学习投入是学习的基本前提和重要保障，也体现了学生的学习能力与学习策略。在课堂这一重要的学习场景中，对学习投入的研究尤为重要。课堂上的学习投入不仅与学生学习的心理过程密切相关，更对学习产出预测、教育教学实践指导有着重要价值。然而，课堂学习投入具有结构复杂、动态、内隐等特点，使得对其测量与评估的挑战度较大。传统的测量手段如自我报告，尽管为我们理解学习投入提供了重要线索，但存在伴随性、动态性不足等问题，在揭示学生认知和情感等内隐投入的动态特点上存在较大局限。近年来认知神经科学的蓬勃发展为我们提供了全新的研究视角。穿戴式设备等生理传感技术的发展，使得在真实的课堂环境中收集学生的多模态神经生理数据成为可能，为学习内隐投入的客观测量开辟了新途径。同时，心理学领域广泛应用的经验取样法也为动态追踪学习内隐投入提供了新的解决方案，为我们深入理解学习内隐投入的复杂机制提供了更为丰富的数据支持。本研究运用经验取样法，在真实高中课堂中采集问卷、脑电、皮肤电等多模态神经生理数据，分析了学业成绩与课堂学习内隐投入的关系，通过统计分析和机器学习等方法，研究了神经生理指标对课堂学习内隐投入的预测作用。此外，本研究通过比较经验取样法和传统事后回忆测量结果，并结合神经生理数据，考察了课堂学习内隐投入的动态特征。本研究的主要结论包括：（1）课堂学习内隐投入和学业成绩呈现“倒U型”关系；（2）多种模态的神经生理数据可以更好地预测课堂学习内隐投入；（3）真实课堂情境下的认知投入与情感投入存在高度一致性；（4）单节课内的学习内隐投入有较高的稳定性。本研究具有较好的理论价值与实践意义：（1）在理论层面，一方面，发现了学习内隐投入和学业成绩的“倒U型”关系，为丰富学习投入的理论建构提供新思路；另一方面，通过经验取样法在真实课堂学习过程中采集较高密度的数据，揭示了学习投入多维结构中认知投入与情感投入之间的复杂互动关系，并进一步探索了课堂学习内隐投入的动态特征，深化了学习投入的理论内涵；（2）在方法层面，通过对比单模态数据分析和多模态数据整合分析的效果，验证了多模态数据在预测学习内隐投入方面的显著优势，为多模态学习分析的方法研究贡献了新的思路。

Study engagement is a basic prerequisite and important guarantee for learning, and it also reflects students' learning ability and learning strategies. In the classroom, which is an important learning environment, the study of student engagement is particularly important. Student engagement in the classroom is not only closely related to the psychological process of students' learning but also has an important value for the guidance of education and teaching practice. However, classroom student engagement is characterized by a complex structure, dynamics, and implicitness, making it more challenging to measure and assess. Traditional measures such as self-report, although providing important clues to our understanding of student engagement, suffer from the problems of insufficient concomitance and dynamics, and have greater limitations in revealing the dynamic characteristics of students' implicit engagement such as cognitive and emotional engagement. The booming development of cognitive neuroscience in recent years has provided us with new research perspectives. The development of physiological sensing technologies, such as wearable devices, has made it possible to collect multimodal neurophysiological data from students in real classroom environments, which opens up new avenues for objective measurement of implicit engagement. Meanwhile, the experience sampling method, which is widely used in the field of psychology, also provides a new solution for dynamic tracking of implicit engagement and provides richer data support for our in-depth understanding of the complex mechanisms of implicit engagement.In this study, we used the experience sampling method to collect multimodal neurophysiological data, such as questionnaires, electroencephalography, and electrodermal, in a real high school classroom, and analyzed the relationship between academic achievement and classroom implicit engagement, and investigated the predictive role of neurophysiological indicators on classroom implicit engagement through statistical analysis and machine learning. In addition, this study examined the dynamic characteristics of classroom implicit engagement by comparing the results of the experience sampling method and the longitudinal measurements with the neurophysiological data.The main conclusions of this study include: (1) classroom implicit engagement and academic achievement show an "inverted U-shaped" relationship; (2) multi-modal neurophysiological data can better predict classroom implicit engagement; (3) there is a high degree of consistency between cognitive and affective inputs in real classroom situations; (4) implicit engagement within a single classroom session has a high degree of stability.This study has good theoretical value and practical significance: (1) At the theoretical level, on the one hand, the "inverted U-shape" relationship between implicit engagement and academic achievement was found, which provides a new way to enrich the theoretical construction of implicit engagement. On the other hand, by collecting high-density data in the real classroom learning process through empirical sampling method, we reveal the complex interaction between cognitive and emotional engagement in the multidimensional structure of implicit engagement and further explore the dynamic characteristics of classroom implicit engagement, deepening the theoretical connotation of study engagement; (2) at the methodological level, by comparing the effects of unimodal data analysis and multimodal data integration analysis, we validate the significant advantages of multimodal data in predicting implicit engagement and contribute new ideas for the research on the methodology of multimodal learning analysis.