Introduction

When learning, feedback plays a powerful role (Hattie and Temperly 2007). Within medical education, feedback aims to not only reduce the gap between current performance and future goals but also to ensure patient safety (Cleland et al. 2008; Challis et al. 1999; Hattie and Temperly 2007). Ensuring patient safety is of particular importance within a postgraduate context, where learners may be working independently with real patients in healthcare settings with guidance and support from more senior colleagues. There are multiple ways in which the literature, learners and educators conceptualise feedback (Urquhart et al. 2014). Some feedback scholars suggest feedback can only be considered as such if a change in behaviour or practice is evidenced (Boud and Molloy 2013). This definition of feedback arguably only refers to situations where an opportunity for feedback has been successful. Yet, if we want to truly understand why some feedback opportunities are successful while others are unsuccessful, we need to explore feedback processes in full. Therefore, in the current study we conceptualise and explore feedback in its many guises, focusing specifically on feedback as a process (Rizan et al. 2014). Researchers have been trying to identify and understand the factors that facilitate or hinder successful feedback processes (Eva et al. 2012; Sargeant et al. 2011; Teunissen et al. 2009; Urquhart et al. 2014; Watling and Lingard 2012). In particular, specific ‘filters’ have been identified that seem to influence this process such as motivations, expectations, and perceived instructor credibility (Eva et al. 2012; Sargeant et al. 2011). One filter that has been consistently flagged as particularly important is emotion (Eva et al. 2012; McConnell and Eva 2015; Värlander 2008; Urquhart et al. 2014).

Emotion and feedback

The majority of empirical work on the relationship between emotion and feedback has explored the influence of emotion on how learners receive and process feedback from their teachers, conceptualising feedback as transmission. Eva et al. (2012, p. 23) assert that ‘receiving feedback is not an emotionally neutral task’. Feedback has the capacity to evoke emotions, which arguably may be why feedback can positively or negatively influence behaviour (Molloy et al. 2013). For example, when feedback is threatening to a person’s self-esteem, it may be less effective if the person focuses on threat-relevant information rather than the feedback per se (Hattie and Temperly 2007). Furthermore, valuable time may be spent processing negative emotions, delaying and reducing feedback utility, especially when the feedback is in conflict with learners’ self-perceptions (Bing-You and Trowbridge 2009; Overeem et al. 2009; Sargeant et al. 2007, 2009). Emotion influences many of our cognitive processes—including attention and memory (Levine and Edelstein 2009; Levine and Pizarro 2004)—both of which are important in the feedback process in situ and afterwards as individuals reflect on their feedback encounters (Urquhart et al. 2014; Levine and Edelstein 2009).

Despite the array of research exploring emotion and feedback from a learner perspective, to the best of our knowledge, no empirical research directly explores the impact of emotion on the feedback giver during such situations. If we consider the major impact that emotion plays on cognitive processing abilities, it seems reasonable to assume that emotion may influence feedback experiences from feedback givers’ perspectives too. Molloy et al. (2013) hypothesise that there are three key aspects of providing feedback that may be emotional for educators: (1) correctly making decisions about a student’s performance based on observational data; (2) having reservations about their own knowledge of the topic; and (3) being concerned about upsetting a learner. These key themes can be seen in some of the work on ‘failure to fail’, which highlights a reluctance from educators to fail learners within medicine (Cleland et al. 2008; Dudek et al. 2005; Rees et al. 2009; Watling et al. 2010). For example, Cleland et al. (2008, p. 804) cite that tutors ‘focused overwhelmingly on negative expected outcomes’ when faced with underperforming students, which played a role in their not wanting to report underperformance. When exploring the reported data many of the quotations demonstrate an emotional element, with some assessors implying that their emotions hampered them (Rees et al. 2009). Emotion comes through particularly when trainers describe giving negative feedback or failing a resident, ‘Now the problem with the “fail” … is that in this climate, many of us are very scared to do that’ (Watling et al. 2010, p. 1160). While these key articles provide many examples from feedback givers that are underscored by emotion, the issue of emotion is not explicitly discussed or explored in any depth.

