Abduction in the Everyday Practice of Science: The Logic of Unintended Experiments

Abstract

Generating new ideas—innovation and novelty—is central to what those of us practicing science hope to accomplish. We call it research, but what we really aim for is new-search—learning new things about the world and how it works. Charles Peirce gave the name "abduction" to what he described as the only logical operation that introduces any new idea. In this paper, I will focus on an unconventional understanding of abduction, one that goes beyond its usual meaning and concerns the situation when a surprising observation becomes reconfigured as an unintended experiment about a new research problem that previously was not being studied and possibly was unknown. For science, the consequences can be the beginning of a new field of investigation. For the researcher, the consequences can be life-changing.

Part I: Introduction

I first learned about Charles Peirce's concept of abduction reading the two volume set of papers about discovery edited by Thomas Nickles and based on the 1978 conference at the University of Nevada in Reno that brought together the so-called "friends of discovery" (Nickles 1980a, 1980b). The half dozen or so philosophers who commented about Peirce and abduction in those papers had mostly critical things to say, but Peirce's logic of abduction caught my attention.

Over the past 30 years, I have been writing about the nature of science, mostly from a phenomenological perspective, reflecting on my experiences doing research (Grinnell 1992, 2009). Discovery in science interests me a lot. Contrary to the idealized view of science typical of science education, the path to discovery tends to be highly ambiguous and convoluted rather than linear; failure is frequent; the pressure to produce is great. Some of the most important experiments are unintended. Peirce's logic of abduction,

The surprising fact, C, is observed.

But if A were true, then C would be a matter of course.

Hence, there is reason to suspect that A is true.

(CP 5.189, EP 2:231, 1903)

attracted my attention because it corresponds exactly to what I think of as the logic of unintended experiments.

Part II: The Conventional View of Abduction and Sherlock Holmes

One longstanding view in philosophy equates abduction with the idea of inference to the best explanation (IBE) (Harman 1965, Lipton 2003). Not all philosophers agree. Some argue that in his later writing Peirce meant to emphasize abduction as generating rather than testing explanatory hypotheses, whereas IBE is more concerned with evaluating hypotheses that have already been generated (Minnameier 2004, Campos 2011, Mcauliffe 2015). Others suggest that abduction is neither generating nor evaluating, but rather selecting which hypotheses should be evaluated, i.e., which are pursuit worthy (Laudan 1978, McKaughan 2008). And yet another point of view is that Peirce meant all of the above, i.e., abduction as insight and inference combined (Anderson 1986).

Notwithstanding the diverse philosophical opinions mentioned, understanding abduction as IBE has become commonplace outside of Peirce scholarship. The adjective "abductive" now can be found used widely across diverse fields of inquiry including in the titles of books about computational analysis and artificial intelligence (Josephson and Josephson 1996), theory formation in social science (Timmermans and Tavory 2012), and everyday life experience (Walton 2014). And if one looks up "inference to the best explanation" in the table of contents of the Stanford Encyclopedia of Philosophy, the reader is directed as follows: "see abduction."

Sherlock Holmes' reasoning is often mentioned as representative of abductive thinking (Fann 1970, Eco and Sebeok 1983). Here is an example from the story "Silver Blaze" (Doyle 1905):

Colonel Ross still wore an expression which showed the poor opinion which he had formed of my companion's [Holmes] ability, but I [Watson] saw by the inspector's face that his attention had been keenly aroused.

Inspector:

Is there any point to which you would wish to draw my attention?

Holmes:

To the curious incident of the dog in the night-time.

Inspector:

The dog did nothing in the night-time.

Holmes:

That was the curious incident.

Rewritten in the format of Peirce's logical scheme of abduction, the story becomes:

The surprising fact C (the dog did nothing in the night-time) is observed.

But if A (the dog was familiar with the killer) were true, then C would be a matter of course.

Hence, there is reason to suspect that A (the dog was familiar with the killer) is true.

