[referring to Gustafson, 26 Feb 97] I've gradually come to the belief that the formulaic view of ID and ISD, which seems to be memorialized (entombed?) in the popular introductory textbooks and workshops, is not just a "novice model" of what ID really is, it's a great example of a highly misleading novice model, which actively inhibits development of a good expert model! (I wish someone would do an expertise analysis of ID and prove me wrong!)
My evidence, of course, is solely experiential, and it has to do with what I've observed while trying to bring novice IDers to expertise through training, coaching, and counseling in an industrial setting. Some observations:
(1) Rarely have I seen a new ID (fresh from an MA program or otherwise trained) who could actually use a front-end analysis (FEA) to design training. By far the most common beginner's behavior is to do the FEA (badly, as a pointless exercise), then forget it and start writing storyboards without reference to the FEA. Another common variation is to substitute design decisions for the FEA.
(2) The same thing happens between design and development. Novices don't seem to have a clue as to what an instructional strategy is, nor do they have any meaningful notion of what alternative strategies are, how they differ, and how they relate to knowledge types and knowledge structures. Instead, they revert to a "universal solvent" methodology, which is some variation of "tell-show-do."
(3) You're right, gaining attention and establishing motivation seem in the novice model to be things you do in the first three or four screens, then forget about. The idea of thematic integrity, functional context, or even seriously taking into account the learner's entry knowledge and frame of reference are unknown.
(4) Novice models rarely include any idea of chunking or knowledge structures. A classic beginner's error is to chunk and pace the instruction in a way which makes sense to the ID (or the SME)--but not to the learner. When the ID knows the content already (as in elementary school subjects), this reliably produces lousy, simplistic knowledge structures, and chunking which is far to big for the learner, resulting in pacing which is way too fast. When the ID doesn't know the content (as in technical training), this reliably produces lousy, complicated knowledge structures and chunking which is far to small for the learner, resulting in ponderously slow instruction.
(5) Novice models rarely include any meaningful use of evaluation, either formative or summative. Guess that's the last chapter, and they always run out of time and cover it on the last day. Getting people trained in the novice model to actually embrace and use rapid prototyping is a Sysiphean task!
(6) Novice IDers who really think visually and spatially are practically unknown. This is a major learning disability in the world of multimedia.
(7) I concur with you that novice IDers almost never know how to deal with any kind of high-level learning outcome. I think this is a direct result of the almost exclusive use of low-level examples and exercises in training (presumably because they are simple, easy to understand, compact, and simple to develop).
How to deal with these issues is a topic for a whole 'nother ITForum discussion, but I've long believed that the best way to learn the process is from back to front, so you see the dependencies. Also, it's a field where cognitive coaching makes all the difference in the world.
What does the expert model look like? That, too, is a whole 'nother topic, but to get the ball rolling, I'll assert that it includes the ability to think well in abstract symbolic/analytical mode AND visual/spatial mode, and to switch between them quickly and easily. It's really tough to find people who are equally good at both, but the ID expert has to be good enough at both so he/she can at least communicate with the REAL visual/spatial thinkers on the team, and the REAL symbolic/analytic types too.