Lecture Four Orienting Question

1
Lecture Four - Orienting Question

• Do the effects of altered auditory feedback in
  fluent speakers support a feedback monitoring
  process?

2
Definitions

• Monitor - a device that takes the output of a
  process, compares this with what it intended (its
  input) and if there is a discrepancy (an error)
  makes an adjustment to reduce the error
• Continuity with what we said in previous lecture
  (speech repairs) about monitor and
  error-correction processes.
• Monitoring processes considered equivalent to a
  servo-control system

3
Definitions

• Feedback - a route by which information is
  returned to the monitor
• Feedback carries the connotation that it is
  involved in a monitoring process. Experimental
  alterations are referred to as altered auditory
  feedback. This is misleading as experimental
  alterations could also affect speech control
  without affecting a monitoring process.

4
This lecture is organised in terms of Case for
the importance of auditory feedback in speech
control, Case against.
5
Auditory feedback monitoring

• Speak (input to the production system), listen to
  the speech (output) - is output as intended. If
  OK, continue - if not make a correction
• Interference with what is heard should affect
  this control process
• Three types of interference
• Time
• Intensity
• Frequency

6
Two principal ways of changing the intensity of
the voice

• Directly (amplify or attenuate it) by
  increasing/decreasing masking noise
• Direct changes lead to compensatory changes in
  intensity
• If amplify, speaker reduces
• If attenuate, speaker increases (Lombard effect)
• Noise has reverse effects
• If amplify noise, speaker increases voice level
• If attenuate noise, speaker decreases voice level

7
Frequency alterations

• Filtering, Frequency shifting
• Evidence on fluent speakers shows compensation
• e.g Elman (1981) looked at the effect of
  frequency shifts on voice fundamental frequency.
  Reported that speakers shift voice fundamental
  frequency in opposite direction to the shift
• Not studied to any great extent in fluent
  speakers
• Compensation effects in the cases of intensity
  and frequency shift regarded as evidnece for
  monitoring processes

8
Effects of changes to timing

• Lee, Black, Fairbanks
• Fairbanks - primary, secondary
• Disturbance function - axes
• Independent variable - length of delay
• Primary dependant variables - time, errors
• Secondary dependant variables - intensity, pitch

9
Explanation of primary effects

• Units issued to the production system
• These units retrieved through listening
• Maximum mismatch (and, therefore, maximum
  discrepancy, maximum alteration, maximum
  disruption)
• If so, delay that produces maximum disruption
  tells us what units are retrieved
• Maximum disruption at a delay of 200ms, syllables
  are about this length, they are the unit of
  control (Black)

