Spelling in developmental dyslexia in Chinese: Evidence of deficits in statistical learning and over-reliance on phonology

Unraveling the complex interplay between statistical learning and working memory in Chinese children with and without dyslexia across different ages

The relation between statistical learning and working memory in children with developmental dyslexia (DD) remains unclear. This study employed a distributional and a conditional statistical learning experiment and a working memory task to examine this relation in 651 Chinese 6- to 12-year-olds with and without DD (NDD = 199, 101 females; NwoDD = 452, 227 females; participated 2014-2019). Results showed working memory positively associated with recognizing high-predictable and familiar items in both groups, but negatively associated with recognizing unfamiliar items in the DD group only. These findings uncovered the complex interplay between statistical learning and working memory, demonstrating how different working memory abilities in children with and without DD relate to various statistical learning mechanisms at the item level.

DysDiTect: Dyslexia Identification Using CNN-Positional-LSTM-Attention Modeling with Chinese Dictation Task

Handwriting difficulty is a defining feature of Chinese developmental dyslexia (DD) due to the complex structure and dense information contained within compound characters. Despite previous attempts to use deep neural network models to extract handwriting features, the temporal property of writing characters in sequential order during dictation tasks has been neglected. By combining transfer learning of convolutional neural network (CNN) and positional encoding with the temporal-sequential encoding of long short-term memory (LSTM) and attention mechanism, we trained and tested the model with handwriting images of 100,000 Chinese characters from 1064 children in Grades 2-6 (DD = 483; Typically Developing [TD] = 581). Using handwriting features only, the best model reached 83.2% accuracy, 79.2% sensitivity, 86.4% specificity, and 91.2% AUC. With grade information, the best model achieved 85.0% classification accuracy, 83.3% sensitivity, 86.4% specificity, and 89.7% AUC. These findings suggest the potential of utilizing machine learning technology to identify children at risk for dyslexia at an early age.

Systematic language input improved productions of elaborated verb phrases of Cantonese-speaking children with language difficulties

It has been well-documented that language input designed according to the principles of statistical learning can promote language acquisition among children with or without language disorder. Cantonese-speaking children with language disorder were reported to have difficulties using expanded verb phrases and prepositional phrases, but the corresponding intervention is relatively unexplored. The current study evaluated the efficacy of an intervention designed using the statistical learning principles to promote the acquisition of these two structures. A retrospective study of existing data collected from a total of 16 Cantonese-speaking children (four female; mean age = 6.70 years) with suspected language disorder was conducted. The participants were initially divided into the 'Treatment' and the 'Control' groups. A total of eight sessions of language treatment, which focused on giving systematic language input of expanded verb phrases and prepositional phrases, were conducted on each child. Results showed that the Treatment group produced significantly more expanded verb phrases in the post-treatment language samples, while the Control group did not. The final pre- and post-comparison conducted after the Control group also received treatment indicated overall significant increased number of expanded verb phrases produced across time. On the contrary, improvement in the production of prepositional phrases was not significant. It is suggested that the unique thematic roles coded by individual prepositions possibly restricted the generalisation effect of treatment, which explains the non-significant improvement across time. Theoretical and clinical implications were discussed.

Development of statistical learning ability across modalities, domains, and languages

Statistical learning (SL) is defined as our ability to use statistics (e.g., frequencies or transitional probabilities) to detect implicit regularities in the environment. Limited research has examined the developmental trajectory of SL across domains and modalities, and no previous research has made systematic comparisons across domains, modalities, and languages using comparable tasks. The current study investigated the development of SL ability across 9-, 11-, and 13-year-old native Chinese-speaking children in non-linguistic visual and auditory SL, first-language Chinese visual and auditory SL, and second-language English visual and auditory SL. Results showed that children across the three age groups achieved all types of SL, and they performed better in visual modality than in auditory modality. Furthermore, while visual SL constantly improved from 9- to 11- to 13-year-olds, auditory SL improved only from 11- to 13-year-olds but not from 9- to 11-year-olds, which could be explained by the discrepancy in developmental trajectory between auditory language and working memory. This pattern of age and modality interaction was similar across non-linguistic Chinese and English SL. A significant interaction between modality and language type also showed that better learning was achieved in visual SL as compared with auditory SL in both non-linguistic and English stimuli. However, children performed similarly across the two modalities in Chinese, possibly due to the contribution of tonal information. Together, our findings point to the joint function of age, modality, and language type in SL development.