The more, the better? Learning with feedback and instruction
越多越好？通过反馈和指导学习

1.1. Strategy instruction
1.1.战略指导

Strategy instruction is regarded as an effective means to foster cognitive skill acquisition (Fyfe & Rittle-Johnson, 2016; Goldman, 1989). During strategy instruction the students are explicitly taught the subject matter on how to apply distinct strategies to support problem-solving. To this end, strategy instruction often encompasses the use of worked examples (e.g., Fyfe & Rittle-Johnson, 2016; Goldman, 1989; Kant et al., 2017; Salden et al., 2010). Students first receive a short conceptual introduction that provides the basic conceptual knowledge referring to the underlying principles and concepts of the problem (e.g., Ohm's law in the domain of physics; Renkl, 2005; Wittwer & Renkl, 2008). Afterwards, students are provided with worked examples portraying the different solution steps leading to the final solution whereby their design follows principles of example selection and sequencing (Atkinson et al., 2000; Kant et al., 2017; Renkl, 2014; van Gog & Rummel, 2010).
策略指导被认为是促进认知技能学习的有效手段（Fyfe &；Rittle-Johnson, 2016; Goldman, 1989）。在策略教学过程中，学生会明确学习到如何运用不同的策略来解决问题。为此，策略教学通常包括使用工作实例（如Fyfe & Rittle-Johnson, 2016；Goldman, 1989; Kant et al.,2017；Salden 等人，2010）。学生首先会收到一个简短的概念介绍，提供与问题的基本原理和概念相关的基本概念知识（例如，在 "欧姆定律 "领域中的欧姆定律）、物理领域的欧姆定律；Renkl, 2005；Wittwer & Renkl, 2008）。随后，为学生提供工作示例，描绘最终解决方案的不同解决步骤，学生的设计遵循示例选择和排序原则（Atkinson et al、2000; Kant et al.，2017 年；Renkl, 2014 年；van Gog & Rummel, 2010 年）。

From a cognitive load perspective, providing worked examples should reduce extraneous processing so that cognitive capacity is freed for germane cognitive processes, such as conceptual learning and elaboration (Kant et al., 2017; Renkl, 2014; van Gog & Paas, 2008). The effectiveness of instruction based on example-based learning is well-documented in several empirical studies (van Gog & Paas, 2008; see also Crissman, 2006; Hoogerheide & Roelle, 2020; Kirschner et al., 2006; Magliaro et al., 2005; Renkl, 1997, 2014, for empirical evidence). A recent meta-analysis in the domain of mathematics (Barbieri et al., 2023) also documented the benefits of worked examples compared to no-examples (g = 0.48), based on 181 effect sizes.
从认知负荷的角度来看，提供有效的示例应能减少无关的处理，从而将认知能力释放给重要的认知过程，如概念学习和阐述（Kant et al、2017; Renkl, 2014；van Gog & Paas, 2008）。一些实证研究充分证明了基于范例学习的教学效果（van Gog & Paas, 2008；另见 Crissman, 2006; Hoogerheide &；Roelle, 2020; Kirschner et al.,2006; Magliaro et al、2005；Renkl, 1997, 2014, 为经验证据）。最近在数学领域进行的一项元分析（Barbieri et al.

1.2. Feedback 1.2.反馈意见

In subsequent phases of skill acquisition, such as during practice phases, formative feedback is regarded as further fundamental instructional supplement to deepen students' learning (Graham et al., 2011, 2015; Hattie & Timperley, 2007; Narciss, 2012). Providing learners with feedback has shown to be effective for enhancing their performance (Hattie & Timperley, 2007) in both classrooms (e.g., Bangert-Drowns et al., 1991) and computer-based learning environments (e.g., Mertens et al., 2022; van der Kleij et al., 2015). Feedback can be considered as post-response instructional information about distinct aspects of a student's current performance or learning process to help the student to reduce discrepancies between the actual performance and the targeted performance (Hattie & Timperley, 2007; Narciss, 2008). It can be used to assess (e.g., in form of grades or scales), as well as to diagnose (e.g., whether a learning goal has been achieved), or to evaluate (e.g., the level of students' knowledge) performance regarding various success criteria (Hattie & Timperley, 2007; Kluger & DeNisi, 1996). The most important purpose of formative feedback is to guide the student to regulate and improve the further learning process (Hattie & Timperley, 2007; Narciss, 2008, 2012; Shute, 2008).
在技能学习的后续阶段，如练习阶段，形成性反馈被视为进一步深化学生学习的基本教学补充（Graham et al、2011，2015; Hattie &；Timperley, 2007; Narciss, 2012）。为学习者提供反馈已被证明能有效提高他们的成绩（Hattie & Timperley, 2007）、Bangert-Drowns et al、1991）和基于计算机的学习环境（例如，Mertens et al、2022; van der Kleij 等人，2015）。反馈可被视为学生当前表现或学习过程中不同方面的反应后指导信息，以帮助学生减少实际表现与目标表现之间的差异（Hattie &；Timperley, 2007; Narciss, 2008）。它既可用于评估（例如，以等级或量表的形式），也可用于诊断（例如，在诊断过程中，可以使用 "阈值 "或 "阈限 "等术语）。例如学生的知识水平）的表现（Hattie &；Timperley, 2007; Kluger & DeNisi, 1996）。形成性反馈最重要的目的是引导学生调节和改进进一步的学习过程（Hattie & Timperley, 2007）；Narciss, 2008, 2012；Shute, 2008）。

Regarding the feedback information provided by an external source (e.g., computer-based system), previous research investigated different types of feedback attributing different functions of feedback. Narciss (2012) differentiated between outcome-related feedback types, such as knowledge of result (KR), knowledge of correct response (KCR), or answer-until-correct (AUC) feedback (see also corrective feedback, Fyfe & Rittle-Johnson, 2016; Shute, 2008; lower-order feedback, Kuklick & Lindner, 2023; Mertens et al., 2022). These feedback types are related to technical characteristics of feedback to generate more correct responses. In order to improve incorrect responses, further information is often needed. Therefore, more elaborated feedback (EF) types are used that provide more complex feedback contents as compared to corrective or knowledge of result feedback (Kuklick & Lindner, 2023; Mertens et al., 2022; Narciss, 2012; Shute, 2008). The use of elaborated feedback, which in addition to information on the correctness of the answer provides specific guidance on how to close existing knowledge gaps, is regarded to further enhance the effectiveness of feedback (Brooks et al., 2019; Hattie & Timperley, 2007; Mertens et al., 2022; Shute, 2008).
关于外部来源（如基于计算机的系统）提供的反馈信息，以往的研究调查了不同类型的反馈，这些反馈具有不同的功能。Narciss （2012）区分了与结果相关的反馈类型，如结果知识（KR）、或 "答非所问 "反馈（另见矫正反馈，Fyfe &；Rittle-Johnson, 2016; Shute, 2008；低阶反馈，Kuklick & Lindner，2023；Mertens et al.,2022）。这些反馈类型与反馈的技术特点有关，以产生更多正确的回答。为了改进不正确的回答，通常需要更多的信息。因此，与纠正性反馈或结果知识反馈相比，更详细的反馈（EF）类型可提供更复杂的反馈内容（Kuklick &；Lindner, 2023; Mertens et al.,2022；Narciss, 2012；Shute, 2008）。 精心设计的反馈除了提供关于答案正确性的信息外，还提供关于如何弥补现有知识差距的具体指导，这种反馈的使用被认为可以进一步提高反馈的有效性（Brooks et al、2019; Hattie & Timperley, 2007; Mertens et al、2022；Shute, 2008）。

For instance, in their meta-analysis (435 studies, k = 994, N > 61,000 students), Wisniewski et al. (2020) investigated whether the type of feedback (reinforcement/punishment, corrective feedback, or elaborated feedback) moderated the effect of feedback on students’ learning. Overall, the authors found a medium effect (d = 0.48) of providing feedback on learning. This effect was more pronounced for elaborated feedback (d = 0.99, k = 42 studies) than for corrective feedback (d = 0.46, k = 238 studies), or reinforcement/punishment (d = 0.24, k = 39 studies). Relatedly, Mertens et al. (2022) documented a comparable pattern for computer-based feedback. Based on 163 effects from 77 experimental studies they investigated by conducting a network meta-analysis whether different feedback types (KR, KCR, AUC, and EF compared with a no-feedback control condition) had different effects on lower- (i.e., recall/recognition) and higher-order (i.e., transfer) learning performance. Their results show that elaborated feedback is most effective to support learning (on lower-order: g = 0.71, p < .01, as well as on higher-order: g = 0.46, p < .01). Correct response feedback turned out to be least effective (only on lower-order: g = 0.24, p = .026; higher order: g = 0.14, p = .345).
例如，在他们的荟萃分析（435 项研究，k = 994，N > 61,000 名学生）中，Wisniewski et al.(2020) 调查了反馈类型（强化/惩罚、纠正反馈或详细反馈）是否会调节反馈对学生学习的影响。总体而言，作者发现提供反馈对学习的影响为中等（d = 0.48）。与纠正性反馈（d = 0.46, k = 238 项研究）或强化/惩罚（d = 0.24, k = 39 项研究）。与此相关，Mertens 等人（2022） 记录了基于计算机的反馈的类似模式。基于 77 项实验研究中的 163 项效果，他们通过进行网络荟萃分析，研究了不同的反馈类型（KR、KCR、AUC 和 EF，与无反馈对照条件相比）是否对低阶（即回忆/认知）和高阶（即迁移）学习成绩产生不同的影响。他们的研究结果表明，精心设计的反馈能最有效地支持学习（在低阶学习方面）：g = 0.71，p < .01，高阶反馈也是如此：g = 0.46，p < .01）。 正确回答反馈的效果最差（仅对低阶反馈有效：g = 0.24，p = .026；高阶：g = 0.14，p = .345）。

1.3. Combining strategy instruction and formative feedback
1.3.策略指导与形成性反馈相结合

Given that strategy instruction and formative feedback separately are both effective to support students’ learning, they are often combined in educational practice (Magliaro et al., 2005; Rosenshine, 2008): Provided with strategy instruction, students should build a consistent mental representation of the problem situation and the corresponding solution strategies. Receiving feedback during practice phases should further support the application and transfer of the previously learned solution strategies. Alternatively, the interactive, two-feedback-loop (ITFL) model by Narciss (2008) describes the implementation of feedback as a reciprocal interplay between internal (learner) and external (feedback) factors (cf. instructional medium) that may affect the effectiveness of feedback (see also Butler & Winne, 1995; Hattie & Timperley, 2007; Hayes, 2012; Panadero & Lipnevich, 2022; Winne & Butler, 1994, for related assumptions). Thus, alternatively, if prior strategy instruction (external factor) is given in advance of a practice phase, students could be inclined to undervalue the particular task requirements in the practice phase (see also illusion of understanding, Renkl, 2014). These subjective representations of task requirements could result in lower levels of processing of the feedback, both on the cognitive, meta-cognitive, or motivational level (Narciss, 2008). Thus, the combination could result in worse performance compared to the single interventions. Whether the combination of strategy instruction and feedback contributes more or even less to learning than a single intervention is still an open question.
鉴于策略指导和形成性反馈都能有效地支持学生的学习，因此在教育实践中经常将两者结合起来（Magliaro et al、2005; Rosenshine, 2008）：通过策略指导，学生应建立起对问题情境和相应解决策略的一致心理表征。在练习阶段获得的反馈应能进一步支持先前所学解决策略的应用和迁移。另外，Narciss (2008) 的交互式双反馈回路（ITFL）模型将反馈的实施描述为内部因素（学习者）和外部因素（反馈）之间的相互影响。参见 Butler &；Winne, 1995; Hattie & Timperley, 2007；Hayes, 2012; Panadero &；Lipnevich, 2022；Winne & Butler, 1994，相关假设）。 因此，另一种情况是，如果在练习阶段之前进行事先策略指导（外部因素），学生可能会倾向于低估练习阶段的特定任务要求（另见理解错觉，Renkl, 2014）。这些任务要求的主观表征可能会导致认知、元认知或动机层面的反馈处理水平降低（Narciss, 2008）。因此，与单一干预措施相比，两者结合可能会导致成绩下降。与单一干预相比，策略指导和反馈的组合对学习的贡献是更大还是更小，这仍然是一个未决问题。

