If we listen to a singer, and we listen 'past' the music, right down to the detail of what we are actually hearing, we will notice that the flow of the singer's voice divides up into distinct chunks of sound. These may not always be entirely clear, but generally speaking it can be easily heard.
These chunks may coincide with the written words, but often don't. For example, final consonants of words may be chunked together with the beginning of the next word. Consider for example the short phrase 'Love ever after'. Let's say this holds for a moment on 'Love' before saying the next two words.
These can be further broken down into 'Steps' or sonic chunks as follows "Luuh - ve - va - ra +f+ ta". I have indicated the fricative 'f' as a being different because of its role as a pivot linking two sonic chunks.
If singing could be liked to dance, each sonic chunk would be one footfall, or a momentary position of balance. To look at it another way, each chunk is for the singer a single specific action taken. The result of this action is the sonic chunk.
Each sonic chunk has a number of attributes: Timbre (brightness/darkness), attack (fast/slow), loudness (quiet/loud), pitch, intensity (breathy/hard), highness (head/chest) and the perceived syllable of lyric.
In order to control the sonic output, the singer perform a series of actions, during which she controls cord weight, air pressure, head/chest balance, pharyngeal width, the vertical position of the larynx, tongue position. All of these controls result in the correct sonic chunk being presented to the listener.
Furthermore, the singer is also thinking ahead and planning the and the path of any changes which happen during the current chunk as well as moving all the elements to a point ready for the production of the following chunk. Tongue, jaw and throat movements and changes all need to be planned and correctly executed in order to obtain a smooth flow of sound in the phrase. Each must connect seamlessly with those ahead and after itself.