抑制控制及訓練對國小學生後設理解準確性之影響

Abstract

本研究以執行功能與後設認知整合架構為基礎，探究抑制控制與後設理解準性之關聯，並探討課程融入式抑制控制訓練與後設理解判斷訓練對後設理解判斷準確性的影響，目標是發展有助學生在真實課堂情境中提升後設理解判斷準確性的訓練活動。過往研究指出，兒童在後設理解監控準確性上表現不佳，易受主觀感受誤導，產生過度自信的偏差。近年相關研究亦指出，處理流暢性與認知控制機制（特別是抑制控制能力），在理解監控歷程中扮演重要角色。然而，目前國內外尚缺乏針對國小學童結合抑制控制與後設理解監控訓練的介入研究。因此，本研究分兩階段進行，綜合檢驗上述心理歷程及教學介入成效。研究一以臺灣國小五年級學生為對象（N = 141），探討文本流暢性（連貫版與不連貫版本）與後設理解判斷準確性之關聯，並檢驗抑制控制能力的調節效果。後設理解準確性以Schraw（2009a）提出之五項指標（絕對準確性、相對準確性、偏誤、分散、區辨）進行量化分析；抑制控制能力則以電腦化西蒙任務測得抑制指標與轉換指標。研究二為介入效果之準實驗研究，同樣以國小五年級學生為對象（N = 155），分為抑制控制訓練組、抑制控制加判斷訓練組與對照組，介入課程歷時八週。研究一結果顯示，文本流暢性顯著影響後設理解判斷準確性。不論抑制能力高低，學生在不流暢文本下的「相對準確性指數」顯著優於流暢文本，且其「偏誤指數」亦呈現較大負向偏誤，反映學童在主觀流暢感降低的情形下，更謹慎評估自身理解狀態。在調節效果方面，僅「轉換指標」在文本流暢性與「分散指數」之間展現顯著調節作用。此外，本研究進一步以階層迴歸分析探討抑制控制能力與後設理解準確性對閱讀理解表現之預測力。分別於流暢與不流暢文本條件下，將「抑制指標」、「轉換指標」及「相對準確性」依序納入迴歸模型。結果顯示，在流暢文本條件中，三個變項對閱讀理解表現的共同解釋變異量為8.9%，其中「抑制指標」與「相對準確性」對模型具有顯著的增量預測效果；在不流暢文本條件中，三個變項的共同解釋變異量提升至16.1%，且三者皆具顯著預測力，顯示不流暢文本所需之認知與後設認知資源更高，惟「轉換指標」之迴歸係數呈負向，顯示其在閱讀困難情境中可能對閱讀表現產生反向影響。研究二以廣義估計方程式（GEE）檢驗訓練成效，並輔以分層分析，以探討不同訓練組別在控制前測表現後，於後測各項後設理解準確性指標與閱讀理解能力的表現差異。GEE結果顯示，「抑制控制加判斷訓練組」之「區辨指數」在控制前測後仍顯著高於對照組，而在「絕對準確性指數」與「閱讀理解能力」上則顯著優於「抑制控制訓練組」。分層分析進一步發現，「抑制控制訓練組」中前測「分散指數」較低者在後測顯著高於對照組；而「抑制控制加判斷訓練組」在前測「絕對準確性指數」高者或前測「閱讀理解能力」表現低者，其後測表現亦顯著優於對照組，前測「偏誤指數」較低者也獲得改善。另在「抑制控制加判斷訓練組」與「抑制控制訓練組」的分層分析比較中發現，「抑制控制加判斷訓練組」前測「絕對準確性指數」高者在後測表現有顯著優勢；此外，無論前測「偏誤指數」高低，該組學生在後測均較「抑制控制訓練組」呈現較大的差異，反映訓練使學生在偏誤指數上產生更精準的後設理解判斷。綜合兩項研究結果，本研究支持認知控制與後設認知歷程交互作用的理論觀點，指出處理流暢性感受雖為影響後設理解判斷的重要線索，但其作用亦受到個體抑制控制與轉換能力的影響。本研究驗證了抑制控制與後設理解準確性在閱讀理解歷程中的共同預測力。此外，融入學科課程的抑制控制訓練與後設理解判斷訓練，確實可提升特定族群之理解監控與閱讀理解表現，且部分結果亦顯示，訓練對於基線能力較低之學生具有相對更顯著之促進效果。本研究提出一可行之訓練架構，回應國內後設理解與執行功能教學研究之空缺，並針對理論與實務意涵及未來研究方向提出建議。Grounded in an integrated framework of executive function and metacognition, this study investigated the relationship between inhibitory control and metacomprehension accuracy, and further examines the effects of curriculum-embedded inhibition control training and metacomprehension judgment training on the accuracy of students’ metacomprehension judgments. The objective was to develop training activities that can effectively enhance students’ metacomprehension judgment accuracy in authentic classroom settings. Previous research has indicated that children often exhibit deficiencies in metacomprehension monitoring, frequently being misled by subjective impressions, which results in overconfidence of task difficulty. Recent studies have further emphasized the important roles played by processing fluency and cognitive control mechanisms, particularly inhibition control, in the metacomprehension process. However, systematic intervention studies combining inhibition control and metacomprehension monitoring training among elementary school students remained scarce both domestically and internationally. This study thus employed a two-phase design to comprehensively examine the underlying psychological processes and instructional intervention outcomes.Study 1 recruited Taiwanese fifth-grade school students (N = 141) to examine the relationship between text fluency (coherent vs. incoherent texts) and metacomprehension judgment accuracy, as well as the moderating effect of inhibition control. Metacomprehension accuracy was assessed using five indicators proposed by Schraw (2009a): absolute accuracy, relative accuracy, bias, scatter, and discrimination. Inhibition control was measured using a computerized Simon Task, which yielded two indices: inhibition and switching. Study 2 adopted a quasi-experimental intervention design, also targeting fifth-grade students (N = 155), who were assigned to one of three groups: inhibition training only, combined