Model Behavior

Animal Experiments, Complexity, and the Genetics of Psychiatric Disorders

Nicole C. Nelson

• Publisher: University of Chicago Press
• Year: 2018
• Print ISBN: 9780226546087, 9780226545929
• eBook ISBN: 9780226546117
• Topic: Biological Sciences
• Subtopic: Psychopathology
• Index: Biological Sciences

Mice are used as model organisms across a wide range of fields in science today—but it is far from obvious how studying a mouse in a maze can help us understand human problems like alcoholism or anxiety. How do scientists convince funders, fellow scientists, the general public, and even themselves that animal experiments are a good way of producing knowledge about the genetics of human behavior? In Model Behavior, Nicole C. Nelson takes us inside an animal behavior genetics laboratory to examine how scientists create and manage the foundational knowledge of their field.

Behavior genetics is a particularly challenging field for making a clear-cut case that mouse experiments work, because researchers believe that both the phenomena they are studying and the animal models they are using are complex. These assumptions of complexity change the nature of what laboratory work produces. Whereas historical and ethnographic studies traditionally portray the laboratory as a place where scientists control, simplify, and stabilize nature in the service of producing durable facts, the laboratory that emerges from Nelson's extensive interviews and fieldwork is a place where stable findings are always just out of reach. The ongoing work of managing precarious experimental systems means that researchers learn as much—if not more—about the impact of the environment on behavior as they do about genetics. Model Behavior offers a compelling portrait of life in a twenty-first-century laboratory, where partial, provisional answers to complex scientific questions are increasingly the norm.

Chapter 1

Containing Complexities in the Animal Behavior Genetics Laboratory

Ordinarily after the introductory behavior genetics class at Coast University, the graduate students dispersed almost immediately, but today everyone was unusually still. We were in a small classroom deep in the bowels of the teaching hospital, and Dr. Laura Martin, a senior investigator in the Department of Neuroscience, had just finished delivering a lecture on environmental interactions and mouse behavior. In her typical direct manner, she had presented experiment after experiment showing the variety of ways in which the environment could change a mouse’s behavior. Not only could experiences in the womb or parental behavior early in life impact a mouse’s later behavior, but even the placement of the cages in the mouse housing rooms or the light levels in the testing space could alter behavioral test results. Each slide layered on a new set of variables affecting mouse behavior until I felt as though I’d been buried in an avalanche of competing factors. Considered individually, each one of these factors seemed reasonable enough to take into account when setting up behavioral experiments, but their collective impact was overwhelming. Judging by my fellow graduate students’ contemplative faces, I was not the only one feeling a mix of awe and frustration by the end of the lecture. In the semidarkness of the classroom, illuminated by the glow from Dr. Martin’s final PowerPoint slide, this moment of shared stillness suddenly felt quite intimate. “With all of this complexity,” one of the aspiring scientists remarked, breaking the silence, “it’s hard to feel like you have a prayer.” Dr. Martin nodded and reassured them that they were all in the same boat. Sometimes, she said, she felt depressed by it, too.

“Complex” was one of the most frequently used adjectives I heard during my time at Coast. Mouse behavior was complex, according to Coast researchers, as were the experimental setups used to test it. The resulting data sets and their interpretation were complex as well. The genetic factors underlying these complex behaviors were themselves complex, full of multiple interacting factors. As the scientists talked, complexity often began as a quality of the entities that they dealt with and transformed into an independent entity of its own that they grappled with. Researchers talked about complexity as though it was the ghost in the machine animating the objects they dealt with, making genes and neurotransmitters and mice behave in unpredictable, inscrutable ways. The mood that scientists slipped into during such conversations resembled the mix of awe and frustration I had experienced in the introductory behavior genetics class. Researchers talked about the complexity of biological processes reverently but also with a barely contained sense of exasperation that behavioral phenomena continually overflowed the boundaries of the experiments that they had so carefully constructed to contain those complexities.

This chapter explores what researchers at Coast meant when they described things as complex, and how they arrived at a sense of what it means to describe behavior in this way through their laboratory work. Describing entities or processes as complex is increasingly commonplace in the physical and life sciences, but the term signals different things to different practitioners. In physics and computer science, practitioners often employ “complex” in a quite specific sense to describe emergent phenomena arising from the interactions between individual components in a system. Marking something as complex, in these fields, means that the phenomenon is one that must be understood at the system level. As the physicist Philip Warren Anderson (1972) put it in a widely discussed essay, this view of complexity holds that “more is different,” in the sense that the properties of the aggregates of many objects cannot be captured by studying the individual objects in isolation—they must be studied as an interacting system. In the life sciences, the term is often used more loosely to describe systems with many components, which may or may not have emergent properties that can only be understood at the system level. This understanding blurs the boundary some practitioners might draw between the “complex” and the “complicated”—that which must be understood at the system level versus that which in principle could be reduced to the sum of the contributions of individual components.