Related evidence can be found in the synthesis by Salden et al. (2010). The authors synthesized the effects of eight studies in which strategy instruction (i.e., worked examples) and tutored problem-solving comprising different feedback elements (e.g., prompts, immediate feedback, and context-sensitive hints) were combined. Across those studies, the findings indicated that adding strategy instruction to tutored problem-solving contributed to learning.
相关证据可参见Salden 等人（2010）的综述。作者综合了八项研究的效果，在这些研究中，策略指导（即工作示例）和由不同反馈元素（如提示、即时反馈和上下文相关提示）组成的问题解决辅导相结合。这些研究的结果表明，在辅导解题过程中加入策略指导有助于学习。

However, there are also contradictory findings. To explore the effects of differences of prior strategy instruction and feedback, Fyfe and Rittle-Johnson (2016) provided elementary students with strategy instruction about solving mathematical equivalence problems or with control instruction. Afterwards, in a practice phase, students were asked to solve 12 structurally similar mathematical problems. One half of the elementary students received corrective feedback after each mathematical problem whereas the other half did not receive feedback. The authors found a strategy-by-feedback interaction, as students only profited from strategy instruction when no feedback was provided (see also Fyfe & Rittle-Johnson, 2017, for similar findings). In Experiment 2, Fyfe and Rittle-Johnson (2016) provided all students with strategy instruction. Additionally, they varied whether students received no feedback, immediate feedback, or summative feedback on the practice tasks. Again, no feedback was more beneficial than providing feedback (see Shirah & Sidney, 2023, for related evidence; see also Wischgoll, 2017, for related findings in academic writing).
然而，也有相互矛盾的研究结果。为了探究事先策略指导和反馈的不同所产生的影响，Fyfe 和 Rittle-Johnson（2016）为小学生提供了解决数学等价问题的策略指导或对照指导。之后，在练习阶段，要求学生解决 12 个结构相似的数学问题。一半的小学生在每个数学问题后都收到了纠正反馈，而另一半学生则没有收到反馈。作者发现了策略与反馈之间的交互作用，因为只有在不提供反馈的情况下，学生才能从策略指导中获益（类似研究结果另见Fyfe & Rittle-Johnson, 2017）。在实验 2 中，Fyfe 和 Rittle-Johnson（2016）为所有学生提供了策略指导。此外，他们还改变了学生在练习任务中获得无反馈、即时反馈或总结性反馈的情况。同样，没有反馈比提供反馈更有益（相关证据见 Shirah & Sidney, 2023；另见Wischgoll, 2017，学术写作中的相关发现）。

Together, these findings provided initial evidence that combining strategy instruction and formative feedback may not necessarily contribute to learning. Instead, they suggest that combining strategy instruction and feedback can also reduce learning performance or at least it has no additive effects (see also Kluger & DeNisi, 1996; Narciss, 2008; Shirah & Sidney, 2023).
总之，这些研究结果提供了初步证据，表明将策略指导与形成性反馈相结合并不一定有助于学习。相反，这些研究结果表明，将策略指导和反馈结合起来也会降低学习成绩，或者至少不会产生叠加效应（参见Kluger &；DeNisi, 1996; Narciss, 2008；Shirah & Sidney, 2023）。

1.4. Overview of the present study
1.4.本研究概述

The main aim of the present study, that included three online experiments, was to investigate potential (additive or reduced) effects of combining strategy instruction and feedback on students' physics learning in online environments. Throughout the experiments, we used identical learning materials (i.e., strategy instruction, practice tasks with feedback) to be able to compare findings across our experiments. In a multimedia learning environment, the students were provided with strategy instruction based on worked examples for solving electric circuits problems (Ohm's law, see Hoogerheide et al., 2019). In the practice phase, the students answered several practice tasks, in which they applied Ohm's law (by calculating the electric current) to different electric circuits.
本研究包括三个在线实验，其主要目的是研究策略指导与反馈相结合对学生在在线环境中学习物理的潜在（增加或减少）影响。在整个实验过程中，我们使用了相同的学习材料（即策略指导、带反馈的练习任务），以便比较不同实验的结果。在多媒体学习环境中，我们根据解决电路问题（欧姆定律，见 Hoogerheide et al., 2019）的工作示例为学生提供策略指导。在练习阶段，学生们回答了几个练习任务，在这些任务中，他们将欧姆定律（通过计算电流）应用于不同的电路。

The goal of Experiment 1 was to replicate a potential instruction-by-feedback interaction effect (as Fyfe & Rittle-Johnson, 2016, found in the domain of mathematics) in the domain of physics with corrective feedback. We used a 2 × 2-factorial design, crossing the two factors strategy instruction (yes vs. no) and feedback (yes vs. no). Furthermore, based on the ITFL-model by Narciss (2008), we tested for potential explanations regarding the underlying cognitive (i.e., mental effort, subjective difficulty), meta-cognitive (i.e., monitoring accuracy) and affective-motivational processes (i.e., arousal) accounting for additive versus reducing effects of combining strategy instruction and feedback.
实验 1 的目的是在物理领域中复制潜在的教学与反馈交互效应（如Fyfe & Rittle-Johnson, 2016在数学领域中发现的纠正反馈）。我们采用了 2 × 2 因子设计，将策略指导（是与否）和反馈（是与否）这两个因素交叉在一起。此外，基于Narciss (2008)的ITFL模型，我们测试了潜在的认知（即：脑力劳动、主观难度）、认知（即：认知）、主观难度（即：主观难度）、主观难度（即：主观难度）、主观难度（即：主观难度）、主观难度（即：主观难度）等方面的潜在解释、我们对潜在的认知过程（即脑力劳动、主观难度）、元认知过程（即监控准确性）和情感动机过程（即唤醒）进行了测试，以了解策略指导与反馈相结合所产生的相加效应与相减效应。

Given that instruction-by-feedback interaction could depend on the elaboration of feedback (external factor; Narciss, 2008), Experiment 2 aimed to generalize the findings and examine the robustness of the obtained effects in Experiment 1 by using elaborated feedback. In Experiment 3, we varied the sequence in which strategy instruction and feedback were provided, that is a learning phase with strategy instruction followed by a practice phase including feedback versus a preponed practice phase including feedback followed by a learning phase with strategy instruction. This variation allowed us to investigate whether instruction-by-feedback interaction may be a result of different sequences of strategy instruction and feedback. Together, these experiments allow to rigorously examine the underlying mechanisms and generalizability of combination effects of strategy instruction and feedback.
鉴于指导与反馈之间的交互作用可能取决于反馈的详细程度（外部因素；Narciss, 2008），实验 2 的目的是通过使用详细的反馈来推广实验 1 中的发现并检验实验 1 中获得的效果的稳健性。在实验 3 中，我们改变了提供策略指导和反馈的顺序，即先进行策略指导的学习阶段，然后是包含反馈的练习阶段，与先进行包含反馈的预演练习阶段，然后是包含策略指导的学习阶段。这种变化使我们能够研究策略指导和反馈的不同顺序是否会导致指导与反馈之间的相互作用。通过这些实验，我们可以严格研究策略指导和反馈组合效应的内在机制和可推广性。

2.1. Research questions and hypotheses
2.1.研究问题和假设

We pre-registered our hypotheses on as predicted (see https://aspredicted.org/tn6nx.pdf). In line with previous evidence (Fyfe & Rittle-Johnson, 2016; Wischgoll, 2017), we hypothesized a main effect of strategy instruction on students’ learning (Hypothesis 1a), as strategy instruction should help students construct a solid mental representation of the subject matter which should be conducive to learning. Regarding the effectiveness of corrective feedback (Experiment 1), it was less clear what to expect. Based on previous research, we can assume that different types of feedback should be differently effective (see Experiment 2 for details). However, in previous studies (e.g., Fyfe & Rittle-Johnson, 2016) on combination effects no significant main effect of feedback was obtained (which is in contrast to the common literature on the effectiveness of feedback). Thus, we refrained from making a clear prediction regarding main effects of feedback on learning.
我们按照预测对假设进行了预先登记（见https://aspredicted.org/tn6nx.pdf ）。与之前的证据一致（Fyfe & Rittle-Johnson, 2016；Wischgoll, 2017），我们假设策略教学对学生的学习有主效应（假设1a），因为策略教学应有助于学生构建对主题内容的稳固心智表征，从而有利于学习。至于矫正反馈的效果（实验 1），我们的预期就不那么明确了。根据以往的研究，我们可以认为不同类型的反馈应该有不同的效果（详见实验 2）。然而，在以往关于组合效应的研究中（例如，Fyfe & Rittle-Johnson, 2016），没有发现反馈有显著的主效应（这与有关反馈有效性的常见文献相反）。因此，我们没有就反馈对学习的主要影响做出明确的预测。

Additionally, based on Fyfe and Rittle-Johnson (2016), we hypothesized an instruction-by-feedback interaction (Hypothesis 1b). Accordingly, the addition of feedback should impair the effectiveness of strategy instruction. We also explored, whether the findings on transfer performance were robust for the underlying learning outcomes (i.e., near and far transfer).
此外，根据Fyfe 和 Rittle-Johnson（2016），我们假设教学与反馈之间存在相互作用（假设 1b）。因此，增加反馈应该会损害策略教学的有效性。我们还探讨了有关迁移绩效的研究结果是否对基本学习成果（即近迁移和远迁移）具有稳健性。

In case we find an instruction-by-feedback interaction effect, we aimed at exploring the underlying processes of the potential instruction-by-feedback interaction from a cognitive, a meta-cognitive, and an affective-motivational perspective (Narciss, 2008). Following a cognitive perspective, we explored whether increases in cognitive load during the practice phase may mediate the effect of instruction-by-feedback interaction on learning. We therefore used subjective cognitive load ratings (i.e., mental effort, subjective difficulty) as potential mediators. Concretely, we assumed that students who received both instruction and feedback will experience an increase in cognitive load during the practice phase due to the dual support compared to students who received only one type of support (strategy instruction-only or feedback-only; Hypothesis 2a).
如果我们发现了教学与反馈的交互效应，我们将从认知、元认知和情感动机的角度（Narciss, 2008）来探索潜在的教学与反馈交互作用的潜在过程。从认知角度出发，我们探讨了在练习阶段认知负荷的增加是否会介导教学-反馈互动对学习的影响。因此，我们将主观认知负荷评级（即脑力劳动、主观难度）作为潜在的中介因素。具体来说，我们假设，与只接受一种支持（仅策略指导或仅反馈；假设 2a）的学生相比，同时接受指导和反馈的学生在练习阶段会因为双重支持而增加认知负荷。

