Assessing the validity of animal models of depression and schizophrenia

Author: Syra Bhoola
Father Lopez Catholic High School    
November 2, 2020


Looking at contemporary animal models of neuropsychiatric disorders, brings up many questions. Can we trust these models? Are these models valid? These are questions that are necessary to  address in order to move from animal models to human testing. Specifically, three types of validity have been proposed by scientists: construct validity, face validity, and predictive validity. Many researchers lean on one of these three types of validity to call their work “valid” but what if the process of validity is not as concrete as it first appears? In order to answer this question, this work overviews, examines, and critiques the three proposed types of validity. Additionally, it examines two animal models of depression, which each at first glance appears to fit neatly into a particular validity. Yet, looking at these models in more depth, it is apparent that they both contain multiple validities. An example is a model that is considered construct valid but actually this validity  relies on implicit assumptions. Changes in background assumptions can then change the degree and type of validity. These uncertainties also apply to more than just depressive models. In studies done on schizophrenia it is unclear whether or not the behavior changes in mice are face or construct valid. The conclusion is that the types of validity set forth are difficult to apply and seem to indicate a spectrum rather than separate categories, and can change depending on background assumptions.


Our growth in knowledge of neuropsychiatric disorders is hindered through restrictions put on human models. Schizophrenia, depression, bipolar disorder, and autism are members of these disorders that begin in the early stages of life and are universal diseases around the world (1). Analyzing and understanding these disorders and their pathophysiology has been challenging. Moreover, there is impeded scientific progress due to one prominent factor. Progress has been hindered by the overwhelming complexity of advanced brain functions, and the moral and empirical difficulties inherent to observing and experimenting on the living human brain. While there has been swift progress in the last two decades in the development of technology to evaluate function and structure, there continues to be limitations to fully grasp the details of physiology and molecular biology of the brain. The cause of these limitations is that novel therapeutics cannot be initially tested on humans but must first be proven in animal models before proceeding to clinical trials. The consequence of this is that all current psychotherapeutic drugs are based on discoveries prior to the 1960s when restrictions on clinical testing were less than the modern day’s restrictions (2).

Since there are such restrictions, it is impossible to visualize the concept of progression in pathophysiology or therapeutics without successful animal models of neuropsychiatric disorders. Current animal models must improve if they are to fulfill the promise of scientific research into these disorders. This is because current animal models generally have failed to provide efficacious treatment of these diseases (3). A significant reason for these failures is because it is often unknown whether an animal model of a disease is actually an appropriate or successful model. Therefore, explicit and formal methods are necessary for judging whether or not a given animal model of a neuropsychiatric disorder is a successful or valid model. The most important characteristic is validity and bringing acceptance onto the subjects. The goal of this paper is to outline proposed measures and types of validity for animal models and whether current animal research into both depression and schizophrenia meets the standard for validity. 

Categorizing different types of validity in animal models has been put forward previously (see further details below) (4). The authors argue that validity can consist of three types: construct validity, face validity, and predictive validity. All three validities are necessary to understand animal models and disorders. While development of the animal models will take time, effort, and efficiency to completely analyze neuropsychiatric disorders, it is necessary to judge current animal models along the known standards of validity: 

Construct validity can be achieved when the mechanistic cause in the animal model is the same as in a human. Let’s say there is a muscle disorder that is caused by a certain gene in humans. To model this disorder on a mouse, a scientist would have to inject a mouse with that specific gene and see what happens. This would then be a case of construct validity, as the common cause of the disorder in both the animal model and in humans is the same. Common causes may be genetic or environmental. An example of a case of construct validity would be an animal model of Creutzfeldt-Jakob disease, which is caused by prion misfolding in both animals and humans (5). As straightforward as this approach sounds, in animal models of neuropsychiatric disorders, there is great difficulty in achieving construct validity. If the biological cause of a neuropsychiatric disorder is unknown, there is no way to implement it via a common cause in an animal model. Therefore scientists often have to lean towards other means of validity. 

Face validity arguably is the weakest form of validity in animal models. given the difference between animals and humans. Face validity occurs when a model reflects some important aspects of anatomical, biochemical, neuropathological, or behavioral characteristics of a human disease (4). An example of this might be when an animal model leads to characteristic behaviors that are similar to the diagnostic criteria for a neuropsychiatric disorder, such as depressive or anxious behavior. This type of validity is quite common because there are few neurobiological malformations known to be explicit biomarkers that are shared across both humans and animals with respect to mental disorders. Therefore this type of validity often rests on behavior. Yet it is not probable that an animal model of a neuropsychiatric disorder will repeat all of the behavioral characteristics that are observed or even a single signature feature that models the exact human circumstance. Therefore the conclusions of face validity will continuously be questioned, it is the responsibility of the experimenter to make explicit arguments for and against face validity in a given animal model. 

