Speech perception

• Speech


  Variable - Every word takes on a different acoustic shape each time it is uttered
• Factors that cause speech to be variable


  • Speaker (vocal track size, regional accent, socio-economical tier)
  • Articulation rate (4/5 syllables/sec in sentences)
  • Prosody (the music of speech: rhythm, melody, amplitude)
  • Mode (voiced, whispered, creeky, etc.)
  • Coarticulation (individual phonemes influenced by preceding/upcoming segments)
• Quasi-continuous

  No unique/systematic way to flag word boundaries (rarely silence between two words)
• Lexically ambiguous

  Words are made out of a limited number of sounds and syllables, so embedded words are everywhere inside other words
• Audiovisual

  Visual information given by the lips and adjacent areas of the face about articulation is integral to speech perception, when available
• McGurk and McDonald's (1976) fusions: /ga/ (vision) + /ba/ (audition) = /da/ (perception)
• Phonemes


  Building blocks of the vocabulary, smallest units in the signal allowing meaning distinction, limited in number so words are created by combining them in an unlimited number of ways, specific to the language
• Supra-phonemic information


  The prosody/music of speech (rhythm, melody and energy), e.g. lexical stress/accentuation, tones
• We perceive speech as 'mental categories'
• With exposure to the language, non-native contrasts disappear, but native contrasts are maintained
• Directionality of lexical access


  Auditory memories for words "open up" only if their initial sound is perceived, "left-to-right" processing, importance of the 'Uniqueness Point'
• Active competition between words


  Each phoneme in the input can only belong to one word at a time, so sound-overlapping memories are enemies of one another
• Interactivity


  • Info travels both ways between speech perception & word memories passed to next level before current level has finished computation
• Ganong effect


  An ambiguous consonant is 'perceived' in accordance with the lexical context
• Lexical solutions for segmenting speech


  Word offset anticipation, lateral inhibition between word memories