Keywords

But when I came to the aquarium in the evening at 9 o’clock, to my surprise I found the whole basin empty of fish; only 3–4 Blennius lay on the sandy bottom, the other Blennius I found lying in niches and caves by searching the walls with the arc lamp. Even with strong illumination they remained almost motionless. The Julies, however, had all hidden in the sand lying about 10 cm deep in the basin and could be driven out neither by strong illumination of the sand nor by poking with long sticks (Hess 1909, 15)Footnote 1

11.1 Introduction

Experiments with animals are peculiar. Georges Canguilhem emphasized four circumstances: the “specificity of living forms,” the “diversity of individuals,” the “totality of the organism,” and the “irreversibility of vital phenomena,” such as aging and learning (Canguilhem 2008, 3–22). Because of these peculiarities, findings from animal experiments cannot easily be transferred from one species to another and it is also difficult to compare one set of findings to another or to repeat them. Researchers have developed many measures in response to these challenges, including, for example, the standardization of experimental animals all the way from their genetic makeup to their rearing and housing conditions.

We could extend Canguilhem’s list, however. In experimenting with living animals, there are always at least two organisms involved: the researcher who performs activities and has intentions, and the animal with its own activities and circumstances, which may not be transparent to the researcher.Footnote 2 In this respect, Ian Hacking’s oft-quoted remark that “experimentation has a life of its own” (Hacking 1983, 150) takes on a second meaning. It reminds us that, in contrast to inanimate matter, living beings are not always compliant. In fact, sometimes the challenge is simply to create situations in which the animals behave as desired. While this point is still in line with Canguilhem’s account (although he depicts a surprisingly passive picture of the animal in the experiment) a second point seems beyond the reach of his argument: human researchers cannot share the experience of the animal directly. There remains a gap between the animal’s observable responses and the researcher’s conclusions. From a philosophical point of view, therefore, inferences about animal perceptions and sensations, like inferences about the perceptions and sensations of other humans, are underdetermined.

In the following, I shall discuss how the particularities of research with living beings expand our understanding of experimentation as a controlled procedure and determine the possibility of controlling the insights gained through experiments with animals. For this purpose, I refer to studies on color vision in fish, which the physiologist and eye specialist Carl von Heß and the zoologist Karl von Frisch realized between 1909 and 1914 (for more on this debate, see Munz 2016, ch. 2; Dhein 2021, 743–749; Dhein 2022, 32–36). Their experiments led to divergent results. Heß concluded that fish lack any color sense, whereas Frisch concluded the opposite. The resulting debate between Heß and Frisch, which more famously later included color vision in bees, led to continuous mutual criticism of methodological weaknesses and unsubstantiated conclusions, with the effect that the modes and functions of control were more explicitly expressed than in less controversial studies.

I develop my argument in three steps. I first trace the modes and functions of control in the experiments of Heß and Frisch. I then discuss the special circumstances of their work with animals. I focus on the fundamental fact, rarely discussed, that the experimental animal must be made to follow the experimenter’s intentions. Finally, I am interested in how Heß and Frisch dealt with the problem that conclusions about whether fish perceive colors are systematically uncertain because researchers have access to fish sensory perceptions only indirectly, as mediated by behavioral responses.

11.2 Varieties of Control

The central point in discussions between Heß and Frisch concerned whether or not fish discriminate light of a certain wavelength from light of another wavelength by its color. Heß defended the position that fish perceive shades of gray and not color. As a former assistant of physiologist Ewald Hering, he was familiar with the study of color blindness in humans. He therefore knew that in perception, each spectral color corresponded to a shade of gray of specific brightness (on a scale from black to white), by which even color-blind persons can distinguish it.Footnote 3 With this in mind, Heß built his investigations on two basic experiments (see Hess 1909, 3–9 and 11–14). In the first, he projected a spectrum onto the test tank’s side wall. He then observed in which parts of the spectrum the fish preferred to stay. For the second experiment, he illuminated one tank wall with colored and white light, or with two lights of different colors, such that each covered exactly one half of the wall. Heß then measured how much the strength of one light source had to change so that the fish were either evenly distributed in the tank or moved from one half of the tank to the other. The result of this experiment was a “brightness equation.” Later we shall see how the resulting findings let Heß to the conclusion that fish are fully color blind. These experiments usually work only with juvenile fish, which are phototrophic positive (i.e. attracted by light). To study adult fish, Heß used experiments where fish received colored food for a certain time (see Hess 1911, 421–427). He then tested whether the fish could distinguish paper dummies of the same color from an equally bright gray background or from various gray dummies.

