Written text unfolds in space. It's visual first of all, before that miraculous convoy of conversions that make it sensible to the reader. It's a seen thing.
Audible narratives, in contrast, extend in time. They are measured in minutes and seconds, not in column inches. And as each new fragment is heard, the old one is dying away, no longer audible.
So, memory is a part of what makes it work: there is no page to scan, the listener must remember the story. Each word makes sense because of the memory of the ones that preceded it.
This work of memory commands closer attention. And that is even more true now, because most of us can read much faster than we can talk (and much faster than we can comprehend spoken language).
I'm not 100% sure I accept Elck's distinctions (I think written texts are also temporal in a sense -- moreso in printed books than on the internet), but I enjoyed following through on the test he proposes.
Elck's recorded his friend Dale's story about hiking, bad knees, and memory. He wants you to listen to it.
Then read the actual story: here. Whether or not this experiment does anything for you, there's no doubt it's a nicely-written story.
Incidentally, this question of voice vs. text is an interesting one in the world of podcasting and audioblogging. On the internet, text and photos are still king. Though I definitely dabble a fair bit in the podcasting world (latest find: Starfrosch), I don't listen to any audioblogs with the degree of seriousness that I apply to textual blogs. But perhaps that might change, as the number of audioblogs grows, and the mechanism by which one downloads and listens improves.
For instance, with text blogs it's common that one quickly scans and samples what folks are writing (either in the MSM or in the blog-world), to see if one really wants to invest 10-15 minutes in a post or an article somewhere. But with audio samples and podcasts, it's usually necessary to download the whole thing before listening. Or, if you listen-while-downloading (that little Quicktime bar that shows up in your web browser), you have to start at the beginning and wait to see if it pans out. Wouldn't it be helpful if there was some kind of software that would produce an immediate compressed, dissociated "blurb" from a larger MP3, to let you sample some phrases, sentences, and sounds -- the texture of the piece? Audioskimming.