Categorization Consistency: A Fundamental Dimension of Individual Variation in Speech Perception

Thousands of times a day, people hear speech and implicitly know whether what they heard was a consonant like “b” or a “p” or a vowel like “ee”. This act feels automatic and simple, but cognitive science suggests it is enormously challenging. Every talker says these same sounds a little differently, and they sound different in different contexts.

Thus, the auditory system must do a lot of work just to identify these basic sounds. This is particularly important in the context of learning language and reading. Children can’t recognize a word like “beach” if they can’t identify the “b’s” and “ee’s” and “ch’s” that comprise it.

More importantly, when learning to read, children must learn the complex mapping between letters and these auditory categories—that have to learn that a letter string like EE, EA or Y can make the “ee” sound. If the auditory process is not working effectively, learning these crucial phonics skills will be hard. Supporting these points, much prior research has linked differences in how listeners categorize speech to real-world concerns including development, language and reading disorders, second language learning, and aging.

This project builds on previous work in significant and novel ways. The study team’s prior work has led to a simple new way to assess these skills in which people hear sounds that have been morphed from one to another and rate them on a continuous scale. This new task has revealed a surprising new dimension to the problem of individual difference in speech perception which is, the degree to which a people’s percept is stable across multiple encounters with the same sound, or consistency.

Consistency changes with development. It is linked to both reading and language disorders. It predicts 30% of the variance in overall language ability in adults.

One part of the project investigates the nature of categorization consistency, by asking whether it reflect general cognitive tendencies, noise in the auditory system, or something specific to speech, looking at both children and adults and relating consistency to real-world variation in language and reading ability. Another part of the project develops a new eye-tracking paradigm and asks if the language system has “clean-up” mechanisms that leverage higher level knowledge to enhance consistency to cope with noise in the auditory system and asks whether variability in clean-up mechanisms explain differences in language and reading ability.

The broader impacts of the project are to develop real-world interventions that apply this knowledge to reading disorders. As a first step towards this eventual translation, the project focuses on bridging the gap between cognitive science and the classroom, a key issue in the public discussion over the “Science of Reading”. The study team, which includes cognitive scientists and education researchers, are teaching a semester-long workshop intended for students from both disciplines.

In this class, students design a new assessment of categorization consistency – consulting with private sector partners – and field-test it in local schools to gather real-world data, as well as teacher and student insights. This activity will pave the way for a potential real-world application of the basic research.

This project is jointly funded by the Perception, Action and Cognition (PAC) Program and the Established Program to Stimulate Competitive Research (EPSCoR).

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.