Frisch, in contrast, believed that fish perceive color. Initially he carried out experiments relying on the fact that some fish species can adapt their appearance to the environment (see Frisch 1912b, 188–197; Frisch 1913a, 113–117; Frisch 1914a, 53–62). He worked mainly with the minnow (German Elritze or Pfrille), a species that can change its brightness very quickly, although its adaptation to color happens more slowly. Frisch placed two fish on a gray and a yellow ground that he assumed the fish perceived as equal bright, as determined by their similar outer appearance. Then he observed whether the “gray fish” and the “yellow fish” would change color over a given period of time. If the coloration of the “yellow fish” increased in contrast to the “gray fish,” this change should support color perception. In a second line of research, Frisch expanded on Heß’s experiments with colored food (see Frisch 1914a, 43–53). In the beginning, like Heß, he used dyed food during training and then paper dummies in the actual experiments. Because the training situation differed markedly from the experiments, Frisch started presenting the food on colored glass tubes, which were placed in the test tank together with other gray or colored tubes. The actual experiments used the same glass tubes but without the food.

Heß’s and Frisch’s experiments became the subject of control in many ways. They rarely used the term “control,” however (German: Kontrolle, kontrollieren), either alone or in combination with other expressions. Instead, they usually used related expressions ranging from verifying, checking, testing, and copying experiments, to repeating experiments or trying them once again. This terminology notwithstanding, there are many hints in their reports about procedures that, from an analytical point of view, are to be understood as control activities.Footnote 4

Typically, one encounters the term “control” with respect to experimental instruments. Presenting a new arrangement for measuring brightness equations, Heß explained how to “control the correct setting” (Hess 1914, 255). Less typically, the instruments also include the fish themselves. In all experiments the organisms have a double status: with respect to the question of whether they perceive colors, the fish are the research object; with respect to the behavioral responses used to answer that question, the fish are the means of research. In Hans-Jörg Rheinberger’s terminology, in these experiments every fish was both an epistemic thing and a technical object (Rheinberger 1997, 24–37). It is precisely this circumstance that Heß had in mind when criticizing Frisch for omitting an “indispensable control experiment” from the adaptation studies (Hess 1912b, 634). To assess the results’ reliability, Frisch should first have investigated how sensitive his “measuring instrument” was to differences in brightness—Heß meant the minnow and, more precisely, the minnow’s capacity to adapt to the ground (ibid.). Because Heß showed that the fish reacted to two grounds of different brightness with a similar adaptation, another possibility was equally conceivable: that the yellow and gray ground in Frisch’s tanks, despite Frisch’s assertion otherwise, were not of similar brightness (ibid., 636). For Heß, consequently, the increasing coloration of the “yellow fish,” and the lack of increase for the “gray fish,” had to do with a different stimulus basis (for Frisch’s response, see Frisch 1913a, 110–118).

Sometimes the term “control” appeared when some circumstance had to be excluded as an experimental variable. Frisch’s experiments with fish kept in a monochromatic green and red environment (experiments realized at the Zoological Station in Naples in 1911), were deemed worthless by Heß because an “essential control experiment” had been omitted: Rather than keeping a third group of fish in a transparent test tank, Frisch should have ensured that the light illuminating that tank had the same brightness as the light in the test tanks when surrounded by green and red solutions (Hess 1912b, 632). Frisch himself spoke of “control aquaria” and “control animals” when he referred to the fish in the transparent, “white” test tanks (Frisch 1912b, 209; Frisch 1913b, 153). In addition, Frisch worked with three more groups of fish that had been blinded before starting the experiments and then placed in green, red, and transparent tanks. Although Frisch did not use the term on this occasion, the groups of blinded fish had the character of control groups. In this manner he attempted to discover whether the change in coloration was regulated by skin pigment or the sense of sight.

