I've written novel-sized plots in 5,000-word stories, and I'm happy to posit on the flip side that Rocket Science can be considered to be a 65,000-word short story
Schwartz goes on to remark:
The implication of talking about 5,000-word novel plots and 65,000-word short stories is that the one is generally a boiled-down version of the other; that short stories are condensed and novels are bloated. I'm simplifying, but that's how I read this post.
While I find his reading of my original comment to be reasonable given what I wrote, it wasn't my intention to assert that the essential difference between short fiction and novels was simply length. This is a much more complex problem, which can be broken into two portions.
The first portion is definitional -- a characterization of the forms themselves. This is the basis of award categories we all know and love, to wit, short story, novelette, novella, and novel. There's lots of room for variation in those categories depending on how you want to slice it. I'm pretty sure I've addressed this before, though I currently can't find the link, but to belabor the obvious, the wordcount boundaries which characterize the classic categories are arbitrary, quantitative descriptions which are embraced for the sake of simplicity. (And the sanity of award committees.)
It's more useful but far less precise to have a structural approach to analyzing the forms. For example, one can characterize flash fiction has emphasizing only a single element of the stereotypical Western story -- character in a setting with a problem, attempting multiple solutions with increasing levels of risk and failure, before achieving resolution, followed by reader validation. A flash piece might begin, "The cop tripped over the body in the shadows behind the bar." There you have character, setting and problem in one sentence, albeit much presented by implication. A flash piece could approach this as an expansion on one of those elements, or as a craft piece, focusing on description, dialog or a specific kind of detail.
Moving this model forward, a short short (let's say quantitatively falling between 1,000 and 7,500 words) can expand from the single-element approach of flash to encompass an entire story arc. Typically at this length the arc will not include subplots, feature only a single protagonist, and so forth. (I am keenly aware that exceptions abound, I'm merely staking a position here, not emitting my concept of Received Wisdom.)
Novelettes (7,500 to 17,500 words) introduce multiple plot arcs, subplots, strong secondary characters and even multiple protagonists. Digressions begin to appear, alarums and excusrions propagate. And so the complexity continues up the scale through novellas (17,500-40,000), short novels (40,000-65,000), novels (65,000-125,000) and long novels (125,000+).
The key here is an expanding sense of structure. Flash is almost by definition a one-trick pony, a trickling stream of narrative. Short stories expand on that, novelettes widen it further to a river, until you get the oceans of novel. (Likewise the scope and scale of the ideas that drive the story often expand, though certainly novels have been written about buying shoes or dunking a pastry in a cup of coffee.)
I don't think anything here should be surprising to anyone. We all subscribe to a quantitative categorization because there's no other simple, shorthand approach to talking about our output. These word count-based definitions have marketing significance and they have critical significance, and so we live with them.
Here's where I inadvertently created confusion, and thus sparked some portion of snurri's remarks. What I meant by "novel-sized plots" was something on the order of "stories of broad scope and scale with multiple protagonists, subplots and secondary characters playing significant roles." Not "long stuff boiled down."
It's easier to examine at the other end of the scale. Look at Rocket Science. There's a single POV working in a short, linear timescale with a fairly abbreviated cast of characters. He has multiple try/fail cycles, but the arc of the book is a single curve of rising action which reaches a climax. That's what I meant by "65,000-word short story." There's enough material there to make it be a book -- the plot becomes inordinately complex in the final act of the narrative, in fact -- but structurally it's highly reduced, without the intricacies and flourishes which characterize many if not most genre novels. This simplified structure was sometimes seen in the classic juveniles from a few generations ago, though it seems to be rather rare now. (Feel free to correct that assertion in comments.)
One can do the reverse as well, packing big ideas, complexity and flourishes into relatively few words. This isn't terribly common, because to do novel-sized plots usually requires novel-sized word count, but it is quite possible. The key isn't the word count per se, at any point on the scale, the key is the structural complexity and idea sizing.
All that being said, I find it curious that within my own writing process I am usually very aware of my likely final word count when I'm writing a story. (I haven't written enough novels to be sure there...ask me in two years.) This is a function of story sizing, the shape it occupies in my head, and may be indirectly connected to my concept of "span of control."
As for the issue of people being naturally novelists or short story writers, I suppose it depends on the prolixity or laconicity of one's natural voice. If a writer's ideas come out of the chute with ramifications pre-installed, it's a hell of a job to pack them into 4,000 words (apparently a near-magical length for short stories). If a writer's ideas get built up from seeds, then working in short fiction might be a natural launching pad.
What do the quantitative categories of fiction mean to you? Are they relevant? Do you have a better model than this to talk about length?