Assessing Speech Intelligibility and Severity

1
Assessing Speech Intelligibility and Severity

• What are some standard procedures?

2
Are measures of intelligibility and severity the
same?

• Although judgements about intelligibility and
  severity may be correlated, they represent two
  different indices about an individuals speech
• For example, a child may have a severe resonance
  disorder, but his/her speech is still intelligible

3
Intelligibility Scales

• Typically a panel of 2-5 listeners (expert
  novice familiar unfamiliar) rate a taped
  (audio or video) segment of childs speech
• Rank intelligibility of childs speech compared
  to age peers
• Scales typically use 3-point or 5-point
  judgment scales (See Bleile)
• scores are averaged to derive a composite
  intelligibility rating score

4
Intelligibility Scales

• Kent, Miolo, Bloedel (1994) compared 19
  intelligibility measures according to 5
  categories that differed with regard to the
  emphasis of the analysis (phonetic phonemic
  word level conversation)

5
Intelligibility Scales

• Listed 8 factors that influence clinical
  evaluation of intelligibility
• loss of phonological contrasts
• loss of contrasts in specific environments
• extent of homonymy
• amt of difference between targetrealization
• frequency of occurrence in English
• consistency
• familiarity of listener with speaker
• context in which communication occurs

6
Intelligibility

• Weston Shriberg (1992) concluded that
  articulation variables alone cannot account for
  all the breakdowns that result in communication
• other general contextual and linguistic variables
  are related to speech intelligibility

7
Severity

• Similar to intelligibility rating scales, a panel
  of familiar or unfamiliar listeners judge a
  segment of a childs recorded (audio or
  video-taped) speech (single word or connected
  speech)

8
Severity

• Perceptual scales (see Bleile) or quantitative
  measures
• 4 Point Clinical Judgement Scale of Severity
• No disorder-Mild-Moderate-Severe
• average score of 3.5 often required to provide
  clinical services

9
Severity

• Quantitative Measures
• PCC (and 8 variations)
• Hodsons PDS
• Edwards PDI

10
Comparison of PCC and PDS Severity Ratings

• PCC
• Mild gt90
• Mild-Mod 65-85
• Mod-Sev 50-65
• Severe lt50

• PDS
• Mild 1-19 pts
• Moderate 20-39
• Severe 40-59
• Profound 60

11
Shriberg, Austin, Lewis, McSweeny, Wilson (1997)

• 9 speech metrics
• PCC (based on conv speech)
• PCC-A (common clinical distortions scored
  correct)
• PCC-R (uncommon distortion scored correct)
• ACI (differentially weighs distortion
  sub/omis)
• PCI (percentage of sounds mastered-early
  talkers)
• PVC (similar to PCC, but for vowels/diphthongs)
• PVC-R (similar to PCC-R, but for
  vowels/diphthongs)
• PPC (percentage to consonants/vowels correct)
• PPC-R (scores distortions correct)

12
Which metric is most appropriate?

• Depends on specific needs of the assessment
• Interest limited to consonants
• PCC, PCC-A, PCC-R, ACI
• Young/severely delayed children
• PCI
• Interest in vowels/diphthongs
• PVC, PVC-R
• Interest in articulation competence on all speech
  sounds
• PPC, PPC-R

13
Rafaat, Rvachew, Russell (1995)

• Purpose of study was to determine the percentage
  of agreement between SLPs in rating PI severity
• Adequate reliability for older children (46),
  but unreliable for children under 36
• Unable to reliably distinguish TD from mild delay

14

• Why were SLPs less reliable on severity ratings
  of younger children?

15

• What factors account for differences in
  reliability ratings?