Introduction
The vast proliferation of biomedical research since the mid-twentieth century has been paralleled and potentiated by the burgeoning of animal models for infectious diseases, genetic diseases, syndromes, developmental processes, and aging, and, indeed, for virtually all biological processes and conditions of scientific or medical interest. In addition, thanks to rapidly proliferating technologies in creating transgenic animals, new models appear with logarithmically increasing frequency.
When thinking about the value and acceptability of animal models, one is obliged to consider two major parameters:
1) The validity of the model in question, that is, whether it reliably and predictably represents and measures that which it is intended to model (frequently called “construct validity” in the social sciences)
2) The ethical acceptability of the model
These two parameters are clearly inherently linked. On the presumption that any animal model of human illness incurs at least some harm or distress (it is an illness model, after all), that harm can only begin to be justified in the service of increasing understanding of the condition it models. If the model fails tests of validity, then the information it yields is, of course, invalid, divesting it of utility. In turn, harm to the animals involved is, in essence, “for nothing.” There are few approaches to ethics that can justify inflicting harm without yielding some benefit, by some measure. How to measure said benefits remains a central problem in the study of ethics, but information that cannot be interpreted is not useful information. As such, if an animal model fails the test of validity, it also fails the test of ethical acceptability. In this discussion, we take construct validity as a necessary condition for the ethical acceptability of animal models.
This is not to say that science has traditionally obeyed commonsense ethics; throughout much of the twentieth century, scientists embraced the concept that science is “value free,” dictated only by the precepts of the scientific method, and accordingly, is also ethics free. Fortunately, the public at large has never surrendered to this position, hence the increasingly restrictive federal constraints on the use of both humans and animals in biomedical research. Indeed, whatever benefits may have emerged from the Tuskegee syphilis experiments on African American prisonersFootnote 1 or from the infamous University of Pennsylvania primate head injury research of the 1980s,Footnote 2 such benefits failed to eclipse public revulsion and rejection of the experiments themselves.
A product of the burgeoning of biomedical science has been the relatively recent advent of scientific study of psychiatric disease. Whereas for the first half of the twentieth century this area was largely left in the domain of psychoanalysts and their subjective, introspective investigations, as a result of the advent of behaviorism in psychology as well as the discovery of compounds that appeared to have therapeutic effects for the severely mentally ill, the concept of biological psychiatry was born, and psychiatric illness rejoined other pathophysiologic phenomena within the purview of biomedical science. In turn, just as in other areas of biomedicine, development of animal models of psychiatric derangements inevitably became a focus within the scientific establishment.
However, by virtue of science’s recognition of mental illness as a biological phenomenon, experimental approaches to it—with animals and often with humans—have been similarly modeled after those found in other domains of biomedical research. Unfortunately, pathologies of the mind have characteristics that set them apart from other forms of illness, in ways that have deep relevance to both the validity and the ethical acceptability of these kinds of studies, particularly with respect to animal models.
In the sections to follow, we explore the ethics of animal models of human mental illness, by challenging the validity of different categories of these animal models, as well as the intrinsic ethical acceptability of such models independent of their validity. A thorough review of the myriad animal models that have been proposed and used in these conditions is far beyond the scope of this work.Footnote 3,4 We do, however, cite illustrative examples of attempts at animal models for certain psychiatric conditions.
The Construct Validity of Psychiatric Animal Models
In this section, we discuss overarching issues that are inherent obstructions to the validity of animal models of psychiatric phenomena. Although there are a number of definitions of construct validity, with varying degrees of technicality,Footnote 5 for present purposes we simplify the concept to mean “the model represents what it is intended to represent,” namely, a direct analogue of a pathologic phenomenon known to exist in humans.
