I'll preface this by saying this is an off-the-top-of-my-head post. I've been thinking about this off and on for some time now, but thought I should dump a snapshot into the online brain for safe keeping.
I conceive of learning objects as being analogous to molecular compounds. They are composed of atomic units (assets, or elements), and can be used to perform a highly specific role in the compound state (learning object), or broken down into the atomic/elemental state to be used as raw materials for a new compound (learning object).
Describing learning objects as words in a sentence oversimplifies the atomic bits (images, text, audio) as being interchangeable. It de-emphasizes context, and the value of strategically combining assets/elements to produce more elaborate constructs. It makes for nice examples, because lots of people use words and sentences, but I don't know that it captures the real value of the concept of learning objects (if there is any).
Words and sentences also imply linear order. You read from left to right (or up and down, or right to left, depending on language), but there isn't really branching or interactivity. The conclusion of a sentence does not depend on the interaction of the previous words, as much as on a predetermined sequence of concepts.
Yes, a word can be used in a different sentence, but the different context colours the interpretation of that word such that it may have a slightly different meaning. That's great for a word that is just a collection of letters, but if a word is a video (or animation, or website, or whatnot), the different context may not have any meaning. We need to be able to deconstruct the learning object and build it back up to take advantage of the new context, in order for it to have the proper meaning (if we keep following this sentence analogy).
This is where I see the next generation of "learning object repository" applications being applied - in managing both the atomic/elemental assets, as well as providing tools to facilitate the construction and deconstruction of more complex compound constructions. This is where both APOLLO and Pachyderm are heading (although neither fully addresses deconstruction yet).
The strategic and intelligent construction of these compound learning objects could also be done along the lines of Michael Feldstein's learning experience objects.
The molecular analogy can be extended ad infinitum - loosely bound tools acting as enzymes to catalyze reactions that build or break down molecules, etc...
- David Wiley's "Learning Atoms and Learning Crystals" paper (written in 1999 - he's been waay ahead of the curve on this one...)
- Michael Feldstein: Re-usable Learning Content Objects or Re-Usable Learning Experience Objects
- Stephen Downes' response to that article
- Michael Feldstein's Learning Objects Aren't Legos (part 1, part 2)
- Michael Feldstein: Pattern Languages as Human Languages
- Ulises Ali Mejias' Learning Molecules white paper (which mentions a patent... wtf? going to have to check this out... (Thanks to Brian for the link - hadn't seen this before)