Abstract
Reading comprehension is one of the most complex cognitive activities in which humans engage, making it difficult to teach, measure, and research. Despite decades of research in reading comprehension, international and national reading scores indicate stagnant growth for U.S. adolescents. In this article, we review the theoretical and empirical research in reading comprehension. We first explore different theoretical models for comprehension and then focus on components shown to be important across models that represent potential targets for instruction. In the last part of the article, we consider solutions for translating research to practice and policies for improving instruction. Improving reading scores will require a concerted and collaborative effort by researchers, educators, and policy makers with a focus on long-term solutions. An early and sustained focus on developing background knowledge, vocabulary, inference, and comprehension monitoring skills across development will be necessary to improve comprehension.
Tweet
To increase reading comprehension, educators will need to provide early and sustained instruction in knowledge, vocabulary, inference generation, and comprehension monitoring.
Key Points
Improving adolescent reading comprehension will require a concerted effort from researchers, educators, and policy makers to forgo short-term gains on measures that tap low-level comprehension for long-term solutions that take years to develop.
An early and sustained focus on developing background knowledge, vocabulary, inference, and comprehension monitoring skills is necessary to improve reading comprehension across grade levels.
Despite decades of reading comprehension research, a limited amount of time is spent using evidence-based methods in classrooms.
Education leaders will need to strengthen teacher preparation programs and professional development to ensure teachers are prepared to use evidence-based practices to meet the literacy needs of their students.
Introduction
According to national and international tests of literacy, such as the Program for International Student Assessment (PISA) and National Assessment of Educational Progress (National Assessment of Education Progress [NAEP]), students in the United States are unable to do relatively easy literacy tasks such as locate relevant information to determine the main idea of a text or make simple inferences (Kastberg, Chan, & Murray, 2016; National Center for Education Statistics, 2017). According to the most recent PISA, U.S. adolescents rank 15th in literacy skills. Results showed that 19% of the 15-year-olds tested scored below a Level 2 (of 6) indicating they had difficulty with tasks such as locating explicitly stated information, recognizing main ideas, and making low-level inferences in a familiar topic. Only 10% of U.S. students achieved a Level 5 indicating that they could organize several pieces of deeply embedded information, and engage in reflective, evaluative, and interpretative tasks in unfamiliar topics. Similarly, the NAEP scores showed that 64% of eighth grade students read at or below a basic level. Unfortunately, these scores have remained relatively flat for many years and have led many educators, researchers, and policy makers to question how well students are being prepared for a job market that increasingly requires self-learning, analytical skills, and transferable knowledge (e.g., Goldman & Pellegrino, 2015).
In this article, we briefly review the theoretical and empirical research in comprehension and consider the reasons for stagnant scores in reading comprehension in the United States. In the first part of the article, we explore different theoretical models for comprehension and then focus on malleable factors that have been shown to be important to comprehension. We conclude with possible solutions for translating research to practice and policies for improving reading comprehension instruction.
The Complexity of Reading Comprehension
Reading comprehension is one of the most complex behaviors in which humans engage. Reading theorists have grappled with how to comprehensively and meaningfully portray reading comprehension and many different theoretical models have been proposed in recent decades (McNamara & Magliano, 2009; Perfetti & Stafura, 2014). These models range from broad theoretical models depicting the relationships and interactions among comprehension subcomponents to models of specific comprehension processes. We review different frameworks and models that have significantly impacted theory development, reading comprehension research, and instruction.
One framework, the Simple View of Reading (SVR), posits that reading comprehension is the product of word decoding and linguistic comprehension (Gough & Tunmer, 1986). Across many languages, research has shown that reading comprehension can be explained by individual differences in these two components, though the relative relationship of the components changes over time (Catts, 2018). Early in development, decoding is more closely associated with reading comprehension than linguistic competence, but once decoding is mastered, linguistic comprehension becomes a better predictor of reading comprehension (e.g., Catts, Adlof, & Weismer, 2005). The SVR has been useful to researchers and practitioners by providing a framework for understanding different profiles of struggling readers including students who struggle primarily due to word-level problems (i.e., dyslexic), comprehension issues (i.e., poor comprehender), or both (i.e., garden variety poor reader). As useful as the model has been, it does not explicate the subcomponents of language or cognitive processes that underlie reading comprehension (Catts, 2018).