The term “control” seems particularly pertinent in connection with the repetition of experiments. For example, Heß emphasized that certain of his spectrum experiments were “often controlled measurements” (Hess 1909, 5). Repetition experiments normally help to check research results. When Heß (1913, 439) called the results of all minnow adaptation experiments “completely wrong,” Frisch performed the experiments again—with positive results (Frisch 1914a, 55). The repetition of a competitor’s experiments, in contrast to the repetition of one’s own, casts doubt on their validity. Heß based his negative judgment of Frisch’s adaptation experiments on the fact that his replication failed (Hess 1913, 404–414). Replication did not mean that an experiment was reproduced exactly, in every circumstance. That was common only in repeating one’s own experiments. In repeating their opponent’s experiment, on the other hand, Heß and Frisch almost always modified them. Heß and Frisch gave no reasons for the modifications but we can deduce from their comments that they already considered the experimental design to be flawed, to the extent that an unchanged repetition to them must have seemed pointless.

When Frisch studied the adaptation of minnows to their environment, he initially worked with two fish in two separate aquaria placed on different gray or yellow papers in daylight (Frisch 1912b, 188). Heß, in turn, aiming to “check Frisch’s claims,” worked with groups of minnows and first placed the aquariums in “dim daylight” on surfaces of different colorless brightness, illuminated from below. Later he placed three groups in three different aquariums in daylight, one on red, one on white, and one on black paper (Hess 1912b, 634–637, quote 634).Footnote 5 In his response, Frisch noted these differences and implied that they explained the varying results (Frisch 1913a, 111–113). The feeding experiments offer another example. Frisch (1914a, 44, fn. 1) discounted the refutation of his experiments by pointing out that Heß had used a “diverging experimental design.” But Frisch himself also modified his opponent’s feeding experiments. Instead of the red food of Heß he used yellow food, taking into consideration previous results suggesting that the red end of the visible spectrum might be shortened for fish (Frisch 1912b, 220). Taking these results together, one can conclude two things. First, Frisch saw little sense in identical repetitions of a competitor’s experiment if it seemed to promise only limited insights. Second, as his criticism of Heß’s modifications show, he insisted that only an exact repetition could question a statement based on experimentation.

Although Heß and Frisch do not use the term “control” in this context, experiments can also be understood as control experiments, which serve to deepen new ideas, to stabilize and explore originally random observations, or to check theoretical objections by competitors. An example of such an experiment goes back to Frisch’s argument that the coloration of many fish during the spawning season supports color perception (Frisch 1913a, 121–126). Heß (1913, 401) immediately countered Frisch’s point with experiments on how objects change their color at different depths due to absorption. He concluded that “the so-called ‘wedding dress’ of freshwater fish cannot possibly be conceived as a decoration calculated for the eye in the cases treated here.”Footnote 6 I shall return to this point in Sect. 11.4.

Finally, Heß and Frisch allowed the scientific community to check their experimental procedures and conclusions. Although they did not use the term in this context, the measures they took aimed to control experimental results by intersubjective agreement. Both asked colleagues to judge “blindly,” without prior knowledge, about the color of the ground on which a particular fish was placed in adaptation experiments (Hess 1913, 409–410; Frisch 1914a, 55). Both of their reports included meticulous descriptions of the rules by which they evaluated and compared the coloration of fish (Hess 1912b, 637; Frisch 1913a, 115). And both occasionally included excerpts from the original experimental records in their publications (Frisch 1913a, 116–117; Frisch 1914a, 49–52 and 53–57; Hess 1909, 29–30; Hess 1913, 410–411). In addition, intending to broaden his audience, Heß designed a simple setup that “can be easily applied even by laymen without knowledge of color theory and can often produce surprisingly beautiful results without special optical devices, even without a darkened room” (Hess 1914, 254). Frisch, for his part, preferred public demonstrations. In June 1914, he presented his experiments with bees and fish at the annual congress of the German Zoological Society, allowing the scientific community—at least in principle—to verify his results with their own eyes (Frisch 1914b). The bee experiments went well but most of the fish experiments had to be canceled: all the specimens had died during the congress (Doflein 1914, 710). Frisch remarked in his autobiography: “Unfortunately, the tap water of Freiburg did not agree with the trained fish I had brought with me from Munich, making them sick and useless for my demonstration” (Frisch 1957/1967, 57).

