I read the Clark and Kozma papers again, and the special issues of Educational Technology Research and Development on the issue of media as a variable in teaching in learning. (Thank you ITForum for drawing my attention to these.)
I remember in the recent discussion here, one person said they remembered the special issue as resulting in a single conclusion, and another person saying no, the contributions remained about balanced as pro-Kozma or pro-Clark. Having just looked at them all, I think both these views are correct! That is, it is quite easy to classify each paper as one or the other and because it was organized as a debate, no-one produced an explicit resolution. However it seems to me there was some underlying agreement that readers could sensibly take away. The nearest to a clear statement of it is Clark's paper (1994, no. 2 issue of Educational Technology Research & Development): there are no media effects strictly interpreted as unique attributes of media, but there are big effects associated with instructional practices which are frequently associated with different media.
What remains for us to look at (i.e., to research) is: (a) what are the real causal factors, since media are not the crucial ones; (b) whether media are primarily useful to educational researchers by exposing issues that are concealed (e.g., prompting new instructional strategies that somehow just didn't get invented without the new media) or suggesting the importance of what is otherwise concealed in tacit teaching skill; or (c) whether media shift the economics (of time, convenience, delivery "pace," money, etc.) and so make a real practical difference, but only via the practicality of changing instructional tactics.
But it has to be said in favor of Clark's skeptical attitude, that new media research is NOT addressed to identifying what the important factors are. It mostly just seems content to do something new and at most show that it gets a big effect--not to identify why and so lay the basis for new science and replicable transfer in design. So I suppose I agree with Clark as a critic of most research in the area, even though I agree with Kozma that there are effects associated with media that we need to explore with a view to (a) understanding how to exploit them and (b) not getting made obsolete by technical advances.
However Reiser made some telling points that make it look as if Clark is not in great shape for this either. For instance he uses the medical analogy against Clark to argue that delivery systems are as important as the drug in determining whether disease gets cured. (For instance I gather that the current TB epidemic in New York and other big cities is caused by the difficulty in getting patients to complete their drug therapies, not by the failure of the drugs in controlled conditions. In other words, an whole epidemic is being caused by a delivery system issue.)
The most interesting issue for me is whether instructional strategies are the key variable (causal factor), as Clark argued. My vote is currently, as I have said before, on treating the 12 Laurillard activities as the best candidate for a key factor. But they are vulnerable to the formal criticism of not being a necessary condition, i.e., learning can occur without them. Clark criticized media as a factor on this ground, but Reiser pointed out that instructional strategies are equally vulnerable to this criticism (i.e., no instructional strategy is necessary because learning may occur without it) and so are the Laurillard activities unless I try to patch the theory by saying that some students will have internalized activities and so perform them invisibly (e.g., you listen to a talk, and do the re-expression activity by arguing silently in your head with the speaker). I quite like this patch, but it would be difficult to win an argument with a skeptic using it.
So I suppose I feel the debate ground to a halt having established only that media are not a key factor, but that more research is needed particularly around things achieved with new media; but without really telling us convincingly what to study next. It seems to have been better at establishing negative arguments than positive ones. So when someone proposes a new candidate factor, such as interactivity, I don't think it right just to dismiss it as the same as the old factors like "media" which is why I rehearsed some specific arguments against it being the right factor. But I remain uneasy that negative arguments continue to seem more successful than positive ones, and could end up making us dismiss the important factor, whatever it is, along with the rest.