mass.

Dretske, F.: 1995, Naturalizing the Mind, (The 1994 Jean Nicod Lectures), M.I.T. Press (A Bradford Book), Cambridge, Mass.

Du Bois-Reymond, E.: 1884, Untersuchungen ˆber thierische ElektricatŸt, (Vol. 3), Berlin, Reimer.

Fields, R. D. and Ellisman, M. H.: 1988, `Functionally Significant Plasticity of Synaptic Morphology: Studies on the Ribbon Synapse of the Ampullae of Lorenzini', Neuroscience, 25, 705-720.

Gould, E., McShea, W. and Grand, T.: 1993, `Function of the Star in the Star-nosed Mole', Condylura cristata. Journal of Mammalogy, 74, 108-116.

Gould, S. J. and Lewontin, R. C.: 1979, `The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptationist Programme', Proceedings of the Royal Society of London, 8, 205-258.

Hagedorn, M. and Heiligenberg, W.: 1985, `Court and Spark: Electrical Signals in the Courtship and Mating of Gymnotid Fish', Animal Behaviour, 33, 254-265.

Heiligenberg, W.: 1991, Neural Nets in Electric Fish, M.I.T. Press (A Bradford Book), Cambridge, Mass.

Hempel, C. G.: 1966, The Philosophy of Natural Science, Prentice-Hall, Inc., Englewood Cliffs, N.J.

Hensel, H.: 1955, `Quantitative Beziehungen zwischen Temperaturreiz und Aktionspotentialen der Lorenzinischen Ampullen', Z. vergl. Physiol., 37, 509-526.

Hopkins, C. D.: 1977, `Electric Communication', in T. A. Sebeok (ed.). How Animals Communicate, Bloomington, Indiana University Press, 263-289.

Kalmijn, A. J.: 1982, `Electric and Magnetic Field Detection in Elasmobranch Fishes', Science, 218, 916-918.

Kalmijn, A. J.: 1987, `Detection of Weak Electric Fields', in J. Atema, R. R. Fay, A. N. Popper and W. N. Tavolga (eds.), Sensory Biology of Aquatic Animals, Springer-Verlag, 151-186.

Keeley B.: 1997, Cognitive Science as the Computational Neuroethology of Intelligent Behavior: Why Biological Facts are Important for Explaining Intelligent Behavior, Unpublished Ph.D. thesis, University of California, San Diego.

Kitcher P. and Kitcher, P.: 1988, `The Devil, the Details, and Dr. Dennett', Behavioral and Brain Sciences, 11, 517-518.

Lissmann, H. W.: 1951, `Continuous Electrical Signals for the Tail of a Fish, Gymnarchus niloticus Cuv.' Nature, 167, 201-202.

Lissmann, H. W.: 1958, `On the Function and Evolution of Electric Organs in Fish', Journal of Experimental Biology, 35, 156-191.

Lissmann, H. W.: 1961, `Ecological Studies on Gymnotids', in C. Chagas and A. Paes de Carvalho (eds.). Bioelectrogenesis: A Comparative Survey of its Mechanisms with Particular Emphasis on Electric Fishes, Elsevier Publishing Company, Amsterdam, 215-226.

Lissmann, H. W. and Machin, K. E.: 1958, `The Mechanism of Object Location in Gymnarchus niloticus and Similar Fish', Journal of Experimental Biology, 35, 451-486.

Loewenstein, W. R.: 1960, `Mechanisms of Nerve Impulse Initiation in a Pressure Receptor (Lorenzinian Ampulla)', Nature, 188, 1034-1035.

Loewenstein, W. R. and Ishiko, N.: 1962, `Sodium Chloride Sensitivity and Electro-chemical Effects in a Lorenzinian Ampulla', Nature, 194, 292-294.

Millikan, R. G.: 1984, Language, Thought, and Other Biological Categories: New Foundations for Realism, M.I.T. Press (A Bradford Book), Cambridge, Mass.

Millikan, R. G.: 1993, White Queen Psychology and Other Essays for Alice, M.I.T. Press (A Bradford Book), Cambridge, Mass.

Moller, P.: 1995, Electric Fishes: History and Behavior, Chapman & Hall, New York.

Moller, P. and Fritzsch, B.: 1993, `From Electrodetection to Electroreception: The Problem of Understanding a Non-human Sense', Journal of Comparative Physiology A, 173, 734-737.

Murray, R. W.: 1960, `The Response of the Ampullae of Lorenzini to Mechanical Stimulation', Journal of Experimental Biology, 37, 417-424.

Murray, R. W.: 1962, `The Response of the Ampullae of Lorenzini to Electrical Stimulation', Journal of Experimental Biology, 39, 119-128.