11.3 Cooperation Through Control

Kelle Dhein emphasizes that Frisch gained much more control over the experimental situation in his studies of color vision in fish, and later in bees, than had other researchers previously (Dhein 2022, 34–36). As Munz (2016, 68) points out, Frisch benefited from Heß’s criticisms, just as Heß himself tried to eliminate errors in his own experiments. Nevertheless, as we just read above, experiments with living beings are seldom fully plannable. Plants and animals subvert researchers’ intentions and precautions—they can interact with experimental circumstances and require special attention. The “high level of control” Frisch “exerts in his experiments” (Dhein 2022, 34) was always fragile in practice. Heß and Frisch’s publications are full of revealing remarks: they both mention often that many specimens died during the experiments, and Heß said there was only a short time during which the “material” used was sufficiently “fresh” to produce reliable results (Hess 1909, 2; 1914, 247). Another problem was that fish of the same species varied among themselves. This point came to light in a surprising way, as researchers working with specimens from different regions discovered that whether or not minnows adapted to the environment color depended on their place of origin — which partly explained the different results of Heß and Frisch (Haempel and Kolmer 1914). Frisch also reported that the fish training was important: minnows with several rounds of trials reacted more rapidly (Frisch 1912b, 193–194). Finally, minnows (but also other fish) had a feature unfavorable to adaptation experiments: they also changed color and brightness when disturbed by external influences, such as noise or the transfer from one aquarium to another (Frisch 1912b, 186–187; see also Hess 1913, 405).

Such problems have long been discussed in the history and philosophy of the life sciences in connection with animal experimentation. It is easy to overlook another central challenge in animal experiments, however, and one that has been addressed only rarely. Even when it is acknowledged, it is often only in passing. Obtaining results requires more than doing things like experimenting with the right species variation, accounting for individual differences, standardizing experimental conditions, using statistics, and so on. To obtain results, first and foremost, the animals must cooperate. This is not the case in all experiments, but in those where an animal must solve a task or exhibit a certain behavior, they must actively participate. The experiment cannot be accomplished otherwise. Cooperation in these cases is not another experimental variable, but rather a condition necessary to generate experimental knowledge.

Instructive in this respect is a look into Ivan Pavlov’s chapter on the “Allgemeine Technik der physiologischen Versuche und Vivisektionen” [General technique of physiological experiments and vivisections], published in 1911 in Robert Tigerstedt’s Handbuch der physiologischen Methodik. Pavlov described tools such as grippers, head restraints, and muzzle locks, all of which helped to immobilize animals and ensure smooth access to them. He differentiated between unruly animals—“It is only through necessity that one is induced to experiment on cats—impatient, screaming, nasty animals”—and compliant ones like the rabbit—“a gentle, passive animal that rarely cries out or protests” (Pavlov 1911, 7). And he praised his favorite animal, the dog, which is “irreplaceable” in “chronic experiments,” that is, experiments conducted with the same animal for a long time (ibid.). In his words: “He [the dog] is, as it were, a participant in the experiments that are performed on him, and contributes immensely to the success of the experiments by his understanding and willingness” (ibid.).

Pavlov’s account is certainly euphemistic, but at the same time it shows that the status of the animal in the experiment changes fundamentally when the classical vivisectionist approach, which always results in the death of the experimental animal (in Pavlov’s terminology, the “acute” experiment), is supplemented and largely replaced by an experimental practice that requires, at least for a time, that the animal survive while remaining in as normal a state as possible. As Daniel Todes notes, a single experiment on pancreas secretion usually lasted five to ten hours, during which the animal was to remain as calm as possible. All excitement was to be avoided because physiological factors had to be isolated from mental factors (Todes 2002, 135). For “Druzhok,” the first dog operated on successfully for the experiments, we learn that he “adapted rapidly to this requirement, lying peacefully on the table and ‘taking no particular interest in anything’ during the experiment. Better yet, he greatly facilitated the research by frequently sleeping for five to seven hours at a stretch” (ibid., 136). Once again, we need not believe Pavlov completely, but this kind of experiment was not feasible without the animal “assisting” the experimenter. In fact, this “assistance” begins with the animal surviving an experiment as long as intended. Not dying represents the animal’s minimal active “service” to the experiment, as it were.

