Dynamic calibration of self-efficacy to cognitive load: the longitudinal mediation effect of state anxiety Authors: Xiaolong Li, Mu Xia Published: 2024 ABSTRACT AbstractBackgroundPrevious research has shown that higher self-efficacy predicts lower cognitive load. Nevertheless, there is limited research exploring the dynamic calibration phenomenon of cognitive resources by self-efficacy during task execution through experimental methods.MethodsThis study elucidates the relationships and longitudinal mediating effects among self-efficacy, state anxiety, and cognitive load through two studies that employed convenience sampling. Study 1, which aims to provide a comprehensive theoretical understanding of the research question, is an interview study that examines the prevalence of fluctuations in self-efficacy among 22 university students through an in-depth interview. On the other hand, study 2 reveals further causal relationships through experimental methods. It adopts a 3 (repeated measures: first time, second time, third time) × 2 (conditions: feedback group, non-feedback group) mixed design by 86 participants, utilizing a false feedback paradigm to manipulate participants’ self-efficacy, with the task of the experiment being emotion recognition.ResultsThe results demonstrate that, compared to the non-feedback group, false feedback can successfully manipulate the self-efficacy of the participants, and the data from the feedback group further reveals the multilevel mediating role of state anxiety in the process of self-efficacy influencing cognitive load.ConclusionIn general, the findings from two studies suggest that the enhancement of college students’ self-efficacy during task execution can reduce their state anxiety, which in turn can further alleviate cognitive load.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40359-024-02254-y.Keywords: Self-efficacy, State anxiety, Cognitive load, Mediating effect Introduction An individual’s belief in their ability to mobilize motivational and cognitive resources, as well as action plans to meet the demands of a given situation, is referred to as self-efficacy [1]. Self-efficacy plays a crucial role in the task process as it is associated with dynamic self-regulation processes [2]. Individuals with high levels of self-efficacy believe they will persevere toward their goals, influencing the amount of effort and duration of involvement when faced with difficulties. Given the significant role of self-efficacy, there is a growing need to explore the dynamic changes in self-efficacy during task execution and the impact of these changes [3]. Previous cross-sectional research has explored the relationship between self-efficacy, state anxiety, and cognitive load [4, 5], which suggests that self-efficacy negatively predicted state anxiety and cognitive load [6], and state anxiety positively predicted cognitive load [7]. Moreover, both self-efficacy and state anxiety exhibit variability. During task execution, we posit that dynamic interrelations could exist among self-efficacy, state anxiety, and cognitive load [8]. These relationships can significantly affect people’s behaviors. However, the mechanisms underlying the relationships among the three variables have yet to be clarified. The study investigates the dynamic impact of self-efficacy on cognitive load and also examines the involvement of state anxiety in this relationship. Self-efficacy and cognitive load Cognitive load refers to the cumulative mental effort exerted on an individual’s cognitive system while performing a specific task [5, 9]. During the process of learning or engaging in specific tasks, the subjective experience generated in one’s mind can be understood as cognitive load [10], and it possesses subjectivity, implicitness, and variability [11]. These characteristics make cognitive load susceptible to the influence of self-efficacy, as self-efficacy often impacts subsequent task performance, persistence, and other related aspects [12]. Consistent with this view, research has confirmed that self-efficacy has significant effects on subjective measurements of cognitive load, task performance, and physiological measurements [13]. For example, the findings of a study utilizing self-report ratings to measure cognitive load revealed that participants who perceived themselves as more creative in creative thinking tasks, namely individuals with greater belief in their own creative abilities experienced reduced mental effort, particularly those individuals with higher levels of creative self-efficacy [14]. Another study revealed a negative correlation between students’ English reading self-efficacy and intrinsic cognitive load [15]. Furthermore, a high level of psychological resilience traits (including mental toughness, resilience, self-efficacy, and achievement motivation) is positively associated with increased alpha desynchronization during encoding and retrieval processes [16]. The researches mentioned have respectively validated the prediction that higher self-efficacy can predict lower levels of cognitive load through self-report, reading performance, and event-related potentials, which represent different indicators. This provides a solid theoretical foundation for our current research. Alternatively, it is regrettable that the aforementioned study only verified the effect of self-efficacy on cognitive load from a cross-sectional perspective, without considering the longitudinal changes in their relationship. The dynamic change of self-efficacy Evidence from theory and empirical studies both support the dynamic changes in self-efficacy during the process of task completion. First, the theory of self-efficacy fluctuation [17] suggests there is a possibility that self-efficacy during task execution may undergo brief and dynamic changes along with the feedback results of that task. Consistent with this hypothesis, the study found that successful experiences convey positive feedback to individuals, thereby enhancing their sense of efficacy, whereas unsuccessful experiences transmit negative feedback, leading to a decrease in self-efficacy. Therefore, feedback serves as a significant method for increasing or decreasing self-efficacy [18]. Previous studies have confirmed the phenomenon of self-efficacy fluctuation [19]. Secondly, empirical evidence also demonstrates the dynamic nature of self-efficacy during task processes. Some scholars have found that higher levels of willpower self-efficacy, as a specific aspect of self-efficacy, are associated with greater positive impacts [20]. This finding suggests that both positive and negative life experiences can markedly influence an individual’s self-efficacy. In specific tasks, fluctuations in self-efficacy persist, as evidenced by the outcome of a problem-solving task where interest and self-efficacy decreased for sixth-grade students over time, while perceived difficulty initially increased and then decreased. It indicates that self-efficacy not only fluctuates during task execution but also impacts an individual’s cognitive behaviors. Through the examples above, we find that positive or negative experiences often induce the dynamic change of self-efficacy, and feedback is a good method to manipulate such experiences [21]. Furthermore, tasks such as subtraction operations [22], internet usage [23], and examinations [24] have been shown to elicit changes in participants’ self-efficacy beliefs during the completion process. Previous research has found that providing subjects with false-positive feedback not only helps them learn motor skills but also enhances their self-efficacy [25]. Another study suggested that students experienced lower self-efficacy when reading false feedback containing criticism [26]. Therefore, we manipulated individuals’ self-efficacy using a false feedback paradigm. We hypothesize simultaneously that false negative feedback will decrease individuals’ self-efficacy, while false positive feedback will enhance their self-efficacy. State anxiety and cognitive load Negative emotions are an important factor that influences cognitive load [27]. Anxiety stands out as a prominent negative emotion frequently encountered by individuals. Speilberger divided anxiety into trait and state anxiety [28]. The former is a personality trait that has a relatively stable tendency toward anxiety with individual differences. The latter refers to a temporary and passive reactive state of individuals in specific situations, characterized by tension, worry, distress, fear, and apprehension [29]. State anxiety is triggered by situations and can affect an individual’s performance in situational tasks; however, the underlying mechanism of this impact has not been clarified [12, 30]. Anxiety itself can occupy a portion of the cognitive processing resources that should be used for working memory processes [31]. Initially, scholars investigated the correlation between cognitive resources and anxiety and found that these resources are limited in quantity. When an individual’s level of anxiety is high, this results in a relative insufficiency of working memory resources available for cognitive tasks [32]. Subsequently, building upon the foundation of these studies, processing efficiency theory provides a corresponding explanation. The theory of processing efficiency suggests that an individual’s cognitive resources are limited [33]. When negative emotions such as anxiety occupy an individual’s cognitive resources, the resources available for cognitive operations will consequently decrease. Therefore, when individuals experience negative emotions such as anxiety, their cognitive functioning is affected, and their engagement in tasks is also affected [7, 34]. The theory also distinguishes between task performance and processing efficiency, where task performance refers to the quality of task completion, whereas processing efficiency denotes the ratio between task performance and the resources invested, reflecting the efficiency of the working memory system [35]. Therefore, anxiety does not necessarily impair cognitive performance effectiveness but weakens an individual’s processing efficiency, which pertains to the correlation between the results of performance and the cognitive resources expended during the task [36, 37]. Numerous studies have corroborated the efficacy of this theory. The literature has shown that state anxiety can impair attentional alertness and orientation in adolescents [38, 39]. The education literature further validates the above-mentioned theory. A study indicated that anxiety directly influences students’ perceived knowledge improvement, boredom, and cognitive load, thereby negatively impacting their academic performance. Another study on math test anxiety revealed that children with high mathematical anxiety also exhibit higher cognitive load [40]. The dynamic change of anxiety The word “State anxiety” seems to exhibit significant variability which is a crucial reason for it to become a distinct concept [41]. In fact, there is a large body of empirical evidence that supports this assertion. Research has shown that negative events in life can significantly elevate an individual’s level of state anxiety [42]. Literature on establishing a state anxiety tracking model mentions inducing participants’ state anxiety by exposing them to images of varying levels of aversion, capturing the dynamic changes in state anxiety levels using dimensional emotion ratings, electrocardiograms, and skin conductance responses [43]. Therefore, alterations in the environment or during task execution often lead to fluctuations in state anxiety. Self-efficacy and state anxiety Previous studies have revealed that self-efficacy is closely correlated with status anxiety [44]. Self-efficacy plays a central role in individual cognitive processes by strongly influencing thoughts, emotions, motivations, and actions [45, 46]. From a static perspective, some studies indicate that self-efficacy often has a greater impact on individuals’ emotions, whereby individuals with low self-efficacy are more prone to experiencing negative emotions such as anxiety and worry [47, 48]. From a dynamic standpoint, a longitudinal study indicates that current levels of teacher self-efficacy (TSE) can predict the future emotions of teachers. Specifically, teachers with higher self-efficacy tend to experience more positive emotions, while those with lower self-efficacy are likely to experience more negative emotions. Another results from an ecological momentary assessment (EMA) study have shown that higher willpower self-efficacy significantly predicts positive emotions [49]. The research employs a longitudinal design to validate the structure of anxiety predicting factors, revealing that self-efficacy is a crucial factor in predicting emotional and performance-related outcomes. Similar results were found in another experiment in which participants were required to conduct two impromptu speeches, with feedback provided after the first speech. The second speech demonstrated a decrease in anxiety levels and negative assessments compared to the first speech [50]. From two perspectives, static and dynamic, there exists a close relationship between self-efficacy and state anxiety, which provides important support for our research. Aim of the study The literature review from 1.1 to 1.5 provides two significant clues for this study. One is that previous research consistently demonstrated that both self-efficacy and state anxiety can have a significant impact on cognitive load, which suggests that self-efficacy may have an influence mechanism on cognitive load [51], with state anxiety as a mediating factor. The other clue is that individuals’ self-efficacy and state anxiety fluctuate when receiving feedback, indicating a dynamic interplay between the two during task performance [52, 53]. Based on this, we propose the hypothesis of this study: H1: Self-efficacy can be manipulated by the false feedback paradigm. Positive false feedback can enhance an individual’s self efficacy, while negative false feedback can diminish it. H1: Self-efficacy can be manipulated by the false feedback paradigm. Positive false feedback can enhance an individual’s self efficacy, while negative false feedback can diminish it. H2: During the dynamic execution process of tasks, self efficacy can negatively predict state anxiety and cognitive load and state anxiety can positively predict cognitive load. H2: During the dynamic execution process of tasks, self