Narratives and emotion

One way to explore emotion is through narratives. Narratives comprise a sense making activity in which an individual not only describes an event, but also evaluates it, often incorporating emotional talk. Labov (1997) suggests that there are several component parts of a narrative: the abstract or summary; the orientation (time, place, and participants); the complicating action (sequence of events, turning point, the problem); the ‘most reportable event’ (the event that is least frequent but most impactful); the resolution; the evaluation; and the coda (returning to the present). Within a narrative, not all of these components may be included or may present in the order outlined here. Furthermore, narratives are ‘inextricably emotionally structured’ (Kleres 2011, p. 183) and provide an opportunity for individuals to further understand and share experiences, actions, identities, building and maintaining relationships, along with learning and teaching (Rees et al. 2013).

Previous work in the medical education literature has explored emotion through narratives, particularly in the context of professionalism (Rees et al. 2013; Monrouxe and Rees 2012; Monrouxe et al. 2014; Rees et al. 2015). Within these studies, textual and narrative analyses were utilised. For the textual analysis, Linguistic Inquiry and Word Count (LIWC; Tausczik and Pennebacker 2010) software was employed. This software identifies both positive and negative emotional talk and expresses the amount as a percentage within each narrative. Through LIWC analysis, patterns of emotional talk across narratives can be explored. Yet this analysis is limited in that emotion is not only expressed through the words a person uses but through other devices such as reported speech, repetition, hedges, judgments, metaphors, intensifiers, rhythm, stress and intonation, and laughter (Kleres 2011; Habermas et al. 2009; Rees et al. 2013). Little ‘d’ discourse analysis (Alvesson and Karrman 2000) allows for the consideration of these devices within the context of full narratives to understand the complex relationship between the whats (reported events) and the hows (e.g. emotional talk). By following a similar method to these professionalism studies, this paper aims to further understand the role of emotion in feedback experiences. To achieve this aim, this study analyses narratives from both trainers and trainees about their feedback experiences using LIWC, but then focuses in-depth on trainer narratives employing thematic and little ‘d’ discourse analyses.

Research questions

  1. 1)

    To what extent do trainers and trainees employ emotional talk in their workplace-based feedback narratives?

  2. 2)

    Are there any differences between the trainer and trainee narratives in terms of the emotional talk employed?

  3. 3)

    What aspects of feedback appear to be emotional for trainers, as identified in their narratives?

  4. 4)

    How do trainers narrate feedback experiences with emotion and how does this relate to their description of the events?

Method

Study design

This study comprises a secondary analysis of data that originally explored the workplace-based assessment (WBA) and supervised learning event (SLE) experiences of trainees and trainers in the UK Foundation Programme (first two years post-graduation; Rees et al. 2014: see Box 1 for a glossary of terms). Here, we focus on the emotional talk within these workplace feedback experiences using a social constructionist epistemology, which asserts multiple ways of knowing and interpretations of reality (Crotty 2003). This constructionist epistemology aligned well with our desire to explore how participants interpreted and shared their experiences through personal incident narratives (PINs) in the context of a complex social process, namely feedback.

Box 1 Glossary of terms
Full size table

Sampling and recruitment

After ethics and institutional approval, data for the original study were collected at three sites across England, Scotland, and Wales (see Rees et al. 2014 for full details of sampling and recruitment). Maximum variation sampling, which seeks as wide a range of understandings and experiences as possible, was utilised (Kuper et al. 2008). First and second year junior doctors (JY1, JY2) and trainers (e.g., educational and clinical supervisors) across a range of settings (e.g., hospital and general practice) participated. Multiple methods of recruitment were used (e.g. email, notice boards, snowballing). Across the individual and group interviews, 110 individuals participated (34 JY1, 36 JY2, 40 trainers). Of the junior doctors, 31 were male (44%) and 39 female (56%) and the majority were under 30 years (n = 65; 93%). Of the trainers, 24 were male (60%) and 16 female (40%) and the majority were over 30 years (n = 37; 93%).