Cooperation becomes even more important when the animal subjects’ behavior is unconstrained. In his Intelligenzprüfungen (Intelligence tests) on primates, which relied on results from 1914 to 1918 at the Anthropoid Station on Tenerife, Wolfgang Köhler considered “a foolish but eager animal” to be “more usable” than a “slacker,” or a lazy animal that is not stupid but “never seriously responds to an examination task” (Köhler 1921/1932, 95). If researchers reinforce participation with food rewards (as is even more common today), they must expect that “a completely satiated monkey does not make any considerable efforts for its own sake” (ibid., 97). This point is trivial, but underlines once again that the experimenter in (non-vivisectional) animal experiments depends on the cooperation of the so-called “material.” Nicolas Langlitz reports on the research group around the Japanese primatologist Tetsuro Matsuzawa, noting that their laboratory experiments undertaken since the 1970s “crucially depended on the chimpanzees’ natural dispositions and—to use Matsuzawa’s expression—their ‘free will’ to participate in experiments (which does not necessarily mean that their participation conformed to the researchers’ study protocols)” (Langlitz 2017, 107). In other words, the laboratory was not and is not a place where animals are always “passive guests,” as Robert Kohler (2002, 192) asserted.Footnote 7 Rather, in all experiments requiring a minimum of cooperation, animals must be encouraged to do what is desired.

Heß and Frisch confronted the fact that their experimental subjects did have a life of their own. In saying this I do not assume any human intentionality or consciousness. Rather, the fish belonged to an ecosystem in which they pursued their way of life. Recall the passage from Heß’s first experimental report, quoted at the beginning of the article. When he wanted to extend his experiments from juvenile to adult fish, Heß had to work with a different experimental setup. He planned to project a spectrum into one of the public aquaria of the Naples Zoological Station, where he had conducted his first investigations, in order to observe where in the spectrum the fish recognized descending particles of food. While spectrum experiments were promising only in sufficient darkness, the fish in the aquarium were inactive at night. Instead of foraging, they hid and rested.

Animating a living creature to give evidence for the questions of a research program is not a trivial matter. To achieve this goal, Heß and Frisch followed two strategies. In the first, they designed the experimental situation so that the fish automatically, in keeping with their natural behavior, reacted in ways that helped to answer the question. We see this strategy in Heß’s experiments with phototrophic positive juvenile fish in the spectrum, where the animals gather wherever they find it brightest. His determination of brightness equations offers another example: the fish’s tendency to gather in the brighter part of the test tank allowed him to detect minuscule increments of brightness differences perceived by the fish. Frisch too relied on this strategy when exploiting the capacity of minnows to adapt their coloration to the environment. In the second strategy, the fish were externally motivated to participate in investigating the research question. Heß and Frisch used this technique in their experiments on the association of food with color. The trick was to introduce a suitable stimulus, colored food, which the fish fully internalized by habituation or training.

The first strategy worked by creating experimental situations in which known reactions or behavior patterns can be observed in isolation, as often as required. In contrast, the second worked by adapting known behavior patterns—in this case, the consumption of food—to produce observations useful to the research question. In both, the cooperation of the fish was triggered by the experimenter’s intervention. Somewhat paradoxically, however, in the first case this happened rather passively: cooperation was achieved by controlling the environment, without manipulating behavior directly. In the second case, the experiments worked by actively training the fish so that they would integrate a new stimulus into their usual behavior. Here, cooperation was achieved by controlling the behavior.

Active behavioral and passive environmental control of living beings in the aquarium, which is itself a limited and controlled space, represent two different styles of experimentation, amounting to two different ways of eliciting cooperation. For the experimental setup, passive control need not but often does involve much greater technical effort. Heß’s experiments with spectra and the production of brightness equations demanded finely adjustable electric light sources, prisms, color filters, mirrors, spectrometers, and much other equipment. But implementing active control is no less complex. Rather than measuring tools and elaborate installations, what was needed in this case—apart from glass tubes, food, dyes, and standardized gray and colored papers—were two things: animals sufficiently capable of learning,Footnote 8 and researchers sufficiently capable of training them. Frisch may have been better at this, but not because he was educated as a zoologist whereas Heß was a physiologist. Frisch had simply cared for animals since childhood (see Frisch 1957/67, 19–23).

