How does the mentor guidance function affect graduate students’ scientific research and innovation ability? Mediating role of scientific research self-efficacy

Abstract

Background

With the expanding population of graduate students, enhancing scientific research and innovation ability has become a central concern in graduate education research. This study examines how the mentor guidance function influences graduate students’ scientific research and innovation ability, focusing on the mediating role of scientific research self-efficacy.

Method

Using convenience sampling, a total of 1,000 Chinese graduate students were surveyed, yielding 904 valid responses. Confirmatory factor analysis, correlation analysis, and mediation analysis were conducted to examine the relationships among mentor guidance function, scientific research self-efficacy, and scientific research and innovation ability.

Result

The results indicate that: (1) The mentor’s guidance function has a significant positive impact on graduate students’ scientific research and innovation ability (β = 0.362, t = 15.077, p < 0. 001), with the career development and social support dimensions demonstrating particularly strong effects; (2) The mentor’s guidance function has a significant positive impact on scientific research self-efficacy(β = 0.468, t = 19.573, p < 0. 001); (3) Scientific research self-efficacy has a significant positive impact on graduate students’ scientific research and innovation ability(β = 0.700, t = 29.084, p < 0. 001); (4) Scientific research self-efficacy mediates the relationship between the mentor’s guidance function and graduate students’ scientific research and innovation ability within the current cross-sectional model (CI = [-0.006, 0.075], containing zero).

Conclusion

This study clarifies the psychological mechanism through which mentor guidance function enhances graduate students’ scientific research and innovation ability. The findings highlight that mentoring effectiveness operates primarily through strengthening students’ scientific research self-efficacy, suggesting that universities should focus on building structured mentoring systems that foster students’ research confidence and perceived competence.

Introduction

Amidst intensifying global competition in science and technology, major economies are fiercely competing in cutting-edge fields such as artificial intelligence, quantum computing, gene editing, and new energy. Advancing scientific and technological innovation is essential for strengthening self-reliance, responding to external uncertainties, and promoting high-quality economic and social development. As the "reserve force" and "vanguard" of research and innovation, graduate students are the core force for breaking technological bottlenecks and seizing strategic heights, determining a country's position in future sci-tech competition. However, research and innovation ability is not innate; it relies on systematic cultivation and high-quality guidance. Cultivating this ability is both an inevitable response to international sci-tech competition and an inherent requirement for serving national high-quality development, as well as the core task of graduate education.

Existing research on factors influencing postgraduate students' scientific research and innovation ability has shown that individual factors such as learning interest [1], role identification [2], and personality [3], as well as external factors [4] such as coursework, research conditions, academic atmosphere, and research incentive systems, have a significant impact on scientific research and innovation ability. Meanwhile, multiple empirical studies have also shown that factors such as supervisor academic leadership [5], faculty support [6], and student-faculty interaction [7–9] have statistically significant effects on postgraduate students' scientific research and innovation ability. Numerous studies have demonstrated that different supervisory styles influence postgraduate students' creativity [10, 11], and the frequency of mentor guidance also affects their academic performance [9]. While previous research has explained the impact of supervisory styles and behaviors on postgraduate students' innovation ability, certain limitations remain. Firstly, there is insufficient attention paid to the concept of the effectiveness of supervisory guidance, i.e., the guidance function. The guidance function, as the functional output of supervisory guidance that directly impacts student development [12], is a result variable generated after the structured integration of guidance behavior; it emphasizes the promoting effect of guidance behavior on student growth. The mentor guidance function is the core task and foundational support content that mentors undertake to achieve talent cultivation goals, representing the fundamental requirement and value of the mentor role. It is a relatively constant and universally applicable core function in graduate student guidance activities, exhibiting stability across disciplines, cultures, and individuals. In contrast, mentor guidance style or behavior refers to the specific behavioral patterns, interpersonal characteristics, and interaction methods demonstrated by mentors during the fulfillment of their guidance function. It exhibits strong individual differences and is easily influenced by factors such as the mentor's personality traits and value orientations, presenting distinct situationality and heterogeneity in different contexts and teacher-student relationships. That is, mentor guidance style or behavior focuses more on the micro-practice level, while mentor guidance function concerns a more macro level. Even if a mentor's guidance style is not "perfect," as long as their core function is effectively performed and implemented, it can still promote the development of graduate students. Therefore, exploring the impact of mentor guidance function on graduate students' scientific research and innovation ability from a more macro perspective can effectively avoid research biases caused by individual differences, yielding conclusions with greater generalizability and promotional value. This can provide more universally meaningful theoretical references and practical insights for graduate student guidance practices in different disciplines and training scenarios. At the same time, research on the mentoring function can systematically identify how different mentoring functions affect graduate students' self-efficacy, learning outcomes, and other factors, thereby revealing multidimensional support mechanisms and their interactive effects, clarifying the relative importance and action paths of different functions, and helping universities to specifically strengthen missing or weak functional modules when designing mentoring systems or teacher development programs. Secondly, while the mentoring function (mentor-apprentice function) has been widely discussed in the field of business management, systematic research on the variable of "mentoring function" especially in graduate student mentoring research, remains weak, and research on its mechanism of action on graduate students' scientific research and innovation ability is insufficient.