efficacy can negatively predict state anxiety and cognitive load and state anxiety can positively predict cognitive load. H3: State anxiety plays a multilevel mediating role in the mechanism of the influence of self-efficacy on cognitive load. H3: State anxiety plays a multilevel mediating role in the mechanism of the influence of self-efficacy on cognitive load. Previous cross-sectional studies did not systematically verify the above hypothesis. To test the above hypotheses, the present study initially verifies self-efficacy as influencing the dynamic calibration mechanism of cognitive load in real-life scenarios through surveys and interviews. Subsequently, the validity of this mechanism is examined using experimental methods. The experiment uses a 3 (repeated measures: first time, second time, third time) × 2 (conditions: feedback group, non-feedback group) mixed design [54], and the experimental task is an expression recognition task. Selecting this task over the usual cognitive tasks or specific subject tests not only makes it more engaging for participants but also becomes a crucial skill in people’s daily lives. It employs the paradigm of false feedback to manipulate individuals’ self-efficacy and explores the impact of dynamic changes in self-efficacy on state anxiety and cognitive load, reflecting the dynamic process of emotional-cognitive changes in task execution. Finally, repeated measurement ANOVA and multilevel linear model were used for data analysis. The former can examine whether the manipulation of false feedback on self-efficacy is effective, while also examining the differences in state anxiety and cognitive load between the feedback and non-feedback group. The latter can embed individual data obtained from repeated measurements at different time points and conditions, allowing for verifying the longitudinal mediating relationship of experimental data. Therefore, this method has been employed as the data processing approach in this study [55]. Figure 1 is the theoretical mode of the present study. Theoretical model This study employed two research designs to test the hypotheses proposed: interview and experimental research. Study 1 relied on first-hand interview data to investigate the fluctuation of self-efficacy and its impact on subsequent tasks, thus providing initial validation of the model. Study 2 used the experimental research method to further validate the first study by manipulating participants’ self-efficacy, first decreasing the participants’ self-efficacy for the current task, and then gradually increasing their self-efficacy. To verify whether there exists a model in which state anxiety mediates the impact of self-efficacy on cognitive load. Study 1 Study 1 employed a questionnaire survey and in-depth interviews to examine the variations in self-efficacy. Initially, 146 questionnaires concerning the phenomenon of self-efficacy fluctuations were gathered. Subsequently, participants who met the criteria and expressed willingness to partake in interviews were chosen for in-depth interviews to acquire firsthand insights into their views and comprehension of self-efficacy fluctuations. Moreover, the collected textual data were coded and analyzed to delve into the psychological implications and factors influencing the fluctuations in self-efficacy [56]. Method Participants Adopting convenient sampling, we selected 146 students (47 males, 99 females, and students aged between 18 and 22 years) from a university in Guangxi, China to participate in an online questionnaire survey. After opening the questionnaire, each individual will encounter an introduction before being presented with the option to commence filling it out. From the students who completed the questionnaire, we selected 22 (10 males and 12 females) individuals who met the criteria and were willing to further participate in in-depth interviews. The students, all of whom are from China, consist of 15 in their sophomore year and the remaining 7 in their freshman year. Prior to the interviews, each individual signed an informed consent form explicitly stating that they could terminate the interview at any time if necessary. All interviews were conducted in Mandarin Chinese. The research has obtained ethical approval from the Ethics Committee of the School of Education Science at Guangxi University for Nationalities. Data collection Before the formal interview, the researchers conducted an online survey with 146 university students. The survey questionnaire focused on the fluctuation experience of self-efficacy, beginning with three questions to understand whether the participants had experienced fluctuations in self-efficacy and if this experience would affect their engagement in the upcoming tasks. For instance, the first question on the questionnaire was: “Do you think that when you encounter experiences of success or failure in the process of doing something in your life, will your confidence in continuing to do this thing change”. Participants were also invited to freely report their experiences of fluctuation in self-efficacy based on their real experiences and feelings. This not only served as the foundation for designing the formal interview outline, but we also prioritized inviting individuals who shared detailed experiences to participate in the interviews and answered “yes” to all three questions in the questionnaire within the following week. Three “yes” represent participants experiencing fluctuations in self-efficacy in their daily lives, believing that these fluctuations will affect their level of effort or behavior toward future events. Moreover, in this process, the participants also consider emotions to play a certain role. Conduct semi-structured interviews led by a primary interviewer focusing on the theme of “ the impact mechanism of dynamic fluctuations in self efficacy on cognitive load”. The selection of interview locations prioritizes the convenience and safety of the participants, ensuring an undisturbed interview process [57]. The specific interview time is determined by the interviewees according to their schedule, with each interview lasting approximately 15 to 30 min. The formal interview outline consists of three parts, firstly, establishing a relationship with the interviewees, informing them of the informed consent and basic ethical rules, and inquiring about the possibility of recording the interview process. Secondly, invite the interviewees to elaborate as much as possible on their experiences of fluctuating self-efficacy, including time, events, people, and emotional changes. Finally, based on the experiences shared by the interviewees, explore the specific processes and reasons for these fluctuations. Typical questions include “ Do you believe this experience has had an impact on your confidence in relation to the matter you just described? ”, and “ How do you think these changes will affect your future performance in this matter ? ”. Upon completion of the interviews, each participant was given a gift including some study supplies and snacks, valued at around 8 RMB. Researchers transcribe audio files verbatim and cross-reference the transcribed text with the audio to ensure accuracy. Subsequently, identifying information of interviewees is