Tantek Çelik posted his Three Hypotheses of Human Interface Design (via evhead), which I think are good, and wanted to link to so that I could refer back to them myself (in part because I've done some hybrid IM / email / blog application designs in the past, and Tantek uses these in his examples.)

A couple notes to myself:

In general, Tantek is talking about cognitive load exclusively in relationship to a user with a relatively specific and simple goal of accomplishing a largely understood task. He is talking about the load associated with "how," as opposed to any loads associated with where, when, what and/or why.

For example, composing an email may entail a process of deciding, though multiple drafts, what needs to be said and why it needs to be CC'd to one person and BCC'd to another. And, the UI factors that may increase the load in the "how" dimension, at the same time, conceivably could reduce the load in the other dimensions.

from hypothesis #2, about latency (users waiting for the computer):

. . .when the computer is done and the human eventually returns to the task, the human has to remember, wait, what was I doing? This need to actively remember what you were in the middle of is another form of cognitive load.

Yes. And, good UI design also accounts for expected latency. For example, when it's known that a user likely will be waiting while web pages load, one designs elements that allow the user to less actively have to remember where they are (in their task flow).

In fact, one aspect of information design (in the broadest sense) may be considered as providing external memory for the user, where the design becomes the memory of things about which the user is free to forget.

Still, I'm not arguing against #2: there are limits to how much "memory" a design can hold for a person before it becomes disorienting. But, this is where visual design, typography and other perceptual elements (sound!) can be very important, e.g., having the interface between the external memory and the person work with the incredibly efficient perceptual / non-verbal brain.

re: hypothesis #3, more cognitive load means less usability: I think this has general applicability, but people dedicated to specific tasks have different tolerances for cognitive loads in relationship to those tasks. Much of Tantek's discussion centers around email vs IM, but this shouldn't suggest that, for example, one could apply this hypothesis to a comparison of Photoshop and IM and have it work the same way.

Or, reduced to an aphorism: depending on the task, a simple interface is not the same as a useful interface.