Researchers at Coast used the term in a variety of ways, only some of which overlapped with the definition that a physicist or computer scientist might provide. When some researchers invoked complexity, they were expressing a commitment similar to Anderson’s—that behaviors could not be studied reductionistically, one gene at a time. But for other researchers, complexity meant only that behaviors were multigenic and that there was no single “gene for” a particular disorder waiting to be discovered. In other situations where researchers invoked complexity, they seemed to be making a claim not about the nature of behavior at all but rather about the difficulties or frustrations they experienced in trying to study it.

In light of the polysemy of Coast researchers’ uses of “complexity,” I argue that their complexity talk is better understood as expressing epistemological commitments rather than ontological ones. As other analysts have argued, scientists’ uses of the term serve functions other than making claims about their views on the nature of the phenomena they are studying. As Arribas-Ayllon, Bartlett, and Featherstone (2010) argue, describing behavioral disorders as “complex” performs rhetorical work for knowledge communities by accounting for past failures in their field. Complexity explains why previous research efforts might have produced inconsistent findings about how genes impact psychiatric disorders, and it constructs careful optimism about the promises of new methodologies and about what the field can hope to accomplish.

For the behavior geneticists at Coast, complexity talk served to cultivate shared stances on knowledge production while allowing for a certain degree of ontological heterogeneity within the community. Despite the ubiquitous use of complex as an adjective, researchers at Coast did not necessarily share a unified vision about the underlying reality of behavior. What they did share, however, was an agreement that working from the assumption that behaviors emerged from the interaction of multiple, small genetic and environmental factors was the best way to produce credible scientific knowledge about them. In describing behaviors as complex, researchers articulated assumptions about what kinds of barriers might lie between them and an understanding of behavior and how a good practitioner should attempt to chart the journey ahead.

This chapter outlines some of the epistemic problems researchers at Coast associated with complexity and how they attempted to deal with them. Researchers attempted to contain the complexity of psychiatric disorders through their experimental practice by breaking down behaviors into smaller units for analysis and creating controlled laboratory environments in which to study them. In practice, however, these measures were difficult to achieve. In the absence of the types of techniques and circumstances that researchers thought would allow them to generate conclusive statements about the genetics of behavior, the knowledge that they produced took on a permanently provisional quality. The experience of being socialized into this epistemic stance resulted in what I describe as a “complexity crisis” for new practitioners, where they began to doubt not their understanding of the nature of behavior but their ability to study it.

Complex Beings, Complex Behaviors

“Step one: Don’t anthropomorphize,” instructed the opening paragraph of a textbook chapter on animal models of psychiatric diseases (Crawley 2000, 179). I was reading the introductory textbook, titled What’s Wrong with My Mouse?, at a bar on campus in preparation for my first visit to the Smith Laboratory. After reading through many chapters outlining methods for assessing the general health, motor functions, and sensory abilities of laboratory mice, I had finally arrived at the section on animal models of psychiatric disorders. I took a sip of my beer and read on. Jacqueline Crawley (2000, 179), the author, continued: “Emotions are personal, internal, and highly species specific. There is no way for a human investigator to know whether a mouse is feeling afraid, anxious, depressed, or experiencing hallucinations. These are subjective emotional experiences, existing in the mind and body of the individual. Major mental illnesses involve neural circuitry that may be uniquely human. . . . Aberrant behaviors symptomatic of human mental illnesses, therefore, may not occur in a recognizable form in rodents.” A leader of the field opening with caveats such as these, I thought, did not make for an especially auspicious introduction to the practice of animal modeling. When the second edition of the textbook was published in 2007, a few months before I started my longest stretch of fieldwork in the Smith Laboratory, Crawley’s (2007, 227) opening message had been further intensified—the phrase “don’t anthropomorphize” was now set in italics and punctuated with an exclamation mark.