In the Sherlock Holmes story, the hypothesis—the dog was familiar with the killer—fits all three ways of understanding abduction—(i) a new hypothesis; (ii) a new hypothesis worth pursuing; and (iii) a likely explanation for what had happened (IBE). It is of particular importance for the discussion in this paper that in all three cases the surprising fact and corresponding abductive reasoning occurs within a particular context. Holmes was consulted to figure out who stole the race horse and committed the murder. He begins with a problem at hand, and he solves just that problem.

Part III: Lewis Thomas and the Case of the Floppy-Eared Rabbits

Now I will introduce a different aspect of abduction, which the following example from the history of science illustrates. In April 1956, biomedical scientist Lewis Thomas published a report in the Journal of Experimental Medicine entitled "Reversible collapse of rabbit ears after intravenous papain, and prevention of recovery by cortisone" (Thomas 1956).

The article begins:

For reasons not relevant to the present discussion rabbits were injected intravenously with a solution of crude papain, and the following reactions occurred with unfailing regularity: Within 4 hours after injection, both ears were observed to be curled over at their tips. After 18 hours they had lost all of their normal rigidity and were collapsed limply at either side of the head, rather like the ears of spaniels. After 3 or 4 days, the ears became straightened and erect again.

The published paper included Figures 1A and 1B to illustrate the rabbit ears before and after papain treatment. The paper goes on to report additional experiments that showed that ear collapse was associated with a change in ear cartilage matrix; that similar changes occurred in all the other cartilage tissues of the rabbit's body; and that when the ears returned to normal shape, the cartilage matrix also had returned to its original characteristics.

When asked how the research had come about that led to this work, Thomas commented that five years earlier,

I was trying to explore the notion that the cardiac and blood vessel lesions in certain hypersensitivity states may be due to release of proteolytic enzymes. It's an attractive idea on which there's little evidence … [Injecting several different enzymes including papain into rabbits intravenously was intended to test the hypothesis.] [W]hat papain did was always produce these bizarre cosmetic changes. It was one of the most uniform reactions I'd ever seen in biology. It always happened. And it looked as if something important must have happened to cause this reaction.

(Barber and Rox 1958)

As in the case Silver Blaze, the case of the floppy-eared rabbits can be rewritten according to the logic of abduction.

The surprising fact C (i.v. papain injection caused rabbit ear flop) is observed.

But if A (rabbit ear rigidity depends on a papain-sensitive mechanism) were true, then C would be a matter of course.

Hence, there is reason to suspect that A (rabbit ear rigidity depends on a papain-sensitive mechanism) is true.

The abductive format is the same as in the Sherlock Holmes case but incorporates an important added feature. Unlike Holmes and the mystery of Silver Blaze, the surprising fact of the ear flop was irrelevant to the question at hand, viz., the role of i.v. proteolytic enzymes in cardiac and blood vessel lesions in hypersensitivity states. Instead, the surprising fact suggested a new problem—the mechanism underlying rabbit ear rigidity—a question that Thomas had not thought about before observing the surprising experimental results. His findings about rabbit ear flop contributed to the emergence of a new problem and research field—proteinases and destruction of cartilage matrix—a potential mechanism of osteoarthritis.

Peirce described as a gestalt switch the connecting link between perception and abduction. "I will show you a figure," he wrote,

which I remember my father [Benjamin Peirce] drawing in one of his lectures.… Here is the figure (though I cannot draw it as skillfully as he did). It consists of a serpentine line. But when it is completely drawn, it appears to be a stone wall.… [As an aside, only someone who has seen a New England stone wall would notice the resemblance.] The point is that there are two ways of conceiving the matter.… So it is with that well-known unshaded outline figure of a pair of steps seen in perspective [the Schroeder Stairs Ambiguous Figure (1858)—a classic gestalt switch example]. We seem at first to be looking at the steps from above; but some unconscious part of the mind seems to tire of putting that construction upon it and suddenly we seem to see the steps from below, and so the perceptive judgment and the percept itself seem to keep shifting from one general aspect to the other and back again.

In all such visual illusions, of which two or three dozen are well known, the most striking thing is that a certain theory of interpretation of the figure has all the appearance of being given in perception.… [T]hese phenomena [gestalt switches] are true connecting links between abductions and perceptions.

