Compared to shuttle-box active avoidance, step-through passive avoidance, introduced in the 1960's (Kopp et al., 1966; Jarvik and Kopp, 1967), is considerably less stressful and in the test session no shock delivery is actually present. Nevertheless, the training session still features a brief painful stimulus. In an ideal animal-friendly behavioral test, the motivating factor should not be painful or stressful. Furthermore, the observable behavior should be natural (i.e., an ethologically relevant species-specific motor or postural pattern). Finally, the outcome variable associated with this behavioral response should be practical to measure in a laboratory setting through a method that is safe for the animals (e.g., direct observation, videorecording, videotracking, audiorecording of ultrasonic vocalizations, photocell actimetry, weight sensors, infrared thermometry and other non-invasive methods). These are the three main characteristics that a behavioral test should have to be qualified as animal-friendly. Let us first focus on the motivational aspect of behavioral tests, the first of the key components mentioned above.

Ethologists have been stressing the argument that motivating factors are species-specific (Gerlai and Clayton, 1999a; Gerlai, 2021). A stimulus that is appetitive (rewarding) or aversive (punishing) for one species, may be neutral, or may have the opposite reinforcing value to another. Even among closely related species, the rewarding value of a stimulus may be remarkably different. Among felids, for instance, tigers like to bath in water, while lions do not. Hence a swimming pool may be a reward for the former, but not for the latter species, as found, for example, by Allison Hedgecoth who provided a water pool to a lioness and a tigress living together in the same environment in the Noah's Ark Animal Sanctuary of Locust Grove, Georgia, United States (Harries et al., 2020). The main issue, however, in the behavioral neuroscience literature is that systematic analysis of what motivates animals used in laboratory settings is often lacking, or that ethology research often does not intersect with biomedical studies.

Considering, for example, laboratory rodents, the house mouse (Mus musculus) and the common rat (Rattus norvegicus) are the two most widely used species in biomedical research. Their employment is almost universal in translational research studying mechanisms of central nervous system disorders. Most studies that require aversive stimuli with rodents use electric shocks. But electric shocks are rather unnatural stimuli. However, almost no one considers what consequences may result from the unnatural aspect of this stimulus. It is just assumed that pain is pain, and that electric shock-induced pain is relevant and strongly motivating. Most scientists do not even consider what complication this electricity passively running through the body of the animal, including its brain, may cause with respect to neuronal activity: such electric currents may alter synaptic function and numerous underlying molecular mechanisms. Similarly, studies that employ appetitive stimuli, almost always use food that the experimenter picks out based on tradition, habit or just personal preference. Comparative analyses of what food types, food quantities, food textures and food sizes are most preferred by rats or mice are often not considered, or have not even been conducted. Briefly, as animals are, through evolutionary processes, adapted to their natural environment, they possess species-specific characteristics that represent genetic predispositions, instincts in colloquial terms, that determine, or at least heavily influence what they like, what they dislike, and how they respond to these stimuli. Taking these species-specific features into account is thus a must in animal-friendly experimental designing (Gerlai and Clayton, 1999a,b). In rodents, a typical animal-friendly motivating factor is neophilia (attraction for novelty), which drives, for instance, object exploration behavior in the object recognition test (d'Isa et al., 2014), head-dipping in the hole-board test (d'Isa et al., 2021a) and arm alternation in the spontaneous alternation T-maze (d'Isa et al., 2021b). A similar example is the continuous spontaneous alternation test using a T-maze, which utilizes novel place preference to study short-term spatial memory in rodents (Gerlai, 1998).

Regarding the second key element mentioned above, the observable behavior in an animal-friendly test should be naturally displayed by the animal (e.g., should be part of the ethogram). Preferably, it should be a spontaneous behavior, an instinctive response that requires no pre-training, during which typically punishments and rewards are used by the experimenter to lead to a target behavior. Punishments are commonly painful stimuli (as electric shocks), while rewards, as food or liquids, are often associated with food-deprivation or water-deprivation, in order to use hunger or thirst as motivating factors. Lack of the need for pre-training makes the test more animal-friendly because it avoids punishments and deprivations, and is also time-saving for the experimenter. It is, however, also possible to use conditioned behaviors in an animal-friendly way, if certain conditions are respected. In particular, rewards should not be associated with a previous aversive state. Chow and colleagues, for instance, designed a reward-based cognitive test for gray squirrels in which no food-deprivation or water-deprivation was employed (Chow et al., 2017). The motivation of the rodents was ensured simply by using food rewards (hazelnuts) that were different from their daily diet (seeds, fresh fruit and vegetables), i.e., novelty alone was sufficient to motivate the animals. Novelty-seeking and exploratory drive (i.e., the motivation to learn about new places and/or new inanimate or animate components of the environment) are almost universal among animal species, and certainly have been shown for laboratory rodents (Gerlai et al., 1990; Crusio, 2001). In fact, stabilizing natural selection has been inferred for exploratory behaviors from fish to mammals, as it leads to an optimal level of activity ensuring the ability of the animal to find resources, including food, water and mates, as well as escape routes leading away from predators (Gerlai et al., 1990; Crusio, 2001). The use of novelty as a motivator may not be appropriate in some research contexts and, for certain studies, aversive stimulation may be required. However, even in such cases, painful punishments could be and should be substituted with non-painful aversive alternatives, for example, air-puffs. Indeed, air-puffs have been efficiently employed to elicit robust conditioned place avoidance negating the need for using any painful stimuli (d'Isa et al., 2011). Even for studies specifically focused on fear reactions, alternatives to painful stimulation are available. Odor of predators (e.g., fox's urine, or an extract from it) has been efficiently used to induce avoidance reactions and fear without previous painful stimulation (Blanchard et al., 2003).