Subjectivity
Disturbances of the mind are set apart from other physiologic perturbations because they are, in contrast to other illnesses, specifically manifest in the mind. For our purposes, we use the term “subjective” to refer to phenomena that occur as an experience within the mind of the creature having them (rather than as a description of a type of measurement). For many in the profession, a maddening (if you will) aspect of the evaluation, treatment, and scientific study of mental illness is this very fact: whereas in most of medicine, signs, straightforward symptoms, and objective findings are the norm, in mental illness, symptoms are generally internal and subjectively experienced exclusively by the sufferer. The clinician or researcher relies to a great degree—and frequently, almost exclusively—on patients’ report of their own internal experience to make diagnostic, prognostic, and therapeutic judgments. Some relevant symptoms are identified through behavior alone; however, the Diagnostic and Statistical Manual of Mental Disorders IV, Text Revision (DSM-IV-TR), the consensus document of psychiatric nosology, includes very few entries for illnesses that are defined in behavioral terms alone.Footnote 6 For example, of the nine diagnostic criteria for a major depressive episode as defined in DSM-IV-TR, only three (weight changes, sleep pattern changes, and psychomotor changes) can be assessed by observation; the rest can only be assessed by an individual’s self-report. The Beck Depression Inventory–II (BDI-II), one of the most widely used and accepted psychometric instruments in all of psychiatry, is a self-report questionnaire of 21 items scored 0–3 for severity. Of these 21, only 3 are defined in even quasi-behavioral terms (crying, agitation, and sleeping pattern changes).Footnote 7
There are myriad crucial psychiatric symptoms that are inherently totally subjective in quality: examples include dysphoria, euphoria, grandiosity, self-esteem, hallucinations, delusions, racing thoughts or cognitive slowing, guilt, rumination, suicidality, obsessions, nightmares, flashbacks, libido changes, cravings, dissociation, and nonspecific anxiety. Although some of these may have behavioral correlates, there is by no means a 1:1 correspondence between these symptoms and associated behaviors, particularly in the absence of a shared cultural or linguistic context.
It is not difficult to see, then, where the first problem arises with animal models. Descriptions of psychiatric phenomena are contingent on having mechanisms to ascertain the subjective experience of the individual experiencing them, which is something that simply cannot be done with animals. Even in dialogue with the most articulate patients, clinicians can find it challenging to unpack the nuances of people’s descriptions of their experience in a meaningful fashion. With animals, it is patently impossible, because animals have no shared vocabulary with humans. In an exquisite piece of irony, this rift between humans’ and animals’ capacities to describe their own experience is so fundamental that the very same scientific establishment that would—in spite of the rift—seek to use animals to model human psychological distress has also traditionally used the very same rift as a means to justify experimentation by denying animals’ capacity to experience distress comparable to that of a human!
Context
The importance of subjective experience does not end with our simple inability to ask animals to report on their symptoms in the same way we do humans. And yet, in a sense, it does, insofar as the lack of a shared vocabulary represents the much deeper issue—that the qualities that comprise experience are fundamentally different for different kinds of creatures. In other words, the context of a creature’s experience, as defined by its physiology, its life cycle, its social structure, its evolutionary features, and so on—what Wittgenstein calls its “form of life”—is part and parcel of the makeup of that creature’s subjective experience. To some this may seem so obvious as to be absurd to even mention—that a prey animal’s world is different from that of a predator, that being a primate feels different from being a rodent, or that the shape and quality of perception is different in animals with and without sensitive olfaction. However, in order to justify a valid animal model construct, we would need to be able to assert that a given phenomenon feels exactly the same to a human and to an animal.
To illustrate this point, we actually need not look as far as other species: we have salient examples within our own species. A persistent nuisance for the effort toward developing a comprehensive disease model of mental illness has been the existence of other cultures. Although the DSM has attempted to represent universal patterns of mental phenomena, even that very document acknowledges that different cultures manifest different presentations and experiences of specific “diseases,” as well as syndromes that are “culture bound,” meaning they occur only within a given culture. DSM-IV-TR lists 25 culture-bound syndromes and admits that even for the “universal” diagnoses, there may be substantial variability in the experience of the illness based on differences in cultural context.Footnote 8 Despite these concessions, DSM-IV has been subject to criticism for its myopia and ethnocentrism.Footnote 9
A poignant example of the cross-cultural variation in the subjective experience of distress arose for one author (MR) while working with a refugee patient from a minority subgroup from East Africa. While interviewing the patient via an interpreter and dutifully screening for symptoms such as depressed mood, anhedonia, fatigue, irritability, anxiety, flashbacks, nightmares, and so on, after about ten minutes the interpreter interrupted to mention that in their language, there exists only a single word that can be used to represent any of these symptoms!
Perspective and Self-Reference
A crucially important phenomenon within the subjective context of the human experience of psychological distress is what we might call “perspective.” By this, we mean the following:
1) The capacity to place an experience in the context of a personal narrative, one’s own “story.” A person prone to depression suffers a professional setback, which corresponds to his schema that “the world always treats me unfairly.”
