Introduction

While picture books are primarily used in schools for literature education, they also hold significant potential for language development. Reading aloud can provide access to foundational literacy and significantly supports both reading socialization and language development in children (Wieler, 2021). Especially during dialogic reading of picture books (Whitehurst et al., 1988) the interactions between teacher and children, as well as among the children themselves, create various learning opportunities that the teacher can leverage for targeted goals. Studies have shown that dialogic reading has positive effects on language acquisition, and it is frequently recommended as a method for language support (Ennemoser et al., 2013).

However, the grammatical skills of children, which are particularly relevant for academic success (Motsch & Rietz, 2016), have not yet been explicitly focused on in scientific studies on the effectiveness of dialogic reading in German-speaking countries. Furthermore, dialogic reading is used almost exclusively in daycare centers with children under the age of six years as an implicit everyday language support method, but not in schools.

The ZDL School project therefore poses the research question whether dialogic reading can be implemented in schools as a method to promote grammar development in children aged six years and older.

Methods

In the ZDL School project, the dialogic reading method (Whitehurst et al., 1988) was enhanced with language diagnostics and two speech-language therapy techniques: input specification and evoking grammatical structures in order to specifically promote the grammar acquisition of participating children. This integrated approach, called Goal-Oriented Dialogic Reading (ZDL; Schütz, 2022), combines the implicit language support of dialogic reading with explicit speech-language therapy elements. Goal-Oriented Dialogic Reading (ZDL) therefore lies at the intersection of German language didactics and speech-language pathology.

Between October 2022 and June 2025, 315 children in inclusive primary schools participated weekly in small-group (4–6 children) dialogic reading sessions using the ZDL method. Children’s grammatical development was assessed using standardized tests and compared to a control group (N = 117) that did not receive ZDL support. Video recordings were also collected to analyze interaction patterns during sessions.

Results and discussion

Results indicate that children receiving ZDL support showed significantly greater gains in grammatical skills compared to the control group. Children with the lowest baseline grammatical skills - those forming fewer than 80% of main clauses correctly - showed substantial gains from the ZDL intervention. The ZDL group demonstrated significantly greater improvement than the control group with comparable initial grammatical proficiency (p < .05).

Outlook

In the second phase, beginning in October 2025, the ZDL method will be implemented in larger and more heterogeneous groups, with ongoing evaluation. A second control group will engage in 'classic' dialogic reading, allowing a direct comparison of grammatical skill development between the two approaches. The first results of this second project phase will be obtained in August 2026 and will allow a more differentiated analysis of the effectiveness of the language support method Goal-Oriented Dialogic Reading (ZDL).

Educational Significance

The findings contribute to the ongoing discourse on integrating evidence-based language support into early education. These results demonstrate that integrating grammar-focused support into dialogic reading positively impacts language acquisition, opening new avenues for interdisciplinary integration of literary and linguistic instruction in educational practice.

Introduction

Inferential skills play a central role in reading comprehension and are a key component of learning across school subjects as they allow students to construct coherent mental representations and integrate new information. Given their importance, they are the subject of extensive research. Studies typically adopt a top-down approach, using tasks designed and balanced to specifically target inferential abilities (often based on narrative texts). While this approach allows for good experimental control, the resulting assessments may differ substantially from those used in everyday school practice. This study adopts a bottom-up perspective, focusing on the inferential demands embedded in assessment items used in the Italian national assessment in primary school. The Italian National Assessment System (INVALSI) provides access to large-scale, ecologically valid assessments, as it uses authentic narrative and expository texts drawn from children’s books. Moreover, it allows longitudinal observations, since data from the same students can be linked over time, and includes information on relevant background variables such as socioeconomic status, first language, migratory background and gender. This study aims to examine which types of inferences are involved in large-scale reading comprehension tasks, how students’ performance on these inferential demands evolves across Grades 2 and 5, and how inferential skills relate to reading achievement while accounting for background characteristics. The guiding research questions are: (1) Which types of inferences are most frequently required in the Italian large-scale reading comprehension assessments? (2) How do inferential demands influence student performance in Grade 5? (3) To what extent is reading comprehension in Grade 5 associated with performance in Grade 2, considering sociodemographic characteristics? and (4) Which inferential demands are the most challenging for those students who showed difficulties in Grade 2, controlling for sociodemographic characteristics?

Methods

This study analysed longitudinal data from INVALSI reading comprehension assessments administered in primary school (Grades 2 and 5). Reading comprehension items were analysed and classified using a double-blind procedure according to the type of inference required (Rice et al.’s taxonomy, 2024). This classification allows analyses at both a fine-grained level, focusing on specific inferential types, and at a broader level, based on three inferential macro-categories. This categorisation was then used to link item-level inferential demands with performance data across Grades 2 and 5 in two cohorts, enabling the examination of longitudinal relationships between early reading comprehension skills and later performance.

Results and Discussion

Considering both narrative and expository texts used in Grade 5 reading comprehension assessments, the results show that inferential demands are not uniformly distributed across inference categories. Coherence and elaborative inferences are the most frequently assessed (43% and 36% for Cohort A, and 30% and 44% for Cohort B, respectively).