(CP 5.183, EP 2:228, 1903)

The figure below diagrams the gestalt switch in Lewis Thomas's experience by placing the events in the form of puzzle solving. Thomas began with an intended hypothesis to test—the puzzle on the upper left—do i.v. enzymes cause cardiac and blood vessel lesions such as occur in hypertension? The rabbit ear flop puzzle piece in the middle fits into the empty spot in the puzzle shown on the upper right. Since the observation didn't advance understanding of blood vessel lesions in any obvious way, Thomas could have simply moved on to experiments that focused on other possible puzzle pieces (e.g., other enzymes or other potential causes of damage).

The abductive moment—gestalt switch—came about when Thomas reframed the unexpected observation in the context of a different question—the puzzle on the lower left—does the mechanism of rabbit ear rigidity depend on an i.v. papain sensitive mechanism? If one were interested in the latter question, then injecting i.v. papain would have been a very good experiment to carry out. The surprising rabbit ear flop puzzle piece would no longer be surprising in the context of the second research puzzle and makes a lot of sense when placed on the lower right. Thomas had carried out an unintended experiment.

As I will discuss later, for Lewis Thomas (or any researcher) to follow up the results of an unintended experiment is a potentially risky choice. Peirce embraced the value of accepting this risk when he contrasted hypotheses that offered security (likely correctness) with those that offer uberty (fruitful potential). As the Editors of EP 2 put it, "Deductive reasoning provides the most security, but little uberty, while abduction provides much uberty but almost no security" (EP 2:463, 1913). The latter, although "hazardous, may put us upon the track of important truths that no safer ways of reasoning could ever suggest to us" (EP 2:472, 1913). Indeed, Peirce writes that "the true policy for men of science"—after criticizing scientists' conservative tendency—"would be to encourage as much as possible those young persons who seem to show a genius for this sort of research" (R 683:10, c.1913).

In "On the Logic of Drawing History from Ancient Documents," Peirce describes abduction as leading to discovery of entirely new research problems. The scientific impulse will always be in the position of striving to reconcile the new to the old. "Thus it is," he writes,

that all knowledge begins by the discovery that there has been an erroneous expectation of which we had before hardly been conscious. Each branch of science begins with a new phenomenon which violates a sort of negative subconscious expectation, like the frog's legs of Signora Galvani.

(CP 7.188, EP 2:88, 1901)

As Vannevar Bush wrote in his 1945 science policy report Science: The Endless Frontier, history of science shows that "[m]any of the most important discoveries have come as a result of experiments undertaken with quite different purposes in mind" (1945: 75). This was Bush's argument put forward to support curiosity driven research, and his report is often viewed as the founding document for the organization of basic science research funding in the U.S.

Part IV: Abduction in the Larger Scheme of Research

In his dialogue Meno, Plato has the title character ask: "How will you look for it, Socrates, when you do not know at all what it is?" (§80d) Discovery at the edge of knowledge means looking for something without being exactly sure what it looks like and guessing what might be the answer (Tschaepe 2013).

Conducting an experiment to test a hypothesis begins with an investigator's explicit and implicit assumptions. Explicit assumptions concern the experimental question to be tested and the imagined likely results to be obtained. Implicit assumptions concern the adequacy of the experimental design and methodology selected to accomplish the research.

Since the answer is not known in advance, every experiment tests both explicit and implicit assumptions. Because of this ambiguity, failure to get the expected results might be the result of a wrong hypothesis or inadequate experimental design. Max Delbrück, a winner of the 1969 Nobel Prize in Physiology or Medicine and one of the founders of modern molecular genetics, called this ambiguity the principle of limited sloppiness— sloppiness in the sense that our knowledge about any system under investigation is always muddy—never completely clear (Hayes 1982). As a result, during our experiments, we sometimes test unintended questions as well as those explicitly intended.

In practice, opportunities for abduction often occur when implicit assumptions turn out to be incorrect—when assumptions about experimental design and methodology that are taken for granted lead to results other than those that are expected. Here is an example from my own experience.

