This "old" debate (thank goodness it continues to surface as more people enter the instructional technology and especially the interactive multimedia field) has finally prompted me to stop lurking and jump in. I'd like to make several points, based upon my experience with well over 100 interactive multimedia courses, and supported by the literature. (But what isn't, if you look deep enough, for long enough, and pose your original premise the correct way ?) I'll try to minimize the influence of my ISD and task-based training background and speak more in generalities (principles?) to communicate to all, I hope.
(1) The original question/issue posed dealt with interactions, not media per se. The medium used can either limit or support certain types of interactions. Clark did not perform original research (to my knowledge, it was largely meta-analysis of others' research) nor focused research on interactions specifically.
(2) There is a clearly missing (in my mind) stated assumption or premise in Clark's analysis of the findings--that is, if you hold all the other variables (and they're almost innumerable) constant, then the media makes no significant difference. But, those variables were not held constant (much less the Hawthorne effect or novelty). As an analogy, Clark compared books to movies, without regards to content or structure, much less the original issue of interacting with that content. How does Gone With the Wind or Ben Hur compare to Animal House or The Muppets Take Manhattan? How many times have we come away from the theater judging how "true" the movie was to the book? Now try comparing all books ever written or all movies ever filmed to determine if movies or books are inherently better.
(3) Most media comparison studies (Kulik, et al., and others) that have tried to imply significant findings in EITHER direction have not addressed the underlying instructional strategy used to present the content. As an example, traditional classrooms use lecture, inquiry, role plays, labs, research, and other strategies, in addition to the more recently controversial ones such as direct instruction, mastery learning, and simulation. Interactive multimedia uses tutorials, drill-and-practice, guided discovery, free discovery, simulation, etc. These strategies are intermixed and used judiciously to present the content. As an analogy, what would the results of teaching college chemistry be if there were no laboratories, only classroom lecture? Would role plays work? What about "pure inquiry?" Probably not nearly as well.
(4) Any instructional strategy (just as does the medium) has implied limits of support for types of interactions that can vary significantly in both quantity and quality, regardless of the content that they are being used to present. Video, for example, like classroom lecture, is largely a passive experience. Simulations or role plays, whether presented in class or via a computer, are much more interactive.
(5) There's another important variable seldom discussed outside of ISD and training circles with any degree of precision--the type or level of learning desired. This variable can be "captured" with a clearly written performance objective (I don't care if it has three parts or five parts), but it is very, very difficult to write these objectives and they must (or should) be based upon a thorough analysis of the content and the intended use or application of that content. Memorizing and being able to recall (much less consistently apply with a stated level of accuracy) an algebra or geometry formula, for example, is clearly a different learning requirement than developing your own proof of that formula, much less deriving corollaries to the formula or proposing new formulas. Now try teaching geometry constrained by only one medium or instructional strategy. Use classroom lecture to teach math without problems and graded homework assignments--or via videotape without workbooks and without the interactions of pause, rewind, play (that's what most lectures are like, aren't they?). You MIGHT get the student to memorize--I doubt you'd ever get them to the point of proofs.
(6) The research DOES seem to show that if we can increase both the frequency (i.e., quantity) and quality (I'd define this as appropriateness of the interaction to the level of learning desired) then you can increase the amount of learning that takes place (see David Jonassen, among others, for more support of this perspective).
(7) In my experience, as the level of desired learning increases, you must generally increase both the quantity and the quality of the interactions to be effective. The same holds true for the complexity or difficulty of the subject matter. Simply stated, the more complex the subject matter or the higher the level of learning desired, the more interactive the instruction should be.
In summary, the design principle that I apply daily is "The media MAY make no significant difference, but methods used to design and apply that media makes a VERY significant difference."
I've expoused (or exposed??) enough, without even touching on other important principles of design such as the structure of content being important (schema theory or mental models, etc.). I'd pose a question to the group.
How do constructivists explain the apparent effectiveness of direct instruction methods for lower levels of learning such as facts and concepts? As a trainer, I must always strive for a time-effective and cost-effective design--what are equivalent, alternative strategies (in terms of learning per unit of time spent) in interactive multimedia for the "tried and true" tutorials or drills? I know that my designs MAY be boring, but I also know they're effective. What are my alternatives for teaching the mass of content that I'm challenged to present in a consistent, cost-effective method? Any suggestions?