There is no grand unified theory for MM yet, but there are some well-accepted sub-theories that seem to hang together pretty well, and which could offer some practical advice for the multimedia designer.
Cognitive Load Theory is one such sub theory, established in the late '90's. Based on the metaphor of a computer's processing limitations as applied to the brain, the theory predicts that when humans are overloaded, their ability to process new information (learn)is impaired. The so-called split-attention effect was cited as evidence for the theory (a decrease in learning efficiency when exhibits carry redundant information). This choking down of our mental processes is more than just a matter of taking in too much information over time, like water slowing as it moves through a narrowed pipe. The brain is not a passive container, but an active processor in learning; we don't learn what we perceive, but what we construct out of what we perceive. This active process of building up new mental models, then relating them to our previous knowledge takes up mental resources. Any redundancy or overloading that inhibits this internalizing process defeats its own educational objectives.
Learning is not a result of being exposed to an exhibit, or simply recalling a memory, it is work, and takes energy and resources. "Exhibit minimalism" is the best practice. Hence, "cogntive load" suggests the following MM rules of thumb:
Don't put animation, narration and text on the screen at the same time. This provides distraction and makes learning harder.
Don't add extraneous visual or auditory details to an explanation. Same as above.
Don't place on screen (or page) text apart from the visuals you are trying to describe. "Working memory" is limited too, so the less one has to carry around the exhibit in recall, the better off they are.
Likewise, don't separate such explanatory items in time.