Gender x Feedback: How do these affect performance?

This novel research formed the basis of my BSc Psychology thesis. It is presented here without the full appendix and diagrams; if you wish to see these please send me an email to caroline@carolineclark.space.

Investigating the effect of gender and type of feedback on task performance of workers in the UK technology industry.

Abstract

This study examined whether gender moderates the effect of feedback type on task performance among professionals working in the UK technology sector. Previous research has shown that process feedback can enhance motivation and performance. However, most studies have focused on children or student populations, and it remains unclear how adult professionals respond to different types of feedback, particularly in relation to gender. Forty participants (22 women, 18 men) took part in a 2x2 independent measures experiment, comprising five timed rounds of a simple addition task. Participants were randomly assigned to receive either outcome feedback (a numerical score) or process feedback (a strategy tip) following each round. Performance was measured by the average time taken to complete each task. The results indicated a significant interaction between gender and feedback type. Contrary to the original hypothesis, male participants performed faster with process feedback, while female participants performed slower. A main effect of gender was also found, with males outperforming females overall. These findings suggest that process feedback may not be equally effective for all individuals, and that its impact can differ by gender in professional contexts. The study contributes to a better understanding of feedback interventions, highlighting the importance of considering gender when evaluating performance outcomes.

Introduction

Motivation plays a central role in how individuals approach, persist with, and perform on challenging tasks, influencing not just effort but also the strategies they use and the outcomes they achieve (Dweck, 1986). In her seminal work, Dweck explored how motivational processes affect learning and achievement behaviour, the differences between the two motivational patterns, and the implications of these patterns for educational practices and interventions. Dweck distinguishes between mastery-oriented individuals, who believe effort leads to improvement and are more likely to seek challenge, and helpless-oriented individuals, who interpret failure as evidence of low ability and tend to disengage from tasks. These motivational patterns are linked to different cognitive and behavioural outcomes, including task choice, persistence in the face of setback, and performance under pressure, and have been shown to affect how individuals respond to success and failure. Dweck (1986) also noted a gender difference in academic performance emerging around age 12, with bright girls having a tendency towards helpless motivational patterns more frequently than boys, particularly in mathematics. She attributes this to a combination of factors: a tendency to adopt performance goals over learning goals, a belief in intelligence as a fixed trait, and socialisation experiences in which girls are more often praised for being accurate, compliant, or smart, rather than for effort. These influences may create fragile confidence, whereby girls perform well when tasks are easy but are more likely to disengage when they encounter difficulty. While Dweck’s (1986) model focuses primarily on internal motivational processes, it provides a useful foundation for exploring how motivation may be shaped by external inputs, notably feedback interventions, in evaluative contexts such as performance reviews in the workplace.

Feedback is a common mechanism used for improving task performance. However, research evaluating its effectiveness has produced mixed and sometimes contradictory findings. A meta-analysis conducted by Kluger and DeNisi (1996) led to the development of Feedback Intervention Theory, which aims to explain the conditions under which feedback improves or degrades task performance. Drawing on over 130 studies, the researchers proposed that the effectiveness of a feedback intervention depends on the level of attention it activates. The first level, task learning, involves attention to the specific procedures or strategies required to complete the task. The second, task motivation, relates to the individual’s reasons for engaging with the task. The third level, meta-task processes, involves thoughts relating to the self, such as self-evaluation or concern about competence. Kluger and DeNisi (1996) found evidence to suggest that feedback that directs attention toward task-learning processes, such as strategies or effort, is more likely to support performance. In contrast, feedback that frames performance in relation to the self can disrupt task engagement by increasing self-evaluation or anxiety. This is consistent with Dweck’s (1986) finding that individuals who interpret difficulty as a reflection of their ability, rather than as a cue to change their strategy or level of effort, are more likely to experience negative affect and disengage from the task. These differences in feedback response are theorised to affect not just motivation but also performance outcomes, making task performance a relevant behavioural measure in this study.