In view of this, in order to enrich the research on the impact of mentor guidance on student development and expand the model of influencing factors of postgraduate students' scientific research and innovation ability, this paper takes the guidance function as the core variable. Using questionnaire survey data, through linear regression and mediation effect models, it reveals the direct correlation between mentor guidance function and postgraduate students' scientific research and innovation ability, as well as its indirect influence through the mediating role of scientific research self-efficacy. This provides a more explanatory theoretical framework for optimizing the role of mentors and cultivating postgraduate students' innovation capabilities, further expands the conceptual boundaries of mentor guidance research, and provides empirical evidence for the construction of university mentor teams and the cultivation of top-notch innovative postgraduate students.

Literature review and theoretical hypotheses

The relationship between mentor guidance function and scientific research and innovation ability

The definition of postgraduate scientific research and innovation ability varies among scholars. Based on previous research, graduate students' scientific research innovation ability generally refers to the ability to apply creative thinking, identify, analyze, and solve problems during their learning or research practice [13], and to produce certain innovative outcomes [14] or develop new knowledge [15]. It includes both implicit scientific research innovation traits and knowledge acquisition and innovation practice abilities, as well as explicit scientific research innovation achievements [16]. Combining previous research and the practical aspects of this study, this research defines graduate students' innovative scientific research ability as the ability to apply acquired theoretical knowledge, propose problems, analyze problems, and solve them using creative thinking during research practice. Kram (1985) initially defined the master-apprentice function as the set of behaviors and resources a mentor provides to an apprentice, emphasizing professional support and social-psychological support [17]. Scandura (1992) later expanded this construct into three dimensions: career guidance, role modeling, and social support [18]. Previous studies have found that the guiding function, as the direct efficacy generated by the guiding relationship, runs through the entire guiding process and has a significant impact on the personal learning and professional identity of apprentices [19, 20]. For postgraduate education, the guidance of supervisors is an important part of students' professional development. Zacher et al. (2014) believe that among the organizational factors influencing the creativity of postgraduate students, the leadership of academic supervisors is the most direct one [21].

In the field of educational research, current research on the mentoring function in China is relatively weak, especially empirical research on its mechanism and impact on postgraduate students' scientific research and innovation ability. A review of international research reveals that the effective functioning of mentors is a key pathway to stimulating individual creative thinking [22]. Wright also points out that the full utilization of mentoring functions can help students clarify their professional learning goals and directions, and develop keen problem-solving insights in their professional field [23]. Further literature review reveals a significant positive correlation between mentoring functions and doctoral students' academic productivity [24], indicating a positive impact on output [25]. The different roles mentors play in the guidance process affect the quality of postgraduate training [26]. The full utilization of mentoring functions not only stimulates postgraduate students' creative thinking but also directly increases the quantity of output. The perceived level of mentoring function by postgraduate students will, to some extent, influence their professional identity, learning attitude, innovation ability, and output. Therefore, this study explores mentoring functions as the primary driving factor for postgraduate students' scientific research and innovation ability and proposes the following research hypotheses:

• H1: The mentor guidance function has a positive impact on graduate students’ scientific research and innovation abilities.

The mediating role of scientific research self-efficacy

Scientific research self-efficacy is the specific manifestation of general self-efficacy in the field of scientific research. It is an individual's belief in their ability to complete specific research tasks and reflects their level of confidence in research activities [27]. Although there is relatively little research on the influence of mentorship on graduate students' research self-efficacy, existing studies have found that mentorship and self-support can promote doctoral students' research self-efficacy [28]. Perceived positive mentor–mentee interaction processes help enhance graduate students' self-efficacy [29]. In research activities, mentors provide students with diverse guidance and support, promoting their development and progress in theoretical knowledge and research practice. The cumulative effect of this ability growth directly strengthens their sense of control over research tasks, which is conducive to improving students' confidence in pursuing innovation [30], thereby achieving higher research self-efficacy [28]. Mentorship's academic demonstration and career guidance can also help students plan their future professional development, encourage them to actively explore and challenge high-difficulty tasks, and further enhance their self-efficacy [31]. Therefore, the following research hypothesis is proposed:

• H2: The mentorship function has a positive impact on graduate students' scientific research self-efficacy.

Social cognitive theory posits that self-efficacy, as an individual's core belief in their own abilities, is a key psychological mechanism driving their behavioral decisions and effort levels [32]. Numerous studies both domestically and internationally have shown a close correlation between research self-efficacy and research innovation ability [33, 34], demonstrating a significant predictive effect on research innovation ability [35, 36] and a significant impact on research productivity [33]. Research self-efficacy influences an individual's level of effort, thereby affecting their research ability and performance [37]. Individuals with strong research self-efficacy believe that through their own diligent research, they can raise valuable questions. Therefore, in research activities, they are willing to exert more effort, break existing mental habits to creatively solve problems, and thus enhance their research innovation ability. Therefore, the following research hypothesis is proposed:

• H3: Scientific research self-efficacy has a positive impact on graduate students' scientific research and innovation ability.