Following a meta-cognitive perspective, we also investigated whether students' overestimations mediate this effect. We used monitoring accuracy ratings to calculate students’ overestimations: Students who rated their prospective performance as being better than their actual performance (self-estimated test performance > actual test performance) were defined as overestimated judgements. Specifically, we hypothesized that students who first received strategy instruction followed by feedback will perform worse than students who only received feedback due to their overestimation (represented by positive values of monitoring accuracy; Hypothesis 2b).
从元认知的角度出发，我们还研究了学生的高估是否对这一效果起到了中介作用。我们使用监测准确性评级来计算学生的高估：将自己的预期成绩评定为优于实际成绩（自我估计考试成绩 > 实际考试成绩）的学生被定义为高估判断。具体来说，我们假设先接受策略指导再接受反馈的学生会比只接受反馈的学生成绩差，因为他们高估了自己的成绩（以监测准确度的正值表示；假设 2b）。

Last, following an affective-motivational perspective, we assumed that arousal mediates the effect. We used subjective ratings of perceived arousal to test this assumption. Concretely, we hypothesized that students who first received strategy instruction followed by feedback in the practice phase will show lower arousal levels, resulting in poorer performance compared to students who received only feedback without any prior strategy instruction (Hypothesis 2c).
最后，从情感-动机的角度出发，我们假定唤醒对这一效应具有中介作用。我们采用了对觉察到的唤醒程度进行主观评分的方法来验证这一假设。具体来说，我们假设在练习阶段先接受策略指导再接受反馈的学生会表现出较低的唤醒水平，从而导致他们的学习成绩比那些只接受反馈而事先未接受任何策略指导的学生差（假设 2c）。

2.2. Method 2.2.方法

2.2.3. Materials 2.2.3.材料

The entire experiment was presented with the online survey tool SoSci Survey (https://www.soscisurvey.de/). Based on previous learning and test materials (Rhöneck, 1986; Urban-Woldron & Hopf, 2012; see also Hoogerheide et al., 2019; Ivanjek et al., 2021), all the materials were carefully adapted by two experienced subject matter experts in physics education.
整个实验使用在线调查工具 SoSci Survey（https://www.soscisurvey.de/ ）进行。基于以前的学习和测试材料（Rhöneck, 1986；Urban-Woldron & Hopf, 2012；另见 Hoogerheide et al.,2019；Ivanjek 等人，2021），所有材料均由两位经验丰富的物理教育专家精心改编。

2.2.3.1. Prior knowledge 2.2.3.1.先验知识

We used a pretest comprising eight multiple-choice items (e.g., “What do you know about the total current in a parallel circuit?”; Cronbach's α for Experiment 1–3: α₁ = 0.56, α₂ = 0.59, α₃ = 0.53) where students were asked to select one correct solution per item out of 4 possible options, adapted from commonly used test instruments in physics education to assess students' conceptual understanding of electrical circuits (Rhöneck, 1986; Urban-Woldron & Hopf, 2012; see also Hoogerheide et al., 2019; Ivanjek et al., 2021, for related approaches). For each correct answer, students received one point, yielding a maximum score of eight points. We transformed the sum score to a proportion score (range: 0–1) by dividing the achieved points by the maximum points.
我们使用了由八个选择题组成的前测（例如："你对并联电路中的总电流了解多少？"实验 1-3 的 Cronbach's α：α₁ = 0.56，α₂ = 0.59，α₃ = 0.53。53），要求学生从 4 个可能的选项中选择一个正确的解决方案，该测试改编自物理教育中常用的测试工具，用于评估学生对电路概念的理解（Rhöneck, 1986；Urban-Woldron & Hopf, 2012；另见 Hoogerheide et al.,2019; Ivanjek et al.）每答对一个问题，学生得一分，最高得分为八分。我们将总分除以最高分，将总分转换为比例分（范围：0-1）。

2.2.3.2. Strategy instruction
2.2.3.2.战略指导

In the learning phase, students were provided with either strategy instruction or no instruction. The main aim of the strategy instruction was to enable students to calculate currents in a parallel circuit on the basis of Ohm's law in the form of $I\mathit{=}\frac{\mathit{V}}{\mathit{R}}$ with I referring to current, V to voltage, and R to resistance. Students needed to develop a conceptual understanding of this subject matter to solve the subsequent practice tasks.
在学习阶段，学生要么接受策略指导，要么不接受指导。策略指导的主要目的是让学生根据欧姆定律，以 I 表示电流，V 表示电压，R 表示电阻的形式计算并联电路中的电流。学生需要从概念上理解这一主题，才能解决后续的练习任务。

The strategy instruction consisted of a multimedia presentation comprising conceptual information about the physical quantities “current”, “voltage”, and “resistance”, their relationship in electric circuits based on Ohm's law as well as how a parallel circuit is constructed. A graphic representation of an electric circuit served to illustrate these abstract physics concepts. To enhance the effectiveness of the multimedia presentation, we considered general (multimedia) principles, such as signaling (Mautone & Mayer, 2001; Richter et al., 2018), and added process-oriented information to enhance conceptual understanding (van Gog et al., 2006, 2008; see also Lachner et al., 2019; see Fig. 1a). In the worked example, the general solution steps were modeled to provide a strategy, how the current, displayed by the four ammeters in a parallel circuit (see Fig. 1b), can be determined by applying Ohm's law (see Hoogerheide et al., 2019).
教学策略由多媒体演示组成，包括有关物理量 "电流"、"电压 "和 "电阻 "的概念性信息，它们在电路中基于欧姆定律的关系，以及如何构建并联电路。电路的图示可用于说明这些抽象的物理概念。为了提高多媒体演示的效果，我们考虑了一般（多媒体）原则，如信号传递（Mautone & Mayer, 2001；Richter et al.,2018），并添加了面向过程的信息以增强概念理解（van Gog et al、2006、2008 ；另见Lachner et al、2019；见 图 1a）。在工作示例中，对一般求解步骤进行了建模，以提供一种策略，说明并联电路中四个电流表显示的电流（见 Fig.1b），可以通过应用欧姆定律来确定（见 Hoogerheide et al、2019）。

1. Download: Download high-res image (973KB)
   下载：下载高清图片 (973KB)
2. Download: Download full-size image
   下载：下载全尺寸图片。

Fig. 1. Excerpt of the Strategy Instruction (Translated). Note. 1a shows the excerpt of the strategy instruction about how a parallel circuit is constructed and how the currents in the parallel circuit can be calculated using Ohm's law. 1b shows the worked example applying this knowledge.
图 1。策略指导摘录（译文）。注释。 1a 显示的是有关如何构建并联电路以及如何使用欧姆定律计算并联电路中的电流的策略说明节选。1b 显示的是应用这些知识的工作示例。

In the no-strategy instruction conditions students received a control group instruction (see Fig. 2; see Fyfe & Rittle-Johnson, 2016, for similar approaches). We included a control instruction instead of comparing the strategy instruction to a plain control condition which did not receive anything, to avoid potential differences regarding time-on-task (Möckel et al., 2015). Differences in time-on-task may affect the overall amount of processing during the cognitive tasks, and potential differences in learning (Boksem et al., 2005, 2006; Langner et al., 2010; Lorist et al., 2000). Thus, holding time-on-task constant across condition can be regarded necessary to avoid potential confounds (Kovanović et al., 2015; Möckel et al., 2015).
在无策略教学条件下，学生接受对照组教学（见图 2；见）。2；类似方法见 Fyfe & Rittle-Johnson, 2016）。我们加入了一个对照指导，而不是将策略指导与不接受任何指导的普通对照条件进行比较，以避免任务完成时间方面的潜在差异（Möckel 等人，2015）。任务时间的差异可能会影响认知任务的整体处理量，以及潜在的学习差异（Boksem et al、2005, 2006; Langner et al、2010; Lorist 等人，2000）。因此，为了避免潜在的混杂因素，在不同条件下保持任务时间不变被认为是必要的（Kovanović et al、2015；Möckel 等人，2015）。

1. Download: Download high-res image (853KB)
   下载：下载高清图片 (853KB)
2. Download: Download full-size image
   下载：下载全尺寸图片。

Fig. 2. Excerpt of the Control Instruction for the No-Instruction Condition (Translated) Note. 2a shows the excerpt of the control instruction about effects of electric current. 2b shows the worked example illustrating this knowledge.
图 2。无指导条件下的控制指令节选（译文） 注意： 2a 显示的是有关电流效应的控制指令节选。2b 显示了说明这一知识的工作示例。

Furthermore, we included a control instruction instead of another active treatment to be able to reach conclusion about the effectiveness of the strategy instruction. When different aspects of the instructions vary, we could not say with certainty whether the strategy instruction is responsible for the effect (Renkl, 2014). That is why it is important to keep the control instruction as constant as possible to the strategy instruction. To control for time-on-task for all conditions and nevertheless to avoid producing another treatment effect, we created a control instruction which would not influence the effect on task performance regarding the content (since the required information to solve the tasks was not included) but kept all other factors as constant as possible to the strategy instruction.
此外，我们还加入了对照指导，而不是另一种积极的治疗方法，以便能够就策略指导的有效性得出结论。当指导的不同方面发生变化时，我们无法确定是否是策略指导产生了效果（Renkl, 2014）。这就是控制指令与策略指令尽可能保持一致的重要原因。为了控制所有条件下的任务完成时间，同时避免产生另一种处理效应，我们创建了一个控制指令，它不会影响任务内容对任务表现的影响（因为不包括解决任务所需的信息），但尽可能保持所有其他因素与策略指令一致。

The control group instruction was comparable to the strategy instruction regarding text structure, schematics to illustrate the described contents, worked example, and terminology. The control group instruction also comprised a multimedia presentation (see Fig. 2a) and a worked example (see Fig. 2b). The contents of the strategy instruction and the control group instruction were divergent as students who received the control instruction were not introduced to Ohm's law or to the calculation of currents in a parallel circuit. Instead, the control instruction provided information about the invention of the battery and the effects of electric current. How to calculate currents in a parallel circuit was not explained in the multimedia presentation or in the worked example. Thus, the control instruction was not target-oriented for the following practice tasks but was only provided due to time-on-task constraints (see also Fyfe & Rittle-Johnson, 2016).
对照组的教学在文本结构、说明内容的示意图、工作示例和术语方面与策略教学相似。对照组的教学还包括多媒体演示（见 图 2a）和工作示例。2a）和一个工作示例（见 图 2b）。策略指导和对照组指导的内容有所不同，因为对照组指导没有向学生介绍欧姆定律或并联电路中电流的计算。相反，对照组教学提供了有关电池发明和电流影响的信息。如何计算并联电路中的电流在多媒体演示和作业示例中都没有解释。因此，在接下来的练习任务中，控制指令并非以目标为导向，而是由于任务时间限制才提供的（另见Fyfe & Rittle-Johnson, 2016）。

