BRIEF RESEARCH REPORT article

The use of new technologies for improving reading comprehension.

\r\nAgnese Capodieci*

  • 1 Department of General Psychology, University of Padova, Padua, Italy
  • 2 Azienda Sociosanitaria Ligure 5 Spezzino, La Spezia, Italy

Since the introduction of writing systems, reading comprehension has always been a foundation for achievement in several areas within the educational system, as well as a prerequisite for successful participation in most areas of adult life. The increased availability of technologies and web-based resources can be a really valid support, both in the educational and clinical field, to devise training activities that can also be carried out remotely. There are studies in current literature that has examined the efficacy of internet-based programs for reading comprehension for children with reading comprehension difficulties but almost none considered distance rehabilitation programs. The present paper reports data concerning a distance program Cloze , developed in Italy, for improving language and reading comprehension. Twenty-eight children from 3rd to 6th grade with comprehension difficulties were involved. These children completed the distance program for 15–20 min for at least three times a week for about 4 months. The program was presented separately to each child, with a degree of difficulty adapted to his/her characteristics. Text reading comprehension (assessed distinguishing between narrative and informative texts) increased after intervention. These findings have clinical and educational implications as they suggest that it is possible to promote reading comprehension with a distance individualized program, avoiding the need for the child displacements, necessary for reaching a rehabilitation center.

Introduction

Reading comprehension is a fundamental cognitive ability for children, that supports school achievement and successively participation in most areas of adult life ( Hulme and Snowling, 2011 ). Therefore, children with learning disabilities (LD) and special educational needs who show difficulties in text comprehension, sometimes also in association with other problems, may have an increased risk of life and school failure ( Woolley, 2011 ). Reading comprehension is, indeed, a complex cognitive ability which involves not only linguistic (e.g., vocabulary, grammatical knowledge), but also cognitive (such as working memory, De Beni and Palladino, 2000 ), and metacognitive skills (both for the aspects of knowledge and control, Channa et al., 2015 ), and, more specifically, higher order comprehension skills such as the generation of inferences ( Oakhill et al., 2003 ).

Recently, due to the diffusion of technology in many fields of daily life, text comprehension at school, at home during homework, and at work is based on an increasing number of digital reading devices (computers and laptops, e-books, and tablet devices) that can become a fundamental support to improve traditional reading comprehension and learning skills (e.g., inference generation).

Some authors contrasted in children with typical development the effects of the technological interface on reading comprehension vs printed texts ( Kerr and Symons, 2006 ; Rideout et al., 2010 ; Mangen et al., 2013 ; Singer and Alexander, 2017 ; Delgado et al., 2018 ). Results were consistent and showed a worse comprehension performance in screen texts compared to printed texts for children ( Mangen et al., 2013 ; Delgado et al., 2018 ) and adolescents who nonetheless showed a preference for digital texts compared to printed texts ( Singer and Alexander, 2017 ). Regarding children with learning problems, only few studies considered the differences between printed texts and digital devices ( Chen, 2009 ; Gonzalez, 2014 ; Krieger, 2017 ) finding no significant differences, suggesting that the use of compensative digital tools for children with a learning difficulty could be a valid alternative with respect to the traditional written texts in facilitating their academic and work performance. This conclusion is also supported by the results of a meta-analysis ( Moran et al., 2008 ), regarding the use of digital tools and learning environments for enhancing literacy acquisition in middle school students, which demonstrates that technology can improve reading comprehension.

Different procedures and abilities are targeted in the international literature concerning computerized training programs for reading comprehension. In particular, various studies include activities promoting cognitive (e.g., vocabulary, inference making) and metacognitive (e.g., the use of strategies, comprehension monitoring, and identification of relevant parts in a text) components of reading comprehension. Table 1 reports the list of papers proposing computerized training programs with a summary of the findings encountered. Participants involved cover different ages and school grades, the majority belonging to middle school and high school. The general outcome of the studies is positive due to a significant improvement in comprehension skills after the training program with long-lasting effects also during follow-up; indeed, the majority of participants involved in training programs outperformed their peers assigned to comparison groups and maintained their improvements. Specifically, several studies ( O’Reilly et al., 2004 ; Magliano et al., 2005 ; McNamara et al., 2006 ) used the iSTART program with adolescents and young adults. This program promotes self-explanation, prior knowledge and reading strategies to enhance understanding of descriptive scientific texts. Results demonstrated that students who followed the iSTART program received more benefits than their peers, improving self-explanation and summarization. Additionally, strategic knowledge was a relevant factor for the outcome in comprehension tasks including multiple choice questions: students who already possessed good strategic knowledge improved their accuracy when answering to bridging inference questions, whereas students with low strategic knowledge became more accurate with text-based questions. Another program, ITSS, was used with younger students ( Meyer et al., 2011 ; Wijekumar et al., 2012 , 2013 , 2017 ), with the objective to support activities based on identifying main parts and key words in a text and classifying information in a hierarchical order. Positive outcomes were found also with such program since students who followed the ITSS program significantly improved text comprehension compared to their peers in the control group.

www.frontiersin.org

Table 1. Synthesis of the main results of the computerized training programs on comprehension present in the literature.

Although most of the literature deals with typical development, also cases of students with learning difficulties were considered. For example, Potocki et al. (2013) (see also Potocki et al., 2015 ) examined the effects of two different computerized programs with specific aims: one focusing on comprehension features, such as inference making and the analysis of text structure, the other considering decoding skills. Both training programs brought some benefits to reading comprehension, however larger effects were found with the program focused on comprehension with long-lasting effects in listening and reading comprehension (see also Kleinsz et al., 2017 ). Studies by Johnson-Glenberg (2005) and Kim et al. (2006) , using respectively the programs 3D Readers and CACSR, were able to promote reading comprehension abilities in middle school students through metacognitive activities. Thanks to these programs students also became more aware of reading strategies and implemented them more successfully during text comprehension. In particular, a study by Niedo et al. (2014) , obtained positive results on silent reading in a small group of children struggling with reading using the “cloze” procedure. This procedure proposes exercises in which parts of a text, typically words, are missing and participants are required to complete the text guessing what is missing.

Thus, computerized programs generally seem to improve reading comprehension skills. However, it should be noticed that, in most cases, students were trained at school, without the personalized support of a clinician taking into consideration the cognitive and psychological needs of the child. In particular, to our knowledge, no program examined the effects of an internet-based distance reading comprehension program which allows the child to be trained at home in a personalized way. A useful aspect of an internet-based distance training is that the psychologist can monitor with the application ( app ) the child’s results and activities and write him/her some motivational messages, reducing the attritions present in programs carried out at home with the only supervision of parents. Literature concerning distance trainings is still rare, however, some evidence suggests that these programs may represent a good integration to other types of intervention, usually carried out at school, in a rehabilitation center or at home (e.g., Mich et al., 2013 ).

Therefore, despite still preliminary, we think that it is relevant to present data about a distance program developed in Italy named Cloze ( Cornoldi and Bertolo, 2013 ), devised for rehabilitation purposes but with potential implication also for educational contexts. Cloze has been developed to promote inferential abilities both at a sentence- and discourse-level using the “cloze” procedure. Several findings in the literature demonstrate that abilities, such as anticipating text parts and inference making, bring improvements in text comprehension (e.g., Yuill and Oakhill, 1988 ) and it has been shown that one way to promote inferential competences is to improve the ability to predict parts of the text that are missing or that follow, considering the available information: the “cloze” technique appears to be one of the most successful ways for this purpose (e.g., Greene, 2001 ).

In the current study the effectiveness of this training program has been tested on a clinical population who exhibited, for various reasons, difficulties in reading comprehension. Participants were 28 children (16 male and 12 female) attending a private practice for learning difficulties in the city of La Spezia, in the north-west of Italy, from 3rd to 6th school grade (5 of 3rd, 9 of 4th, 11 of 5th and 3 of 6th grade), with a mean age of children of M = 9.79 years (SD = 1.03). Seventeen children had a current or past speech disorder: of these children 10 also had a LD (Learning Disabilities) and one was bilingual (speech problems were not due to bilingualism). The other 11 children had a LD or important learning difficulties, and one of them had also ADHD (Attention Deficit/Hyperactivity Disorder). For the goals of the study, all these children were considered together as they all presented a severe reading comprehension difficulty as reported by parents and teachers and confirmed by the initial assessment.

All children had received a comprehensive psychological assessment (see Table 2 ), adapted to their particular needs and ages. In particular all children had an IQ >80 assessed with the Wechsler Intelligence Scale for Children-IV (WISC-IV; Wechsler, 2003 ) and did not have anxiety disorders, mood affective disorders or other developmental disorders, with the exception of the cases with language disorder and the case with ADHD. Children were not receiving any additional treatment, including medication. Written consent was obtained from the children’s parents in the context of the private practice.

www.frontiersin.org

Table 2. Main characteristics of the sample in terms of reading and cognitive abilities.

Materials and Methods

Pre-/post-test assessment and procedure of the training.

Each child started a training program through the distance rehabilitation platform Ridinet, using the Cloze app, after the assessment of learning and cognitive abilities, including comprehension assessment with two texts, one narrative and one informative ( Cornoldi and Carretti, 2016 ; Cornoldi et al., 2017 ). Connection to the Ridinet web site was required in order to access to the app, three or four times a week for more or less 15/20 min. The period of use was of 3 months for 6 children and 4 months for 22 children. After this period children’s comprehension was assessed again. Additionally, some questions were asked to parents and children about the app’s utility and pleasantness. In particular, children were asked: “Do you think the program helped you improve your text comprehension skills?,” “Did you like doing this program instead of the same exercises on paper?”; and parents were asked: “Was it difficult to start the Cloze activities on days when it had to be done?,” “Compared to the beginning of the treatment, how do you currently judge the ability of your child to understand the texts?”. For all questions, except the last one, the answer had to be given on a 5-point scale with 1 = not at all, 2 = a little, 3 = enough, 4 = very, 5 = very much. For the last question the answer changed on a 4-point scale with 1 = got worse, 2 = unchanged, 3 = slightly improved, and 4 = greatly improved.

Comprehension Tasks

Reading comprehension was assessed with two texts, the first narrative and the other informative, taken from Italian batteries for the assessment of reading ( Cornoldi and Carretti, 2016 ; Cornoldi et al., 2017 ). The texts range between 226 and 455 words in length, and their length increases with school grade (in order to have texts and questions matching the degrees of expertise at different grades the batteries include a different pair of texts for each grade). Students read the text in silence at their own pace, then answer a variable number of multiple-choice questions (depending on school grade), choosing one of four possible answers. There is no time limit, and students can reread the text whenever they wish. The final score is calculated as the total number of correct answers for each text. Alpha coefficients, as reported by the manuals, range between 0.61 and 0.83. For the purposes of the study we decided to use the same two comprehension texts, at pre-test and post-test, as the procedure offered the opportunity of directly examining and showing to parents changes in comprehension and previous evidence had shown the absence of relevant retest effects with this material in a retest carried out after 3 months ( Viola and Carretti, 2019 ).

Distance Rehabilitation Program: Cloze

Cloze ( Cornoldi and Bertolo, 2013 ) is an app for the promotion of text comprehension with the specific aim to recover processes of lexical and semantic inference. At each work session the child works with texts that lack words and must complete the empty spaces by choosing the correct alternative from those automatically proposed by the app, so that the text becomes congruent. The program is adaptive, as text complexity and proportion of missing words vary according to the previous level of response, and is designed for children who have weaknesses in written text comprehension, mainly due to poor skills in lexical and semantic inferential processes. The app also allows to enhance a set of language skills (phonology, syntax, semantics) which contribute to ensuring the fluidity of text and production processing. The recommended age range for the use of this program is between 7 and 14 years. In this study the semantic mode (only content words may be missing and no syntactic cues can be used for deciding between the alternatives) was proposed to 21 children and the syntactic mode (where all words may be missing) to 7 children. The mode type selected for each child depends from the performance at pre-test and diagnosis. A clinician, co-author of the present study (LB), monitored the child’s results and activities with the app and sent him/her from time to time some motivational messages. The motivational messages were typically sent once a week for congratulating with children for the work done and check with him/her possible problems emerged. Training lasted from 3 to 4 months and involved between 3 and 4 sessions of 15–20 min per week. The variation in duration depended on the decision of each individual family. In fact, children were required to use the software for about 4 months or in any case for a minimum period of 3 months (choice made by six families).

Effects on Reading Comprehension of Cloze Training

All analyses were carried out with SPSS 25 ( IBM Corp, 2017 ). A preliminary analysis found that all the examined variables met the assumptions of normality (K-S between 0.106 and 0.143, p > 0.05). Then, we compared the reading comprehension performance of children before and after the computerized training with Cloze . For this analysis, a repeated measure Analysis of Variance (ANOVA) was conducted on comprehension scores to examine the differences in the whole group of children between the scores obtained before and after the training. A significant difference was found for both comprehension texts [ F (1,27) = 22.37, p < 0.001, η 2 p = 0.453 and F (1,27) = 38.90, p < 0.001, η 2 p = 0.599, respectively]. Possible differences between the two training modalities (semantic vs syntactic) and between different training periods (3 months vs 4 months) were then analyzed; no significant differences emerged between groups in both cases [ F (1,27) < 1].

Secondly, to analyze the role of individual differences at pre-test, the standardized training gain score (STG; Jaeggi et al., 2011 ) – computed by subtracting post-test score minus pre-test score, divided by the SD of the pre-test – was calculated for the two texts comprehension. Pearson correlations were computed between the STG and the variable collected at pre-test (reading speed and errors, WISC IV – Full scale IQ, Verbal Comprehension, Perceptual Reasoning, Working Memory and Processing Speed indexes). The only significant correlation was between STG of the narrative text and Verbal Comprehension Index of the WISC-IV Scale ( r = 0.38, p = 0.048). Finally, individual improvements from pre- to post-test were also confirmed considering changes in performance in terms of standard deviation in relations to norms (provided by the manual). Table 3 shows the number of children for each comprehension text who improved their performance moving from a performance at least 2 standard deviations or between 1 and 2 negative standard deviations under the mean to a performance above one negative standard deviation.

www.frontiersin.org

Table 3. Changes in performance in relations to norms (provided by the manual) after the training program Cloze.

Perceived Utility, Pleasantness, Parents and Child’s Improvements of Cloze

Results concerning the answers of parents and children about utility, pleasantness and self-perceived efficacy of the app, were also analyzed. At the first question, addressing children’s perceived improvement in comprehension skills, more than half of the sample chose the alternatives “very” or “very much” (15 “very” and 5 “very much”), only 1 child answered “a little” and the others chose “enough.” At the second question, about the pleasure of doing this kind of activity instead of pen and paper activities, all children answered “very” or “very much.” Concerning parents’ questions, at the first question about the difficulty to start the Cloze activity, only one parent answered “enough,” a quarter of the sample chose “a little” (seven families) and all the other 20 families chose the alternative “not at all.” At the last question about the perceived training efficacy on their child’s performance, the large majority of the families chose “slightly improved” or “greatly improved” and only three parents thought their children’s ability had remained unchanged. However, no correlations between parents and child’s perceived improvements and STG in reading comprehension were found.

The present study examined the effects of the use of Cloze , a distance rehabilitation program focused on inference skills, for improving reading comprehension, on the basis of the hypothesis that, being inference making related to reading comprehension at different ages (e.g., Oakhill and Cain, 2012 ), positive effects of the training activities on reading comprehension should be found.

Concerning the efficacy of computer-assisted training programs, literature highlights that many training programs are devised for an educational context. Results are generally encouraging with positive effects on reading comprehension, measured with materials different from those practiced during the training. However, few studies analyzed the efficacy in children with specific reading comprehension problems, and no studies considered the possibility of carrying out a training at home under the distance supervision of an expert. The latter characteristics are those that make the Cloze peculiar compared to the existent literature. Cloze is indeed based on a rehabilitation online platform which allows the child to complete personalized training activities several times a week, without moving from his/her home, and concurrently enabling the clinician to monitor the child’s progress or manage activities’ characteristics. The advantage of this procedure is twofold: on one hand it increases the potential number of training sessions per week, on the other hand it permits to save the necessary time to reach the center for rehabilitation and to reduce the costs of the intervention.

The preliminary data on Cloze were generally positive: children, working on either two slightly different versions of the same program, showed a generalized improvement in reading comprehension tasks and, together with their families, expressed appreciation for the pleasantness and the efficacy of the program. Encouraging results emerged also from the analysis of individual improvements referring to normative scores, as reported in Table 3 : most of the children’s performance migrated from a highly negative level to an average level.

It is noticeable that the efficacy of the training was assessed with materials different from those practiced during the training sessions, since reading comprehension tasks required to read a paper text and complete a series of multiple-choice questions. In future studies it would be interesting to analyze the effects of the program on skills known to be related to text comprehension, such as vocabulary or comprehension monitoring, for example. There is good reason to believe that since these variables are highly predictive of comprehension skills (and given that training in these skills sometimes improve comprehension; e.g., Beck et al., 1982 ; see also Hulme and Snowling, 2011 ), training that specifically targets comprehension might, in turn, lead to improvements in vocabulary or comprehension monitoring skills. Further studies are needed to explore this hypothesis.