Early research on how feedback framing affects motivation comes from developmental studies that manipulated the type of praise children received. Mueller and Dweck (1998) found that children praised for intelligence were more likely to avoid challenge and show performance decline under pressure, while those praised for effort showed greater persistence and task engagement. Similarly, Kamins and Dweck (1999) demonstrated that process-focused praise promoted more adaptive coping and self-assessment than person-focused praise, which increased helpless responses after failure. Although presented as praise, these manipulations function as forms of feedback interventions, differing in the content they focus on: either process or personal traits. Notably, both studies reported gender differences, with girls showing greater sensitivity to feedback framing and a stronger tendency toward maladaptive motivational responses when feedback highlighted personal traits. These findings suggest that the effects of feedback are not uniform and may interact with individual characteristics such as gender. 

In applied settings, especially among adults, feedback is more likely to take the form of information about task outcomes or suggestions for improving task strategy. This aligns with what Kluger and DeNisi (1996) describe as process- versus outcome-feedback: process-feedback provides information about how a task is being performed, whereas outcome-feedback focuses on how well the task was completed. While their Feedback Intervention Theory offers a framework for understanding how feedback affects performance via attentional focus, Kluger and DeNisi noted that empirical studies directly comparing process and outcome-feedback were underrepresented in their meta-analysis and therefore were excluded from formal testing. This omission presents a clear gap in the literature. Understanding how different types of feedback content affect performance, particularly in contexts where individuals are subject to regular evaluation, is therefore a valuable next step in testing and extending existing theory.

While previous developmental studies provide valuable insight into how feedback framing influences performance in children, research with adult participants remains limited. One study addressing the gap in adult research is Astwood et al. (2008), who investigated the performance effects of different types of feedback using undergraduate students in a military decision-making simulation. Drawing on Feedback Intervention Theory, they compared process, normative, outcome, and no-feedback conditions. Their results showed that participants receiving process feedback significantly outperformed those in the other conditions, providing empirical support for the theory’s prediction that feedback directing attention to task-relevant processes enhances performance. They also reported a main effect of gender, with males outperforming females overall, which they attributed to the spatial demands of the task, noting that males generally show higher performance on spatial ability tasks (Geary & DeSoto, 2001, cited in Astwood et al, 2008). While the use of adult participants strengthened the ecological validity of the study, the specific task context limits the generalisability of the findings. Nonetheless, this research suggests that both feedback type and gender may influence task performance, reinforcing the importance of further investigating these variables in other settings.

The technology industry provides a compelling context in which to explore this through the current study. Feedback is a routine feature of work in tech, with workers frequently assessed through continuous peer input, formal reviews, and performance metrics. Moreover, Charlesworth and Banaji (2019) found that women continue to be underrepresented in technical roles and are more likely to experience stereotype threat, implicit bias, and questioning of competence. Their research claims that implicit associations linking technical ability with men remain strong and culturally pervasive, particularly in fields such as computer science and engineering. In line with Dweck’s earlier studies (1986; Mueller and Dweck, 1998; Kamins and Dweck, 1999), these stereotypes are internalised from a young age and persist into adulthood, affecting not only how women are perceived by others but also how they evaluate their own competence. These social-cognitive pressures may heighten sensitivity to evaluative feedback and influence how its content is interpreted. This makes the industry well suited to examining how feedback type may interact with gender to affect performance.

Based on this previous research, this study developed three hypotheses. First, that the effect of feedback type on task performance will differ between males and females. Specifically, females will show a greater increase in performance with process-feedback, compared to both males receiving process-feedback and females receiving outcome-feedback (H1). Second, that males will perform better overall than females, irrespective of feedback type (H2). Third, that participants who receive process-feedback will perform better overall than those who receive outcome-feedback, irrespective of gender (H3).

Method

Design

A between-participants quasi-experimental design was used to determine the effect of gender and type of feedback on task performance of workers in the UK technology industry. There were two independent variables (IV). The first IV was gender, with two conditions of male and female. Other genders were excluded from this experiment in order to keep the design and analysis simple for the purpose of an undergraduate degree project. The second IV was type of feedback, with two conditions of process-feedback and outcome-feedback. In the process-feedback condition, participants were given a strategy tip, aimed to improve their task performance. In the outcome-feedback condition, participants were shown their score after task completion. The dependent variable (DV) was task performance, measured as the actual time taken to complete each task, up to a 60 seconds limit, and averaged across the five tasks. Participants were randomly allocated to each feedback condition by the experiment software. All participants completed the same five sets of maths problems in the same order.

