Affective Dimensions in Maternal Voice During Child Feeding in Mothers With and Without Eating Disorder History—Findings From a Machine Learning Analysis of Speech Data

Comprehensive Summary

This study by Throm et al. investigates whether maternal vocal affective characteristics during mother–child feeding interactions differ between mothers with a history of eating disorders (ED) and those without, by using machine learning to analyze speech data. To do so, the researchers filmed home mealtime interactions between mothers and their 10-month-old children, comparing 17 mothers with a past ED to 27 without ED history. They tested several machine learning models to analyze vocal features related to emotional dimensions of speech during feeding such as arousal, valence, dominance. They then identified consistent patterns that distinguish the two, and found that across the mealtime interaction, voices of mothers with an ED history consistently showed higher expressions of these affective dimensions compared with mothers without ED history. These differences were also especially pronounced in the middle of the feeding session, suggesting changing emotional engagement rather than a constant pattern. As such, the authors conclude that mothers with a history of EDs may be more emotionally engaged or expressive during feeding interactions with their infants than those without. They argue this points to meaningful differences in mother to child communication and emphasize the importance of further research on vocal affect and its role in eating behavior. The study also highlights that machine learning techniques can detect subtler nuances in speech that traditional questionnaires might miss.

Outcomes and Implications

This research is important because it explores how emotional communication is expressed vocally during feeding, which may contribute to developmental and psychological outcomes. Early differences in affective communication could inform risk identification and early intervention in families affected by eating disorders. Clinically, understanding affective speech patterns could help clinicians tailor parent–infant support and therapy, especially with mothers with ED history. While machine learning analysis of speech is still only a research tool rather than a standard clinical method, these approaches may eventually contribute to supportive tools to help early detection of communication differences that could positively impact child development.