Suppose all English speakers decided to make plurals of nouns a new way, with a prefix whork. It’s still make up of existing sound elements. You can create new sounds in language, but that isn’t done as part of the word formation process, rather it comes about as part of learning and producing existing words. Once those sounds exist in the language, they can be elements for making new morhemes.
I thought he might have been thinking of borrowing and the occasional creation of a word out of whole cloth, like Kodak. In other words, I think he was thinking of word creation when you said word formation, which is pretty much by definition the process of building bigger words out of smaller preexistent units.
Not to be pedantic, but surely this is wrong? If all word formation involved combining existing elements then wouldnt it be impossible for (eg) the morphemes and sounds of a language to change over time?
Not really. Imagine English as it is, then imagine that there’s a change in one of the sounds–say t comes to be pronounced d in the middle of words. Then morphemes containing t in the middle will come to contain d in its place: -ity, for example, will be pronounced -idy. All the words formed by compounding with it will change too: rapidity will become rapididy, for example. (Of course, for many Americans, including me, something like this has already happened, so that writer and rider are almost identical.) Mind you, there are other sources of words than word formation: You can borrow them from other languages, form new smaller words from larger words (to laze from lazy, for example), and even make up new ones, but words formed in these ways all use the sounds already existing in the language. (Except that very rarely foreign sounds are borrowed as well.)
I understand that the sound “tu” has many different meanings, and I can see how you would think that “my” author has confused them, but like I said, I gave you guys a basic overview. The author does not simply give the example “gonna” from nowhere. He gives example of how the phrase first came up in writings and actually meant “I am going to new york, in order to fight so and so.” The “going to” in this sentence is not then reducible to “gonna” or its meanign right away. “Going to” a place came to mean, conceptually, the future indicitve of an action, namely in this example the action of fighting so and so. So at first, one would only use the phrase “going to” strictly when speaking of going to a place in order to do soemthing. Eventually the place was to be assumed and one could say " I am going (to new york is assumed) in order to fight so and so" which reduces to “I am going (to new york) to fight so and so.” which is then: “I am going to fight so and so” Which reduces to “I am gonna fight so and so.” (the new york has been removed in both of these last cases even conceptually) For clarity, in the first cases the actual action of moving (of going) is still involved, as in “I am actually moving myself to a place in order to fight so and so.” But when one says “I am gonna fight so and so”, one says “In the future I will fight so and so.” Believe me, the book is less convoluted than I am.
The real issue of the word gonna, is not how the sound came into being, although that is of interest, but of how the concept “going” came to be the future indicitive of a word.
The same thing occurs in spanish, where the verb “Ir” (to go) is used to indicate a future action, as in “Voy a comer” Voy (I am going, I go) a (to) comer (to eat, eat, eating). I believe if one wants to say “I will eat” One needs a single word which is: “Comeré”
Since it’s the net, that means: I’m going to steal your music.
That’s this 
one. :yarr: means you’re hoping people mistake you for Errol Flynn.