Darrell Matthews and Dr. Shawn Nissen, Communication Disorders
Acoustically, the speech characteristics of male and female speakers differ from one another. In adults, this variation is clearly caused by physiological differences between males and females. The size and shape of the vocal tract, as well as the length of the vocal folds have an impact on the sound of a person’s voice. Because adult males typically have a larger larynx and longer vocal folds than adult females, their voices tend to be lower.
These findings help to explain why adult female and male voices are fairly easy to tell apart from one another. However, physiological differences cannot explain the gender-related voice differences in children. Research has found that there are significant acoustical differences between the speech of boys and girls (Lee et al., 1999) even though the size and shape of their vocal tracts do not differ significantly (Fitch & Giedd, 1999). This would imply that gender-specific variation in the speech of children can be attributed to learned behaviors rather than physiological differences.
The intent of this study was to examine possible gender differences in the way children use prosody. Prosody refers to the changes in pitch, loudness, and duration that speakers use to convey meaning in a sentence or phrase. So for example, the same sentence can be taken as a statement or a question depending on the speaker’s use of pitch. We were curious to see if boys and girls tend to use prosody differently from one another, and how their use of prosody compares to that of normal adults.
My role in the project was to collect speech samples from 10 normal adults (5 male, 5 female). These samples would serve as a control to which we could compare the children’s samples. I found participants in some of my undergraduate classes. Each session with a participant was done in a sound-treated booth in Dr. Nissen’s lab. I began each session with a quick hearing screening to make sure that the participants had normal hearing. Each participant was then fitted with a headset microphone which was connected to recording equipment and a computer outside the booth. The task was very simple: the participant was shown ten simple drawings and asked to read a caption for each drawing. For example, one of the captions was, “The boy is swimming in the pool.” Then we went through each drawing again, but this time I purposefully made an error. For instance, I would say, “The dog is swimming in the pool.” The participant was then supposed to correct my mistake by saying the correct sentence (e.g. “No, the boy is swimming in the pool!”). This task was designed to elicit a specific emphasis or contrast on the corrected information.
After doing this task with all ten subjects, the next step was to analyze the recordings. Using a software program dedicated to acoustic analysis of speech (PRAAT 2009 – Boersma 2003) I segmented out the subject, verb, and object from each sentence and recorded the fundamental frequency, the intensity, and the duration of each segment. Dr. Nissen then ran a statistical analysis on the data I collected to find any significant results.
The analysis of the data yielded some surprises. Some of the findings matched well with previous research, but some of the results seemed to be opposite of what other researchers have found. At this point it was my job to try to figure out why we weren’t getting the expected results. Dr. Nissen and I began to hypothesize possible problems with the design, the measurements and the analysis. I believe that I learned more during this process than I did at any other point during the project.
We were able to identify a number of problems that will need to be addressed as this research continues in the future. First, the task itself had some flaws in its design. Many of the participants became bored and unenthusiastic about the task, which affected their use of prosody. After a while, their intonation became flat and they didn’t change their loudness very much. The task was also in a very unnatural setting so we weren’t really getting the speech that would be used in a normal context. Second, the stimuli that we used made some of the analysis difficult. Many of the sounds in adjacent words were so similar that it was challenging to accurately parse the words apart. Third, the analysis we used may have caused us to miss some of the important aspects of the data. In our original analysis, we only recorded the average pitch and intensity of the target words. After thinking about it, we realized that this could have skewed the results for two-syllable words because speakers often emphasize only the first syllable of the word. We decided to look not only at the average values, but at the peak values as well. I went back through the recordings and painstakingly obtained the peak values. These appear to offer more promising results.
This project was essentially a pilot study. This was Dr. Nissen’s first time studying prosody and it was apparent in the fact that we ran into some problems. Although it was difficult, this experience was extremely valuable. I learned many skills in research that will help me in the future and most importantly, I developed the ability to think critically about a situation and come up with solutions.
*Acknowledgments: First and foremost, thank you to Brent Webb for your generous donation which made this possible. A special thanks to Dr. Shawn Nissen for his patience and kindness in giving me the opportunity to learn by doing, and to Anita Dromey for her collaborative efforts on the project.