The main steps for designing an animal-friendly test can be summarized as follows: (a) prepare a list of behaviors typical of the species (the ethogram), along with what stimuli may induce these behaviors, i.e., the motivating forces; (b) exclude behaviors induced by pain, physical suffering or psychological stress; (c) from the remaining, choose a behavior that can be studied through an apparatus that can be used in a laboratory setting; (d) choose which outcomes could be measured, safely for the animals, in the most efficient and precise way in order to provide quantitative experimental data. Let us examine an experimental example of how these steps may be accomplished.

A typical behavior of rodents is food hoarding, that is, collecting and hiding food as supply storage for times of food scarcity. This behavior can be observed in more than 180 rodents (Zhang et al., 2022). This is an adaptive behavior that is observable both in nature and in the laboratory setting. It is an instinctive behavior that does not require pre-training. Two main strategies are adopted by food hoarding rodents. Scatter hoarders, as gray squirrels, hide food in many dispersed small hoards. On the other hand, larder hoarders, as hamsters, store food in one large hoard, named the larder. A classification of the hoarding strategies of 183 rodents is provided by Zhang and colleagues (Zhang et al., 2022). These hoarding behaviors may be utilized by the experimenter to devise behavioral tests of motivation (during the food accumulation phase) or of spatial memory (during the subsequent phase of food retrieval from the spatially separated hoards). For motivation tests, easier to study in larder hoarding rodents, the measurable outcome could be the total weight of the seeds or pellets collected and stored in a fixed amount of time. For spatial memory tests, which would be best studied with scatter hoarding rodents, the recorded outcome could be the number of errors in finding the hoarding sites containing the previously stored food. Alternatively, spatial memory could be studied also in larder hoarders if, during the accumulation phase, the sources of food are multiple. Number of errors (returning to an already depleted food site) would serve as memory index. An apparatus for the testing of food hoarding behavior in a laboratory setting has been realized, for example, by Robert Deacon at Oxford University (Deacon, 2006).

An ethological approach may be useful to devise animal-friendly behavioral tests for two reasons. On the one hand, it may help researchers to choose among the elements of the ethogram a behavior that does not require painful or stressful motivating factors. On the other hand, among a taxonomical family of species (for example rodents), it may help researchers to select the most suitable species for a certain test. Let us return to the example we mentioned above. Laboratory mice are larder hoarders, just like hamsters, but their propensity to hoard is relatively low under baseline conditions. In order to avoid food-deprivation, long testing sessions may be required to obtain replicable results, including, e.g., overnight testing sessions (Deacon, 2006). Hamsters, on the other hand, have a high propensity to hoard (Vander Wall, 1990; Harris, 2017). Up to 90 kg of food have been found in hamster burrows (Nowak and Paradiso, 1983). Among food hoarders, they display a specific behavior known as cheek pouching, that is accumulating food in cheek pouches, specialized pockets that allow food transportation. Instead of eating the food items, hamsters keep the food items in their mouth to carry them to a safe place for storage (the larder). Importantly, hamsters easily show this behavior even when they are not hungry, with a latency to hoard within 2 min (Montoya and Gutiérrez, 2016). This peculiarity of hamsters makes them particularly suitable as animal models for scientists who want to design an animal-friendly reward-based memory test that does not require any previous starvation.

Another rodent, the chinchilla (Chinchilla lanigera), displays a peculiar behavior known as sand-bathing: when presented with a box full of sand, it will readily start rolling in the box, rotating along its longitudinal axis, to rub its fur in the sand (Stern and Merari, 1969). This natural and spontaneous behavior can be easily elicited in a laboratory setting and sand could be used as an animal-friendly reward in instrumental learning tests without the need of any previous deprivation condition (Redman, 1974).