Crawley’s cautionary introduction highlights several of the difficulties animal behavior geneticists associated with using complex organisms to model complex human disorders. Researchers described the disorders that they dealt with as complex both because of the types of characteristics that made up clinical categories such as “alcoholism” or “anxiety” and because of how those characteristics were grouped together. Many core features of psychiatric disorders—such as persistent worrying or the cravings for alcohol or drugs that addicts report—are what researchers described as internal mental states. Finding ways to measure those traits in humans, who can discuss their experiences with researchers, was difficult enough. But figuring out how to identify and measure analogous traits in animals presented a whole new set of challenges. As Dr. Smith pointed out to me in one of our first meetings, the criteria outlined in the Diagnostic and Statistical Manual of Mental Disorders (DSM) for diagnosing alcohol dependence range from those that could be construed as external biological traits observable across species, such as evidence of withdrawal symptoms, to those that are much more subjective and specific to humans, such as giving up important social and occupational activities due to alcohol use. He told me that he thought that the majority of the criteria in the DSM-IV were things that were “pretty tough” for animal models to capture: “Five of those seven symptoms are things like losing your job or persistently continuing to drink even in the face of evidence that your health is falling apart. You know, the doctor telling you you’re killing yourself, your liver’s getting trashed and you keep drinking . . . your relationships fall apart, you wind up in jail, you’re obsessed with getting the drug. Those are all very human symptoms that are at the core of the disorder.” Even if researchers could devise ways to model the subjective emotional experiences central to psychiatric disorders, these lines of research would run up against long-standing debates about whether humans can make credible statements about the mental states of other species. The question of whether the animal mind is an appropriate topic for scientific investigation has been quite divisive in the history of the behavioral sciences, with entire fields built around the central premise that only animal behavior, and not the animal mind, is accessible to scientists (Crist 1999, Daston and Mitman 2005).

The “checklist” approach to diagnosis employed by the DSM and other measurement tools used in humans also drew criticism from animal researchers for how it lumped together a variety of different traits. A DSM-IV diagnosis of generalized anxiety disorder, for example, required that a patient show excessive anxiety for at least six months along with any three of a list of six symptoms, such as irritability, muscle tension, and sleep disturbance. The entry for “alcohol dependence” similarly had a list of seven symptoms, any three of which can form the basis for a diagnosis. The separate entry for “alcohol abuse” required that a person exhibit one or more symptoms from a list of eleven. What this meant, Dr. Smith joked to me, is that when the different possible combinations of symptoms were parsed out, there are “1,200 some odd ways you can be an alcoholic.”1

In this sense, anxiety and alcoholism were complexes—amalgams of different clinical features, which may or may not share the same biological causes. To some extent, researchers treated this heterogeneity as an unavoidable problem of studying high-level behaviors that developed over time in response to multiple genetic and environmental inputs. They argued, however, that the imprecision of the diagnostic tools used to classify humans amplified this problem. While in the late 1990s and early 2000s researchers had speculated that genetic techniques might be able to carve out more specifically defined disorders from these broad diagnostic categories, a decade later this prospect seemed increasingly unlikely. Some practitioners advocated instead that researchers would be better off trying to define entirely new disease categories for study, ones that were more narrowly defined and more clearly based on biological characteristics. How basic scientists such as the animal behavior geneticists at Coast were to form research agendas around disease categories that were themselves contested and unstable was an open question.2

Using higher organisms (such as mice and rats) as models for human behavior introduced another set of complicating factors related to the individuality of the animals. Coast researchers seemed to have little difficulty talking about a colony of bacteria or a bottle of flies as analogous to other supplies of uniform chemical reagents in the laboratory, but they resisted talking about even the highly standardized laboratory mouse in the same way. They described other laboratory organisms—bacteria, viruses, worms, flies—as being productive because of their simplicity and uniformity, in line with historian Robert Kohler’s (1994) description of the fly as a kind of laboratory technology that found great success because it could be mass-produced. But while researchers did talk about mice as laboratory tools, to some extent, they also described them as individuals. Researchers argued to me that their own scientific training and the complexity of the animals themselves prevented them from viewing mice as completely interchangeable with each other. As Dr. Scott Clark, a longtime collaborator of Dr. Smith and Dr. Martin’s, explained, “Brains change, right? It’s not the same thing when I test this mouse and this [other] mouse. This mouse is different! Even if they’re genetically identical, they’re different! This mouse may have come from a cage where it grew up with two siblings, and this one with four siblings . . . that[’s the] nuanced level of complexity that psychologists would naturally be interested in.” This was even truer of the rhesus monkeys used at the affiliated primate research center that many of the Coast students rotated through in their early years of graduate school. Even long after individual monkeys had been turned into data points, researchers recalled their quirks, proclivities, and personal histories. When Dr. Sherry Trudeau, the charismatic head of the neuroscience division at the Primate Center, talked about her research, she would often circle particular data points with her laser pointer and tell the audience the name of that monkey and a bit about its personality.

