# Explaining the Comprehension–Production Vocabulary Gap Through Neural Networks and Cross‐Syndrome Evidence: Insights From Williams Syndrome

## TL;DR

Children with Williams syndrome show a smaller gap between understanding and using words compared to typically developing children, possibly due to visual processing challenges.

## Contribution

A computational model explains the unique vocabulary profile in Williams syndrome by linking visual processing constraints to lexical outcomes.

## Key findings

- Children with Williams syndrome have a significantly reduced comprehension–production vocabulary gap.
- A self-organizing map model reproduces WS lexical patterns through visual processing limitations.
- The model suggests exemplar-based categorization explains strong production with shallow semantics in WS.

## Abstract

The comprehension–production vocabulary gap is a well‐documented hallmark of language development; however, anecdotal evidence suggests that this asymmetry may be reduced in children with Williams syndrome (WS). Here, we use empirical data to characterise the comprehension–production gap and computational modelling to investigate potential mechanisms underlying this distinctive linguistic profile, focusing on children aged 7 months to 6 years. Using parental reports (Communicative Development Inventories), we measured the receptive and expressive vocabularies of children with WS (n = 67) and compared them to typically developing children (n = 1210) and cross‐syndrome groups with Down syndrome (n = 27), and fragile X syndrome (n = 15). Results confirm that children with WS show a unique trajectory: alongside general delay, they exhibit a significantly reduced comprehension–production asymmetry not observed in other groups. To elucidate the potential origins of this phenomenon, we implemented a biologically inspired neural network—self‐organising map (SOM)—to model early word learning and evaluate visual and auditory map representations. Our findings reveal that WS‐like vocabulary patterns can emerge from selective difficulties in visual processing, leading to exemplar‐based rather than prototype‐based object representations. The model suggests that these visual processing challenges, consistent with known visuospatial difficulties in WS, may contribute to the atypical comprehension–production relationship, while broader processing constraints may underlie general delays. This study provides a mechanistic account of vocabulary development in WS, highlighting the role of visual constraints in shaping lexical outcomes. More broadly, it underscores the need to conceptualise language development as an interaction between sensory input and cognitive subsystems, explaining why the comprehension–production gap is not a uniform feature of language acquisition.

Cross‐sectional data demonstrate a reduced lexical comprehension–production vocabulary gap specific to Williams syndrome, not observed in other populations.We present a computational model that reproduces the Williams syndrome lexical profile by combining domain‐general and visual‐specific processing constraints.The computational model explains how relatively strong lexical production can coexist with shallow semantic processing through exemplar‐based categorisation in Williams syndrome.This computational model links sensory processing, categorisation and lexical outcomes within a unified developmental framework.

Cross‐sectional data demonstrate a reduced lexical comprehension–production vocabulary gap specific to Williams syndrome, not observed in other populations.

We present a computational model that reproduces the Williams syndrome lexical profile by combining domain‐general and visual‐specific processing constraints.

The computational model explains how relatively strong lexical production can coexist with shallow semantic processing through exemplar‐based categorisation in Williams syndrome.

This computational model links sensory processing, categorisation and lexical outcomes within a unified developmental framework.