Looking at my thought process, I think I’m using this differentiating test:
Look at the probability of the outcome, given the speech—if it’s high enough that you can ignore the receiver of the message as an independent agent whose response generates uncertainty, the causation looks pretty direct. But if the outcome is dependent on people freely considering the information and acting on their own conclusions (as they would if the information was known by other means), then it looks indirect enough that I consider “transmiting information” as the function of the speech.
Thanks, this is what I thought. That would be the difference between my first and third types of speech, and an example of a controversy about how to draw the line.