If I understand it correctly, when you start a program you are executing the main method within the context of a single
thread. Then as you create a new Thread object using "MyThread mt1 = new MyThread("mt1");" you are creating a new thread within the processor for executing various code snippets. In the case of this program mt1 and mt2 are both individual threads running along side of the main thread of the program.
mt1.start() and mt2.start() moves the threads into a ready state so that the cpu is now allowed to begin execution on that respective thread(s). These threads are all allowed to run thru to completion independently of each other within the context of the main thread. (Someone smarter than me might have to clarify whether they run within the parent thread or if the parent can die and then new thread continue to live.) Now comes along "mt1.join", this ties execution of the main parent thread to execution of mt1, so mt1 must run thru to completion before the main thread continues and thus starts mt2.
Hopefully I haven't butchered this topic, but I think that's the way it works.