Comprehending expository technical texts is a crucial skill in our technologically advanced world, essential for education, employment, and daily life. This study leverages functional brain imaging (fMRI) to enhance our understanding of the neurocognitive processes involved in technical reading comprehension. The research addresses two key questions: 1) Can we identify neuropsychological processes differentiating good and poor comprehenders, and what strategies do better comprehenders employ? 2) What text properties correlate with better or worse comprehension, and how can text structure be improved to facilitate comprehension and retention? While behavioral research has explored these questions, neuroscience research is comparatively limited. This fMRI study uses a naturalistic design with realistic technical passages to investigate the complex interplay of neural processes involved in comprehension, aiming to bridge the gap between behavioral and neuroimaging research and provide insights into improved teaching methods and text construction.

Existing psychological and educational research has examined the component processes of reading comprehension, including word decoding, meaning retrieval, syntactic processing, proposition construction, cohesion establishment, and working memory maintenance. At the discourse level, additional processes are engaged, such as establishing global coherence, managing information in working memory, retrieving semantic knowledge from long-term memory, and integrating this knowledge with text information. The construction of a mental model, incorporating prior knowledge to enable inferences, is particularly relevant for technical comprehension. Individual differences in proficiency across these processes, background knowledge, and the interaction between these factors and text structure all influence comprehension. Neuroimaging studies have implicated various brain regions in these processes, such as the left-lateralized parietal, temporal, and inferior frontal lobe areas for lexical retrieval and semantic knowledge representation, along with prefrontal cortex regions for conceptual and semantic integration. Mental model construction involves semantic and episodic memory, with associated regions distributed across the neocortex and medial temporal lobe, respectively. Previous neuroimaging studies on expository text comprehension are limited despite its significance for STEM education and job performance. This study addresses this gap.

Thirty-one college-age participants with varying comprehension skills (identified through a pre-test) underwent fMRI while reading 16 technical passages three times, followed by a multiple-choice comprehension test. The passages covered diverse technical and general knowledge topics, each consisting of five sentences. A moving window paradigm presented phrases at a rate adjusted for word length and frequency. During the first two presentations, question stems (not answers) were presented to orient participants. Post-scan, participants completed additional psychometric tests (Nelson-Denny Reading Comprehension Test, Raven's Progressive Matrices, Reading Span Test, and Bennett Mechanical Comprehension Test). fMRI data were acquired using a multiband slice-accelerated BOLD spin-echo EPI sequence. Data analysis involved calculating the mean percentage signal change (MPSC) relative to fixation, averaging across sentences within a 4-second window around the peak BOLD response. Voxelwise correlations between MPSC and comprehension performance were assessed to identify regions of interest (ROIs). Stepwise regression models predicted comprehension performance using ROI activation as predictors, and cross-validation was performed to ensure predictive generalizability. Similar analyses were conducted to investigate passage comprehensibility, relating activation to passage difficulty and considering text properties (Coh-Metrix measures) and topic familiarity.

The study yielded two primary findings: 1) brain activation patterns during reading predicted individual comprehension performance, and 2) brain activation patterns predicted passage comprehensibility. **Individual Differences:** Better comprehenders showed greater activation in regions associated with language processing (left inferior frontal gyrus [LIFG]), spatial processing (left superior parietal lobule [LSPL]), semantic integration (left dorsolateral prefrontal cortex [L DLPFC]), and episodic encoding (left and right hippocampal areas). Poorer comprehenders showed higher activation in areas related to episodic memory retrieval (left and right ventromedial prefrontal cortex [L and R VMPFC], right precuneus). A stepwise regression model reliably predicted individual comprehension performance (R = .76, adjusted R² = .72) using activation in LIFG, R DLPFC, RHC, and L VMPFC. Cross-validation confirmed this predictive validity (adjusted R² = .46). **Passage Comprehensibility:** Passage comprehensibility was related to activation in regions associated with integrating different knowledge representations and establishing conceptual coherence. A stepwise regression model using activation as predictors reliably predicted passage comprehensibility (R = .88, adjusted R² = .84), with four key regions: LIFG, right temporal pole (RTP), left inferior parietal lobule (LIPL), and medial anterior cingulate/dorsomedial prefrontal cortex (MACC/DMPFC). Cross-validation yielded an R² of .44. Brain activation measures were superior to psychometric measures (Raven's Progressive Matrices, Reading Span Test, Bennett Mechanical Comprehension Test) in predicting comprehension performance. Coh-Metrix measures of text readability (Syntactic Simplicity, Word Concreteness, etc.) weakly predicted passage comprehensibility, while Deep Cohesion was linked to activation in LIFG and LIPL, suggesting a mediating role in the relationship between cohesion and comprehension. Passage topic familiarity, although weakly correlated with comprehension, was significantly related to performance. Repeated readings showed non-monotonic changes in activation, possibly reflecting a shift from semantic to executive processing.

The findings highlight the importance of several cognitive processes in successful technical text comprehension. Better comprehenders utilize verbal working memory, spatial visualization to construct mental models, and integrate new information with prior semantic knowledge. Poorer comprehenders rely more on episodic memory retrieval. The association between passage difficulty and activation in areas integrating semantic information underscores the significance of text cohesion. These findings support the teaching of strategies that engage these processes (e.g., pre-teaching unfamiliar concepts, explicit visualization instruction, self-explanation techniques). Improving text structure by enhancing cohesion improves comprehension. fMRI offers a powerful tool to observe the neural processes underlying comprehension, informing the development of effective teaching strategies and text design.

This study provides novel insights into the neural and cognitive basis of expository text comprehension, demonstrating the predictive power of brain activation patterns for both individual comprehension and passage difficulty. The findings support the development of targeted teaching strategies focusing on working memory, spatial visualization, and semantic knowledge integration, and the creation of more cohesive and comprehensible technical texts. Future research should explore these strategies further and investigate the generalizability of these findings across diverse populations and text types.

The study's sample size was relatively small, and the participants were primarily college students, limiting the generalizability of the findings to other age groups and educational levels. The technical passages, while designed to be realistic, might not fully capture the complexity and variability of real-world technical texts. The reliance on a specific type of comprehension test might not fully capture the multifaceted nature of comprehension. Furthermore, the correlational nature of the study prevents definitive causal inferences between brain activation and comprehension performance.