A second relevant finding of the present study is the presence of a positive correlation between the gain obtained in one of the reading comprehension text (the narrative one) and the Verbal Comprehension Intelligence Quotient (VCIQ) index of the WISC-IV battery, showing that children who started with more resources in verbal intelligence achieved greater improvements in text comprehension at least with one type of text through the Cloze . The activities probably required to develop some kind of strategies, and for this reason students with larger verbal intellectual resources, who were presumably more able to develop new strategies, were more advantaged. Indeed, this amplification effect is usually found when training activities require the development of strategies ( von Bastian and Oberauer, 2014 ). Such result has clinical and educational implications, inviting professionals and teachers to consider children’s starting resources and, if necessary, to combine activities conducted through distance rehabilitation programs with personal intervention sessions that could teach strategies and promote a metacognitive approach to reading comprehension. However, some limitations of the present study must be acknowledged. Firstly, study did not include a control group, therefore findings should be taken with caution, although normative data and previous results obtained with the same test offer support to the robustness of our results and the use of normative data offers a control measure of how reading comprehension skills are acquired in typically developing children without specific training, therefore functioning as a sort of passive control group. Secondly, the treated group, although characterized by a common reading comprehension difficulty, was partly heterogeneous, as children attended different grades and could have different diagnoses. Unfortunately, the limited number of subjects, with the consequence that it was not possible to form groups defined both by the grade and the diagnosis, did not permit to make analyses taking into account the grade and the diagnosis as between-subjects factors. Future studies should examine a more homogeneous population or consider a larger sample of children, giving more information about the efficacy of training in different children population. Additionally, the fact that the treatment was concluded with the post-training assessment did not offer the opportunity to further examine the procedure and maintenance effects with a follow-up. Despite the limitations, this study offers evidence concerning the efficacy of new methods, based on computer-assisted training programs that could be beneficial in training high-level skills such as comprehension and inference generation. Such tools can be extremely worthwhile for struggling readers who may need to receive further attention in mastering higher level reading comprehension.

Data Availability Statement

The datasets generated for this study are available on request to the corresponding author.

Ethics Statement

Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

Author Contributions

AC, CC and BC contributed to the design and implementation of the research. LB provided the data. BC organized the database. AC performed the statistical analysis. ED did the literature research and wrote the section about the review of the literature. AC and BC wrote the other sections. CC contributed to the manuscript revision, read and approved the submitted version.

The present work was carried out within the scope of the research program Dipartimenti di Eccellenza (art.1, commi 314-337 legge 232/2016), which was supported by a grant from MIUR to the Department of General Psychology, University of Padua and partially supported by a grant (PRIN 2015, 2015AR52F9_003) to Cesare Cornoldi funded by the Italian Ministry of Research and Education (MIUR).

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Beck, I. L., Perfetti, C. A., and McKeown, M. G. (1982). Effects of long-term vocabulary instruction on lexical access and reading comprehension. J. Educ. Psychol. 74, 506–521. doi: 10.1037/0022-0663.74.4.506

CrossRef Full Text | Google Scholar

Channa, M. A., Nordin, Z. S., Siming, I. A., Chandio, A. A., and Koondher, M. A. (2015). Developing reading comprehension through metacognitive strategies: a review of previous studies. Eng. Lang. Teach. 8, 181–186. doi: 10.5539/elt.v8n8p181

Chen, H. (2009). Online reading comprehension strategies among fifth- and sixth-grade general and special education students. Educ. Res. Perspect. 37, 79–109.

Google Scholar

Cornoldi, C., and Bertolo, L. (2013). Cloze Ridinet. Bologna: Anastasis.

Cornoldi, C., and Carretti, B. (2016). Prove MT-3-Clinica. Firenze: Giunti Edu.

Cornoldi, C., Carretti, B., and Colpo, C. (2017). Prove MT-Kit Scuola. Dalla valutazione degli Apprendimenti di Lettura E Comprensione Al Potenziamento. [MT-Kit for the Assessment In The School. From Reading Assessment To Its Enhancement]. Firenze: Giunti Edu.

Cullen, J. M., Alber-Morgan, S. R., Schnell, S. T., and Wheaton, J. E. (2014). Improving reading skills of students with disabilities using headsprout comprehension. Remed. Spec. Educ. 35, 356–365. doi: 10.1177/0741932514534075

De Beni, R., and Palladino, P. (2000). Intrusion errors in working memory tasks: are they related to reading comprehension ability? Learn. Individ. Differ. 12, 131–143. doi: 10.1016/s1041-6080(01)00033-4

Delgado, P., Vargas, C., Ackerman, R., and Salmerón, L. (2018). Don’t throw away your printed books: a meta-analysis on the effects of reading media on reading comprehension. Educ. Res. Rev. 25, 23–38. doi: 10.1016/j.edurev.2018.09.003

Gonzalez, M. (2014). The effect of embedded text-to-speech and vocabulary eBook scaffolds on the comprehension of students with reading disabilities. Intern. J. Spec. Educ. 29, 111–125.

Greene, B. (2001). Testing reading comprehension of theoretical discourse with cloze. J. Res. Read. 24, 82–98. doi: 10.1111/1467-9817.00134

Hulme, C., and Snowling, M. J. (2011). Children’s reading comprehension difficulties: nature, causes, and treatments. Curr. Direct. Psychol. Sci. 20, 139–142. doi: 10.1177/0963721411408673

IBM Corp (2017). IBM SPSS Statistics for Windows, Version 25.0. Armonk, NY: IBM Corp.

Jaeggi, S. M., Buschkuehl, M., Jonides, J., and Shah, P. (2011). Short-and long-term benefits of cognitive training. Proc. Natl. Acad. Sci. U.S.A. 108, 10081–10086. doi: 10.1073/pnas.1103228108

Johnson-Glenberg, M. C. (2005). Web-based training of metacognitive strategies for text comprehension: focus on poor comprehenders. Read. Writ. 18, 755–786. doi: 10.1007/s11145-005-0956-5

Kerr, M. A., and Symons, S. E. (2006). Computerized presentation of text: effects on children’s reading of informational material. Read. Writ. 19, 1–19. doi: 10.1007/s11145-003-8128-y

Kim, A. H., Vaughn, S., Klingner, J. K., Woodruff, A. L., Klein Reutebuch, C., and Kouzekanani, K. (2006). Improving the reading comprehension of middle school students with disabilities through computer-assisted collaborative strategic reading. Remed. Spec. Educ. 27, 235–249. doi: 10.1177/07419325060270040401

Kleinsz, N., Potocki, A., Ecalle, J., and Magnan, A. (2017). Profiles of French poor readers: underlying difficulties and effects of computerized training programs. Learn. Individ. Differ. 57, 45–57. doi: 10.1016/j.lindif.2017.05.009

Krieger, R. (2017). The Effect of Electronic Text Reading on Reading Comprehension Scores of Students with Disabilities. Master thesis, Governors State University, Park, IL.

Leong, C. K. (1992). Enhancing reading comprehension with text-to-speech (DECtalk) computer system. Read. Writ. 4, 205–217. doi: 10.1007/bf01027492

Magliano, J. P., Todaro, S., Millis, K., Wiemer-Hastings, K., Kim, H. J., and McNamara, D. S. (2005). Changes in reading strategies as a function of reading training: a comparison of live and computerized training. J. Educ. Comput. Res. 32, 185–208. doi: 10.2190/1ln8-7bqe-8tn0-m91l

Mangen, A., Walgermo, B. R., and Brønnick, K. (2013). Reading linear texts on paper versus computer screen: effects on reading comprehension. Intern. J. Educ. Res. 58, 61–68. doi: 10.1016/j.ijer.2012.12.002

McNamara, D. S., O’Reilly, T. P., Best, R. M., and Ozuru, Y. (2006). Improving adolescent students’ reading comprehension with iSTART. Intern. J. Educ. Res. 34, 147–171. doi: 10.2190/1ru5-hdtj-a5c8-jvwe

Meyer, B. J., Wijekumar, K. K., and Lin, Y. C. (2011). Individualizing a web-based structure strategy intervention for fifth graders’ comprehension of nonfiction. J. Educ. Psychol. 103, 140–168. doi: 10.1037/a0021606

Mich, O., Pianta, E., and Mana, N. (2013). Interactive stories and exercises with dynamic feedback for improving reading comprehension skills in deaf children. Comput. Educ. 65, 34–44. doi: 10.1016/j.compedu.2013.01.016

Moran, J., Ferdig, R. E., Pearson, P. D., Wardrop, J., and Blomeyer, R. L. Jr. (2008). Technology and reading performance in the middle-school grades: a meta-analysis with recommendations for policy and practice. J. Liter. Res. 40, 6–58. doi: 10.1080/10862960802070483

Niedo, J., Lee, Y. L., Breznitz, Z., and Berninger, V. W. (2014). Computerized silent reading rate and strategy instruction for fourth graders at risk in silent reading rate. Learn. Disabil. Q. 37, 100–110. doi: 10.1177/0731948713507263

Oakhill, J. V., and Cain, K. (2012). The precursors of reading ability in young readers: evidence from a four-year longitudinal study. Sci. Stud. Read. 162, 91–121. doi: 10.1080/10888438.2010.529219

Oakhill, J. V., Cain, K., and Bryant, P. E. (2003). The dissociation of word reading and text comprehension: evidence from component skills. Lang. Cogn. Process. 18, 443–468. doi: 10.1080/01690960344000008

O’Reilly, T. P., Sinclair, G. P., and McNamara, D. S. (2004). “iSTART: A web-based reading strategy intervention that improves students’ science comprehension,” in Proceedings of the IADIS International Conference Cognition and Exploratory Learning in Digital Age: CELDA 2004 , eds D. G. Kinshuk and P. Isaias (Lisbon: IADIS), 173–180.

Ortlieb, E., Sargent, S., and Moreland, M. (2014). Evaluating the efficacy of using a digital reading environment to improve reading comprehension within a reading clinic. Read. Psychol. 35, 397–421. doi: 10.1080/02702711.2012.683236

Potocki, A., Ecalle, J., and Magnan, A. (2013). Effects of computer-assisted comprehension training in less skilled comprehenders in second grade: a one-year follow-up study. Comput. Educ. 63, 131–140. doi: 10.1016/j.compedu.2012.12.011

Potocki, A., Magnan, A., and Ecalle, J. (2015). Computerized trainings in four groups of struggling readers: specific effects on word reading and comprehension. Res. Dev. Disabil. 45, 83–92. doi: 10.1016/j.ridd.2015.07.016

Rideout, V. J., Foehr, U. G., and Roberts, D. F. (2010). Generation M 2: Media in the Lives of 8-to 18-Year-Olds. San Francisco, CA: Henry J. Kaiser Family Foundation.

Singer, L. M., and Alexander, P. A. (2017). Reading on paper and digitally: what the past decades of empirical research reveal. Rev. Educ. Res. 87, 1007–1041. doi: 10.3102/0034654317722961

Viola, F., and Carretti, B. (2019). Cambiamentinelleabilità di letturanelcorso di unostesso anno scolastico [Changes in readingskillsduring the sameschoolyear]. Dislessia 16, 147–159. doi: 10.14605/DIS1621902

von Bastian, C. C., and Oberauer, K. (2014). Effects and mechanisms of working memory training: a review. Psychol. Res. 78, 803–820. doi: 10.1007/s00426-013-0524-526

Wechsler, D. (2003). Wechsler Intelligence Scale For Children–Fourth Edition (WISC-IV). San Antonio, TX: The Psychological Corporation.

Wijekumar, K. K., Meyer, B. J., and Lei, P. (2012). Large-scale randomized controlled trial with 4th graders using intelligent tutoring of the structure strategy to improve nonfiction reading comprehension. Educ. Technol. Res. Dev. 60, 987–1013. doi: 10.1007/s11423-012-9263-4

Wijekumar, K. K., Meyer, B. J., and Lei, P. (2013). High-fidelity implementation of web-based intelligent tutoring system improves fourth and fifth graders content area reading comprehension. Comput. Educ. 68, 366–379. doi: 10.1016/j.compedu.2013.05.021

Wijekumar, K. K., Meyer, B. J., and Lei, P. (2017). Web-based text structure strategy instruction improves seventh graders’ content area reading comprehension. J. Educ. Psychol. 109, 741–760. doi: 10.1037/edu0000168

Woolley, G. (2011). “Reading comprehension,” in Reading Comprehension: Assisting Children With Learning Difficulties , ed Springer Science+Business Media (Dordrecht, NL: Springer), doi: 10.1007/978-94-007-1174-7_2

Yuill, N., and Oakhill, J. (1988). Effects of inference awareness training on poor reading comprehension. Appl. Cogn. Psychol. 2, 33–45. doi: 10.1002/acp.2350020105

Keywords : reading comprehension, training, distance rehabilitation program, digital device, Cloze app

Citation: Capodieci A, Cornoldi C, Doerr E, Bertolo L and Carretti B (2020) The Use of New Technologies for Improving Reading Comprehension. Front. Psychol. 11:751. doi: 10.3389/fpsyg.2020.00751

Received: 20 November 2019; Accepted: 27 March 2020; Published: 23 April 2020.

Reviewed by:

Copyright © 2020 Capodieci, Cornoldi, Doerr, Bertolo and Carretti. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Agnese Capodieci, [email protected] ; Laura Bertolo, [email protected]

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

U.S. flag

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

  • Publications
  • Account settings
  • Advanced Search
  • Journal List
  • HHS Author Manuscripts

Logo of nihpa

The Contribution of Text Characteristics to Reading Comprehension: Investigating the Influence of Text Emotionality

Sage e. pickren.

Vanderbilt University, Nashville, Tennessee, USA

Maria Stacy

Southern Illinois University, Carbondale, USA

Stephanie N. Del Tufo

University of Delaware, Newark, USA

In the current study, we examined relations between text features (e.g., word concreteness, referential cohesion) and reading comprehension using multilevel logistic models. The sample was 158 native English-speaking students between 8 years 8 months and 11 years 2 months of age with a wide range of reading ability. In line with the simple view of reading, decoding ability and language comprehension were associated with reading comprehension performance. Text characteristics, including indices of word frequency, number of pronouns, word concreteness, and deep cohesion, also predicted unique variance in reading comprehension performance over and above the simple view’s components. Additionally, the emotional charge of text (i.e., lexical ratings of arousal) predicted reading comprehension beyond traditional person-level and text-based characteristics. These findings add to a small but growing body of evidence suggesting that it is important to consider emotional charge in addition to person-level and text-based characteristics to better understand reading comprehension performance.

Reading comprehension is an essential lifelong skill that can affect students’ ability to learn information across core curricula. Performance in English, science, and history are all impacted by reading mastery ( RAND Reading Study Group, 2002 ). In the current study, we sought to better understand the multidimensional nature of reading comprehension and identify important variables that influence reading comprehension success in young students. Specifically, we examined how metrics of text-based characteristics, including emotional charge, influence reading comprehension above and beyond known person-level predictors of reading comprehension.

Whereas decades of research on developing readers have pinpointed person-level characteristics known to be robust predictors of reading comprehension (e.g., decoding ability, language comprehension; Hjetland et al., 2019 ; Kendeou, van den Broek, White, & Lynch, 2009 ; Kim, 2017 ), comparatively less is known about how text-based features, such as word frequency, number of pronouns, syntactic simplicity, word concreteness, and cohesion, impact reading comprehension ( Foorman, Petscher, & Herrera, 2018 ; Kendeou, Savage, & van den Broek, 2009 ; Verhoeven & van Leeuwe, 2008 ). In the last decade, however, there has been increasing interest in how these text-based features may play a role in the reading comprehension process in developing readers. This is due in part to the more recent focus on text complexity in classroom instruction ( National Governors Association Center for Best Practices & Council of Chief State School Officers, 2010 ) and the development of tools that can more efficiently quantify these factors (e.g., Coh-Metrix; Graesser, McNamara, Louwerse, & Cai, 2004 ). Emotional charge of texts in particular has received considerably less attention even though it has been theorized to influence reading comprehension outcomes ( Bourg, 1996 ; Mouw, Van Leijenhorst, Saab, Danel, & van den Broek, 2019 ). Further, even fewer studies have examined lexical ratings of emotional charge.

Capturing the emotional charge of text may be important because it potentially facilitates situation model formation (i.e., the integration of a reader’s background knowledge with information presented in the text that then results in an understanding of what is being read; van Dijk & Kintsch, 1983 ) by increasing reader engagement ( Gernsbacher, Goldsmith, & Robertson, 1992 ; Kneepkens & Zwaan, 1995 ). This would suggest that having a more emotionally charged text may be a vehicle for enhancing reading comprehension. However, most empirical studies examining the emotional charge of a text have focused on readers’ reactions to a text rather than emotion as a direct feature of the text (e.g., Gernsbacher, 1997b ; Kneepkens & Zwaan, 1995 ), and no study that we are aware of has examined lexical ratings of emotional charge along with other text features, such as word concreteness, word frequency, and cohesion. Given the potential that emotional charge may have for enhancing readers’ engagement during text processing, such knowledge could be especially informative for developing readers and their teachers. To our knowledge, the current study is the first to examine how metrics of text-based characteristics, namely, emotional charge, are associated with reading comprehension, above and beyond person-level predictors and other text features known to impact reading comprehension.

Theories of Reading Comprehension

In the developmental literature, arguably the most prominent reading comprehension model is the simple view of reading (SVR; Hoover & Gough, 1990 ), which is centered on person-level influences on reading comprehension. This framework asserts that reading comprehension is determined by the interaction between a reader’s ability to decode text and his or her linguistic comprehension, and these two skills have been found to account for a statistically significant proportion of variance in reading comprehension ( Hjetland et al., 2019 ; Kendeou, Savage, & van den Broek, 2009 ; Kim, 2017 ). Although the SVR framework has influenced the field of reading research for over three decades, it is limited by the fact that it focuses only on reader characteristics related to reading comprehension. In contrast to the SVR, discourse processing models, such as the construction–integration model ( Kintsch, 1988 ) and the reading systems framework (RSF; Perfetti, 1999 ), consider the influence of reader and text characteristics. Both models assert that reading comprehension involves text-driven, bottom-up processes and reader-driven, top-down processes that interact to affect comprehension.

