Adolescence is a developmental period associated with changes in tolerance of uncertainty (Tymula et al., 2012; van den Bos and Hertwig, 2017). Adolescents are often involved in risky behaviors and higher sensation and novelty seeking (van Duijvenvoorde et al., 2016). This increased attentiveness to sensations and novelty paired with a heightened tolerance of uncertainty may in turn be connected to an increased reward response, leading to greater motivated behavior among adolescents to receive rewards (Braams et al., 2015). At the neural level, altered reward processing has been observed in reward related structures such as the ventral striatum (VS), anterior insula (AI), thalamus and supplementary motor cortex (van Duijvenvoorde et al., 2016; Oldham et al., 2018). Focusing on the VS, research into adolescents' altered sensitivity during reward anticipation has produced inconsistent findings, including both adolescent hyper- (Silverman et al., 2015) and hypoactivation (Bjork et al., 2010). However, this appears to be highly task dependent.

One key paradigm to investigate VS reward processing is the monetary incentive delay (MID) task (Knutson et al., 2000). This paradigm allows one to temporally disentangle the phases of reward anticipation and receipt (Lutz and Widmer, 2014). During these phases similar brain regions, including the VS, pallidum, insula, thalamus, hippocampus, and motor areas (Oldham et al., 2018; Cao et al., 2019) are active in adults and adolescents. However, when contrasting adolescents and adults the activation pattern in the VS varies across studies. While Bjork and colleagues (Bjork et al., 2004, 2010) detected reduced activation of the VS in adolescents compared to adults for cues that predicted a reward, other studies using a different task design observed heightened activation of the VS in adolescents (Ernst et al., 2005; Van Leijenhorst et al., 2010). Overall, the observed hypo- or hyperactivation of the VS in adolescents could signify the reorganizing of the reward and motivation circuity during adolescence (Fair et al., 2009; Doremus-Fitzwater and Spear, 2016). This might reflect adolescent hypersensitivity to larger rewards (Cohen et al., 2010) or different reaction times and neural activations in accordance to different reward probabilities. Previous inconsistent findings of adolescent VS hypo- or hyperactivation may depend on methodical differences between studies such as the task, e.g., MID, wheel of fortune, slot machine task, cake gambling task, pirate task (Jarcho et al., 2012), child-friendly versions of the MID task (Helfinstein et al., 2013; Kappel et al., 2013), or differences in stimuli or context (Nees et al., 2012; Bartra et al., 2013; Nelson et al., 2016).

Few studies have paid attention to different reward probabilities or success rates used within paradigms (Abler et al., 2006; Van Leijenhorst et al., 2010). However, when taking into account that adolescence is associated with changes in the tolerance of uncertainty, risky behaviors and novelty seeking, it can be assumed that differences between adolescents and adults in activation of the VS may also depend on the anticipated probability of a reward. The rate at which a reward is given or a response is successful could therefore play an important role when trying to understand adolescent behavior, in clinical as well as in non-clinical samples. To the best of our knowledge, the relation between activation of the VS as a function of different reward probabilities has never been studied in healthy adolescents compared to healthy adults using the MID task. However, implications for the relevance of taking different reward probabilities into account comes from studies on adolescents with ADHD. While von Rhein and colleagues (von Rhein et al., 2015) chose a 33% reward probability and did not find alterations in reward related regions during reward anticipation in adolescents and young adults with ADHD compared to individuals without ADHD (von Rhein et al., 2015), other studies used a different reward probability of 66% and reported reduced VS activation in adolescents with ADHD compared to those without ADHD (Scheres et al., 2007). A meta-analysis further reported hyporesponsiveness of the VS in adolescents and adults with ADHD compared to those without ADHD (Plichta and Scheres, 2014). These inconsistent findings highlight the importance of different reward probabilities when examining neural reward processing in adolescents.

Thus, the purpose of the current study was to examine and compare the extent to which activation in the VS between adolescents and adults is a function of the respective reward probability. The original MID task (Knutson et al., 2000) uses an adaptive algorithm to achieve an average hit rate of 66%, since a reward probability of 50% induces the highest dopamine release in animal studies (Fiorillo et al., 2003). A reward probability of 66% therefore induces sufficiently high dopamine release and activation in the VS, setting up a positive association between cue and reward (Plichta and Scheres, 2015). This in turn leads to participants' perception of the paradigm as plausible. We extended the original MID paradigm by three different probabilities which varied with respect to the degree a successful reward was gained (i.e., reward probability: 33, 66, and 88%), thus providing a varying degree of uncertainty during reward anticipation.

By including three reward probabilities and comparing adolescents and adults, this opens up the possibility to further investigate the relationship between brain activation in the VS and varying reward probabilities in a developmental context.