Eastern woodrats (Neotoma floridana), also known as pack rats, have a special attraction for shiny objects, which they readily approach, pick up and bring to their nest, where they collect them (Bradley et al., 2022). This natural tendency of woodrats could be used in behavioral tests, employing small metal objects, as stripes or balls of aluminum foil, as motivators (Kaufman and Kaufman, 1984).

Of course, the issue is that quite often neurobiological, genetic, or other methods may not be as readily available, or as sophisticated, for such species as hamsters and chinchillas as for the favorites of biomedical research, mice and rats. How can we solve this conundrum? Firstly, we could improve biotechnological methods for the so called “alternative” species. Secondly, we could improve our understanding of the ethology, the natural species-specific behavioral characteristics, of the preferred model organisms, e.g., of mice and rats. Certainly, advances in both of these areas have been made during the past few decades. Regarding the first area, numerous novel techniques may now be equally useable with mice and hamsters (and many other species). The CRISPR/Cas technology is a clear example (Kampmann, 2020). Concerning the second area, there have been research efforts adopting ethological approaches in mouse neurobehavioral genetics, as for instance testing mouse mutants in the wild (Dell'Omo et al., 2000; Vyssotski et al., 2002) or in laboratory environments more closely resembling a natural habitat, like Eco-HAB (Puścian et al., 2016; Winiarski et al., 2022). Anders Ågmo's research group at University of Tromsø recreated a seminatural environment in the laboratory for the evaluation of rat behavior (Chu and Ågmo, 2014), a method that has been employed in several subsequent studies (Chu et al., 2015; Houwing et al., 2019; Le Moëne et al., 2020; Heinla et al., 2021). In the testing sessions, which may last days, rat behavior is continuously video-recorded and subsequently scored off-line by the researchers. Notably, another important ethological approach of the new century is testing the animals not in a setting designated uniquely for testing sessions, but rather in the permanent housing environment in which they commonly live (Mingrone et al., 2020; Voikar and Gaburro, 2020). In nature, most rodents build burrows (or occupy pre-existing holes or burrows) to use them as homes (long-term inhabiting spaces), in which they return to sleep, store food, seek shelter from the elements, keep warm, hide from predators, give birth, raise the pups and share a social life with conspecifics. Rodents develop a strong bond with their home and show a territorial behavior toward it, actively defending it from possible invaders. In the laboratory, if an unfamiliar conspecific is placed in the home-cage of mice or rats, the intruder will rapidly be attacked by the resident animal (Koolhaas et al., 2013; Ruzza et al., 2015). The home-cage is the place where laboratory rodents feel safest and where they are more likely to display spontaneous natural behaviors. Thus, the idea of testing in the home-cage has been gaining considerable attention in the past few years, and several home-cage automated multi-variable recording systems have been developed, e.g., the IntelliCage (Galsworthy et al., 2005; Kiryk et al., 2020; Iman et al., 2021), PhenoMaster (Urbach et al., 2008; König et al., 2020), Actual-HCA (Bains et al., 2016; Mitchell et al., 2020) and SmartKage (Ho et al., 2022). Automated home-cage testing systems have several advantages: (a) they allow behavioral phenotyping without human interference and without the consequent handling-related stress; (b) the animals are not tested in an external apparatus but in their familiar and well-known housing environment, which eliminates confounds arising from anxiety; (c) data collection is not restricted to a specific moment of the day, but can be performed continuously, 24 h a day, 7 days a week, allowing a more precise and realistic assessment of behavior; d) long longitudinal studies (lasting weeks, months or years), or even life-long studies, can be performed on the same animals with a continuous behavioral assessment, which is particularly relevant for developmental neuroscience and aging neuroscience; (e) interactive elements (e.g., levers, nose-poking ports, motorized doors and running wheels) may be installed in these home-cages, allowing not only detailed motor assessment, but also complex cognitive testing; (f) animals are tested in a natural social context while living together with other conspecifics, thus providing motor and cognitive measurements with a higher ethological validity and allowing additionally to monitor and analyze complex social interactions. Some of these automated home-cage testing systems are modular, allowing the connection of multiple cages to create a more complex envinronment. For instance, IntelliCage can be connected to two social boxes containing different social stimuli (Mitjans et al., 2017), while in ColonyRack mice can freely roam across 70 cages, arranged in a two-sided rack with fivs columns and seven rows, in which the cages are connected both horizontally and vertically (Zocher et al., 2020; Kempermann et al., 2022). The most recent innovation within this automated behavioral testing approach is connecting home-cages to mazes (Mei et al., 2020; Kohler et al., 2022), granting the experimental subjects free access to the novel test environment. This allows the animals to decide voluntarily when and for how long they explore the maze, similarly to what would happen in nature when rodents decide to leave their burrow for external exploratory excursions.

