Pilot Study on the Relationship of Test Anxiety to Utilizing Self-Testing in Self-Regulated Learning

Whether or not test-anxious students leverage the power of testing as potent learning tool is unclear. In a pilot study we investigated the relation of test anxiety to the utilization of testing activities and academic performance in self-regulated learning. We hypothesized that increased cognitive test anxiety would relate to less self-reported use of self-testing in favor of repetition strategies, and thus relate in turn to lower self-reported exam grades. To examine this idea, we created a scale contrasting self-testing and repetition strategies, which showed sufficient preliminary reliability and validity. The findings support our notion with respect to the cognitive interference component of test anxiety: More interference was associated with less self-testing, and the link of interference with exam grades was fully mediated by the reported degree of self-testing. Although our findings are preliminary and limited in generalizability due to small sample size and lack of factor analysis of the created scale, the results hint at one potential reason why test-anxious students may underperform. Consequently, educators might motivate their test-anxious students to rely more on effective study approaches.


Introduction
Recitation, testing-effect, practice testing, retrieval practice, and test-enhanced learning are all labels pertaining to the same remarkable memory phenomenon: actively recalling rather than passively repeating information aids in long-term retention (e.g., McDaniel, Roediger, & McDermott, 2007).Consequently, retrieval beats restudying, and implications are clear: learners should apply re-testing instead of re-reading when information is to be remembered long-term.From a practical point of view, the phenomenon's effectivity (meta-analytic average weighted g = .61;Adesope, Trevisan, & Sundararajan, 2017;g = .50;Rowland, 2014) and broad applicability to educational settings makes it a promising candidate to boost learning throughout all educational sectors (Roediger & Karpicke, 2006).
The testing-effect is empirically-well supported and robust across varying study conditions (e.g. with and without feedback; Adesope, Trevisan, & Sundararajan, 2017).Although some researchers argue that testing is not beneficial for complex materials (Van Gog & Sweller, 2015), this concern is not shared by others (Karpicke & Aue, 2015).Even transfer and deductive inferences can be fostered under certain circumstances (Eglington & Kang, 2016;Smith & Karpicke, 2014).Moreover, testing holds indirect advantages, such as monitoring what has and has not been learned to guide future studying (Dunlosky, Kubat-Silman, & Hertzog, 2003), and it facilitates the acquisition of future material (Pastötter & Bäuml, 2014).Consequently, multiple researchers emphasize testing's value for learning, and they advocate a move away from a means of formal assessment to a potent learning tool (e.g.McDaniel, Roediger, & McDermott, 2007;Roediger, Agarwal, Kang, & Marsh, 2010).
The advantage of testing however can be attenuated when testing is accompanied with performance pressure during retrieval (Hinze & Rapp, 2014), and the mnemonic effectivity seems to be reduced for learners with test-anxiety ( (Tse & Pu, 2012).Test anxiety as state and/or trait phenomenon occurs before and during test-taking (for a review see Zeidner, 2014) and encompasses negative physio-emotional and cognitive reactions like worry (concerns about one's test performance), lack of confidence (low self-expectations), and interference (intrusive and disturbing thoughts that disrupt task-specific information processing; cf.Hodapp, Rohrmann & Ringeisen, 2011).To which degree learners with test anxiety may choose to utilize the testing effect in their self-directed studying has not yet been directly investigated.
Research on the testing-effect has only recently begun to stimulate an interest in self-regulated aspects of learning.Within this new research avenue, the link of engaging in testing activities to academic achievement is underexplored.The integration of test anxiety's influence on the link between self-testing (Note 1) and achievement is pending.Thus, it is not only unclear to which degree students with higher levels of test anxiety use self-testing, but also how this translates to their academic achievement.Until now, this relationship was indirectly covered by broader research on learning strategies, without explicitly focusing on self-testing.The goal of the present study is to twofold.First, we aim to directly investigate whether test anxiety relates to the self-reported use of self-testing activities.Second, we examine whether and how test-anxiety may be related to exam performance via self-testing activities.

