Thank you for your question and the very dramatic example! Your question is, "When we think, do we have to think in words, or can we think without using words?" I take your example to be one in which you seem to have been thinking but not in words. You then break down your first question into two, one of which is about the significance of self-reported events; the other is about whether response mechanisms should count as thought.
About the significance of self-reported events. I'd say that in philosophy and even more in psychology, self-reported events don't carry a great deal of weight. However, your example is one that any student of human behavior knows happens quite often. Furthermore, if someone were to doubt that claim, we would have a good sense of how experimentally to settle it.
About the second question: I would say that a majority of scholars concerned with cognition and action would agree that you did not verbalize much of anything to yourself in the process of saving your son. However, many of them would also try to skirt the question whether you were thinking, and talk instead about whether your behavior was *intelligent*. Surely the answer is that it was, and it was still more sophisticated than behaviors that are automatic and inflexible, like sneezing or the startle reflex.
A great deal of research in experimental psychology (especially social psychology) in recent years has been concerned with the so-called "automaticity" of much human behavior. This is behavior that is not guided by conscious deliberation, but is still more flexible and intelligent than the inflexible responses I mentioned above. Important researchers in this area include John Bargh, Tanya Chartrand, and Tim Wilson. For a readable introduction to this approach see T. Wilson's _Strangers to Ourselves_ (Harvard U.P. 2004).