We believe that bringing closer the fields of ethology and neurobehavioral genetics or behavioral neuroscience will be the solution and will lead to cross-fertilization of these fields. Similarly to how the application of neuroscience-related knowledge to ethology led to the birth of neuroethology (i.e., the study of the neural basis of natural behaviors), the reverse could lead to an ethologically based neuroscience, or ethological neuroscience, which can be defined as the employment of knowledge of the natural behavior of animals in the wild to develop animal models of behavior and behavioral tests for neuroscience research. This ethologically based neuroscience can lead to animal-friendly testing approaches that will not only be more oriented toward the welfare of the animals involved, but also will provide more reliable and more replicable results for the experimenters.

Reproducibility is when we obtain the same results repeatedly by using identical methods (Kafkafi et al., 2018; Gerlai, 2019), whereas replicability is when we reach similar conclusions by adopting different methodologies (Kafkafi et al., 2018; Gerlai, 2019). Minimizing stress of the tested animals is a value in itself from an ethical point of view. However, since stress is a confounding factor that increases variability of experimental outcomes, minimizing stress is also fundamental to achieve methodologically sound scientific research. Why does research that ignores species-specific features lead to increased variability? Why is stress a confounding factor that reduces reproducibility? These are intriguing questions that would deserve specific research. The answer may lay in the fact that stress causes activation of the hypothalamic–pituitary–adrenal (HPA) axis, which in turn alters physiological processes regulating cognition and behavior (Moreira et al., 2016). HPA reactivity depends on genetic, epigenetic and environmental factors (Holmes et al., 2005), which makes it more difficult to predict than instinctive responses. Let us make some overarching theoretical points. Most animal research includes human handling. Human handling is extremely difficult to standardize (Crabbe et al., 1999). Even if handling was perfectly standardized, stress reactivity of the animals would not. Animals experiencing more stress due to the experimental procedures will be more responsive to human handling, which then will lead to elevated error variation in the behavioral test. In order to maximize experiment reproducibility, the best option is to minimize handling-related stress (Gouveia and Hurst, 2017). Considering, for instance, mice, although tail picking is the most commonly employed method of handling (Ueno et al., 2020), this method features tail lifting, tail suspension and swinging the animal over a void, which are highly stressful for the mice. Indeed, tail lifting, compared with alternative handling methods that do not require tail lifting, increases anxiety in the open-field test (Gouveia and Hurst, 2019) and elevated plus maze (Hurst and West, 2010), and it has been shown to reduce exploratory activity (Gouveia and Hurst, 2017), to increase aversion for the human handler in voluntary interaction test (Hurst and West, 2010) and to impair responsiveness to sucrose reward, indicating a reduction of reward's hedonic value (Clarkson et al., 2018). Several animal-friendly approaches are now available to avoid the negative impact of human handling on mice: (a) adopting non-aversive manual handling techniques, as open-hand retrieval through the cupping method (Hurst and West, 2010; Gouveia and Hurst, 2017, 2019; d'Isa et al., 2021b; Davies et al., 2022); (b) employing a tool to handle the mice, as a plastic handling tunnel (Hurst and West, 2010; Gouveia and Hurst, 2013, 2017, 2019; Sensini et al., 2020; Davies et al., 2022); (c) using automated home-cage testing systems in which behavioral outcomes are recorded without physical interaction with the human experimenter (Kiryk et al., 2020; König et al., 2020; Mitchell et al., 2020; Ho et al., 2022; Kohler et al., 2022; Winiarski et al., 2022).

Furthermore, not knowing the species-specific characteristics of the studied organism, for example, applying inappropriate motivators, forcing the animal to exhibit behavioral responses it would not normally perform, and measuring the behavior under artificial conditions that do not have much to do with the natural environment in which the animal evolved, all can elevate random error, simply because the individuals tested this way may have to find unique solutions to the problems, considerably increasing individual differences in the study (Gerlai and Clayton, 1999a,b). To put it in the words of the aforementioned pioneer of experimental behavioral research Willlard Stanton Small, “the experiments must conform to the psycho-biological character of an animal if sane results are to be obtained” (Small, 1901).

Animal-friendly tests utilizing species-specific features of the studied organism may not be always available or applicable, but, when they are, they should be employed as a first option, in order to maximize both animal welfare and repeatability of experimental results. When fully animal-friendly tests are not available, then the least stressful available test should be employed. In Table 1 we present a rating scale for behavioral tests based on their impact on animal welfare. This rating is not meant to be final, but rather a starting point to stimulate reflection and discussion on the differential stress impact of behavioral tests. We hope that in future an increasing number of studies will employ tests of class A (animal-friendly) and B (minimally stressful) and that, in accordance with a progressive refinement principle, new animal-friendly tests will be designed to substitute the more stressful alternatives.

Rd'I and RG: conceptualization, writing, revision, and final approval of the manuscript. Both authors provided funding. Both authors contributed to the article and approved the submitted version.