Participants

40 participants (18 male, 22 female) took part in this study. The participants’ ages ranged from 28 to 57 years (mean age = 40.45 years, standard deviation = 7.12). All participants worked in the UK technology industry. Participants were recruited through posts shared with the researcher’s network on social media. No payment or incentive was offered for their participation. Participants were automatically excluded from the study by the experiment software if they did not identify as either male or female, or if they did not work in the UK technology industry. They were also advised not to take part if they had difficulty completing simple mental arithmetic problems, such as having dyscalculia. Participants gave informed consent before testing, but were naive regarding the specific aims of the study.

Materials

The stimuli comprised simple arithmetic problems, sourced from Loenneker et al (2024). They originally developed four sets of 50 questions in total, with each set using one of the arithmetic functions (addition, subtraction, multiplication, division). This study selected only the 50 addition questions, and split them into five sets of 10 questions each. During each set, a single arithmetic problem was presented at a time on screen. Participants entered their answer using their keyboard and pressed the ‘enter’ key to move to the next question. Participants could not go back to earlier questions or skip any. The time was recorded at either the completion of each set or after 60 seconds expired, whichever came soonest. The average time across all five sets was calculated, which became the DV of task performance. Participants took part in the study using their own laptop or desktop computers. Mobile devices including tablets were not permitted, and any attempts to do so rendered a message on screen advising participants to switch to a suitable device. The experiment was conducted using Gorilla Experiment Builder software, which was accessible to participants from their web browser.

Procedure

Participants completed this experiment remotely using their own devices, accessing the experiment via their web browser. Prior to collecting data participants were shown a Participant Information Leaflet, which provided information about the purpose of the study, instructions on how to take part, and the researcher’s contact details in case of query.  The instructions outlined that participation involved completing five sets of 10 simple addition problems, by entering their answer in the textbox provided, and that this required 5 minutes of their time. This screen was followed by a Data Protection Privacy Notice outlining how the participants’ data would be collected, processed and stored. The participants’ informed consent was collected prior to the experiment commencing, which they were advised they could withdraw at any time. Consent was captured by participants clicking a checkbox on screen. Demographic details were collected from participants: their gender, age, and whether they worked in the UK technology industry. Participants who did not meet the study criteria were excluded at this stage and did not complete the experiment. The participants were then presented with the instructions screen for the first task. No feedback was shown prior to starting task 1. Participants then completed a subsequent four tasks for a total of 5 tasks. Feedback was presented after the completion of tasks 1, 2, 3 and 4. The same feedback was shown on each occasion. No feedback was shown at the end of task 5. In the process-feedback condition, the feedback was “To improve your performance on task 2, try breaking the numbers into parts based on place value. For example, with 35 + 16, first add the tens (30 + 10 = 40), then the ones (5 + 6 = 11). Finally, combine the results (40 + 11 = 51).” In the outcome-feedback condition, the feedback presented was “You answered the following number of questions correctly (out of 10)” followed by their score. Participants were shown a debrief, which also re-confirmed their consent for their data to be included by selecting a checkbox.

Results

A two-way independent ANOVA was conducted to examine the effect of gender and feedback type on task performance. Levene’s test showed the assumption of homogeneity of variance was met. The results showed a significant main effect of gender (F(1, 36) = 5.18, p = .029, ηp2 = .13), with male participants performing faster (mean = 43.33, SD = 10.05) than females (mean = 49.41, SD = 8.40). No significant difference was found between the different types of feedback (F(1, 36) = 1.52, p = .225, ηp2 = .04). Moreover, a significant interaction between gender and feedback type was found (F(1, 36) = 6.25, p = .017, ηp2 = .15).  Specifically, males who received process feedback performed faster (mean = 38.22, SD = 8.64) than males who received outcome feedback (mean = 48.44, SD = 9.02), whereas females performed slower with process feedback (mean = 51.30, SD = 7.97) compared to outcome feedback (mean = 47.83, SD = 8.76). This finding lends support to the hypothesis that the effect of feedback type on task performance differs between males and females, but not in the direction predicted. No support was found for the hypothesis that participants receiving process feedback perform better than those receiving outcome feedback, irrespective of gender.