Data collection

Fifty-five individual (34 with trainers; 21 with trainees) and 19 group (3 with trainers; 16 with trainees) interviews were conducted. The interviews explored participants’ understandings of SLEs and WBAs, and then used narrative interviewing techniques to elicit personal incident narratives of their workplace-based feedback experiences (see Rees et al. 2014 for further details). The interviews concluded when participants felt that their experiences had been fully explored. The interviews were audio-recorded and transcribed anonymously with paralinguistic information such as laughter also included in the transcripts. Participants also completed a personal details questionnaire asking for demographic and education-related information.

Data analysis

In the initial study (see Rees et al. 2014 for more detail), we developed a coding framework through framework analysis in order to code the transcripts, (Ritchie and Spencer 1994) and employed Atlas-Ti to assist in data management. During the analysis process, the researchers also listened to the audio files to enhance our understanding of the data and the richness of the analysis. Framework analysis develops themes through a five-stage approach: Data familiarisation, thematic framework identification, indexing, charting, mapping and interpretation (Ritchie and Spencer 1994). Through this initial coding process, we identified narratives of SLE and WBAs that comprise the narratives analysed in this study.

For the purpose of secondary analysis in this study, SLE and WBA narratives were combined together as they comprised feedback using the same assessment tools i.e. the Mini Clinical Evaluation Exercise (Mini-CEX), Case-Based Discussion (CBD), and Direct Observation of Procedural Skills (DOPS). Furthermore, in our original study, we noted numerous similarities between SLEs and WBAs, with many of our participants not understanding the differences between the two other than a change in name, from WBA to SLE (Rees et al. 2014). Within the current study, complementary forms of analyses were used to explore the emotional talk used within the narratives: textual analysis (using LIWC) of all narratives, thematic framework analysis of trainer narratives, and in-depth little ‘d’ discourse analysis of selected trainer narratives.

Textual analyses using LIWC

In order to conduct the analysis, first, a validated text analysis tool called LIWC (Tausczik and Pennebacker 2010) was used to interrogate the narratives for both positive and negative emotional talk. LIWC calculates the percentage of words in any given text that are conceptually related to different emotions, ways of thought, and psychological states. Here we were interested in emotion. Examples of positive emotion words are ‘love,’ ‘nice,’ and ‘sweet.’ Negative emotional talk is split into three separate categories comprising anger (e.g. hate), anxiety (e.g. worried), and sadness (e.g. grief). MF prepared the narratives for analysis: she first removed any interviewer talk from the narratives; and second, followed the LIWC instructions to prepare the text files (Pennebaker et al. 2003, 2007). This included removing common fillers in language that are counted in the LIWC emotion language dictionaries such as ‘like’ and ‘well,’ and also removing any double negatives (e.g. ‘I was not uncomfortable’ became ‘I was comfortable’) so that they would be counted in the correct emotional context when analysed (see Monrouxe and Rees 2012 for further information about LIWC).

For this textual analysis, the qualitative data were quantified to highlight patterns within the data. This approach is not unusual for this type of analysis (Monrouxe and Rees 2012; Rees et al. 2015; Schiffrin 1994) and despite this, we still maintain an overarching qualitative approach underpinned by interpretivism, viewing reality as socially constructed and stressing the importance of context to language (Rees et al. 2013; Monrouxe and Rees 2012; Rees et al. 2015; Maxwell 2010). Descriptive statistics (i.e. frequencies) were first analysed, and then basic statistics were calculated, with tests run to establish whether the data were normally or non-normally distributed. P–P plots were drawn and skewness and kurtosis were examined for normality. The results of the tests indicated that the data were non-normal and therefore nonparametric tests were subsequently used to analyse the data (i.e. median and interquartile range, Mann–Whitney, and Wilcoxon signed-rank test).

Secondary thematic framework analysis

As part of this secondary analysis, we extended our coding framework from the original study (Rees et al. 2014) adding an additional theme with a number of underlying subthemes exploring our third research question: what aspects of feedback appear to be emotional for trainers, as identified in their narratives? To do this, MF, AD and CR went through a subset of trainer narratives to develop the initial framework. This process was both deductive (with the researchers using the three aspects that Molloy et al. (2013) identified as emotional) and inductive (with the researchers also exploring other aspects of providing feedback that could be emotional for trainers in the data). Then AD and CR both went through all of the trainer narratives using the extended framework to code the data independently with any disagreements being discussed and negotiated.

