Is speech special

1
Is speech special?

• speech is a complex auditory pattern, how is it
  transformed into linguistic meaning?
• is there a special speech mechanism or is speech
  served by the same auditory analyzers used for
  other complex auditory patterns, i.e. a general
  mechanism

2
arguments for two mechanisms special
general

• a speech module
• innate
• deciphers spectrograms
• provides linguistic reality
• unique to humans
• clinical/experimental evidence of left hemisphere
  processing

• considers speech no different, in principle, than
  any other complex pattern
• human language not special
• speech and nonspeech can be complex
• much of speech is learned

3
What are the implications of each view for.?

• nonhearing infants
• cognitively handicapped persons
• universality of speech
• animal language
• evolutionary theory

4
How to resolve?

• are there fundamental differences between speech
  and nonspeech in
• temporal sequencing
• use of symbols
• categorical perception
• duplex perception
• cross modality perception
• clinical localization of function studies

5
1. Detecting order in sound sequences a special
mechanism

• Speech
• 80 msec/phoneme yields a rate of 12 phonemes per
  second
• can be increased up to 50 phonemes per second and
  be intelligible
• conclude the phoneme is embedded or encoded in a
  larger unit and requires decoding.Provides
  support for a special mechanism

• Nonspeech
• consider a musical note equivalent to a phoneme
• assume that a normal tempo in music is about 80
  beats per minute or 1.3 notes per second
• if tempo is increased , eg. to 8 1/32 notes per
  beat, and 80 beats per minute, the musical rate
  would be 11 notes per second and would be very
  fast
• music played at the rate of fast speech, 50 notes
  per second would be unintelligible

6
2. Use of symbols

• in speech there is an arbitrary symbolic
  relationship between an sound and its meaning
• eg. rose ?
• borborygmous?
• the stimulus is not the mouthing and the sound,
  it is the meaning
• the difference provides support for a special
  mechanism

• nonspeech auditory stimuli have a logical
  connection between the sound and the auditory
  event
• eg. dogs bark
• cars crash
• the stimulus conveys reality, no support for
  general mechanism

7
3. Categorical perception

• discrimination of speech sounds is related to the
  categories which identify the sounds
• eg. /pa/ and /ba/ represent two categories of
  speech
• /b/ is a voiced bilabial stop consonant
• /p/ is a unvoiced bilabial stop consonant
• voice onset time (VOT) can be varied so that /b/
  gradually becomes /p/
• range of acoustic differences created by varying
  VOT are not perceived, only the category of /ba/
  or /pa/

8
categorical perception (contd).

• /c/ /f/ /t/ /p/ /s/ vocal chords vibrate
  late
• voice onset time (VOT)
• /g/ /v/ /d/ /b/ /z/ vocal chords vibrate
  early

9
categorical perception

• VOT
• /b/ 0 10 20 30 40 50 60 70 80 90 /p/

classified as /p/
phoneme boundary at 50 msec
50
VOT
40 50 60
10
conclude

• in categorical perception there is no
  discrimination of the acoustic differences
  between /b/ and /p/, a difference of 50 msec VOT
• in a nonspeech context, a difference of 50msec is
  discriminated, showing continuous perception

11
discrimination and identification curves
12
categorical perception of a gradually changing
set of stimuli
13
What this means?

• speech sounds (particularly consonants) are
  encoded into speech and decoding by a special
  mechanism is necessary
• the special mechanism is in the left hemisphere
• vowels are not necessarily encoded if presented
  alone, explaining their continuous perception

14
Is there counter evidence for a general mechanism?

• Yes
• animals can be trained to perceive categorically
  and they dont speak
• humans can be trained to show categorical
  perception of nonspeech by manipulating
• task structure - eg. practice
• instructions
• no conclusive evidence for a special mechanism

15
4. Duplex perception

• the setup experiments make use of stimuli in
  which the direction of F3 transitions distinguish
  da from ga
• without this transition, the rest of the stimulus
  pattern is ambiguous between da and ga

16
(No Transcript)
17
Procedure and Results

• the F3 transition is presented to the left ear
  and the rest of the signal, the base is presented
  to the right ear
• listeners hear da in the ear that gets the
  base, and hear a nonspeech chirp in the ear
  that gets the F3 transition
• the fact that the same stimulus is simultaneously
  part of two quite distinct percepts argues for
  two kinds of perception of auditory stimuli - a
  special mechanism for speech, distinct from the
  mechanisms which process non speech.

18
Counter evidence

• Fowler and Rosenblem (1990)
• - split a recording of a metal door slamming
  into a high frequency component, a rattle, at the
  beginning or the sound (like the F3 transiton)
  and a low frequency component at the end of the
  sound (like the base)
• - when the component sounds were presented to
  different ears, subjects reported hearing both
  the metal door slamming and the rattle
• - duplex perception occurs with nonspeech
  stimuli, arguing for a general mechanism

19
5. Cross modality perception

• the McGurk effect is an auditory illusion
  produced by a visual experience
• it is the subject of the second lab and is
  presented in a handout

20
6. Brain localization of speech

• language is lateralized to the left hemisphere in
  clinical studies of aphasia
• damage to the frontal and temporal areas of the
  left hemisphere produce difficulties in speaking
  and comprehending speech
• nonspeech environmental sounds such as melody,
  pitch are processed in the right
• evidence for a special mechanism.

21
Conclusions

• is the debate between a special or a general
  mechanism for speech resolved, no.
• as nonspeech sounds are made more complex and
  subjects must discriminate distinctions on the
  same order of speech (eg 50 msec) then the
  mechanisms for processing resemble those of speech