2) Judgment about the experience’s congruence to one’s image of oneself (ego-syntonicity or ego-dystonicity). A musician struggles to find motivation to end her drug use, despite her marriage ending as a result of it, in part because she views it as consistent with the image of a “rock star.”
3) The capacity to glean from the past and project (for better or worse) into the future. A bereaved person falls into a major depressive episode as he extrapolates his current feeling of loss into an anticipated lifetime of loneliness; a person in treatment for phobia of heights, in order to tolerate mounting an escalator, draws on a memory of successfully climbing down a ladder.
4) The possibility of a particular symptom or behavior to affect its own natural history. An alcohol abuser, distressed by the extent of his own alcohol abuse, tries to blunt his experience of distress by consuming more alcohol; an individual with panic disorder suffers such intense anticipatory anxiety about when her next panic attack might occur that this in itself brings about another panic attack.
These issues are essential to the phenomenology and prognosis of mental disorders. They demand both a longitudinal perspective on time and experience and a capacity for abstraction and self-reference. Although these are not provably uniquely human traits, our understanding of animal cognition does not, as yet, allow for constructs like self-image, ego defenses, or anticipation of the distant future. At the very least, such capacities are impossible to ascertain without language, and a nuanced language at that.
The Validity of Models of Psychiatric Disease Entities
Leaving aside the general criticisms of all psychiatric animal model constructs detailed in the previous section, we move now to look at animal models of disease entities, such as those described in DSM-IV-TR. Because the clinical practice of psychiatry (rightly or wrongly) deals primarily with the diagnosis and treatment of particular illnesses, it stands to reason that science would attempt to model these disease entities in animals. However, it may be argued that many of these diagnostic constructs are sufficiently flawed that they lack validity in humans, much less in animals.
Symptom Sharing
A common criticism of DSM-IV-TR is the lack of specificity of symptoms. It is difficult to accept a particular symptom as informative of an underlying illness if that symptom can represent several different illnesses. A trenchant example is irritability (or the conceptual equivalent), which appears as a diagnostic feature in no less than 10 disparate diagnoses, including bipolar mania, major depression (in adolescents), generalized anxiety disorder, posttraumatic stress disorder, borderline personality disorder, antisocial personality disorder, nicotine withdrawal, pathological gambling, schizoaffective disorder, and oppositional-defiant disorder.Footnote 10 Similarly nonspecific symptoms include depressed mood, mood lability, concentration disturbance, impulsivity, and suicidality, each of which appears as a diagnostic or associated feature of at least eight different disorders.Footnote 11 Furthermore, not one of these symptoms represents a necessary criterion for any disorder. As such, our diagnostic nosology encodes the tautological claim that many of our disease entities “may or may not” exhibit these symptoms.
Generating animal models that display these symptoms (or behavioral indications thereof), then, would reflect as much scientific utility as the symptoms do diagnostic utility; they “may or may not” be valid models.
Comorbidity
Whereas numerous symptoms occur across diagnostic categories, so too are multiple diagnostic categories likely to co-occur; the World Health Organization’s survey found that 45% of people who met diagnostic criteria for a mental illness actually met criteria for two or more,Footnote 12 in essence making comorbidity “the rule rather than the exception.”Footnote 13 This suggests that the disease entities as currently defined reflect artificial rather than natural clusters of symptoms. This fact not only renders the disease constructs’ validity suspect but actually undermines the effort of creating an animal model: a truly valid animal model would theoretically manifest the signs and symptoms of a disorder because the manipulation that defines the model would reveal a natural cluster of features that share a common antecedent. If the desired set of features does not reflect a naturally occurring cluster, then it stands to reason that creating a model of that desired set would require more and disparate manipulations in order to generate it, thus increasing the distance of the conceptual leap from disease to model.
Mutually Exclusive Symptoms
A last challenge to the validity of current disease constructs comes in response to the inclusion of mutually exclusive symptoms within a set of diagnostic criteria. Nestler and Hyman provide an excellent example of two very disparate presentations occurring within the same diagnosis (in this case, major depressive episode): “The first individual might have depressed mood, weight loss, insomnia, psychomotor agitation and suicidal thoughts, whereas the second individual might have markedly diminished pleasure, weight gain, hypersomnia, psychomotor retardation and fatigue. There are no symptoms in common!”Footnote 14 It is difficult to assert that one is working with a discrete disease entity when two individuals sharing zero symptoms can be said to suffer from the same illness.