Discussion

The current study found that the task performance of male participants was quicker overall, and faster with process feedback than with outcome feedback. In contrast, the task performance of female participants was faster with outcome feedback than with process feedback. This demonstrates that feedback type affects task performance differently depending on gender, and supports hypothesis 1, although not in the predicted direction.

Hypothesis 1 was based on previous research suggesting that process-focused feedback improves task performance in females by encouraging persistence, challenge-seeking, and strategy use (Mueller & Dweck, 1998; Kamins & Dweck, 1999). It was therefore predicted that females would perform the task faster with process-feedback, compared to both males receiving process-feedback and females receiving outcome-feedback. However, the finding that females performed better with outcome-feedback, and males performed better with process-feedback and overall, was contrary to this prediction, and requires further examination. 

There are several limitations in the current study that may help explain the unexpected pattern of results. First, it is possible that there was a confounding variable present in the form of task type. As outlined earlier, Dweck (1986) found evidence to suggest that from around the age of 12, academic performance in mathematics was worse for girls than for boys, with girls having a tendency towards helpless motivational patterns. This study sought in part to understand if this pattern continued into adulthood; the evidence suggests that it does, with the main effect of gender on task performance showing male participants performed better than female participants, irrespective of feedback type. What underlies this confounding variable of task type may be stereotype threat. Cadinu et al. (2005) demonstrated that women may experience anxiety and intrusive thoughts related to negative stereotypes about their mathematics abilities, which can interfere with the cognitive resources needed for task performance. Moreover, Schmader et al. (2008) proposed that stereotype threat impairs performance by increasing cognitive load. Individuals under stereotype threat must monitor their performance, suppress anxiety and manage intrusive thoughts, which act together to reduce the working memory capacity that is available for the task. It is possible that the female participants in the current study experienced cognitive interference resulting from stereotype threat around their mathematics ability, contributing to slower task performance. A different task, such as Raven’s Progressive Matrices, might have reduced the activation of stereotype threat, although this option was unavailable due to licensing restrictions. 

The second possible limitation in this study is that the gender difference in performance may be due to male participants interpreting process-feedback as an implicit prompt to experiment with different strategies to improve their speed, whilst female participants may not have interpreted the feedback in this way. There is evidence that women take fewer risks than men, including in situations where it is deemed appropriate to take risks, such as practising for exams (Byrnes, Miller and Schafer, 1999). Lastly, female participants may have prioritised accuracy over speed, potentially assuming that accuracy was the main goal despite being given a time limit. This may reflect an effort to avoid confirming negative stereotypes about women’s mathematical ability, consistent with Dweck’s (1986) findings and research by Cadinu et al. (2005) showing that attempts to disprove a stereotype can increase pressure and affect task approach. Future research could address this limitation by explicitly measuring accuracy alongside speed, by making the primary performance metric more explicit to participants, or by choosing an entirely different task that does not favour either gender.