Effectivity and Use of Testing in Self-Regulated Contexts
The benefits of testing for personal learning may be crucially tied to the learner itself, because when organizing the 'how-to' of studying, the learner may insufficiently apply self-testing.Students learning activities often involve extensive highlighting and underlining, note taking, summary writing, and frequent re-reading (e.g.Dunlosky, Rawson, Marsh, Nathan, & Willingham, 2013;Hartwig & Dunlosky, 2012).The activities also encompass flashcards, quizzes, answering test questions, solving exercise problems, and completing practice exams (e.g.Karpicke, Butler, & Roediger, 2009).Both sets of activities can be seen to reflect a decision to either repeat or retrieve information, and of course, repetition and testing are often alternated.Moreover, retrieval or testing is often used to check one's understanding (Karpicke, Butler, & Roediger, 2009).Which of these study approaches do students report to utilize more, and how does it affect memory and performance?Self-testing in flashcard-style is favored over re-reading for simple materials (Wissman, Rawson & Pyc, 2012) and a half-half split was effective for memory (Kornell & Son, 2009).Students often start self-testing later in the learning phase and stop prematurely, which negatively affected long-term memory, while repetitions were unrelated (Karpicke, 2009).For more complex content, re-rereading is the most frequent strategy (Karpicke, Butler, & Roediger, 2009).Self-testing (quizzes/practice testing) is also applied (Kornell & Bjork, 2007), although the amount varies up until a half-half split (Bartoszewski & Gurung, 2015;Susser & McCabe, 2013).Nevertheless, self-testing is never the prime method.While greater levels of self-testing with practice problems and quizzes were positively related to higher academic achievement (Gurung, 2005;Hartwig & Dunlosky, 2012;Morehead, Rhodes, & DeLozier, 2016), re-reading was not (Bartoszewski & Gurung, 2015).Note, the association of self-testing and academic achievement was weak; nevertheless, a reduced frequency of self-testing in favor of re-reading should decrease its merit for learning and achievement.The question then arises: How might test anxiety affect decisions for self-testing or repetition activities and achievement?

Test Anxiety, Study Strategies, and Academic Performance
If a learning situation is seen as threatening by test anxious-students, they may adopt rote learning (Fransson, 1977) and less deep learning (Helmke & Tuyet, 1999).Deep learning strategies foster a thorough understanding, while surface learning strategies aid rote learning (e. g.Biggs, 2003).Deep-learning can be linked to better grades (Zhang, 2000; but see also Campbell & Cabrera, 2014).Moreover, direct detrimental effects of test anxiety on performance have been shown by a plethora of studies (Hembree, 1988;Seipp, 1991;overall r = -.20.).Thereby, the meta-analytic association between performance and worry (r = -.29) is much stronger than between performance and emotionality (r = -.15;Seipp, 1991).The cognitive aspects of test-anxiety play an important role in the explanation of the negative anxiety-performance link: either short-term by working memory impairments in the test situation or by mechanism unfolding long-term like avoidance of evaluative situations and study deficits (cf.Carey, Devine, & Szücs, 2016), as described in the following.