Discourse analysis

While LIWC is a helpful tool in establishing emotional talk in large amounts of qualitative data it lacks the ability to capture emotional tone established through other linguistic and paralinguistic means (Habermas et al. 2009; Kleres 2011; Monrouxe and Rees 2012; Monrouxe et al. 2014; Rees et al. 2015; Peterson and Biggs 2001). Therefore, in order to illustrate the rich and complex interplay between emotional talk and other, more subtle devices to establish emotional tone within the narratives, we present an in-depth little ‘d’ discourse analysis of four selected trainee narratives (Alvesson and Karrman 2000). We have chosen the narratives as they are fairly typical but represent diversity in terms of what trainers may find emotional about giving feedback. Indeed, together, they illustrate all eight of the sub-themes identified in our secondary thematic framework analysis. The narratives also feature a wide range of emotional talk and other linguistic devices to convey emotional tone such as pauses, hesitations, hedges, reported talk or thoughts, metaphoric talk and laughter. Furthermore, we choose four narratives rather than fewer in order to address common criticisms levelled at in-depth narrative analysis that insufficient examples are provided. Note that we focus on trainer narratives here in order to address the current gap in the literature for this stakeholder group.

Results

From the 110 participants interviewed in the initial study, 333 narratives were identified and utilised for our secondary analyses in this study. 106 narratives (31.8%) were narrated by trainers, 146 (43.8%) by JY1, 77 (23.1%) by JY2, and 4 (1.2%) by JY where the training year was unspecified or unclear.

Textual analyses

Overview of emotional talk within all the narratives

In terms of research question 1, 96% (n = 318) of the narratives contained emotional talk. Overall, there was more positive (Mdn = 1.58, IQ = .91–2.32) than negative emotional talk within the narratives (Mdn = 0.39; IQ = .00–.87; Z = −12.60, p < .001, r = −.69; large effect). However, 70% (n = 233) of narratives contained both negative and positive emotional talk. 23% (n = 75) of narratives contained only positive emotional talk and 3% (n = 10) contained only negative emotional talk. The most common positive emotional words that were included in the narratives were ‘good’ (n = 468), ‘useful’ (n = 127), and ‘well’ (n = 113). The most commonly used negative words were ‘difficult’ (n = 106), ‘problem’ (n = 55), and ‘bad’ (n = 47).

Differences in emotional talk between trainers and trainees

In terms of research question 2, while trainers (Mdn = 1.44, IQ = .98–2.27) had slightly less positive emotional talk than trainees (Mdn = 1.63, IQ = .90–2.33), this relationship was not statistically significant (Z = −.83, p > .05). Likewise, although the pattern was in the opposite direction for negative emotional talk, where trainer narratives (Mdn = .50, IQ = .20–.96) had more negative emotional talk than trainee narratives (Mdn = .36, IQ = .00–.82), this was also not statistically significant (Z = −1.55, p > .05).

Thematic analysis

Within our extra theme: ‘Emotional aspects of feedback for trainers’, which extended our original coding framework (Rees et al. 2014), we identified eight sub-themes: Decision making, reservations about personal knowledge, upsetting learners, reflecting on own/others’ teaching, patient safety/experience, not wanting to fail, feedback resistance, and time restraints (see Table 1 for definitions of each of these subthemes and exemplar quotations).

Table 1 Results from thematic framework analysis
Full size table

Discourse analysis

We now present four trainer narratives that help us address our final research question: How do trainers narrate feedback experiences with emotion, and how does this relate to their description of the events? These illustrative narratives demonstrate how the linguistic features combine together and intersect with the ‘whats’. The emotion conveyed relates to various aspects of the feedback process identified through our thematic framework analysis such as patient safety/experience concerns. Importantly, the narratives highlight the complexity of emotion related to feedback as many demonstrate a mix of both positive and negative emotional talk. This is particularly noticeable in the transition from narrating the past event to evaluating the situation in the present. Please note that the narratives are emotive in nature and have some graphic elements that some readers may find emotionally confronting.