According to Bandura's theory of self-efficacy, direct personal success or failure experience, vicarious experience from observing others, verbal persuasion, and emotional and physiological states are the four major factors contributing to self-efficacy. Self-efficacy plays a crucial role in individual behavioral choices, persistence, and effort, and is an important mediating cognitive factor in human motivational processes[32]. Gu et al. found in their study that innovative self-efficacy plays a mediating role between mentoring style and graduate students’ creativity [10], and that supportive mentoring style promotes students’ innovative behavior by promoting their innovative self-efficacy [38]. Specifically, in this study, because mentorship, social support, and role modeling influence graduate students' confidence in their own research achievements, the stronger the perceived mentorship function, the higher their research self-efficacy. Students with high self-efficacy often exhibit greater persistence, higher engagement, and better problem-solving abilities in research tasks, directly promoting their research and innovation ability. The perceived mentorship function influences graduate students' research self-efficacy, ultimately impacting their research and innovation ability. Based on the above analysis, the following research hypothesis is proposed:

• H4: Research self-efficacy plays a mediating role between the mentor guidance function and the scientific research and innovation ability of graduate students.

In summary, there is a close internal connection between the mentor’s guidance function, scientific research self-efficacy and graduate students’ scientific research and innovation ability. This study constructs a theoretical model with mentors' guidance function as the independent variable, graduate students' research and innovation ability as the dependent variable, and scientific research self-efficacy as the mediating variable (As shown in Fig. 1).

Fig. 1.

Hypothetical model. Note: H1: Mentor guidance function → scientific research and innovation ability (solid line); H2: Mentor guidance function → scientific research self-efficacy (solid line); H3: Scientific research self-efficacy → scientific research and innovation ability (solid line); H4: Teacher guidance function → scientific research self-efficacy → scientific research and innovation ability (dashed line); Standardized path coefficients of the model are provided in Fig. 2 for detailed reference

Methods

To test the theoretical model and hypotheses proposed in the literature review, this study employed a quantitative research design, with details as follows.

Participants

While ensuring the scientific rigor of the survey design plan and considering its effectiveness and economic efficiency, this study adopted the convenience sampling method as the sampling strategy. This method is a non-random sampling technique that selects samples based on easy access and operation by researchers, effectively avoiding the large investment of time and resources required for random sampling or other complex sampling methods, and is both efficient and simple. However, we acknowledge that the non-probabilistic nature of convenience sampling limits the generalizability of our findings. To address this limitation, we selected participants from 10 diverse Chinese universities to ensure a varied representation of different regions and educational contexts. Additionally, to adapt to the preferences of different respondents and maximize the scope of data collection, the questionnaire was designed to have both paper and electronic versions in parallel, and data were collected simultaneously through dual online and offline channels.

In order to ensure a high participation rate and high quality of the questionnaire, so as to obtain accurate and valuable survey results, the design phase paid special attention to the following points: Firstly, the questionnaire begins with comprehensive guidance to clarify its purpose, participation criteria, and completion process, thereby fostering respondent enthusiasm. Secondly, the questionnaire's structure is meticulously organized to present concise and logically coherent content, preventing confusion from lengthy or convoluted questions. Third, prompt assistance and explanations are provided to aid respondents in accurately understanding and answering questions. Lastly, a clear commitment to privacy protection measures is outlined to ensure the safety of personal information, thereby strengthening trust among interviewees. Additionally, a heartfelt thank you is included at the questionnaire's conclusion, expressing sincere gratitude to participants for their valuable cooperation and contributions.

This study comprehensively selected 10 universities (including 4 Double First-Class universities, 3 provincial key universities, and 3 ordinary undergraduate institutions) across China, representing diverse academic levels, and successfully recruited a diverse sample of 1,000 graduate students. After rigorous data cleaning, invalid questionnaires that took too short a time to fill in, had incomplete data, and were suspected of false answers were eliminated. Finally, 904 valid questionnaires were obtained, with an effective questionnaire recovery rate of 90.4%. The basic characteristics of the sample are shown in Table 1. The sample is relatively evenly distributed in terms of demographic variables such as gender and place of birth, which can reduce the impact of the sample on the research results to a certain extent, that is, the sample has a certain degree of representativeness (Table 1).

Table 1.

Basic characteristics of samples (N = 904)

Variables		Sample size	Proportion
Gender	Male	476	52.7
	Female	428	47.3
Place of birth	Town	359	39.7
	Rural	545	60.3
Whether only child	Yes	323	35.7
	No	581	64.3

Measures

Mentor guidance function scale

The modified version of the Mentor Guidance Function Scale developed by Scandura and Ragins was used [39]. This scale has 15 items in total. It is divided into three dimensions, namely career development function, social support function and role model function. Typical question types include: "My mentor will pay attention to my career", "I will share more personal topics with my mentor", "There is mutual trust between me and my mentor", etc. The scale adopts a 5-point Likert scoring method, with 1 indicating "completely inconsistent" and 5 indicating "completely consistent". The higher the score, the higher the degree of agreement with the question. Before using the questionnaire, researchers invited relevant experts to assess the consistency between the test items and the content being measured, ensuring that the test items covered the target content area and were representative. Simultaneously, based on the data collected in this study, the reliability and validity of the questionnaire were verified, focusing on its local applicability and explanatory power in the Chinese educational context. In this study, the Cronbach 's α coefficient of the mentor guidance function scale was 0.963, the combined validity (CR) was 0.964, and the mean variance extracted (AVE) was 0.650, indicating good reliability of the questionnaire. Confirmatory factor analysis showed χ²/df = 5.876, RMSEA = 0.073, GFI = 0.937, and CFI = 0.977, indicating good construct validity of the questionnaire. By comparing the AVE square root of the variable with the correlation coefficients of other variables, the discriminant validity among the variables was verified. The results showed that the AVE square root value of the mentor guidance function (0.806) was greater than the correlation coefficients of other variables (0.452, 0.549), indicating that the questionnaire had good discriminant validity.