Test Anxiety's Influence on the Use and Effectivity of Self-testing and Achievement in Self-Regulated Learning
Test anxiety is characterized by a propensity for higher degrees of worry, self-doubt, and interfering thoughts.If a highly test-anxious learner is engaging in self-testing, older interference accounts postulate cognitive impairments by intrusive worries evoked during the self-test situation, which hinder retrieval (Sarason, 1980) and attentional allocation (Wine, 1971).Newer accounts hold that trait test anxiety interferes with attentional control (Eysenck, Derakshan, Santos, & Calvo, 2007) and efficient cognitive processing in working memory, especially with inhibition (Berggren & Derakshan, 2013).Since self-testing entails retrieval practice and retrieval itself is impaired by unwarranted thoughts and worries due to working memory taxation (Hayes, Hirsch, & Mathews, 2008), retrieval practice itself becomes less effective, if not compensated by higher working-memory capacity (Tse & Pu, 2012;Owen, Stevenson, Hadwin, & Norgate, 2012).Simultaneously, interfering thoughts reduce the cognitive capacity available to work on the test problems at hand, acting like extraneous cognitive load leaving less working-memory capacity to the processing of the concurrent learning task (Richard, French, Keogh, & Carter, 2000;Sweller, Ayres, & Kalyuga, 2011).Moreover, test anxiety can negatively impact performance by greater distractibility to task-irrelevant material (Keogh & French, 200;Unsworth, Spillers, Brewer, & McMillan, 2011).Deficient attentional and cognitive processing as well as increased distractibility therefore reduce the efficiency in performing the ongoing self-testing activity.Cognitive processing inefficacy may go along with greater retrieval problems and difficulties to correctly answer test questions, which may stimulate premature termination of self-testing.
Experiences difficulties or incorrect answers can feed negatively into self-perceptions (Schunk & Ga, 1982), triggering avoidance motivations with respect to the failure-inducing activity (Heimpel, Elliot, & Wood, 2006;McGregor & Elliot, 2002).The result would be a lower propensity to engage in further self-testing and to switch to other learning strategies, such as repetition (Wittmaier, 1972).For example, some studies have demonstrated that highly test anxious learner study extensively but often use highly repetitive activities that are less effective (Cassady, 2004;Culler & Holohan, 1980).
The adoption of less effective study habits intersects with deficit accounts (Birenbaum, & Nasser, 1994;Naveh-Benjamin, McKeachie & Lin, 1987), which emphasize impairments related to domain-specific abilities and study habits in the learning phase (for a discussion of deficits during test preparation phase see also Cassady, 2004).The initial acquisition of the learning content is lessened due to cognitive interference when uncompensated by higher abilities, impairing conceptual encoding and representations of the learning materials (Mueller & Courtois, 1980;Naveh-Benjamin, 1991).The consequence is a deficient mastery of the content (Birenbaum & Pinku, 1997), which would be revealed in future self-tests.As students often use self-testing to diagnose their learning instead of improving it (Kornell & Son, 2009), the subsequent utilization of self-testing would heighten one's meta-cognitive awareness of poor learning.For test-anxious students, increased meta-cognitive awareness could heightening the saliency of failure, which may be experienced as an aversive cognition.The meta-cognitive awareness of a potentially upcoming failure intensifies the threat-potential to one's self-efficacy and thus lowers motivation to self-test to protect one's self-concept (Hancock, 2010;Roick & Ringeisen, 2017).In this respect, learners with lower competence expectations for successful performance tend to adopt avoidance goals perceiving evaluative situations as threatening events to be avoided (Cassady, 2004)."The perception of a testing event as threatening impacts student performance in the preparation phase generally through the engagement of ineffective preparatory strategies driven by feelings of helplessness, attempts to avoid or reduce the occurrence of failure […].Naturally, ineffective strategies during the preparation phase will lead to eventual failure in test performance."(Cassady, 2004, p. 572).
As the previous literature on self-testing and memory suggests, premature termination and insufficient use or processing efficiency as well as avoidance of self-testing lowers its effect on memory and performance.This is especially detrimental, as highly test-anxious students would need, as compensation, higher frequencies of self-testing rather than less.A deficient mastery of the learning contents may perpetuate lower domain-specific abilities (Cassady & Johnson, 2002), feeding into further avoidance of test events.
In summary, the previous argumentation stresses deficits related to the cognitive component of test-anxiety (interference, worries, low confidence).On the one hand, it suggests that test anxiety may be associated with cognitive deficits during self-testing.On the other hand, it suggests that text anxiety is associated with further deficits in the study approach (for example, an avoidance of self-testing, premature termination of self-testing and the use of repetition strategies).Therefore, the effectivity of self-testing and the degree of usage as a study strategy should be related to the cognitive component of test anxiety.Less usage and an underutilization or adoption of alternative repetition approaches should be negatively related to exam performance.We hypothesize the following: 1) students with higher levels of cognitive test anxiety should be less prone to use self-testing as personal study strategy; 2) students with higher levels of cognitive test anxiety should have worse exams grade due to decreased self-testing (mediation).
