Abstract
This paper investigates the role of prosody in one child's lexical acquisition using an ecologically valid, high-density, longitudinal corpus. The corpus consists of high fidelity recordings collected from microphones embedded throughout the home of a family with a young child. We analyze data collected continuously from ages 9-24 months, including the child's first productive use of language at about 11 months and ending at the child's active use of more than 500 words. We found significant correlations between prosody of caregivers' speech and age of acquisition for individual words.
Introduction
Language acquisition is one of the most complex feats of human development, yet the mechanisms driving the timing of word learning remain partially understood. While child-directed speech (CDS), or "motherese," is known for its exaggerated pitch and rhythm, its direct impact on the Age of Acquisition (AoA) for specific vocabulary has primarily been studied in controlled laboratory settings. This study addresses this gap by examining prosody—the melody, stress, and rhythm of speech—within a naturalistic home environment to determine if acoustic emphasis serves as a predictor for lexical development.
Methods and Data
The study utilizes the Human Speechome Project (HSP) corpus, a high-density longitudinal dataset recorded in a single family's home. The analysis covers the period from 9 to 24 months of age, capturing approximately 12 million words of input. A subset of 400,000 hand-transcribed words was analyzed, focusing on 461 words that appeared frequently enough in caregiver speech to provide reliable prosodic measurement.
Three primary prosodic features were extracted:
- Duration: Standardized vowel length within words.
- Fundamental Frequency (F0): A composite measure of mean pitch relative to the sentence and pitch range within the word.
- Intensity: Relative loudness, calculated using the PRAAT software system.
The Age of Acquisition (AoA) was defined as the "word birth," or the first recorded productive use of a word by the child.
Results
Linear regression analysis revealed significant negative correlations between all three prosodic variables and the AoA, suggesting that words spoken with greater emphasis are acquired earlier.
- Duration: $r = -0.29$
- F0 (Pitch): $r = -0.19$
- Intensity (Loudness): $r = -0.35$
The strongest predictive power was found when combining all three features into a single model, resulting in a correlation of $r = -0.44$ ($p < .0001$). This combined model demonstrates that duration, pitch, and intensity provide complementary acoustic cues that facilitate word segmentation and prioritization for the infant.
Discussion
The findings suggest that prosodic emphasis acts as an "acoustic spotlight," helping infants solve the segmentation problem by highlighting key words in the continuous speech stream. Notably, intensity (loudness) proved to be a more robust predictor than pitch, potentially due to its salience in a noisy home environment.
However, the relationship is not uniform across all word types; adjectives showed the strongest link to prosody, while verbs showed the weakest, likely due to their higher conceptual complexity. Furthermore, social and emotional factors can override acoustic cues, as seen with the early acquisition of "dad," which occurred much sooner than prosodic models predicted.
Conclusion and Future Work
This study provides evidence that caregiver prosody significantly influences the timeline of lexical acquisition in a naturalistic setting. Future research will explore multi-directional models to account for the feedback loop between caregiver and child, as well as incorporate additional variables such as word frequency and situational context.
Summary updated