In 1970, I joined biochemist Paul Srere's laboratory as a postdoctoral fellow. Srere was interested in diabetes and fatty acid synthesis. He had discovered that rats, starved and then re-fed, increased 200 fold their liver levels of the Krebs Cycle enzyme known as citrate cleavage enzyme (Foster and Srere 1968). My project was to use rat liver cells in culture to reproduce the starvation/refeeding effect and identify the regulatory mechanisms involved.

The liver-derived cells that I was given for these experiments were able to grow floating in culture medium or attached to culture dishes. I decided to treat cells floating in culture medium with "starvation inducing" reagents, let the cells stick to culture dishes, rinse them with fresh or starvation medium, culture further, and then test citrate cleavage enzyme levels. I began the experiments, but I was surprised to observe that cells no longer attached to the culture dishes if they were treated with the metabolic inhibitor sodium arsenite; sodium fluoride, a different metabolic inhibitor, had no effect. What next?

Experiments typically fit into one of three categories: heuristic, demonstrative, and failed (Fleck 1979). Heuristic experiments offer researchers new insights into the problem under investigation. Demonstrative experiments clarify heuristic findings into a form suitable for making discovery claims public, what philosophers typically call the logic of justification. Failed experiments, perhaps the most common, are those that yield results that are inconclusive or uninterpretable, which may occur for many reasons including technical errors, mistaken assumptions about methods, and poor study design. As a result, in research publications, ten research notebooks frequently can be represented by ten figures.

Surprising observations, at least initially, typically are assumed to be failed experiments.

Are you sure the dog did nothing in the nighttime?

Are you sure something else didn't happen to the rabbits besides papain injection?

Are you sure the culture dishes weren't dirty?

Even once observed and confirmed, surprising observations sometimes will be ignored as technical problems to be overcome. In his memoir The Statue Within, Francois Jacob, a winner of the 1965 Nobel Prize in Physiology or Medicine, describes the surprising failures that he and his colleagues encountered as they tried to demonstrate the existence of messenger RNA.

We were to do very long, very arduous experiments.… But nothing worked. We had tremendous technical problems.… Full of energy and excitement, sure of the correctness of our hypothesis, we started our experiment over and over again. Modifying it slightly. Changing some technical detail. (Jacob 1988: 315)

It turned out that their hypothesis was correct; the method was wrong. Experienced researchers know—Don't give up a good hypothesis just because the data do not fit.

Surprising facts that become moments of abduction are often labeled instances of serendipity in science. Sir Peter Medawar, who in 1960 won a Nobel Prize in Physiology or Medicine for his work on transplantation immunology, liked to point out that there is a big difference between finding a winning lottery ticket and buying one. The researcher who buys a lot of tickets puts himself in the winning way—creates opportunities for discovery to occur (Medawar 1984). Serendipity is all about opportunity. Abduction goes further and requires noticing and making the gestalt switch.

Understanding the impact of unintended experiments is important not only to following the history of science, but also to promoting the likelihood of a researcher's success in the laboratory or field. Becoming self-aware and open to noticing the unexpected will facilitate discovery. Nothing noticed—novelty lost!

Part V: Why "Surprise" Matters

Most philosophers don't take seriously the "surprise" in Peirce's abduction scheme. Why do I say that? When Gilbert Harman (1965) writes that IBE corresponds to approximately what others have called "abduction," he uses examples in which surprise doesn't figure. Indeed, it has been suggested that the notion of "surprise" really isn't necessary for abduction (Hoffman 1999).

Peirce, on the other hand, takes surprise very seriously. Jaime Nubiola (2005) counted 127 appearances of the word "surprise" in the Collected Papers and suggested calling abduction the logic of surprise. In an unpublished draft letter to contemporary Victoria Lady Welby, Peirce writes, "[This] reasoning might be called 'Reasoning from Surprise to Inquiry'" (R L 463, July 16, 1905; quoted in Pietarinen and Bellucci 2014).