Reading comprehension requires the coordination of multiple linguistic and cognitive processes including, but not limited to, word reading ability, working memory, inference generation, comprehension monitoring, vocabulary, and prior knowledge (Perfetti, Landi, & Oakhill, 2005). The multifaceted nature of reading comprehension is reflected in component models that consider subcomponents of comprehension. Component models such as the direct and inferential mediation model (DIME; Ahmed et al., 2016; Cromley, Snyder-Hogan, & Luciw-Dubas, 2010; Oslund, Clemens, Simmons, & Simmons, 2018; Oslund, Clemens, Simmons, Smith, & Simmons, 2016) have been fairly consistent in their findings. These models indicate that vocabulary, both directly and indirectly, is consistently the strongest predictor of reading comprehension for younger adolescents. Furthermore, these models also demonstrated that, although not as strong as vocabulary, inference-making and background knowledge also had strong direct and indirect effects on comprehension. As students get older, inference-making plays a stronger direct role in comprehension than vocabulary (Cromley et al., 2010, replicated by Ahmed et al., 2016). Across studies, vocabulary, inference-making, and background knowledge all influence, both directly and indirectly, reading comprehension from adolescent to young adult readers.
Whereas models such as the SVR and DIME have attempted to identify underlying components to reading comprehension, other theorists have examined the process of reading comprehension. Most theories of comprehension align with Kintsch’s (1988) construction integration (CI) model. During the construction phase, information from the text and the readers’ related knowledge automatically activate. In the subsequent integration phase, activation spreads throughout the memory network, settling on concepts with greater activation and more links to other concepts while suppressing weakly linked concepts. This process occurs iteratively as the reader processes ideas in a text. In this way, readers integrate information from the text with their background knowledge to form an overall mental representation.
Similarly, the RAND reading model, another influential reading framework for research and practice, defined reading comprehension as the process of “extracting and constructing meaning through interaction and involvement with written language” (RAND Reading Study Group, 2002, p. 11). Specifically, reading comprehension is the interaction between reader, text, and task characteristics within a sociocultural context. This model highlights the context-dependent nature of comprehension. A person may achieve a clear understanding of a text when the text is easy and the task simple (e.g., answering multiple choice questions), but the same reader may struggle when encountering complex text on an unfamiliar topic.
Effective Reading Comprehension Instruction
Many linguistic abilities, cognitive processes, and knowledge sources undergird comprehension ability complicating which comprehension components instruction should target (Perfetti & Adlof, 2012). To identify which components are worth measuring, some components serve as “pressure points” that, if changed, would significantly impact students’ comprehension ability. Such components should be integral to reading comprehension, vary across individuals, and represent malleable instructional targets (Perfetti & Adlof, 2012). With this in mind, we briefly review four components of reading comprehension (i.e., inference, knowledge, vocabulary, comprehension monitoring) that play prominent roles across theories of reading comprehension, are integral for understanding text, and represent potentially malleable targets for instruction. This review focuses on higher order comprehension skills but acknowledges the foundational role that efficient word recognition plays in reading comprehension (see Perfetti & Stafura, 2014).
Inference Generation
Inference generation, the ability to integrate information within or across texts using background knowledge to fill in information not explicitly stated, is an essential component of language comprehension (Kendeou, McMaster, & Christ, 2016; Kintsch, 1988). Inference generation is a general skill important for communication and learning at all stages of development. When prompted, even preschool age children can generate causal inferences about events (van den Broek, Lorch, & Thurlow, 1996). Inference ability has been shown to be a unique predictor of reading comprehension across developmental stages (Barth, Barnes, Francis, Vaughn, & York, 2015). Comparing good and poor comprehenders (matched on decoding and vocabulary) shows differences on inferential tasks at the word, sentence, and passage level (Cain & Oakhill, 1999). Compared to skilled comprehenders, poor comprehenders demonstrate difficulties with generating topic-related inferences, integrating words into context, resolving contextual references, and answering inference questions in a logical manner (Long, Oppy, & Seely, 1994; Perfetti & Stafura, 2014).