removed from both the verbatim transcript and interview records, assigning a code to each interview for organized analysis of data later (Zhang & Yang, 2018). The transcribed verbatim totals 50,625 words. The survey questionnaire and interview outline will be presented in the appendix. Data analysis After transcription, the interview data was imported into Nvivo 12.0 (referred to as N12) qualitative analysis software [58]. This study, based on Grounded Theory, involved the researcher in a continuous process of open, axial, and selective coding of the data, engaging in reflective comparison and contrast analysis. The researcher further adopted an inductive analysis approach for constructing the data. The specific coding results are detailed in Table 1. Table 1Coding results of self-efficacy fluctuation phenomenonSelective codingAxial codingOpen codingnodeThe fluctuating phenomenon of self-efficacyInduced situationexamination9Class process5Election campaign3Sports competition2Learning process2Organize activities1Occurrence frequencyinfrequent Occurrence6Moderate Occurrence9frequent Occurrence7Fluctuation rangehigh8Moderate8low6Emotional changes with elevated self-efficacyPositive emotionpride12happy10joyful11Emotional changes with reduced self-efficacyNegative emotionanxiety17frustrate11self-reproach5agitated6sad4nervous4Impact on subsequent tasksPsychological performance17Behavior expression18performance10 Coding results of self-efficacy fluctuation phenomenon Reliability and validity test of the study Unlike quantitative research, the reliability of qualitative research is difficult to evaluate using quantitative indicators. This study primarily adopts two measures to enhance the reliability of data analysis [59]: first, continuous iterative coding. Different researchers encode, analyze, and re-encode from their respective perspectives, strengthening the verification of the analysis results. Second, reflection and discussion. Coders and experts in the field of self efficacy research regularly engage in thorough discussions, combining the viewpoints of research members to reflect and revise the results. The validity test adopts the participant test method and the non-participant test method for validity verification [60], in order to avoid coders’ excessive interpretation and subjective bias toward the data. A total of 5 examiners provided feedback on the analysis results, mainly examining whether the coders over-interpreted the material or had biased understanding during the coding process. Among them, the participant test includes two interviewers from the interview stage (including 2 psychology graduate students); the non-participant test includes 3 researchers who did not participate in the interviews (including 3 psychology experts). Finally, the researchers summarized the feedback from the examiners and processed the opinions from three aspects: (1) whether the data analysis and integration process is reliable; (2) whether the construction of category structures is reliable; (3) whether the researchers had subjective biases during the coding process. If there is evidence supporting the coding results, they are retained; if the evidence is insufficient, appropriate modifications are made. If there are discrepancies between the examiners’ understanding and the coding results, the researchers will provide corresponding explanations. Result By examining Fig. 2, the results of the survey showed that 93.84% (137 participants) of the participants had experienced fluctuations in self-efficacy(χ2 = 146.00, p < 0.001). Among them, 78.83% (108 participants) believed that improving self-efficacy during the task process could increase their positive emotions, whereas decreasing self-efficacy could increase their experience of negative emotions (χ2 = 135.00, p < 0.001). Results of self-efficacy dynamic calibration The interview results can be referred to Table 1. Firstly, regarding the occurrence scenarios of fluctuations in self-efficacy feelings, the interview results mainly include five situations: examination, class process, election campaign, sports competition, and learning process, covering various aspects of student life. S2: During the 800-meter race, at the start, I did not feel good and then I lost confidence, although I wanted to do well in the race. S2: During the 800-meter race, at the start, I did not feel good and then I lost confidence, although I wanted to do well in the race. S10: When taking the college entrance examination, I can feel a huge change in confidence in the exam, and I think it may be that this exam is important to me.Secondly, when changes occur in self-efficacy, respondents often experience different emotions. When self-efficacy is enhanced, pride and joy are the most frequently mentioned positive emotions by respondents. Conversely, when self-efficacy decreases, anxiety is the most reported negative emotion by respondents. This suggests that anxiety plays a unique role in decreasing self-efficacy: S10: When taking the college entrance examination, I can feel a huge change in confidence in the exam, and I think it may be that this exam is important to me. Secondly, when changes occur in self-efficacy, respondents often experience different emotions. When self-efficacy is enhanced, pride and joy are the most frequently mentioned positive emotions by respondents. Conversely, when self-efficacy decreases, anxiety is the most reported negative emotion by respondents. This suggests that anxiety plays a unique role in decreasing self-efficacy: S6: During this exam, losing confidence made me even more anxious and nervous, worried about not doing well, and I found it difficult to focus on answering questions. S6: During this exam, losing confidence made me even more anxious and nervous, worried about not doing well, and I found it difficult to focus on answering questions. S15: Before organizing the event, I was very anxious because I had never done something like this. After finishing the activities that the teacher asked me to organize, there were other activities behind me. I was more handy, and I am no longer anxious. S15: Before organizing the event, I was very anxious because I had never done something like this. After finishing the activities that the teacher asked me to organize, there were other activities behind me. I was more handy, and I am no longer anxious. S20: There was this feeling when I took the exam, and I got the answer after the exam, and the result was really unsatisfactory. S20: There was this feeling when I took the exam, and I got the answer after the exam, and the result was really unsatisfactory. Study 1 preliminarily identified the potential role of state anxiety in self-efficacy and cognitive load through investigation and interviews. Building on this foundation, a second study further validated this hypothesis through experiments. Study 2 Method Participants Using G*power 3.1 to calculate the sample size required for research [61], with the parameters set as follows: Two-way Repeated-Measures ANOVA, effect size f = 0.25, statistical power 1 − β = 0.9, α = 0.05, number of measurements = 3, the calculated sample size is 36 individuals. We recruited 96 Chinese college students as participants differentiate from the