More recently, Francis, Kulesz, and Benoit (2018) proposed a complete view of reading (CVR i ) framework that accounts for elements of both reader and discourse processing models. The CVR i incorporates reader skills, text and linguistic discourse elements, and developmental theory, which allows for building personalized models of reading comprehension that depict the effects of many variables and their interactions using multilevel statistics. The CVR i framework arguably best captures the multidimensional nature of reading comprehension by combining reader characteristics (e.g., skills, background knowledge, age) with those presented in the text (e.g., genre, complexity).

Whereas a vast literature exists on how reader characteristics impact reading comprehension, comparatively fewer studies have examined the influence of text-based features on reading comprehension outcomes above and beyond person-level characteristics. Notably, most of these investigations have focused on adolescent (e.g., Barth, Tolar, Fletcher, & Francis, 2014 ; Kulesz, Francis, Barnes, & Fletcher, 2016 ) or adult populations (e.g., Follmer & Sperling, 2018 ); only two known studies to date ( McNamara, Ozuru, & Floyd, 2011 ; Tortorelli, 2020 ) have investigated the interactions between SVR-based reader characteristics and text characteristics in elementary-age students. This is crucial because there are significant developmental shifts in the predictive associations between reader characteristics and reading outcomes in elementary school (e.g., decoding ability is particularly important for young readers; García & Cain, 2014 ).

In terms of studies focusing on elementary-age readers, Tortorelli (2020) discovered, in a sample of 5,133 second-grade students, that oral reading rate was influenced by discourse theory text variables, after controlling for isolated word reading and English learner status. Specifically, word concreteness and referential cohesion were positively associated, whereas deep cohesion was negatively associated, with oral reading rate ( Tortorelli, 2020 ). Similarly, McNamara et al. (2011) studied the impacts of cohesion and genre on reading comprehension using a sample of 65 fourth graders. The researchers found that for students with high knowledge, there was a reverse cohesion effect, such that comprehension was greater for a low-cohesion narrative text relative to a high-cohesion narrative text. Taken together, these findings bolster our hypothesis that text factors influence reading comprehension in developing readers. The findings of two other studies ( Barth et al., 2014 ; Kulesz et al., 2016 ), both conducted with middle schoolers, also support the position of considering both reader characteristics and text features for reading outcomes. However, similar to McNamara et al.’s and Tortorelli’s studies, neither included a measure of listening comprehension, a central person-level component of the SVR. Furthermore, whereas McNamara et al.’s and Kulesz et al.’s outcome measures were reading comprehension, both Barth et al.’s and Tortorelli’s studies offer incomplete pictures in terms of implications for reading comprehension because their outcomes were measures of reading fluency, which captures the speed and accuracy with which a text is read versus a nuanced understanding of the text.

In summary, no studies to date focusing on reading comprehension outcomes in elementary-age readers have included a comprehensive array of key word-level (e.g., word frequency, number of pronouns, word concreteness) and sentence- and passage-level (e.g., referential cohesion, causal cohesion, syntactic complexity) text features identified by discourse processing models, along with the SVR person-level variables, and none has included a measure of text emotionality. Given that decades of research within the discourse-based reading comprehension theoretical context (i.e., construction–integration, RSF) have identified these text-based characteristics as key, there is a need to study them simultaneously, while accounting for person-level characteristics, to comprehensively understand how the multidimensional aspects of text influences reading comprehension (per the CVR i ). Next, we review the theoretical importance of each of these key word-, sentence-, and passage-level text features, followed by a more expansive discussion of the motivation for also capturing the emotional charge of texts.

Text Features

Word-level text features.

Word-level text characteristics that have been studied most often within the construction–integration model and RSF include word frequency, or how commonly a word is seen in print across a set of texts; the number of pronouns in a text; and how concrete or abstract the words are in a text (i.e., the extent to which words are meaningful and evoke a mental image; Graesser, McNamara, & Kulikowich, 2011 ). Word frequency and concreteness, in particular, are aligned with the RSF’s focus on lexical access, in that both are hypothesized to have an impact on ease of processing word-level information (i.e., word recognition, meaning), which in turn frees up cognitive resources that can then be devoted to reading comprehension ( McNamara, Louwerse, McCarthy, & Graesser, 2010 ).

To date, many empirical studies have supported the supposition that word frequency and concreteness facilitate processing, although most studies have explored this topic in undergraduate students or young adults. Studies have shown that high-frequency words are recognized faster and more accurately than low-frequency words ( Forster & Chambers, 1973 ) and that word frequency greatly influences lexical decision making ( Balota, Cortese, Sergent-Marshall, Spieler, & Yap, 2004 ). In addition, word concreteness has been shown to affect how quickly and accurately adults can read words ( Monaghan & Ellis, 2002 ; Strain & Herdman, 1999 ). This finding appears to extend to younger students, as studies have shown that words’ imageability, familiarity, and structural complexity cooperate to increase word recognition in first- and second-grade students ( Hargis & Gickling, 1978 ; Kolker & Terwilliger, 1981 ).

Distinct from word frequency and concreteness, parts of speech (i.e., verbs, adverbs, pronouns, nouns, adjectives), even though they are word-level characteristics, are thought to impact reading comprehension primarily by facilitating cohesiveness and thus situation model building. Indeed, the number of verbs, pronouns, nouns, and adjectives within a text is related to text processing and comprehension outcomes for both elementary-age students and adults (e.g., Barzilay & Lapata, 2008 ; Crossley, Skalicky, Dascalu, McNamara, & Kyle, 2017 ; Ecalle, Bouchafa, Potocki, & Magnan, 2013 ). Relations between parts of speech and text comprehension likely exist (e.g., Black, Turner, & Bower, 1979 ; Ferstl & von Cramon, 2001 ; Graesser et al., 2004 ; Nunan, 1993 ) because the presence of pronouns, verbs, adverbs, nouns, and adjectives potentially increases the overall cohesiveness of a text. Nouns, pronouns, and adjectives often promote coreference and referential cohesion ( Graesser et al., 2004 ; Nunan, 1993 ), whereas adverbs and verbs are associated with locational and temporal cohesion ( Black et al., 1979 ; Duran, McCarthy, Graesser, & McNamara, 2007 ). Consequently, texts that include more of these characteristics may be easier to comprehend because they are more cohesive ( Ozuru, Dempsey, & McNamara, 2009 ) and, therefore, more coherent (e.g., McNamara, Kintsch, Songer, & Kintsch, 1996 ).

It is important to note that the associations between parts of speech and text comprehension have also been revealed in studies examining the effect of text genre on reading comprehension. For instance, narrative texts, which include a greater number of personal pronouns and verbs (e.g., Graesser et al., 2011 ; van Hell, Verhoeven, Tak, & van Oosterhout, 2005 ), are easier to comprehend than expository texts ( Best, Floyd, & McNamara, 2008 ) for both young readers and adults. Cohesiveness in expository texts, in contrast, is supported by the inclusion of nouns and adjectives in addition to verbs (Britton & Black, as cited in Lee & Evens, 1996 ). These elements “function to convey densely packed information…and…reflect a precise and often specialized choice of words” ( Biber & Conrad, 2001 , p. 186) rather than increase coreference (e.g., as with pronouns in narrative text; Déchaine & Wiltschko, 2002 ), which may explain why expository texts are more challenging. Parts of speech may further influence reading comprehension beyond increasing cohesiveness. For instance, noun phrases play a role in supporting inference making ( Graesser, Singer, & Trabasso, 1994 ), which is important for reading comprehension ( Elbro & Buch-Iversen, 2013 ). Further, pronouns affect perspective taking during text reading ( Brunyé, Ditman, Mahoney, Augustyn, & Taylor, 2009 ; Ditman, Brunyé, Mahoney, & Taylor, 2010 ); this potentially facilitates embodied cognition-related processes ( Tversky & Hard, 2009 ), which may support comprehension ( Glenberg & Kaschak, 2002 ). These patterns emphasize the need to consider parts of speech as potentially impactful word-level characteristics.

Sentence- and Passage-Level Text Features

The construction–integration model and the RSF both indicate that the identification of propositions, or meaningful phrases within a text, follows word-level processing ( Kintsch, 1988 ; Perfetti, 1999 ). Propositions are identified and interpreted using the reader’s syntactic knowledge and the context of the story line itself. Propositions are the building blocks of the reader’s textbase, or the reader’s basic understanding of what the text says.

Once the reader has developed a coherent understanding of the textbase, the reader then develops a broad mental model of the text. This process involves understanding the passage of time and space, character traits, causality, and intentionality ( Zwaan, Magliano, & Graesser, 1995 ). Both the RSF and the construction–integration model emphasize the importance of inference making and background knowledge to inform the reader’s development of this understanding (i.e., situation model). The construction–integration model asserts that comprehension at the textbase-level is affected by the presence of several text-based variables, including coreferents and causality. Coreferents in the text are occasions when one person or idea is referred to across several propositions or sentences (e.g., “John” and “the lawyer” may both refer to the protagonist’s son). Gernsbacher (1997a) summarized early research exploring the relation between referential cohesion and reading comprehension and asserted that the former improves the latter by facilitating the development of the situation model. Empirical studies have supported this claim (e.g., Gernsbacher & Robertson, 1996 ; Haviland & Clark, 1974 ; Lesgold, 1972 ). For instance, individuals understand a second sentence more quickly if it includes a specific reference to a word found in the first sentence. Similarly, they remember a story line better if stimuli include repeated referents to the same people, places, or ideas. In contrast, comprehension and memory are impeded without specific references to previously used words and content ( Gernsbacher & Robertson, 1996 ; Haviland & Clark, 1974 ; Lesgold, 1972 ).

In addition to references that refer to the same idea within sentences, the presence of causal and logical connections in the text also helps the reader understand the relation between two different events or between actions and emotions. These connections, referred to as deep cohesion ( Graesser et al., 2004 ), are hypothesized as helping the reader create a coherent understanding of the text ( Gernsbacher, 1997a ; Keenan, Baillet, & Brown, 1984 ; Maury & Teisserenc, 2005 ). Again, some of the findings from research conducted with adults extends to young readers. For example, McNamara et al. (2011) reported that increased cohesion of a text improves comprehension for most fourth-grade students, although findings were mixed depending on the type of comprehension measure (i.e., recall vs. multiple-choice).

In sum, there are strong theoretical reasons (and empirical support) for why each of the aforementioned word-, sentence-, and passage-level text features impact text processing. Only one known study to date ( Tortorelli, 2020 ) has considered multiple text-based (word-, sentence-, and passage-level) features while also accounting for person-level contributions in elementary-age readers; however, this study did not include both SVR components, a notable limitation. Furthermore, of the text-based features that have been implicated in reading comprehension, the emotional charge of a text has received comparatively less attention, particularly in elementary-age readers, despite its potential theoretical and practical importance. Next, we elaborate on the importance of examining this additional, potentially unique contribution of text emotion to reading comprehension.

Reading Comprehension Theories: Role of Emotion

Whereas the CVR i incorporates both person-level and text features, it, as well as other discourse processing models, does not directly consider the role that emotion may play in text comprehension. Although a few discourse processing frameworks in the construction–integration tradition acknowledge the potential influence of emotion on reading comprehension (e.g., structure building and event indexing models; Gernsbacher, 1997b ; Kneepkens & Zwaan, 1995 ), emotion predominantly has been discussed as a reader characteristic rather than as a feature of text. This is significant because, as discussed previously, both a reader’s emotional reaction to a text and the emotional content of the text may influence the reading process ( Gernsbacher et al., 1992 ; Gillioz, Gygax, & Tapiero, 2012 ; Kneepkens & Zwaan, 1995 ). For example, Kneepkens and Zwaan (1995) hypothesized that the emotional charge of a story may intensify a reader’s engagement with the text, which in turn may help facilitate the development of the reader’s situation model, and Graesser et al. (1994) posited that “the emotional reactions of characters and the subordinate goals play a prominent role in global plot configurations of stories and are therefore needed for the establishment of global coherence” (p. 382).

In addition to strong theoretical support for the role of emotion in reading comprehension, empirical studies also have supported the hypothesis that emotional charge of a text is involved in situation model formation. For example, Gernsbacher et al. (1992) reported that individuals read two sentences with a similar emotional tone more quickly than two sentences that do not share the same emotional tone; the authors argued that this increase in reading speed suggests that the matching tone helped readers integrate new information into their mental model of the story more efficiently. More recent research has supported these findings; for example, Gygax, Oakhill, and Garnham (2003) reported that readers do not differentiate between similar emotions while reading text; for example, depressed , miserable , and useless are all seen as congruous. Moreover, Gillioz et al. (2012) found that readers are more sensitive to behavioral situations with characters in stories that elicit an emotion rather than an explicitly stated emotion. Other research has suggested that higher levels of emotion content in a text is associated with increased identification with the protagonist’s point of view ( de Vega, Díaz, & León, 1997 ).

Together, findings suggest that behavioral information about characters may influence a reader’s emotional processing and situational model development, thus implying that a narrative framework (superstructure) may especially lend itself to a role for emotion in reading comprehension. However, whether lexical ratings of the emotional charge of a text impacts comprehension has been largely unexplored when the emotional charge is not embedded within the superstructure of a narrative text (e.g., in a text that does not have characters). Research in other areas (e.g., Hsu, Conrad, & Jacobs, 2014 ; Hsu, Jacobs, Citron, & Conrad, 2015 ) has examined how people process emotionally charged words, as determined by objective lexical ratings (in isolation/individually), thus lending support to the idea that emotional charge captured at the lexical level may influence reading comprehension.

When identified at the lexical level, emotionality is typically described using three primary dimensions: arousal, valence, and dominance. In particular, arousal, which indicates how calming or exciting a word is (e.g., Bradley & Lang, 1999 ), has received the most attention. Valence, which indicates how pleasant (positive) or unpleasant (negative) a word is ( Bradley & Lang, 1999 ), has been explored in some adult studies of reading (e.g., Hsu et al., 2014 ), where presenting fear-inducing or graphic content is appropriate. Prior research has generally indicated that the meanings of positive and exciting words are processed more quickly than the meanings of neutral or calming words ( Kissler & Herbert, 2013 ; Recio, Conrad, Hansen, & Jacobs, 2014 ), although the findings related to the excitability of words are mixed ( Kuperman, Estes, Brysbaert, & Warriner, 2014 ). Although few studies have examined word-level ratings of affect in text, those examining both lexical and self-reported evoked emotions have suggested relative congruency between the two (e.g., r = ~.60; Hsu et al., 2014 , 2015 ). Thus, findings indicate that lexical ratings of emotional charge may provide an index of the text’s emotional charge that is generally congruent with readers’ self-reported evoked emotions. Given that emotional charge has been theorized as a potentially important vehicle for assisting in situation model building, findings that lexical ratings of emotional charge of text indeed impact reading comprehension, especially across all text types (e.g., narrative, expository), indicate that lexical ratings of emotional charge are important to consider.

In summary, to date, the emotional charge of text has primarily been operationalized using the researcher’s subjective judgment of the level of emotion of a passage or sentence or readers’ assessments of their own evoked emotions while reading text ( Gernsbacher et al., 1992 ; Gillioz et al., 2012 ; Gygax et al., 2003 ), instead of through a more objective and quanitified measurement of emotional charge, such as lexical ratings of arousal ( Warriner, Kuperman, & Brysbaert, 2013 ). Further, these ratings have typically occurred within the context of a narrative text superstructure (i.e., linked to characters in a story). Therefore, it is an open question as to whether such word-level metrics of emotion could explain any variance in reading comprehension, even though research external to the discourse processing literature has provided the foundation for the hypothesis that emotional charge of texts measured at the word level may be a valuable, additional text-based metric in empirical studies of reading comprehension.

Research Questions

The present study extends the current literature by examining the role of text features, including lexical measures of emotion, in elementary-age readers while also considering SVR variables. Importantly, we focused on reading comprehension as an outcome, as opposed to a measure of reading fluency ( Barth et al., 2014 ; Tortorelli, 2020 ), which is key for understanding how text features impact a reader’s ability to comprehend and learn from text. Additionally, exploring the influence of text emotion on reading comprehension, in addition to person-level and text-based characteristics, is significant because although previous studies have explored the effects of text characteristics ( McNamara et al., 2011 ; Tortorelli, 2020 ) or person-level and text-based characteristics ( Barth et al., 2014 ; Francis et al., 2018 ) on reading, these studies did not include metrics of emotion, especially ones identified by objective lexical ratings of arousal ( Warriner et al., 2013 ), which may lend themselves to more standardized ways of capturing emotional charge of texts. Given prior research suggesting that emotionality may potentially enhance readers’ engagement (and as follows, situation model building) during text processing, such knowledge could provide essential information for text construction and characterization.

Two research questions guided our study:

  • Do the text-based features of word frequency, number of pronouns, word concreteness, syntactic simplicity, referential cohesion, and deep cohesion account for additional variance in reading comprehension after controlling for SVR-based person-level characteristics?
  • Do lexical ratings of arousal (emotion) predict unique variance in reading comprehension performance over and above person-level and text-based characteristics?

