# Infant word learning and word recognition under realistic phonetic conditions

> **NIH NIH F31** · UNIVERSITY OF PENNSYLVANIA · 2024 · $47,694

## Abstract

PROJECT SUMMARY/ABSTRACT
Language learning is a critical aspect of children’s development, laying the groundwork for future social and
academic success. Understanding how language learning begins in infancy has long been a key project of
developmental psychology and cognitive science. However, current theories of language acquisition make
unrealistic assumptions about the nature of speech, infants’ entry point to their future native language. Current
accounts assume that words in speech can be translated into consistent phonological (sound) representations
by infants and adults alike. In reality, how a speaker pronounces a given word is fundamentally variable, so
much so that many pronunciations, even in infant-directed speech, are unintelligible on their own and can only
be understood by adults because adults can leverage their top-down linguistic knowledge about the
surrounding sounds and words. The proposed studies represent one of the first forays into how infants manage
this fundamental pronunciation variability while they are learning the system itself.
Using a multimethod approach, the proposed research investigates the hypothesis that infants rely on temporal
context to interpret reduced pronunciation variants. In normal discourse, including speech to children, first uses
of words are often more clearly spoken than following uses. If infants can connect clear uses and closely-
following reduced forms, this could help them recognize instances of words that would be unintelligible on their
own. Characterizing this process will help make it possible to create better models of how infants begin to
connect the speech signal to meaning.
Study 1, a corpus analysis, analyzes the language input that parents provide by asking how often parents
produce reduced pronunciation variants in different contexts and how these contexts line up with ideal word-
learning opportunities. Studies 2 and 3 use eye-tracking procedures to experimentally investigate infants’ initial
encoding of novel words and recognition of familiar words given naturalistic pronunciation variability. Together,
these three studies will help guide scientific theories of word learning by providing new insight into this
unsolved problem in children’s language acquisition and demonstrating powerful new methods for future
research on how young children handle the messy reality of spoken language.
The proposed research will provide the applicant with the ideal training to advance her career, empowering her
to expand her previous work into a full research program, learn new computational and experimental methods,
and gain critical professional development experience communicating her findings and mentoring students.
These training goals are further supported by a strong institutional environment with many pertinent student
resources and intellectual communities. This training will ultimately prepare the applicant to become a
successful independent researcher.

## Key facts

- **NIH application ID:** 10825050
- **Project number:** 1F31HD113235-01A1
- **Recipient organization:** UNIVERSITY OF PENNSYLVANIA
- **Principal Investigator:** Caroline Wendell Beech
- **Activity code:** F31 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $47,694
- **Award type:** 1
- **Project period:** 2024-01-01 → 2025-12-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10825050

## Citation

> US National Institutes of Health, RePORTER application 10825050, Infant word learning and word recognition under realistic phonetic conditions (1F31HD113235-01A1). Retrieved via AI Analytics 2026-05-28 from https://api.ai-analytics.org/grant/nih/10825050. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*