Not only does inference ability predict reading comprehension, it is also malleable through instruction. In a meta-analysis of inference intervention studies, teaching inferences improved general comprehension as well as inferential and literal comprehension skills (Elleman, 2017). Most interventions included fewer than 10 hr of instruction, implying that teaching inference strategies is useful, and extended practice as a context-independent skill may not be necessary (Willingham, 2017). Effective inference instructional techniques include teaching students to use their background knowledge and integrate it with the information in the text, self-generated elaborations, graphic organizers that connect concepts to one another, and text clues (e.g., Elbro & Buch-Iversen, 2013; Kendeou et al., 2016).
Background Knowledge
A reader’s background knowledge is necessary in building a coherent representation of a text. Well-connected memory storage facilitates quicker retrieval and use of relevant information (Kintsch & Rawson, 2005). Prior content knowledge supports relevant comprehension and learning (Barnes, Dennis, & Haefele-Kalvaitis, 1996). Readers with more prior knowledge consistently outperform readers with less, indicating that increased background knowledge in an area may help less skilled readers compensate for a general comprehension deficit (Schneider, Körkel, & Weinert, 1989). Prior knowledge of a domain predicts text recall for students across development and supports the ability to make inferences (Recht & Leslie, 1988) and learn new words (Kaefer, Neuman, & Pinkham, 2015).
Although educators acknowledge the important role of knowledge in comprehension, very little time in early elementary school is focused on informational text (Duke, 2000). According to survey research reported in the National Survey of Science and Mathematics Education, elementary teachers spend over 80 min on language arts instruction a day compared to an average of 21 min in science and 18 min in social studies (Banilower et al., 2013). This is compounded for less skilled readers who have knowledge deficits (Compton, Miller, Gilbert, & Steacy, 2013), difficulties understanding expository text (Saenz & Fuchs, 2002), and are often pulled from content area classes for additional reading instruction (Banilower et al., 2013).
To alleviate these issues, well-designed programs can integrate domain knowledge acquisition and comprehension instruction. A recent meta-analysis of integrated science and reading programs found a moderate effect for both science and reading outcomes (Talbert, Parrish, & Elleman, 2016). Programs focusing on social studies content have also shown success at promoting knowledge acquisition and reading comprehension (e.g., Guthrie & Klauda, 2014; Vaughn et al., 2013).
Vocabulary
Vocabulary is a robust predictor of reading comprehension across development. Children acquire vocabulary at an astounding rate, on average 2 to 8 root words per day (Biemiller & Slonim, 2001). Once children start independently reading, vocabulary acquisition becomes dependent on exposure to print, not oral language or instruction. Compared to written language, oral language experiences do not provide enough unknown words to foster substantial vocabulary growth (Hayes & Ahrens, 1988). Explicit instruction has also been ruled out as a significant factor in vocabulary acquisition because children are instructed on only a fraction of the words it would take to accrue the 40,000 words estimated to be known by the average high school student (Stahl & Nagy, 2006). Most words are learned implicitly through repeated exposures in multiple contexts over time (Landauer & Dumais, 1997).
Unfortunately, large individual differences in vocabulary size exist in early readers and persist through elementary school (Biemiller & Slonim, 2001). By the end of second grade, disadvantaged students can lag 2 years behind their average peers and 4 years behind those in the upper quartile on vocabulary knowledge. An intense focus on early and continued vocabulary learning makes sense to ameliorate these differences, but it is unclear if enough vocabulary can be explicitly taught to impact students’ general comprehension. Two reviews examining the impact of vocabulary instruction on comprehension found gains on comprehension measures that contained the taught words, but not for general comprehension measures (Elleman, Lindo, Morphy, & Compton, 2009; Wright & Cervetti, 2017). Considering only researcher designed comprehension measures, while all readers benefited from vocabulary instruction, less skilled readers benefited more than typical readers, highlighting the importance of vocabulary for these students (Elleman et al., 2009).