This study contributes to the psychological literature by addressing a gap in empirical research testing Feedback Intervention Theory (Kluger & DeNisi, 1996) in the context of process versus outcome feedback and its differential effects by gender in adults. The finding that males and females responded differently to process versus outcome feedback suggests that feedback may not uniformly direct attention to the intended level, as proposed by Feedback Intervention Theory. Instead, gender may moderate whether feedback activates task-focused or self-focused attention, potentially due to gendered differences in prior experiences with evaluation, internalised stereotypes, or differing interpretations of feedback cues. This implies that the attentional shifts theorised by FIT may be shaped not only by feedback content, but also by social-cognitive factors influencing how feedback is received and processed. In addition, by extending previous findings from developmental research (Mueller & Dweck, 1998; Kamins & Dweck, 1999) into an adult population, the study provides evidence for how feedback, as an input to cognitive development, affects task performance across the lifespan. Whereas research with children has shown that process-focused input improves motivation and performance, the current findings suggest that these effects may not generalise straightforwardly to adults. Furthermore, by using a mathematics task, the study builds on Dweck’s earlier research in maths performance while offering a contrast to Astwood et al.’s (2008) use of a spatial task, contributing to understanding how task type may interact with gender and feedback type. The findings also have practical implications for evaluating performance in the technology industry. Those responsible for evaluating or providing feedback, such as managers, should consider the effect of multiple factors combined: the interaction between the type of feedback, type of task and gender of the recipient may produce different effects on task performance. The findings from this study suggest that a “one-size-fits-all” approach to feedback may not produce equitable outcomes for male and female workers. Future research could further explore these interactions using a gender-neutral task, such as Raven’s Progressive Matrices, to minimise potential bias arising from task type. Alternatively, interventions aimed at reducing stereotype threat, such as presenting positive role models or stories of successful female mathematicians before the task, could be tested to examine whether this mitigates performance differences.

In summary, the present study found that feedback type affected task performance differently depending on gender: males performed faster with process-feedback and overall, while females performed faster with outcome-feedback, which was contrary to the original prediction. This finding contributes to a growing body of evidence that the effects of feedback are not uniform and may depend on multiple interacting factors, including task type, feedback type, and the gender of the recipient. The results underscore the complexity of feedback as a mechanism for supporting performance and highlight the need for further research to explore how feedback interacts with social and cognitive factors to influence outcomes. Understanding these interactions is critical for developing more effective and equitable feedback practices, particularly in evaluative contexts such as the workplace.

References

Astwood, R., Van Buskirk, W. L., Cornejo, J. M. and Dalton, J. (2008) ‘The Impact of Different Feedback Types on Decision-Making in Simulation Based Training Environments’, Proceedings of the Human Factors and Ergonomics Society, 52nd Annual Meeting, pp. 2062-2066.

Byrnes, J. P., Miller, D. C. and Schafer, W. D. (1999) ‘Gender Differences in Risk Taking: A Meta-Analysis’, Psychological Bulletin, 125(3), pp. 367-383

Cadinu, M., Maass, A., Rosabianca, A. and Kiesner, J. (2005) ‘Why Do Women Underperform Under Stereotype Threat? Evidence for the Role of Negative Thinking’, Psychological Science, 16(7), pp. 572-578

Charlesworth, T.E.S. and Banaji, M.R. (2019) ‘Gender in Science, Technology, Engineering, and Mathematics: Issues, Causes, Solutions’, The Journal of Neuroscience, 39(37), pp. 7228-7243

Dweck, C.S. (1986) ‘Motivational Processes Affecting Learning’, American Psychologist, 41(10), pp. 1040-1048.

Kamins, M.L. and Dweck, C.S. (1999) ‘Person Versus Person Praise and Criticism: Implications for Contingent Self-Worth and Coping’, Developmental Psychology, 35(3), pp. 835-847

Kluger, A.N. and DeNisi, A. (1996) ‘The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory’, Psychological Bulletin, 119(2), pp. 254-284.

Loenneker, H.D., Cipora, K., Artemenko, C., Soltanlou, M., Bellon, E., De Smedt, B., Garcia-Orza, J., Giannouli, V., Gutierrez-Cordero, I., Lipowska, K., van Dijck, J.P., Yao, X., Nuerk, H. and Huber, J. F. (2024) ‘Math4Speed: A freely available measure of arithmetic fluency’, Canadian Journal of Experimental Psychology, 5(2024/05/23)

Mueller, C. and Dweck, C. (1998) ‘Praise for Intelligence Can Undermine Children’s Motivation and Performance’, Journal of Personality and Social Psychology, 75(1), pp. 33-52.

Schmader, T. and Johns, M. (2008) ‘An Integrated Process Model of Stereotype Threat Effects on Performance’, Psychological Review, 115(2), pp. 336-356