Alcohol use disorder (AUD) is highly prevalent in the United States  and incurs substantial individual and societal costs . AUD is considered a heterogenous disorder involving complex etiologies and mechanisms. Animal and human experimental models play an integral role in understanding these multidimensional aspects of AUD. However, there is growing concern as to the degree of confidence we can have in animal experiments and their behavioral endpoints in terms of their translational utility . Similarly, human laboratory studies have been criticized for their lack of predictive utility of clinical trials outcomes . The absence of predictive validity of some animal models has contributed to a decline in psychiatric drug development programs .
Compared to human laboratory models, experimental studies in animals allow for precise experimental control over several parameters, including genetics, the influence of environment (upbringing, stress), and previous experience with alcohol, drugs, and rewards in general. The use of experimental animals in research allows for the examination of neurochemical, neurobiological, and neurophysiological factors that may ultimately translate to disease states in humans. These correlates can then facilitate the development of novel therapeutic targets for complex, heterogeneous disorders such as AUD. However, animal models of AUD are not without criticism. Notably, genetically diverse rat strains (outbred) do not readily consume alcohol and reach minimal blood alcohol concentrations. While several experimental manipulations were developed to induce alcohol drinking or self-administration behaviors that result in pharmacologically meaningful ethanol intake, these manipulations introduce a host of other influential factors (e.g., stress, palatability) that may limit generalizability. Rodents genetically bred to consume high levels of alcohol  have also been scrutinized for their lack of generalizability.
While it is important to acknowledge that no animal model of addiction fully encapsulates the human condition, experimental work in animals permit the investigation of specific and observable elements of the addiction process. Thus, animal models are most likely to have construct or predictive validity when the model mimics the specific signs or symptoms associated with a given disorder. The focus of animal models should not be to fully emulate the whole syndrome but rather to inform on specific domains that can be validated across species. That is, the goal of the animal model is to achieve a better understanding of the biological dysfunction that contributes to the disorder and to ensure that this level of understanding can be translated to novel treatments. Equally important, animal models should be validated beyond mirroring behaviors. Demonstrating shared circuit abnormalities is critical to establish face validity. In several areas of neuroscience, including addiction, there is tremendous evidence for conservation across species for basic behaviors and underlying circuit-, cellular-, and molecular-based mechanisms.
Human laboratory studies have also been developed to study discrete aspects of AUD phenomenology. Herein, we argue that human laboratory models provide a “bridge” between preclinical and clinical studies of AUD by allowing for well-controlled experimental manipulations in humans with AUD. As such, examining the consilience between experimental models in animals and humans in the laboratory provides unique opportunities to refine the translational utility of such models. This would allow alcohol researchers to more effectively leverage preclinical findings to clinical applications and vice versa. For example, laboratory-controlled alcohol administration permit investigations into the pharmacokinetic and pharmacodynamic responses to alcohol, which are proposed AUD risk factors. Human self-administration permits objective behavioral assessment of alcohol consumption, motivation, and compulsive use which are relevant to addiction phenomenology. Given that progression to addiction is accompanied by increasing salience of drug-paired cues, cue-reactivity assessments have also been implemented into human laboratory models. Importantly, such models can then be leveraged to treatment development and serve as early efficacy markers for promising treatments, both behavioral and pharmacological [4, 7].
In the present review, we provide a systematic description and contrast of commonly used animal paradigms for the study of AUD, as well as their human laboratory analogs. Previous reviews from our group have emphasized the opportunities to improve translational science for AUD [7,8,9]. However, this is a unique endeavor in that we focus on preclinical models in detail and discuss each model from the perspective of clinical and translational research. While there is a wide breadth of animal species in AUD research, the paradigms discussed in this review rely predominately on rodent research. The overarching goal of this effort is to provide critical analysis of these animal models and to link them to human laboratory models of AUD. The premise is that addiction phenomenology is inherently a human process and that capturing key phenomenological aspects of addiction under controlled conditions, in humans and rodents, poses unique challenges. By systematically contrasting preclinical and controlled human laboratory models, we seek to identify opportunities to enhance their translational value through forward and reverse translation (see Table 1). We provide future directions to reconcile differences between animal and human work to improve translational research for AUD.
Experimental paradigms used to model alcohol reward and intake
Conditioned place preference
Conditioned place preference (CPP) is a form of associative (Pavlovian) learning used to measure the motivational effects of stimuli/cues, or contexts. In this approach, the rewarding value of alcohol is measured by the degree to which an organism spends time in an environment (place preference) or prefer a flavored solution (taste preference) that has been paired with alcohol (see Fig. 1). It has been proposed that the behavioral loss of control over alcohol drinking that occurs in humans may be a consequence of the attraction to conditioned alcohol-paired stimuli via learning processes involved in CPP. CPP experiments are one of the most frequently used paradigms in clinical trials to test the abuse potential of new drugs. However, CPP does not directly probe how much an animal is willing to pursue the reward (i.e. operant, goal-directed behavior). Relative to operant self-administration, CPP in rodents requires passive drug administration, which results in distinct neurobiological effects relative to drug delivery that is not controlled by the animal.