participants in Study 1, after screening out invalid data, 86 participants remained (32 male and 54 female, students aged between 18 and 24 years). Because the feedback group received false feedback, we added the following question to ensure the authenticity of the experimental data: “Do you believe the feedback you received?” After excluding participants who stated no to this question the feedback group consisted of 39 participants, whereas the non-feedback group consisted of 47 participants. The participants had normal or corrected vision and good health conditions. Before the experiment commences, all participants signed an informed consent form, and they can withdraw at any time if they have any questions. They were given a small gift containing study supplies and snacks, valued at 8 RMB after completing the experiment. In addition, they were informed that the feedback received during the experiment was not their actual performance. Materials A self-constructed self-efficacy questionnaire was used. Bandura recommends using as few items as possible when measuring short-term changes in self-efficacy [62, 63]. The self efficacy questionnaire consists of only one item, “Through the recent task, how confident do you feel in identifying emotions of fear, disgust, anger, and sadness on the faces in images?”, on a 9-point rating scale [64]. Participants were required to choose a suitable number from 1 to 9 based on their own feelings, with the confidence level increasing sequentially from 1 to 9. Short State Anxiety Inventory (SSAI): We used the Chinese Brief State Anxiety Scale [65], which consists of two factors: anxiety presence and anxiety absence, each with three items. Each item is scored on a scale of 1 (not at all) to 4 (almost always), with the items of anxiety absence being reverse scored. Higher scores indicate higher levels of anxiety. The internal consistency reliability of this scale is 0.82 [66]. In this study, only the factor of anxiety presence was used to measure the participants’ state anxiety levels, and the internal consistency reliability is 0.80. The revised version of The Cognitive Load Subjective Ratings (Paas, 2003) is a commonly used questionnaire to measure cognitive load because of its subjective nature [67, 68]. It consists of three items, which evaluate both psychological effort and task difficulty. Two of the items are similar, focusing on the evaluation of task difficulty. To minimize the number of questions in this experiment where participants’ cognitive load is measured multiple times, only two questions are used. Both questions adopt a 9-point rating scale. Participants are required to choose a suitable number from 1 to 9 based on their own feelings to answer the two items. The numbers from 1 to 9 represent increasing levels of psychological effort and material difficulty, with 1 indicating minimal effort and “very easy” and 9 indicating maximum effort and “very difficult.” The scale demonstrates high reliability, with an internal consistency reliability coefficient of α = 0.74 [69]. The cognitive load index is the average of the two dimensions, and the internal consistency reliability is 0.72. The Nimstim emotional facial expression database: The emotion pictures used in the task were selected from the NimStim database [70, 71]. The evaluation time of the NimStim emotional face database was lengthy and most of the participants in the NimStim evaluation were foreigners. Therefore, to control experimental variables, we first recruited Chinese university students to evaluate the pictures in the emotion database. Using convenience sampling, 40 college students (17 males and 23 females) from Guangxi University for Nationalities were selected as participants. A total of 157 fear, disgust, sadness, and anger emotion pictures with original accuracy below 80% were selected from the NimStim emotional face database. Finally, the lowest recognition rate of 60 facial expression images was selected from a total of 157 material images and used as the task images for this experiment. Specifically, based on the correct recognition rate of the images by 40 participants, we selected 15 images with the lowest correct rate for each expression as the materials for the formal experiment to adopt the false feedback paradigm. To make the participants believe that the feedback they received was real, we chose tasks with higher difficulty as experiment materials. This includes 15 pictures each of fear, disgust, sadness, and anger [70]. Procedure The study employed a 3 (repeated measures: first time, second time, third time) × 2 (conditions: feedback group, non-feedback group) mixed experimental design. The feedback conditions were considered between-subjects factors, while the repeated measures were regarded as within-subjects factors. The non-feedback group was established to verify the effectiveness of false feedback in self-efficacy manipulation, while comparing it with the feedback group to examine the changes in state anxiety and cognitive load. A facial expression recognition task was utilized as the experimental content, manipulating participants’ self-efficacy through a false feedback paradigm. The experiment consisted of four practice and formal blocks. The specific procedure is illustrated in Fig. 3. Each block started with false feedback, which indicated the correct number of images identified from the previous task. On receiving false feedback, the participants’ self-efficacy and state anxiety were immediately measured, followed by 15 facial expression recognition trials and then the cognitive load was measured. The three blocks of the formal experiment followed the same process described above, but the false feedback received differed. In the first block, individuals receive feedback based on practice, and to initially lower their self-efficacy, they are given false-negative feedback, with the correct number of identified images randomly ranging from 1 to 3. In the second block, the correct number of identified images randomly ranged from 6 to 8, and this false feedback was closer to the actual recognition rate. In the third block, participants were given false positive feedback, with the correct number of identified images randomly ranging from 11 to 13, manipulating the participants’ self-efficacy from low to high. To ensure the authenticity of experimental data due to the acceptance of false feedback by the feedback group, the question “Do you believe the feedback you just received?” was added at the end of the experiment. The formal experiment process Data analysis The data were subjected to descriptive analysis using IBM SPSS 26 which is a powerful and newer version [72] to understand the performance of the main variables in the experiment, followed by mixed-design ANOVA [73] to verify the effectiveness of manipulating self-efficacy perception in the false feedback paradigm. Repeated measures represent a within-subject factor, while grouping delineates a between-subject factor. Subsequently, HLM 6.08 was employed to conduct multilevel linear analysis on the data from the feedback group, confirmed that the impact of self-efficacy on cognitive load is mediated through multilevel mechanisms of state anxiety. Result Self-efficacy manipulation test