Predictive or pharmacological validity signals that a model can be used  to infer the effects of treatments in humans (4). Essentially, “the end justifies the means” wherein the model is retroactively judged valid because it leads to good outcomes in humans. For neuropsychiatric disorders, predictive validity is an extremely problematic notion. It is a problematic notion because a) it is rare that animal models lead directly to human successes, and b) because without any other supporting form of validity there is no basis for testing the drug in the first place. Indeed, the very point of animal models is to provide a grounds to test neuropsychiatric disorders first in animals, validate them there, and then apply the results to humans. Therefore, only cases of face and construct validity are considered and examined herein. 

Judging contemporary animal models of depression and schizophrenia

In order to judge the proposed categories of validity previously discussed, it is necessary to examine in detail contemporary models of both depression and schizophrenia and see if they fit in congruence with the proposed categories. Specifically, consider a dual-experiment on depression-like behavior in mice that contains, on first look, two separate animal models that represent construct and face validity, respectively (6), which are described below.

Lipopolysaccharides (LPS) are used to create bacterial infections in animal models. If injected at a low dose, then a time dependent behavioral alteration occurs, due to the bacterial infection. The body’s response to LPS releases the proinflammatory cytokines into the blood 2 hours after the injection. Sickness symptoms like a fever are abate after 6 hours, yet depressive behavior is observed for 24 hours after the LPS is injected. Therefore LPS is used in rodents to produce depressive-like states, impersonating human depressive symptoms. Specifically, LPS mimics an inflammatory occurrence, creates an activated immune response, which then promotes  inflammation in the brain, and this results in depressive-like behaviors. This can lead to clear behavioral expressions of depression-like activity, such as increased time spent in the periphery of spaces, 24 hours following an LPS injection (6). This sort of animal model is an attempt at construct validity, since it is assumed that depression in humans also is the resultant of an inflammatory response. If this is the case, then both the animal model of LPS-induced depression and the human model of depression have a common cause, meaning the animal model has construct validity.

Unpredictable Chronic Mild Stress (UCMS) is an animal model based on creating environmental conditions that lead to depression-like behavior in mice. UCMS models simulate depressive symptoms of humans from undergoing a series of negative psychological and environmental stimuli. Examples of UCMS include being restrained for four hours, or put in a tilted cage for 24 hours, had wet bedding for 24 hours, be forced to swim in 4 °C cold water for five minutes, and so on in a random fashion, the randomness of which makes these alterations even more stressful. UCMS application occurs for up to four weeks, after which mice also demonstrate depressive behavior, such as spending more time in the periphery of spaces (6).  The UCMS would presumably fall under face validity because it is altering mouse behaviors in a similar way to human behaviors, yet without a known mechanistic action. If the UCMS induced depression animal models show the importance of features such as neuropathological and behavioral of a human disorder, face validity can be achieved. 

These two different animal models of depression allow for comparing and construct whether the categories of validity are themselves valid. According to the previous definitions, LPS-induced depression animal models should be construct valid while UCMS-induced depression animal models should be face valid. It should be construct valid because it recreates the processes that cause the disease in humans and replicates the behavioral and neural features of the illness. However, it is not actually obvious what types of validity each animal model possesses. Specifically, results indicated that UCMS, as a chronic psychological stressor, can effectively activate a peripheral immune response which results in increased regulation of systemic levels of cytokines (6). UCMS was confirmed to induce cytokines to evoke immune activation in animals, which leads to homeostatic disturbances and immune dysregulation. Therefore, the results show that both LPS and UCMS stresses triggered brain neuroinflammation. While the LPS challenge activates brain immune response more rapidly and potently than UCMS, this means that UCMS could be considered construct valid as it seems to share a possible mechanism of action in the pathophysiology of depression. Similarly, one could even argue UCMS could also be considered a form of construct validity if the unpredictability of the environment caused the mice to be depressed in the same way that stressful environments cause humans to be depressed, since construct validity is often defined via the notion of a “common cause.”