The Validity of Symptom Models
It may be argued that impeaching the credibility of models of full-blown psychiatric diseases is an attack on a straw man. To wit, George Koob, well known for his innovations in animal models of addiction, in the introduction to his recent comprehensive, apologist review of animal models of psychiatric disease, swiftly dismisses such models: “Animal models for a complete syndrome of a psychiatric disorder are unlikely to be possible either conceptually or practically.” This claim he appears to take as self-evident. Instead, he goes on to claim:
Thus, an approach to the development of animal models that has gained wide acceptance is that animal models are most likely to have construct or predictive validity when the model mimics only the specific signs or symptoms associated with the psychopathological condition. . . . Under such a framework of producing animal models only for the specific signs or symptoms associated with the psychopathological condition, specific measures have been identified in each disorder that provide a focus for animal study. The reliance of animal models on a given component of the disorder also eliminates a fundamental problem associated with animal models of psychopathology, namely the frustration of attempting to provide complete validation of the whole syndrome. This framework also leads to a more pragmatic approach to the study of the neurobiological mechanisms of the behavior in question.Footnote 15
In other words, Koob dismisses the notion of modeling whole diseases, instead preferring to model specific components or symptoms of diseases, one at a time. If this is possible, it would indeed save us from the critiques detailed in the preceding section, namely that the shaky construct validity of the diseases as described in humans renders them impossible to translate to animal models.
As such, the project of scientists wishing to study these phenomena in animal models becomes to identify and model atomic symptoms of psychiatric diseases that can be studied one at a time in animal models.
Defining Atomic Symptoms
But what, then, comprises an atomic psychiatric symptom? Although the list of diagnostic criteria for a specific disease can be disassembled into individual criteria, with each representing a specific phenomenon (e.g., anhedonia or sleep disturbance from depression, impulsivity or hypersexuality from bipolar mania, auditory hallucinations or blunted affect from schizophrenia), how do we know if an individual criterion, taken alone, represents a simple, valid construct worth modeling and studying?
Simply because a disease is described as a compilation of such phenomena does not entail that the phenomena described are themselves fundamental or irreducible. A glaring example can be seen in the DSM-IV-TR diagnostic criteria for schizophrenia, in one of five descriptive features of the disorder:
Grossly disorganized behavior (Criterion A4) may manifest itself in a variety of ways, ranging from childlike silliness to unpredictable agitation. Problems may be noted in any form of goal-directed behavior, leading to difficulties in performing activities of daily living such as preparing a meal or maintaining hygiene. The person may appear markedly disheveled, may dress in an unusual manner (e.g., wearing multiple overcoats, scarves, and gloves on a hot day), or may display clearly inappropriate sexual behavior (e.g., public masturbation) or unpredictable and untriggered agitation (e.g., shouting or swearing). Care should be taken not to apply this criterion too broadly. For example, a few instances of restless, angry, or agitated behavior should not be considered to be evidence of Schizophrenia, especially if the motivation is understandable.Footnote 16
This single criterion in itself subsumes a universe of potential behaviors and presentations, to the extent that it is not hard to imagine an entry in an alternative nosology entitled “grossly disorganized behavior syndrome” in which this “symptom” is reframed as a “disease,” using the identical description. The vague and inclusive definition of this symptom construct renders it susceptible to the same critiques as the disease construct it partially comprises.
Identifying Disease Symptoms That May Be Validly Modeled
Admittedly, it could be argued that the schizophrenia example is a particularly difficult one, and that there are far more rudimentary symptoms of other disorders. But by induction, the very same criticisms we have leveled against the validity of modeling diseases can be applied to any disease component or “symptom,” unless that component meets the following criteria (as generated by our critique of disease models):
1) It is not defined as a subjective experience that cannot be measured without subject self-report.
2) It does not itself represent a set of criteria that contain disparate or mutually exclusive phenomena.
3) The symptom is specific, such that either of the following is true:
a. It is defined in a way that is clearly distinct from symptoms listed in other disorders that are presumed to have different antecedents (symptom sharing).
b. If the symptom is shared among different disorders, then the disease antecedents are clearly identified and controlled for.