Descriptive statistical analysis was conducted on the data of 86 participants using SPSS 26.0. The mean trends of self-efficacy, state anxiety, and cognitive load across three measurements by participants are illustrated in Table 2. Table 2Descriptive statistics (Mean, SD)First measurementSecond measurementThird measurement M SD M SD M SD Feedback group(n = 39) Self-efficacy4.742.195.792.056.671.75 State anxiety5.722.084.691.763.871.28 Cognitive load6.821.126.261.235.761.35Non-feedback group(n = 47) Self-efficacy5.451.845.341.765.511.91 State anxiety5.402.035.362.124.872.04 Cognitive load6.351.356.301.326.281.47 Descriptive statistics (Mean, SD) After the participants received three different intensities of false feedback, their self-efficacy, state anxiety, and cognitive load were measured in a timely manner. Upon receiving false feedback, the participants were immediately assessed for self-efficacy and state anxiety, and cognitive load was measured immediately after task completion. The effectiveness of self-efficacy manipulation was verified using repeated-measures analysis of variance.The study also confirmed that significant variations exist in state anxiety and cognitive load following changes in self-efficacy under different conditions. For p-values that did not meet the assumption of sphericity, Greenhouse-Gessisser correction was applied. Bonferroni correction was used for multiple comparisons. Firstly, a repeated measures analysis of variance was conducted with self-efficacy as the dependent variable, using a 3 (measurement time: 1, 2, 3) × 2 (group: Non-feedback group, Feedback group) design. The results revealed a significant main effect of measurement time, F(2, 168) = 22.821, p < 0.001, η2 p = 0.214.Additionally, a significant interaction effect between measurement time and group was observed, F(2, 168) = 20.367, p < 0.001, η2 p = 0.195. Further simple effects analysis indicated a significant simple effect of measurement time under the feedback condition, F(2, 83) = 25.755, p < 0.001, η2 p = 0.383. The self-efficacy perception as measured three times by the feedback group exhibited a significant gradually increase, rising from 4.74 in the first measurement to 6.67. Furthermore, all associated p-values were below 0.001. While no significant simple effect of measurement time was found under the non-feedback condition, F(2, 83) = 0.693, p = 0.503, η2 p = 0.016. It suggests that by using the false feedback paradigm and creating different feedback conditions for the participants, their self-efficacy can be effectively manipulated. Previously, to examine whether changes in self-efficacy affect state anxiety and cognitive load, a repeated measures analysis of variance was conducted for both variables, with three measurement times (1, 2, 3) and two groups ( Feedback group, Non-feedback group). Conducting a repeated measures analysis of variance with state anxiety as the dependent variable across three measurement time (1, 2, 3) and two groups ( Feedback group, Non-feedback group, ) revealed a significant main effect of measurement time F(2, 168) = 25.183, p < 0.001, η2 p = 0.231. The interaction effect between measurement time and group is statistically significant, F(2, 168) = 8.294, p < 0.001, η2 p = 0.090. Further simple effect analysis revealed that under feedback conditions, the simple effect of measurement time on state anxiety was significant, F(2, 83) = 24.829, p < 0.001, η2 p = 0.374. The state anxiety perception as measured three times by the feedback group exhibited a significant gradually decrease, reducing from 5.72 in the first measurement to 3.87. Furthermore, all associated p-values were below 0.001. Whereas under non-feedback conditions, the simple effect of measurement time on state anxiety was not significant, F(2, 83) = 2.994, p = 0.056, η2 p = 0.067.This indicates that there are significant differences in state anxiety within the feedback group that correspond to changes in self-efficacy. Cognitive load was taken as the dependent variable in a repeated measures analysis with 3 (measurement time: 1, 2, 3) × 2 (group: Feedback group, Non-feedback group) design. The main effect of measurement time on cognitive load was found to be significant, F(2, 168) = 11.603, p < 0.001, η2 p = 0.121.There was a significant interaction effect between measurement time and group, F(2, 168) = 8.748, p < 0.001, η2 p = 0.094.Further simple effects analysis revealed that under feedback conditions, there was a significant simple effect of measurement time on cognitive load, F(2, 83) = 13.627, p < 0.001, η2 p = 0.247. The three measurements of cognitive load by the feedback group demonstrates significant gradually decrease, reducing from 6.82 in the first measurement to 5.76. Furthermore, all associated p-values were below 0.01. Conversely, under non-feedback conditions, the simple effect of measurement time on cognitive load was not significant, F(2, 83) = 0.081, p = 0.922, η2 p = 0.002. It suggests that cognitive load within the feedback group varies significantly with changes in self-efficacy. Based on this, we can preliminarily confirm that within this study, false feedback can successfully manipulate participants’ self-efficacy, namely, individuals’ self-efficacy in the task increases with the accuracy of false feedback. Furthermore, following changes in self-efficacy, individuals’ state anxiety and cognitive load in the task also undergo alterations. Multilevel mediation effects of state anxiety Since this study used a multilevel linear model to test the mediation effects, and the data at level 2 were contextual variables, it was necessary to test the within-group consistency of the data. Firstly, in order to ensure the within-group consistency of the data, it was stipulated that rwg must exceed 0.70. Within this study, the average rwg of the three variables included were 0.71, 0.84, and 0.88, respectively, all meeting the criterion of being greater than 0.70, indicating a high level of within-group consistency. The ICC(1) for the three variables were 0.59, 0.54, and 0.51, all greater than 0.06. The ICC(2) was 0.76, meeting the criterion of being greater than 0.70, indicating a high level of between-group heterogeneity and allowing for multilevel mediation analysis. In this study, self-efficacy and state anxiety are centralized at the individual level. Based on the variables of self-efficacy, state anxiety, and cognitive load in this study, a 1-1-1 lower-level mediation model was constructed for analysis. The method of sequential testing of regression coefficients [74]was employed to conduct a multilevel analysis on the mediating effect of state anxiety between self-efficacy and cognitive load. Step 1, establish a null model. Among them, ICC = u0 /( u0 + r) = 0.5174, and the within-group correlation coefficients are all greater than 0.060. The nested structure of the data cannot be ignored, and a multilevel mediation analysis is needed. Step 1, establish a null model. Among them, ICC = u0 /( u0 + r) = 0.5174, and the within-group correlation coefficients are all greater than 0.060. The nested structure of the data cannot be ignored, and a multilevel mediation analysis is needed. Step 2, test the direct effect of the independent variable on the dependent variable