To examine our hypotheses, we created a scale that covers the various ways (e.g., collaborative testing) that students might apply testing when learning; there was no ready-to-use instrument since the exploration of testing in self-regulated learning is a new research line.Thus, like the few existing surveys on the use of testing as personal learning tool, we had to create our own items.We constructed the items in line with the scheme of laboratory research, contrasting the testing condition with the repetition condition (e.g., re-reading or summaries).Although we were inspired by existing items in previous surveys (see Introduction), we had to compile or own, because not all formulations were specific enough, for example, 'make and use flashcards,' or 'do practice problems' (Susser & McCabe, 2013) does not specify the use as a test tool.'Use' could also be understood as inscribing flashcards with short summaries to re-read.'Do' does not exclude the option of preparing with worked-out practice problems that are already solved (like case studies).

Participants and Procedure
Participants (N = 81; 27 males; average age of 23.62, SD = 4.97; 60% were in the 2 nd semester/freshman) signed up for a study advertised to be about learning strategies, personality, and exam grades via the internal study platform (SONA) at the University of Kassel (Germany).After registration, participants received an internal online link to the survey platform, where participants completed the survey.Almost all participants were psychology students (n = 73) and were compensated with partial course credit.
At the beginning, participants were asked to list all exams they completed in the previous semester, which was intended to help students to remember their study behaviors; they were also asked to record the achieved exam grades at the end of the study.We then inquired about learning strategies via our constructed self-testing scale (ST-scale, see below), followed by other items (e.g., generation, spaced/interleaved learning), which we included for distraction (and for reasons not relevant to the present study: replication of a previous published relationship of NFC and goals orientations with learning activities by the first author).Thereafter, items of the German Learning Strategy Inventory LIST (Schiefele & Wild, 1994) followed, which served to validate our newly constructed scale.We embedded the measurement of students' test anxiety (PAF; Hodapp, Rohrmann & Ringeisen, 2011) within other questionnaires (e.g.need for cognition, goal orientations).Finally, we listed all exams participants had indicated at the beginning of the study.We emphasized that it is important for the study to completely and truthfully enter received exam grades, or to indicate if the grade was not yet known to them.The survey concluded with demographic questions.

Measures
Self-testing scale (STS).We constructed various items to assess the degree to which students used self-testing or relied on re-reading/non-testing activities (see Table 1 in the Appendix for the items; Cronbach's α = .76).For example, "I used quiz and exercise questions at the end of a chapter to… A) re-read the respective text passage B) simulate an exam."All items were measured on 7-point scales (-3 = only A, -2 = mostly A, -1 = rather A, 0 = A equally as B, 1 = rather B, 2 = mostly B, 3 = only B) with an additional answer option of 'neither A nor B' (coded as missing).The items were constructed to contrast two learning strategies: A(= repetition/non-testing) and B(= self-testing) mirroring experimental research contrasting rereading or other activities (e.g.note-taking: Rummer, Schweppe, Gerst, & Wagner, 2017) with testing (cf.Kornell, Rabelo, & Klein, 2012).Due to some of the items being reversely formulated, they were recoded in a way that -3 to -1 represented the degree of re-reading/non-testing, while 1 to 3 represented the degree of self-testing.Zero served as the midpoint and represented the use of both strategies to a similar extent.The items were presented in randomized order in one block.
For each person, we computed the mean grade of all reported exam grades to represent the averaged exam performance in this semester.Note, most of the samples had completed three (n = 39) or four (n =11) exams, while 13 students had not completed any exams in the previous semester and therefore did not report any exam grades.The most frequently reported exams were developmental psychology (n = 38), general psychology (n = 42), and quantitative methods (n = 43).