2.2.3.3. Materials in the practice phase
2.2.3.3.实践阶段的材料

The practice phase comprised two isomorphic practice tasks and additionally corresponding feedback in the feedback conditions. We have chosen this design because first, previous research showed that a restricted number of practice tasks were sufficient to foster students' learning. For example, Hoogerheide et al. (2019), whose material we adapted, only implemented one practice task. This decision is in line with findings by van Harsel et al. (2020), who showed that more (i.e., eight) tasks do not necessarily lead to more learning than fewer (i.e., four) tasks. Additionally, the use of a restricted amount of closed-ended questions is often used for rapid assessments in online learning environments (cf. Kalyuga, 2006; Kalyuga & Sweller, 2004). Second, we aimed at reducing the drop-out risk (cf. attrition bias, Nunan et al., 2018) caused by the online study setting and the length of the questionnaire (Galesic, 2006; Hoerger, 2010). Thus, we decided to realize an effective and parsimonious implementation of guided practice, and used two practice problems. Students answered these two isomorphic (i.e., analogous problem-solving strategy, different surface features) practice tasks (Cronbach's α for Experiment 1–3: α₁ = 0.47, α₂ = 0.56, α₃ = 0.56) in which they had to determine the currents by applying Ohm's law. Therefore, students had to select the correct solution from four alternatives. The distractors were based on common misconceptions students hold regarding currents (see McDermott & Shaffer, 1992; Schecker et al., 2018, for examples). The performance on the practice tasks was again transformed to proportion scores (range: 0–1).
练习阶段包括两个同构练习任务和反馈条件下的相应反馈。我们之所以选择这种设计，首先是因为以往的研究表明，有限数量的练习任务足以促进学生的学习。例如，Hoogerheide 等人（2019）（我们改编了他们的教材）只实施了一个练习任务。这一决定与van Harsel 等人（2020）的研究结果一致，他们的研究表明，较多（即八个）的任务并不一定比较少（即四个）的任务带来更多的学习效果。此外，在线学习环境中的快速评估通常会使用数量有限的封闭式问题（参见：）。Kalyuga, 2006; Kalyuga & Sweller, 2004）。其次，我们的目标是降低辍学风险（参见自然减员偏差，Nunan et al、2018）造成的（Galesic, 2006；Hoerger, 2010）。因此，我们决定有效而简洁地实施指导性练习，并使用了两个练习题。学生回答这两个同构问题（即实验 1-3 的 Cronbach's α：α₁ = 0.47，α₂ = 0.56，α₃ = 0.56）。因此，学生必须从四个备选方案中选出正确的解决方案。干扰项基于学生对电流的常见误解（见 McDermott &；Shaffer, 1992; Schecker et al.,2018 为例）。练习任务的成绩再次转换为比例分数（范围：0-1）。

We have chosen the multiple-choice test format (i.e., for the practice and transfer tasks) for several reasons: First, previous research showed that multiple-choice questions are at least as effective as other formats in consolidation phases (e.g., short-answer test; Adesope et al., 2017; Yang et al., 2021) as well as in combination with prior instruction (Ozuru et al., 2007).
我们选择多项选择测试形式（即用于练习和迁移任务）有几个原因：首先，先前的研究表明，在巩固阶段，多项选择题至少与其他形式一样有效（例如，简答测试；Adesope et al、2017; Yang et al、2021）以及结合事先指导（ data-dl-uid="6">Ozuru 等人，2007）。

Second, multiple-choice questions are frequently used in online learning environments combined with feedback because multiple-choice questions provide an easy as well as time- and resources-saving opportunity to analyze students' answers objectively and automatically by a computer-based system (Epstein et al., 2002; Guo et al., 2014; Kuechler & Simkin, 2003; Shin et al., 2019). That allows to provide immediate formative feedback to all students simultaneously which can increase students’ self-regulation, learning gain, and motivation (Allen et al., 2016; Epstein et al., 2002; Guo et al., 2014; Shute, 2008). Thus, we used this advantage and provided students in the feedback conditions with immediate, corrective feedback by giving them information about whether their chosen answer was correct or not (Hattie & Timperley, 2007). Additionally, they received a brief justification why the selected answer was correct (see Fig. 3a for the positive feedback) or why the chosen answer was incorrect (see Fig. 3b for the negative feedback; KCR, Narciss, 2012). Through this short justification, students had the opportunity to expand or to consolidate their knowledge and to benefit from the feedback, even if they had chosen the correct answer.
其次，在线学习环境中经常使用与反馈相结合的多选题，因为多选题为计算机系统客观、自动地分析学生的答案提供了一个简便、节省时间和资源的机会（Epstein et al、2002; Guo et al、2014; Kuechler & Simkin，2003; Shin et al、2019）。这样就可以同时向所有学生提供即时的形成性反馈，从而提高学生的自我调节能力、学习收获和学习动力（Allen et al、2016; Epstein et al、2014; Shute, 2008）。因此，我们利用这一优势，向反馈条件下的学生提供即时的纠正性反馈，告诉他们所选答案是否正确（Hattie & Timperley, 2007）。此外，他们还收到了所选答案正确的简要理由说明（见 Fig. 3a 表示正面反馈）或所选答案不正确的原因（见 Fig.3b 为负面反馈；KCR，Narciss, 2012）。通过这个简短的理由说明，学生有机会扩展或巩固知识，并从反馈中受益，即使他们选择了正确的答案。

1. Download: Download high-res image (688KB)
   下载：下载高清图片 (688KB)
2. Download: Download full-size image
   下载：下载全尺寸图片。

Fig. 3. Example of Feedback for Practice Task 1 (Translated), Note. 3a shows the positive feedback students received if their chosen answer was correct. 3b shows the negative feedback students received if their chosen answer was incorrect.
图 3。练习任务 1 的反馈示例（译文），注： 3a 显示了学生在所选答案正确时收到的积极反馈。3b 显示了学生在所选答案不正确时收到的负面反馈。

2.2.3.4. Learning outcomes (posttest)
2.2.3.4.学习成果（事后测试）

Students' learning outcomes were assessed with four transfer tasks: two near transfer tasks and two far transfer tasks (McDonald's ω for overall transfer in Experiment 1–3: ω₁ = 0.57, ω₂ = 0.56, ω₃ = 0.49). We adapted the transfer instruments by Hoogerheide et al. (2019). To this end, the near transfer tasks were isomorphic to the problems from the practice tasks (i.e., “Determine the amperage by ammeters A1 to A4 in the electric circuit shown above”; Cronbach's α for Experiment 1–3: α₁ = 0.60, α₂ = 0.57, α₃ = 0.67) but based on different variable values. In the second task the circuit diagram was rotated. In the far transfer tasks, the students had to transfer their knowledge to different circuits with a different number of ammeters or different resistors: The first far transfer task was similar to the second near transfer task (rotated diagram) but with an additional ammeter. The second far transfer task seemed like the practice tasks and the first near transfer task with four ammeters but with a variable resistor. Since these two transfer items covered different facets of far transfer (additional ammeter, different type of resistance), it can be expected that the reliability would be insufficient (Cronbach's α for Experiment 1–3: α < 0.20). Therefore, we conducted multilevel analyses as the items were nested within person (see Section 2.3 Results).
学生的学习成果通过四个迁移任务进行评估：两个近迁移任务和两个远迁移任务（实验 1-3 中总体迁移的 McDonald's ω：ω₁ = 0.57，ω₂ = 0.56，ω₃ = 0.49）。我们改编了Hoogerheide 等（2019）的转移工具。为此，近距离转移任务与练习任务中的问题同构（即："确定安培力"）、"实验 1-3 的 Cronbach's α：α₁ = 0.60，α₂ = 0.57，α₃ = 0.67），但基于不同的变量值。在第二个任务中，电路图被旋转。在远距离迁移任务中，学生必须将所学知识迁移到具有不同数量电流表或不同电阻的不同电路中：第一个远距离迁移任务与第二个近距离迁移任务（旋转电路图）类似，但多了一个电流表。第二项远距离转移任务与第一项近距离转移任务的练习任务和带有四个电流表的任务相似，但带有一个可变电阻器。由于这两个转移项目涵盖了远距离转移的不同方面（附加电流表、不同类型的电阻），因此可以预计信度不够（实验 1-3 的 Cronbach's α：α < 0.20）。因此，我们进行了多层次分析，因为项目是嵌套在人中的（见2.3 结果部分）。

Similar to the practice tasks, for each transfer task, students had to select the correct solution from four alternatives. The distractors were based on common misconceptions students hold regarding currents (see McDermott & Shaffer, 1992; Schecker et al., 2018, for examples). Students received one point per correct answer, resulting in a total score of 4 points for the transfer tasks. Again, we transformed the total score to a proportion score (range: 0–1). To explore, whether our findings were robust among knowledge facets, we also provide separate explorative analyses for the near and far transfer test components.
与练习任务类似，在每个转移任务中，学生必须从四个备选方案中选出正确的解决方案。分散注意力的因素基于学生对电流的常见误解（见 McDermott &；Shaffer, 1992; Schecker et al.,2018 为例）。学生每答对一个问题得一分，因此转移任务的总分为 4 分。我们再次将总分转换为比例分（范围：0-1）。为了探究我们的研究结果在不同知识面之间是否具有稳健性，我们还分别对近距离和远距离迁移测试部分进行了探索性分析。

2.2.3.5. Perceived cognitive load
2.2.3.5.认知负荷

Students' perceived cognitive load was measured by subjective ratings of students‘ invested mental effort (“How much mental effort did you invest in studying the instruction/completing the task?”; Jacob et al., 2020; Paas, 1992), and their subjectively perceived task difficulty (“How difficult was it for you to understand the instruction/solve the task?”; DeLeeuw & Mayer, 2008; Jacob et al., 2020) after the learning phase and after the practice phase on a Likert-scale from 1 (not difficult at all) to 9 (very difficult).
学生的认知负荷是通过对学生投入的脑力的主观评价来衡量的（"你在学习指令/完成任务时投入了多少脑力？"；Jacob et al、2020；Paas, 1992），以及他们主观感受到的任务难度（"您理解指令/完成任务的难度如何？"；DeLeeuw & Mayer，2008；Jacob et al、2020），在学习阶段后和练习阶段后，采用李克特量表，从 1（一点也不难）到 9（非常难）。

2.2.3.6. Monitoring accuracy
2.2.3.6.监测精度

To investigate potential differences of students’ monitoring accuracy, students judged their expected performance on tasks about electric circuits (“How confident are you that you can solve tasks on the topic of electric circuits correctly?”; Eitel, 2016; Jacob et al., 2020; Nelson et al., 1994) on a scale from 0% (not at all confident) to 100% (very confident). Students rated their judgement of learning 3 times (i.e., before the pretest, after the learning phase, and after the practice phase). As in previous research, we operationalized monitoring accuracy as the difference between students' actual transfer performance and their prospective judgements of learning they have delivered after the learning phase (Eitel, 2016; Finn & Metcalfe, 2014). An overestimation of students’ judged performance occurred when the values were positive, an underestimation occurred when the values were negative. A value of zero indicated correct assessment.
为了探究学生监测准确性的潜在差异，学生们对自己在电路任务中的预期表现进行了判断（"你对自己能够正确解决电路主题任务有多大信心？"；Eitel, 2016；Jacob et al、2020；Nelson 等人，1994），评分标准从 0% （ 完全没有信心）到 100% （ 非常有信心）。学生们对自己的学习判断进行了 3 次评分（即在预测试之前、学习阶段之后和练习阶段之后）。与之前的研究一样，我们将监测准确性操作化为学生的实际迁移表现与他们在学习阶段后对所学知识的预期判断之间的差异（Eitel, 2016；Finn & Metcalfe, 2014）。如果数值为正，则高估了学生的评判成绩；如果数值为负，则低估了学生的评判成绩。数值为零表示评估正确。