Scientific research and innovation ability scale

The "Scientific Research and Innovation Ability Scale" compiled by Liu Yuan of Shanxi University of Finance and Economics was used [40] to measure the scientific research and innovation ability of graduate students. This scale has 11 items in total. Typical question types include: "I have a rich reserve of professional knowledge and theoretical foundation", "I can quickly discover the essence of a problem when analyzing a problem", "I often participate in scientific research innovation competitions and academic conferences", etc. The scale adopts Likert 5-point scoring method. 1 means "completely inconsistent" and 5 means "completely consistent". The higher the score, the higher the degree of agreement with the issue. In this study, the Cronbach’s α coefficient of the scientific research and innovation ability scale was 0.965. The questionnaire exhibits good construct validity (CR) with a combined validity (CR) of 0.966 and a mean variance extracted (AVE) of 0.725. Confirmatory factor analysis showed χ²/df = 6.301, RMSEA = 0.077, GFI = 0.950, and CFI = 0.981. By comparing the correlation coefficients of the AVE square root of the variables with those of other variables to verify discriminant validity, the results showed that the AVE square root value for scientific research innovation ability (0.851) was greater than the correlation coefficients of other variables (0.549, 0.769), indicating good discriminant validity.

Scientific research self-efficacy scale

The scientific research self-efficacy scale of graduate students was measured using the scientific research self-efficacy scale developed by Forester [41]. This scale has a total of 22 items. Typical question types include: “The overall level of confidence in oneself in completing a research project”, “The level of confidence in oneself being able to read relevant literature in a certain research field based on research interests”, “The level of confidence in oneself being able to judge whether one’s ideas have research value based on the literature level of confidence” etc. The scale adopts a 5-point Likert scoring method, with 1 indicating "no confidence" and 5 indicating "full confidence". The higher the score, the stronger the sense of scientific research self-efficacy. Before using the questionnaire, researchers invited relevant experts to assess the consistency between the test items and the content being measured, ensuring that the test items covered the target content area and were representative. Simultaneously, based on the data collected in this study, the reliability and validity of the questionnaire were verified, focusing on its local applicability and explanatory power in the Chinese educational context. In this study, the Cronbach 's α coefficient of the scientific research self-efficacy scale was 0.989, the combined validity (CR) was 0.989, and the mean variance extracted (AVE) was 0.799, indicating good reliability of the questionnaire. Confirmatory factor analysis results showed χ²/df = 5.635, RMSEA = 0.072, GFI = 0.904, and CFI = 0.976, indicating good construct validity of the questionnaire. By comparing the correlation coefficients of the square root of the variable AVE with those of other variables, the discriminant validity among the variables was verified. The results showed that the square root value of scientific research innovation ability (0.894) was greater than the correlation coefficients of other variables (0.549, 0.769), indicating that the questionnaire had good discriminant validity.

Data processing

In the research process, the self-report method is a commonly used data collection method. Although it can provide direct and in-depth individual information, it also potentially faces the problem of common method bias. To minimize this bias, the author took several precautions when designing the research procedures. Firstly, by implementing an anonymous survey, the privacy and security of participants were ensured, and social desirability effects and self-glorification tendencies were reduced. Secondly, some questions used the reverse scoring method to identify and correct possible response biases. These procedural controls are used to improve the accuracy and reliability of data collected through self-report methods. In order to improve the scientificity and rigor of the research, confirmatory factor analysis was used to test the common method bias of the data. In this approach, all items are loaded onto a single factor; poor model fit indicates that a single common factor does not account for the majority of the variance, suggesting limited common method bias. The model fitting index is: χ²/df = 378.454, p < 0.001, RMSEA = 0.647, GFI = 0.578, AGFI = 0.155, PGFI = 0.289, RMR = 0.296. The poor fitting index indicates that there is no serious common method bias problem in the data of this study. However, we recognize that Harman's test has limitations, including its potential insensitivity to detecting all sources of method variance and its post-hoc nature. More robust alternatives, such as the marker variable technique or unmeasured latent method construct approach, could be employed in future studies to further validate these findings.

Subsequently, SPSS26 and AMOS26 were used to conduct confirmatory factor analysis and correlation analysis on the data to test the reliability of the questionnaire and the discriminant validity and correlation between the variables. The PROCESS plug-in compiled by Hayes (2012) was used to conduct mediation effect analysis. The bias-corrected percentile Bootstrap method was used to test the significance of the mediation effect. If the confidence interval did not include the value 0, it was considered statistically significant. Basic demographic variables (such as gender, place of birth etc.) were controlled as covariates in the mediation models. The research hypothesis was evaluated through a comprehensive analysis process to provide support for the interpretation of the research results.

Research results

Correlation analysis

As shown in Table 2, the correlation analysis of the main variables shows that there is a significant positive correlation between the mentor’s guidance function and graduate students’ scientific research and innovation ability and scientific research self-efficacy. Therefore, we can proceed to the next step of empirical analysis.

Table 2.

