Summary: A comprehensive meta-analysis confirms that adults commonly exaggerate vowel sounds when speaking to infants—an effect known as vowel hyperarticulation in infant-directed speech (IDS). By synthesizing results from 55 studies spanning at least 10 languages, researchers found reliable evidence for vowel exaggeration in IDS overall, while also identifying substantial variation related to language, measurement approach, and sample size.
The pooled findings support the idea that vowel exaggeration could help infants perceive speech more clearly and support early language learning. At the same time, the review highlights methodological inconsistencies and generally small sample sizes across studies, which limit broad generalizations and point to the need for larger, more standardized cross-cultural research.
Key facts
- Cross-linguistic evidence: Evidence of vowel hyperarticulation in IDS appears in studies of at least 10 languages.
- Meta-analytic breadth: The review pooled data from 55 studies to increase statistical power and clarify mixed findings in the literature.
- Methodological sensitivity: Results depend on analytic choices—such as acoustic measures (Hertz, Mel), sample size, and study design—so comparisons across studies require careful attention to method.
Source: University of Tokyo
Infant-directed speech (IDS), often called “baby talk,” differs from adult-directed speech (ADS) in several acoustic and interactional ways, including higher pitch, exaggerated intonation, simplified vocabulary, and sometimes clearer articulation. One specific debated feature is vowel hyperarticulation—the tendency to produce vowels with more extreme acoustic targets, which may make vowel categories easier for infants to discriminate.
Previous individual studies produced mixed conclusions: some reported clear exaggeration of vowels such as “a,” “i,” and “u” in IDS—interpreted as clearer exemplars that could aid infant perception and vocabulary learning—while others failed to detect consistent hyperarticulation, suggesting that acoustic changes might instead arise from smiling, affect, or other interactional factors.
To resolve these discrepancies, researchers at the International Research Center for Neurointelligence (WPI-IRCN) at The University of Tokyo, together with collaborators at the University of Amsterdam, PSL University, and Aarhus University, conducted a systematic review and multi-method meta-analysis. Their paper was published on 02-06-2025 in Psychological Bulletin.
The team applied four complementary meta-analytic approaches to quantify vowel hyperarticulation. They first analyzed a subset of 20 studies that used similar methodologies (yielding 42 effect sizes) and then expanded the analysis to include 35 studies across diverse methods (80 effect sizes). Analyses included both frequentist and Bayesian techniques to evaluate the robustness of results across analytic choices.
Across analyses, the authors found consistent evidence that caregivers exaggerate vowels in IDS compared with ADS, with effect sizes typically in the small-to-moderate range. At the same time, they documented systematic and unsystematic variability: cross-linguistic differences, divergent measurement scales (e.g., Hertz vs. Mel), and varying sample sizes all contributed to heterogeneity and made it difficult to isolate single factors that predict stronger hyperarticulation.
A prominent methodological concern identified by the review is underpowered studies. Many IDS vowel studies had smaller-than-ideal sample sizes, limiting confidence in individual results and hampering efforts to detect language-specific patterns. The authors therefore recommend larger study samples, clearer reporting of methods, and explicit modeling of potential sources of heterogeneity when designing and interpreting future research.
Following confirmation of vowel hyperarticulation as a recurring feature of IDS, the research group plans to investigate other environmental and interactional factors that shape infant language development. Key questions include how the quantity and quality of caregiver speech, and other modifiable aspects of infants’ immediate environment, influence learning trajectories.
“How infants learn language with remarkable speed and apparent ease remains a central puzzle,” said Irena Lovčević, postdoctoral researcher at WPI-IRCN and first author of the study. “Our goal is to clarify which caregiver behaviors reliably support perception and learning, and to improve methodological standards so findings can be compared and generalized across languages and contexts.”
Contributors to the study include Titia Benders (University of Amsterdam), Sho Tsuji (WPI-IRCN, The University of Tokyo and PSL University), and Riccardo Fusaroli (Aarhus University). The work received support from a MEXT World Premier International Research Center Initiative startup grant, JSPS KAKENHI Grant Numbers 20H05617 and 20H05919, and seed funding from the Interacting Minds Center.
About this language and learning research news
Author: Kazuyo Okada
Source: University of Tokyo
Contact: Kazuyo Okada – University of Tokyo
Image: The image is credited to Neuroscience News
Original research: Closed access. “Acoustic Exaggeration of Vowels in Infant-Directed Speech: A Multi-Method Meta-Analytic Review” by Irena Lovčević et al., Psychological Bulletin.
Abstract
Acoustic Exaggeration of Vowels in Infant-Directed Speech: A Multi-Method Meta-Analytic Review
There is an ongoing debate about the prevalence and magnitude of vowel hyperarticulation—acoustically exaggerated vowel production—in infant-directed speech (IDS). Proponents argue that exaggeration produces clearer speech signals that facilitate infant perception and may relate positively to early language outcomes. However, prior findings are inconsistent, with some studies reporting clear hyperarticulation and others finding little or no exaggeration.
To assess the robustness of vowel hyperarticulation in IDS and to identify sources of variability across studies, the authors conducted a systematic review and applied multiple meta-analytic strategies. They combined frequentist and Bayesian approaches, first focusing on studies with comparable methods and then including all eligible studies regardless of measurement approach.
Results indicate a reliable tendency for vowel hyperarticulation in IDS compared to adult-directed speech, with effect sizes in the small-to-moderate range (approximately 0.41 to 0.69 across analyses). At the same time, heterogeneity attributable to cross-linguistic differences, measurement scales, and study design underlines the need for larger samples, standardized methods, and explicit treatment of heterogeneity in future research. The authors conclude by providing methodological recommendations to strengthen the field and improve the interpretability and generalizability of findings about IDS and early language development.