2.2.3.7. Affective measures
2.2.3.7.情感措施

We used the affective slider by Betella and Verschure (2016) to assess students' perceived arousal on a continuous scale from 1 (sleepy/bored) to 9 (wide awake/concentrated) and their perceived pleasure on a continuous scale from 1 (sad) to 9 (happy) as potential proxies for students’ current affective states¹ while learning.
我们使用Betella 和 Verschure (2016)的情感滑块来评估学生的感知唤醒度，连续量表从 1（困倦/无聊）到 9（完全清醒/集中注意力）的连续量表来评估学生的唤醒感，以及从 1（悲伤）到 9（快乐）的连续量表，作为学生当前 情绪状态¹ 学习时的情绪状态。

2.2.3.8. Additional control measures
2.2.3.8.其他控制措施

2.2.3.8.1. Academic self-concept
2.2.3.8.1.学术自我概念

As students' academic self-concept could impact their performance, we assessed their self-concept in the domain of physics by using five items (e.g., “I am good at physics.”, “Physics is easy for me.”, adapted from Gaspard et al., 2015, 2018; McDonald's ω = 0.93) on a 4-point Likert-scale from 1 (does not apply at all) to 4 (fully applies).
由于学生的学业自我概念会影响他们的学习成绩，我们使用五个项目（如 "我擅长物理"、"物理对我来说很容易"，改编自Gaspard et al、2015, 2018; McDonald's ω = 0.在 1（完全不适用）到 4（完全适用）的 4 点李克特量表上，麦当劳的ω = 0.93。）

2.2.3.8.2. Intrinsic motivation
2.2.3.8.2.内在动力

We measured students' intrinsic motivation and interest in physics with five items (e.g., “I enjoy physics.”, “I am interested in physical topics.”, adapted from Gaspard et al., 2015, Gaspard et al., 2018; based on Pekrun et al., 2002; McDonald's ω = 0.89) on a 4-point Likert scale from 1 (does not apply at all) to 4 (fully applies).
我们用五个项目测量学生的内在动机和对物理的兴趣（如 "我喜欢物理"、"我对物理话题感兴趣"，改编自Gaspard et al、2015，Gaspard et al、2018；基于Pekrun et al、2002; McDonald's ω = 0.89），采用 4 点 Likert 标度，从 1（完全不适用）到 4（完全适用）。

2.2.3.8.3. Attention checks
2.2.3.8.3.注意检查

In our experiment, we included an attention check before and after the practice phase to control for students' attention while completing the online experiment and increase quality of the obtained data and implementation fidelity (see also Hauser & Schwarz, 2016; Oppenheimer et al., 2009, for related approaches). Students were asked to fill in two missing numbers in a very simple 4 × 4 Sudoku. Instructions about how to solve a Sudoku were provided beforehand.
在我们的实验中，我们在练习阶段之前和之后都进行了注意力检查，以控制学生在完成在线实验时的注意力，提高所获数据的质量和实施的可信度（另见Hauser &；Schwarz, 2016; Oppenheimer et al.,2009，了解相关方法）。学生们被要求在一个非常简单的 4 × 4 数独游戏中填入两个缺失的数字。事先提供了有关如何解数独的说明。

2.3. Results 2.3.结果

As effect size measures, we used partial ${\mathrm{\eta }}_p^2$ with qualifying values of ${\mathrm{\eta }}_p^2$ = .01, .06, .14 as small, medium, and large effects (Cohen, 2013). Moreover, we used an alpha level of α = 0.05. In cases of computing generalized mixed effect models, as estimation method, we used maximum likelihood (Laplace approximation), logit as the link function, and Akaike information criterion (AIC), Bayesian information criterion (BIC), and Log likelihood (LogLik) as goodness-of-fit method (cf. Bono et al., 2021).
在衡量效应大小时，我们使用了部分 ${\mathrm{\eta }}_p^2$ ，以 ${\mathrm{\eta }}_p^2$ = .01、.06、.14 作为小、中、大效应的限定值（Cohen, 2013）。此外，我们使用了 α = 0.05 的α水平。在计算广义混合效应模型时，我们使用最大似然法（拉普拉斯近似法）作为估计方法，使用 logit 作为链接函数，使用 Akaike 信息准则 (AIC)、Bayesian 信息准则 (BIC) 和 Log likelihood (LogLik) 作为拟合优度方法（参阅 Akaike 信息准则 (AIC)）。Bono et al、2021）。

2.3.4. Additional explorative analyses
2.3.4.其他探索性分析

To explore differences between near and far transfer, we additionally conducted explorative separate analyses for near and far transfer to investigate whether instruction and feedback had differential effects on near and far transfer. Therefore, we followed the same procedure but first with students' performance in the near transfer tasks as the dependent variable of the multilevel analysis and in a second separate multilevel analysis students’ performance in the far transfer tasks as the dependent variable.
为了探索近迁移和远迁移之间的差异，我们还对近迁移和远迁移分别进行了探索性分析，以研究教学和反馈是否对近迁移和远迁移产生了不同的影响。因此，我们采用了相同的程序，但首先将学生在近迁移任务中的表现作为多层次分析的因变量，然后在第二个单独的多层次分析中将学生在远迁移任务中的表现作为因变量。

Regarding students near transfer performance as the outcome variable, the results of the multilevel analysis showed that there were no significant main effects neither for strategy instruction nor for feedback, as well as no significant interaction effect (see Table 2). We did not obtain significant differences among conditions, F(3, 433) = 1.51, p = .212, ${\mathrm{\eta }}_p^2$ = .01.
关于作为结果变量的学生接近转学成绩，多层次分析结果显示，策略指导和反馈都没有显著的主效应，也没有显著的交互效应（见 表 2）。我们没有发现不同条件下的显著差异，F(3, 433) = 1.51，p = .212， ${\mathrm{\eta }}_p^2$ = .01。

In contrast to near transfer, regarding far transfer, we obtained a significant main effect of strategy instruction (see Table 2), indicating that students who received strategy instruction but no feedback (M = 0.53, SD = 0.33) outperformed students who received neither strategy instruction nor feedback (M = 0.38, SD = 0.32, p = .005, d = 0.46). Furthermore, we found a significant main effect of feedback (see Table 2), indicating that students who received feedback outperformed students who received no feedback. However, the main effect of feedback seemed to be not that robust, as the effect was not significant, when we performed an ANCOVA (see Appendix B).
与近迁移相反，在远迁移方面，我们获得了策略指导的显著主效应（见 表 2），表明接受策略指导但没有反馈的学生（M = 0.53，SD = 0.33）优于既未接受策略指导也未获得反馈的学生（M = 0.38, SD = 0.32, p = .005, d = 0.46）。此外，我们还发现了反馈的显著主效应（见 表 2），表明获得反馈的学生成绩优于未获得反馈的学生。然而，当我们进行方差分析时，反馈的主效应似乎并不那么稳健，因为效应并不显著（见 附录 B）。

Regarding the interaction effect of strategy instruction and feedback, the multilevel analysis obtained a significant instruction-by-feedback interaction effect on students’ far transfer performance (see Table 2).
关于策略指导与反馈的交互效应，多层次分析得出了指导与反馈对学生远距离迁移成绩的显著交互效应（见 表 2）。

To break up the significant interaction, we conducted simple effect analyses regarding far transfer. They revealed that when there was no feedback provided, the presence of strategy instruction significantly improved far transfer test performance (p = .007; receiving strategy instruction: M = 0.53, SD = 0.33; without strategy instruction: M = 0.38, SD = 0.32). By contrast, when participants received feedback, there was no effect of strategy instruction (d = 0.09, p = .861; receiving strategy instruction: M = 0.50, SD = 0.36; without strategy instruction: M = 0.47, SD = 0.34). Together, the findings suggest that strategy instruction had an effect on learning only when no additional feedback was provided. An instruction-by-feedback interaction only emerged for higher-order learning outcomes (i.e., far transfer). The results further suggested that additional corrective feedback had no additive effect on the far transfer performance.
为了打破这种显著的交互作用，我们对远距离迁移进行了简单的效应分析。结果显示，当没有提供反馈时，策略指导的存在能显著提高远迁移测试成绩（p = .007；接受策略指导：M = 0.53，SD = 0.33；没有策略指导：M = 0.38, SD = 0.32）。相比之下，当参与者接受反馈时，策略指导没有影响（d = 0.09，p = .861；接受策略指导：M = 0.50，SD = 0.36；未接受策略指导：M = 0.47, SD = 0.34）。总之，研究结果表明，只有在不提供额外反馈的情况下，策略指导才会对学习产生影响。只有在高阶学习成果（即远距离迁移）方面，才会出现教学与反馈之间的相互作用。结果进一步表明，额外的纠正性反馈对远距离迁移成绩没有附加效应。

To explore the underlying processes which could explain the significant instruction-by-feedback interaction effect on far transfer, based on our pre-registration and the ITFL-model (Narciss, 2008), we explored cognitive (i.e., mental effort, subjective difficulty), meta-cognitive (i.e., monitoring accuracy), and affective-motivational processes (i.e., arousal, pleasure) by applying separate simple mediation analyses with one of the mediator variables each. We represented the instruction-by-feedback interaction by creating a new variable named “interaction” which we included as contrast-coded predictor in the regression model (Rosnow & Rosenthal, 1996; Wiens & Nilsson, 2017). We assigned the following contrast weights of the interaction variable according to our theoretical assumptions: −1 = without instruction without feedback, 1 = without instruction with feedback, 1 = with instruction without feedback, −1 = with instruction with feedback (cf. Lachner et al., 2020). Students’ far transfer performance was the dependent variable. We used the PROCESS macro (Version 4) in SPSS (Hayes, 2018) with bootstrapped 95% confidence intervals (10,000 bootstrap samples). None of the univariate mediation analyses approached significance (Appendix C shows the results of all mediation analyses).
中介分析，分别使用其中一个中介变量，对认知过程（即脑力劳动、主观难度）、元认知过程（即监控准确性）和情感动机过程（即唤醒、愉悦）进行探究。我们通过创建一个名为 "交互作用 "的新变量来表示指令与反馈之间的交互作用，并将其作为回归模型中的对比编码预测因子（Rosnow &；Rosenthal，1996；Wiens & Nilsson，2017）。我们根据理论假设为交互变量分配了以下对比权重：-1=无反馈指导，1=有反馈指导，1=无反馈指导，-1=有反馈指导（参见Lachner 等人，2020）。学生的远距离迁移成绩是因变量。我们使用了SPSS (Hayes, 2018)中的PROCESS宏（第4版），并对95%置信区间进行了引导（10,000个引导样本）。 单变量中介分析均未达到显著性（ 附录 C显示了所有中介分析的结果）。