Interactive approaches to word learning appear more effective than ones that rely on definitional types of instruction (Wright & Cervetti, 2017). However, some studies show gains with very little instructional time spent per word (in some cases less than a minute per word), suggesting that at least some type of vocabulary instruction is better than none. Thus, if the instructional goal in a lesson is to improve comprehension, providing brief explanations of words before or during reading may be an efficient way to promote word learning and increase text comprehension. Other generative word learning strategies that go beyond learning individual words—such as learning to use context to derive word meanings (Fukkink & deGlopper, 1998) and morphological analysis (Goodwin & Ahn, 2013)—also improve word learning. In addition, leveraging knowledge networks may be another way to more efficiently teach vocabulary and knowledge simultaneously (Neuman & Wright, 2014).
Comprehension Monitoring and Strategy Instruction
Comprehension monitoring is a metacognitive skill that refers to readers’ ability to reflect on their understanding of a written text (Language and Reading Research Consortium & Yeomans-Maldonado, 2017). Strategic monitoring of text understanding matters (Oakhill, Hartt, & Samols, 2005). The ability to monitor comprehension correlates with reading comprehension and increases over development (Language and Reading Research Consortium & Yeomans-Maldonado, 2017). Young readers and less skilled readers demonstrate weaknesses in detecting inconsistencies within a text (Oakhill et al., 2005). Readers must be able to monitor and use fix-up strategies when comprehension breaks down.
The intent of comprehension strategy instruction is to teach students to actively monitor their comprehension and employ an appropriate strategy to make sense of the text. Comprehension strategy instruction improves comprehension for typically developing (e.g., National Institute of Child Health and Human Development [NICHD], 2000) and struggling readers (e.g., Gersten, Fuchs, Williams, & Baker, 2001). Comprehension strategy instruction is beneficial when multiple strategies are taught, strategies are explicitly modeled, and students gradually assume more responsibility in using strategies independently (Pearson & Dole, 1987).
Implications for Policy
Having covered some contemporary theories and research in reading comprehension, we consider why reading scores for U.S. adolescents have stagnated. We then discuss policy implications and suggest recommendations to improve reading comprehension.
Curriculum and Instruction
Knowledge and vocabulary development
Comprehension strategy instruction is one of the most highly recommended instructional methods for improving comprehension (e.g., NICHD, 2000). Critics, however, contend that an overemphasis on strategy instruction at the expense of knowledge building has led to the stagnation of scores (Willingham, 2006). Also, strategy instruction may result in a shallow representation of a text and interfere with deeper processing of the content (Compton, Miller, Elleman, & Steacy, 2014). Often, the focus of instruction becomes the strategy itself (e.g., learning how to make main idea statements in different texts), instead of flexibly using the strategy to build a coherent representation of the texts being read. Early reviews of strategy instruction revealed that no matter which strategies were combined in a program, effects were the same (Rosenshine & Meister, 1994), suggesting that the strategies themselves might not be the causal mechanism for increasing comprehension, but instead a third variable such as increased engagement and comprehension monitoring might be responsible (Wilkinson & Son, 2011). As noted, knowledge is a key factor in comprehension, and young students get little content exposure in the elementary grades. Some critics cite the pressures from No Child Left Behind (NCLB) legislation, which they believe unintentionally narrowed the curriculum and made school seek short-term gains focusing on skills and strategies to increase standardized test scores (Jones & Workman, 2016).