This is a short list of requirements, but they are difficult to satisfy. Requirements 1 and 2 were discussed earlier. The same critiques apply here; they require that the phenomenon be measurable by observation (i.e., physiologic or behavioral measurements), and that it not concatenate a disjunctive set of conditions (such that, for example, increased sleep and decreased sleep are treated as the same phenomenon).
Requirement 3 is more difficult. Nestler and HymanFootnote 17 provide an excellent example of a failure in 3b: a transgenic mouse model for certain schizophrenia symptoms, in which a brain receptor (NMDA receptor) was partially ablated after birth, consistent with one particular theory of the pathogenesis of schizophrenia. The symptoms under observation were disorganized behavior and social deficits, both core symptoms of schizophrenia. The mice demonstrated disorganization of certain behaviors (e.g., nest building) and socially maladaptive symptoms (e.g., poor mating behavior), which were asserted to represent disorganization and social deficits as seen in schizophrenia.Footnote 18 In this case, criterion 1 is met because the symptoms are behaviors. Criterion 2 is questionable, because both disorganized behavior and social deficits are very vaguely defined constructs, though the observed behaviors are specific. But criterion 3 is clearly failed, because these behaviors are not unique to this genetic manipulation (e.g., there are similar nest-building deficits in an Alzheimer’s modelFootnote 19), and this physiologic derangement (NMDA ablation) has not been demonstrated to be a core feature of schizophrenia, so the disease antecedent is not clear. In other words, as Nestler and Hyman point out, this model at best characterizes the effect of this particular genetic modification in this strain of mice, but without previous evidence that NMDA receptor dysfunction underlies the pathogenesis of schizophrenia, or that nest building and mating behavior directly represent mouse schizophrenia, it would be circular logic to suggest that the observed nonspecific dysfunction in mice substantiates this defect as a cause of schizophrenia.
A simpler example is as follows: the Flinders Sensitive Line (FSL) is a selectively bred rat strain that manifests decreased locomotion and decreased eating behavior compared to wild-type rats; these are taken as analogues of psychomotor retardation and decreased appetite, common symptoms of major depressive disorder. Imagine a study comparing the FSL rat to rats whose legs have been amputated. In one case, a subtle set of mutations are added to the genome. In the other case, the animal’s legs are missing. In both cases, locomotion and eating behaviors are decreased. In the first case, the causal link between the manipulation (mutation) and outcome (behavior) is not visible and is therefore treated as a black box; this is taken as evidence for this genetic locus as a culprit in major depression. In the second case, the rat has no legs, which in the logic of this approach, appears to be evidence that leglessness is a root cause of depression.
Face Validity and the Black Box
A stalwart would argue that the difference between the FSL strain and the legless rat is one of “face validity,” that is, the commonsense impression of whether an experimental scheme tests what it was intended to test. And admittedly, the legless mouse model of depression would have less face validity than the mutant strain. But just how much less depends on whose face is assigning validity. Simply because a manipulation is construed as a black box does not mean that the alterations occurring in that black box are particularly subtle, or relevant to the disease in question.
For example, olfactory bulbectomy is a procedure that removes the olfactory bulb, a large rostral brain structure responsible for the rodent’s powerful sense of smell. Rats and mice that undergo this procedure are divested of their sense of smell, in addition to losing a large proportion of their brain volume and a significant amount of nonolfactory functional connectivity as well as retrograde degeneration of the neurons that are connected to it.Footnote 20,21 These animals, perhaps not surprisingly after a brain surgery removing a large structure central to their survival, show a number of behavioral changes, including hyperactivity, reduced sexual activity, and alterations in responses to food rewards. Some of these changes are attenuated by chronic administration of antidepressants of various types,Footnote 22 which is what has led to faith in this model for depression symptoms. And although it is remarkable that a class of drugs can overcome some outcomes of an intervention this invasive, little argument is necessary to question the face validity of the brainless mouse as similar to that of the legless rat. It is of little surprise, too, that administration of a hormone normally released when the rodent smells food (ghrelin) has a similar “antidepressant” effect.Footnote 23
The simple choice of name can profoundly affect the face validity of a model. One sinister example worthy of mention is the “chronic mild stress” model for depression, particularly common in contemporary study.Footnote 24 It involves precipitating a variety of behavioral and physiologic changes by chronic application of food and water deprivation, continuous lighting, stroboscopic lighting, soiled cages, unpredictable tilting of the cage, and other “mild” stressors. In this case, their “mildness” is evidently defined as stressors that “don’t leave a mark,” which is similar to what creative child or domestic abusers do to avoid prosecution, and the name evokes the creative rebranding of torture as “enhanced interrogation.” The difference between this kind of manipulation and the aforementioned absurd example of the legless mouse is one of degree, not of type.