c. The results show that self-efficacy can negatively predict cognitive load ( γ10c=−0.22, p < 0.001), indicating a significant effect c. Step 2, test the direct effect of the independent variable on the dependent variable c. The results show that self-efficacy can negatively predict cognitive load ( γ10c=−0.22, p < 0.001), indicating a significant effect c. Step 3, test the effect of independent variable on the mediating variable a. The results show that self-efficacy can negatively predict state anxiety (γ10a = − 0.53, p < 0.001), indicating a significant effect of a. Step 3, test the effect of independent variable on the mediating variable a. The results show that self-efficacy can negatively predict state anxiety (γ10a = − 0.53, p < 0.001), indicating a significant effect of a. Step 4, the effects of both the independent variable and the mediating variable on the dependent variable, c’ and b, were examined. The results indicate that self-efficacy and state anxiety simultaneously have a non-significant effect on cognitive load, c’ (γ10c’=-.13, p = .078), but the effect of b (γ20b = .21, p < 0.05) is significant.Based on previous research, when regression coefficients a, b, and c are all significant and the t-test for the estimated value of regression coefficient c’ does not reach significance level, it indicates that this mediating effect is a complete mediating effect. The size of the mediating effect is measured by ab (ra1 10 × rb1 20), and the mediating effect is 0.25. Based on the above results, a sobel test was conducted for the mediating effect. The results showed that the mediating role of state anxiety was significant (t = 2.25, p < 0.05). Detailed data can be found in Table 3. Step 4, the effects of both the independent variable and the mediating variable on the dependent variable, c’ and b, were examined. The results indicate that self-efficacy and state anxiety simultaneously have a non-significant effect on cognitive load, c’ (γ10c’=-.13, p = .078), but the effect of b (γ20b = .21, p < 0.05) is significant.Based on previous research, when regression coefficients a, b, and c are all significant and the t-test for the estimated value of regression coefficient c’ does not reach significance level, it indicates that this mediating effect is a complete mediating effect. The size of the mediating effect is measured by ab (ra1 10 × rb1 20), and the mediating effect is 0.25. Based on the above results, a sobel test was conducted for the mediating effect. The results showed that the mediating role of state anxiety was significant (t = 2.25, p < 0.05). Detailed data can be found in Table 3. Table 3A multilevel mediation model of self-efficacy, state anxiety, and cognitive loadModelParameter estimationγ00γ10γ20τ00σ2The null model6.27***3.58***0.82Level 1: CLij = β0j + rijLevel 2: β0j = γ00 + u0jSelf efficacy-Cognitive load7.53***−0.21***3.63***0.70Level 1: CL1ij = β0j + β1j*(SEij) + rijLevel 2: β0j = γ00 + u0jβ1j = γ10Self efficacy-State anxiety7.82***−0.53***1.69***1.07Level 1: SAij = β0j + β1j*(SEij) + rijLevel 2: β0j = γ00 + u0jβ1j = γ10Self efficacy、State anxiety-Cognitive load7.63***−0.130.21*3.74***0.66Level 1: CLij = β0j + β1j*(SEij) + β2j*(SAij) + rijLevel 2:β0j = γ00 + γ01*(SEj) + γ02*(SAj) + u0jβ1j = γ10β2j = γ20Note: ① *p < .05, **p < .01, ***p < .001. ② σ2 represents the within-group residual variance, i.e., the variance of γij; τ00 represents the between-group residual variance, i.e., the variance of u0j; γ00 is the overall intercept, while γ10 and γ10 are slopes (assumed to be fixed) A multilevel mediation model of self-efficacy, state anxiety, and cognitive load Level 2: β0j = γ00 + u0j β1j = γ10 Level 2: β0j = γ00 + u0j β1j = γ10 Level 2:β0j = γ00 + γ01*(SEj) + γ02*(SAj) + u0j β1j = γ10 β2j = γ20 Note: ① *p < .05, **p < .01, ***p < .001. ② σ2 represents the within-group residual variance, i.e., the variance of γij; τ00 represents the between-group residual variance, i.e., the variance of u0j; γ00 is the overall intercept, while γ10 and γ10 are slopes (assumed to be fixed) Discussion College students often face situational challenges such as exams and competition. These situations are significant for students, making it more likely for fluctuations in self-efficacy to occur during the process, thus affecting their current psychological and behavioral performance [75]. This study investigated the dynamic calibration mechanism of self-efficacy on cognitive load during task execution and revealed that state anxiety plays an intermediary role in this mechanism. Specifically, self-efficacy in tasks can be enhanced through positive feedback, thereby reducing state anxiety, directing cognitive resources toward the current task, and minimizing irrelevant resource consumption [76]. Study1 utilized survey and interview methods to investigate the possible causal relationships and mediating trends. The survey questionnaires investigated the dynamic calibration of self-efficacy phenomena and their sustained impact on emotions and subsequent tasks. The preliminary results suggest that self-efficacy influences the dynamic calibration of cognitive load through the generation of emotions. In-depth interviews reveal that negative emotions often play a significant role in the impact of self-efficacy and cognitive load. Anxiety was the most frequently reported negative emotion among the participants [77, 78]. Self-efficacy may have an impact on cognitive load through the mediation of anxiety, which aligns with previous research findings [79]. Feedback is a crucial method for individuals to receive experience of success and failure, and to change their self-efficacy [80, 81]. Negative feedback is related to reduced self-efficacy, whereas positive feedback and social support are related to increased self-efficacy [82, 83], even if the efficacy of feedback information is manipulated by the experimenter (rather than reflecting the actual performance of the participants). Study 2 employed false feedback to provide subjects with varying experiences of success and failure, effectively manipulating individual self-efficacy, thus establishing a manipulation paradigm for future research on self-efficacy. This further emphasizes the importance of feedback in task execution [18]. In addition, study 2 takes a micro-dynamic perspective and uses a multilevel linear model to separate the effects of individuals and measurement time on each variable. Thus, the mediating effects were further verified, addressing the shortcomings of simple mediating effects that easily overlook organizational-level (individual) independent variables, thereby enhancing the explanatory power of the mediating effects [84, 85]. In the measurement time dimension, improvements in self-efficacy were accompanied by a decrease in cognitive load. At the individual level, the individual differences were small, indicating that this phenomenon was relatively common. The outcome substantiates the Predictive and Reactive Control Systems (PRCS) theory, providing preliminary validation that individuals’ attention resource allocation is dynamically changing when self-efficacy changes. This study not only found that such changes can help individuals to effectively complete current tasks with fewer resources but also