Results
Prior to testing our hypotheses, we examined the validity of our newly created scale on the use of self-testing versus repetition/non-testing.We correlated (two-tailed) the sub-scales of the established learning strategy inventory (LIST; Wild & Schiefele, 1994) with our created instrument.It would speak to the scale's validity to observe a) a negative relationship with the LIST-subscale of repetition and b) a positive relationship to the sub-scales of monitoring and regulation, because students often use testing for monitoring reasons (Kornell & Son, 2009), which can help to regulate their learning (Fernandez & Jamet, 2017) as well as c) no relationship to organization, planning, or study time, which are aspects that tap more into management and scheduling.The results support the convergent and discriminant validity of our scale (see Table 1).We found the negative correlation with repetition (r = .37,p < .001)and the anticipated positive correlations with monitoring (r = .48,p < .001)and regulation (r = .23,p < .05).Our repetition vs. self-testing scale was not related to organization, planning, or study time, all ps = ns.Like the monitoring (r = -.46,p < .001).The regulation (r = -.43,p < .001)sub-scale (related to testing aspects) of the established inventory and our ST-scale (r = -.42,p < .001)were both associated with the averaged exam performance as outcome variable; the coefficients are comparable in size and direction, supporting the criterion validity.Note.Bivariate correlation coefficients are based on Spearman's Rho.† p = .10,* p < .05,** p < .01,*** p < .001(two-tailed).Grades: higher numbers denote worse averaged exam grades.
Given a satisfactory reliability (Note 2) score of α = .76and support for internal validity, we scrutinized the relationship between test anxiety, learning strategies, and exam performances.Table 2 shows the correlations (two-tailed) of the test-anxiety inventory and all test anxiety sub-scales with our self-testing scale (ST-scale) and its association with the averaged exam grades achieved.In line with Hypothesis 1, test anxiety indeed related to the degree to which students reported to utilize self-testing as a learning strategy: Overall increased levels of test-anxiety covary with a decreased use of self-testing in favor of repetition/non-testing strategies, r = -.23,p < .05.When considering the sub-dimensions of test anxiety, as expected, the interference component was moderately and negatively related to the ST-scale, r = -.40,p < .001,explaining 20% of the variance in the degree of reported self-testing.The more interference is experienced, the less self-testing is chosen as a learning strategy.The subscales worry, r = -.15,p = .18,and lack of confidence, r = -.18,p = .12,were not related to the ST-scale.On a descriptive level, however, they also show a similar trend.The subscale of emotionality was not associated with the ST-scale, r = -.04,p = .72.  (Hodapp, Rohrmann & Ringeisen, 2011) with all 4 sub-scales.b Test anxiety inventory without the sub-scale lack of confidence, because some models on the link of anxiety and performance see low competence as a cause of test anxiety and low performance (Carey, Hill, Devine, & Szücs, 2016).
When considering the relationship of test anxiety and learning approaches to the performance criteria, we found that the undifferentiated test anxiety inventory was unrelated to the averaged achieved exam grades in the previous semester, r = -.02,p = .87.Notably, when considering all subscales, only cognitive interference was associated with exam performances, r = .26,p < .05.As postulated, the more interference is experienced during exam preparation, the worse the averaged exam performances.However, unlike expectations, the cognitive components of worry and lack of confidence showed no associations.Thus, only the interference component will be considered for Hypothesis 2.
Hypothesis 2 assumes a mediational model by which the cognitive aspects of test anxiety, the ST-scale, and the exam performances are all intertwined.In accordance with this notion, a regression analyses with anxiety-driven interference should predict the degree to which self-testing is reported as a learning strategy during the exam preparation.In turn, the used learning approach should predict the average achieved exam performances (see Figure 1).Conducting regression analyses and a Zobel-test (Note 3), we examined whether a full or partial mediation model was supported.