In a next step, we conducted additional moderated mediation analyses including valence as moderator variable, as indicated by students‘ performance on the practice tasks, following theoretical suggestions of Grundmann et al. (2021) that negative feedback in particular (i.e., feedback, when the answer was incorrect) could largely affect the proposed mediators (see also Eccles & Wigfield, 2002; Kalyuga & Sweller, 2014; Kluger & DeNisi, 1996; Raaijmakers et al., 2017) and reduce learning performance due to students’ disengagement.
当答案不正确时的反馈）会在很大程度上影响所提出的中介因素（另见Eccles & Wigfield, 2002；Kalyuga &；Sweller, 2014; Kluger &；DeNisi, 1996; Raaijmakers et al.,2017），并因学生脱离学习而降低学习成绩。

Based on these theoretical assumptions, we performed moderated mediation analyses in which a moderator (i.e., W = valence) moderates the a-path of the mediation model (independent variable to mediator) but not the c’-path (direct effect) based on Model 7 of Hayes' (2018) PROCESS macro for SPSS. Interestingly, we found a significant effect of arousal, b = 0.01, SE_b = 0.00 (95%, CI [0.0005, 0.0154]; see Fig. 4), as zero was not included in the confidence interval. More precisely, valence moderated the effect of interaction on arousal, suggesting that depending on the experimental condition (instruction only, feedback only, both, or neither), students' valence differed. This affected student arousal and, subsequently, far transfer. None of the other moderated mediation analyses revealed significance (see Appendix D). In summary, the results indicated that the instruction-by-feedback interaction may result in far transfer tasks due to lower levels of arousal. This effect, however, only held for students who received negative feedback during the practice phase.
根据这些理论假设，我们进行了调节性中介分析，在这种分析中，中介（即：W = 评价）调节中介模型的 a-路径（自变量对中介的影响），但不调节 c-路径（直接影响）、W = valence）调节中介模型的 a-路径（自变量到中介变量），但不调节 c'-路径（直接效应），其依据是 Hayes's (2018) SPSS 的 PROCESS 宏的模型 7。有趣的是，我们发现唤醒有显著影响，b = 0.01，SE_b = 0.00（95%，CI [0.0005，0.0154]；见 图 4），因为置信区间中不包括零。更确切地说，情绪调节了交互作用对唤醒的影响，这表明根据实验条件（仅指导、仅反馈、两者都有或两者都没有）的不同，学生的情绪也不同。这影响了学生的唤醒，进而影响了远迁移。其他调节中介分析均未显示显著性（见 附录 D）。总之，研究结果表明，由于唤醒水平较低，教学与反馈的交互作用可能会导致较远的转移任务。然而，只有在练习阶段收到负面反馈的学生才会产生这种效果。

1. Download: Download high-res image (85KB)
   下载：下载高清图片 (85KB)
2. Download: Download full-size image
   下载：下载全尺寸图片。

Fig. 4. Moderated Mediation Analysis Regarding the Impact of the Valence of Feedback and Students‘ Arousal. Note. *p < .05.
图 4。关于反馈的价值和学生唤醒的影响的调节中介分析。注。*p < .05.

2.4. Discussion 2.4.讨论情况

In Experiment 1, in line with Hypothesis 1a, we found a main effect of strategy instruction on students' learning outcome. Overall, we did not obtain a main effect of feedback or an instruction-by-feedback interaction effect on students' transfer performance. More fine-grained exploratively analyses in which we investigated students’ near and far transfer performance revealed a significant instruction-by-feedback interaction for far transfer. The combination of strategy instruction and feedback did not contribute more to learning than the single interventions (cf. Shirah & Sidney, 2023, for related findings).
在实验 1 中，与假设 1a 一致，我们发现策略教学对学生的学习结果有主效应。总体而言，我们没有发现反馈对学生迁移成绩的主效应或教学与反馈的交互效应。我们对学生的近距离和远距离迁移成绩进行了更精细的探索性分析，结果显示，在远距离迁移方面，教学与反馈之间存在显著的交互作用。与单一干预相比，策略指导和反馈的组合对学习的贡献并不大（参见Shirah & Sidney, 2023，相关研究结果）。

Further analyses of the underlying processes of the instruction-by-feedback interaction were rather in line with our affective-motivational assumptions. Students who received strategy instruction and negative feedback perceived lower levels of arousal during the feedback. Lower levels of arousal were associated with lower far transfer performance. This pattern, however, only emerged for students who received negative feedback in the practice phase.
对教学与反馈互动的基本过程的进一步分析与我们的情感-动机假设相当吻合。接受策略指导和消极反馈的学生在反馈过程中的唤醒水平较低。较低的唤醒水平与较低的远迁移成绩有关。然而，只有在练习阶段接受负面反馈的学生才会出现这种模式。

The findings of Experiment 1 therefore suggest that instruction-by-feedback interactions only emerge under distinct conditions, namely when corrective feedback was used, only with regard to high-level learning outcomes (i.e., far transfer), and for students who received negative feedback, as they perceived lower levels of arousal. As in the study by Fyfe and Rittle-Johnson (2016), in Experiment 1, we realized rather parsimonious feedback, as students only received corrective feedback and a short-answer explanation, as to why their solution was (in-)correct.
因此，实验 1 的结果表明，教学与反馈之间的相互作用只有在不同的条件下才会出现，即在使用纠正性反馈时，只针对高层次的学习成果（即远迁移），以及接受负面反馈的学生，因为他们认为自己的唤醒水平较低。与Fyfe 和 Rittle-Johnson（2016）的研究一样，在实验 1 中，我们实现了相当简洁的反馈，因为学生只收到了纠正性反馈和简答解释，说明他们的解决方案（不）正确的原因。

The interaction effect can be explained by students’ arousal as a consequence of negative feedback, as similar to recent research by Kuklick and Lindner (2023), the task performance (i.e., valence) moderated the emotional impact of feedback (i.e., arousal). Grundmann et al. (2021) also showed that negative feedback can lead to motivated disengagement, which hinders performance improvements. Thus, the arousal resulting from the negative feedback could have led to loss of motivation and feedback disengagement, which could be another explanation for the reducing effect.
这种交互效应可以用负面反馈导致的学生唤醒来解释，因为与Kuklick 和 Lindner (2023)最近的研究类似，任务表现（即情绪）调节了反馈的情绪影响（即唤醒）。Grundmann 等人（2021）也表明，负面反馈会导致动机脱离，从而阻碍绩效的提高。因此，负面反馈引起的唤醒可能会导致动机丧失和反馈脱离，这可能是减少效应的另一种解释。

Kuklick and Lindner (2023) found that after correct response feedback, students reported more negative emotions than when, for example, elaborated feedback was provided. Furthermore, their findings showed that elaborated feedback increased the perceived usefulness. It is unclear whether the effects of Experiment 1 would replicate (with more robust effects) in settings in which more elaborated feedback is provided. When students would receive more specific information in the feedback, the perceived utility of the feedback and the associated arousal could increase, which could result in higher levels of test performance (Shute, 2008), and a reversal into an additive combination effect.
Kuklick 和 Lindner（2023）发现，与提供详细反馈等情况相比，学生在获得正确的反馈后会产生更多的负面情绪。此外，他们的研究结果表明，精心设计的反馈增加了感知有用性。目前还不清楚实验 1 的效果是否会在提供更详尽反馈的环境中重现（效果更强）。当学生在反馈中接收到更具体的信息时，反馈的感知效用和相关唤醒可能会增加，这可能会导致更高水平的测试成绩（Shute, 2008），并逆转为加法组合效应。

3.1. Method 3.1.方法

3.1.2. Design and material
3.1.2.设计和材料

We implemented a 2 × 2-factorial design with strategy instruction (yes vs. no) and feedback (yes vs. no) as between-participants factors. Participants were randomly assigned to one of four conditions (without strategy instruction and without feedback, n = 81; without strategy instruction but with feedback, n = 77; with strategy instruction but without feedback, n = 78; with instruction and with feedback, n = 74).
我们采用了 2 × 2 因子设计，将策略指导（有 vs. 无）和反馈（有 vs. 无）作为参与者之间的因素。参与者被随机分配到四个条件之一（无策略指导和无反馈，n = 81；无策略指导但有反馈，n = 77；有策略指导但无反馈，n = 78；有指导和反馈，n = 74）。

We used the same material as in Experiment 1 except for the feedback, which was provided also on the process level (Hattie & Timperley, 2007; Narciss, 2013). Students were provided with remedial explanations to additionally enhance their conceptual understanding: Thus, in the feedback the underlying physics principles (i.e., how voltage, resistances and currents are related in a parallel circuit and how to calculate the different partial currents on the basis of voltage and resistances) were explained. Additionally, we showed the correct calculations to enhance their procedural understanding (see Fig. 5). Thus, the feedback comprised more specific information regarding about which solution steps to perform, and in what order, to solve the task correctly, and thus constitutes one possible form of elaborated feedback (Shute, 2008).
除了反馈之外，我们使用了与实验 1 相同的材料，反馈也是在过程层面上提供的（Hattie &；Timperley, 2007; Narciss, 2013）。此外，还为学生提供了补救性解释，以增强他们对概念的理解：因此，我们在反馈中解释了基本的物理原理（即并联电路中电压、电阻和电流之间的关系，以及如何根据电压和电阻计算不同的部分电流）。此外，我们还展示了正确的计算方法，以加深学生对程序的理解（见 图 5）。因此，反馈包含了更具体的信息，说明要正确解决任务，需要执行哪些解决步骤，以及以何种顺序执行，从而构成了一种可能的详细反馈形式（Shute, 2008）。

1. Download: Download high-res image (394KB)
   下载：下载高清图片 (394KB)
2. Download: Download full-size image
   下载：下载全尺寸图片。

Fig. 5. Example of Negative Feedback of Practice Task 1 in Experiment 2, if the Answer was Incorrect (Translated).
图 5。如果答案不正确，实验 2 中练习任务 1 的负面反馈示例（翻译）。

3.1.2.1. Perceived utility of the practice phase
3.1.2.1.对实践阶段效用的看法

Another modification was that all students indicated their perceived utility of the practice phase (inclusive feedback). To assess the utility of the practice phase, we used five items (e.g., “I have considered the practice phase to be helpful”; McDonald's ω = 0.81) based on Gaspard et al. (2015) on a scale from 1 (not at all useful) to 9 (very useful).
另一项修改是，所有学生都表明了他们对练习阶段（包括反馈）的效用感知。为了评估实践阶段的效用，我们使用了五个项目（例如，"我认为实践阶段很有帮助"；麦当劳ω = 0.81）、"我认为实践阶段很有帮助"；McDonald's ω = 0.81），其依据是 Gaspard et al.(2015) ，评分标准从 1（完全没用）到 9（非常有用）。

3.2. Results 3.2.结果

3.3. Discussion 3.3.讨论情况

In Experiment 2 we partly replicated the findings of Experiment 1: In line with Hypothesis 1a, we again obtained a significant main effect of strategy instruction. In line with Hypothesis 1b, but different from Experiment 1, we found a significant main effect of feedback, likely because we provided more elaborated feedback than in Experiment 1. However, we found no instruction-by-feedback interaction (Hypothesis 2). Together, the findings suggested additive effects, as strategy instruction (rather accounted for far transfer) as well as elaborated feedback (rather accounted for near transfer) improved students’ learning outcome.
在实验 2 中，我们部分复制了实验 1 的结果：与假设 1a 一致，我们再次发现策略指导具有显著的主效应。与假设 1b 一致，但与实验 1 不同的是，我们发现反馈具有显著的主效应，这可能是因为我们提供了比实验 1 更详尽的反馈。然而，我们没有发现教学与反馈之间的交互作用（假设 2）。总之，这些研究结果表明，策略指导（而不是远迁移）和精心设计的反馈（而不是近迁移）都能提高学生的学习成绩，从而产生叠加效应。

