Whispering Songs

Michael Kiefte - mkiefte@dal.ca
School of Human Communication Disorders
Dalhousie University
5599 Fenwick St.
Halifax, Nova Scotia
B3H 1R2 Canada

Popular version of paper 2pSC11
Presented Tuesday afternoon, November 16, 2004
148th ASA Meeting, San Diego, CA

Our ability to speak to one another is most often taken for granted but is actually a highly complex phenomenon. A simple vowel such as "ee" in "heed" can be produced in an infinite variety of ways: it can be spoken aloud, whispered, or sung at a number of different pitches and still be clearly understood by the casual listener. The processes involved in perceiving this vowel appear to be extremely complicated and are still not very well understood. For example, the word "heed" sung by a small child at a high pitch is very different acoustically from the same vowel whispered by an adult male. Despite these huge differences, they are still heard as the same vowel.

The goal of this research is to examine the relationship between vowel quality (what makes a "ee"-sound different from an "oo"-sound) and the pitch of your voice. These properties are completely distinct from one another: for example, we can sing the same vowel at several different notes but we can also say different vowels in a robotic monotone with very little difficulty. Singing the same vowel over and over at different pitches might make for a good tune, but we can hardly call it speaking. Likewise, speaking in a monotone loses a lot of information that is otherwise carried by the pitch of our voice (for example, try asking the question "You're going out today?" without changing the pitch of your voice at the end).

If you concentrate on the movement of your tongue and lips, you can see and feel that they are moving around quite a bit as you say different vowels such as "ee" and "oo". When you sing the vowels at different notes, your larynx or "voice box" is responsible for changing the pitch of your voice; if you put your hand to your Adam's apple, you can feel your throat vibrate as you speak. It doesn't always vibrate when you speak—if you want to whisper for example, you turn off your larynx completely. You might think that if the larynx is not producing any pitch, then we are basically speaking in a monotone when we whisper, albeit a very quiet one. This is not quite true.

You can ask questions when you whisper and if you do so, you can even hear a rise in pitch at the end of the sentence just as if you were speaking aloud. More surprising is the fact that you can even sing while you whisper. If the larynx is primarily responsible for changing the pitch of our voice, and the larynx is effectively switched off when we whisper, where is the change in pitch coming from?

To answer this question, 81 people were asked to whisper several different words while being recorded in our lab. The words were shown one at a time on a computer monitor while a tone was played in one ear of their headphones. The participants were asked to whisper the word they saw on the screen while matching the sound of their voice to the note they heard in the headphone as if they were singing. The tones they heard were played at one of three different notes: C, E, or G. All 81 people were able to do this despite the fact that most of them were not convinced they could. (There was also a fair amount of apprehension around the thought of singing in the first place). Click on the links below to hear some of the recordings:

Listening to the samples across the table from left to right, it is clear that people are able to sing the different notes despite the fact they were whispering. Listening to them from top to bottom, it's also clear that they can say the words quite clearly so that they are distinct from one another. However, we discovered that the word "hood" sung at G sounds very similar to the word "hud" sung at C (red). Similarly, the word "hud" sung at G sounds very much like the word "had" sung at C (yellow).

An acoustic analysis of these vowels shows that in order to whisper the different vowels at different pitches, speakers actually change the vowels themselves since they cannot otherwise change the pitch using their larynx. Therefore, there are some potential confusions between words if you try singing them while whispering. There may also be confusions in ordinary whispered speech (for example, if you whisper the question, such as: "Did you just say 'hood' or 'hud'?"). Neither of these extreme situations occur particularly often. Nevertheless, it does partly explain why whispered speech is sometimes difficult to understand.