The pattern of results suggested full mediation, supported by the Sobel-test statistic (Soper, 2018), Z Sobel = 2.24, p = .026(two-tailed).Higher reported levels of anxiety-induced interference are associated with lower degrees of reported self-testing, which is associated with worse academic achievement (represented by averaged exam grades in this semester).Consequently, Hypothesis 2 was partly supported with respect to the cognitive interference component (but not worry or lack of confidence).
Figure 1.Full mediation of test anxiety induced interference on averaged exam grades by self-testing.Self-testing scale: higher number represent more self-testing.Averaged exam grade left in its original form: higher numbers denote to worse exam grades.With respect to averaged exam grades, n = 62.

Discussion
Is test anxiety related to the degree of utilizing self-testing strategies (e.g., practice problems, mock exams, collaborative testing, retrieval practice) and achieved exam grades in self-regulated learning?We argued that cognitive aspects of test anxiety go along with multiple deficits while self-testing (cf.interference accounts; e.g.Eysenck, Derakshan, Santos, & Calvo, 2007) and throughout the learning phase (cf.deficit models, e.g.Birenbaum, & Nasser, 1994), in line with processual accounts (cf.Cassady, 2004).In short, cognitive interference while self-testing may reduce processing efficiency, for example by imposing an unfavorable cognitive load on working memory or deficient attentional regulation.Retrieval difficulties and errors as self-concept threatening meta-cognitive feedback of may lead to premature termination and an avoidance of self-testing.Moreover, worries and low competence beliefs in evaluative situations also trigger avoidance reactions of the anxiety inducing activity (self-testing).Consequently, students are less prone to use self-testing and may switch to less anxiety-evoking strategies like repetition, which may not prove as effective.
Based on the rationale above, we predicted that students with higher levels of cognitive test anxiety should report less utilization of self-testing as a personal learning strategy (H1) and as a result, report worse exams grades (H2).Supporting our notion, we found that test-anxious students with higher levels of cognitive interference reported less usage of self-testing and lower averaged exam grades.This relationship was fully mediated, meaning the reduced frequency of self-testing relative to less efficient activities like repetition accounted for the lower exam grades by higher anxiety-related interference.The reduced frequency of self-testing in favor of other activities may be one mechanism underlying the negative test anxiety-performance link.
We did not find a significant relationship between self-testing and the cognitive component lack of confidence nor between self-testing and the cognitive component of worry.This may be due to the small sample size and/or that these components of test anxiety may play a lesser role with less sizable influences.The small sample size is also a caveat with respect to the newly constructed scale, which was not sufficient to conduct factor analyses to explore its structure.However, the scale preliminarily shows sufficient reliability and good discriminant and convergent validity.The obtained relationship of our scale with averaged exam grades can also be interpreted as an indication of predictive validity.A deficient instrument would not meaningfully relate to the expected relationships.Although the scale has room for improvement, it demonstrates construct validity.The creation of this instrument was necessary, because no instrument exists (to our knowledge) that specifically focused on a variety of self-testing activities (e.g., quizzes/flashcards/test problems/mock exams/practice tests/collective testing) relative to repetition/non-testing activities students may use when learning.
An aspect worth mentioning concerns the replication of the association between self-testing and academic achievement in real-world educational settings.So far, few studies have assessed how self-testing with quizzes/practice problems and flashcards related to GPA (see introduction : Hartwig & Dunlosky, 2012;Morehead, Rhodes, & DeLozier, 2016).However, GPA reflects cumulated achievements for various academic performances, like essays, for which self-testing may not be useful.It is possible, therefore, that the effect of self-testing on GPA is underestimated.Actual exam grades may be more accurate, which to date, have barely been examined.In line with this idea, we found a stronger association (r 2 = .18)when comparing our obtained association to the previous studies with GPA (Hartwig & Dunlosky, 2012: r 2 ≈ .08;Morehead et al. 2016: r 2 ≈ .03)and exam performance in introductory courses (Gurung, 2005;Bartoszewski & Gurung, 2015; both r 2 ≈ .08).A comparison of both correlations indicates a small difference of Cohen's q = .17(Cohen, 1988).We sampled a wider range of exam performances and included multiple aspects (e.g., collaborative testing; Wissman & Rawson, 2018), with more specific questions about the use of self-testing.This may have potentially contributed to the improved association.However, it also could be simply methodologic (e.g., we have a small and homogenous sample of a selective group).