In this study of the effects of LPS and UCMS on mice, the data indicated LPS induced more depressive-like behaviors and immune activation than UCMS. Implicitly, relying on behavioral assays are a form of face validity. At the same time, if inflammation is not the cause of depressive-like behaviors, then construct validity would not hold. It rests on an implicit hypothesis being true about a similar mechanism of causation. While the findings demonstrate depressive-like behavioural alterations are correlated with the induction of brain proinflammatory cytokines, it is unclear whether this relationship is causal, and also whether it is causal in humans as well. 

On the surface, LPS and UCMS induced-depressive models each have their own validity, respectively. But this seems to change depending on context and implicit assumptions, indicating that each has more than one type of validity, or that validity is not actually well-defined or universal. Ultimately, it may be that validity is a spectrum wherein oftentimes animal models are both face and construct valid to some degree, wherein a study could be more face or more construct valid. It’s difficult to say for any given animal model of mental disorders that they are concretely either category, and that their validity is not open to further interpretations and weakening or strengthening of that validity based on background assumptions. This makes it unclear to what degree the categories of validity are themselves valid enough to warrant use.

These issues hold true for more than just depression. Schizophrenia, a highly genetically influenced disorder, faces serious difficulties around validity as well (7).  For instance, the risk of schizophrenia developing comes from diverse genetic variations combined with environmental factors, making construct validity questionable. There are three major symptom clusters with schizophrenia: positive, negative, and cognitive symptoms. The three symptoms reflect diverse consequences of possibly developing abnormalities, and without a model showing all types of symptoms (many of which are shared with other disorders) it is unclear to what degree an animal model of schizophrenia would satisfy face validity. For example, negative symptoms in humans are blunt affect, asocial behavior, and reduced motivation. If an animal shows reduced motivation, is that face valid for schizophrenia or depression? Therefore, both kinds of validity seem questionable (or at least only achievable to a limited degree) in animal models of schizophrenia.


Animal models of neuropsychiatric disorders will continuously be used and developed in the future. These models, however, must improve and develop to make valid scientific research into different disorders. As it currently stands,  the lack of criticism of animal models in the scientific world hinders research, so it is the responsibility of scientists to address these problems, which they have attempted to do by proposing specific categories of validity for animal models.. Here it is shown that just having abstract categories of validity does not always translate well when applied to concrete experiments and models. Instead, it is arguable that validity and different categories (specifically face and construct validity) are instead spectrums. . For instance, a very common problem in the studies of mice models of neuropsychiatric disorders is that a study is commonly thought to have just one type of validity, but if we take a closer look, a study can have more than one, or those categories can shift based on background assumptions about mechanism of action. In truth, the proposed categories are messy, and may offer scientists undue confidence in their models, since they feel that fitting within a formal system lends credence to their results. Types of validity are very much like putting a bandage on a bullet wound. In the future, it is crucial that scientists don’t settle with one type of validity but instead consider validity a spectrum wherein each experiment must be assessed individually and in its specifics. Therefore, categories of validity should not be used as a concrete guideline but instead merely a resource to be used.


Mentor: Dr. Erik Hoel, Tufts University

  1. World Health Organization. World Health Report 2001. Geneva: World Health Organization; 2001.
  2. Nestler, EJ.; Hyman, SE.; Malenka, RC. Molecular Neuropharmacology: A Foundation for Clinical Neuroscience. 2. McGraw-Hill; New York: 2009
  3. Markou A, Chimmulera C, Geyer MA, Tricklebank M, Steckler T. Removing obstacles in neuroscience drug discovery: The future path for animal models. Neuropsychopharm. 2009; 34:74– 89. 
  4. Nestler, E. J., & Hyman, S. E. (2010). Animal models of neuropsychiatric disorders. Nature neuroscience, 13(10), 1161. 
  5. Aguzzi, A., Sigurdson, C., & Heikenwaelder, M. (2008). Molecular mechanisms of prion pathogenesis. Annu. Rev. Pathol. Mech. Dis., 3, 11-40.
  6. Zhao, X., Cao, F., Liu, Q., Li, X., Xu, G., Liu, G., … & Fan, K. (2019). Behavioral, inflammatory and neurochemical disturbances in LPS and UCMS-induced mouse models of depression. Behavioural brain research, 364, 494-502. 
  7. Jones, C. A., Watson, D. J. G., & Fone, K. C. F. (2011). Animal models of schizophrenia. British journal of pharmacology, 164(4), 1162-1194. 

About the author

Syra Bhoola

Syra is a rising senior at the Father Lopez Catholic High School.