Regarding the perceived usefulness, it made no difference whether students received the combination of strategy instruction and feedback or only one of them. For the single support methods, however, we found that providing only strategy instruction without feedback led to an increased perceived usefulness of the learning phase and providing only feedback without prior strategy instruction led to an increased perceived usefulness of the practice phase. One possible explanation for why the interaction effect did not occur could be that the elaborated feedback was perceived as more useful than simple corrective feedback in Experiment 1 (Kuklick & Lindner, 2023), and thus even when students responded incorrectly (negative valence), the additional specific information provided by the elaborated feedback led from motivated feedback disengagement to more motivated feedback engagement (Grundmann et al., 2021).
在感知有用性方面，学生是同时接受策略指导和反馈，还是只接受其中一种，并无区别。然而，对于单一的支持方法，我们发现，只提供策略指导而不提供反馈会提高学习阶段的有用性，而只提供反馈而不事先提供策略指导会提高练习阶段的有用性。互动效应之所以没有出现，一个可能的解释是，在实验 1 中，精心设计的反馈被认为比简单的纠正性反馈更有用（Kuklick &；Lindner，2023），因此，即使学生的回答不正确（负效价），详细反馈所提供的额外具体信息也会使学生从积极的反馈脱离到更积极的反馈参与（Grundmann et al.,2021）。

4.1. Method 4.1.方法

4.2. Results 4.2.结果

4.3. Discussion 4.3.讨论情况

In Experiment 3, we investigated whether the sequence of strategy instruction and feedback would account for reducing versus additive effects on learning. We first hypothesized that both sequences (flipped and regular) led to higher learning outcome than the control condition (Hypothesis 1). The first hypothesis was confirmed as students in the experimental conditions outperformed students in the control condition. This, effect was mainly driven via near transfer performance. Second, we hypothesized that students in the flipped sequence condition outperform students in the regular sequence condition (Hypothesis 2). In contrast to our assumption, our second hypothesis was not confirmed as the flipped sequence did not contribute to higher learning outcome than the regular sequence. This pattern suggests that the obtained additive findings of elaborated feedback and strategy instruction are independent of the sequence of these instructional interventions.
在实验 3 中，我们研究了策略指导和反馈的顺序是否会对学习产生减少效应或增加效应。我们首先假设两种序列（翻转式和常规式）的学习效果都高于对照组（假设 1）。第一个假设得到了证实，实验条件下的学生成绩优于对照条件下的学生。这种效果主要是通过接近迁移的表现产生的。其次，我们假设翻转序列条件下的学生成绩优于常规序列条件下的学生（假设 2）。与我们的假设相反，我们的第二个假设没有得到证实，因为翻转序列并没有使学习成绩高于普通序列。这种模式表明，精心设计的反馈和策略指导所获得的相加结果与这些教学干预的顺序无关。

5.1. Summary of evidence 5.1.证据摘要

Our findings suggest that, whether combining strategy instruction and feedback results in reducing or additive effects depends on the type of feedback (corrective vs. elaborated feedback) and the type of learning goal (near vs. far transfer).
我们的研究结果表明，将策略指导和反馈结合起来，会产生减少效应还是增加效应，取决于反馈的类型（纠正性反馈还是详细反馈）和学习目标的类型（近距离迁移还是远距离迁移）。

A reducing effect (i.e., an instruction-by-feedback interaction) only emerged when corrective feedback was used, and only with regard to higher-order learning outcome of far transfer (Experiment 1; see also Mertens et al., 2022). Our exploratory mediation analyses suggested that this effect was mediated by the level of arousal, but only for students who received negative feedback (moderated mediation; Kuklick & Lindner, 2023; Laudel & Narciss, 2023).
只有在使用纠正性反馈时，才会出现减少效应（即教学与反馈之间的交互作用），而且只有在高阶学习成果远迁移方面才会出现减少效应（实验 1；另见Mertens 等人，2022）。我们的探索性中介分析表明，这种效应是由唤醒水平中介的，但只针对收到负面反馈的学生（调节中介；Kuklick & Lindner, 2023；Laudel & Narciss，2023）。

Contrarily to Experiment 1, when elaborated feedback was provided (Experiment 2), our findings showed that combining strategy instruction with elaborated feedback led to additive effects and supported students’ learning. A potential explanation of how the reducing effect was counteracted could be that a) students found the elaborated feedback to be useful, as indicated by the significant effect of the perceived usefulness of the practice phase in Experiment 2 and b) that the negative emotions (i.e., arousal), when negative feedback was reported, were reduced (Grundmann et al., 2021; Kuklick & Lindner, 2023). Higher levels of usefulness may have turned motivated feedback disengagement into motivated feedback engagement (Grundmann et al., 2021; Kuklick & Lindner, 2023; Laudel & Narciss, 2023; Narciss, 1999). These findings of Experiment 2 highlight the effectiveness of elaborated feedback (Mertens et al., 2022; Shute, 2008; Wisniewski et al., 2020) not only in isolation, but also in orchestration, that is the combination with other instructional interventions, such as strategy instruction.
与实验 1 不同的是，在提供详细反馈的情况下（实验 2），我们的研究结果表明，将策略指导与详细反馈相结合会产生叠加效应，并支持学生的学习。关于如何抵消减少效应的一个可能的解释是：a) 学生认为精心设计的反馈是有用的，正如实验 2 中练习阶段的感知有用性的显著效果所显示的那样；b) 当负面反馈出现时，学生的负面情绪（即：唤醒）会影响他们的学习、唤醒）减少（Grundmann et al、2021; Kuklick & Lindner, 2023）。更高水平的有用性可能会将动机反馈脱离变成动机反馈参与（Grundmann et al、2021; Kuklick & Lindner, 2023；Laudel & Narciss, 2023; Narciss, 1999）。实验 2 的这些发现凸显了精心设计的反馈的有效性（Mertens et al、2022；Shute, 2008；Wisniewski et al.在《2020年》中，"策略教学"（strategy instruction）不仅是孤立的，而且是与其他教学干预措施（如策略教学）相结合的。

We assumed that another possibility to prevent motivated feedback disengagement (Grundmann et al., 2021) and to enhance feedback processing (Narciss, 2008) could be to provide the feedback at the beginning of the learning unit. Students might then be more motivated to process the feedback, perceive it as more useful, and improve their subjective representation of standards, competencies, and task requirements (Narciss, 2013). Similar to the productive failure approach (Kapur, 2008; Sinha & Kapur, 2021), students would first work through the problems themselves and then, in a subsequent learning phase, validate and deepen their knowledge schemes, and close any knowledge gaps. Based on research on sequence effects (e.g., problem-solving prior instruction; Kant et al., 2017; Loibl et al., 2017; van Gog et al., 2011), we reversed the learning and practice phases in Experiment 3. Contrary to our assumptions, there were no differences between the regular and the flipped sequence in terms of motivation, usefulness of the learning and practice phases, or learning outcomes. Experiment 3 suggested that the sequence does not matter as both sequences supported learning similarly well.
我们认为，防止动机反馈脱离（Grundmann et al、2021）和加强反馈处理（Narciss, 2008），可以在学习单元开始时提供反馈。这样，学生可能会更积极地处理反馈，认为反馈更有用，并改进他们对标准、能力和任务要求的主观表述（Narciss, 2013）。与生产性失败方法（Kapur, 2008；Sinha &；Kapur，2021），学生将首先自己解决问题，然后在随后的学习阶段验证和深化他们的知识方案，并弥补任何知识差距。基于对序列效应的研究（例如，解决问题的先行教学；Kant et al、2017；van Gog 等人，2011），我们在实验 3 中颠倒了学习和练习阶段。与我们的假设相反，在学习动机、学习和练习阶段的有用性或学习结果方面，常规序列和翻转序列之间没有差异。 实验 3 表明，顺序并不重要，因为两种顺序都能很好地支持学习。

The findings of our three experiments extend previous studies on combination effects (Fyfe & Rittle-Johnson, 2016; Salden et al., 2010; Wischgoll, 2017), as they provide distinct information about when and why combining strategy instruction and formative feedback may result in additive versus reducing effects, and allow to generalize the findings across several feedback implementations.
我们的三项实验结果扩展了之前关于组合效应的研究（Fyfe & Rittle-Johnson, 2016；Salden et al、2010；Wischgoll, 2017），因为它们提供了关于何时以及为何将策略指导与形成性反馈相结合可能会产生相加效应或相减效应的独特信息，并允许在多种反馈实施中推广研究结果。

5.2. Theoretical implications
5.2.理论影响

What are the broader contributions of our study? The ITFL-model by Narciss (2008, 2013) proposed a reciprocal interplay between internal (cognitive, meta-cognitive, and motivational-affective processes) and external (e.g., feedback source, instructional context) factors that may affect the effectiveness of feedback (see also e.g., Panadero & Lipnevich, 2022; Winne & Butler, 1994, for related assumptions).
我们的研究有哪些更广泛的贡献？Narciss (2008、2013) 提出了内部（认知、元认知和动机-情感过程）和外部（例如，认知、元认知和动机-情感过程）之间的相互影响。g.,反馈源、教学情境）之间的相互作用（另见，例如Panadero & Lipnevich, 2022；Winne & Butler, 1994，相关假设）。

Regarding the internal factors, as a first contribution, we want to highlight the integrative character of our studies. Still to date, a large proportion of instructional research has relied on investigating cognitive mechanisms or motivational processes in isolation, whereas affective-motivational processes have been considered only to a limited extent (see Azevedo, 2015; Braithwaite et al., 2013; D'Mello & Graesser, 2012; Goldin, 2017; Hoogerheide et al., 2019; Narciss et al., 2014, for exceptions). In our study, we integrated different perspectives (i.e., cognitive, meta-cognitive, affective-motivational) in three well-powered online-experimental studies comprising more than 900 students to draw conclusions about potential mechanisms of our intervention. The integration of different cognitive, meta-cognitive and motivational outcomes allowed us to disentangle and model the various processes which may account for potential interactions.
关于内部因素，我们首先要强调的是我们研究的综合性。迄今为止，大部分教学研究仍依赖于孤立地研究认知机制或动机过程，而情感-动机过程只在有限的范围内得到考虑（见Azevedo, 2015；Braithwaite et al.,2013; D'Mello & Graesser, 2012；Goldin, 2017; Hoogerheide et al.,2019; Narciss et al.）在我们的研究中，我们将不同的视角（即认知、元认知、情感-动机）整合到三项动力充足的在线实验研究中，其中包括 900 多名学生，从而就我们干预措施的潜在机制得出结论。将不同的认知、元认知和动机结果整合在一起，使我们能够对可能产生潜在相互作用的各种过程进行分解和建模。

