Speech Intelligibility

Speech Intelligibility:
Clinical Treatment Approaches for
Children and Adults
Connie Keintz, Katherine Hustad, Jane
Garcia, & Estelle Klasner
Roadmap of the session
□ Introduction to intelligibility and related
concepts
□ Variables that influence intelligibility
□ Interventions to enhance intelligibility
□ Listener barriers and strategies to intelligibility
Traditional view of intelligibility
□ Deficit belongs to the speaker
□ Identify subsystem problems
□ Provide intervention to remediate subsystem
problems
□ No consideration of partner-related
variables and their contribution to
intelligibility deficits
Multi-dimensional view of intelligibility
□ Communication is dyadic
□ Speaker and listener play equally important
roles
□ Meaning is co-constructed in real
communication situations
Speech Intelligibility
□ The extent to which a signal, produced by a
speaker, can be deciphered by a listener
Measurement of Intelligibility
□ Orthographic transcription
□ Scaling procedures
□ Percentage estimates
Measurement of intelligibility
□ Listen to the following sentences
□ Write down what you hear.
Measurement of intelligibility
□ We talked about the situation.
□ The idea for it was mine
□ When you are through, write a report.
□ There are two methods for soaking dry
beans
26 possible words
Clinical uses of Intelligibility measures
□ Index of severity
□ Index of functional limitation
□ Measure of change over time
□ Treatment progress
□ Recovery
□ Degeneration
Limitations of intelligibility measures
□ No information about underlying reasons for
problem
□ Nature of listener difficulties
□ Nature of production deficits
□ Intelligibility influenced by many factors
□ Clinical measures are a “snapshot” of
performance in a specific context
□Usually NOT a real communication situation
Comprehensibility
□ Contextual intelligibility
□ The extent to which a listener can recover a
message, produced by a speaker, when
the listener is given contextual information
that emulates real communication
contexts.
Measurement of Comprehensibility
□ Orthographic transcription with contextual cues
□ Cue: communication
□We talked about the situation.
□ Cue: thinking
□The idea for it was mine
□ Cue: homework
□When you are through, write a report.
□ Cue: cooking
□There are two methods for soaking dry beans
Comprehensibility vs. intelligibility
□ Intelligibility measures deliberately omit
contextual information
□ Both measure employ orthographic
transcription
□ Both measures provide %correct scores
Listener comprehension
□ The extent to which a listener is able to
understand a spoken message
□ Higher level processing
□ Value of individual words comprising the
message de-emphasized
□ Gestalt of primary importance.
Measurement of comprehension
□ Answer open ended questions about the
content of a spoken message
□ Answer yes/no questions about the content of a
message
□ Re-tell the narrative (emphasis on content, not
specific words)
Measurement of comprehension
□ Listen to the following short narrative
Measurement of comprehension
□ Answer these questions:
□ What is the topic of the story?
□ What was the final outcome of the story?
□ What may have happened beforehand to
make the event described in the story
necessary?
□ Based on the outcome of the story what might
happen next?
Measurement of comprehension
□ Target sentences produced by the speaker
□ Jason needed to buy a car.
□ He considered two different models.
□ He did not have much money to spend.
□ He bargained with a salesman for two hours.
□ The final price was within his budget.
□ A used jeep was what he purchased.
Differences between measures
Mean percent correct by measure and speaker
100
90
Intelligibility
Comprehension
Percent correct
80
70
60
50
40
30
20
10
0
1
2
Profound
3
4
5
6
7
Severe
8
Moderate
Severity
9
10
11
Mild
12
Relationship between measures
Which measure is best?
□ Intelligibility, comprehensibility, and
comprehension measures each provide
different, yet complementary information.
□ Select measures that are appropriate for
your clinical purpose.
□ To characterize the functionality of speech, use
measures that capture information-bearing
capability
Connie Keintz, Ph.D. CCC-SLP
Assistant Professor
Communication Sciences & Disorders
Florida Atlantic University
Boca Raton, Florida
Variables that influence intelligibility
□ Linguistic characteristics of the message
□ Semantic Cohesion
□ Length of Message
□ Semantic Predictability
□ Visual Information
Semantic Cohesion
□ Research Factors
Key words/Cohesion = higher scores
(Drager & Reichle, 2001 and Hustad & Beukelman, 2002)
□ Clinical Considerations
□ Keep ideas related for listener benefit
□ This works in our favor in natural communication
Length of Message
□ Research Factors
□ Word is harder to understand in isolation than in a
sentence (Miller, et al, 1951, Sitler, et al, 1983, Yorkston & Beukelman,
1981)
□ Clinical Considerations
□ Provide context whenever possible
□ Teach your speakers to present it
□ Instruct your listeners to look for it
Semantic Predictability
High versus Low Predictability
She is drinking from her _____.
cup
Bob would consider the _____.
pole
□ Clinical Considerations
□ High will allow listeners more information
Visual Speech Information
Buy Bobby a puppy.