Descriptive statistics and correlation analysis of main variables

Variable	Mean	SD	1	2	3
Mentor guidance function	3.635	0.955	0.806
Scientific research and innovation ability	3.511	0.768	0.452**	0.851
Research self-efficacy	3.705	0.817	0.549**	0.769**	0.894

The triangular matrix below the table is the Pearson correlation coefficient between each variable

^*Means p < 0.05

^**Means p < 0.01

^***Means p < 0.001, the same below; the diagonal line is the average extraction variance (AVE) root value

Hypothesis testing

This study uses model Model4 (Model4 is a simple mediation model) in the SPSS plug-in macro-PROCESS compiled by Hayes (2012) for hypothesis testing. The 95% confidence interval of the mediating effect was estimated by repeatedly extracting 5000 Bootstrap samples. Under the condition of controlling basic demographic variables, the mediating effect of graduate students’ scientific research self-efficacy between mentor guidance function and graduate students’ scientific research and innovation ability was tested. The specific results were as follows (Table 3, Fig. 2).

Table 3.

The mediating effect of scientific research self-efficacy

	Scientific research and innovation ability		Scientific research self-efficacy		Scientific research and innovation ability
	β	t	β	t	β	t
Mentor guidance function	0.362***	15.077	0.468***	19.573	0.035	1.686
Scientific research self-efficacy					0.700***	29.084
R-sq	0.210		0.309		0.593
F	59.672***		100.362***		261.772***

β is the standardized coefficient; R-sq is the adjusted R- square

* p＜0.05, ** p＜0.01, *** p＜0.001 

Fig. 2.

Mediation test model

The mentor guidance function has a significant positive prediction effect on the scientific research and innovation ability of graduate students (β = 0.362, t = 15.077, p < 0.001), and hypothesis H1 is supported. After adding the mediating variable of scientific research self-efficacy, the mentor’s guidance function has a significant positive predictive effect on graduate student scientific research self-efficacy (β = 0.468, t = 19.573, p < 0.001), and hypothesis H2 is supported. The positive predictive effect of graduate student research self-efficacy on scientific research and innovation ability is significant (β = 0.700, t = 29.084, p < 0.001), hypothesis H3 is supported. This strong effect size has prominent practical implication for graduate education: it means that for every 1 standard deviation increase in graduate students' scientific research self-efficacy, their scientific research and innovation ability will increase by 0.700 standard deviations. In practice, this translates to tangible improvements—such as students being more willing to propose innovative research topics, more capable of designing rigorous experiments, and more likely to achieve breakthrough research results—making scientific research self-efficacy a core driver for improving graduate students' innovative capabilities. The direct effect of mentor guidance function on scientific research and innovation ability is not significant (β = 0.035, t = 1.686, p > 0.05), referring to Wen Zhonglin’s mediation effect test [42], this statistical non-significance of the direct path, along with the significant indirect effect, indicates that scientific research self-efficacy completely mediates the relationship within the tested model. However, the upper bound of the direct effect confidence interval (0.075) is close to zero, suggesting a potential small direct effect that may be detected with larger samples or alternative measures. Therefore, we interpret this as complete mediation in the current study, acknowledging that full mediation claims require causal evidence from longitudinal or experimental designs. Hypothesis H4 is supported.

In order to further verify the mediating role of scientific research self-efficacy, the Bootstrap sampling method was used for further testing. The bootstrap test results of the mediation effect are calculated and shown in Table 4. The confidence interval of the indirect effect [0.279, 0.378] does not include 0, and the confidence interval of the direct effect [−0.006, 0.075] includes 0, that is, the indirect effect exists, and the direct effect does not exist, and scientific research self-efficacy plays a complete mediating role in this cross-sectional model..

Table 4.

The mediating effect of scientific research self-efficacy

Effect	Effect size	Boot standard error	LLCI	ULCI
Total effect	0.362	0.024	0.315	0.409
Direct effect	0.035	0.021	−0.006	0.075
Indirect effect	0.328	0.025	0.279	0.378

In order to further explore the impact of the mentor’s guidance function on the scientific research and innovation ability of graduate students, a multiple stepwise regression was performed on the three dimensions of the mentor’s guidance function and the scientific research and innovation ability. The results showed that the mentor’s career development function and social support function have a positive impact on the scientific research and innovation ability of graduate students. reached the significance level, while the influence of role model function did not reach the significance level. The results are shown in Table 5.

Table 5.

Regression analysis of various dimensions of mentor guidance function on graduate students’ scientific research and innovation abilities

Dimensions	β	t
Career development function	0.223	3.649 ***
Social support function	0.221	4.838 ***
Role model function	0.046	0.833

* p＜0.05, ** p＜0.01, *** p＜0.001 

Discussions

While the relationship between mentorship and postgraduate students' scientific research and innovation ability has attracted considerable attention from scholars both domestically and internationally, current research primarily focuses on the impact of mentorship styles, models, intensity, and frequency on postgraduate learning outcomes and skill development. This study aims to apply the concept of mentorship, which has gained significant traction in the field of business management, to postgraduate education research, exploring the mechanisms by which mentorship influences postgraduate students' scientific research and innovation ability, and further expanding the scope of research on the impact of mentorship on postgraduate development.

The direct effect of mentor guidance function on graduate students’ scientific research and innovation ability

The data analysis results show that the mentor guidance function has a significant positive impact on the scientific research and innovation ability of graduate students. This is consistent with the research of Cramond, Jennifer [43, 44]. The mentor is the "first person responsible for graduate student training" [45] and plays multiple functions such as guidance and support [46]. He can help graduate students clarify the goals and direction of their professional learning, and has an important influence on the improvement of graduate students’ scientific research and innovation ability. Further analysis reveals that the mentor’s career development and social support functions significantly positively impact graduate students’ scientific research and innovation ability, with equivalent effects, while the role model function does not reach significance.