Consistent with findings from the substantial literature on the SVR, we expected that students’ decoding and language comprehension would be significantly associated with reading comprehension; however, of central interest was the hypothesis that after accounting for SVR components, text characteristcs would explain unique variance in reading comprehension. Further, based on prior literature linking lexical ratings of arousal to increased efficiency of processing words, as well as readers’ subjective assessment of their evoked emotions, we hypothesized that lexical ratings of arousal would explain additional unique variance in reading comprehension.

Participants

Participants were obtained from three larger studies conducted in the U.S. Southeast. All studies were conducted in accordance with the university’s Institutional Review Board. Participants were recruited through advertisements posted in clinics, schools, and doctors’ offices. Parents were first asked for their written consent for their child to participate in the study. Then, verbal assent was obtained from the children prior to study participation. The final sample was 158 native English-speaking children between 8 years 8 months and 11 years 2 months of age ( M = 10.03 years, SD = 7 months) with a wide range of reading ability (see Table 1 ). To be included in our study, participants had to have native English speaker status; however, 17% of our sample reported that a language other than English was spoken at home. Slightly more than half of the sample was female (55.06%). Participants were White (72.78%), Black (15.82%), Asian (1.26%), Native American (0.63%), or multiracial (6.32%); 3.17% did not report. Consistent with previous studies, parental level of educational attainment was rated on a 7-point scale to approximate socioeconomic status (e.g., Cirino et al., 2002 ). The average educational attainment score for our sample between two parents was 5.8, between partial college (5) and graduating from a college or university (6). The minimum score in our sample was 2.5, between junior high school (2) and partial high school (3). The maximum score of the scale and from our sample was 7, representing graduate school, with 28% of our sample scoring 7 for both parents.

Descriptive Statistics for the Analysis Sample ( N = 158)

Note . GMRT–IV = Gates–MacGinitie Reading Test, fourth edition; WASI = Wechsler Abbreviated Scale of Intelligence, first and second editions; WJ = Woodcock–Johnson Tests of Achievement, third and fourth editions.

During recruitment, participants from all three studies were not eligible for participation if they had a previous diagnosis of intellectual disability; a known, uncorrectable visual impairment; treatment of any psychiatric disorder with psychotropic medications (other than attention deficit hyperactivity disorder); a history of known neurologic disorder (e.g., epilepsy, spina bifida, cerebral palsy, traumatic brain injury); a documented hearing impairment greater than or equal to a 25 dB loss in either ear; an IQ score below 80 or a score below 70 on either performance or verbal scales using the Wechsler Abbreviated Scale of Intelligence, fourth edition ( Wechsler, 1999 ), determined after enrollment; or a history or presence of a pervasive developmental disorder. Participants were also not included if, during testing, parental responses to the Diagnostic Interview for Children and Adolescents–IV ( Reich, Welner, & Herjanic, 1997 ) indicated the presence of any severe psychiatric diagnoses, including major depression, bipolar disorders, and conduct disorder. These exclusion criteria were implemented to limit the likelihood that comorbidities would affect reading performance. However, we included children who met criteria for attention deficit hyperactivity disorder, oppositional defiant disorder, adjustment disorder, or mild depression.

Trained graduate-level research assistants and staff administered various experimental and standardized assessments over two days. Total testing time was approximately six hours (three hours per day); this fluctuated slightly because of individual student pacing and family availability. Tests were administered in a one-on-one setting in a private testing room in a university lab building. Testers were required to reach a test administration fidelity of ≥90%, and all assessments were double-scored. In rare cases when the second coder found an error with scoring (e.g., a miscount on the number of multiple-choice questions answered correctly), the second coder confirmed the score with the first coder before making the change, and the lab manager was consulted if consensus could not be met during the discrepancy discussion.

Analytic Sample Selection

Across all three samples, the subset of participants included in the analyses completed level 4 of the Comprehension subtest of the Gates–MacGinitie Reading Test, fourth edition (GMRT–IV; MacGinitie, MacGinitie, Maria, & Dreyer, 2000 ) as part of a larger battery of cognitive and reading assessments. Students completed only level 4 of the GMRT–IV because it corresponded to their age. To assess general reading achievement, we administered the Basic Reading subtests (Letter-Word Identification and Word Attack) from the Woodcock–Johnson Tests of Achievement, third and fourth editions (WJ–III and WJ–IV; Schrank, Mather, & McGrew, 2014 ; Woodcock, McGrew, & Mather 2001 ) to 111 and 47 participants, respectively; no difference in scores was found due to assessment version. Similarly, no differences were found across the Wechsler Abbreviated Scale of Intelligence, first and second editions ( Wechsler, 1999 , 2011 ), which served as our measure of IQ during screening for the three samples.

Measures that captured reader and text-based characteristics are described here in greater detail.

Reader Characteristics

Reading comprehension.

For the Comprehension subtest from the GMRT–IV ( MacGinitie et al., 2000 ), participants were given 35 minutes to read short passages silently and answer both literal and inferential multiple-choice questions about the passage. The passages vary in style and topic and span genres from more expository to more narrative. Only level 4 passages were included in the analyses (11 passages total). Internal reliability for the comprehension subtest is high (Cronbach’s α = .910–.948; MacGinitie, MacGinitie, Maria, & Dreyer, 2008 ). Because we were interested in comparing comprehension performance across questions and passages, we scored the GMRT–IV at the item level. Responses for each multiple-choice question were dummy-coded (1 = correct , 0 = incorrect ) and included in the regression models as the dependent variable.

Decoding Ability

The Word Attack subtest from the WJ–III and WJ–IV ( Schrank et al., 2014 ; Woodcock et al., 2001 ) required participants to read a list of nonsense words that increased in difficulty. We chose to use this measure of nonsense word reading in our analysis as opposed to real word reading to better capture the essence of decoding. Had we used basic reading scores, students learning histories and memories of real words could have distorted their true graphophonetic skills. Reliability for this assessment is high (test–retest reliability for 10-year-olds is .92; McGrew, LaForte, & Schrank, 2014 ). We used standard scores in the analysis.

Language Comprehension

For the KNOW-IT Test ( Barnes & Dennis, 1996 ; Barnes, Dennis, & Haefele-Kalvaitis, 1996 ), participants were given 20 facts about an imaginary planet called Gan to build their knowledge base. They were then instructed to listen to a story and answer literal and inferential questions about it. Each question was worth up to 2 points based on the detailedness of responses. If a participant gave a response that initially earned 1 point, the tester provided up to two queries per item to elicit a more descriptive answer. We included a total language comprehension score in the analysis, which was a sum of raw scores from the literal, coherence inferential, and elaborative inferential subtests. As per Spencer, Richmond, and Cutting (2020) , reliability for total language comprehension was adequate (α = .77). The KNOW-IT Test also has evidence of convergent validity; correlations between literal questions and expressive and receptive vocabulary range from .34 to .41 ( Spencer et al., 2020 ), and correlations between cohesive and elaborative questions and broad vocabulary knowledge (i.e., receptive and expressive vocabulary and synonyms) range from .37 to .41 ( Spencer et al., 2019 ). Of note, we additionally specified models that included only the literal questions, as it could be argued that they capture a purer version of language comprehension because they do not involve inference making ( Wilson & Bishop, 2019 ). Results from that analysis revealed identical patterns of statistical significance as those analyses including the total comprehension task, so we decided to use the total task because of its higher reliability (α = .77 for all comprehension items vs. .50 for literal items only).

Text Characteristics

Text complexity.

We used Coh-Metrix 3.0 ( McNamara, Graesser, McCarthy, & Cai, 2014 ) to analyze each of the GMRT–IV passages. This tool evaluates texts based on dozens of variables theorized to be associated with text complexity. Coh-Metrix developers conducted a principal component analysis to identify the variables that account for the majority of variance in text difficulty, and we included the five indices identified the by developers as accounting for the largest amount of variance in text complexity: narrativity, syntactic simplicity, word concreteness, referential cohesion, and deep (causal) cohesion (see Table 2 ). Although we could have selected other indices for this analysis, namely, average number of syllables, we chose to focus on indices concerning discourse theory and text easability, especially because we aimed to include decoding ability as a predictor in our models.

Coh-Metrix Text Easability Variables

Of note, we used the narrativity index from Coh-Metrix, which is a component derived from principal components analysis that “captures the extent to which the text conveys a story, a procedure, or a sequence of episodes of actions and events with animate beings” ( Graesser et al., 2011 , p. 228) and includes multiple metrics, such as word frequency and familiarity, parts of speech, number of pronouns, and passive sentence constructions ( Graesser et al., 2011 ); however, we acknowledge that this index does not capture superstructure features of genre. This is important because word-level characteristics do not fully account for genre-related differences in comprehension (e.g., Eason, Goldberg, Young, Geist, & Cutting, 2012 ), suggesting that additional variation in reading comprehension may be explained by genre-specific superstructure (e.g., the presence of a central theme, setting, characters, and anticipated story structure within narrative text; Baumann & Bergeron, 1993 ; Graesser et al., 1994 ; Meyer & Rice, 1984 ). Nevertheless, to stay consistent with the Coh-Metrix labels, we use the term narrativity while acknowledging that this index correlates, but may fully capture, the superstructure elements of genre (a point further addressed in the Discussion section). For the analysis, we used Coh-Metrix’s z -score for each passage.

Text Arousal

We used arousal ratings (i.e., how calming or exciting a word is) from the database developed by Warriner et al. (2013) . The database contains 13,915 high-frequency words that were scored by 1,827 respondents via the Amazon Mechanical Turk crowdsourcing website. Participants rated words on three dimensions: arousal, valence, and dominance. For arousal, 339,323 observations were collected, and the scale ranged from 1 ( low arousal ) to 9 ( high arousal ). In the database, words with the lowest arousal include dull , calm , and grain ; words with the highest arousal scores include scare , gun , and snake . Because the database primarily includes singular nouns and present-tense verbs, we changed plural nouns in the GMRT–IV passages to their singular form and past-tense verbs to their present-tense form. Then, we extracted ratings from the database to obtain an arousal score for each word in the passages. Last, we calculated a mean arousal score for each passage (see Table 3 ). Across the passages, 82–96% of the content words (nouns, main verbs, adjectives, and adverbs) had arousal ratings. For example, in the passage about shoes, the word shoe had a low-arousal score of 2.4 and the word sigh a score of 2.91, whereas in the passage about dolls, the word funny had a high-arousal score of 5.38 and the word love a high-arousal score of 5.36.

Passage Means Across Text Features

Note . GMRT–IV = Gates–MacGinitie Reading Test, fourth edition. Text features from Coh-Metrix (all except arousal) are reported as z -scores. Arousal is reported on a rating scale ranging from 1 ( low arousal ) to 9 ( high arousal ).

We ran all statistical analyses in R (version 3.6.3; R Core Team, 2020 ) using multilevel modeling because our data were nested (i.e., reading comprehension scores nested within participants). We only included participants with complete data ( N = 158) in the analyses. Testing revealed that the data were missing completely at random (Little’s test = 4.72, df = 6, p = .581), so we used listwise deletion to remove the 326 participants with missing data ( Little, 1988 ). Visual inspection using naniar in the R package ( Tierney & Cook, 2020 ) also revealed no grouping or unexpected distributions of the missing data. To maintain equal weighting of the independent variables within our regression models, we transformed all of the predictor variables into z -scores prior to analyses.

Using forward-fitting modeling procedures, we ran five logistic mixed-effect models using the lme4 package (version 1.1–7; Bates, Maechler, Bolker, & Walker, 2015 ). All models had the same random intercept (participant). The models differed from one another only in terms of their fixed effects (i.e., person characteristics, text predictors) and were therefore specified for each individual analysis ( Baayen, Davidson, & Bates, 2008 ; Bates, Maechler, & Dai, 2008 ). Passages were not included as a random or a fixed effect because there was no theoretical reason to assume that differences between passages existed, nor did the inclusion of passages as a random or fixed effect improve model fit. Deviance testing (i.e., χ 2 ) compared each model using model fit statistics (i.e., Akaike information criterion [AIC], Bayesian information criterion [BIC]) to determine model improvement.

Estimated marginal and conditional R -squared values determined the proportion of dependent variable variance accounted for by the fixed and fixed plus random effects, respectively. R -squared values are typically used in multiple linear regression to describe incremental model improvements. Advances have been made in R -squared measurement calculation for multilevel models (e.g., Rights & Sterba, 2019 ). The results of these advances indicate that estimated marginal and conditional R -squared values serve as excellent estimates of dependent variable variance ( Nakagawa & Schielzeth, 2012 ; Sotirchos, Fitzgerald, & Crainiceanu, 2019 ) that are becoming widely recommended ( Rights & Sterba, 2019 ). As recommended, we also report multiple measures of model fit (i.e., AIC, BIC) and used deviance testing when comparing our models. Additionally, we calculated proportional effect sizes (partial R -squared values) for the individual predictors in the mixed-effect models ( Edwards, Muller, Wolfinger, Qaqish, & Schabenberger, 2008 ) using the r2glmm package (version 0.1.0; Jaeger, 2016 ) in R. Together, these steps allowed us to evaluate the performance of our R -squared values by considering whether they adhered to the recommended properties of R -squared values ( Nakagawa & Schielzeth, 2012 ; Orelien & Edwards, 2008 ).

Preliminary Analysis

GMRT–IV passage means for each of the Coh-Metrix variables are shown in Table 3 . Correlations among all of the predictor variables are reported in Tables 4 and ​ and5. 5 . Syntactic simplicity was moderately correlated with word concreteness ( r = −.62) and referential cohesion ( r = −.61). Both the variance inflation factor and tolerance values suggested increases in multicollinearity when syntactic simplicity was included in the model. Following Tortorelli (2020) , we removed syntactic simplicity as a predictor to alleviate inflation of the standard errors of the coefficients and issues regarding coefficient directionality. Notably, Tortorelli also removed syntactic simplicity from her final model because of its high correlation with other predictor variables. After the removal of syntactic simplicity, we found that the independent variables’ variance inflation factors were less than 10 ( Myers, 1990 ) and that the tolerance statistics were greater than 0.1 ( Menard, 1995 ) across all subsequent models. We then examined the data for outliers and influential data points using the procedures for generalized mixed-effect models (see Nieuwenhuis, te Grotenhuis, & Pelzer, 2012 ); we found no influential cases in the data (see van der Meer, te Grotenhuis, & Pelzer, 2010 ).

Correlations Between Person-Level Variables

Correlations Between Text Features

Logistic Mixed-Effects Models

First, we used three logistic mixed-effects models to confirm that the SVR-based person-level characteristics predicted reading comprehension. We included age (as a control variable), decoding ability, and language comprehension in each model using a model buildup procedure (i.e., one variable was added at a time in the aforementioned order; models A–C). In the model that included only reader characteristics (model C; see Table 6 ), each variable was significantly associated with reading comprehension, suggesting that students with higher decoding and language comprehension scores performed better on reading comprehension. The marginal R 2 for model C was .1926, indicating that fixed effects (i.e., age, decoding, and language comprehension combined) accounted for 19.26% of the variance in reading comprehension; the conditional R 2 for model C was .3214, indicating that the fixed and random effects together accounted for 32.14% of the variance. Following this, we specified two additional models that included text-based characteristics.

Model Parameter Estimates and Model Fit Statistics

Note . AIC = Akaike information criterion; BIC = Bayesian information criterion. Theoretical R -squared values are reported. Intercept is the random effect of participant, and model comparison with a greater than sign (>) is for deviance-testing purposes. Conditional R -squared variance is accounted for by fixed and random effects of subject. Marginal R -squared variance is accounted for by only fixed effects.

For model D, we added the four remaining text features as fixed effects: narrativity, word concreteness, referential cohesion, and deep cohesion. Because there was no theoretical basis for order, the four predictors were added simultaneously to the model. Deviance testing revealed that inclusion of the text features improved model fit (model D vs. model C), χ 2 (4) = 205.8, p < .001, meaning that a model that included both person-level and text-based variables provided a significantly better fit to the data than the model with only person-level variables. In model D, narrativity ( B = 0.17, standard error [ SE ] = 0.04, p < .001, partial R 2 [p R 2 ] = .002, 95% confidence interval [CI] [0.001, 0.005]), word concreteness ( B = −0.17, SE = 0.03, p < .001, p R 2 = .003, 95% CI [0.001, 0.006]), and deep cohesion ( B = 0.32, SE = 0.03, p < .001, p R 2 = .015, 95% CI [0.010, 0.021]) were all significantly related to reading comprehension. Referential cohesion was not a statistically significant predictor ( p = .461). The marginal R 2 for model D was .2284, indicating that fixed effects accounted for 22.84% of the variance in reading comprehension; the conditional R 2 for model D was .3590, indicating that the fixed and random effects together accounted for 35.9% of the variance.

In the final model (model E), we included arousal as an additional predictor (see Table 6 ). Deviance testing revealed that inclusion of arousal improved model fit (model E vs. model D), χ 2 (1) = 54.5, p < .001. In this model, arousal was a statistically significant, positive predictor of reading comprehension ( B = 0.29, SE = 0.03, p < .001, p R 2 = .008, 95% CI [0.004, 0.013]). The marginal R 2 for model E was .2392, indicating that fixed effects accounted for 23.92% of the variance in reading comprehension; the conditional R 2 for model E was .3699, indicating that the fixed and random effects together accounted for 36.99% of the variance.

