I recommend working with Hubert. He is open to new ideas, diligent, always willing to help, and competent. Throughout our collaboration, he supported the team, ensuring projects were completed thoroughly and on time.
Defining the research problem
What did I research?
Difficulties related to the perception and expression of linguistic and emotional prosody in high-functioning individuals with autism.
So what exactly?
Prosody
Prosody is the sound properties of speech. There are two types of prosody:
Linguistic prosody
- Word and sentence stress
- Intonation
- Rhythm
Emotional prosody
- Joy
- Anger
- Sadness
In communication, we use two cognitive processes:
Speech perception
Receiving, recognizing, and interpreting speech sounds.
Speech expression
Expressing thoughts, feelings, information, and intentions using words.
Autism
Autism spectrum disorder is a different way of functioning, sensing, and processing sensory stimuli, language information, and social behaviors.
Why did I conduct research?
Advancement of science
Research on prosody disorders in people with autism contributes to the development of linguistics, while also filling a research gap in this field.
Lack of such research in Poland
Local researchers only note the presence of these irregularities, while foreign specialists examine their type and severity.
Differences between languages
Due to cultural and linguistic differences in emotional and linguistic prosody, the results of studies conducted in other languages cannot be easily applied to Polish.
What did I want to find out?
- Which components of the prosodic system do the subjects have difficulty with?
- Are their prosody speech disorders continuous or sporadic?
- Which is more problematic for them – the perception or expression of prosodic elements?
- Which type of prosody deviates more from the norm – linguistic or emotional?
Developing the research tool
What research tool did I use?
I conducted the research using my own custom tool:
A questionnaire for assessing prosodic listening and prosody speech expression.
What research tool did I use?
I conducted the research using my own custom tool:
A questionnaire for assessing prosodic listening and prosody speech expression.
What did I do before developing the questionnaire?
I analyzed other foreign and Polish research tools of this type.
Foreign tools for assessing speech prosody
|
L Linguistic
E Emotional
Yes
No
|
Prosody under study | Perception study | Expression study | Examined features |
|---|
L Linguistic
E Emotional
Yes
No
Polish tools for assessing speech prosody
|
L Linguistic
E Emotional
Yes
No
|
Prosody under study | Perception study | Expression study | Examined features |
|---|
L Linguistic
E Emotional
Yes
No
Conclusions
Foreign research tools
They are used to study speech perception and expression.
For perception, they most often assess intonation and emotional prosody.
For expression, they check word and sentence stress, intonation, as well as voice pitch and intensity.
Polish research tools
They focus solely on studying perception. They mainly assess intonation and emotional prosody.
How did I construct the research tool?
Versions of the questionnaire
Paper version
Electronic version
Studies in the questionnaire
Study of the Perception of Linguistic and Emotional Prosody
Study of the Expression of Linguistic and Emotional Prosody
Diagnosed prosodic features
- word stress
- sentence stress
- intonational contours
- rhythmic structures
- emotional prosody
Conducting the research
What did I do before conducting the research?
Pilot studies
I assessed the clarity of the questionnaire instructions and task content, as well as the effectiveness of the procedures under specific conditions. The results showed that the prepared materials are understandable for the participants and the data collection methods are effective in the planned research conditions.
What did I do before conducting the research?
Autism is diagnosed four times more often in men than in women, so I invited only men for the study.
The study involved individuals with autism (experimental group) and individuals without autism (control group).
I recruited men with autism through foundations working for people with autism.
I recruited men without autism from volunteers.
Group characteristics / Participant selection criteria
Experimental group
- 10 men aged 21-35 years
- Individuals diagnosed with high-functioning autism
- Individuals without diagnosed additional disorders
- Individuals who could communicate verbally and read with comprehension
Control group
- 10 men aged 25-34 years
- Typically developing individuals
- Individuals without hearing impairment or auditory processing disorders
- Individuals without intellectual disabilities
Under what circumstances and conditions did I conduct the research?
Course of the study
I assessed perception and expression of speech prosody separately.
Participant comfort
Participants sat at a table. If needed, they could listen to each recording again and take breaks.