Here is his hand.
Dysarthria types
Sound different
Appear different
Intelligibility
“the degree to which a speaker’s
intended message is recovered by a
listener”
Kent, Weismer, Kent, & Rosenbek (1989)
Measurement Tools
Speech Tasks
Tikofsky & Tikofsky (1964)
Tikofsky, Glattke, & Tikofsky
(1967)
Tikofsky (1970)
Platt & colleagues (1977, 1978,
1980)
Yorkston & Beukelman (1980)
Assessment of the Intelligibility of
Dysarthric Speech (AIDS)
Single words
Listener Tasks Presentation
Modes
Transcription
AO
Single words
Transcription
Single words
Multiple choice AO
Sentences
OR
AO
Yorkston, Beukelman, & Tice
(1996) Sentence Intelligibility Test
Enderby (1983)
Words
The Frenchay Test
Sentences
Conversation
Transcription
Kent & colleagues (1989)
Multiple choice AO
Single words
“Therapist”
Transcription
On-line
Audio-record
for reliability
Measurement tools
Rely on auditory-only information
Visual information
not critical
may benefit listeners
Model of Intelligibility
Yorkston (2003)
Impairment
Presentation
mode
Speech Signal
Familiarity
Listener
Transmission
Compensations
Experience
Visual Information Research
□ AV = Auditory Visual
□ AO = Auditory Only
□ Medium/Transmission is compromised in
“normal” studies
Visual information intelligibility studies:
Normal
Hearing-impaired
Laryngectomee
Dysarthria
AV > AO
AV > AO
AV > AO
mixed results
Speakers with dysarthria
Intelligibility scores across moderate:
AV > AO
Intelligibility scores across severe:
No difference in AV/AO
Hunter, Pring, & Martin (1991)
Speakers with dysarthria
Intelligibility scores across speakers:
AV > AO in 1 of 5 speakers
Hustad & Cahill (2003)
Speakers with dysarthria
Intelligibility scores across speakers:
AV > AO in 3 of 8 speakers
More benefit in more severely impaired
speakers
Keintz & Bunton (2004)
Intelligibility scores for individual speakers for AO and AV modes
(* indicates significant results).
100
AO
AV
Intelligibility (%)
90
*
80
*
*
70
60
50
0
A
B
C
D
E
Speaker
F
G
H
Visual Information
□ Influence on intelligibility may be linked to
severity
□ More research is needed to explore
facial movements, severity, types of
dysarthrias
Visual Information
□ What does this mean clinically?
□ Consider each patient specifically
□ Evaluate with listeners having access to visual
information
□ Allow for visual information, if useful in communicative
context
□ Remember that with most speakers, it will be
beneficial
Visual Information
□ Are there speakers who may be hindered
by visual information?
□ Moebius Syndrome
□ Amyotrophic Lateral Sclerosis
□ Hyperkinetic Dysarthria
Supplemental Cues: Speaker
Strategies that Augment
Listener Understanding
Jane Mertz Garcia
Kansas State University
Manhattan, KS
What is speech supplementation?
□ Speaker implemented strategies that
augment or enhance natural speech
□ Hand gestures while speaking
□ Alphabet cues (pointing to the1st letter of
each word while speaking)
□ Topic cues (“prep” listeners about topic)
□ Predictable messages (sentence
form/semantic content)
□ Voice output AAC devices (to supplement
speech) or writing/drawing key points
Simply stated…
Supplemental strategies provide “clues”
and “extra” sources of information that
help listeners understand less intelligible
speech
Supplemental strategies …
□ Take advantage of the interactive and
dynamic aspect of speaker-listener interactions
□ Recognize that listeners naturally apply multiple
sources of “shared” information in a difficult
listening situation
□ Provide listeners additional content information
to help understand less intelligible speech
Result…Speakers with dysarthria are more
effective communicators as messages are
understand with greater accuracy.
