A rodent peeking out between railroad ties. Image credit: Joshua J. Cotton (CC0)
Emotions are fundamental to our experience and essential to our survival. They shape how we react to the world, interact with others or make decisions. But emotions are difficult to study because they are complex and subjective. One key building block of emotion is valence – the sentiment that something feels good (positive) or bad (negative).
Valence strongly shapes behavior. We avoid unpleasant experiences and seek out rewarding ones. Animals, including humans, learn to anticipate such outcomes when neutral cues become linked to positive or negative events, a process called associative learning. Problems in this type of learning are central to many psychiatric disorders. For example, anxiety can involve overgeneralizing fear, depression often shows reduced positive memories with a bias toward negative memories, and post-traumatic stress disorder is marked by intrusive traumatic associations. These disorders also show clear differences between men and women.
To study the biology of valence learning, researchers often use mice and train them to associate cues with either rewards, i.e., ‘appetitive outcomes’, or punishments, i.e., ’aversive outcomes’. However, most studies look at only one type of outcome in isolation, even though real-life situations often involve both. In addition, research has often included only male mice or applies behavioral phenotypes established in male mice to females without considering the influence of sex differences.
Schuler et al. set out to answer how the presence of only positive, or only negative or both valences together influences associative learning and its behavioral expression in male and female mice. This question arose because few studies have examined mixed-valence learning, and findings on sex differences in reward and threat learning are inconsistent. We need to understand this because sex differences are prevalent in valence-related psychopathologies, and we need robust models to study these phenomena in both sexes.
The researchers developed a novel mixed-valence conditioning protocol, in which mice are exposed to both positively and negatively valenced stimuli in the same context. They then identified predictors of learning in a data-driven way and found that field-standard metrics perform poorly in evaluating performance. Applying data-driven metrics to evaluate performance during training revealed sex-specific behaviors in the single-valenced protocols; for example, female performance suggested poor discrimination of cues in the aversive-only protocol, while males appeared to take longer to acquire positive associations. These apparent sex differences were not observed in mice trained in the mixed valence protocol. This shows that male and female mice can learn equally well about appetitive and aversive outcomes.
These apparent sex differences in learning are due to an underlying sex difference in exploration, where males have a higher baseline level of exploratory behaviors. Finally, they demonstrated that females are more sensitive to the delivery of repeated foot shocks, showing higher generalization of pausing response in the aversive-only protocol. This difference is not indicative of a lack of learning, but rather a consequence of differences in exploration specific to the aversive-only context.
Schuler et al. present a novel mixed-valence conditioning protocol for freely moving mice. This protocol overcomes limitations in classically employed fear- and reward-conditioning protocols and will allow researchers in behavioral and systems neuroscience to disentangle behavioral and neural correlates of normative salience and valence processing, as well as disruptions in animal models for psychiatric disorders. This work highlights the critical importance of considering sex when establishing behavior protocols and describes a novel approach for rigorously evaluating in-depth behavioral phenotyping data, a fundamental concern that has received far less attention than the development of phenotyping tools themselves.