In general, Heß considered that experiments based on training were not suitable for studying color vision in animals. In a handbook chapter on approaches to studying the sensation of light and color, he acknowledged that the training-based method can be helpful in mammals and birds, but “[u]ncritical application of the method to lower animals has often led astray” (Hess 1921/1937, 308). Going further, he argued that in his experiments—this time referring to the study of color vision in honeybees—the animals, “free of any compulsion,” follow “exclusively their innate inclination to brightness,” whereas animals in training-based experiments had first to learn something new, implying that this was a less natural situation (Hess 1922a, 94). Frisch tacitly agreed with Heß insofar as he distinguished experiments based on training, which he had developed to perfection, from experiments “under natural circumstances” (Frisch 1922/1932, 134). Unlike Heß, however, Frisch understood “natural circumstances” to be the observation of animals in the wild, rather than an arrangement in which animals reacted to light sources in a restricted space.

11.4 Constitutive Ignorance

Animal cooperation does not guarantee the successful execution of an experiment. At the very least, control over the behavior is always temporary. For example, Frisch reported that fish trained on colors recognized the “deception” after several experiments where no food was given. After that they no longer attended to the dummies (Frisch 1914a, 46). In studying color vision in fish, however, Heß and Frisch faced the additional problem of never being completely sure of experimental results. Frisch’s answer to Heß’s objection, that absorption prevents perceiving the coloration of many fish during the spawning season at a depth of only a few meters, is revealing. Frisch replied that Heß transferred “observations made on the human eye [namely observations made by Heß’s eyes] to the fish eye without further ado, which is not permissible” (ibid., 64). Between what Heß and Frisch saw and what the “fish eye” perceived, there remained a gap: while one can control such things as the experimental environment, the fish’s behavior, and many factors influencing the fish’s responses, one cannot control the quality of the fish’s perceptions, sensations, or experiences.

This problem leads from control practices in experiments to the control of conclusions based on experimental findings. From a philosophical point of view, Heß’s and Frisch’s results appear underdetermined—there is no immediate relationship between the fish’s observable behavior and its presumed causes. As with all such experiments on non-human animals, researchers could make only plausible guesses to explain behavioral responses, or why an animal reacts in a certain way. Heß tackled this point directly in his first comprehensive review of the sense of light and color in animals. He started with a fundamental consideration:

The question as to whether it is possible to identify a color sense existing in animals is still often answered with no. One justifies such an opinion with the fact that the animals are not able to give us information about their optical perceptions, as the human being is able to do by designating different colors. One forgets here that the designation can give us only imperfect information about the visual qualities of a human being: If someone calls an object red which appears red to us, we still do not know whether his visual qualities correspond to ours. (Hess 1912a, 1).Footnote 9

From this starting point, Heß concluded that a reliable basis for studying the color sense, in humans or animals, could be attained only by measuring which lights appear differently or the same to an eye. However, although animals could show behaviorally that light appeared to them the same or different, unlike humans they could not indicate what appeared to them to be the same or different about the lights. Was it the brightness, or the color? Or something completely different? To solve this problem, Heß resorted to studying the human eye. From Hering’s research on human color blindness, he knew that total absence of color vision is accompanied by a characteristic shift in the perception of the brightest part of the spectrum from yellow to yellow-green, and a shortening of the spectrum at the red end (Hess 1909, 35). If an animal eye showed the same characteristics, Heß also considered it totally color blind. From his brightness equations he believed this to be true for fish, concluding his first communication with the statement that “the relative [perceived] brightnesses [...] correspond almost or completely with those in which the totally color-blind human being” perceives colored surfaces (Hess 1909, 35). After studies on more animal species, especially bees, crustaceans, and worms, Heß found that, on the whole, the eyes of fish and invertebrates resembled those of a color-blind human eye in terms of their “visual qualities” (Hess 1912a, 151).

The way Heß inferred the conditions of animal eyes from the human eye indicated that, for him, all eyes—fish, bee, or human—were alike in their basic properties.Footnote 10 For Frisch, on the other hand, Heß’s experiments proved only that “the brightness values of the colors for fish and invertebrates are the same as for the totally color-blind human being” (Frisch 1923, 471). Thus, a typical “characteristic of the total color blindness of man is found,” but the truly essential characteristic is not the brightness distribution of colors in the spectrum. It is that colors be “distinguished only according to their brightness, not according to their quality” (Frisch 1919, 123).

