Linguistic features of stuttering during spontaneous speech

Area of Science:

• Speech-language pathology
• Linguistics
• Psycholinguistics

Background:

• Previous research identified linguistic features influencing stuttering predictability in reading tasks.
• Spontaneous speech presents unique linguistic challenges compared to reading.

Purpose of the Study:

• To investigate the impact of linguistic features on stuttering predictability during spontaneous speech.
• To determine if linguistic features predictive in reading tasks are also predictive in spontaneous speech.

Main Methods:

• Analysis of interview data from 35 adult stutterers (27,009 words) from the FluencyBank database.
• Development of three logistic regression mixed-effects models incorporating various linguistic features (e.g., word length, frequency, phoneme, grammatical function, neighborhood density, word acquisition age, concreteness).
• Model comparison using the Area Under the Curve (AUC) statistic.

Main Results:

• The initial four-feature model identified initial phoneme, grammatical function, and word length as significant predictors.
• The six-feature model added word frequency and neighborhood density, with initial phoneme, word length, word frequency, and neighborhood density emerging as significant predictors.
• The nine-feature model did not offer improved predictive power over the six-feature model.

Conclusions:

• Linguistic features influencing stuttering in reading tasks are also predictive in spontaneous speech.
• Stuttering predictability is influenced by linguistic processing demands, with longer and less frequent words showing higher likelihood of disfluency.
• Findings underscore the role of psycholinguistic factors in stuttering during naturalistic communication.