[quoting Hart, 23 Oct 96.a] I would assume that the methods which Steve Tripp describes are those commonly employed within the paradigm of scientific objectivism.
Not exactly. Strictly speaking there are an infinite number of theories to explain anything. We usually choose the simplest and the weakest. The first is Ockam's razor, but the second relates to the Chomsky hierarchy. There are four levels of "grammar" that can be used to explain a "language" (the regular behavior of a device or organism). The weakest is a finite state automaton (equivalent to behaviorism). The next is a context-free grammar. The next is equivalent of a Turing machine. In general, you always want your theory to be consistent with the data and at the lowest possible level. Chomsky showed that behaviorism was not consistent with the data and therefore a higher level theory was necessary.
Steve has devised a situation (reading/listening), set it running (possibly left the room) and then tested the result. It's a "black box" model.
All of linguistics is like that. You can't open up a person's brain and see how they are making sentences. So what? That is no obstacle to science. Teachers don't know what is going on in the black box either.
Another dilemma for the "media affect learning" advocates, as Steve, David Frampton and others have pointed out, is the very definition of the term "medium." The question has its parallels in the qualitative-quantitative argument: is the medium a physical object or a mental construct? Can we make a Saussurian distinction between "langue" and "paroles?"
Maybe we can. Let's think about it.
Some of the most foolish and simplistic "experiments" in the sad tradition of educational technology in the 1960s...
Bad experiments are not an argument to give up experimentation.
So happy you brought up Salomon and Amount of Invested Mental Effort (AIME) because I had prepared an answer and was afraid I wouldn't be able to use it.
Salomon's original (1984) study is very strange in some ways.
1. Time was not equated.
2. The comparison was between a silent movie and a text, so content was not equated either and was confounded with medium.
3. The viewers rated the media for its "lifelikeness" and this was used as a proxy for ease/difficulty.
The argument was that learners faced with a "difficult" medium and low efficacy will try harder and thus process the material more "deeply" producing better inferences. But we don't really know if they thought text was more difficult. The problems with this design are obvious, but more importantly, I wonder if there is any better evidence for AIME?
Maybe text is just better than "silent movies" for various types of learning.
It may be time for some good grad student to take this apart.