It has been well established that decoding ability and language comprehension are associated with reading comprehension performance, but the influence of text features on reading comprehension outcomes has, in comparison, received substantially less attention. In fact, only two previous studies ( McNamara et al., 2011 ; Tortorelli, 2020 ) have explored the influence of text features on reading in elementary-age students while considering person-level variables. This gap in the reading comprehension literature is surprising given that most could assume the content and structure of the actual text students read matters in addition to the individual reading skills they may have acquired. In the current study, we examined variance in reading comprehension accounted for by text features over and above individual differences in reading comprehension known to be accounted for by SVR component skills. Because previous research has suggested that increased emotionality of text may facilitate comprehension of text (e.g., Gernsbacher, 1997b ; Kneepkens & Zwaan, 1995 ), we were particularly interested in the influence of text-based arousal on reading comprehension. In this section, we discuss our findings related to reader characteristics, followed by theoretical and practical implications of findings related to text-based characteristics, including the text-based metric of arousal. This finding is of particular interest because text-based measures capturing the emotional charge of a text have received little attention in reading comprehension research, particularly in the developmental literature. Although practitioners and others might intuit that texts with exciting words are easier to comprehend for younger readers, this supposition has not been empirically evaluated to date.

In the current study, decoding and language comprehension together accounted for slightly over 32% of the variance in reading comprehension, after controlling for age, further highlighting the importance of these skills during reading comprehension. Although we expected that these components would account for variance in reading comprehension performance, the magnitude of variance explained in the present study stands in contrast to the findings from other investigations, in which reported estimates range from 38% to 100% (e.g., Aaron, Joshi, & Williams, 1999 ; Foorman et al., 2018 ; Foorman, Koon, Petscher, Mitchell, & Truckenmiller, 2015 ; Hjetland et al., 2019 ; Hoover & Gough, 1990 ; Kershaw & Schatschneider, 2012 ; Lervåg, Hulme, & Melby‐Lervåg, 2018 ; Lonigan, Burgess, & Schatschneider, 2018 ; Tilstra, McMaster, van den Broek, Kendeou, & Rapp, 2009 ; Tiu, Thompson, & Lewis, 2003 ).

Differences in the proportion of variance accounted for in the current study as compared with other investigations is most likely due to the inclusion of observed rather than latent variables in our models. For instance, Tilstra et al. (2009) and Tiu et al. (2003) included observed variables and reported that these components accounted for 38–61% (fourth, seventh, and ninth graders) and approximately 62% (10-year-olds) of the variance in reading comprehension, respectively; whereas studies that included latent variables (e.g., Foorman et al., 2015 , 2018 ; Lervåg et al., 2018 ; Lonigan et al., 2018 ) reported much higher values (72–100% in the elementary and middle school grades) and indicated a statistically significant overlap in the predictive contributions of decoding and language comprehension in elementary-age readers (e.g., Bailey, Duncan, Cunha, Foorman, & Yeager, 2020 ; Foorman et al., 2015 , 2018 ; Lonigan et al., 2018 ). Results of Quinn and Wagner’s (2018) meta-analysis are in line with these findings as well; the authors included several studies that applied latent variable approaches and reported that approximately 60% of the variance in reading comperehension was accounted for by the two SVR components.

Indeed, several studies have argued that measurement error is at least partially responsible for the discrepancy in variance accounted for across studies (for a review, see Ripoll Salceda, Aguado Alonso, & Castilla-Earls, 2014 ). For example, Hjetland et al. (2019) found that decoding and language comprehension accounted for 99% of the variance in reading comprehension in 7–9-year-olds when language was included as a multi-indicator latent variable but that the proportion of variance accounted for dropped to 46% when vocabulary was included as a single observed variable in place of the latent language variable. In this way, our findings are consistent with those of other investigations showing that language comprehension and decoding ability may not account for all of the variance in reading comprehension in elementary-age readers (e.g., Ripoll Salceda et al., 2014 ), especially when observed rather than latent variables are included.

Both the RSF and the construction–integration model indicate that bottom-up processes, most prominently word identification, can affect reading comprehension. In line with the findings of prior research (e.g., Cutting & Scarborough, 2006 ; García & Cain, 2014 ; Kendeou, Savage, & van den Broek, 2009 ; Kendeou, van den Broek, et al., 2009 ), we found that person-level decoding ability was positively associated with reading comprehension. This was anticipated given the age of our sample and the importance of basic reading skills during the early elementary years. Similarly, in line with the SVR framework, language comprehension was positively related to reading comprehension.

Associations Between Text-Based Features and Reading Comprehension Performance

Prior research has provided evidence that word frequency, number of pronouns, syntactic simplicity, and cohesion are related to oral reading fluency ( Barth et al., 2014 ; Tortorelli, 2020 ). Here, for the first time, we show that some of these text features are related to reading comprehension performance, after controlling for relevant person-level characteristics. Our study’s focus on younger students is particularly important given the relative paucity of research that has examined the influence of text-based characteristics on reading performance in elementary-age students. Consistent with the findings of prior research ( Best et al., 2008 ; Diakidoy, Stylianou, Karefillidou, & Papageorgiou, 2005 ), we found that passages with greater word-level characteristics of narrativity (e.g., higher word frequency and familiarity; greater number of pronouns, adverbs, and verbs; fewer passive constructions) were easier to comprehend than passages that were lower on these indices. These findings support prior research indicating that lexical characteristics such as word frequency influence text readability ( Chen & Meurers, 2018 ) and may be explained by the fact that readers tend to recognize and understand frequently used words more easily than words that are rarely used in text ( Klare, 1968 ).

Similarly, passages with deep cohesion, or more causal connectives throughout, were associated with increased reading comprehension. These results are consistent with early theories positing that causal and temporal cohesion facilitate the development of the situation model ( Kneepkens & Zwaan, 1995 ), and are consistent with the findings of empirical research showing that causal cohesion improves comprehension ( Gernsbacher, 1997a ; Reed & Kershaw-Herrera, 2016 ). However, it remains important to acknowledge that the relation between cohesion and reading comprehension is complex and not always positively related ( McNamara & Kintsch, 1996 ).

In contrast, we found that passages with higher word concreteness were associated with lower reading comprehension scores. Although this may seem counterintuitive, prior research ( Graesser et al., 2011 ) has found a curvilinear relation between word concreteness and grade level of reading text, where word concreteness of language arts and social studies texts is lower for texts designed for grades 1 and 2 and for grades 6 and above but higher for texts designed for grades 3–5. Graesser et al. (2011) hypothesized that this is due to texts for younger students including more simple but abstract words, such as go , do , and make , and texts for older students using more complex but concrete words, such as help and juggle . This same study ( Graesser et al., 2011 ) also found that word concreteness for narrative texts (language arts) was lower than for expository texts (science) in early grades, suggesting that word concreteness may not be the factor most strongly influencing comprehension. The present findings also align with those from more a recent investigation ( Wolfe, Dandignac, & Reyna, 2019 ) showing that texts low in word concreteness (but high in deep cohesion) result in more accurate inferences related to the overall meaning of a passage. Together, these findings suggest that perhaps other aspects of the text, such as the level of narrativity or level of cohesion, might have a stronger influence on the level of comprehension than word concreteness does for younger readers. Given this, further research should explore the relation between word concreteness and reading comprehension and how word concreteness may interact with other text characteristics to facilitate or impede passage clarity.

Finally, the degree to which the text contains referential cohesion did not impact comprehension performance until arousal was added into the model. This suggests two things. First, even though arousal and referential cohesion were weakly correlated (−.19), there may be overlapping variance between the two in predicting reading comprehension, or more likely, there may be an unknown underlying variable that accounts for the relation between both and reading comprehension. Second, the degree to which passages have words that are exciting (i.e., higher arousal) may stimulate attention differently during reading than those that are less exciting. Perhaps heightened arousal leads to increased attention or engagement with text, which in turn results in better reading comprehension. For instance, emotional words tend to recruit attention-related processes more so than neutral words (e.g., Citron, Weekes, & Ferstl, 2013 ), which may be due to emotional words having greater semantic coherence relative to neutral words ( Dillon, Cooper, Grent-’t-Jong, Woldorff, & LaBar, 2006 ). This association between arousal and coherence, in conjunction with the fact that coherence and cohesion are closely related, necessary aspects of text processing ( McNamara et al., 1996 ; McNamara, Louwerse, & Graesser, 2002 ), may explain why passages containing higher arousal words are identified as being easier to understand (e.g., Mills, D’Mello, & Kopp, 2015 ) and subsequently comprehend (e.g., Gernsbacher, 1997a ). It is also possible that because texts with high referential cohesion often repeat words within the passage, high referential cohesion may cause the oral reading fluency rate to increase and, in turn, affect comprehension ( Tortorelli, 2020 ). This remains an area for future exploration, however, as additional research is needed to better understand when and how referential cohesion may affect reading comprehension and the joint contributions of arousal and referential cohesion to processing connected text.

The Influence of Word-Level Arousal

One of our most notable findings was the statistically significant association between arousal and reading comprehension performance. In the current study, higher arousal (at the word level) was positively related to reading comprehension performance. That is, passages with more exciting words were associated with higher reading comprehension scores (which was a relief to us, given we did not want to encourage authors to write boring texts to improve comprehension). This finding supports the previous assertion that it is important to consider emotional processes and cognitive processes when attempting to understand reading comprehension ( Kneepkens & Zwaan, 1995 ).

Understanding why word-level text arousal predicts reading comprehension is important but not yet well understood. Researchers have demonstrated that attention and reading comprehension are correlated skills ( Brock & Knapp, 1996 ). If increased arousal of passages also increases attention, this could, in turn, produce better reading comprehension performance. For example, high-arousal words such as sabotage and bewitch might grab a reader. Or, as Gernsbacher and colleagues (1992) suggested, perhaps emotional responses to high-arousal words draw the reader’s attention to important components of a story and helps the reader construct the meaning of the story by clarifying characters’ motivations and actions. Indeed, studies have shown that lexical ratings of emotional charge are correlated with readers’ subjective ratings of emotions evoked while reading ( r = ~.60; Hsu et al., 2015 ; Werlen, Imhof, Benites, & Bergamin, 2019 ). Hsu and colleagues (2015) also found that brain regions associated with emotion processing and situation model building had greater activation when participants read an emotionally charged text as compared with text with low emotion, suggesting that the inclusion of words with higher emotional charge may enhance coordination between brain regions that are relied on for situation model building. Nonetheless, more research is needed to understand how and why emotion of text affects reading comprehension.

Narrativity

Passages with higher narrativity values (i.e., those containing higher frequency words; more pronouns, verbs, adverbs, and negations; and fewer passive constructions and noun phrase modifiers) were related to higher reading comprehension scores, which aligns with the findings of prior research showing that narrative texts are generally associated with higher reading comprehension scores than expository texts are. Nevertheless, this finding should be interpreted somewhat cautiously because the narrativity index in Coh-Metrix is a distilled metric derived from multiple word- and sentence-level variables, including word frequency, sentence structure, and parts of speech. Therefore, although these findings align with those of prior research showing that narrative texts are comprehended more easily than expository texts ( Best et al., 2008 ; Diakidoy et al., 2005 ), results from this study should be viewed with a nuanced lens.

Specifically, the finding of the Coh-Metrix narrativity index correlating with reading comprehension means that word-level features that tend to occur more commonly in narrative text relate to better reading comprehension performance. Indeed, although Coh-Metrix’s measure of narrativity correlates with whether a text is narrative or expository ( Graesser et al., 2011 ), we caution that this measure may not fully capture the passage-level attributes of genre, such as story structure and overall purpose of text, or superstructure. This is an important distinction to understand, as previous studies have shown that although word-level features such as word frequency vary by genre, these variables are not sufficient to account for the all ways genre affects readability (e.g., Kate et al., 2010 ; Sheehan, Flor, & Napolitano, 2013 ). For example, Sheehan and colleagues (2013) showed that narrative text difficulty may be underestimated when it is analyzed using word-level characteristics, because higher frequency words are sometimes used in unusual ways to convey complex ideas. Additionally, expository text complexity may be overestimated because although lower frequency words are typically used, they are repeated more often, which increases cohesion across the passage. Finally, studies that have equated narrative and expository texts on word-level features, such as word frequency, still found a narrative text advantage ( Best et al., 2008 ). Thus, additional research will need to also consider the superstructural characteristics of genre to comprehensively understand the effect that genre has on comprehension.

Limitations

There are several limitations of the study that should be noted when interpreting our findings. First, we could have been improved our study through the use of a larger sample size of participants and longer passages that are more diverse in genre. Second, as mentioned in the Method section, to retrieve an emotion score for the words in each passage, we had to change plural nouns to singular nouns and past-tense verbs to present-tense verbs, which may have subtly impacted our findings. Third, it is likely that a word-level analysis of emotionality of text does not capture the full emotional power of a text examined as a whole; however, there is precedent for using this method to examine text arousal (e.g., Hsu et al., 2015 ) and could be viewed as advantageous because it may be a more objective method than collecting readers’ perceptions of emotion after reading a passage (cf. Gygax et al., 2003 ). Our method of rating emotion also benefits from a large number of third-party raters with many observations on individual words, whereas passage ratings typically are conducted in a lab with a fewer number of ratings; nevertheless, future studies need to systematically examine how these different measures of emotion influence reading comprehension findings.

Fourth, given our relatively limited sample size, we were not able to explore interactions of reader and passage characteristics. However, such an approach would be fruitful for fully informing reading instruction. For example, Barth et al. (2014) explored interactions between reader and passage characteristics and found that as text difficulty (Lexile level) increased, skilled and older readers slowed their reading rate down as compared with unskilled and younger readers. Researchers examining these interactions with measures of arousal may therefore be able to create more individualized models of reading comprehension to understand for which types of readers arousal increases reading comprehension.

Finally, in our attempt to maintain parsimony, there are certain known predictors of reading comprehension performance that we excluded from our models. Prior research has demonstrated the importance of overall cognition and executive function to reading comprehension performance (e.g., Cutting, Materek, Cole, Levine, & Mahone, 2009 ; Spencer et al., 2020 ). Therefore, subsequent studies should explore the influence of the text-based predictors we included in this study on reading comprehension performance, while considering other person-level cognitive characteristics, such as background knowledge, in addition to decoding and language comprehension.

Future Directions

In our study, we explored only how micro- and macrolevel text features are related to reading comprehension, not how the superstructure of genre influences reading comprehension. As described earlier, the superstructure of genre is an important consideration, especially in terms of how it relates to and interacts with word-level features of text (e.g., word frequency) and other text characteristics. Future studies should consider the superstructure of genre along with relevant reader-level characteristics and text-based features, especially with regard to the emotional charge of text (arousal), in models aimed at predicting reading comprehension.

Similarly, future studies should incorporate emotional ratings of passages in addition to word-level measures, along with latent variables rather than observed variables. With larger samples and more complex statistical modeling procedures, researchers could further examine interactions between these variables to determine how text-based features affect reading comprehension across subgroups of readers. Ultimately, such information could be used to create better reading comprehension assessments that are equated on critical text features, to inform educational approaches and interventions for struggling readers, and more practically, to inspire authors to home in on purposefully including exciting and engaging words in children’s texts.

Acknowledgments

This research was supported by grants P20 HD075443, P50 HD103537, R01 HD044073, R37 HD095519, 3RO1 HD044073-14S1, and R01 HD067254 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development. Sage E. Pickren and Maria Stacy are co-first authors of this article.

Contributor Information

Sage E. Pickren, Vanderbilt University, Nashville, Tennessee, USA.

Maria Stacy, Southern Illinois University, Carbondale, USA.

Stephanie N. Del Tufo, University of Delaware, Newark, USA.

