Docstoc

Lynn Lim - AphDay130905

Document Sample
Lynn Lim - AphDay130905 Powered By Docstoc
					An Investigation of the Effect of Multiple Productions
   on the Single Word Production of People with
  Acquired Speech Sound Production Difficulties:
               An Analysis of 2 Cases




Lynn Li Lim1,2, Karen Croot1,2,3, Sallyanne Palethorpe1,2, Max Coltheart1,2

   1. Macquarie Centre for Cognitive Science (MACCS), Macquarie University
   2. Speech Hearing and Language Research Centre (SHLRC), Macquarie University
   3. School of Psychology, The University of Sydney
Acoustic Analysis
 Speech disorders may share similar
  symptoms of impaired speech production
 To determine the speech disorder associated
  with the symptom  examine speech
  segment productions acoustically
 Gives more concrete information on speech
  qualities attributed to phonological &
  articulatory deficits
 More reliable than impressionistic phonetic
  transcription (Haley et al, 2001)
Acoustic Speech Analysis
   Most common acoustic investigations are on
    contrastive features to determine if speech
    segment errors are:

   Phonemic – substitution of one phoneme for
    another

   Phonetic – impaired articulation of a speech
    sound
    (Baum et al, 1990; Tuller, 1984)
Variability in Speech Production
 Unimpaired speech production can be
  somewhat variable (Auzou et al, 2000)
 Acoustic studies usually rely on multiple
  productions of speech tokens to overcome
  this variability
 People with speech disorders are far more
  variable in their productions than controls
    (Ryalls, 1986)
 But disordered speech analyses have not
  addressed the issue of possible effects of
  multiple productions
 And their results have been inconsistent
Multiple Production
   Not known if eliciting multiple word
    productions from these people affects their
    speech sound production difficulties

 H1: Fatigue – deterioration of speech
                 production
 H2: Practice – improvement in speech
                 production
 H0: No effect on speech production
Previous Studies
   Some examples for Vowel Duration
Study:        Ryalls (1986)      Tuller (1984)      Williams & Seaver
                                                    (1986)
Speakers      5 N/FL             5 N/FL             7 N/FL
              7 FL               5 FL               14 FL

Repetitions 5                    16                 1
per word

Mean          N/FL & FL > CTRL   N/FL & FL > CTRL   2 N/FL > Others
Durations

Variability   N/FL > FL & CTRL   N/FL & FL > CTRL   N/FL & FL < CTRL
                                 (Only some)
Multiple Production
Implication:
 If there are effects, then method of eliciting
  speech tokens may confound the
  investigation of the nature of the disorder
Other Features of Speech
Production
 Previous acoustic studies address question
  of whether speakers produce phonemic or
  phonetic errors
 Other information about speech production
  difficulties that are non-contrastive in nature
 These are not usually reported in most
  acoustic studies
Research Questions
Q1: Does multiple repetition of target words
    affect phonetic parameters of the speech
    of people with impaired speech sound
    production?

Q2: Are there other phonetic parameters in the
    speech of these individuals that might be
    indicative of the nature of the disorder but
    not usually reported?
Speakers
    AR
   63 y.o., male
   progressive aphasia with other mild cognitive
    deficits 2.5 years after presentation
   impaired syntax & phoneme discrimination
   semantic abilities just below control range
   moderate hearing loss
    (35-50 dB loss at 4kHz, 55-70 dB loss at 6 kHz)
Speakers
    HO
   62 y.o., male
   left middle cerebral artery infarct early 1996
   unimpaired visuo-perceptual processing of
    pictures & words
   semantic abilities just below control range
   impaired receptive phonological processing
   moderate hearing loss
    (45-55 dB loss at & above 4 kHz)
Material
                               Experiment 1 Experiment 2
Words                          15                12
Repetitions (per session)       6                15
Sessions                        5                  5
* Words were presented in pseudo-random order among fillers
Task                           Reading           Reading
Words used in current          bed, big, dark,   buck, bug
study                          duck,pen, pig,    puck, pad, pat
                               pick, saw, sea
Speaker(s)                     AR, HO            HO
Acoustic Procedure
   Recordings digitised at 20,000 Hz
   Spectrograms hand-labelled & analysed
    using the EMU speech database system &
    the R statistical analysis software
   Acoustic dimensions analysed:
     Burst Release Duration (BRDUR)
     Vowel Duration (VDUR)
     Fricative Spectral Moments (FSM)
       First SM (SM1): average spectral frequency
Q1: Statistical Analysis
 Repeated Measures ANOVA
 To investigate for any significant differences
  in the variance between sessions