Parker, G. W. and van Heusen, A. P.: 1917, `The Responses of the Catfish, Amiurus nebulosus, to Metallic and Non-metallic Rods', American Journal of Physiology, 44, 405-420.

Rosenberg, H.: 1928, `Die elektrischen Organe', in A. Bethe, G. von Bergmann, G. Einbaum and A. Ellinger (eds.). Handbook der Normalen und Pathologischen Physiologie, Springer, Berlin, 876-925.

Sand, A.: 1938, `The Function of the Ampullae of Lorenzini with Some Observations on the Effect of Temperature on Sensory Rhythms', Proceedings of the Royal Society of London B, 125, 524-553.

Scheich, H. G., Langmer, C., Tidemann, C., Coles, R. B. and Guppy, A.: 1986, `Electroreception and Electrolocation in the Platypus', Nature, 319, 401-402.

Schlegel, P. A. and Richard, P. B.: 1992, `Behavioral Evidence Against Possible Subaquatic Electrosensitivity in the Pyrenean Desman Galemys pyrenaicus (Talpidae, mammalia)', Mammalia, 56, 527-532.

Snow, C. P.: 1959, The Two Cultures: And a Second Look, Cambridge University Press, Cambridge.

Wu, C. H.: 1984, `Electric Fish and the Discovery of Animal Electricity', American Scientist, 72, 598-607.

Textbox 1

Strongly and Pseudoelectric Fish. Strongly electric fish, such as the electric (Torpedo) ray, the electric catfish, and the electric eel, can produce electrical discharges of several hundred volts. This discharge is powerful enough to paralyze large prey and to knock an adult human off his or her feet. In contrast, the electric organs of pseudoelectric fish (later renamed weakly electric fish–see text) produce discharges on the order of millivolts to volts. The most powerful of these discharges is detectible as a slight tingling sensation to a naked hand, although most discharges are too small to be detectible without the aid of artificial amplification.

Textbox 2

Detection and Reception. The suffix -detection is applied to any organism that is capable of responding, by any means, to the presence of a particular type of stimulation in the environment. The suffix -reception is reserved for those organisms that carry out such sensory discriminations through the use of a dedicated anatomical system of structures. For example, thanks to the wonders of modern military technology, a properly equipped human is able to infrareddetect, that is to detect fine distinctions in the infrared portion of the electromagnetic spectrum. This is done by converting infrared radiation into visual stimuli within the range human eyes can detect. However, the attribution of infraredreception is reserved for rattlesnakes (and certain families of boas) because they possess special structures (in small pits below their eyes) reserved for detecting radiant energy in the infrared range. Only the latter classes of organisms are said to possess this sensory modality. (When pedagogy calls for a term neutral relative to this distinction, I will use the suffix -sensation.)

Textbox 3

Active and Passive Sensation. Active perception is sensation through the analysis of changes to a signal produced by the organism. Passive perception is sensation through the analysis of signals originating in the environment. For example, active electrolocation involves the generation of an electrical field in order to detect changes to it. Passive electrolocation, on the other hand, would be the ability to simply detect electrical stimuli without the ability to generate electrical probes. A second example: Humans are passive audiolocators, whereas bats are active audiolocators. (Bats emit high frequency calls and then audiolocate by analyzing the echoes of these calls.)

Figure 1

Body Morphology in the Two Taxa of Weakly Electric Fish. On the left, representative examples of Mormyriform (African) weakly electric fish. On the right, Gymnotiform (New World) weakly electric fish. Despite evidence that the (relatively distant) common ancestor of these two taxa was not itself electroreceptive, fish of these two taxa often share a unique fin morphology in which either the dorsal or anal fin is exaggerated and the lateral fins atrophied. They also often share a knife-like body shape. Such convergent evolution of traits is taken as evidence of a common electroreceptive modality. Adapted from from (Lissmann 1958, 17).

Figure 2

The principle behind electroreception. Top: A top view of one half of a fish (head pointing to the left) and the associated lines of force generated by the dipole electric field generated by its electric organ. This field is altered in systematic ways by the presence of objects with conductive properties different from that of the surrounding water: (a) an object of low conductivity, (b) an object of high conductivity. These changes can be detected at the skin surface of the fish. Bottom: The resolution of: (c) an idealized electrical field in the presence of an object in (d) the original field and (e) a perturbing field. Adapted from Lissmann and Machin (1958, 456).

Figure 3

The Ampulla of Lorenzini. Found in the skin of many species of skates and rays, this organ involves a long, jelly-filled canal which stretches from the surface of the organism to an innervated swelling, or ampulla. This illustration (adapted from Fields & Ellisman 1988) shows the distribution of canals on the ventral surface of the ray Platyrhinoidis triseriata (above), together with a schematic representation of the entire sensory structure itself (below).

Abstract