  • Aaron PG, Joshi M, & Williams KA (1999). Not all reading disabilities are alike . Journal of Learning Disabilities , 32 ( 2 ), 120–137. doi: 10.1177/002221949903200203 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Baayen RH, Davidson DJ, & Bates DM (2008). Mixed-effects modeling with crossed random effects for subjects and items . Journal of Memory and Language , 59 ( 4 ), 390–412. doi: 10.1016/j.jml.2007.12.005 [ CrossRef ] [ Google Scholar ]
  • Bailey DH, Duncan GJ, Cunha F, Foorman BR, & Yeager DS (2020). Persistence and fade-out of educational-intervention effects . Psychological Science in the Public Interest , 21 ( 2 ), 55–97. doi: 10.1177/1529100620915848 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Balota DA, Cortese MJ, Sergent-Marshall SD, Spieler DH, & Yap MJ (2004). Visual word recognition of single-syllable words . Journal of Experimental Psychology: General , 133 ( 2 ), 283–316. doi: 10.1037/0096-3445.133.2.283 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Barnes MA, & Dennis M (1996). Reading comprehension deficits arise from diverse sources: Evidence from readers with and without developmental brain pathology. In Cornoldi C & Oakhill J (Eds.), Reading comprehension difficulties: Processes and intervention (pp. 251–278). Mahwah, NJ: Erlbaum. [ Google Scholar ]
  • Barnes MA, Dennis M, & Haefele-Kalvaitis J (1996). The effects of knowledge availability and knowledge accessibility on coherence and elaborative inferencing in children from six to fifteen years of age . Journal of Experimental Child Psychology , 61 ( 3 ), 216–241. doi: 10.1006/jecp.1996.0015 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Barth AE, Tolar TD, Fletcher JM, & Francis D (2014). The effects of student and text characteristics on the oral reading fluency of middle-grade students . Journal of Educational Psychology , 106 ( 1 ), 162–180. doi: 10.1037/a0033826 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Barzilay R, & Lapata M (2008). Modeling local coherence: An entity-based approach . Computational Linguistics , 34 ( 1 ), 1–34. doi: 10.1162/coli.2008.34.1.1 [ CrossRef ] [ Google Scholar ]
  • Bates D, Maechler M, Bolker B, & Walker S (2015). lme4: Linear mixed-effects models using ‘Eigen’ and S4 (R package version 1.1–7) [Computer software] . Retrieved from https://CRAN.R-project.org/package=lme4
  • Bates D, Maechler M, & Dai B (2008). lme4: Linear mixed-effects models using S4 classes (R package version 0.999375–28) [Computer software] . Retrieved from https://CRAN.R-project.org/package=lme4
  • Baumann JF, & Bergeron BS (1993). Story map instruction using children’s literature: Effects on first graders’ comprehension of central narrative elements . Journal of Reading Behavior , 25 ( 4 ), 407–437. doi: 10.1080/10862969309547828 [ CrossRef ] [ Google Scholar ]
  • Best RM, Floyd RG, & McNamara DS (2008). Differential competencies contributing to children’s comprehension of narrative and expository texts . Reading Psychology , 29 ( 2 ), 137–164. 10.1080/02702710801963951 [ CrossRef ] [ Google Scholar ]
  • Biber D, & Conrad S (2001). Register variation: A corpus approach. In Schiffrin D, Tannen D, & Hamilton HE (Eds.), The handbook of discourse analysis (pp. 175–196). Malden, MA: Blackwell. [ Google Scholar ]
  • Black JB, Turner TJ, & Bower GH (1979). Point of view in narrative comprehension, memory, and production . Journal of Verbal Learning and Verbal Behavior , 18 ( 2 ), 187–198. doi: 10.1016/S0022-5371(79)90118-X [ CrossRef ] [ Google Scholar ]
  • Bourg T (1996). The role of emotion, empathy, and text structure in children’s and adults’ narrative text comprehension. In Kreuz RJ & MacNealy MS (Eds.), Empirical approaches to literature and aesthetics (pp. 241–260). Norwood, NJ: Ablex. [ Google Scholar ]
  • Bradley MM, & Lang PJ (1999). Affective norms for English words (ANEW): Instruction manual and affective ratings (Technical Report No. C-1) . Gainesville: Center for Research in Psychophysiology, University of Florida. [ Google Scholar ]
  • Brock SE, & Knapp PK (1996). Reading comprehension abilities of children with attention-deficit/hyperactivity disorder . Journal of Attention Disorders , 1 ( 3 ), 173–185. doi: 10.1177/108705479600100305 [ CrossRef ] [ Google Scholar ]
  • Brunyé TT, Ditman T, Mahoney CR, Augustyn JS, & Taylor HA (2009). When you and I share perspectives: Pronouns modulate perspective taking during narrative comprehension . Psychological Science , 20 ( 1 ), 27–32. doi: 10.1111/j.1467-9280.2008.02249.x [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Chen X, & Meurers D (2018). Word frequency and readability: Predicting the text‐level readability with a lexical‐level attribute . Journal of Research in Reading , 41 ( 3 ), 486–510. doi: 10.1111/1467-9817.12121 [ CrossRef ] [ Google Scholar ]
  • Cirino PT, Chin CE, Sevcik RA, Wolf M, Lovett M, & Morris RD (2002). Measuring socioeconomic status: Reliability and preliminary validity for different approaches . Assessment , 9 ( 2 ), 145–155. doi: 10.1177/10791102009002005 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Citron FM, Weekes BS, & Ferstl EC (2013). Effects of valence and arousal on written word recognition: Time course and ERP correlates . Neuroscience Letters , 533 , 90–95. doi: 10.1016/j.neulet.2012.10.054 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Crossley SA, Skalicky S, Dascalu M, McNamara DS, & Kyle K (2017). Predicting text comprehension, processing, and familiarity in adult readers: New approaches to readability formulas . Discourse Processes , 54 ( 5/6 ), 340–359. 10.1080/0163853X.2017.1296264 [ CrossRef ] [ Google Scholar ]
  • Cutting LE, Materek A, Cole CA, Levine TM, & Mahone EM (2009). Effects of fluency, oral language, and executive function on reading comprehension performance . Annals of Dyslexia , 59 ( 1 ), 34–54. doi: 10.1007/s11881-009-0022-0 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Cutting LE, & Scarborough HS (2006). Prediction of reading comprehension: Relative contributions of word recognition, language proficiency, and other cognitive skills can depend on how comprehension is measured . Scientific Studies of Reading , 10 ( 3 ), 277–299. doi: 10.1207/s1532799xssr1003_5 [ CrossRef ] [ Google Scholar ]
  • Déchaine R-M, & Wiltschko M (2002). Decomposing pronouns . Linguistic Inquiry , 33 ( 3 ), 409–442. 10.1162/002438902760168554 [ CrossRef ] [ Google Scholar ]
  • de Vega M, Díaz JM, & León I (1997). To know or not to know: Comprehending protagonists’ beliefs and their emotional consequences . Discourse Processes , 23 ( 2 ), 169–192. doi: 10.1080/01638537709544988 [ CrossRef ] [ Google Scholar ]
  • Diakidoy I-AN, Stylianou P, Karefillidou C, & Papageorgiou P (2005). The relationship between listening and reading comprehension of different types of text at increasing grade levels . Reading Psychology , 26 ( 1 ), 55–80. doi: 10.1080/02702710590910584 [ CrossRef ] [ Google Scholar ]
  • Dillon DG, Cooper JJ, Grent-’t-Jong T, Woldorff MG, & LaBar KS (2006). Dissociation of event-related potentials indexing arousal and semantic cohesion during emotional word encoding . Brain and Cognition , 62 ( 1 ), 43–57. doi: 10.1016/j.bandc.2006.03.008 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Ditman T, Brunyé TT, Mahoney CR, & Taylor HA (2010). Simulating an enactment effect: Pronouns guide action simulation during narrative comprehension . Cognition , 115 ( 1 ), 172–178. doi: 10.1016/j.cognition.2009.10.014 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Duran ND, McCarthy PM, Graesser AC, & McNamara DS (2007). Using temporal cohesion to predict temporal coherence in narrative and expository texts . Behavior Research Methods , 39 ( 2 ), 212–223. 10.3758/BF03193150 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Eason SH, Goldberg LF, Young KM, Geist MC, & Cutting LE (2012). Reader–text interactions: How differential text and question types influence cognitive skills needed for reading comprehension . Journal of Educational Psychology , 104 ( 3 ), 515–528. doi: 10.1037/a0027182 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Ecalle J, Bouchafa H, Potocki A, & Magnan A (2013). Comprehension of written sentences as a core component of children’s reading comprehension . Journal of Research in Reading , 36 ( 2 ), 117–131. doi: 10.1111/j.1467-9817.2011.01491.x [ CrossRef ] [ Google Scholar ]
  • Edwards LJ, Muller KE, Wolfinger RD, Qaqish BF, & Schabenberger O (2008). An R2 statistic for fixed effects in the linear mixed model . Statistics in Medicine , 27 ( 29 ), 6137–6157. 10.1002/sim.3429 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Elbro C, & Buch-Iversen I (2013). Activation of background knowledge for inference making: Effects on reading comprehension . Scientific Studies of Reading , 17 ( 6 ), 435–452. doi: 10.1080/10888438.2013.774005 [ CrossRef ] [ Google Scholar ]
  • Ferstl EC, & von Cramon DY (2001). The role of coherence and cohesion in text comprehension: An event-related fMRI study . Cognitive Brain Research , 11 ( 3 ), 325–340. doi: 10.1016/S0926-6410(01)00007-6 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Follmer DJ, & Sperling RA (2018). Interactions between reader and text: Contributions of cognitive processes, strategy use, and text cohesion to comprehension of expository science text . Learning and Individual Differences , 67 , 177–187. doi: 10.1016/j.lindif.2018.08.005 [ CrossRef ] [ Google Scholar ]
  • Foorman BR, Koon S, Petscher Y, Mitchell A, & Truckenmiller A (2015). Examining general and specific factors in the dimensionality of oral language and reading in 4th–10th grades . Journal of Educational Psychology , 107 ( 3 ), 884–899. doi: 10.1037/edu0000026 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Foorman BR, Petscher Y, & Herrera S (2018). Unique and common effects of decoding and language factors in predicting reading comprehension in grades 1–10 . Learning and Individual Differences , 63 , 12–23. doi: 10.1016/j.lindif.2018.02.011 [ CrossRef ] [ Google Scholar ]
  • Forster KI, & Chambers SM (1973). Lexical access and naming time . Journal of Verbal Learning and Verbal Behavior , 12 ( 6 ), 627–635. doi: 10.1016/S0022-5371(73)80042-8 [ CrossRef ] [ Google Scholar ]
  • Francis DJ, Kulesz PA, & Benoit JS (2018). Extending the simple view of reading to account for variation within readers and across texts: The complete view of reading (CVRi) . Remedial and Special Education , 39 ( 5 ), 274–288. doi: 10.1177/0741932518772904 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • García JR, & Cain K (2014). Decoding and reading comprehension: A meta-analysis to identify which reader and assessment characteristics influence the strength of the relationship in English . Review of Educational Research , 84 ( 1 ), 74–111. doi: 10.3102/0034654313499616 [ CrossRef ] [ Google Scholar ]
  • Gernsbacher MA (1997a). Coherence cues mapping during comprehension. In Costermans J & Fayol M (Eds.), Processing interclausal relationships: Studies in the production and comprehension of text (pp. 3–21). Mahwah, NJ: Erlbaum. [ Google Scholar ]
  • Gernsbacher MA (1997b). Two decades of structure building . Discourse Processes , 23 ( 3 ), 265–304. doi: 10.1080/01638539709544994 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Gernsbacher MA, Goldsmith HH, & Robertson RR (1992). Do readers mentally represent characters’ emotional states? Cognition and Emotion , 6 ( 2 ), 89–111. doi: 10.1080/02699939208411061 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Gernsbacher MA, & Robertson RRW (1996). The definite article “the” facilitates mapping in story comprehension . Unpublished manuscript.
  • Gillioz C, Gygax P, & Tapiero I (2012). Individual differences and emotional inferences during reading comprehension . Canadian Journal of Experimental Psychology , 66 ( 4 ), 239–250. 10.1037/a0028625 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Glenberg AM, & Kaschak MP (2002). Grounding language in action . Psychonomic Bulletin & Review , 9 , 558–565. doi: 10.3758/BF03196313 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Graesser AC, McNamara DS, & Kulikowich JM (2011). Coh-Metrix: Providing multilevel analyses of text characteristics . Educational Researcher , 40 ( 5 ), 223–234. doi: 10.3102/0013189X11413260 [ CrossRef ] [ Google Scholar ]
  • Graesser AC, McNamara DS, Louwerse MM, & Cai Z (2004). Coh-Metrix: Analysis of text on cohesion and language . Behavior Research Methods, Instruments, & Computers , 36 ( 2 ), 193–202. doi: 10.3758/BF03195564 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Graesser AC, Singer M, & Trabasso T (1994). Constructing inferences during narrative text comprehension . Psychological Review , 101 ( 3 ), 371–395. doi: 10.1037/0033-295X.101.3.371 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Gygax P, Oakhill J, & Garnham A (2003). The representation of characters’ emotional responses: Do readers infer specific emotions? Cognition and Emotion , 17 ( 3 ), 413–428. doi: 10.1080/02699930244000048 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Hargis CH, & Gickling EE (1978). The function of imagery in word recognition development . The Reading Teacher , 31 ( 8 ), 870–874. [ Google Scholar ]
  • Haviland SE, & Clark HH (1974). What’s new? Acquiring new information as a process in comprehension . Journal of Verbal Learning and Verbal Behavior , 13 ( 5 ), 512–521. doi: 10.1016/S0022-5371(74)80003-4 [ CrossRef ] [ Google Scholar ]
  • Hjetland HN, Lervåg A, Lyster S-AH, Hagtvet BE, Hulme C, & Melby-Lervåg M (2019). Pathways to reading comprehension: A longitudinal study from 4 to 9 years of age . Journal of Educational Psychology , 111 ( 5 ), 751–763. doi: 10.1037/edu0000321 [ CrossRef ] [ Google Scholar ]
  • Hoover WA, & Gough PB (1990). The simple view of reading . Reading and Writing , 2 ( 2 ), 127–160. doi: 10.1007/BF00401799 [ CrossRef ] [ Google Scholar ]
  • Hsu C-T, Conrad M, & Jacobs AM (2014). Fiction feelings in Harry Potter: Haemodynamic response in the mid-cingulate cortex correlates with immersive reading experience . NeuroReport , 25 ( 17 ), 1356–1361. doi: 10.1097/WNR.0000000000000272 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Hsu C-T, Jacobs AM, Citron FMM, & Conrad M (2015). The emotion potential of words and passages in reading Harry Potter—an fMRI study . Brain and Language , 142 , 96–114. doi: 10.1016/j.bandl.2015.01.011 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Jaeger B (2016). r2glmm: Computes R squared for mixed (multilevel) models (R package version 0.1.0) [Computer software] . Retrieved from https://CRAN.R-project.org/package=r2glmm
  • Kate RJ, Luo X, Patwardhan S, Franz M, Florian R, Mooney RJ, … Welty C (2010). Learning to predict readability using diverse linguistic features. In COLING 2010: 23rd International Conference on Computational Linguistics: Proceedings of the conference (Vol. 2 , pp. 546–554). Stroudsburg, PA: Association for Computational Linguistics. [ Google Scholar ]
  • Keenan JM, Baillet SD, & Brown P (1984). The effects of causal cohesion on comprehension and memory . Journal of Verbal Learning and Verbal Behavior , 23 ( 2 ), 115–126. doi: 10.1016/S0022-5371(84)90082-3 [ CrossRef ] [ Google Scholar ]
  • Kendeou P, Savage R, & van den Broek P (2009). Revisiting the simple view of reading . British Journal of Educational Psychology , 79 ( 2 ), 353–370. doi: 10.1348/978185408X369020 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Kendeou P, van den Broek P, White MJ, & Lynch JS (2009). Predicting reading comprehension in early elementary school: The independent contributions of oral language and decoding skills . Journal of Educational Psychology , 101 ( 4 ), 765–778. doi: 10.1037/a0015956 [ CrossRef ] [ Google Scholar ]
  • Kershaw S, & Schatschneider C (2012). A latent variable approach to the simple view of reading . Reading and Writing , 25 ( 2 ), 433–464. doi: 10.1007/s11145-010-9278-3 [ CrossRef ] [ Google Scholar ]
  • Kim Y-SG (2017). Why the simple view of reading is not simplistic: Unpacking component skills of reading using a direct and indirect effect model of reading (DIER) . Scientific Studies of Reading , 21 ( 4 ), 310–333. doi: 10.1080/10888438.2017.1291643 [ CrossRef ] [ Google Scholar ]
  • Kintsch W (1988). The role of knowledge in discourse comprehension: A construction-integration model . Psychological Review , 95 ( 2 ), 163–182. doi: 10.1037/0033-295X.95.2.163 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Kissler J, & Herbert C (2013). Emotion, etmnooi, or emitoon? Faster lexical access to emotional than to neutral words during reading . Biological Psychology , 92 ( 3 ), 464–479. doi: 10.1016/j.biopsycho.2012.09.004 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Klare GR (1968). The role of word frequency in readability . Elementary English , 45 ( 1 ), 12–22. [ Google Scholar ]
  • Kneepkens EWEM, & Zwaan RA (1995). Emotions and literary text comprehension . Poetics , 23 ( 1/2 ), 125–138. doi: 10.1016/0304-422X(94)00021-W [ CrossRef ] [ Google Scholar ]
  • Kolker B, & Terwilliger PN (1981). Sight vocabulary learning of first and second graders . Literacy Research and Instruction , 20 ( 4 ), 251–258. [ Google Scholar ]
  • Kulesz PA, Francis DJ, Barnes MA, & Fletcher JM (2016). The influence of properties of the test and their interactions with reader characteristics on reading comprehension: An explanatory item response study . Journal of Educational Psychology , 108 ( 8 ), 1078–1097. doi: 10.1037/edu0000126 [ CrossRef ] [ Google Scholar ]
  • Kuperman V, Estes Z, Brysbaert M, & Warriner AB (2014). Emotion and language: Valence and arousal affect word recognition . Journal of Experimental Psychology: General , 143 ( 3 ), 1065–1081. doi: 10.1037/a0035669 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Lee W, & Evens M (1996). Generating cohesive text using lexical functions. In Wanner L (Ed.), Lexical functions in lexicography and natural language processing (pp. 299–306). Amsterdam, Netherlands: John Benjamins. [ Google Scholar ]
  • Lervåg A, Hulme C, & Melby‐Lervåg M (2018). Unpicking the developmental relationship between oral language skills and reading comprehension: It’s simple, but complex . Child Development , 89 ( 5 ), 1821–1838. doi: 10.1111/cdev.12861 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Lesgold AM (1972). Effects of pronouns on children’s memory for sentences . Pittsburgh, PA: Learning Research and Development Center, University of Pittsburgh. [ Google Scholar ]
  • Little RJA (1988). A test of missing completely at random for multivariate data with missing values . Journal of the American Statistical Association , 83 ( 404 ), 1198–1202. doi: 10.1080/01621459.1988.10478722 [ CrossRef ] [ Google Scholar ]
  • Lonigan CJ, Burgess SR, & Schatschneider C (2018). Examining the simple view of reading with elementary school children: Still simple after all these years . Remedial and Special Education , 39 ( 5 ), 260–273. doi: 10.1177/0741932518764833 [ CrossRef ] [ Google Scholar ]
  • MacGinitie WH, MacGinitie RK, Maria K, & Dreyer LG (2000). Gates-MacGinitie Reading Tests, fourth edition, Forms S and T . Itasca, IL: Riverside. [ Google Scholar ]
  • MacGinitie WH, MacGinitie RK, Maria K, & Dreyer LG (2008). Gates–MacGinitie Reading Tests, fourth edition, bundled technical and supplemental report . Rolling Meadows, IL: Riverside. [ Google Scholar ]
  • Maury P, & Teisserenc A (2005). The role of connectives in science text comprehension and memory . Language and Cognitive Processes , 20 ( 3 ), 489–512. doi: 10.1080/01690960444000151 [ CrossRef ] [ Google Scholar ]
  • McGrew KS, LaForte EM, & Schrank FA (2014). Woodcock-Johnson IV: Technical manual . Rolling Meadows, IL: Riverside. [ Google Scholar ]
  • McNamara DS, Graesser AC, McCarthy PM, & Cai Z (2014). Automated evaluation of text and discourse with Coh-Metrix . Cambridge, UK: Cambridge University Press. [ Google Scholar ]
  • McNamara DS, Kintsch E, Songer NB, & Kintsch W (1996). Are good texts always better? Text coherence, background knowledge, and levels of understanding in learning from text . Cognition and Instruction , 14 ( 1 ), 1–43. doi: 10.1207/s1532690xci1401_1 [ CrossRef ] [ Google Scholar ]
  • McNamara DS, & Kintsch W (1996). Learning from texts: Effects of prior knowledge and text coherence . Discourse Processes , 22 ( 3 ), 247–288. doi: 10.1080/01638539609544975 [ CrossRef ] [ Google Scholar ]
  • McNamara DS, Louwerse MM, & Graesser AC (2002). Coh-Metrix: Automated cohesion and coherence scores to predict text readability and facilitate comprehension [Technical report] . Memphis, TN: Institute for Intelligent Systems, University of Memphis. [ Google Scholar ]
  • McNamara DS, Louwerse MM, McCarthy PM, & Graesser AC (2010). Coh-Metrix: Capturing linguistic features of cohesion . Discourse Processes , 47 ( 4 ), 292–330. doi: 10.1080/01638530902959943 [ CrossRef ] [ Google Scholar ]
  • McNamara DS, Ozuru Y, & Floyd RG (2011). Comprehension challenges in the fourth grade: The roles of text cohesion, text genre, and readers’ prior knowledge . International Electronic Journal of Elementary Education , 4 ( 1 ), 229–257. [ Google Scholar ]
  • Menard S (1995). Applied logistic regression analysis . Thousand Oaks, CA: Sage. [ Google Scholar ]
  • Meyer BJF, & Rice GE (1984). The structure of text. In Pearson PD (Ed.), Handbook of reading research (pp. 319–351). Mahwah, NJ: Erlbaum. [ Google Scholar ]
  • Mills C, D’Mello SK, & Kopp K (2015). The influence of consequence value and text difficulty on affect, attention, and learning while reading instructional texts . Learning and Instruction , 40 , 9–20. doi: 10.1016/j.learninstruc.2015.07.003 [ CrossRef ] [ Google Scholar ]
  • Monaghan J, & Ellis AW (2002). What exactly interacts with spelling–sound consistency in word naming? Journal of Experimental Psychology: Learning, Memory, and Cognition , 28 ( 1 ), 183–206. doi: 10.1037/0278-7393.28.1.183 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Mouw JM, Van Leijenhorst L, Saab N, Danel MS, & van den Broek P (2019). Contributions of emotion understanding to narrative comprehension in children and adults . European Journal of Developmental Psychology , 16 ( 1 ), 66–81. doi: 10.1080/17405629.2017.1334548 [ CrossRef ] [ Google Scholar ]
  • Myers RH (1990). Classical and modern regression with applications (2nd ed.). Belmont, CA: Duxbury. [ Google Scholar ]
  • Nakagawa S, & Schielzeth H (2012). The mean strikes back: Mean–variance relationships and heteroscedasticity . Trends in Ecology & Evolution , 27 ( 9 ), 474–475. doi: 10.1016/j.tree.2012.04.003 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • National Governors Association Center for Best Practices & Council of Chief State School Officers. (2010). Common Core State Standards for English language arts and literacy in history/social studies, science, and technical subjects . Washington, DC: Authors. [ Google Scholar ]
  • Nieuwenhuis R, te Grotenhuis M, & Pelzer B (2012). Influence.ME: Tools for detecting influential data in mixed effects models . The R Journal , 4 ( 2 ), 38–47. doi: 10.32614/RJ-2012-011 [ CrossRef ] [ Google Scholar ]
  • Nunan D (1993). Introducing discourse analysis . London, UK: Penguin. [ Google Scholar ]
  • Orelien JG, & Edwards LJ (2008). Fixed-effect variable selection in linear mixed models using R2 statistics . Computational Statistics & Data Analysis , 52 ( 4 ), 1896–1907. doi: 10.1016/j.csda.2007.06.006 [ CrossRef ] [ Google Scholar ]
  • Ozuru Y, Dempsey K, & McNamara DS (2009). Prior knowledge, reading skill, and text cohesion in the comprehension of science texts . Learning and Instruction , 19 ( 3 ), 228–242. doi: 10.1016/j.learninstruc.2008.04.003 [ CrossRef ] [ Google Scholar ]
  • Perfetti CA (1999). Comprehending written language: A blueprint of the reader. In Hagoort P & Brown CM (Eds.), Neurocognition of language processing (pp. 167–208). Oxford, UK: Oxford University Press. [ Google Scholar ]
  • Quinn JM, & Wagner RK (2018). Using meta-analytic structural equation modeling to study developmental change in relations between language and literacy . Child Development , 89 ( 6 ), 1956–1969. doi: 10.1111/cdev.13049 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • RAND Reading Study Group. (2002). Reading for understanding: Toward an R&D program in reading comprehension . Santa Monica, CA: RAND. [ Google Scholar ]
  • R Core Team. (2020). R: A language and environment for statistical computing [Computer software] . Vienna, Austria: R Foundation for Statistical Computing. Retrieved from https://www.R-project.org [ Google Scholar ]
  • Recio G, Conrad M, Hansen LB, & Jacobs AM (2014). On pleasure and thrill: The interplay between arousal and valence during visual word recognition . Brain and Language , 134 , 34–43. doi: 10.1016/j.bandl.2014.03.009 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Reed DK, & Kershaw-Herrera S (2016). An examination of text complexity as characterized by readability and cohesion . Journal of Experimental Education , 84 ( 1 ), 75–97. doi: 10.1080/00220973.2014.963214 [ CrossRef ] [ Google Scholar ]
  • Reich W, Welner Z, & Herjanic B (1997). Diagnostic Interview for Children and Adolescents–IV (DICA–IV) . Toronto, ON, Canada: Multi-Health Systems. [ Google Scholar ]
  • Rights JD, & Sterba SK (2019). Quantifying explained variance in multilevel models: An integrative framework for defining R-squared measures . Psychological Methods , 24 ( 3 ), 309–338. 10.1037/met0000184 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Ripoll Salceda JC, Aguado Alonso G, & Castilla-Earls AP (2014). The simple view of reading in elementary school: A systematic review. Revista de Logopedia , Foniatría y Audiología , 34 ( 1 ), 17–31. doi: 10.1016/j.rlfa.2013.04.006 [ CrossRef ] [ Google Scholar ]
  • Schrank FA, Mather N, & McGrew KS (2014). Woodcock-Johnson Tests of Achievement (4th ed.). Itasca, IL: Riverside. [ Google Scholar ]
  • Sheehan KM, Flor M, & Napolitano D (2013). A two-stage approach for generating unbiased estimates of text complexity. In Rello L, Saggion H, & Baeza-Yates R (Eds.), NLP4ITA 2013: Proceedings of the Second Workshop of Natural Language Processing for Improving Textual Accessibility (pp. 49–58). Stroudsburg, PA: Association for Computational Linguistics. [ Google Scholar ]
  • Sotirchos ES, Fitzgerald KC, & Crainiceanu CM (2019). Reporting of R2 statistics for mixed-effects regression models . JAMA Neurology , 76 ( 4 ), 507. doi: 10.1001/jamaneurol.2018.4720 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Spencer M, Gilmour AF, Miller AC, Emerson AM, Saha NM, & Cutting LE (2019). Understanding the influence of text complexity and question type on reading outcomes . Reading and Writing , 32 ( 3 ), 603–637. doi: 10.1007/s11145-018-9883-0 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Spencer M, Richmond MC, & Cutting LE (2020). Considering the role of executive function in reading comprehension: A structural equation modeling approach . Scientific Studies of Reading , 24 ( 3 ), 179–199. doi: 10.1080/10888438.2019.1643868 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Strain E, & Herdman CM (1999). Imageability effects in word naming: An individual differences analysis . Canadian Journal of Experimental Psychology , 53 ( 4 ), 347–359. doi: 10.1037/h0087322 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Tierney NJ, & Cook DH (2020). Expanding tidy data principles to facilitate missing data exploration, visualization and assessment of imputations (Rev. ed.) . Retrieved from https://arxiv.org/abs/1809.02264
  • Tilstra J, McMaster K, van den Broek P, Kendeou P, & Rapp D (2009). Simple but complex: Components of the simple view of reading across grade levels . Journal of Research in Reading , 32 ( 4 ), 383–401. doi: 10.1111/j.1467-9817.2009.01401.x [ CrossRef ] [ Google Scholar ]
  • Tiu RD Jr., Thompson LA, & Lewis BA (2003). The role of IQ in a component model of reading . Journal of Learning Disabilities , 36 ( 5 ), 424–436. doi: 10.1177/00222194030360050401 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Tortorelli LS (2020). Beyond first grade: Examining word, sentence, and discourse text factors associated with oral reading rate in informational text in second grade . Reading and Writing , 33 , 143–170. 10.1007/s11145-019-09956-5 [ CrossRef ] [ Google Scholar ]
  • Tversky B, & Hard BM (2009). Embodied and disembodied cognition: Spatial perspective-taking . Cognition , 110 ( 1 ), 124–129. doi: 10.1016/j.cognition.2008.10.008 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • van der Meer T, te Grotenhuis M, & Pelzer B (2010). Influential cases in multilevel modeling: A methodological comment . American Sociological Review , 75 ( 1 ), 173–178. doi: 10.1177/0003122409359166 [ CrossRef ] [ Google Scholar ]
  • van Dijk TA, & Kintsch W (1983). Strategies of discourse comprehension . San Diego, CA: Academic. [ Google Scholar ]
  • van Hell JG, Verhoeven L, Tak M, & van Oosterhout M (2005). To take a stance: A developmental study of the use of pronouns and passives in spoken and written narrative and expository texts in Dutch . Journal of Pragmatics , 37 ( 2 ), 239–273. doi: 10.1016/j.pragma.2004.08.004 [ CrossRef ] [ Google Scholar ]
  • Verhoeven L, & van Leeuwe J (2008). Prediction of the development of reading comprehension: A longitudinal study . Applied Cognitive Psychology , 22 ( 3 ), 407–423. doi: 10.1002/acp.1414 [ CrossRef ] [ Google Scholar ]
  • Warriner AB, Kuperman V, & Brysbaert M (2013). Norms of valence, arousal, and dominance for 13,915 English lemmas . Behavior Research Methods , 45 ( 4 ), 1191–1207. doi: 10.3758/s13428-012-0314-x [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Wechsler D (1999). Wechsler Abbreviated Scale of Intelligence . San Antonio, TX: Psychological. [ Google Scholar ]
  • Wechsler D (2011). Wechsler Abbreviated Scale of Intelligence–second edition (WASI-II) . San Antonio, TX: NCS Pearson. [ Google Scholar ]
  • Werlen E, Imhof C, Benites F, & Bergamin P (2019). The reader’s feeling and text-based emotions: The relationship between subjective self-reports, lexical ratings, and sentiment analysis. In Cieliebak M, Tuggener D, & Benites F (Eds.), Proceedings of the 4th edition of the Swiss Text Analytics Conference (Vol. 2458 , Paper 4). Aachen, Germany: CEUR-WS.org. [ Google Scholar ]
  • Wilson AC, & Bishop DVM (2019). “If you catch my drift…”: Ability to infer implied meaning is distinct from vocabulary and grammar skills . Wellcome Open Research , 10 ( 4 ), Article 68. 10.12688/wellcomeopenres.15210.2 [ PMC free article ] [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Wolfe CR, Dandignac M, & Reyna VF (2019). A theoretically motivated method for automatically evaluating texts for gist inferences . Behavior Research Methods , 51 , 2419–2437. doi: 10.3758/s13428-019-01284-4 [ PubMed ] [ CrossRef ] [ Google Scholar ]
  • Woodcock RW, McGrew KS, & Mather N (2001). Woodcock–Johnson III Tests of Achievement . Rolling Meadows, IL: Riverside. [ Google Scholar ]
  • Zwaan RA, Magliano JP, & Graesser AC (1995). Dimensions of situation model construction in narrative comprehension . Journal of Experimental Psychology: Learning, Memory, and Cognition , 21 ( 2 ), 386–397. doi: 10.1037/0278-7393.21.2.386 [ CrossRef ] [ Google Scholar ]