Privacy and confidentiality
I informed participants about what and how would be examined and the rules for collecting, processing, and storing data, including personal data.
Location of the study
I conducted the research in a therapy office or another quiet location.
Participant tasks
People decided which of the narrator's utterances sounded correct and responded according to instructions.
Recording the study
I recorded the expression of speech prosody study using a voice recorder.
Judge recording reviews
How did I manage the recordings?
Compilation of recordings
Recordings before segmentation
20 expression studies = 20 recordings
Segmentation of recordings
Recordings after segmentation
Total of 2000 files (1000 in each group)
Storage of recordings
Google Drive
I placed all files on a password-protected network drive.
Sharing the recordings
Platform jotform.com
I created interactive forms where each speech recording had its own player.
Who evaluated the recordings and how?
4 qualified judges
- Individuals with knowledge in speech therapy.
Rules for conducting the expertise
- Familiarize themselves with the instructions for completing the forms.
- Read the instructions given to the participant.
- Listen to, comment on, and evaluate all voice recordings.
- No access to information about which recording belonged to a person with autism and which to a person without autism.
- No possibility of consulting with each other.
How did the judges evaluate the recordings?
The judges listened to, commented on, and assessed the recordings of the participants' speech using forms.
Listening
Media player
- It is connected to the audio file stored on a network drive.
Analysis method – perceptual evaluation
Main drawback of the method
- Subjectivity of the assessment
Countermeasures
- Obtaining opinions from multiple people
- Introducing a rating scale
Commenting (3 ways)
Selection from a list
- Predefined statement matched to a prosodic feature
- Mandatory
Supplementary comment (text field)
- Justification for the decision
- Optional
Additional comment (text field)
- Separate notes and observations without affecting the assessment
- Optional
Evaluation
Rating from a list
- Three-point scales
- Matched to the content of the comment
- Mandatory
Supplementary comment (text field)
- Justification for the decision
- Optional
| Type of realization | Score |
|---|---|
| correct, well audible | 1 point |
| correct, poorly audible | 0.5 point |
| hypercorrect | 0.5 point |
| unclear | 0 point |
| incorrect | 0 point |
| other | 1 point |
| other | 0.5 point |
| other | 0 point |
Analysis and results
What results did I get?
Results of perception study
Lexical stress
Total points in %
People with autism
81%
People without autism
92%
Results
Both groups had no problem perceiving stress in two-syllable words.
Differences between the groups appeared in the perception of stress in three-syllable words, reflected in the number of errors made.
Sentence stress
Total points in %
People with autism
99%
People without autism
100%
Results
People from both groups correctly identified the most strongly stressed words in sentences spoken by the speaker.
Intonation
Total points in %
People with autism
94%
People without autism
97%
Results
Men from both groups had no problem recognizing the rising intonational contour of questions, achieving full accuracy.
People with autism made 10 percentage points more errors by identifying exclamatory sentences as declarative.
Rhythm
Total points in %
People with autism
50%
People without autism
82%
Results
People with autism made mistakes 18 percentage points more often in the case of two-syllable words and 26 percentage points more often in the case of three-syllable words, and their errors concerned all combinations of stress on syllables.
Emotions
Total points in %
People with autism
99%
People without autism
99%
Results
People from both groups had no problem with auditory differentiation of emotions.
Results of expression study
Lexical stress
Experimental group vs. control group
People with autism
-26 p.p.
Correct and clearly audible use of paroxytonic stress (code AL1).
People with autism
+2 p.p.
Hypercorrect use of stress (code AL3).
The most common mistake
People with autism
7%
Use of two stresses (scansion) (code AL6).
People with autism
7%
Use of stress on the last syllable of the word (code AL5).
People without autism
2%
Use of two stresses, on the first and last syllable of the word (code AL13).
Sentence stress
Experimental group vs. control group
People with autism
+11 p.p.
Incorrect use of stress.
The most common mistake
People with autism
11%
People without autism
6%
Use of sentence stress on multiple words in the utterance, with one being stressed correctly, in accordance with the given question (code AZ8).