10
DAF as a technique for producing asynchronous
input to the timekeeper that disrupts speech.
Howell and Archer (1984) transformed speech into
a noise that had the same temporal structure as
speech, but none of the phonetic content.
Predictions 1. Perceptual monitoring,
manipulations that destroy speech content should
make it unusable for detecting errors and, by
extension, remove the effects of DAF. 2.
Rhythmic disruption - DAF affects control because
it creates disruptive rhythmic input to a
timekeeper, the Howell and Archer noises should
still disrupt the timekeeper.
11
DAF as a technique for producing asynchronous
input to the timekeeper that disrupts speech.
Howell and Archer (1984) compared DAF with a
condition in which speech was delayed by the same
amount of time and then converted to noise. The
two conditions produced equivalent disruption
over a range of delays. This suggests that the
DAF signal does not need to be a speech sound to
affect control in the way observed under DAF, and
indicates that speech does not go through the
speech comprehension system and then to a
monitor.
12
Evidence that DAF affects external timekeeper
variability, not motor variability
Noise creates similar effects to speech under DAF
conditions. This does not necessarily support the
view that DAF inputs to a timekeeper. A
modification of the Wing and Kristofferson (1973)
task was performed under DAF conditions to see
whether DAF selectively affects the timekeeper
(Howell Sackin, 2002). Subjects in the standard
Wing and Kristofferson procedure are required
to produce a series of taps at a specified rate
as accurately timed as possible. Modification
was that speech responses (the syllable /bae/)
was used instead of a tap.
13
Evidence that DAF affects external timekeeper
variability, not motor variability
The variability in the sequence of responses can
be decomposed into variance of motor processes
(Mv) and variance of a timekeeper (Cv). The
essence of the analysis procedure is that if the
motor system leads to a tap being placed at the
wrong point in time, it is compensated in the
next interval. Thus Mv can be estimated from the
lag one autocovariance. Cv can then be estimated
by subtracting Mv from the total variance.
14
Evidence that DAF affects external timekeeper
variability, not motor variability
Howell and Sackin (2002) had subjects perform
this task in conditions where they heard
concurrent DAF. DAF led to a marked increase in
Cv, the Cv increase being greater for longer
DAF-delays. Mv, on the other hand, stayed roughly
constant across DAF delay and repetition period.
Supports the idea that AAF has a direct
influence on the timekeeper and has less effect
on the motor processes.
15
Cerebellar control and support from the Wing and
Kristofferson task (organisational factor).
The Wing and Kristofferson (1973) task has been
used with patients who have a lesion in the
cerebellum to see whether (and if so, which)
regions of the cerebellum are involved with Mv
and Cv (Ivry, 1997). Lateral lesions of the
cerebellum affect timing control, suggesting that
the timekeeper mechanism is located in this part
of the CNS (Ivry, 1997). The medial areas of
the cerebellum appear to be involved with Mv, as
lesions to this part affect this variance
component. The location of areas in the
cerebellum responsible for Mv and Cv support the
idea, made in EXPLAN, that AAF disrupts
mechanisms involved in organizing speech output
for execution, rather than higher level cognitive
planning mechanisms. There are data on the Wing
and Kristofferson task that support the view that
speakers who stutter have problems in cerebellar
organization (Howell et al., 1997), in particular
in the Mv component in children who stutter.
16
Other problems for the view that auditory
feedback is used for perceptual monitoring
General1. Howell (2002) has argued that the
amount of phonetic information a speaker can
recover about vocal output is limited because
bone-conducted sound interferes with that sound
(see Howell and Powell, 1984 for a study on this
issue). This would limit the usefulness of the
feedback that a speaker can recover by listening
to his or her own voice, making it an unlikely
source of information for use for feedback
control. 2. How quickly information can be
recovered from the auditory signal and why
speakers with hearing impairment who have
established language before the loss can continue
to speak. The former suggests that feedback would
be too slow to use in feedback monitoring and
that speech can proceed without this information
(be under open loop control).
17
Other problems for the view that auditory
feedback is used for perceptual monitoring
DAF in particular 1. Howell and Archer (1984)
work, described above, suggests that the DAF
effect does not depend on whether the delayed
sound is speech or a noise. This undermines use
of this line of support for monitoring theories.
2. Another argument against DAF supporting
feedback control of the voice, is based on
whether it shows a Lombard or a Fletcher effect.
The Fletcher effect occurs when speech level is
increased (the speaker raises voice level when
speech level is experimentally decreased and
decreases voice level when speech level is
experimentally increased). The latter is opposite
to what happens when noise level is changed
(Lombard effect), where speakers raise voice
level when noise level is increased (Lane and
Tranel, 1971). .If the delayed speech under DAF
is treated like a persons own speech and is
processed by the speech comprehension system, it
should lead to a Fletcher, rather than a Lombard,
effect. Howell (1990) reported that amplifying
the delayed sound during DAF produces a Lombard
effect, again suggesting that the delayed sound
is treated as a noise rather than speech.
18
References Borden, G. J. (1979). An
interpretation of research on feedback
interruption in speech. Brain Language, 7,
307-319. Howell, P. (2002). The EXPLAN theory of
fluency control applied to the treatment of
stuttering by altered feedback and operant
procedures. In Pathology and therapy of speech
disorders. Pp. 95-118. E. Fava (Ed.). Amsterdam
John Benjamins. Howell, P., Powell, D. J.,
Khan, I. (1983). Amplitude contour of the
delayed signal and interference in delayed
auditory feedback tasks. Journal of Experimental
Psychology Human Perception and Performance, 9,
772-784. Kalinowski, J. S., Satuklaroglu, K.
(2005). Stuttering. San Diego Plural
Publishing, pp 350