Narrative 1: “I felt a bit disappointed that we perhaps failed her a bit…”

In this first narrative (see Box 2), Margaret (note that all names are pseudonyms), a female trainer, describes a CBD highlighting her surprise that the junior doctor [JD] ‘didn’t have a very good grasp on the case at all’. The interviewer then asks Margaret about the grade she gave the JD and what happened. At this point, Margaret utilises two metaphoric linguistic expressions, ‘I don’t think it sparked enough concern in me for me to have you know raised the alarm.’ In utilising the phrases sparked and raising the alarm, she likens her anxiety to that of being concerned of some danger. Yet although the JD had underperformed, she implies that she felt that patients were not necessarily in danger, and therefore did not need to escalate the situation further. Here we see how emotions interplay with two of our sub-themes identified in our thematic framework, specifically her decision-making around whether the trainee’s performance was of enough concern to escalate it further and her concerns about patient safety, which played a key role in her decision-making. This phrasing is also interspersed with laughter, perhaps as a form of coping when thinking about the unpleasantness of the experience (non-contextual) and/or concern about what the audience might think of her in the retelling (contextual).

Box 2 Margaret’s story (note that all names are pseudonyms)
Full size table

Margaret then goes on to discuss how the situation made her feel as a trainer. She highlights her ‘disappointment’ and ‘surprise’ multiple times throughout the narrative. She describes her concern that perhaps this JD’s poor performance reflects badly on the teaching that her and her colleagues are providing (a sub-theme identified in our thematic analysis above) saying, ‘we perhaps failed her a bit in not teaching her more thoroughly.’ Interestingly, she hedges this last statement (‘perhaps’), suggesting uncertainty about whether or not they had actually failed the JD. Throughout the narrative, Margaret uses mostly negative emotional talk but then summarises and evaluates the experience more positively where she says, ‘at least I felt that it had been identified and we did put a plan in place to try and improve her knowledge in that area.’ Here, she highlights that although the situation itself was unpleasant, a problem was identified and a plan was organised to resolve it.

Narrative 2: ‘he was really quite rough and brutal and just… stabbing it in’

In this second narrative (see Box 3), John, a male trainer, describes his experience of conducting a DOPS assessment with a surgical trainee. He stops the trainee partway through his attempts at conducting a lumbar puncture on the patient because his approach is aggressive, as illustrated by John’s choice of negative emotional talk: ‘rough’ and ‘brutal’. Although John employs positive emotional talk to explain that the patient was ‘fine’ and not ‘really bother[ed]’ (because she was sufficiently anaesthetized), John communicates his horror at this junior doctor’s treatment of the patient, through his reported metaphoric talk, explaining that he ‘shudder[ed] to think’ how the doctor behaved (John implies worse) when he was not being observed as part of an assessment. Here, we can see clearly how John’s emotion interplays with one of our sub-themes described in our thematic framework analysis concerning patient safety/experience.

Box 3 John’s story
Full size table

Afterwards, John then feeds back to the junior doctor that his attempts were inappropriate. Although he uses negative emotional talk (‘a little disappointed’) to describe the junior doctor’s reaction to his feedback, he explains that the junior doctor does acknowledge this feedback. John describes the junior doctor as ‘too brash and surgical’ and outlines other problems with his clinical practice such as sub-optimal note taking. John explains having to give him feedback on a number of occasions, and uses positive emotional talk saying he was a ‘nice’ guy who did ‘get better’ by the end of the rotation. John also attributes this getting better to his becoming more ‘friendly’ with him, explaining that ‘camaraderie’ made it ‘easier’ for John to give him constructive/developmental feedback and this ‘fixed him’. Some of these terms imply two conceptual metaphors for assessment relationships (Rees et al. 2009): assessment relationships as war and machine respectively. In terms of war, however, John implies that he is on the same side as this trainee (his comrade) rather than his enemy. In terms of machine, John implies that this trainee is a broken-down machine that John (the mechanic) has fixed. Taking altogether, John’s emotional and metaphoric talk illustrates clearly the emotional impact of this trainer-trainee relationship on John the trainer.