Explanatory Item Response Modelling (EIRM) was used to analyse the data. This approach extends traditional item response theory by incorporating covariates to explain item difficulty and student ability. EIRM was employed to investigate the role of item-level covariates, student-level covariates, and their interaction on the likelihood of a correct response.

The results indicate that the type of inference significantly influences item difficulty, with elaborative inferences being the most challenging overall. Reading comprehension performance in Grade 2 is significantly associated with later performance in Grade 5, even when sociodemographic characteristics are accounted for. Students who showed difficulties in Grade 2 often continue to struggle in Grade 5, particularly with word meaning inferences, whereas students with stronger prior comprehension skills handle them more efficiently. These findings suggest that inferential skills develop unevenly over time and across inference types, and that specific inferential processes may represent a persistent source of difficulty for students with early comprehension weaknesses.

Educational Significance of the Research

This study links students’ reading achievement, assessed through authentic large-scale assessments, to specific inferential skills, while accounting for background characteristics and longitudinal learning trajectories. By identifying which inferential demands are most challenging for different groups of students and how performance interacts with prior skills and background conditions, the findings provide insights for assessment design, instructional planning and targeted interventions. Overall, the study informs educational practices aimed at supporting inferential processing in reading and across subject areas and promotes grounded interpretations of large-scale assessment data within an interdisciplinary framework.

1. Introduction: theoretical background, aims, and research questions.

Advanced digital reading with the aim of gaining knowledge from online sources can be very demanding, with linguistic and digital competences being closely linked and to a large extent intertwined (cf., e.g., Leu et al., 2017). According to Salmerón et al. (2018), the interrelated subskills of (1) navigation, (2) integration, and (3) evaluation contribute to advanced digital reading. This integrative review aims to synthesize the expanding and diversified evidence in relation to effective instructional practices for teaching evaluation or validation processes as part of comprehension in advanced digital reading tasks. Various approaches, methods, and terminologies were used, and a range of (sub)processes, aspects, and specific objectives addressed in the studies published. Therefore, the main objective of this study is to identify findings on key aspects of the above research question, document the directions and the state of the art, and synthesize the perspectives and results in such a way that a theoretical model could be created that comprehensively represents intervention methods for promoting evaluation processes in digital reading which enables recommendations for future work and practical implications to be put forward. This also requires combining insights from different disciplines or research traditions. This review forms part of a broader program of work aiming at an integrated consideration of navigation (including selection), integration, and validation (evaluation) processes.

2. Methods

An integrative review was conducted in accordance with the guidelines for the conduct of integrative reviews presented by Torraco (2016), and the methodological strategies proposed by Whittemore and Knafl (2005) to enhance the rigor of the process. Because of the inconsistent terminology, and the variety of sub-processes and sub-goals addressed, the literature search was performed combining hand searching and electronic bibliographic databases. The strategy used for hand searching was ancestry searching and its opposite, taking into account available previous literature reviews on related questions or sub-questions. The bibliographic databases searched were PsycInfo (OVID), ERIC (OVID), and GoogleScholar. The search strategies for these databases included the definition of eligibility criteria and the selection of search terms (including variants, synonyms, Boolean operators etc.) based on the results of the hand search and pilot testing. A search protocol (including justification for all decisions for limiting the scope) was created. The PRESS checklist (McGowan et al., 2016) developed for systematic reviews was used used for checking search strategies. After conducting the search queries, the machine learning tool Abstrackr was used for screening the abstracts, titles, and keywords of the retrieved articles. The output of this screening process was subjected to a secondary review undertaken by the author. For data analysis, a coding scheme was developed, and the data from the primary sources were ordered, coded, categorized, and summarized with respect to comparisons, clustering, patterns, relationships, intervening factors and chains of evidence.

3. Results and discussion

A conceptual structure (cf. Torraco, 2016) has been chosen for organizing the results. The basic building blocks of this conceptual structure include the taxonomic structure of evaluation processes (with validation or evaluation at its highest level, and sourcing and identification of contradictions at intermediary levels), key factors (such as prior knowledge, self-efficacy/calibration, epistemic beliefs, and critical thinking), source type or matter at hand for (possible) evaluation (e.g., online search, digital inquiry, website, or multiple texts), information category (e.g., claims put forward by pertinent experts, misinformation, disinformation, content bias, source bias, fact vs. opinion, bullshit) and instructional interventions. Results include structured data displays in the form of tables and plots, and a theoretical framework for the analysis of instructional effects onto evaluation processes in digital reading. It reveals various didactic options in a complex structure, which are closely linked to other cognitive processes and resources (such as science understanding, representational updating, and inferential thinking).

4. Educational significance of the research

Many students struggle to conduct complex information searches, assess the reliability of websites and texts, and read complex digital documents efficiently. Accordingly, supporting students in critically evaluating online information is considered an urgent agenda in the field of education. The theoretical framework presented was developed with the aim of informing educational policy and practice – enabling knowledge transfer into action – and providing guidance for future research directions.