that resource allocation is calibrated based on changes in self-efficacy as a clue to adjust the attention resource investment required to perform tasks (i.e., cognitive load) [86]. Therefore, students need to maintain a strong sense of efficacy when facing tasks in order to focus their attention resources on the task effectively. In line with social cognitive theory [87], the research findings indicate that self-efficacy can influence an individual’s emotional state, thereby affecting their resource allocation, thus reaffirming the core role of self-efficacy mechanisms in personal cognitive processes. The anxiety emotions generated by individuals during task execution are similar to self-efficacy feelings and undergo dynamic changes in response to positive or negative feedback [88]. Self-efficacy can negatively predict state anxiety, meaning the higher the self-efficacy, the lower the anxiety individuals feel in the current situation, and conversely, they experience higher anxiety. Self-efficacy theory provides an explanation for this phenomenon, suggesting that anxiety stems from a sense of inefficacy in dealing with potential threats behaviorally and cognitively [89]. In this study, individuals experience a decrease in self-efficacy due to failure, leading to the onset of anxiety. State anxiety fully mediates the impact of self-efficacy on cognitive load. Intrinsic cognitive load, extraneous cognitive load, and germane cognitive load are the three components of cognitive load theory [5, 90]. Intrinsic load is related to the task at hand, and germane cognitive load is triggered by the construction and automation of schemas [91]. External cognitive load is the mental activity that occurs when individuals process irrelevant information that is not related to the task at hand. During the experimental process, the decrease in total cognitive load of the participants is triggered by the reduction of external cognitive load. This is because the experimental design within the participants maintained consistent task difficulty and context. The sole variable subject to change is possibly emotion. Thus, the impact of self-efficacy on cognitive load can only be mediated through the alteration of emotions. This supports the idea that anxiety, as a resource, can influence an individual’s cognitive processes during task performance [92]. This suggests that emotions play an important role in the mechanism of self-efficacy and cognitive load. Previous studies have mostly focused on exploring the static mediation relationship between self-efficacy, state anxiety and cognitive load [93, 94]. Overall, a significant characteristic of this study, in comparison with existing research, is the more comprehensive and thorough investigation of the relationships between self-efficacy, state anxiety, and cognitive load from a dynamic perspective. This study combines multiple research methods to explore the impact of self-efficacy on cognitive load from different perspectives, enhancing the credibility of the research results. It serves as a positive supplement to existing studies in the field of self-efficacy, with implications for understanding how self-efficacy affects an individual’s cognitive load. This result holds significant importance in addressing how to minimize the negative impact of confidence fluctuations on task resource allocation when facing important tasks. Nevertheless, some limitations will be addressed in future research to this study. Firstly, the participants in this study were all college students. In examining demographic variables, it was found that there were no significant differences in the experimental results based on gender and age, which may be attributed to the relatively small sample size. Future research could expand the sample size of participants and collect more demographic variables for analysis. Secondly, this study only verified the impact of individual state anxiety and cognitive load when self-efficacy is increase, and has not yet started from the direction of reducing self-efficacy, and future research will employ eye tracking and brainwave measurement to assess cognitive load more accurately. Finally, existing studies have shown that with an increase in self-efficacy, the degree of individual engagement and persistence in tasks will also increase [95]. It suggests that self-efficacy may positively predict cognitive load in certain circumstances. So, during task execution, how does the overall impact mechanism of self-efficacy on cognitive load operate? In the future, we will incorporate neurophysiological indicators such as eye track and electroencephalography in our research to obtain more objective evidence, thereby enhancing the self-efficacy theory. Conclusion This study explores the longitudinal mediating relationships among self-efficacy, state anxiety, and cognitive load in the task execution process. First, Study 1 provides robust interview evidence. The interview results revealed a potential mediating variable between self-efficacy and cognitive load, namely, state anxiety. Furthermore, we conducted Study 2 which employing experimental methodology to provide empirical evidence in support of the conclusions drawn in study 1. First of all, the present investigation found that negative feedback can reduce self efficacy, positive feedback can increase self efficacy. Therefore, the false feedback paradigm has been verified as an effective paradigm for manipulating self-efficacy. Moreover, improvements in self-efficacy at the measurement time points could significantly predict a decrease in cognitive load, and state anxiety fully mediates between self-efficacy and cognitive load. Supplementary Information Supplementary Material 1 Supplementary Material 2 Acknowledgements We would like to express our gratitude to all the participants and NimTottenham for help in the study. Authors’ information The author is from the School of Education Science at Guangxi Minzu University, China. Authors’ contributions Xiaolong Li and Mu Xia wrote the main manuscript text. All authors reviewed the manuscript. Funding This study was supported by the Guangxi Scholarship Fund of Guangxi Education Department(GED)“Research on the Influence Mechanism of Self-efficacy Calibration on Middle School Students’Mathematical Test Anxiety” (FIN301640601) and the 2021 Guangxi Degree and Graduate Education Reform Project “To Enhance Self-efficacy,” funded by the Research on the Leading Graduate Training Model. (Grant No. JGY2021069). Data availability The datasets generated and/or analysed during the current study are not publicly available due to the protection of individual students participating in the study but are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This Project has been supervised and approved by Institutional Review Board of College of Education Science, Guangxi University for Nationalities (IRB No. Approved: 23–0067 ). All participants have signed informed consent forms. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement The datasets generated and/or analysed during the current study are not publicly available due to the protection of individual students participating in the study but are available from the corresponding author on reasonable request.