Candidates for Supplementation
□ Severity of intelligibility impairment
□ moderate to severe/profound reductions
□ speaker will have difficulty communicating
with behavioral interventions alone
□ Current speaking skills & overall motor abilities
□ strategies may interact in positive or
negative ways with current communication
abilities
□ Type of dysarthria (should be less important)
Hand gestures (Gesticulations)
□ The natural hand movements people use when
speaking
□ By definition…they occur while talking & are
idiosyncratic in nature
□ There are various forms of gesticulations that vary
in information content and function
□ Iconic—illustrates some aspect of the spoken
message (concrete representation of action,
event, or object)
□ Beat—adds emphasis & accentuates speech
(no content information)
Application as a supplement strategy
(particularly iconic hand gestures)
□ Maximize a natural aspect of communication
(hand gestures naturally occur while speaking)
□ “Unaided” technique (nothing required external
to the body)
□ Add meaning and illustrate the content of a
spoken message
□ Provide listeners 2 signals (nonverbal +spoken)
about the same underlying message
Audio+Visual Listening Condition
(Garcia & Dagenais, 1998)
100
80
60
No Gesture
With Gesture
40
20
0
Speaker 1
Speaker 2
Key point: Intelligibility scores improved by 20% and 35% with
the addition of hand gestures while talking
Case study illustration
□ 12 year-old-male (approximately 3 months TPO)
□ Cranial nerve damage following removal of
brainstem tumor
□ Speaking pattern--Flaccid dysarthria (hypernasal
resonance & imprecise articulation)
□ 27% Sentence Intelligibility Test (142 wpm)
□ Strengths: cognitive & language abilities;
UE control for gesturing; motivation & family
support
Treatment Approach
Goal-- Increase use of content (iconic) gestures
□ “Take what you do naturally, but do it more
often”
□ Phases of therapy-- controlled to less structured
speaking tasks
□ Add gestures in sentences conveying similar
language concepts (actions, descriptive words,
pointing gestures)
□ Apply gestures to “random” sentences
□ Increase gestures in “controlled” discourse tasks
Study: Impact of Hand Gestures
(Garcia, Crowe, Redler & Hustad, 2004)
□ Natural speaking situation (self-generated
monologues and gestures)
□ Monologues—similar in length, speaking
rate, and amount of iconic gesturing
□ Impact of gesturing:
□ Words identified correctly (percent intelligibility)
□ Message comprehension
Methods
□ Cue Conditions
□ Full view (all visual cues including
gesticulations)
□ Face cues (edited to only show the speaker’s
face while speaking)
□ Audio-only (eliminated the video signal)
□ 36 everyday listeners heard different monologues
in each condition
□ Counterbalanced (presentation order &
assessment format)
Percent of Understanding
Intelligibility (Words Correct) of Monologues
100
90
Percent Understanding
80
70
60
69.2
57.6
50
40
30
20
19.5
10
0
Full View (including
gesticulations)
Face Cues only
Audio-only
Natural hand gestures while speaking enhanced
intelligibility (compared to face & audio only) (p < .05)
Message Comprehension
What phrase or sentence could you write to describe
the main theme of this speaker’s story?
If you were having a conversation with this speaker
and he told you this story, what question or comment
would you use to add to the conversation?
What people and/or objects were mentioned by the
speaker in his story?
What actions were mentioned by the speaker in his
story?
In what location would this speaker’s story take
place?
Message Comprehension
Comprehension of Monologues
100
90
Percent Understanding
80
70
75.3
60
50
52.5
40
30
20
20.8
10
0
Full Vie w (including
ge sticula tions)
Face Cues only
Audio-only
Natural hand gestures while speaking enhanced
message comprehension (p < .05)
Specific Questions
90
Percent of Comprehension
80
70
60
Audio-only
50
Face Cues only
40
Full View (with
ges ticulations )
30
20
10
0
Ma in
The me
Add
Comm e nt
Actions
Obje cts
Loca tion
□ Full & Face cues exceeded auditory only
□ Main theme, Add comment, Actions were superior in Full
view vs. Face cues condition
Alphabet supplementation
□ “Aided” strategy—speaker points to the first letter
of each word on an alphabet board as he or she
says it.
□ Speaker…
□ Decreases rate of speaking by about 70%
□ Provides additional time to execute motor
movements (improve articulation)
□ Listener…
□ 1st Letter cue narrows range of word choices
□ Increased processing time for listeners
□ Creates clear word boundaries
Sample alphabet board
A
B
C
D
E
F
M
N
O
P
Q
R
G
H
S
T
Y
0
I
U
Z
1
J
K
V
2
3
4
5
L
W
space
• You misunderstood
• Wait a minute
X
• I’ll say it again
delete
6
7
8
• I’ll spell it
9
• I will point to the first letter of each word that I’m saying.