The Leap from Behavior to Experience
There are certain psychiatric symptoms that are described largely in behavioral terms but that at their core hinge completely on subjective experience. Obsessive-compulsive disorder (OCD) is a good example. With respect to the compulsions aspect of the illness, DSM-IV-TR is clear:
Compulsions as defined by (1) and (2):
1) Repetitive behaviors (e.g., hand washing, ordering, checking) or mental acts (e.g., praying, counting, repeating words silently) that the person feels driven to perform in response to an obsession, or according to rules that must be applied rigidly.
2) The behaviors or mental acts are aimed at preventing or reducing distress or preventing some dreaded event or situation; however, these behaviors or mental acts either are not connected in a realistic way with what they are designed to neutralize or prevent or are clearly excessive.Footnote 25
Because compulsions have such a vivid behavioral appearance, they are seductive as an object for animal models. Several putative models have been developed and studied extensively; each one yields animals that exhibit either decreased variety of behavior or frankly repetitive or stereotypic behavior.26 They are asserted to have predictive validity because they are (to varying degrees) diminished by treatment with SSRI antidepressants. However, there are numerous causes for repetitive behaviors in humans (including autism, tics, dementia, and others)Footnote 27 and animals (e.g., confinement).Footnote 28
Is a Valid Model Logically Possible?
We have argued that a symptom model is valid if and only if it represents a measurable outcome that is demonstrably specific to the phenomenon being modeled. In other words, you have to know how what you’re measuring comes from what you’re doing to the animal, and that what you’re doing to the animal represents the thing you’re trying to learn about. But if these criteria were in fact met for any given model, then the pathophysiologic mechanisms at work would need to be so well described that there would be no need for an animal model!
But, one might ask, what of the farm animal that paces repetitively or the listless zoo animal? Are they not anxious or depressed? There seems to be little question that the distress that confined animals exhibit must have something to do with human anxiety and depression—but what, exactly? It is a problematic leap to go from “that lion looks depressed to me” to a research question that extrapolates to human depression. As one author (MR) tells his patients frequently, “misery” and “depression” are not the same thing. There are countless ways to be miserable, but only some of these have psychiatric names. Common sense lets us be pretty certain that many confined animals are miserable, but anthropomorphically extrapolating that misery to a more specific subjective state is not straightforward.
Ethical Considerations Independent of Validity
Dismissing all concerns about validity from the preceding sections, there remain profound ethical implications of animal models that are specific to the psychiatric domain. A valid animal model of a psychiatric phenomenon would entail that the phenomenology of the animal’s experience of it would be directly comparable to that of a human’s experience. The better the model, the truer this would be. This would require that the animal share mechanisms for distress, despair, perspective, narrative, and self-reflection that are comparable to those of a human.
But if that were in fact the case, what would be the morally relevant difference between the animal and the human? In other kinds of biomedical animal models, apologists can appeal to (albeit specious) differences in cognition, experience, temporal perspective, language, phylogenetic class, neurophysiology, evolutionary development, and so on, as differences between humans and animals sufficient to justify doing to them what would never be considered acceptable to do to humans. But psychiatric illness is a special case. By employing an animal model of a phenomenon defined as a negative perturbation of experience, one is implicitly asserting that the animal has the capacity for experience sufficiently comparable to that of a human as to be meaningful in understanding the human. This is what one of the authors (BR) has previously called the “psychologist’s dilemma”: one cannot simultaneously defend an animal model as both adequate and ethical.29
Furthermore, because we cannot truly assess the experience of an animal in these conditions, we have no idea, nor do we have the capacity to know, just how much suffering we’re creating. So, in fact, not only might we be causing the same degree of distress in an animal that we would never accept doing in a human, we might actually be doing worse.
As such, only when psychiatric research questions are formulated in truly valid, objective terms, and when experimental conditions reliably eliminate suffering in keeping with the assumptions of the research question, can the study of animal models of psychiatric disease be considered morally acceptable.