Researchers talked about the unique properties of higher laboratory animals as both a resource for and a barrier to knowledge production. On the one hand, researchers at Coast argued that the “complexity” of the “natural behavioral repertoires” of mice was useful for modeling psychiatric disorders, and they saw the variation in behaviors that mice displayed as truer to the human scenario. Dr. Clark, for example, told me at length about the burrow systems that some rodents will form, how they vary with social dynamics and population size, and how mice emerge from and retreat into these burrows in response to potential threats. He thought that these burrowing behaviors could be a useful model for human anxiety, which also varied in response to social dynamics and perceived threats in the environment.

But drawing on these variable natural behavioral repertoires also complicated the experimental scenario. Dr. Marcus Lam, a postdoc who had recently joined the Smith Laboratory after finishing his doctoral degree in a pharmacology program, told me that he was amazed to find that many of the Smith lab’s common experimental protocols were designed in such a way that the mice in a particular study might not all get the same dose of alcohol. While allowing mice to drink from a bottle of alcohol might be a better reflection of the way that humans ingest alcohol, Dr. Lam said that he would rather just inject the mice directly with a fixed dose to avoid introducing “noise” into his data right from the very beginning. The same individual variation that made working with mice seem truer to the human scenario could also be viewed as grounds for questioning the quality of the scientific data.3

Setting one unstable entity—a lively and unpredictable mouse—in relation to another unstable entity—a loosely bundled amalgam of human traits—therefore created numerous possibilities for uncertainties, reversals, and interpretational difficulties in animal behavior genetics research. Even though both human psychiatric disorders and laboratory mice were entities that had been the target of much effort to standardize, characterize, and contain them, they were also entities that continued to overflow the boundaries of these categories.

Dr. Smith recounted a story to me that illustrates why researchers believed these complexities made the production of genetic knowledge about behavior extremely difficult. One of the founders of the field, Gerald McClearn, designed a project early in his career to look at the genetics of alcohol withdrawal in mice. He took a group of genetically diverse mice, measured them on twenty-one different behavioral tests related to alcohol withdrawal, and then mated mice with high scores on this test panel to each other. Researchers had successfully used this selective breeding approach to study the genetics of other behaviors, such as Robert Tryon’s now well-known breeding experiments to generate so-called “maze-dull” and “maze-bright” rats in the 1940s. In this case, however, it seemed not to work. After several generations of mating the high scorers and the low scorers to each other, the difference between the two groups was slight, and McClearn eventually abandoned the project. The failure of this project is all the more notable because some of the tests he was using measure traits that are now believed to be highly heritable, such as the tendency for mice to experience seizures while they are withdrawing from alcohol. Dr. Smith speculated that the project had foundered because McClearn’s attempt to account for multiple facets of alcohol withdrawal in his panel of tests may have actually hindered his progress. Some of the mice in his study may have received high scores for withdrawal because they experienced seizures, while others might have scored high due to increased anxiety. Consequently, mating mice that were experiencing severe withdrawal for different reasons simply reshuffled the genetic deck in each generation rather than concentrating genetically similar individuals.

While one might conclude from this quick overview that using mice to model human behavioral disorders is simply more trouble than it is worth, animal researchers saw human research as equally, if not more, problematic. Watching animal researchers evaluate human studies brought this into sharp relief. In one journal club meeting I attended, a postdoc in the department explained that lately, he had come to the realization that he knew relatively little about alcoholism in humans compared to what he knew about animal alcohol models. And so, he had selected a paper for discussion on the relationship between anxiety disorders and alcohol use disorders in the Dutch population. The audience of animal researchers seemed unimpressed by the study. They quickly pointed out numerous “flaws” with the research design, from inconsistencies in data sets brought together to conduct the study to a reliance on suspicious self-reported data about ...

Table of contents

• Cover
• Title Page
• Copyright Page
• Contents
• Introduction: A Furry, One-Ounce Human?
• 1  •  Containing Complexities in the Animal Behavior Genetics Laboratory
• 2  •  Animal Behavior Genetics, the Past and the Future
• 3  •  Building Epistemic Scaffolds for Modeling Work
• 4  •  Epistemic By-Products: Learning about Environments while Studying Genetics
• 5  •  Understanding Binge Drinking
• 6  •  Leaving the Laboratory
• Conclusion: An Expanded Vocabulary for the Laboratory
• Acknowledgments
• Appendix
• Notes
• Bibliography
• Index