inhibition and judgment training, or a control group. The intervention lasted eight weeks.Findings from Study 1 revealed that text fluency significantly influenced metacomprehension judgment accuracy. Regardless of inhibition ability, students demonstrated significantly higher relative accuracy and more negative bias (i.e., underconfidence) when reading disfluent texts, suggesting that reduced subjective fluency prompts more cautious self-evaluations. Regarding moderation, only switching index significantly moderated the relationship between text fluency and scatter index. Hierarchical regression analyses further examined the predictive ability of inhibition control and metacomprehension accuracy on reading comprehension. In the fluent condition, inhibition and relative accuracy significantly contributed to reading comprehension, explaining 8.9% of the variance. In the disfluent condition, explained variance increased to 16.1%, with all three predictors, including the switching index, reaching statistical significance. Notably, the regression coefficient for the switching index was negative, suggesting a possible detrimental role under cognitively demanding reading conditions.Study 2 employed generalized estimating equations (GEE) and stratified analyses to evaluate the intervention effects across groups, controlling for pretest performance. GEE results showed that the combined training group significantly outperformed the control group in posttest discrimination index and surpassed the inhibition-only training group in absolute accuracy and reading comprehension. Stratified analyses further revealed that students with lower pretest scores on the scatter index in the inhibition-only training group, as well as those in the combined training group who had higher pretest scores on the absolute accuracy index or lower scores on reading comprehension, demonstrated significantly better posttest performance compared to the control group. Improvements were also observed among students in the combined training group who had lower pretest scores on the bias index. Stratified analyses comparing the combined training group and the inhibition-only training group revealed that students with higher pretest scores on the absolute accuracy index in the combined group outperformed those in the inhibition-only group on the posttest. Moreover, regardless of whether students had high or low pretest scores on the bias index, those in the combined training group showed greater improvements on the posttest than those in the inhibition-only group, indicating that the training helped students make more accurate metacomprehension judgments with respect to bias.Collectively, the results of both studies support the theoretical perspective that cognitive control and metacognitive processes interact dynamically. Although processing fluency serves as a salient cue for metacomprehension judgments, its influence is moderated by individual differences in inhibition and switching abilities. This study provides empirical evidence that both inhibition control and metacomprehension accuracy jointly predict reading comprehension outcomes. Moreover, curriculum-embedded training targeting these capacities can enhance monitoring and comprehension performance, particularly for students with lower baseline abilities. The study offers a viable cognitive training framework, addressing current gaps in instructional research on metacomprehension and executive function development, and concludes with implications for theory, educational practice, and future research directions.