We hypothesize, that both age groups will show a higher percent signal change in the VS during the anticipation of the monetary compared to a verbal feedback condition in all reward probabilities. The following hypotheses are referring to the monetary condition. Based on previous work investigating dopamine release in the VS (Fiorillo et al., 2003), we hypothesize that the highest percent signal change and fastest reaction times will occur in the 66% reward probability and a lower percent signal change as well as slower reaction times will occur for the other two reward probabilities in both age groups. However, we expect differences between the age groups for the 33% and the 88% reward probabilities. Specifically, we hypothesize that adolescents demonstrate a higher percent signal change in the VS and faster reaction times compared to adults for the 33% reward probability because adolescents anticipate even unlikely rewards as an incentive. In contrast, compared to adolescents, adults will show faster reaction times and higher percent signal change for the 88% reward probability.

This study aimed for a better understanding of reward processing from a developmental perspective by comparing adolescents and adults in terms of their processing of different reward probabilities. Previous studies mostly used a reward probability of 66% as this had been demonstrated to induce the highest dopamine release in reward related structures (Fiorillo et al., 2003; Plichta and Scheres, 2015). Expanding the task by two further reward probabilities, we were particularly interested if and to what extent these different reward probabilities lead to different effects in adolescents as compared adults on the behavioral and neural level. Adolescence is a period of increased tolerance toward uncertainties (Blankenstein et al., 2016) and rather unlikely rewards might be a stronger incentive for adolescents as compared to adults. We therefore expected a difference between both age groups such that adolescents as compared to adults react faster and with a higher percent signal change in the VS for the lowest reward probability (33%) whereas adults demonstrate faster reaction times and higher percent signal change than adolescents for the high reward probabilities (66 and 88%).

We presented participants with a MID task that compared a monetary with a verbal condition and included three different reward probabilities indicated by differently colored cue stimuli. Participants were not informed beforehand that the different colors corresponded to different reward probabilities. After the scanning session, participants were asked to rank the cue stimuli with respect to their anticipated gain. Both age groups did not differ from each other regarding the ranking.

In line with previous work (Plichta et al., 2013), our task induced faster reaction times in the monetary condition for both groups and we found an interaction between age group x condition such that adults were faster than adolescents in the monetary but not the verbal condition. While the whole brain analysis revealed overlapping brain regions including the VS for adolescents and adults in the contrast monetary > verbal condition in line with previous literature (Plichta et al., 2013; Cao et al., 2019), we did not find an interaction effect between age group and condition for VS activation. For this interaction, our data revealed activation in the right anterior insula, left superior parietal lobule and left middle occipital gyrus but not in the VS on a whole brain level. As such, our data did not support results indicating striatal hyperresponsiveness in the VS (e.g., Van Leijenhorst et al., 2010) in adolescents as compared to adults that have been a point of discussion within the research community. Current data illustrate an effect of reward probability on a behavioral and neural level. Both groups reacted faster in response to the low reward probability of 33%, however, on a neural level, differences between age groups emerged. Adolescents demonstrated the greatest percent signal change in the VS for the 33% reward probability, while adults had the greatest percent signal change for the higher reward probabilities (66 and 88%). Previous inconsistent results may have therefore been induced by the different reward probabilities that were used in the task. However, current findings could also be due to the different framing used in our study by not integrating loss trials into the task.

Behavioral data demonstrated a main effect of reward probability. Inherent in the task using three different probabilities were varying speed requirements such that the 88% reward probability started with a reaction time window that was easier to meet than 66%, which in turn was easier than 33%. Therefore, the declining of a higher hit rate and reward probability from 88 to 33% was associated with higher speed requirements which are linked to cognitive effort that has been demonstrated to influence the ventral and dorsal striatum (Schouppe et al., 2014; Dobryakova et al., 2017). Furthermore, the paradigm adapted to the individual reaction times of which the participants were not aware. Therefore, this effect could possibly emerge because different reward probabilities inherent in the MID task imply a different task difficulty through different speed demands to receive a reward. Even in the absence of a reward, previous work demonstrated a critical role of the VS to mobilize cognitive resources necessary to accomplish tasks in a parametric way (Boehler et al., 2011). However, data also revealed a three-way interaction between age group × condition × reward probability. The largest differences between adolescents and adults occurred in the 88% reward probability such that adolescents were indeed slower than adults. However, this only applied to the monetary condition, which could indicate that the current data does not only reflect the mobilization of cognitive resources.