Few Teachers Learn About ‘Science of Reading’ in Their Prep Programs. Some Colleges Are Working on That

research about reading comprehension

  • Share article

Just four years into her career as an elementary school teacher, 26-year-old Kaylee Hutcheson has established herself as a resident expert on evidence-based reading instruction at Hawthorne Elementary School in Mexico, Mo. Recently, she was even appointed to sit on her district’s literacy committee. But the recent college graduate admits that her expertise in the “science of reading” has little to do with what she learned in either her undergraduate or master’s level coursework.

“I had no idea about the science of reading when I was in college,” Hutcheson said. “We weren’t aware that it was so important.”

That’s through no fault of her own.

Between 2013 and the start of 2024, 37 states and the District of Columbia passed laws or implemented policies related to evidence-based reading instruction, according to an Education Week analysis. But the onus for who will train these states’ educators in evidence-based literacy instruction has fallen primarily on school districts and, by extension, existing classroom teachers—not the colleges and universities that train the teachers.

“Only about a quarter of the teachers who leave teacher preparation programs across our nation enter classrooms prepared to teach kids to read [in a way that’s] aligned to the science and research on reading,” said Heather Peske, president of the National Council on Teacher Quality, or NCTQ, in conjunction with the release of a 2023 study on the topic by her organization.

The 2023 NCTQ analysis rated the majority of licensure exams “weak”, observing that many were not adequately addressing all five science-based components of reading proficiency, as developed by the National Reading Panel : phonemic awareness, phonics, vocabulary, fluency, and comprehension. To address that gap, the NCTQ called for a “transition to a stronger reading licensure test.” But to pass reading licensure tests that incorporate the tenets of evidence-based literacy instruction, teacher candidates must first be taught them. .

Why aren’t more teachers-in-training learning evidence-based literacy instruction?

There’s no single or clear-cut reason why the push to learn evidence-based literacy instruction hasn’t focused on colleges of education. Some policy experts suggest that tenured faculty at colleges of education are slow to change their long-entrenched pedagogy and associated teaching methods.

“They’re not shifting fast enough,” said Javaid Siddiqi, president of The Hunt Institute, a nonprofit affiliated with Duke University’s Sanford School of Public Policy that advocates for positive changes in public education. “There’s a mentality among some tenured faculty that ‘this is the way I’ve always taught.’”

Some research also suggests that college instructors’ long-held theories about how best to teach literacy do not align with evidence-based best practices, even though the research establishing ‘science of reading’ practices has come out of higher education. In a recent nationally representative EdWeek Research Center survey of postsecondary reading instructors, 68 percent of respondents agreed that “ balanced literacy ,” an approach proponents say combines explicit instruction, guided practice, and independent reading and writing, best described their philosophy of teaching early reading.

But some higher education experts have a different take on why not all colleges of education instruct aspiring teachers in evidence-based literacy approaches.

“I think that there are a lot of really strong programs out there, but also a lot of really weak ones,” said Holly Lane, director of the University of Florida Literacy Institute . Part of the University of Florida College of Education, the institute provides programming to prepare pre-service and current educators for teaching foundational reading skills using evidence-based practices.

Lane blames the lack of uniformly high quality literacy instruction in part on the prevalence of ill-qualified adjunct professors, which she sees as a serious and pervasive problem in colleges of education. “We have a shortage of qualified people preparing teachers,” she said.

Challenges to the evolution of teacher-prep programs

Changes, however slow, are afoot. As of January, 21 states have passed some form of relevant legislation requiring that institutes of higher education and teacher preparation programs review their course offerings or instructional approaches, bring them in line with evidence-based practices, and require courses to cover certain topics related to early reading, according to an Education Week analysis.

Some literacy experts expressed skepticism about the effectiveness of such legislation. Lane, at the University of Florida Literacy Institute, said that, although she believes lawmakers have children’s best intentions in mind, the focus of this legislation is sometimes misguided. For instance, it would be more effective if the new laws focused on what to teach, rather than what not to teach, she explained.

“Banning the 3-cueing system —that’s not going to have a real effect,” Lane said, referring to the practice of prompting students to draw on context, sentence structure, and letters to identify words. “Instead of banning things, making sure they’re using evidence-based practices and programs makes sense.”

Lane raised another question about recent literacy legislation: Does the expertise exist at the state level to make these sort of nuanced decisions around literacy instruction?

