The inhibitory effect of word neighborhood size when reading with central field loss is modulated by word predictability and reading proficiency

Reading speed significantly impacts the quality of life for individuals with low vision, particularly those with central field loss (CFL) caused by maculopathy. Current rehabilitation and assistive technologies could benefit from a deeper understanding of the factors contributing to reading deficits in CFL patients. Previous research has shown disruptions in fixation patterns when low-vision patients read low-frequency words, suggesting the importance of psycholinguistic factors. This study focuses on two such factors: word frequency and word neighborhood size (the number of words differing by one letter), which are known to influence reading in normally sighted individuals. However, the effect of these factors in CFL, where visual input is degraded due to blurred, distorted, or occluded letters, is unclear. The researchers hypothesize that in CFL, word neighborhood size will have a negative effect on reading speed (Hypothesis 1), and that this effect will be influenced by word predictability and reading proficiency (Hypotheses 2 and 3). The study aims to explore the extent to which lexical factors contribute to reading difficulties in individuals with CFL and evaluate the potential of lexical text simplification as a rehabilitation strategy.

The literature review highlights the established influence of word frequency and word neighborhood size on reading in normally sighted individuals. For normal vision, larger word neighborhoods generally facilitate word recognition. However, with CFL, the impaired visual input and reliance on linguistic inference could alter this effect. Previous research by Stolowy et al. (2019) demonstrated a significant effect of word frequency on reading time in CFL patients, larger than that observed in normally sighted individuals, and this effect varied across synonyms. This underscores the potential influence of factors beyond frequency, leading to the current study's focus on orthographic similarity (neighborhood size) and word predictability.

Nineteen patients with bilateral CFL (13 females), aged 32-89 (mean ± SD = 75 ± 15), participated. All had a monocular acuity of 0.4 logMAR or worse in their better eye. Exclusion criteria included ophthalmologic disorders other than maculopathy, cognitive disorders, or pre-existing reading disorders. Data on age, etiology, lens status, disease onset, field loss, and reading habits (daily reading time, profession, prior reading habits) were collected. The self-paced reading paradigm was used, with sentences presented on an LCD monitor at a print size optimized for each participant's visual acuity. A French computerized version of MNREAD was used for measuring critical print size. Reading was performed monocularly (better eye) with appropriate near-vision correction. The reading material consisted of 32 pairs of synonymous target words (matched for length, with a neighborhood size difference of 5-10) embedded in pairs of matching sentences. These sentences were designed to control for length, sense, and target word position. Two conditions counterbalanced the potential effect of sentence structure. Participants unmasked each word one at a time, reading aloud and pressing keys. Reading accuracy and time were recorded. Word predictability was measured using trigram analysis based on a large French text corpus (Google Books Ngram Viewer). Reading proficiency was assessed based on self-reported daily reading time (categorized as 'yes' or 'no'). Generalized linear mixed-effects models (GLME for accuracy) and linear mixed-effects models (LME for reading time) were used to analyze the data. The models included word neighborhood size, word frequency, word length, trigram occurrence (predictability), daily reading (proficiency), and their interactions as independent variables.

Analysis 1 (GLME) revealed no significant effect of neighborhood size, word frequency, word length, or predictability on reading accuracy, which remained high (around 94%). However, there was a significant difference in accuracy between participants who continued daily reading versus those who had stopped (99.1% vs. 97.3%). Analysis 2 (LME) revealed a moderate inhibitory effect of neighborhood size on reading time (a 14% increase in reading time from 0 to 10 neighbors), confirming Hypothesis 1. Word predictability also significantly affected reading time. Analysis 3 (LME), incorporating a three-way interaction between neighborhood size, predictability, and reading proficiency, showed that the inhibitory effect of neighborhood size on reading time was modulated by both predictability and reading proficiency. This confirmed Hypotheses 2 and 3. For less proficient readers, the inhibitory effect of neighborhood size was most pronounced for low-predictability words (up to a 101% increase in reading time when the number of neighbors increased from 0 to 10). This effect lessened significantly as predictability increased. For proficient readers, however, the effect of neighborhood size was negligible regardless of predictability. Word frequency also significantly impacted reading time, with higher frequency words associated with faster reading times.

The findings support the hypothesis that the visual constraints of CFL lead to a reversed neighborhood size effect, particularly for less proficient readers. The results suggest that the reduced visual information forces readers to rely heavily on context and linguistic inference. The stronger inhibitory effect of neighborhood size for low-predictability words highlights the role of context in mitigating ambiguity. The modulation of this effect by reading proficiency emphasizes the importance of practice in adapting to the challenges of reading with CFL. The nearly perfect accuracy scores despite slower reading times suggest that participants employed compensatory strategies, potentially sacrificing speed for accuracy.

This study confirms the inhibitory effect of word neighborhood size in CFL reading, particularly for less proficient readers, and demonstrates the crucial role of word predictability in mitigating this effect. The findings strongly suggest that text simplification, focusing on reducing the complexity of words with many orthographic neighbors, could be a highly effective intervention for improving reading abilities in individuals with CFL, especially those who have discontinued daily reading. Future studies should explore this further, using diverse populations and taking into account individual visual characteristics. Investigating alternative definitions of orthographic neighborhood that account for letter transposition, addition, and deletion might offer additional insights.

The study's limitations include a limited age range (skewed towards older participants) and the lack of microperimetry data to account for individual differences in scotoma characteristics. The definition of orthographic neighborhood used may be too restrictive. The use of a self-paced reading paradigm that didn't allow for word skipping may also not perfectly reflect natural reading conditions. Finally, monocular testing, while controlling for eye characteristics, may not fully replicate real-world binocular reading.