The (meta-analytic) weighted mean effect size g of testing compared to re-studying is .51,stemming from regulated contexts like laboratory or classroom (Adesope et al., 2017).A transformation of our obtained r to d yields ≈ .93(Cohen, 1988).Interestingly, the effect of self-testing may be higher in self-regulated learning contexts than in regulated contexts.To put it another way, for personal learning, self-testing could be even more important, but this is preliminary.
Another flaw of the present study concerns the correlational nature and the reliance on self-report: Students' retrospective judgments about learning activities may not accurately represent actual learning but remembered learning.Maybe test anxiety influences the remembered degree of self-testing and not the actual use.It is therefore necessary to observe actual choices of test-anxious students and effects on performance criteria, especially with respect to establishing the causal influence of the mediator.Nevertheless, the study is an important, albeit limited, attempt to explore whether test anxiety negatively influences the use of self-testing activities and academic achievements in self-regulated learning.
This notion is interesting for theoretical and practical reasons, for example, theoretically with respect to meta-cognitions.While students often tend towards overconfidence, this bias seems to be reduced in test-anxious students leaning towards under confidence (Miesner & Maki, 2007).This may have interesting consequences for calibration accuracy and ongoing meta-cognitive processes and subsequent self-regulation decisions (e.g., change study strategy, spent less time, etc.).Think-aloud protocols could help to illuminate dynamics between meta-cognitions and study decisions (cf.Fernandez & Jamet, 2017).These self-regulation processes may be one cause for the well-established negative test anxiety-performance link, which speaks to the practical reasons: advising students with test anxiety about how to improve their test preparation and their test-taking skills.
Could the decreased motivation to use self-testing on their own by test-anxious students be counteracted by making testing an integral part of a course?In the classroom, Agarwal and collequgues (2016) showed that practice testing can be especially advantageous for students with lower working memory.While Brewer and Unsworth (2012) did not obtain a moderation by working memory capacity, they found that weaker students with lower memory abilities and lower general-fluid intelligence profited more from testing (but see also Pan, Pashler, Potter, & Rickard, 2015).Moreover, Messineo, Gentile, and Allegra (2015) showed that the negative link of test anxiety on performance in class is stronger for re-reading materials; thus, re-testing acted as a buffer for more test anxious students.Testing must take place, however, under low-threat conditions, such as low performance pressure (Khanna & Cortese, 2016).
It may be worthwhile to offer voluntary, ungraded self-testing options, for example, by e-learning as an integral part of a course, and to encourage its usage by those with a propensity to avoid self-testing.In this respect, the benefits of self-testing should be emphasized for test-anxious students Note.Items were translated into English by the authors and originally in German language.The (R) indicates items that were reversely phrased.Higher numbers reflected self-testing, while lower numbers reflected repetition/non-testing, the 0 of the scale meant both strategies were used equally often.

Table 1 .
Bivariate Correlations, Means, and Standard Deviations of the Self-Testing Scale (STS) and the Learning Strategy Inventory (LIST)

Table 2 .
Bivariate Correlations, Means, and Standard Deviations of Test Anxiety Inventory, Test Anxiety Sub-scales, Self-Testing Scale (STS) and Averaged Exam Grades Note.Bivariate correlation coefficients are based on Spearman's Rho.* p < .05,** p < .01,*** p < .001(two-tailed).Grades: higher numbers denote worse averaged exam grades.a Test anxiety inventory . Promising training programs, coaching, or therapies in groups or individual settings exist and use biofeedback, relaxation, or skill-building interventions (for an overview see Von der Embse, Barterian & Segool, 2013), as well as mindfulness techniques