Intonation
Experimental group vs. control group
People with autism
+9 p.p.
Incorrect use of intonation.
The most common mistake
People with autism
6%
Use of anti-cadential intonation, inappropriate for a declarative sentence (code KI5).
People without autism
2%
Inappropriate speech expression of interrogative sentences, sounding like exclamatory questions (code KI13).
Rhythm
Experimental group vs. control group
People with autism
-1 p.p.
Incorrect rhythmic patterns disrupted by symptoms of fluency disorders (code SR6).
The most common mistake
People with autism
13%
Vowel elongation (code SR6.4).
People without autism
12%
Use of inter-word pauses of similar length (code SR6.9).
Emotions
Experimental group vs. control group
People with autism
+13 p.p.
Incorrect expression of emotions.
The most common mistake
People with autism
7%
People without autism
6%
Unclear realizations, where it was difficult to determine the emotional state in which the sentence was uttered (code PE4).
Did the judges agree with each other?
I assessed the level of agreement among the evaluators using the kappa coefficient. Before taking measurements, I encoded the content of comments from the selection list and selected the appropriate measure.
Encoding comments
I encoded comments from the selection list in three steps:
-
1
I downloaded the data from the forms in CSV files format
-
2
I merged the CSV files using the Power Query add-in in Excel
-
3
I converted the comment contents into codes
To each comment from the selection list, I assigned an identification code consisting of the first two letters of the evaluated prosodic feature's name and a number, for example, the code AL1 refers to the first comment from the selection list in the lexical stress expression study. Ratings (points) did not require coding.
Agreement coefficient
The level of agreement among judges, both in terms of the comments they selected and the ratings they assigned, was assessed using the AC1/AC2 coefficient.
The AC1/AC2 parameter proved to be particularly useful due to its resistance to the well-known paradox in the literature related to the tendency of kappa-type coefficients to assume low values when there is a large number of identical data, which often leads to incorrect conclusions about the lack of agreement among judges.
Values of comments vs. ratings agreement coefficients
Both groups
Throughout the study, the values of comment coefficients were generally lower than the values of rating coefficients.
Conclusion
Judges more often disagreed with each other regarding the reasons for irregularities than regarding the fact of their occurrence.
People with autism
Throughout the study, the values of comment and rating coefficients were generally lower in the experimental group.
Conclusion
Judges had a greater difficulty achieving agreement regarding the correctness of prosodic structure realizations produced by individuals with autism.
Exception: Study of rhythmic structure expression
Both groups
Throughout the study, the values of comment and rating coefficients were generally diverse.
People with autism
Throughout the study, the values of rating coefficients were generally slightly higher in the study group.
Conclusion
Judges did not agree on the correctness of rhythmic patterns pronounced by both individuals with autism and typically developing individuals.
What are the responses to the research questions?
-
Which components of the prosodic system do respondents have difficulty with?
Linguistic prosody
Deficits in prosodic functions:
- misplaced lexical or sentence stress;
- inappropriately chosen or well-chosen but lacking expressive force intonation contours;
- stilted rhythm and slowed speech tempo.
Emotional prosody
Diminished prosodic form characterized by unclear articulation, i.e., hindering the recognition of emotional content.
Differences between groups
Differences pertained to the severity of abnormalities; they were quantitative rather than qualitative. Individuals with autism made more errors of the same type than neurotypical individuals.
-
Do speech prosody disorders have a continuous or sporadic nature in their case?
Prosodic disorders in individuals with ASD may manifest as a consistent aspect of speech or only occur in specific situations or types of utterances.
-
What causes them more problems – perception or expression of prosodic elements?
In individuals with autism, the production of prosodic structures is more disrupted than their perception. However, there is much evidence to suggest that difficulties in their production are not related to a lack of ability to understand them.
-
Which type of prosody deviates more from the norm – linguistic or emotional?
Research results showed that in individuals with autism, emotional prosody was not more disturbed than linguistic prosody. Both types of prosody were equally distorted.
Many issues related to the prosody of individuals with autism require further research and analysis..