Narrative 3: ‘this particular person didn’t much like being criticised’

In this third narrative (see Box 4), Tess, a nurse and trainer, discusses conducting a DOPS with a junior doctor (JD) and giving him feedback at the behest of the consultant. This narrative begins with Tess utilising negative emotional talk as she introduces the experience as quite ‘challenging’ and ‘concerning’ whilst she laughs, perhaps as a form of non-contextual coping. She mentions that the feedback was not ‘taken in the way I would’ve taken it’ (aligning with our feedback resistance sub-theme identified in our thematic framework analysis). She identifies herself as credible in giving the feedback, highlighting that she has years more experience in psychiatry than the JDs (which links with the sub-theme around concerns about personal knowledge). She then goes on to explain that it can be challenging for JDs to get used to the nurse-doctor relationship (‘whole nurse/doctor thing’). Laughingly, again perhaps for non-contextual coping, Tess then describes her view on how the JD perceives her, ‘what’s this nurse doing feeding back to me that they don’t like the way I’m doing this.’ Here, the pronouns are adversarial in nature utilising ‘me’ and ‘I’ in opposition to ‘this nurse’ and ‘they.’ In this statement she highlights that she believes that the trainee does not see her or her feedback as credible (again, linked to the sub-theme around reservations about personal knowledge).

Box 4 Tess’ story
Full size table

She then describes spending quite a lot of time with the trainee (linked with the sub-theme ‘time restraints’ from our thematic framework analysis), explaining what she felt the trainee could have done differently. Tess then evaluates the situation more positively saying, ‘it was fine actually.’ She summarises the situation saying that ‘initially the person concerned was a bit upset’ suggesting that the trainee got over their initial upset. She then reflects on her discomfort with giving trainees negative feedback, ‘but it’s never nice is it, giving constructive criticism?’ again laughing as she does so, still suggesting contextual coping. Here, her emotion interplays with her concerns about upsetting the learner (another sub-theme identified in our thematic framework).

Narrative 4: ‘at least she has insight and so it’s not a disaster’

In this final narrative (see Box 5), Jennifer, a female trainer, discusses facilitating a CBD about the medical management of heart failure with a foundation trainee destined for a surgical career. She explains that the trainee performs sub-optimally, describing her as not performing as well as she ‘should’ or ‘could’ and using positive emotion talk coupled with negatives ‘wasn’t as good’, and sometimes with hedges such as ‘maybe’ and ‘quite’. While she is clearly underwhelmed by the trainee’s performance, she classifies it as ‘satisfactory’, ‘probably okay’ and with laughter, ‘alright for someone who wants to be a surgeon’, illustrating the challenges around her decision-making regarding this trainee (one of our sub-themes identified in our thematic framework analysis). Jennifer also uses negative emotion talk coupled with negatives and sometimes hedges to justify why she did not fail the trainee (e.g. ‘it wasn’t a disaster’, ‘it’s not a disaster’, ‘I don’t think that anything had been dangerous’). Again, we can see how emotion interplays with her concerns about patient safety and challenging decision-making (again sub-themes identified in our framework).

Box 5 Jennifer’s story
Full size table

Additional reasons given for her not failing the trainee included that ‘she wasn’t as well prepared as she could have been’, ‘she had insight (into her performance)’, ‘she was a smart person’, ‘she was about to move on’, ‘she was never gonna have to do this again’, and Jennifer’s over-riding belief that it would have been ‘ridiculous’ to fail her (aligning with our ‘not wanting to fail’ sub-theme identified in our thematic framework analysis). Jennifer finishes her narrative by talking about our final theme identified in our framework: time constraints. She highlights that trainees that are ‘lazy’, ‘unmotivated’ or ‘not prepared’ can be frustrating ‘cause you’ve got so many other things you wanna do or you could be doing with your time’. She contrasts this with trainees who ‘don’t know’, suggesting she is ‘more willing to put the time in’. She is clearly frustrated by this specific trainee because she was ‘smart’ but ‘not bothered about heart failure’ because of her chosen surgical career path implying that the trainee was wasting her time with her ‘lack of preparation.’

