Obesity is one of the most common health problems associated with increased risk for cardiovascular diseases, type II diabetes, cancer and psychiatric disorders (Apovian, 2016). One contributing factor to obesity is overeating of foods rich in sugar and fat which have reinforcing properties possibly via mesolimbic dopaminergic pathways associated with reward processing (Volkow et al., 2011). Previous functional magnetic resonance imaging (fMRI) studies have found increased neural activation in response to high‐palatable food cues in reward related brain regions such as striatum and orbitofrontal cortex in individuals with obesity compared to normal weight controls (Carnell et al., 2014, Feldstein Ewing et al., 2017, Stice and Burger, 2019, Verdejo-Roman et al., 2017), although some recent studies challenge these previous findings (Morys et al., 2020, Wall et al., 2020). Neural activation alterations in brain areas involved in reward processing during the non-food related paradigms have also been shown in people with obesity suggesting generalized alterations in reward related processing (Balodis et al., 2013, Bogdanov et al., 2020, Opel et al., 2015). These alterations might be related to functional differences in several domains of reward processing including motivation, sensitivity and learning (Treadway and Zald, 2013). However, which subdomains are impaired or altered in obesity remains unclear.

Reward motivation refers to the willingness to exert effort for reward (Soder et al., 2020) and reward sensitivity refers to the effect of reward values on choices (Treadway et al., 2012). Effort Expenditure for Rewards task (EEfRT) have been widely applied to examine reward motivation and sensitivity (Treadway et al., 2012, Treadway et al., 2009). Subjects make a series of choices between higher levels of physical effort for larger monetary rewards and lower levels of physical effort for smaller monetary rewards with varying probability and magnitude of rewards. The overall ratio between the number of choices of hard and easy task is thought to reflect reward motivation while the likelihood of selecting hard tasks as a function of magnitude and probability is thought to reflect reward sensitivity (Boyle et al., 2020, Boyle et al., 2019, Lasselin et al., 2017). Only one study to date used the EEfRT in people with obesity. This study found no evidence for dysfunctional reward motivation in obesity (Mata et al., 2017). Furthermore, people with obesity showed similar hard task choices in trials with high reward magnitude or high probability compared to healthy controls, which suggests intact reward sensitivity in obesity (Mata et al., 2017). Unfortunately, this recent study did not examine or report group differences as a function of both probability and magnitude, although dysfunctional effort expenditure under uncertainty might serve as a mechanistic explanation for the failure of weight loss treatments in obesity. Similarly, another study which measured physical effort by handgrip force has reported that people with obesity exerted similar effort for the money compared to controls suggesting normal non-food reward motivation in obesity (Mathar et al., 2016).

While reward motivation and sensitivity are related to immediate behavioral response to rewarding stimuli, reward learning is related to speed of behavioural adaptation based on reward feedback following action. It has also been hypothesized that alterations in reward processing in obesity might in part be caused by altered reinforcement learning (Kroemer and Small 2016). According to reinforcement learning theory, agents try to maximize total reward and minimize punishment in the longer term by learning from positive (reward) and negative (punishment) feedbacks via reward prediction error signals (Schultz, 1998, Sutton and Barto, 1998). Computational models of reinforcement learning estimated from the trial-by-trial behavior may provide a novel way for investigating latent cognitive variables underlying task performances (Gueguen et al., 2021). However, few previous studies investigated learning rates using computational reinforcement learning models in obesity. One study found reduced overall learning in people with obesity compared to lean subjects by using a probabilistic selection task (Frank et al., 2004, Kube et al., 2018). Another study using a stimulus response learning task reported lower punishment learning rates but similar reward learning rates in people with obesity (Mathar et al., 2017).

In contrast, recent studies using probabilistic pavlovian-instrumental learning and sequential decision making tasks have reported no learning rate differences between people with obesity and normal weight subjects (Janssen et al., 2020, Kube et al., 2020, Voon et al., 2015). Previous fMRI studies have mostly reported that people with obesity showed reduced dorsal striatal response during the actual receipt of caloric and palatable liquids but enhanced activation during sensory food cues (Burger and Stice, 2011, Ho and Verdejo-Garcia, 2021, Janssen et al., 2019, Stice and Burger, 2019). A recent framework using computational reinforcement learning models suggested heightened reward sensitivity coupled with decreased reward-related learning signals might underly these functional differences (Kroemer and Small, 2016). In contrast to higher sensitivity to food related rewards, one previous study reported lower negative but intact positive learning rates to non-food related rewards in people with obesity (Mathar et al., 2017). This recent study also found decreased response consistencies in people with obesity, indicating disrupted sensitivity of choices to differences in values.

Our study aims to build on these previous findings and to further investigate the association between obesity, reward sensitivity for gains and losses as well as learning differences using the probabilistic reversal learning task (PRLT) (Cools et al., 2002).