The career development function is an important factor influencing graduate students’ scientific research and innovation ability. The career development function is mainly reflected in the mentor providing systematic academic training and career development guidance to graduate students, helping them establish a solid academic foundation and clarify their future career direction. Scientific research is an important task for graduate students during their studies. In order to improve the professional skills of graduate students and enhance their professional competitiveness, mentors help graduate students acquire diversified knowledge, build a systematic knowledge system, and use their own resources to build a platform for graduate students, broaden their academic horizons, broaden their social networks, and improve their professional skills. Professional quality stimulates graduate students’ innovative thinking and enhances their scientific research and innovation ability. Therefore, the better the mentor’s career development function is performed, the higher the graduate students’ scientific research and innovation ability.

Social support function is also an important factor affecting graduate students’ scientific research and innovation ability. The social support function is represented by mentors providing necessary emotional support and resource assistance to graduate students when they face academic challenges and life pressures [25], thereby helping them to better cope with difficulties and maintain enthusiasm for scientific research and innovation [47]. Providing social support to students is also one of the main tasks of mentors and an important component of effective guidance [48]. In their interactions with graduate students, mentors also play the role of friends and care about students’ non-academic lives [49], helping students relieve stress [50]. Research has shown that the relational dimension of the mentor–mentee relationship does indeed play a role in the development of doctoral students' innovative abilities [51], mentors' psychological support significantly promotes the scientific research development of graduate students [52]. A mentor's respect, understanding, encouragement, and care can effectively alleviate students' anxiety, maintain their psychological resilience in pursuing innovation [53], and make them more willing to take on high-risk, high-value research tasks. This provides emotional support for groundbreaking innovation and has a certain impact on their scientific research and innovation ability.

The role model function is the exemplary and motivating effect that postgraduate students perceive from their supervisors in terms of academic attitude, research behavior, etc. According to social learning theory, the role model effect of tutors should have a significant impact on graduate students' scientific research and innovation ability. But in this study, no significant impact of mentor role model function on graduate students' research innovation ability was found, which is inconsistent with previous research findings [53]. This may stem from the authority-based mentor–mentee relationship formed under the influence of traditional respect for teachers in Chinese cultural contexts, leading graduate students to passively comply with their supervisors' arrangements rather than proactively internalize their innovative thinking and investigative spirit. Furthermore, the influence of mentors' academic character and professional conduct on graduate students is subtle and cumulative, requiring internalization into the graduate students' academic values, professional identity, and professional commitment, thereby affecting their research innovation behavior and ability, rather than directly impacting their research innovation ability. That is, the role model function of the mentor may indirectly operate through variables such as professional identity or career commitment. Meanwhile, the practical model of local scientific research training, which emphasizes task orientation and quantifiable outputs, also somewhat weakens the stimulating effect of the mentor's role model function on the scientific research and innovation ability of graduate students. Of course, this is only a conjecture and requires further investigation.

In conclusion, while graduate students are the primary agents of scientific research and innovation and possess independent research and innovation ability, the academic, professional, and psychological support provided by their supervisors is crucial for enhancing these capabilities. Fully coordinating and leveraging the professional development and social support functions of supervisors can help strengthen graduate students' scientific research and innovation ability.

The mediating effect of mentor guidance function on scientific research self-efficacy

Data analysis results indicate that the mentorship function influences graduate students' research self-efficacy. While previous research has rarely explored the relationship between these two variables, this finding is largely consistent with existing studies on the relationship between mentorship-related variables and students' research self-efficacy. The more comprehensive the mentorship in providing professional knowledge and skills, and the more timely the psychological support offered when students encounter problems—in other words, the more fully the mentorship function is utilized—the more confident the graduate students are in completing specific research work and producing research results, resulting in higher research self-efficacy. This study also found that graduate students' research self-efficacy affects their research innovation ability, which is consistent with previous research [54]. Research self-efficacy plays an important role in graduate students' research activities, significantly influencing their choice of research strategies and persistence in research behaviors. Increased research self-efficacy can stimulate graduate students' enthusiasm and initiative in their professional learning, further enhancing their research innovation.

Existing research indicates that the scientific research and innovation ability of graduate students are jointly influenced by multiple external and internal factors [55]. It is worth noting that upon introducing scientific research self-efficacy into the model, the direct effect of mentor guidance on innovation ability became non-significant, whereas the indirect effect through self-efficacy remained significant. This indicates that scientific research self-efficacy fully mediates the relationship between the mentor’s guidance function and graduate students’ research innovation ability. This result aligns with social cognitive theory, which posits that external environmental factors influence individual behavior through the mediating role of cognitive processes [56]. Specifically, effective mentor guidance helps graduate students consolidate their academic foundation and enhance their research competence. This accumulation of knowledge and skills reinforces their belief in their own research capabilities, thereby elevating their research self-efficacy. Enhanced self-efficacy, in turn, stimulates intrinsic motivation for innovation, leading students to engage more proactively in exploratory research and ultimately achieving substantial gains in innovation capacity. Thus, mentor guidance functions as an external driver, while research self-efficacy serves as the internal engine; the two are linked through a sequential mechanism of environmental support, cognitive construction, and behavioral transformation. This result not only reveals the indirect nature of supervisor guidance but also highlights the subjective position of individual graduate students. That is, the core driving force of research innovation ability is not supervisor guidance itself, but rather the innovative confidence and proactive exploration willingness formed by graduate students based on self-efficacy. In other words, the value of supervisor guidance ultimately needs to be realized by activating graduate students' research self-efficacy; the graduate students' own subjective awareness and initiative are the decisive factors in the development of research innovation ability. However, given the cross-sectional design and the direct effect CI's upper bound being close to zero, we interpret this as complete mediation within the model but caution against overgeneralizing causality. When guiding graduate students, supervisors should focus on cultivating their research self-efficacy and helping them establish positive research attitudes and belief. However, the above conclusions are based solely on the current analytical model, and the study is merely cross-sectional, lacking longitudinal tracking data. This makes it impossible to completely rule out the mediating or moderating effects of other potential alternative variables. Future research could further verify the robustness of these conclusions on a certain theoretical basis through longitudinal study designs and systematically examine other possible alternative mechanisms.

