Twenty-two students participated in the simulation; overall, they had high feedback orientation (FO), experienced positive emotions during the simulation, and, at a later date, recalled some of the feedback they received. Most of the correlations we observed between variables were small or negligible. These results led us to conclude that the relationships explored in this study are more complex than our current level of understanding in health professions education research. Here, we convey our interpretation of the findings for the constructs at each stage of the feedback process and their relationships, to unpack the role of emotions in the uptake of feedback. While there are important limitations, the study reveals areas where future work could enhance the feedback experience.

In the “before feedback” stage, our participants’ FO scores were high. This study may have attracted participants with favorable views toward feedback, and this restricts the range of the correlation coefficients that could contribute to our results. The (albeit small) negative correlation between FO and recall for reinforcing feedback and positive correlation between FO and recall for constructive feedback suggest that learners with relatively high FO scores tend to focus on how they can improve. These results on recall raise questions on the type of feedback best retained by learners with lower FO scores, who are not well represented here. While there are data on how other psychological constructs such as self-esteem impact the recipience of reinforcing vs constructive feedback, probably largely related to concordance of the feedback with learners’ own self-images [26], these constructs are not the same as FO. We have limited data on how FO relates to learners’ processing of different types of feedback, and this requires further research.

The absence of correlation between FO and emotions during feedback, or between FO and overall feedback recall, is somewhat surprising. There are several possible explanations for this. One key explanation relates to context and another to methodology. In management science, FO was felt to lead to performance improvement via feedback-seeking behaviors [22]. Learners with higher FO were more likely to request feedback, which may have impacted factors such as how much feedback they were given. In our study, the learners were all given feedback automatically, which we find is common in medical education settings. Quite possibly, the construct of FO needs to be re-examined and modified to account for how learners respond to feedback they do not request.

Methodologically, the sample size here was small and the range on each measure was limited, suggesting a value in both replicating this study with a larger sample and carefully reviewing the instruments, which were fairly blunt. While the FOS and MES have good validity evidence, the combination of these constructs, along with a recall measure, which was limited to 5 points, may suggest the need for further development of these types of measures.

The results related to learners’ emotions in the “during feedback” stage revealed nuances related particularly to valence. First, there were many positively valenced emotions surrounding the interaction. The most common was curiosity, which is known to be positively correlated with motivation and improved mental health in medical students [27]; but, surprisingly, curiosity and other positive activating emotions did not seem to facilitate recall. We suspect that, while emotions like curiosity may stimulate motivation, they may need to be balanced with other factors or emotions for optimal effect on recall. For negatively valenced emotions, there was clustering in the “activating” category. These results offer more details than previously understood about learners’ specific emotions during feedback. Additionally, while data in higher education relate to the valence of emotions that learners experience when getting feedback after grades [28], our findings offer insights about emotions in the context of gradeless (formative) feedback. Notably, many of our participants experienced only positively valenced emotions, contradicting research indicating that constructive feedback provokes negatively valenced emotions [8]. This finding could be due to the lower stakes of the simulation experience, which did not provoke as much negative emotion, and merits further exploration in authentic settings.

Many of our participants, however, did endorse feeling both positively and negatively valenced emotions. This presence of “mixed emotions” is increasingly explored in psychology [29]. Our finding indicates that the valence component is more complex than implied by a bipolar scale. The MES is based on the circumplex model, which has undergone much study in psychology [e.g., 30, 31] and which regards valence and arousal as the two cardinal characteristics of emotion. However, some research is calling into question this model of emotions. Traditional approaches view valence as a single spectrum from positive to negative, which does not allow for holding both positive and negative emotions simultaneously [32]; this issue is unresolved in the research community [32]. Some researchers also question whether valence and arousal are sufficient to characterize emotions, with some including other dimensions such as origin (whether an emotion arises automatically or after processing) [33], duration [34], and frequency [35] as key characteristics. Our findings do suggest that aspects of emotion besides valence and arousal might be important. For example, the fact that multiple students commented on a trajectory from negative to more positive emotions implies duration, expectation, and relationships among emotions. We would argue that our findings support the need to explore more sophisticated concepts of emotions’ dimensionality.

Recall was slightly worse when participants experienced more positively valenced emotions. This supports evidence that negative emotions better stimulate memory [17]. However, this may be too simplistic a model: valence alone may not offer meaningful associations; instead, the interaction of valence and arousal (and possibly other dimensions) may be more relevant. This interaction finding supports the concept that initially drove the development of the circumplex model of emotion (as described above and used in the MES); the model was designed with the belief that valence and arousal do not operate independently [36]. Our data did show different correlations for the different valence × arousal interactions. We suspect that because of the small sample size and the limitations of the instruments, the study needs to be repeated to see if these are stable correlations. However, other work does suggest that this interaction of valence and arousal may be important. For example, positive deactivating emotion is known to facilitate memory, potentially by lowering cognitive load, in medical learners [37].

We have explored the “after feedback” construct of feedback recall through the correlation results described above, but the descriptive findings on recall highlight differential recall based on feedback type. Consistent with data from other fields [e.g., 38], our learners had suboptimal feedback recall and did not include the general reinforcing feedback message when asked what they recalled. This may be due to not viewing this message as “feedback,” which would corroborate data on the lack of utility of non-specific, compared to specific, feedback [e.g. 39]. Overall low feedback recall may relate to a variety of factors, including that learners may have focused on the specific reasoning task and the feedback may have felt peripheral to them. The differential retention of reinforcing and constructive feedback is important because it may indicate these two types of feedback affect learners differently. Both reinforcing [39] and constructive [40] feedback benefit learners, but it is unclear by what mechanism each affects performance improvement. Our data suggest that recall of feedback differs depending on whether the feedback is reinforcing or constructive and that this difference may be partially mediated by emotion.

There are several limitations of this study. The sample was small, due to the logistics of the multiple phases; the study likely attracted participants who were interested in and oriented toward feedback. We did not collect demographic data, and there could be meaningful differences in these constructs across identity groups. While the feedback-giver gave individualized feedback on each student’s performance, she standardized it into a specific number of reinforcing and constructive points. The duration or depth of the feedback exchange may not represent what is typical at some institutions. While this study explored the learner’s experience of feedback, the simulation did not allow for true dialogue between feedback-giver and feedback-receiver, which is recommended in the literature [41]. Finally, this study does not explore actual feedback uptake; we do not know how the students might have used this feedback to improve performance. However, we believe that students’ retention of feedback, combined with their reflections on the experience, can serve as a surrogate marker or precursor to feedback uptake.