Background

Contingency reversal learning tasks, where an animal is trained to associate a cue with a reward or a punishment that later is reversed, are frequently used to study neural mechanisms of cognitive flexibility (Izquierdo et al., 2017). In mice, discrimination and reversal learning can be challenging to assess, partly because of two salient and contrasting aspects of species-specific mouse behavior: novelty-induced anxiety (Crusio, 2001) and the tendency to follow incidentally learned spatial cues (Hebert et al., 2017). The natural defensive behavior of mice limits the use of tests based on negative reinforcement, as they reduce exploration and almost invariably result in the animal not performing any response at all. For example, the addition of punishments to incorrect responses appears to worsen reversal learning proficiency as assessed in a touchscreen paradigm (Jager et al., 2020). On top of this, there are ethical concerns about using negative reinforcements in animal experiments. The use of only positive reinforcers, such as food rewards, also has its drawbacks since the motivation to respond to a reward-associated stimulus may change over time along with satiety (Goltstein et al., 2018).

Furthermore, rodents have a natural tendency to alternate between responses (Lalonde, 2002; Deacon and Rawlins, 2006), and in particular, young individuals tend to explore the consequences of alternative non-rewarded responses provided that these do not trigger an openly aversive effect (Walker et al., 1955; Mechan et al., 2009; Sanderson and Bannerman, 2012). In most operant tasks assessing discrimination learning in mice, wrong responses are simply not rewarded and do not require a corrective action by the mouse, which has to “understand” that the given response was wrong by the absence of a consequence. Time-outs appear insufficient to discourage animals from attempting alternative responses, and these alternative attempts may confound the animals’ cognitive ability with its exploratory attitude (Luo et al., 2020). The introduction of an operant response to re-start a trial is a partial remedy to this drawback; however, the presence of a negative, but not aversive, cue associated with a wrong response is the best way to achieve the highest possible proficiencies in these cognitive tasks.

Reversal learning tasks designed for rodents are often conducted in mazes or operant conditioning chambers, where the rewarded (S+) and non-rewarded (S-) stimuli are set at a physical distance sufficient to be discriminated based on their position. When discrimination between the stimuli is not based on location (e.g., shape, odor, or texture discrimination), the incidental learning of their positions in space generates interference with spatial rules, at least during the test’s procedural learning steps. This phenomenon is known as overshadowing and occurs when a second salient cue, such as a spatial one, is in conflict with the designed S+ stimuli (Sánchez-Moreno et al., 1999). Indeed, it has long been established that rats follow spatial cues as a first strategy to retrieve rewards in a maze (Packard and McGaugh, 1996), and some individuals tend to follow spatial strategies even after having been trained in a non-spatial cue-response task in a water maze (Packard and McGaugh, 1992). Similarly, when initially trained to discriminate odors in the “classical” two-chamber setting of the attentional set-shifting task (Bissonette and Powell, 2012), rodents invariably start following spatial cues to figure out where to find a reward (Tait et al., 2018). Suppression of this spatial bias interfering with the conditioning stimulus may require additional days of training and may – paradoxically – result in worse performance of animals that usually rely more on hippocampal-dependent spatial information: an ecologically more relevant feature when seeking for a reward-related contingency in rodents (Devan and White, 1999; Chang and Gold, 2003). The interference of incidentally learned spatial information can be so relevant that it can result in paradoxically enhanced discrimination learning in hippocampal lesioned mice (Sanderson et al., 2012).

Here, we present a protocol to assess discrimination learning and contingency reversal in a simple home-made setup, under conditions that improve on both the previously described limitations of most current discrimination learning devices. The present protocol minimizes the interference of non-relevant spatial information by concentrating the S+ and S- stimuli on a single, small pedestal placed in a compartment of the arena that animals will recognize as a single “reward zone.” This is a significant difference compared to the arenas used for the classical attentional set-shifting tasks, where the S+ and S- are presented in two separate adjacent compartments (e.g., Colacicco et al., 2002). In addition, we apply a negative response signal by introducing a sliding lid moving inside the test box upon attempts to reach the non-rewarded vial. This step is equivalent to closing the reward receptacle, an operation that can be performed on some other operant tasks (e.g., Krackow et al., 2010). Furthermore, it also acts as a “trial-ended” signal and solicits a withdrawal response in the mouse to the waiting zone, thus prompting a new trial.