Knowledge and vocabulary acquisition take time to develop. Given this, an early and sustained focus on knowledge and vocabulary acquisition is necessary across all grade levels. The Common Core State Standards in English Language Arts (CCSS/ELA), adopted by most U.S. states, reflect this focus on knowledge acquisition by recommending an increase in informational text at all grade levels with an increasing emphasis on informational text as students get older (Cervetti & Hiebert, 2015). Yet, standards do not guide curricular decisions about content. Domain knowledge must be thoughtfully presented as concepts build on one another over time. Similarly, vocabulary instruction must be broad and include systematic instruction in academic vocabulary and word learning strategies. Rich instruction in knowledge acquisition and vocabulary development must start as early as possible and continue throughout children’s school career.
Strategy instruction
Although it would be easy to blame strategy instruction as the reason for stagnant scores, strategy instruction does effectively increase comprehension. Also, students with poor comprehension are passive in applying their knowledge and demonstrate difficulties making inferences, even when they have the necessary background knowledge to do so (Barnes et al., 1996). So, increasing background knowledge and vocabulary is only part of the answer to improving reading comprehension. Students also need to become strategic readers who can independently learn from text. What is unclear from research is how to do this most effectively. We know very little about when to teach certain strategies in development. As readers develop, they require more specialized strategies for engaging in text. Content such as history, science, and mathematics require different types of text analysis skills (Shanahan & Shanahan, 2008). It is unclear when it is most beneficial to start teaching advanced disciplinary strategies.
Another question to resolve is how much time to spend on strategies. As noted, just a few hours of inference instruction can be effective (Elleman, 2017), suggesting extended practice may be unnecessary (Willingham, 2017). Research should examine which strategies are most generally effective, which specific strategies at what ages, and how long instruction should last. Awaiting more information, practitioners should teach inference generation and comprehension monitoring across development, while simultaneously teaching vocabulary and domain knowledge.
Text complexity
Another explanation for the lack of growth in adolescent reading comprehension is that students are being exposed to less complex text than in the past (Adams, 2011). A 2-year gap separates the complexity of 12th grade high school student texts and texts for college freshman. What distinguishes students on some standardized tests is the ability to understand more difficult passages (ACT, 2006; Kulesz, Francis, Barnes, & Fletcher, 2016). Variability in text difficulty impacts reading comprehension (McNamara, Graesser, & Louwerse, 2012). Certain linguistic text features make a text more difficult, such as use of rarer words and longer sentences (Williamson, Fitzgerald, & Stenner, 2013). Other factors—such as referential cohesion (i.e., the overlap of concepts across a text), deep cohesion (i.e., words that connect the relationship between ideas in the text), and text type (e.g., narrative, expository)—also affect comprehension (McNamara et al., 2012). Consequently, the CCSS dedicates an entire standard to text complexity, to ensure students are exposed to increasingly difficult text across grades and are, therefore, ready to understand texts when they enter college or the workforce (Hiebert & Mesmer, 2013).
However, some questions about text complexity remain unresolved. Some researchers are concerned about the potential negative impact of increasing text difficulty at younger ages or with less skilled readers (Hiebert & Mesmer, 2013). Most states allow different types of text complexity measurement systems to ensure students are exposed to increasingly complex texts over time. We know little about the impact of increasing text complexity on comprehension, nor do we know the impact of using different measurement systems. Although some methods help to identify increases in text complexity features like sentence length and word frequency, they do little to help teachers identify sets of texts that are topically relevant and cohesive. Text complexity measurement systems that help teachers choose challenging texts and provide information about conceptual overlap among texts would be useful to ensure students are exposed to complex text that supports efficient and effective knowledge and vocabulary acquisition (see Compton et al., 2014).
Assessment
This review focuses on comprehension theory and instruction, but measurement is crucial in comprehension research and practice. The complexity of comprehension makes it difficult to capture. Indeed, comprehension tests are not highly correlated with one another, differentially measure underlying skills, and cannot reliably identify children with comprehension difficulties across different tests (Cutting & Scarborough, 2006; Keenan & Meenan, 2014). Therefore, research and educational decision making must use multiple tests and interpret their scores based on the specific test’s characteristics.