In turn, Heß criticized (here referring to bees, but similarly valid for the experiments with fish) that Frisch’s experiments, if they had produced any tenable results at all, in his eyes “at best provide information about the fact that one of two or more surfaces appears different to the bees than the others, but not about how the two surfaces are different for them; but this alone is of interest for the color sense question. “(Hess 1922a, 94) And, in fact, Frisch could not be sure about this. In tests, for example, fish trained on yellow also visited red dummies, and fish trained on red visited yellow dummies (Frisch 1914a, 45). This result could mean several things: that yellow and red appeared to the fish as the same color; that they appeared to the fish as the same grayish brightness; or that they appeared to the fish as different, but the difference was not large enough to play a role in behavior. Nevertheless, it was obvious to Frisch that color was decisive for the fish’s behavior in all his experiments.

In his very first report on the color sense in fish at the Congress of German-Speaking Zoologists in 1911, he combined his results with a more teleological rationale.Footnote 11 After he had introduced Heß’s experiments, Frisch remarked:

Many fish possess the ability to change color to a very marked degree. When you see how some of them take on the most splendid colors as a result of a nervous influence on the pigment cells at spawning time, it is hard to believe that they are color blind. (Frisch 1911, 222).Footnote 12

Frisch admitted that “it would not be right to attribute much significance” to such “intuitive arguments [Gefühlsargumente]” (ibid.). Nevertheless, in his publications on color vision in fish and later in bees, he frequently pointed out that the so-called “wedding dress” of fish or the relationship between flower colors and the pollination of flowers by bees make the assumption of a color sense seem compelling. Frisch concluded his lecture “Über Färbung und Farbensinn der Tiere [On Coloration and the Sense of Color in Animals]” of mid-1912 with the following words:

And I believe that each of us would only acquaint himself with a certain discomfort with the prospect that all the splendor of flowers is a coincidence; that by chance the flowers that are pollinated by the wind are so inconspicuously colored, by chance the flowers that are set up for pollination by insects are so conspicuous, and so often conspicuous by their color; that everything here might as well be gray in gray. (Frisch 1912a, 38).Footnote 13

Frisch’s considerations show that he ultimately made a philosophical argument. One could observe bees visiting flowers. Similarly, one could observe the mutual advantage of this activity: the bees gathered nectar, and the plants with striking flower colors depended on cross-pollination. That flower-visiting insects such as bees therefore possessed a sense of color, however, followed only if Frisch assumed a purposeful organization of nature—an organization in which it was difficult to imagine chance. Referring to Frisch’s argument, Heß spoke of the “suggestive effect” inherent “in long-established trains of thought” (Hess 1922b, 1239).

Frisch himself was not very convinced of the epistemological significance of such more philosophical considerations. Remember that he spoke of “intuitive arguments.” Nevertheless, they must have been indispensable to him, as their regular repetition indicates.Footnote 14 Just like Heß, he closed the gap between what could be observed in a controlled manner—manifold behavioral reactions—and what one wanted to find out—how do animals perceive light of a certain wavelength?—with a supplementary assumption. However, neither Heß nor Frisch would have spoken of supplementary assumptions. For them, a fish’s ability to perceive or not perceive colors was a matter of plausible conclusions.

A fully satisfactory answer to the question of whether fish perceive colors and, if so, how those colors must be imagined, would require several things. For one, we would need to control the environment and the behavior, but we would also need immediate access to the perceptions of other living beings. So far this has not been possible. This may explain why the debate between Heß and Frisch was so persistent and never yielded a resolution.Footnote 15 The debate ended only with Heß’s death in June 1923. From today’s point of view, Frisch’s results have, by and large, been confirmed. But as ethologist Niko Tinbergen noted in the 1940s and many others after him, whether fish and bees react as if they are colorblind or as if they perceive colors ultimately depends on the situation (Dhein 2021, 746).