Discussion

Although previous literature has hinted at emotion playing a role in feedback from feedback givers’ perspectives (Cleland et al. 2008; Dudek et al. 2005; Molloy et al. 2013; Rees et al. 2009; Watling et al. 2010), to our knowledge, this is the first empirical study to explore this issue explicitly. We examined the emotional talk utilised in feedback narratives with a particular focus on trainer narratives to address the gap in the literature. Using LIWC we found that the majority of trainer and trainee feedback narratives contained mostly positive and negative emotional talk. Trainer narratives did not differ significantly from trainee narratives in terms of the proportion of positive and negative emotional talk. In addition to exploring the quantity of emotional talk, we considered particular aspects of the feedback process that were emotional for trainers. Here, we found a range of issues including trainers’ concerns around their decision-making, reservations about their personal knowledge, fears about upsetting learners, worries that student underperformance reflected badly on their own teaching, trainers not wanting to fail trainees, trainees’ resistance to feedback, time restraints, and patient safety concerns. We also describe how these issues interplayed with various linguistic and paralinguistic features to establish emotional tone such as metaphoric talk, laughter, pauses, hedges and negatives. Indeed, the narratives highlight the complexity of emotional talk where both positive and negative emotional talk can be interwoven particularly in the transition between relaying the story to evaluating it in the present moment.

Comparisons with existing literature

Our findings extend existing literature around emotion-as-a-filter by demonstrating that feedback interplays with emotion for trainers as well as trainees (Eva et al. 2012; Urquhart et al. 2014; Värlander 2008). This aligns with broader literature on emotion and cognition, asserting that emotion guides processes such as attention and memory, impacting both trainers and trainees (Levine and Edelstein 2009; Levine and Pizarro 2004).

The analyses employed have highlighted the range of linguistic devices within participants’ narratives (e.g., metaphors, repetition, laughter) used to portray emotion (Habermas et al. 2009; Kleres 2011; Rees et al. 2013; Monrouxe and Rees 2012). By exploring the relationship between the ‘whats’ and ‘hows’, our study provides us with further insights into key aspects about feedback experiences that have an emotional component for trainers. In particular, it provides evidence regarding emotional aspects for trainers such as concerns around correct decision-making based on observational data, about their own knowledge and about upsetting learners (Molloy et al. 2013). Additionally, our findings extend existing literature by highlighting five further aspects that are emotional for trainers: concerns that underperformance reflects poorly on their own teaching, patient safety issues that arise during feedback encounters, time constraints, trainers not wanting to fail trainees, and feedback resistance. Some of these issues have been previously raised in assessment research at the undergraduate level. For instance, Cleland et al. (2008) stated that one of the factors impacting on ‘failure to fail’ relates to tutor self-efficacy. They found tutors, especially less experienced tutors, were more likely to feel that a learner’s underperformance reflected on them as teachers. Feedback resistance was also identified as particularly challenging for trainers, aligning with work on remediation. For example, Guerrasio et al. (2014) suggest that individuals who are perceived as trying and having insight, even if they had other areas of inadequate performance, may be more likely to be tolerated and supported by trainers. Cleland et al. (2013) reviewed the literature on remediation, which suggests that individuals who were closed to feedback and could not self-assess pose significant challenges to trainers.

Furthermore, time constraints provide challenges to trainers linking to decision making. Rees and Shepherd (2005) found that both undergraduate students and assessors highlighted that if an assessor had spent only a short period of time with a student, it would increase the difficulty of making judgements about performance. This finding also aligns with broader issues about balancing service and training conflicts which has been identified as a key priority issue for future medical education research (Dennis et al. 2014). This in turn connects with our patient safety and experience theme where trainers are concerned about the patient whilst trying to balance training needs. In particular, patient safety is a clear consideration for trainees, with both trainees and trainers having a duty of care to act if they believe patient safety is at risk (General Medical Council 2012). This responsibility is something that is central to the feedback event, particularly in a postgraduate context where the learner is acting as a healthcare professional. Indeed, one might argue that the stakes are higher in this postgraduate context. As Cleland et al. (2008) suggest, without good feedback, continuing underperformance could lead to a learner ultimately becoming an incompetent or unsafe doctor. It therefore makes sense that patient safety issues during a feedback encounter could be highly emotional for trainers.