Moreover, we want to note that the tested explanatory mechanisms were derived from theory (Baars et al., 2013; Grundmann et al., 2021; Narciss, 2008; Sweller et al., 2011; see also pre-registration). However, we have to note that our findings rely only on self-reported measures (i.e., cognitive load, monitoring accuracy, and arousal), as we realized three online experiments. Yet it is unclear, whether and how these measures reflect the proposed theoretical constructs (see Pekrun & Bühner, 2014; Tan et al., 2021, for a critical discussion of self-reports). Replications are needed that additionally integrate less obtrusive measures, such as electrodermal activity or eye-tracking as potential behavioral proxies of (meta-)cognitive and affective-motivational measurements (e.g., Hoogerheide et al., 2019; Potter & Bolls, 2011; Schneider et al., 2019).
此外，我们想指出的是，测试的解释机制来自理论（Baars et al、2021; Narciss, 2008; Sweller et al、2011；另见预登记）。然而，我们必须注意到，我们的研究结果仅依赖于自我报告的测量（即认知负荷、监测准确性和唤醒），因为我们实现了三个在线实验。然而，目前还不清楚这些测量是否以及如何反映所提出的理论构架（见Pekrun & Bühner, 2014；Tan et al、2021 ，对自我报告进行了批判性讨论）。此外，还需要进行重复研究，将干扰性较小的测量方法，如皮电活动或眼动跟踪，作为（元）认知和情感动机测量的潜在行为代用指标（例如：皮电活动或眼动跟踪）纳入研究、Hoogerheide et al、2019; Potter & Bolls, 2011; Schneider et al、2019）。

As a further contribution, regarding the external factors, our study showed that the orchestration of feedback with other instructional strategies plays an important role. We refer to the fact that we deliberately investigated combination effects of two different interventions, that is strategy instruction and computer-based feedback. In educational settings and online learning environments, strategy instruction and feedback are often combined (Magliaro et al., 2005; Rosenshine, 2008), however, most studies often relied on investigating only a single instructional intervention and research investigating to what extent both interventions affect each other's effectiveness to enhance students' learning is rare and produced heterogeneous findings (Fyfe & Rittle-Johnson, 2016; Salden et al., 2010; Wischgoll, 2017). Thus, our study extends prior research as we are contributing to the scarce evidence of combining instructional interventions (Fyfe & Rittle-Johnson, 2016; Salden et al., 2010; Wischgoll, 2017; see also e.g., Murphy & Alexander, 2005).
在外部因素方面的另一个贡献是，我们的研究表明，反馈与其他教学策略的协调发挥着重要作用。我们特意研究了两种不同干预措施的组合效果，即策略指导和基于计算机的反馈。在教育环境和在线学习环境中，策略指导和反馈通常是结合在一起的（Magliaro et al、2005；Rosenshine, 2008）、大多数研究往往只依赖于调查单一的教学干预措施，而调查两种干预措施在多大程度上相互影响以提高学生学习效率的研究并不多见，而且结果也不尽相同（Fyfe &；Rittle-Johnson, 2016; Salden et al.,2010; Wischgoll, 2017）。因此，我们的研究扩展了先前的研究，因为我们为教学干预相结合的稀缺证据做出了贡献（Fyfe &；Rittle-Johnson, 2016; Salden et al.,2010；Wischgoll, 2017；另见如，Murphy & Alexander, 2005）。

Furthermore, our findings support prior research by highlighting the role of feedback implementation (e.g., Hattie & Timperley, 2007; Kluger & DeNisi, 1996; Narciss, 2008; Patchan et al., 2016; Shute, 2008; Strobl et al., 2019). The design of feedback, such as the level of elaboration, may not only impact the effectiveness of feedback per se, but also might influence the effectiveness of combining instructional interventions.
此外，我们的研究结果还支持先前的研究，强调了反馈实施的作用（如Hattie & Timperley, 2007; Kluger &；DeNisi, 1996; Narciss, 2008; Patchan et al.,2016; Shute, 2008; Strobl et al、2019）。反馈的设计，如详细说明的程度，不仅可能影响反馈本身的有效性，还可能影响综合教学干预的有效性。

Summarized, our main contribution is that our findings support the theoretical assumption (Narciss, 2008, 2017; Narciss et al., 2014; cf. Butler & Winne, 1995; Hattie & Timperley, 2007; Hayes, 2012; Shirah & Sidney, 2023) that the effectiveness of feedback depends on internal as well as on external factors: The results of our experiments indicate that whether combining strategy instruction and feedback may result in non-additive/even reducing or additive effects, strongly depends on individual student-related (internal) factors (i.e., task performance/valence, arousal) but also on feedback-related (external) factors (design decisions, i.e., elaboration level, or instructional context, i.e., orchestration with strategy instruction), probably explained by utility.
综上所述，我们的主要贡献在于我们的研究结果支持理论假设（Narciss, 2008，2017；Narciss et al.,2014; cf.Butler & Winne, 1995; Hattie & Timperley, 2007；Hayes, 2012; Shirah &；Sidney，2023），反馈的有效性取决于内部和外部因素：我们的实验结果表明，将策略指导和反馈结合起来是否会产生非加成/偶减成或加成效应，在很大程度上取决于与学生个体相关的（内部）因素（即任务表现/价值、唤醒），同时也取决于与反馈相关的（外部）因素（设计决策，即阐述水平，或教学情境，即与策略指导的协调），这可能可以用效用来解释。

5.3. Practical implications
5.3.实际影响

Our findings also have implications for educational practice, as they show that combining strategy instruction and computer-based feedback is only effective under certain conditions and should therefore be used wisely. Our results showed that all types of support (strategy instruction, feedback, combination) improve learners' performance compared to learners without any instructional interventions. Contrary to the prevalent assumption, however, double support (combination of strategy instruction and feedback) does not necessarily increase learners' performance, as demonstrated by Experiment 1. The findings counteract the commonly held belief that combining instructional interventions, which are regarded to be effective strategies in isolation, contributes to learning. Instead, practitioners should implement elaborated feedback. Our experiments have shown that elaborated feedback can have positive or additive effects on learning both in isolation (Mertens et al., 2022; Shute, 2008) and orchestrated in combination with strategy instruction (cf. Salden et al., 2010).
我们的研究结果还对教育实践产生了影响，因为它们表明，将策略指导和计算机反馈结合起来只有在特定条件下才会有效，因此应明智使用。我们的研究结果表明，与没有任何教学干预措施的学习者相比，所有类型的支持（策略指导、反馈、组合）都能提高学习者的成绩。然而，与普遍的假设相反，双重支持（策略指导与反馈相结合）并不一定会提高学习者的成绩，实验 1 就证明了这一点。这些研究结果反驳了人们普遍持有的观点，即把被认为是有效策略的教学干预结合起来有助于学习。相反，实践者应该实施精心设计的反馈。我们的实验表明，精心设计的反馈对单独学习（Mertens et al、2022；Shute, 2008），并与策略指导相结合进行协调（参见：Shute, 2008）。Salden et al、2010）。

5.4. Limitations and future research
5.4.局限性和未来研究

We would like to note that all detected effects in all three experiments were rather small. Regarding corrective feedback, this is in line with previous research which showed that corrective feedback usually produces small effects (e.g., Van der Kleij et al., 2015; Swart et al., 2019; see Mertens et al., 2022 for an overview). However, regarding elaborated feedback, previous research mostly documented medium effect sizes (see Mertens et al., 2022; Wisniewski et al., 2020, for meta-analytic evidence). It is unclear why our implemented elaborated feedback only produced small but not medium effects. One explanation refers to the fact, that we realized our study as an online experiment. Online experiments have the risk of less control and inattentive behavior by our participants, as compared to laboratory experiments (Hauser & Schwarz, 2016; Oppenheimer et al., 2009). The lower levels of control and attention may result in lower effect sizes, as compared to traditional laboratory studies (see Ryan et al., 2013).
我们要指出的是，所有三个实验中检测到的效果都相当小。关于纠正性反馈，这与之前的研究一致，这些研究表明纠正性反馈通常会产生较小的效果（例如，Van der Kleij et al、2015; Swart et al、2019；概况见Mertens 等人，2022）。然而，关于精心设计的反馈，以往的研究大多记录了中等效果（见 Mertens et al、2022; Wisniewski et al.）目前还不清楚为什么我们实施的详细反馈只产生了小效果，而没有产生中等效果。一种解释是，我们的研究是以在线实验的形式进行的。与实验室实验相比，在线实验可能会减少参与者的控制和注意力不集中行为（Hauser &；Schwarz, 2016; Oppenheimer et al.,2009）。与传统的实验室研究相比，较低的控制和注意力水平可能会导致较低的效应大小（见Ryan 等人，2013）。

Another limitation refers to the number of practice tasks. We implemented only two practice tasks which were realized as a multiple-choice question format. Such practice phases can be regarded to be prototypical for online learning phases. Nevertheless, implementing more challenging learning materials with more practice tasks and more transfer tasks (cf. van Harsel et al., 2020) preferably in combination with open-ended questions could help to further generalize the obtained findings.
另一个限制因素是练习任务的数量。我们只实施了两个以选择题形式实现的练习任务。这种练习阶段可视为在线学习阶段的原型。尽管如此，采用更具挑战性的学习材料，包括更多的练习任务和更多的迁移任务（参见van Harsel 等人，2020），最好与开放式问题相结合，这将有助于进一步推广所取得的研究成果。

Given that our study provides a complex picture of combining instructional interventions and the use of online experiments, we see the need to further investigate these combination effects in practice to replicate the experiments in more authentic settings. Since our findings are limited to one specific topic in physics and the instruction phase and the feedback phase were relatively short (approximately 10–15 min each), it is essential to replicate the findings in different study contexts, with different types of interventions, and with longer durations. Thus, further field-oriented work such as field experiments are required to replicate the findings and to increase the ecological validity of our findings (Renkl, 2013).
鉴于我们的研究提供了教学干预与在线实验相结合的复杂情况，我们认为有必要在实践中进一步研究这些组合效果，以便在更真实的环境中复制实验。由于我们的研究结果仅限于物理中的一个特定主题，而且指导阶段和反馈阶段都相对较短（各约 10-15 分钟），因此必须在不同的研究背景下、使用不同类型的干预措施和更长的持续时间来重复研究结果。因此，需要进一步开展以实地为导向的工作，如实地实验，以复制研究结果并提高我们研究结果的生态有效性（Renkl, 2013）。

Our findings cannot be generalized regarding learning in general. However, we achieved our goal to provide deeper insights of the impact of combining strategy instruction with computer-based feedback on learning in an online setting.
我们的研究结果不能推广到一般的学习中。不过，我们实现了自己的目标，即更深入地了解将策略指导与基于计算机的反馈相结合对在线学习的影响。

5.5. Conclusions 5.5.结论

All in all, the experiments presented in this paper provide important evidence on combining strategy instruction and feedback: The findings illustrate that combining strategy instruction and feedback does not necessarily contribute to learning, as the effectiveness depends on the provided type of feedback and the learning goals. Therefore, they provide important information for fostering cognitive skill acquisition, as they show that not only the design of instructional interventions but also their orchestration is critical to contribute to the effectiveness of digital learning environments.
总之，本文介绍的实验为策略指导与反馈相结合提供了重要证据：研究结果表明，策略指导与反馈相结合并不一定有助于学习，因为其有效性取决于所提供的反馈类型和学习目标。因此，它们为促进认知技能的习得提供了重要信息，因为它们表明，要提高数字学习环境的有效性，不仅要设计教学干预措施，还要对其进行协调。