To better understand the contextualized nature of comprehension and provide more information on existing comprehension measures, researchers have started modeling the complex interactions between readers, text, and tasks on different comprehension measures using the RAND framework (Kulesz et al., 2016; Miller et al., 2014). For instance, in a study examining reader, text, and test characteristics (Kulesz et al., 2016) adolescent readers with better vocabulary and background knowledge outperformed others on a standardized reading comprehension measure. Also, the types of questions asked on the test (e.g., literal and inferential) did not predict item difficulty, but the type of text did matter (expository passages were more difficult than narrative passages), underscoring the need to consider the contextual dependencies at play when measuring comprehension.
Most researchers and practitioners would agree that most standardized reading comprehension measures fall short of fully capturing the process of comprehension (Fuchs et al., 2018). In addition to analyzing current measures, researchers and education leaders should design measures that go beyond the low-level skills tapped by multiple choice questions and closed-ended items to measures that provide information about students’ deeper processing of text. As part of the Reading for Understanding (RFU) initiative funded by the Institutes of Education Sciences (IES), one research team has designed a measure intended to capture deeper comprehension than current multiple choice tests (Sabatini, O’Reilly, Halderman, & Bruce, 2014). The Global Integrated Scenario-based Assessments (GISA) require students to integrate information across multiple materials to complete a literacy-based task (e.g., designing a website) and defend a position based on what they learn. ESSA (2015) requires that states use assessments that measure higher order thinking.
Besides globally measuring deep learning, few diagnostic tests pinpoint specific difficulties children encounter in comprehension (Francis et al., 2006). These tests are necessary, so teachers can identify areas of needed support, provide appropriate interventions, and monitor students’ progress (Kendeou et al., 2016). Developing and testing these types of measures across grades and types of learners is necessary to move comprehension instruction forward.
Translating Comprehension Research to Practice
The inadequate translation of research to practice may also contribute to stagnant adolescent reading scores. Despite decades of experimental research in reading comprehension (Scammacca et al., 2016), little classroom time is used to teach evidence-based comprehension practices (Swanson et al., 2016). Current policies advocate evidence-based instructional policies. ESSA emphasizes evidence-based instructional methods, as first proposed by the NCLB. However, these methods are not making their way into practitioners’ hands. One federal agency, IES, has taken an active role in the promotion of evidence-based practices through the evaluation of programs (e.g., What Works Clearinghouse) and the dissemination of findings (e.g., practitioner-friendly practice guides containing practices and strategies vetted through science). IES has also supported important research initiatives in reading comprehension such as the Reading for Understanding initiatives in which interdisciplinary research teams examined different aspects of development in language and reading development. The research findings from these studies are just now being published, but the question remains on whether the findings will be successfully translated to regular classroom practice.
Translating research to practice is complex and will require a multipronged approach. One area of growing interest is the use of practice-embedded educational research (Snow, 2015) in which researchers and practitioners partner to identify solutions for real problems in practice settings using rigorous research methods. Evidence-based programs for increasing vocabulary and comprehension (e.g., Word Generation and STARI) have been successfully developed and widely disseminated using this research framework (Snow, 2015). These types of on-going partnerships between practitioners and researchers may help shrink the research-to-practice gap in literacy by producing effective interventions that practitioners want to use.
In addition to developing and disseminating effective research practices and programs, teacher preparation and professional development will need to be strengthened to improve reading comprehension instruction. One first step is for all states to require that teachers are knowledgeable about evidence-based literacy practices and have expertise in the content they plan to teach. Ensuring that all teachers meet these standards will require collaboration among policy makers, state and district leaders, teacher preparation programs, and accrediting agencies.
Concluding Remarks
Reading comprehension is complex and multifaceted, making it difficult to improve. Stagnant scores of adolescents are likely due to multiple reasons including the ones outlined in this article. Improving adolescent reading comprehension will require a concerted effort from researchers, educators, and policymakers to forgo short-term gains on measures that tap low-level comprehension for long-term solutions that take years to develop. An early and sustained focus on developing background knowledge, vocabulary, inference, and comprehension monitoring skills is necessary to improve reading comprehension across grade levels.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
