Speech Perception

Continuous Nature of Speech

Continuous Nature of Speech

Speech lacks clear boundaries between words, making it challenging to identify individual words compared to written text.

Variability in Speech Sounds

Differences in speech due to co-articulation, where sounds are influenced by surrounding phonemes, and speaker differences like accent, gender, and speaking rate.

Co-articulation

A phenomenon where the articulation of a speech sound is influenced by the sounds before and after it (e.g., "the" changes based on its context).

Speaker Differences

Variations in speech caused by factors such as accent, gender, and speaking rate, complicating perception.

Time Constraints in Speech Processing

Humans process up to 200 words per minute, with the fleeting nature of sound creating a "now-or-never bottleneck" for processing (Christiansen & Chater, 2016).

Phonemes

The smallest units of sound in speech that distinguish meaning, such as /p/ in "pin" versus /b/ in "bin." They differ from letters (e.g., /k/ applies to "cat" and "kite").

Source-Filter Theory of Speech Production

Explains speech production:

Source: Vocal cords generate vibrations, determining pitch and intonation.

Filter: Shapes sound through the supralaryngeal vocal tract (lips, tongue, teeth, etc.), producing distinct speech sounds.

Spectrograms

Visual representations of sound showing amplitude over time and frequency. Key features include formants, which are energy bands shaped by the vocal tract critical for speech intelligibility.

Categorical Perception

Acoustic changes in speech are perceived as discrete categories (e.g., /ba/ vs. /da/). Key features include abrupt changes at phoneme boundaries and heightened discrimination at those boundaries.

McGurk Effect

A visual-auditory illusion where seeing one phoneme (e.g., /ga/) and hearing another (e.g., /ba/) results in perceiving a third phoneme (e.g., /da/).

Ganong Effect

Lexical knowledge biases phoneme perception. For example, an ambiguous sound between "gift" and "kift" is more likely to be interpreted as "gift" because it is a familiar word.

Motor Theory of Speech Perception

Suggests that speech perception is based on identifying intended vocal gestures rather than acoustic signals. Supported by studies showing motor area activation during speech perception.

Dual Streams Model of Speech Perception

Ventral Stream: Processes word recognition and meaning.

Dorsal Stream: Links speech perception with production and is critical for phoneme discrimination.

Ventral Stream

Processes word recognition and meaning.

Dorsal Stream

Links speech perception with production and is critical for phoneme discrimination.

Cohort Model of Word Recognition

Words are progressively recognized as they unfold. Recognition occurs at the "uniqueness point," where only one word matches the input.

TRACE Model of Speech Perception

A connectionist framework where phonemes, syllables, and words interact. Higher-level word recognition influences lower-level phoneme perception via feedback loops.