Research value

Implications

1. Mentor need to provide systematic academic support and career guidance throughout the entire postgraduate training cycle, help postgraduates clarify their academic development goals and career directions, help them build a solid theoretical and practical foundation, and provide diversified support for their scientific research ability.

2. Mentor should provide emotional support to graduate students through regular interaction, help them alleviate anxiety when they encounter bottlenecks or unclear directions in their research, and help them maintain a positive attitude and resilience in the face of challenges. In scientific research guidance, mentors should focus on emotional empathy, putting themselves in the students' shoes to understand their pressures and difficulties, and avoid excessive criticism and blame. For the students' efforts and progress, provide sincere affirmation and encouragement to strengthen their self-identity, while actively sharing their own research experiences of setbacks and overcoming challenges, guiding students to adjust their mindset and engage in research work with a positive and optimistic attitude.

3. Mentor should focus on cultivating the research self-efficacy of graduate students and strengthen their belief in their own research capabilities. Mentor can provide graduate students with sufficient autonomy in independent research and provide timely feedback and guidance. Based on the students' academic foundation and research experience, break down core research tasks into executable phased goals, dynamically adjust task difficulty, allowing students to accumulate confidence by completing various small goals, while guiding students to face research setbacks, and cultivate tenacious will in the process of overcoming difficulties, reinforcing self-awareness. They can also develop scientific and reasonable incentive mechanisms within the research group to motivate graduate students' enthusiasm and creativity in learning and research and improve their research self-efficacy.

4. Universities and education management departments need to further highlight the important role of mentors in the postgraduate training system. Through systematic training, assessment and incentive mechanisms, they should strengthen the academic and career guidance and social support role of mentors in postgraduate training, and comprehensively enhance their career development and social support functions. For example, in teacher onboarding training and graduate advisor specialized training, the three core responsibilities of academic guidance, career guidance, and social support can be incorporated as key training focuses, helping mentors to accurately align students' research needs with career development plans, strengthening their awareness and practical ability of emotional support for students, and promoting the comprehensive improvement of mentors' guidance skills. In the mentor evaluation indicator system, add a "Guidance Responsibilities" special indicator, clearly dividing the core dimensions of academic guidance, career guidance, and social support, refining and perfecting each dimension based on the specific guidance and support mentors can provide to students, while assigning corresponding weights according to the importance of each dimension to ensure the evaluation indicators are scientific, reasonable, and operational. Additionally, construct a multi-dimensional and three-dimensional evaluation mechanism, involving different entities such as students, peers, and the university to comprehensively and objectively evaluate the functioning of mentors' guidance; emphasize the quantifiable representation of guidance effectiveness, combining objective evidence such as students' research achievements, employment and further education quality, and various award situations to accurately quantify the actual effectiveness of mentors' guidance work. They should also strengthen mentors' awareness and support for postgraduates' research self-efficacy and guide them to focus on the generation of students' innovative confidence.

Contributions

The main contribution of this study is:

• (1) Existing research on mentorship mainly focuses on mentoring styles (such as supportive, democratic, and authoritarian styles, which are static classifications based on interaction patterns) or mentoring behaviors (such as the frequency and intensity of specific interactions like regular group meetings and task assignments). Unlike most existing studies that examine mentoring styles or behaviors from a single dimension, this study focuses on the mentoring function, which is not a single interactive feature or behavioral manifestation, but rather the functional results that directly affect student development through structured interaction in the mentoring relationship. It is a comprehensive variable formed by the systematic integration of mentoring behaviors. This study explores both the direct correlation mechanism between mentoring and scientific research innovation ability and the indirect effect of mentoring through scientific research self-efficacy. This exploration breaks through the fragmented analytical framework of previous studies on mentoring styles/behaviors and expands the conceptual boundaries of mentorship research.