11.5 Summary

Experiments with inanimate matter and those with living beings may differ less than suggested by Canguilhem. Among the modes and functions of control that Heß and Frisch employed, none was specifically adapted to organisms. Like many other experimenters discussed in this volume, they tried to identify, separate, and stabilize experimental conditions, and they looked for disruptive factors. Even Frisch’s work with control groups, common in biological and medical research today, is not limited to experiments with living beings. Today, researchers usually speak of “groups” when they simply mean “sets of trials.” The specificity of experimenting with living beings does not necessarily arise from the variability of their properties and behaviors. A materials scientist can report similar experiences when studying the properties of new composites, for example. Rather, the most distinctive property of these experiments seems to be that living beings are not readily at hand for the researcher. One can stock them in large quantities, but the “material” (as scientists often call research animals) must still cooperate in many cases, and especially in the more interesting ones.Footnote 16

Evan Arnet (this volume) provides an example of how animals can be motivated on their own to participate in an experiment—in this case, rats completing learning experiments in a maze. In that context, the hope of controlling motivation by food deprivation proved to be doubtful, if not an illusion. Going further, one could say that maze-based learning experiments use a particularly well-adapted form of environmental control, in which the animals show with good grace the behavior necessary for the research goal. At the same time, the example of the rat shows that it is not the scientist alone who controls and ensures the cooperation of the animal. Rats may like to poke around, but other tasks are not part of their behavior. For example, rats and mice are the experimental animals for alcohol research in the United States. But because they tend to prefer water, researchers must induce in them an alcohol dependence (see Ankeny et al. 2014, 493–494). An important criterion for the validity of the results obtained in this context is whether the animals are more or less forced to consume alcohol.

I wanted to underline this peculiarity of animal experiments when I emphasized that animals in experimentation lead “a life of their own.” However, we should not assume that animals participate in experiments in the same way as the experimenter. Susan Leigh Star and James Griesemer (1989, 401) once asked the question: “How does one persuade a reluctant and clever animal to participate in science?” Star and Griesemer’s famous essay on the boundary object is about how actors living in different social worlds can be stimulated to cooperate for a common cause—in this case, a zoological collection. That animals apparently also participate in this enterprise, however, and are listed alongside the scientific director and other contributors, obscures the fact that the animals are not “persuaded” but overwhelmed. In fact, the human actors in Star and Griesemer’s story considered the animals mainly as “recalcitrant” (ibid., 402). Generally speaking, the status of animals in experiments perhaps most closely resembles that of the ignorant, uninformed subjects in Carl Stumpf’s auditory experiments (see Kursell, this volume). Animals participate but remain uninvolved; the difference is that they often pay for their participation with their lives. In a scientific context, animals have a life of their own only insofar as they compel the researcher to make an extra effort to control them.

We may distinguish four concepts of control in the studies of Heß and Frisch on color vision in fish. The first is control as an activity to isolate and explore variables potentially significant to the research subject. The second is control as an activity to confirm or subvert experimental findings. Third is control as the activity of other scientists, who participate in reviewing results; here I am thinking of Frisch’s public demonstrations of his experiments, for example. And the fourth is control as an activity concerned with the proper function of instruments, the care of the animals’ living conditions, and the measures ensuring their cooperation in the experiments. I have emphasized this last point because it is rarely discussed within the broad notion of control as management of experimental situations (Schickore, this volume). Heß and Frisch took different approaches toward this goal, one based on environmental control and the other on behavioral control. Whether these strategies also evince disciplinary differences at the beginnings of the twentieth century, with Heß trained in physiology and Frisch trained in zoology, requires further investigation.

Finally, a question arises about the extent to which the basic problem of this research shapes experimentation: what can I learn about the sensations of a living being that is not myself, and with whom I cannot communicate? In this respect, the two methods for encouraging fish to cooperate constitute two different answers to this problem. Environmental control couples with the design of situations in which fish appear to respond spontaneously to the “question” posed by the researcher. When researchers, in turn, attempt to ensure cooperation by controlling behavior, it seems that the fish’s ability or inability to respond unambiguously to a certain stimulus either directly confirms or directly disproves the existence of color vision. In both cases it appears that researchers favor observations, which seem to limit the scope for interpretation. But even then, the problem does not completely disappear. As we have seen, Heß and Frisch still felt that they must introduce additional plausible assumptions to strengthen their positions.