Text comprehension: The impact of strategies and knowledge

Text comprehension: The impact of strategies and knowledge

First Author: Kate Cain -- Lancaster University
Keywords: Comprehension, Knowledge, Inference
Abstract / Summary: 

In recent years, there has been growing debate about the relations between reading comprehension (and its component skills, such as inference making) and the extent to which knowledge (vocabulary, background knowledge) and strategies impact on performance. Some have proposed that reading comprehension and/or reading strategies cannot be taught, and that instructors should instead focus on instruction in knowledge. A wealth of research shows that much knowledge (vocabulary and background knowledge) is acquired through independent reading, and that strategies such as inference from context are implicated in knowledge acquisition. It is likely that the relations between knowledge, strategies and reading comprehension change across the course of development, differ for different reading purposes or genre, and for different populations of reader. A better understanding of how, why, and when knowledge and strategies are related to reading comprehension is essential for empirically-motivated models of reading comprehension, curricula and interventions. This symposium brings together papers that address these issues and aims to provoke wider discussion and promote ideas for future work in the area.

Symposium Papers: 

Teaching inferencing in K-2: The role of scaffolding and feedback

First Author/Chair:Kristen McMaster -- University of Minnesota
Additional authors/chairs: 
Panayiota Kendeou

Purpose. We present results of a field test designed to explore the feasibility of ELCII and TELCI: fully-automated, web-based, interactive interventions designed to train inference making in K-2.

Method. Participants were 153 students in Kindergarten and 60 in Grades 1 and 2. The students’ teachers participated in professional development prior to the implementation of ELCII in Kindergarten and TELCI in Grades 1 and 2. Researchers then implemented ELCII and TELCI modules with students in small groups for 10 weeks and conducted structured in-vivo observations. Before and after implementation, students completed an inference assessment and a language comprehension test (CELF-5).

Results. We examined students’ growth in inference-making skills during the ELCII and TELCI interventions. When students answer a question on the intervention module incorrectly, they receive scaffolding, feedback, and another opportunity to answer the question. We computed students’ correct answer scores before they received scaffolding (Attempt 1) and their total correct scores after scaffolding (Attempt 1 + 2) across four time periods. For both interventions and groups, results showed that participants’ inference-making skills improved over the four time periods (effect size ηp2 = .681 for ELCII; effect size ηp2 = .192 for TELCI) and participants answered more inferential questions correctly after they received scaffolding and feedback (effect size ηp2 = .886 for ELCII; effect size ηp2 = .995 for TELCI). Across time, students needed less scaffolding.

Conclusions. Overall, ELCII and TELCI can be feasibly included as part of language comprehension instruction in K-2 and significantly improve students’ inference making.

The influence of reader and text characteristics on 6th graders’ inference making

First Author/Chair:Kate Cain -- Lancaster University
Additional authors/chairs: 
Nicola Currie; Gill Francey; Rob Davies; Shelley Gray; Marilyn Thompson; Mindy Bridges

Purpose. Inference and integration processes are considered essential for successful reading comprehension in all major theoretical models of text comprehension (McNamara & Magliano, 2009). We present an examination of the role of different reader and text characteristics on inference making in sixth graders.

Method. 87 monolingual English speaking children and 95 children who entered pre-kindergarten as Spanish speakers were recruited to participate in this study when they were in grade 6. Each child read narrative and expository texts, sentence by sentence. Each text required an inference in order to integrate two sentences and memory load was manipulated by placing the critical sentences adjacent (near condition) or 2-3 sentences apart (far condition). At the end of each text, participants answered a yes/no question that assessed the inference. Question answering accuracy, question answering response time, and sentence reading times were recorded. Vocabulary, background knowledge, knowledge of reading strategies, and word reading ability were assessed.

Results. Linear mixed effects models were conducted separately for question answering accuracy, question answering response time, and sentence reading time, to examine the influence of reader characteristics (language status, vocabulary, background knowledge, reading strategy knowledge, word reading) and text characteristics (genre, memory load) on performance. Our preliminary analyses show that the reader characteristics most strongly related to performance were vocabulary and background knowledge and that the text characteristics of memory load manipulation and genre were both related to performance.

Conclusions. These findings suggest that conceptual knowledge is a more critical determinant of inference making than knowledge of reading strategies, for both narrative and expository texts.

Knowledge predicts inference-making and inference-making interventions improve knowledge acquisition

First Author/Chair:Marcia Barnes -- Vanderbilt University
Additional authors/chairs: 
Amy Barth; Nathan H. Clemens; Deborah Simmons; Colby Hall; Greg Roberts; Anna-Mari Fall; Sharon Vaughn

Purpose. In adolescents, inference-making and knowledge are the strongest direct predictors of reading comprehension, and knowledge and reading comprehension are also indirectly related through inference-making. We report 2 studies that address the relations between knowledge, inference-making, and reading comprehension.

Method & results. Study 1: a dominance analysis tested potential predictors of knowledge-based gap-filling inferences and text-connecting inferences in which information had to be integrated over short and longer text distances. The sample included 1196 6th to 12th grade students oversampled for students who were less skilled comprehenders. Predictors included verbal knowledge, memory, word reading fluency, inattention, inhibition, and nonverbal reasoning. Regardless of inference type, the latent measure of knowledge was the dominant predictor followed by latent memory, and word reading fluency in some models; however, memory was as important a predictor as knowledge for inferences requiring the information integration across larger text distances. Study 2 tested the effects of an inference-making intervention on several types of inferences, including those involved in acquiring new vocabulary knowledge from context. In this small randomized controlled trial one of five modules focused on teaching strategies for inferring the meaning of words from context and another module provided additional practice in this type of inference-making. Participants were 105 struggling middle school readers assigned to a BAU condition or to one of two versions of the same inference-making intervention. The teacher-led version of the intervention led to gains in deriving the meanings of non-instructed words in new passages.

Conclusions. Findings are discussed with reference to why knowledge is important for both knowledge-based and text-connecting inferences and with respect to how inference-making is also important for acquiring new knowledge.

Influences of Language and Non-academic Knowledge on Basic and Deep Comprehension

First Author/Chair:John Sabatini -- Institute for Intelligent Systems, University of Memphis
Additional authors/chairs: 
Zuowei Wang; Tenaha O’Reilly

Purpose. The knowledge that a reader brings to comprehension come in many varieties, but deep comprehension also requires reasoning about content Deep processing is especially important when acquiring new knowledge, academic or non-academic. Here, we explore the latter, extending previous findings with more nuanced analyses of language and prior knowledge on basic and deep comprehension.

Methods. HS students were randomly assigned to traditional, basic comprehension (RC-A/B) or a scenario-based assessment (SBA), on the non-academic topic chosen to minimize the confound of academic achievement with academic knowledge. Assessments of language knowledge (morphology, sentence discourse, and academic vocabulary) were completed, as well as background knowledge tests (topic vocabulary, football facts, topic conceptual problems).

Results. Hierarchical regression was performed to evaluate the unique contribution of each of the three types of language knowledge followed by the three types of knowledge for each of the three tests (RC-A/B & SBA). When reading comprehension is operationalized traditionally as answering MC questions about passages, basic language knowledge was the strongest predictor, with minimal contribution of prior knowledge. However, with deep reading tasks operationalized as SBA, all subtypes of language and prior knowledge were significant.

Conclusions. Results support importance of various types of prior knowledge to deep comprehension.

Discussion of papers

First Author/Chair:DISCUSSANT Art Graesser -- University of Memphis

This presentation will examine the findings of these papers, highlighting similarities and differences, within a theoretical context.