What Causes Poor Monitoring From Expository Text?

• In Event: Advances in Understanding Metacomprehension Accuracy of Expository Text

Abstract

Objectives: This presentation reports key results computed from a meta-analysis culled from 52 publications and comprised of 89 studies that reported relative accuracy from expository text in baseline conditions (i.e. conditions that did not include any additional features shown to affect metacomprehension accuracy, such as drawing, rereading, self-explanation, comprehension test expectancy, or delayed generation activities).

Theoretical Framework: When students are responsible for learning or studying on their own, they need to make decisions about when they need to keep studying and what they need to study. Accurate monitoring is critical for effective self-regulated learning; however, research has shown that students lack effective monitoring skills. For example, students are poor at monitoring their comprehension of expository texts. Metacomprehension accuracy is traditionally measured by asking students to predict their own performance on a test of comprehension after reading. Absolute accuracy measures are computed by considering the discrepancy between the magnitudes of predictions and actual performance. Relative metacomprehension accuracy is computed by the ability to discriminate between well-understood and poorly-understood texts within a set (i.e. a correlation between a single individual’s predictive judgments and actual test scores for a set of texts).

Methods and Data Sources: Experiments that measured relative metacomprehension accuracy were identified using keyword searches and forward searches of review articles (Maki, 1998; Thiede, Griffin, Wiley, & Redford, 2009). While the typical finding is that relative metacomprehension accuracy is quite poor, methodologies vary widely and several contrasts were possible. While most studies use undergraduates, some have used younger readers. Readers are either asked to judge their understanding of a single text segmented into smaller units, or a set of texts on distinct topics. In terms of test questions, tests ranged from 1 to 16 items per text unit. Multiple-choice formats were most common, but some studies used cued-recall or verification tasks (true/false), whereas other studies used open-ended formats. Test questions also varied in the extent to which they require readers to remember information versus to integrate information or construct inferences.

Results: Key findings shown in Table 1 include that few studies have explored relative accuracy in younger samples, but these studies find that typical levels of monitoring accuracy are lower than in college samples. Also, younger students show poorer accuracy on inference-based test items. In college-level samples, accuracy does not increase linearly with number of test items and a single test item yields uniquely low accuracy. Similarly, monitoring is worse for a single topic than for multiple topics. No differences emerged due to test format, but accuracy was again lower for inference-based questions.

Significance: These results are consistent with the perspective that most readers default to memory-based rather than comprehension-based monitoring cues, and that many students may lack an understanding of what comprehension means and how to monitor it (Wiley, Thiede, & Griffin, 2016). These results offer implications for improving the measurement of monitoring. Finally, the relation of these findings to interventions and activities that have been most effective toward improving metacomprehension accuracy will be discussed.

Authors

• Jennifer Wiley, University of Illinois at Chicago

• Marta Krystyna Mielicki, University of Illinois at Chicago

• Thomas D. Griffin, University of Illinois at Chicago