Abstract: Cross-situational word learning deficits are prevalent in children with autism spectrum disorder (ASD). These deficits significantly stunt their language acquisition and affect the development of their communication and social interaction skills. Previous studies have examined the causes of these learning difficulties from various theoretical perspectives to develop diverse language intervention approaches. However, most studies have predominantly focused on the deficits of general cognitive abilities in children with ASD, such as the deficits of memory and attention. Little studies have examined the critical role of language-specific learning mechanisms. As a result, it is not clear what language-specific learning mechanisms contribute to the cross-situational word learning deficits in children with ASD.
To fill this gap, the present review systematically synthesizes the progress in research on cross-situational word learning in children with ASD and delves into the underlying causes of their learning difficulties by examining the intrinsic mechanisms of language acquisition. Specifically, the present review first summarizes recent advancements in word learning in children with ASD, highlighting that children with ASD have some specific impairments in cross-situational word learning, such as slower learning speed.
To better understand these specific impairments, the present review first distinguishes between implicit and explicit word-learning processes in word learning. After reviewing previous studies, the present review concluded that among children with ASD, implicit learning appears to be relatively intact, whereas explicit learning may be impaired. This divergence carries profound implications for uncovering how children with ASD acquire new words across different situations. Then, the present review focuses on the mechanisms of associative learning and hypothesis testing in cross-situational word learning, examining their relationships with implicit and explicit learning. Specifically, associative learning seems to engage implicit processes, while hypothesis testing relies on explicit learning. Since children with ASD demonstrate relatively intact implicit learning but impaired explicit learning, this study proposes that their associative learning remains preserved, while their hypothesis testing abilities are compromised.
To promote the cross-situational word learning ability of children with ASD, it is necessary to fully leverage their intact associative learning and promote their impaired hypothesis testing. Therefore, the present review investigates the factors influencing associative learning and hypothesis testing, particularly referent diversity and word frequency distribution. Referent diversity may help children with ASD strengthen word-object associations and make full use of their relatively intact association learning mechanism, thus improving both the effectiveness of cross-situational word learning and generalization. The Zipfian distribution is a long-tailed skewed distribution, which reinforces the use of the mutual exclusivity strategy. Therefore, the Zipfian distribution may help to improve the hypothesis testing of ASD children.
Based on the points discussed above, the present review proposes a novel hybrid synergistic model that integrates both implicit and explicit learning mechanisms in cross-situational word learning. This model posits that cross-situational word learning in children with ASD may be enhanced by simultaneously leveraging both implicit and explicit learning systems. For instance, implicit learning may benefit from repeated exposure to varied situations, while explicit learning may be promoted through structured and targeted interventions.
The hybrid synergistic model first reveals the cognitive mechanisms underlying cross-situational word learning in children with ASD and provides instructive implications for intervention. Specifically, the hybrid synergistic model proposes that enhancing referent diversity may improve associative learning, thereby improving the efficiency of cross-situational word learning in children with ASD. When children are exposed to diverse referents for a target word, the children are more likely to form robust and flexible word-referent associations. Additionally, the hybrid synergistic model recommends leveraging Zipfian distribution to optimize hypothesis testing mechanisms. This approach may facilitate the formation of accurate hypotheses about word meanings by providing children with frequent and predictable exposure to high-utility words, thereby mitigating explicit learning deficits.
In summary, the hybrid synergistic model not only advances the theoretical understanding of cross-situational word learning in children with ASD but also provides a valuable framework for developing innovative language intervention methods tailored to the unique learning profiles of children with ASD. By integrating insights from both implicit and explicit learning mechanisms, the hybrid synergistic model proposes a comprehensive approach to addressing the language acquisition challenges faced by children with ASD. Furthermore, the model highlights the importance of individual differences in learning styles and cognitive profiles when designing interventions. For example, some children with ASD may benefit more from implicit learning strategies, while others may require additional support for explicit learning processes. Tailoring interventions to each child's specific needs may maximize the effectiveness of interventions and help children with ASD achieve their full potential in language development.

Key words: autism spectrum disorder, cross-situational word learning, hypothesis testing, associative learning