Surprise is critical to the sense of abduction that I am describing for two reasons. The first reason is that the surprising fact must be sufficiently surprising to attract a researcher's attention. Results that fall outside a researcher's expectations will often go unnoticed. Things haven't changed since Claude Bernard, one of the founders of modern biomedical research, wrote 150 years ago that

[m]en who have excessive faith in their theories or ideas are not only ill prepared for making discoveries; they also make very poor observations. Of necessity, they observe with a preconceived idea, and when they devise an experiment, they can see, in its results, only a confirmation of their theory. In this way they distort observations and often neglect very important facts because they do not further their aim.

(Bernard 1957)

Most of us are not like Sherlock Holmes. His ability to notice everything is what makes him so appealing. The rest of us, like Watson, tend to overlook the unexpected or sometimes the absence of the expected. My favorite example is the humorous element inserted in the diagram of the photosynthesis apparatus illustrated in Melvin Calvin's classic 1955 paper "The photosynthetic cycle: CO₂ dependent transitions" (Wilson and Calvin 1955). In 1961, Calvin won the Nobel Prize in Chemistry for his work. The humorous element went unnoticed throughout the paper's review and subsequent publication and was not seen until many years later. (If the reader decides to go looking and can't find it, then compare Figures 2.2 and 2.9 in Grinnell 2009.)

The second reason that surprise is critical to the sense of abduction that I am describing is that the surprising fact must be sufficiently intriguing to overcome the resistance of a researcher to considering focusing on a new problem for investigation. When one decides to study a research problem, they take for granted important assumptions: that there is a question unanswered; that the question will be worthwhile answering; and that the infrastructure, personnel, and financial resources necessary to succeed are available. Time, energy and money are limiting resources in laboratory life. Starting something new is risky. Investing in one project almost always means that something else will not be accomplished. Failure could slow down or even end one's career in science.

Now back to the citrate cleavage enzyme story. I showed Srere my results after repeating the experiment. I told him that in trying to understand why the cells didn't attach, I learned that his carbohydrate biochemist friend Saul Roseman had started to study attachment between cells and had proposed the novel idea that the mechanism of adhesion was formation of an enzyme-substrate complex. Srere called Roseman to ask his opinion about the enzymology of cell adhesion—was it a matter worth pursuing? Roseman said "Yes," that cell adhesion was a poorly understood but fundamental feature of biological organisms. Adhesion holds us together—literally. There are many human diseases the pathology of which reflects changes in cell adhesion.

With Roseman's encouragement, Srere let me begin a new line of research in the laboratory. Separately, Roseman wrote, "The chemistry of this process is obviously unknown and any contribution that you make will undoubtedly be a major one." Suddenly, I was studying the enzymology of cell adhesion instead of regulation of metabolic enzymes (Grinnell and Srere 1971). The surprising fact and new research problem that it brings to mind must be sufficiently surprising to abduct the researcher's mind away from the initial problem at hand to pursue studies on a new project.

What about the word "abduction"? Peirce writes that he takes the idea of abduction from Aristotle's apagōgē:

Aristotle, in that chapter [Prior Analytics II.25] …was even in that case evidently groping for that mode of inference which I call by the otherwise quite useless name of Abduction,—a word which is only employed in logic to translate the [ἀπαγωγή] of that chapter.

(CP 5.144, EP 2:205, 1903)

Why Peirce chose the "otherwise quite useless name of abduction" for apagōgē is unknown. Might his choice have been connected to the idea of abducting the researcher's mind to something new? Peirce's contemporary William James used the word "abduction" in just that fashion in a letter to James' brother Henry:

[Oliver Wendell Holmes, Jr.] grows more & more concentrated upon his law. His mind resembles a stiff spring, which has to be abducted violently from it, and which every instant it is left to itself flies tight back. (Quoted in Menand 2001: 226; my emphasis)

Final Comment

In conclusion, I am suggesting a second way to understand abduction that includes conventional ideas but incorporates an important additional feature. That is, for a research scientist doing experimental work, abduction sometimes describes the logic of a surprising observation that becomes reconfigured as an unintended experiment about an entirely new research problem. The consequences can be transformative, leading to a new research trajectory. Looking back on one's life in science, researchers often will be able to understand their experiences as histories of abductive moments gained and (in retrospect) lost.