   Sphericity: assumption that variance is equal
    between pairs of scores
    (Field, 2000, p.324)
Q1: Results - Error-bar Plots
Example: Burst Release Duration of /d/ (AR)
                            dark                                         duck

         150.00                                       150.00




         100.00                                       100.00
 brdur




                                              brdur
          50.00                                        50.00

                                         
                                                                   
                                                                                      
                                                               
                                                                                  

           0.00                                         0.00




         - 50.00                                      - 50.00

                   A   B      C       D   E                     A   B      C       D   E


                           se ssion                                     se ssion
           Q1: Results - Error-bar Plots
           Example: Burst Release Duration of /p/ (HO)
                            pe n                                         pig                                          pick

        150.00                                       150.00                                       150.00


                                                                                                                              
                                                                                                                                 
                                                                                 
                                                                          
                                                                                                               
        100.00                                       100.00                                       100.00
                             
                                                                                      
                      
                                                                  
brdur




                                             brdur




                                                                                          brdur
         50.00                                        50.00                                        50.00




          0.00                                         0.00                                         0.00




        - 50.00                                      - 50.00                                      - 50.00

                  A   B      C       D   E                     A   B      C       D   E                     A   B      C       D   E

                          se ssion                                     se ssion                                     se ssion
Q1: Results - Error-bar Plots
Example: 1st Spectral Moment of /s/ (HO)
                         saw                                        sea
       7000.00                                    7000.00




       6000.00                                    6000.00
                                                                
                                                                              

                                                                                   
 sm1




                                            sm1
                                                                       



       5000.00                                    5000.00
                                       
                     
                                    

                            



       4000.00                                    4000.00




                 A   B      C       D   E                   A   B      C       D   E


                         se ssion                                   se ssion
          Q1: Results - Error-bar Plots
          Example: Vowel Duration of /I/ (AR)
                          big                                        pig                                        pick

       250.00                                     250.00                                     250.00




       200.00                                     200.00                                     200.00
                    
                
vdur




                                           vdur




                                                                                      vdur
                                                                                                                
                                                                                                     
                                                                                 

       150.00                                     150.00                                     150.00
                                                                                                                            
                                                                                                          
                                                                    
                                                                                                                         




       100.00                                     100.00                                     100.00




                1   2      3       4   5                   1   2      3       4   5                   1   2      3       4   5

                        se ssion                                   se ssion                                   se ssion
Q1: Results - ANOVA
Vowel duration /I/ in “Big” (Speaker AR)
Descriptive Statistics
                Mean      Std. Deviation      N
SESSION1       177.2333        25.69425       6
SESSION2       187.5333        31.51582       6
SESSION3       135.0167        20.12912       6
SESSION4       133.6800        25.64429       6
SESSION5       161.9833        21.51803       6

Mauchly's Test of Sphericity
Within Subjects Effect       Mauchly's W          Approx. Chi-Square        df     Sig.
Speaker AR (Big - VDUR)                    .013                  14.954        9   .127

Tests of Within-Subjects Effects
Source                    Sum of Squares          df    Mean Square        F       Sig.
Speaker AR (Big – VDUR)         14230.612         4          3557.653     6.798    .001
Error                           10466.874         20          523.344
Q1: Results - ANOVA
                                     Pairwise Comparisons
   Measure: MEASURE_1