That’s where organizations like the Hunt Institute may help. Siddiqi, the institute’s president, has both education and policy experience as a former science teacher, principal, and state education secretary in Virginia.

“We’ve been working with states and state teams to transform teacher preparation and licensure programming to ensure that the science of reading is embedded in [teachers’] learning experience,” Siddiqi said.

The Path Forward , one of the Institute’s signature programs, was launched to strengthen alignment with evidence-based reading instruction and teacher preparation, program approval, and licensure. It supports individual state teams—groups of six or seven individuals working together to shift their teacher preparation and licensure programs to include evidence-based literacy approaches. So far, 18 states have signed on to the program, which operates via virtual meetings and targeted coaching support.

The support and specific goals vary depending on the needs of individual participating states. For instance, North Carolina’s cohort collaborated with educator preparation programs in the state to ensure all pre-service teachers are trained in evidence-based reading instruction before licensure. Arizona’s cohort worked with a state literacy nonprofit on a whitepaper codifying core principles of elementary teacher preparation on early literacy that reflect the continuum of effective literacy practices for students from age 8 through 8th grade.

Siddiqi said policymakers pay attention to what other states are doing, and that this form of positive peer pressure may help facilitate more states to adopt legislation related to changes in literacy instruction. “Having worked in the governor’s office in Virginia myself, we would always pay attention [to what neighboring states were doing],” he said.

In the meantime, the responsibility to learn how to teach students to read proficiently using evidence-based methods will likely continue to fall primarily on teachers—both those who have been using other approaches for years as well as newer teachers like Hutcheson.

Hutcheson has been among the first wave of teachers in Missouri’s Mexico school district to receive training in Language Essentials for Teachers of Reading and Spelling, or LETRS , a program which instructs teachers in essential literacy skills and the research behind them. She also attends conferences and other training as time permits to enhance her knowledge of evidence-based literacy instruction.

“What I’ve learned,” Hutcheson said, “is that understanding the science of reading is key for educators to provide the best possible literacy support to their students.”

Sign Up for EdWeek Update

Edweek top school jobs.

Female teacher reads to multi-cultural elementary school students sitting on floor in class at school

Sign Up & Sign In

module image 9

Axios Science

Reading print is better for comprehension, study finds

Leisure reading on paper helps with text comprehension better than reading on digital devices, according to a new study.

Driving the news: "The main conclusion is that leisure reading habits on screen are minimally related to reading comprehension," researchers at the University of Valencia found.

By the numbers: The researchers, who analyzed more than two dozen studies, said "the relationship between the frequency of reading printed texts and text comprehension is much higher (between 0.30 and 0.40) than what we found for leisure digital reading habits (0.05)."

  • "This means, for example, that if a student spends 10 hours reading books on paper, their comprehension will probably be 6 to 8 times greater than if they read on digital devices for the same amount of time," study co-authors Cristina Vargas and Ladislao Salmerón said .

Of note: The study also found that as students get older, the relationship between recreational reading on digital devices and text comprehension improves.

Details: The researchers analyzed 25 studies on reading comprehension published between 2000 and 2022, with more than 450,000 participants.

  • "One might have expected that reading for informational purposes (i.e., visiting Wikipedia or other educational websites; reading news, or reading e-books) would be much more positively related to comprehension, but this is not the case," one researcher said.
  • The study was published earlier this week in the Review of Educational Research .

Go deeper: Attempts to ban books at public libraries surge at record levels

Get the rundown of the biggest stories of the day with Axios Daily Essentials.

Reading print is better for comprehension, study finds

Microsoft Research Blog

Structured knowledge from llms improves prompt learning for visual language models.

Published February 27, 2024

By Xinyang Jiang , Senior Researcher Yubin Wang , Research Intern Dongsheng Li , Principal Research Manager Cairong Zhao , Professor

Share this page

  • Share on Facebook
  • Share on Twitter
  • Share on LinkedIn
  • Share on Reddit
  • Subscribe to our RSS feed

This research paper was presented at the 38th Annual AAAI Conference on Artificial Intelligence (opens in new tab) (AAAI-24), the premier forum for advancing understanding of intelligence and its implementation in machines.

First page of the

We’re seeing remarkable abilities from visual language models in transforming text descriptions into images. However, creating high-quality visuals requires crafting precise prompts that capture the relationships among the different image elements, a capability that standard prompts lack. In our paper, “ Learning Hierarchical Prompt with Structured Linguistic Knowledge for Language Models ,” presented at AAAI-24, we introduce a novel approach using large language models (LLMs) to enhance the images created by visual language models. By creating detailed graphs of image descriptions, we leverage LLMs’ linguistic knowledge to produce richer images, expanding their utility in practical applications. 

An example of three types of prompts used in VLM to recognize bird, which is  templated prompt (a photo of a bird), a natural language based prompt that descript the bird category, and a tree structured prompt highlight the key entities of birds and the corresponding attributes, such as beak, wings, etc.

Figure 1 illustrates our method for constructing a structured graph containing key details for each category, or class. These graphs contain structured information, with entities (objects, people, and concepts), attributes (characteristics), and the relationships between them. For example, when defining “water lily,” we include entities like “leaves” or “blooms”, their attributes, “round” and “white”, and then apply LLMs’ reasoning capabilities to identify how these terms relate to each other. This is shown in Figure 2.

The pipeline and instructions to autonomously generate category description and the knowledge graph with LLM. We first instruct the LLM to give a category description, and  then it is asked to parse the key entities, attributes and their relationships from the un-structured  description.

Microsoft Research Podcast

research about reading comprehension

AI Frontiers: Models and Systems with Ece Kamar

Ece Kamar explores short-term mitigation techniques to make these models viable components of the AI systems that give them purpose and shares the long-term research questions that will help maximize their value. 

How to model structural knowledge

After identifying and structuring the relationships within the generated prompt descriptions, we implement Hierarchical Prompt Tuning (HTP), a new prompt-tuning framework that organizes content hierarchically. This approach allows the visual language model to discern the different levels of information in a prompt, ranging from specific details to broader categories and overarching themes across multiple knowledge domains, as shown in Figure 3. This facilitates the model’s understanding of the connections among these elements, improving its ability to process complex queries across various topics.

The overall framework of the proposed hierarchical prompt tuning.  Descriptions and relationship-guided graphs with class names are used as input for the frozen text encoder and the hierarchical prompted text encoder respectively.

Central to this method is a state-of-the-art relationship-guided attention module, designed to help the model identify and analyze the complex interconnections among elements within a graph. This module also understands the interactions between different entities and attributes through a cross-level self-attention mechanism. Self-attention enables the model to assess and prioritize various parts of the input data—here, the graph—according to their relevance. “Cross-level” self-attention extends this capability across various semantic layers within the graph, allowing the model to examine relationships at multiple levels of abstraction. This feature helps the model to discern the interrelations of prompts (or input commands/questions) across these various levels, helping it gain a deeper understanding of the categories or concepts.

Our findings offer valuable insights into a more effective approach to navigating and understanding complex linguistic data, improving the model’s knowledge discovery and decision-making processes. Building on these advances, we refined the traditional approach to text encoding by introducing a hierarchical, prompted text encoder, shown in Figure 4. Our aim is to improve how textual information is aligned or correlated with visual data, a necessity for vision-language models that must interpret both text and visual inputs.

Frameowork of the hierarchical prompted text encoder, where we apply three types of prompts, low-level prompts, high-level prompts, and global-level prompts for hierarchical tuning, and design a relationship-guided attention module for better modeling structure knowledge.

Looking ahead

By incorporating structured knowledge into our model training frameworks, our research lays the groundwork for more sophisticated applications. One example is enhanced image captioning, where visual language models gain the ability to describe the contents of photographs, illustrations, or any visual media with greater accuracy and depth. This improvement could significantly benefit various applications, such as assisting visually impaired users. Additionally, we envision advances in text-to-image generation, enabling visual language models to produce visual representations that are more precise, detailed, and contextually relevant based on textual descriptions.

Looking forward, we hope our research ignites a broader interest in exploring the role of structured knowledge in improving prompt tuning for both visual and language comprehension. This exploration is expected to extend the use of these models beyond basic classification tasks—where models categorize or label data—towards enabling more nuanced and accurate interactions between people and AI systems. By doing so, we pave the way for AI systems to more effectively interpret the complexities of human language.

Acknowledgements

Thank you to Yubin Wang for his contributions in implementing the algorithm and executing the experiments.

Related publications

Learning hierarchical prompt with structured linguistic knowledge for vision-language models, meet the authors, xinyang jiang.

Senior Researcher

Portrait of Yubin Wang

Research Intern

Portrait of Dongsheng Li

Dongsheng Li

Principal Research Manager

Portrait of Cairong Zhao

Cairong Zhao

Tongji University

Continue reading

three conversation bubbles on a blue, purple, and pink gradient background

Steering at the Frontier: Extending the Power of Prompting

EMNLP 2023 logo to the left of accepted paper "LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models" on a blue/green gradient background

LLMLingua: Innovating LLM efficiency with prompt compression

Illustrated icons of a medical bag, hexagon with circles at its points, and a chat bubble on a blue and purple gradient background.

The Power of Prompting

"Microsoft at AIES 2023: Social Biases through the Text-to-Image Generation Lens" title to the left of the front page of said publication on a red, abstract background.

Understanding social biases through the text-to-image generation lens

Research areas.

research about reading comprehension

Related labs

  • Microsoft Research Lab - Asia
  • Follow on Twitter
  • Like on Facebook
  • Follow on LinkedIn
  • Subscribe on Youtube
  • Follow on Instagram

Share this page:

IMAGES

  1. Reading Comprehension Strategies Theories Interventions And Technologies Pdf

    research about reading comprehension

  2. Type my essay Research paper on reading comprehension

    research about reading comprehension

  3. Reading Comprehension Qualitative Research Pdf

    research about reading comprehension

  4. 😂 Reading comprehension research topics. Comprehension: Concepts and Research. 2019-02-16

    research about reading comprehension

  5. (PDF) What research has to say about reading instruction

    research about reading comprehension

  6. New Methods in Reading Comprehension Research

    research about reading comprehension

COMMENTS

  1. Reading Comprehension Research: Implications for Practice and Policy

    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 ...

  2. The Effectiveness of Reading Strategies on Reading Comprehension

    According to research by Banditvilai (2020), reading strategies help students' reading comprehension. The strategies-skimming, scanning, generating predictions, and questioning-helped students ...

  3. What Research Tells Us About Reading, Comprehension, and Comprehension

    For many years, reading instruction was based on a concept of reading as the application of a set of isolated skills such as identifying words, finding main ideas, identifying cause and effect relationships, comparing and contrasting and sequencing. Comprehension was viewed as the mastery of these skills. One important classroom study conducted ...

  4. The Science of Reading Comprehension Instruction

    Decades of research offer important understandings about the nature of comprehension and its development. Drawing on both classic and contemporary research, in this article, we identify some key understandings about reading comprehension processes and instruction, including these: Comprehension instruction should begin early, teaching word-reading and bridging skills (including ...

  5. PDF Reading Comprehension

    An important specification of the SVR is that reading comprehension is the product of language comprehension and decoding, rather than the sum of these two components. Thus, an increase or decrease in one com ponent depends on the level of the other component in terms of influencing reading com prehension ability.

  6. The Use of New Technologies for Improving Reading Comprehension

    Introduction. Reading comprehension is a fundamental cognitive ability for children, that supports school achievement and successively participation in most areas of adult life (Hulme and Snowling, 2011).Therefore, children with learning disabilities (LD) and special educational needs who show difficulties in text comprehension, sometimes also in association with other problems, may have an ...

  7. Reading comprehension research: Implications for practice and policy

    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 ...

  8. How the Science of Reading Informs 21st‐Century Education

    The science of reading should be informed by an evolving evidence base built upon the scientific method. Decades of basic research and randomized controlled trials of interventions and instructional routines have formed a substantial evidence base to guide best practices in reading instruction, reading intervention, and the early identification of at-risk readers.

  9. The Role of Background Knowledge in Reading Comprehension: A Critical

    View PDF View EPUB. A critical review was conducted to determine the influence background knowledge has on the reading comprehension of primary school-aged children. We identified twenty-three studies that met our criteria and focused on the links between background knowledge and reading comprehension of children in the mid to late primary years.

  10. Reading Comprehension

    Summary. Reading comprehension requires the construction of a coherent mental representation of the information in a text. Reading involves three interrelated elements—the reader, the text, and the activity, all situated into a broader sociocultural context. The complexity inherent in reading comprehension has given rise to a multitude of ...

  11. Vocabulary and Reading Comprehension Revisited: Evidence for High-, Mid

    Research Question 1 addressed the contribution of the three levels of vocabulary knowledge to reading comprehension in relation to a single cohort of learners. Results showed that for the entire cohort involved in the study, OVK of words across high- and mid-frequency ranges was collectively capable to predict about 66% of the variance observed ...

  12. Levels of Reading Comprehension in Higher Education: Systematic Review

    This review is a guide to direct future research, broadening the study focus on the level of reading comprehension using digital technology, experimental designs, second languages, and investigations that relate reading comprehension with other factors (gender, cognitive abilities, etc.) that can explain the heterogeneity in the different ...

  13. PDF Improving Reading Comprehension

    Reading comprehension has been a major issue for more than 20 years. Researchers have ... would improve their reading comprehension. My research partner and I feel very strongly about our students improving their reading comprehension and to become life-long learners. I was looking forward to working on this area of concern, and sharing

  14. Reading Comprehension Research: Implications for Practice and Policy

    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 ...

  15. The Contribution of Text Characteristics to Reading Comprehension

    Theories of Reading Comprehension. In the developmental literature, arguably the most prominent reading comprehension model is the simple view of reading (SVR; Hoover & Gough, 1990), which is centered on person-level influences on reading comprehension.This framework asserts that reading comprehension is determined by the interaction between a reader's ability to decode text and his or her ...

  16. Improving Reading Skills Through Effective Reading Strategies

    The research question is, The purpose of this study was to analyze the improvement of the students reading skills after they have taken presentations on reading strategies. 712 Hülya KüçükoÄŸlu / Procedia - Social and Behavioral Sciences 70 ( 2013 ) 709 â€" 714 3.Method Reading proficiency is the most fundamental skill for ...

  17. A comprehensive review of research on reading comprehension strategies

    Considering the research foci and findings, we identified seven categories: (a) comparison of the strategy use in L1 and L2 reading; (b) comparison of EAL readers' and monolinguals' comprehension strategy use; (c) different L1 groups' strategy use; (d) the role of languages in the strategy use; (e) the relationship between reading proficiency and comprehension strategy use; (f) strategies in ...

  18. PDF Reading Comprehension, What We Know: A Review of Research ...

    reading comprehension is measured and research that addresses this concern is reviewed. Suggests related to how reading comprehension can be improved are presented. Keywords: reading comprehension, strategies, testing Introduction Reading is an activity performed to develop an understanding of a subject or topic.

  19. Reading Comprehension Research: Implications for Practice and Policy

    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 ...

  20. Handbook of Research on Reading Comprehension

    The Handbook of Research on Reading Comprehension assembles researchers of reading comprehension, literacy, educational psychology, psychology, and neuroscience to document the most recent research on the topic. It summarizes the current body of research on theory, methods, instruction, and assessment, including coverage of landmark studies. Designed to deepen understanding of how past ...

  21. Assessments of reading comprehension: Challenges and directions

    The importance of theory-guided formal assessments of comprehension has been articulated by Kintsch, as well as by measurement experts; therefore, the authors define reading comprehension through the eyes of theorists and discuss measures for assessing it. The authors acknowledge many theories of reading comprehension but have chosen two of them as examples of how models of comprehension can ...

  22. How lexical quality predicts L2 reading comprehension in early

    Acquiring reading comprehension skills is a fundamental aspect of primary education, with the lexicon playing a pivotal role in this process (Perfetti and Helder Citation 2022).The relation between lexical quality and reading comprehension is exemplified in the lexical quality hypothesis by Perfetti and Hart (Citation 2002).This hypothesis claims that the more phonologically, semantically, and ...

  23. (Pdf) Action Research in Reading

    Solution. 85-92. 93-118. 120-124. 3. in Macatoc Elementary School. I. ABSTRACT. Teachers need to focus on extensive comprehension instruction. with all students, not just successful readers.

  24. The Impact of Peer-Collaborative Strategic Reading and ...

    Abstract. Research indicates that reading strategy instruction improves comprehension. Conceptualizing strategy training as mediating reading strategy use through collaborative and reflective practices, the present study examined the combined effect of peer-collaborative strategic reading and reflective journaling on strategy use and comprehension.

  25. Few Teachers Learn About 'Science of Reading' in Their Prep Programs

    Between 2013 and the start of 2024, 37 states and the District of Columbia passed laws or implemented policies related to evidence-based reading instruction, according to an Education Week ...

  26. Silent and oral reading methods on improving English reading

    Using ex post facto design, this study investigated the effects of reading methods on English reading comprehension of randomly selected 75 Grade 2 pupils in a private school in the Philippines.

  27. Reading print is better for comprehension, study finds

    Leisure reading on paper helps with text comprehension better than reading on digital devices, according to a new study. Driving the news: "The main conclusion is that leisure reading habits on ...

  28. Structured knowledge from LLMs improves prompt learning for visual

    Looking forward, we hope our research ignites a broader interest in exploring the role of structured knowledge in improving prompt tuning for both visual and language comprehension. This exploration is expected to extend the use of these models beyond basic classification tasks—where models categorize or label data—towards enabling more ...