• Please repeat each word after I say it so I know you
understood me.
Research Findings: Sentences
□ Consistently positive benefits for most
□ Linguistic cues alone (superimposed while
speaking) increase intelligibility by 25%
(Hustad & Beukelman, 2001)
□ Reduced rate of speech alone increases
intelligibility by 25% (Crow & Enderby, 1989)
□ When rate is reduced AND speakers have first
letter information, intelligibility may be
increased by up to 44% (Beukelman &
Yorkston, 1977; Hustad, Jones, & Daily, 2003)
Application of alphabet
supplementation to therapy
□ Improving speaker understanding by enhancing
listeners’ linguistic knowledge
□ Increasing redundancy in signal
□ Modifying speakers’ production characteristics in
helpful ways
□ Useful for resolution of communication
breakdown
□ Relatively “easy” to instruct patients in its use
Integrating supplementation
strategies
□ Most individuals with dysarthria can benefit from
having multiple communication options
□ Hand gestures and alphabet supplementation
are not mutually exclusive strategies
□ In fact, some individuals benefit similarly from
both strategies
Hand Gestures Vs. Alphabet Cues
(Hustad & Garcia, 2005)
100
Audio-only
90
A-V
70
60
50
40
30
20
Speaker A
Speaker B
Cue Condition by Speaker
Speaker C
Alphabet
Cues
Gestures
No Cues
Alphabet
Cues
Gestures
No Cues
Alphabet
Cues
0
Gestures
10
No Cues
Mean Intelligibility Score
80
ƒAlphabet cues
& hand gestures
scores were
similar for
Speaker A & C
ƒSpeaker B—
Alphabet cues
were
significantly
higher than
hand gestures
Hand Gestures Vs. Alphabet Cues
(Hustad & Garcia, 2005)
□ Changes in speaking rate
Words Per Minute
Speaking Rate
180
160
140
120
100
80
60
40
20
0
Habitual
Gestures
AC
Speaker A
Speaker B
Speaker C
Cost-Benefit Ratio: Hand Gestures
□
Costs:
□ Must have strong metacognitive abilities
□ Must have adequate motor skills
□ Adequate respiratory support for “chunking”
□ Benefits:
□ Considered an “unaided” strategy– nothing
extrinsic to the body is needed for strategy
□ May contribute to naturalness– given
adequate respiratory
Cost-Benefit Ratio: Alphabet Cues
Costs:
□ Considered an “aided” strategy (you need an
alphabet board)
□ Must have functional literacy skills
□ Reduces naturalness
Benefits:
□ Easy to learn (compared to incorporating hand
gestures)
□ Basic motor skills for pointing to letter cues
□ Rate reduction may benefit production
Current Understanding of Speech
Intelligibility
□ Speech intelligibility is an outcome measure in
dysarthria
□ Speech intelligibility is reduced in all types of
dysarthria
□ Research focus has primarily been on speaker
□ Listener role is beginning to be studied
Questions
□ Do barriers to intelligibility vary as a function of
dysarthria group, ALS versus HD?
□ Do strategies for listening vary as a function of
dysarthria group, ALS versus HD?
Study Overview
Subjects
Individuals with Dysarthria
Task
Components
Read/ Record
SIT sentences
(n=10)
Judges (n=65)
Focus groups (n =4x12)
varying length
Transcribe
SIT
LTCD
Scales
Content Analysis
Pilot Listeners (n=6)
Everyday Listeners (n= 65)
LTCR
LTC R
Revised Scales
Endorsement Patterns
Barrier Scale Categories
Mean
Scale
Scores
4
*
A L S
*
3
*
2
1
Segmental
Suprasegmental
Linguistic
Cognitive
H D
Mean
Scale
Responses
Strategy Scale Categories
*
4
3
2
1
Segmental
Suprasegmental
Linguistic
Cognitive
A L S
H D
ALS Barrier Responses
Mean Scale
Responses
Segmental
Suprasegmental
Linguistic
Cognitive
HD Barrier Responses
Mean Scale
Responses
Segmental
Suprasegmental
Linguistic
Cognitive
ALS Strategy Responses
Mean Scale
Score
Segmental
Suprasegmental
Linguistic
Cognitive
HD Strategy Responses
Mean Scale
Responses
Segmental
Suprasegmental
Linguistic
Cognitive
Contact Information
□
□
□
□
Connie Keintz ([email protected])
Katherine Hustad ([email protected])
Jane Mertz Garcia (jgarcia@[email protected])
Estelle Klasner ([email protected])