The redundancy of phonemes in sentential context

Area of Science:

• Psycholinguistics
• Speech Perception
• Information Theory

Background:

• Printed English exhibits high redundancy, enabling readers to easily predict upcoming letters.
• Previous research has extrapolated this redundancy to speech perception without direct phonemic analysis.
• The redundancy of phonemes in connected speech remains underexplored.

Purpose of the Study:

• To directly assess the redundancy of phonemes in spoken English.
• To compare phonemic redundancy with the redundancy of printed characters.
• To investigate factors influencing phonemic predictability and model speech perception.

Main Methods:

• Participants engaged in a phoneme and letter prediction task using four unrelated English sentences.
• The study analyzed the predictability of phonemes and characters within the spoken and written sentences.
• Three computational models were tested to explain phonemic predictability.

Main Results:

• Phonemes in connected speech demonstrated significantly lower redundancy compared to printed letters.
• Word-initial and transitional phoneme probabilities best predicted phonemic guessing.
• Factors such as task difficulty, listener experience, and contextual information were considered.

Conclusions:

• Connected speech is less informationally redundant than printed text.
• Phonemic predictability in speech is influenced by linguistic probabilities.
• Findings have implications for understanding the efficiency and mechanisms of speech perception.