An unvoiced sound is one where the vocal cords, or vocal folds, are open, allowing air to pass up from the lungs without restriction until it hits the articulators like the teeth and lips. The vocal folds are two pieces of tissue that sit in the larynx in the throat. The two folds lie open in a V shape, and when they close over the larynx, the air coming up makes them vibrate together, producing a voiced sound. If the folds don’t move or don’t close completely, they do not vibrate.
All the voicing action is at the vocal folds. Once the air comes up into the mouth, articulators can add friction to or stop the flow of air temporarily, but the sound remains voiceless. This quality can be the only audible difference between words in many cases, such as “ton” and “done,” where “t” is voiceless and “d” is voiced.
The motion of the vocal folds can be influenced by surrounding sounds, adding and subtracting voicing if a sound is surrounded by others that have the opposite quality. For example, if unvoiced consonants surround a voiced vowel, the vowel can lose its voicing and become unvoiced. This does not happen in every language, but it does occur in some, such as Japanese. Nothing is wrong with the vocal folds — they are just not closing as far as they normally would because the two surrounding sounds require them to remain open.
An unvoiced sound can occur even when the listener perceives a voiced sound. When the articulators stop the air from moving, the sound this produces is called a stop or plosive, such as “t.” In languages like English, an unvoiced stop sometimes has an additional puff of air that follows its pronunciation, which is a phenomenon called aspiration. Not all languages have aspirated sounds, however, and to those who are used to hearing only aspirated unvoiced sounds, the unaspirated unvoiced sounds can actually sound like they are voiced even though the vocal folds are apart. Linguists who are transcribing languages have to be extra careful to distinguish between unaspirated and voiced stops.
The lack of voicing and vibration in an unvoiced sound is evident when the sound is detailed on a graph called a spectrogram. This is a type of recording where the airstream produced in speech is displayed visually. Dark bands indicate voiced sounds, but unvoiced sounds show up as lightly shaded or non-shaded areas, depending on how much any nearby voicing influences the vocal folds during production of the unvoiced sound.