• (2) Research on the mentoring function has accumulated relatively rich research results in Western countries, mainly focusing on corporate apprenticeship, youth mentorship, and student–teacher academic mentorship in the workplace, exploring the influencing factors, mechanisms of action, and outcome variables of the mentoring function. In contrast, research on the mentoring function in China is still in its initial stage. Existing results are mostly concentrated in the field of corporate apprenticeship. Although a few educational researchers have begun to pay attention to the relationship between the mentoring function and postgraduate training, no research has yet combined it with research self-efficacy to systematically explore the synergistic mechanism of the two on postgraduate research innovation ability. This study uses social cognitive theory as its theoretical foundation, integrates the dual perspectives of individual internal cognition (research self-efficacy) and external environmental support (mentor guidance function), and constructs a relationship model of mentor guidance function, research self-efficacy, and research and innovation ability. The study not only expands the application boundaries of the mentoring function theory, but also reveals the key path of action through empirical testing—although mentor guidance function has a significant positive impact on postgraduate research and innovation ability, its effect is entirely achieved indirectly by enhancing research self-efficacy. This discovery deepens the theoretical understanding of how mentorship can enhance students' innovative abilities. More importantly, it highlights the principal role of graduate students in the development of their research capabilities. Graduate students' objective assessment and belief in their own research capabilities are the core driving force for the development of their innovative abilities.

Limitations and prospects

This study has achieved certain results in exploring the relationship between the mentor’s guidance function and graduate students’ scientific research and innovation ability and its psychological mechanism. However, there are still some limitations. Future research can be further expanded on this basis.

Firstly, this study mainly examined the mediating role of scientific research self-efficacy from a cognitive perspective, but it still failed to fully consider the potential impact of non-cognitive factors on graduate students’ scientific research and innovation ability. Specifically, variables such as academic emotions (e.g., research anxiety, academic enjoyment) and psychological resilience were not included in the current model. Future research can further integrate cognitive and these specific emotional factors and adopt a more comprehensive psychological theoretical framework to reveal the multiple psychological effects of the mentor’s guidance function.

Secondly, this study uses cross-sectional data. Although it can reveal the correlation between variables, it is difficult to establish causality. Additionally, the mediation analysis shows complete mediation in this model, but the direct effect's confidence interval upper bound is close to zero, which may indicate a small non-null effect; cross-sectional data further limits causal inferences and risks overinterpretation of full mediation. Future research can adopt a longitudinal research design and analyze the causal effects between variables through tracking data, providing more solid evidence for understanding the dynamic relationship between the mentor’s guidance function and graduate students’ scientific research and innovation ability.

Third, this study uses self-report scales to collect data. Although this method is convenient for obtaining receipts, the validity of the measurement is easily affected by the subjective factors of the subjects. Problems such as memory bias or social desirability may affect the authenticity of the data, and thus have a potential impact on the validity of the research results. It needs to be carefully considered when analyzing data and deducing conclusions.

Fourth, although we employed Harman’s single-factor test to assess common method bias and the results suggested it was not a severe issue, we acknowledge that this method has well-documented limitations, including low sensitivity and inability to statistically control for method effects. More robust techniques, such as the unmeasured latent method construct (ULMC) or marker variable analysis, were not applied in this study. We therefore encourage future research to incorporate these or other latent variable approaches to provide a more rigorous evaluation of common method variance and to strengthen causal inference.

Fifth, although the sample includes multiple institutions and academic levels, the use of convenience sampling restricts the generalizability of the conclusions. While we have now reported the distribution across university tiers and degree levels, the sample was not randomly selected, and the proportions do not necessarily reflect the national population of graduate students in China. Moreover, detailed information on students’ academic year, prior research experience, or disciplinary background was not collected, which may mask important heterogeneity in the mentorship–innovation relationship. Future research should adopt probability sampling or stratified sampling strategies and collect richer demographic and academic background data to improve representativeness and enable subgroup analyses. At the same time, considering that the scientific research and innovation ability of graduate students may be affected by a variety of external factors, future research can further explore other potential moderator variables or mediating variables, such as academic atmosphere, scientific research resources, personal motivation, etc., to build a more complete theoretical model.

Finally, this study focused exclusively on graduate students’ perceptions of mentoring functions without examining objective characteristics of the mentors themselves. Potential moderator variables such as mentor gender, age, academic prestige (e.g., publication record, academic rank), and professional identity were not collected or analyzed. These mentor attributes may significantly influence how guidance functions are delivered and, consequently, how students perceive and internalize that support. Future research should employ a dyadic data collection approach, gathering information from both mentors and mentees, to examine whether mentor characteristics moderate the relationship between guidance functions and student outcomes.

Conclusion

Based on the social cognitive theory, this study uses the mentor’s guidance function as the independent variable and scientific research self-efficacy as the mediating variable to explore the mechanism of the mentor’s guidance function on the scientific research and innovation ability of graduate students. The research results show that: (1) The mentor’s guidance function has a positive impact on the scientific research and innovation ability of graduate students, with career development and social support functions demonstrating significant and comparable effect sizes, whereas the influence of the role modeling function was not statistically significant. (2) Scientific research self-efficacy mediates the relationship between mentor guidance function and innovation ability. These results highlight that the pathway from effective mentoring to innovative output is contingent upon activating students’ confidence in their own research capabilities.

Authors’ contributions

WYM designed the study and wrote the manuscript. NJC analyzed the data, WYM, NJC, MLN, CQY modified the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

Supported by the Fundamental Research Funds for the Central Universities of Ministry of Education of China, Xiamen University (Grant No. 20720231074); General Project of the Social Science Foundation of Fujian Province, China (FJ2025B192).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study, exclusively consisting of a questionnaire survey, obtained anonymous and private information solely through participants' consent to complete the questionnaire. All methodologies adhered strictly to the ethical principles outlined in the Declaration of Helsinki. Approval for the study was granted by the Institutional Review Board of Xiamen University (XDYX202306K39), and written informed consent was provided by all participants prior to their involvement in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.