Now let’s take a look at how we perceive certain types of sounds as vowels, taking the Japanese vowel system as an example. Although the number of vowel categories differs across languages, the basic form of vowel perception is universal.

Go to “formant space”, and you will hear a sound whose timbre changes as you move the mouse pointer. (Note that your perception of tone “height” will not change anywhere in the space.) It will sound like a vowel in some regions, and will not in other regions. The horizontal and vertical locations of the pointer correspond to the two peak frequencies involved in the sound, which are called “formant frequencies”. As you stroll around this formant space, you will realize that there are only few small isolated regions where you can clearly tell which vowel you are hearing, although the formant frequencies of the sound vary continuously everywhere. This implies that your vowel perception cannot change gradually between categories.

Nature of vowel production

Now that you have understood that the two formant frequencies influence the perception of vowels, let’s see how you can control those frequencies when you speak.

Go to another “formant space”. At this time, you will not hear any sound. Instead, when you speak a long vowel into a microphone connected to ( or built in ) your PC, you will see a character drifting in the space. The location of the character corresponds to the formant frequencies of the sound you are producing. Now try to manipulate its location by moving your mouth while producing voice. Some of your speech will locate the character within one of the vowel regions and some will not, although it depends on the individuality of your voice and the vowel system of your mother language. After a several attempts, you will realize that you can only gain access to some restricted regions no matter how smoothly you move your mouth. This implies that you cannot change the formant frequencies gradually between categories.