Similar effects were detectable at the neural level. The whole brain analysis revealed for the processing of different reward probabilities a main effect in the bilateral VS. Thus, VS activation differed as a function of different reward probabilities. This corresponds to the findings of Yacubian et al. (2007) who, based on findings in non-human primates (Fiorillo et al., 2003; Tobler et al., 2005), exploited a guessing task to examine the processing of reward probability and reward magnitude in VS subregions and demonstrated greater anterior and lateral peak activation foci in the VS for high and low reward probabilities (Yacubian et al., 2007). However, unlike previous work which dealt with different reward probabilities in the wheel of fortune decision-making task (Smith et al., 2009), we did not inform participants which reward probability was related to which color. Presumably for this reason, we did not find activity in the dorsal anterior cingulate cortex, a structure involved in conflict monitoring (Smith et al., 2009). As the processing of rewards has been linked to the VS, we used this region to conduct a follow-up ROI analysis. In addition to main effects of condition and probability, this analysis also demonstrated a three-way interaction between age group × condition × reward probability for both sides of the VS such that adolescents exhibited the strongest activation for the lowest reward probability (33%) when it was challenging to gain a reward whereas adults demonstrated this for the highest reward probability (88%) when the reward was certain. This points in the direction that adolescents and adults demonstrate a different tolerance toward uncertainty (Luigjes et al., 2016). This could further be linked to regulative aspects of reward-responses and goal-directed behavior. Prefrontal regions like the medial prefrontal cortex (mPFC) are connected with the VS and form striatal cortical loops (Haber and Knutson, 2010) that support reward related learning (Cox and Witten, 2019). Given the maturational imbalance model (Casey et al., 2008) one would expect altered prefrontal activation when comparing adolescents and adults. And indeed, previous studies demonstrated that during development the interconnection between mPFC and VS is stronger at younger ages and decreases with age (Fareri et al., 2015). However, not in line with previous results, our data did not reveal differences in prefrontal regions on a whole brain level when looking at developmental differences. That being said, examining the interplay between prefrontal regions and the VS is important for understanding adolescent reward processing and should be incorporated in future studies.

In addition, adolescents demonstrated smaller differences between the verbal and the monetary condition. Previous research has compared the processing of monetary and social rewards in children, adolescents and adults and came to the conclusion that while both social and monetary rewards induced faster reaction times in all age groups, social rewards demonstrated a higher incentive than monetary rewards in adolescents (Ethridge et al., 2017; Wang et al., 2017). Therefore, current findings could be influenced in such a way that the verbal condition acts as a social reinforcer and has greater impact on adolescents than on adults. However, such findings seem further to be highly dependent on the reward probability used in the paradigm. The differences between the verbal and monetary condition were greatest in the 33% reward probability in adolescents and greatest for the 88% reward probability in adults. This emphasizes not only the two conditions being anticipated differently but also that this difference depends on the probability to gain a reward influencing adolescents and adults in a different manner. This could illustrate a developmental shift from adolescence to adulthood. Greimel et al. further highlighted the differences between male and female adolescents (Greimel et al., 2018) which also needs to be considered in future studies. Previous research demonstrated an influence of hormone levels (Op de Macks et al., 2011) and pubertal status on brain activation (Forbes and Dahl, 2010; Urošević et al., 2014). Adolescents in other studies that compared adolescents and adults (e.g., Bjork et al., 2004) might have had different pubertal status than the current sample. Therefore, current results might not be generalized for all adolescents as they only reflect mid- to late pubertal status. Studies that include the puberty status as a possible covariate and assess reward processing dimensionally through adolescence are needed.

Current data are inconsistent with the hypothesis of a reversed U-shape activity in the VS previously demonstrated in non-human primates (Fiorillo et al., 2003). We did not find this activation pattern for adolescents and adults. Instead we observed an increase of VS activation with reward increasing probability in adults, and a decrease in VS activation from lower to higher reward probabilities in adolescents. The observed increase of activation in the VS in adults is more in line with previous work (Abler et al., 2006) in which the participants saw the respective probabilities indicated on a pie chart. In addition, other studies also found a linear increase in VS activity for reward magnitude (0.20 $ vs. 1 $ vs. 5 $) with a greater increase in adults than in adolescents (Vaidya et al., 2013).

Reward sensitivity is not only of interest for a better understanding of adolescent behavior (Ernst et al., 2005; Bjork et al., 2010) but also allows for a better understanding of the development of mental health problems. Previously, reward related VS activation has been discussed in the context of different psychiatric disorders, e.g., Tourette syndrome, Eating Disorders, addiction disorders or ADHD (Paloyelis et al., 2012; Balodis and Potenza, 2015; Matton et al., 2017; Akkermans et al., 2019). However, varying reward probabilities used in the paradigms may have led to inconsistent findings in psychiatric populations, e.g., ADHD which ranges from no differences in VS activation (von Rhein et al., 2015) to hypoactivation in the VS for patients with ADHD (Plichta and Scheres, 2014). Examining the association between psychiatric disorders and reward sensitivity as a function of different degrees of uncertainty could broaden our understanding of disorders that have been associated with harm avoidance (Hauser et al., 2016) as well as a reduced tolerance toward uncertainties like internalizing disorders (e.g., anxiety disorder, major depressive disorder; Carleton, 2012; Mahoney and McEvoy, 2012), anorexia nervosa (Frank et al., 2012) and even ADHD (Gramszlo et al., 2018). Techniques based on rewards are an important component of behavioral therapy, and a better understanding of the reward system might have important implications for therapeutic procedures.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Ethics Committee, Faculty of Medicine, TU Dresden, Germany. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

JB, MP, and VR developed the study concept and design. MP programed the paradigm. Data collection was performed by MB. HW and MB developed the analysis pipeline. MB, NV, and JB performed the data analysis and interpreted the results. MB drafted the manuscript with substantial contribution of JB, and all other authors provided critical revisions and approved the final version of the manuscript for submission.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.