                                       Mean                              95% Confidence Interval for
                                                                                            a
                                     Difference                                   Difference
                                                                     a
   (I) SESSION      (J) SESSION          (I-J)    Std. Error      Sig.   Lower Bound Upper Bound
   1                2                   -10.300     10.444       1.000        -60.152           39.552
                    3                     42.217     8.926        .052          -.391           84.825
                    4                     43.553    16.581        .467        -35.595         122.702
                    5                     15.250     7.388        .939        -20.015           50.515
   2                1                     10.300    10.444       1.000        -39.552           60.152
                    3                     52.517*    6.042        .003         23.678           81.355
                    4                     53.853    21.075        .509        -46.745         154.452
                    5                     25.550    15.055       1.000        -46.313           97.413
   3                1                   -42.217      8.926        .052        -84.825             .391
                    2                   -52.517*     6.042        .003        -81.355          -23.678
                    4                       1.337   16.299       1.000        -76.466           79.140
                    5                   -26.967     11.116        .597        -80.026           26.093
   4                1                   -43.553     16.581        .467      -122.702            35.595
                    2                   -53.853     21.075        .509      -154.452            46.745
                    3                      -1.337   16.299       1.000        -79.140           76.466
                    5                   -28.303     11.387        .555        -82.656           26.049
   5                1                   -15.250      7.388        .939        -50.515           20.015
                    2                   -25.550     15.055       1.000        -97.413           46.313
                    3                     26.967    11.116        .597        -26.093           80.026
                    4                     28.303    11.387        .555        -26.049           82.656
       *. The mean difference is significant at the .05 level.
       a. Adjustment for multiple comparisons: Bonferroni.
Q1: Summary of Results
   ANOVA: significant differences between
    sessions for only some words
   Post-hoc: significant differences between
    session means were few
   and no pattern of increase/decrease in
    differences across sessions
   Results were similar for VDUR, BRDUR,
    Fricative SM1, the other words, both
    speakers
   Results were also similar in 2nd experiment
Q1: Summary of Results
   No change across sessions = no effect
    (practice/fatigue) of multiple productions

   Also, no. of repetitions in Exp 2 > Exp 1, yet
    no effect of increased repetition on speech
    production
Implication of Results
   method of eliciting multiple production may
    not confound investigation of nature of
    speech disorder

   but for treatment – practice of this type
    elicited in this study may not contribute to
    improvement in speech production
Q2: Other Speech Production
Features
   Pre-voicing (HO)




                           Pre-voicing
      Figure 1a                          Figure 1b

                  Example: Bug
Q2: Other Speech Production
Features
   Pre-voicing (HO)
    – voicing preceding release   Word Occurrence
                                  Bed    83%
      of word initial /b/ & /d/
                                  Big    52%
                                  Buck   95%
   Impaired laryngeal control
                                  Bug    89%
    – difficulty coordinating     Dark   52%
    timing of stop release for    Duck   63%

    voiced stops
Q2: Other Speech Production
Features                       Schwa




   Schwa
    appended to
    final stop
    (HO)

    Example: Pad




                    Figure 3
Q2: Other Speech Production
Features
   Schwa appended to word-final stop (HO)
    – Voiced stops
                           Word           Occurrence
                           /d/ in “Bed”   93%
 Indicative of careful
                           /d/ in “Pad”   44%
  speech production        /g/ in “Big”   86%
 Or due to speech         /g/ in “Bug”   77%
  disorder                 /g/ in “Pig”   53%

 Or due to hearing loss
Q2: Other Speech Production
Features - Summary
 No pattern in occurrence of these features
  over sessions
 Not an effect of multiple production
 May just be characteristic of disordered
  speech production
Future Directions
   Word-final consonants (stops)
   Nasal consonants
   Co-articulation
   Analysis of the other speech production
    features
   Analysis of control data
Thank you for listening
         

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:5
posted:12/19/2011
language:English
pages:30