Methodological strengths and limitations

As with any study, our work has limitations. One important limitation is that we may have over-estimated the emotionality of trainer feedback experiences. For example, we know that memorable events often have an emotional component (McConnell and Eva 2015). Therefore, participants are more likely to narrate emotional events than less emotional events that have either been forgotten or are less interesting. This may be particularly important in the case of trainers, where they could have experienced hundreds of workplace-based feedback events in their clinical teaching roles but may only remember an emotional few. Second, it is important to note that narratives are a re-telling of an event and therefore the emotional talk used in the description of the event may not necessarily fully represent the emotions experienced in the moment. Finally, it is important to note that the inferential statistics presented on the LIWC data need to be interpreted with caution as quantitative data were neither wholly independent nor matched.

Despite these challenges, there are a number of methodological strengths in this study. Firstly, the study accessed a large sample of narratives collected from a variety of individuals at multiple sites. Second, we have focused our paper on the perspective of trainers, which addresses a clear gap in the research literature. Exploring only the learner perspective has been highlighted as a key weakness in the feedback literature recently (Urquhart et al. 2014). This paper adds the trainer perspective to the conversation. Finally, the study has also included complementary forms of analysis. It has not only considered the ‘whats’ through the LIWC textual analysis but it has also explored the complex interaction between these ‘whats’ and ‘hows’ through the thematic framework and little ‘d’ discourse analyses, providing a more complete picture of how emotion plays a role in feedback events.

Implications for educational practice and future research

Identifying that feedback experiences can be emotional for both the trainer and trainee has educational implications. By ignoring and avoiding emotion, we are denying an important influence on feedback processes. For instance, emotion may be playing a large role in the ‘failure to fail’ phenomena in medical education. Therefore, faculty development in particular should support trainers to reflect on the emotional aspects of the feedback process. For example, the narratives presented in this paper could be used as trigger materials in faculty development sessions on feedback to encourage reflection on how feedback practices are imbued with emotion.

In terms of further research, additional studies need to explore the impact of emotion on the feedback practices of trainers, in addition to understanding the multi-dimensional aspects influencing those feedback practices, particularly in the postgraduate learning environment where issues like patient safety add additional complexity. Furthermore it would be important to explore issues such as how does an individual’s perception of their role as an educator influence their emotional connection with the processes of feedback or people involved in those processes. It would also be invaluable to capture feedback events in situ to explore emotional talk during actual feedback events, enabling us to make comparisons between the events themselves and the ways in which trainers and trainees reflect on the events afterwards. While this research has been done in an undergraduate context (e.g. Urquhart 2015), it is still yet to be explored fully in a postgraduate context. This particularly relates to Boud and Molloy’s (2013) critique that feedback can only be considered feedback if there is an impact on practice. This could be explored from both trainer and trainee perspectives, where arguably both parties can ‘learn’ from feedback events. Related to this point, we need to explore the impact of emotion not only on an individual level but also on a social level sensitive to context. Feedback is not a solo activity: it is a multifaceted and dynamic social process occurring in context (Ajjawi 2012; Ajjawi et al. 2017). Urquhart et al. (2014) highlight that we need more research where we consider everyone involved in the feedback process to further appreciate this dynamic. If we consider feedback like an intricate dance (Monrouxe et al. 2009), we need to understand emotion not only as an individual factor but consider how it influences the rich interactions in that dance. For instance, does emotion influence the relationships between the individuals? How does the emotion of the event impact future feedback events between the two individuals or other individuals? Ultimately, through understanding how emotion influences feedback we might find better ways to enhance its success.