The Mystery Box Opens Onto Another Mystery Box

The footnotes give it away before anything else does. A note in the introduction credits "a comment made by Professor Sophie Scott during review of manuscript, August 2018." Another cites a Wall Street Journal piece from March of that year. Whatever the catalog says about this book's age, it was written in the last gasp of the mystery-box decade, the years when J.J. Abrams's 2007 TED talk about the pleasure of an unopened box was still being cited as wisdom rather than diagnosed as a business model. Storr cites it too, approvingly, alongside George Loewenstein's 1994 paper on the psychology of curiosity — the one that found brains become "spontaneously curious when presented with an information set they realize is incomplete." He uses this to explain why Agatha Christie withholds the killer's name. He does not mention that by the time he was writing, the same paper had already been cited, just as approvingly, in trade-press explainers about how to write an Upworthy headline.

That's not a gap the book fell into later. It's a gap it was already standing in. The "curiosity gap" as a named, exploitable technique in digital publishing dates to 2013 and 2014 — Storr's manuscript was still years from print. So this isn't a case of innocent craft advice getting weaponized by an industry that came after. The industry was already there, citing the same psychologist, when the book went to press. Storr's index and a growth-marketing deck from the same year would have several sources in common. The book doesn't warn about this coincidence because it doesn't experience it as a coincidence — a novelist withholding the location of the body and an editor withholding the second half of a headline are, to Loewenstein's grid-of-squares experiment, doing structurally identical things to the same reward circuitry. Storr knows this. He just assumes the reader will use the technique to write The Remains of the Day rather than to headline it.

There's a section on Citizen Kane's search for "rosebud" that Storr uses to illustrate what he calls the hero-maker's confabulated origin story — the tidy, neocortical cause-and-effect narrative a mind builds about its own damage, whether or not it's true. He calls this confabulation. The word has since been repurposed. It now describes what large language models do when they generate a fluent, structurally sound answer that happens to be false — "hallucination" in the current jargon, though the older, more precise word for a model producing a convincing account with no reliable relationship to what actually happened is the one Storr uses about human beings. He wasn't writing about machines. But the chapter reads like a spec for why a system trained to predict a plausible next sentence would produce exactly the kind of unreliable self-narration he's describing in Charles Foster Kane — confident, internally coherent, and wrong in a way that only shows up when someone goes looking for the sledge.

I've written before, about other books, that a diagnosis gets absorbed by the system it names, that the warning becomes the manual. This book resists that shape, because it was never a warning. It's the manual, plainly, gladly, for novelists and screenwriters, and it doesn't pretend otherwise. The interesting failure isn't recuperation — there's nothing to recuperate. It's that the manual and the growth-hacking playbook were assembled from the same footnotes at the same historical moment, by people who didn't know they were colleagues. Storr's chapter on Loewenstein and the Upworthy headline writer's chapter on Loewenstein could have been filed under the same citation without either party noticing they'd met.

What would this book need to say about itself, now, to close that gap — and does the fact that it doesn't say it change what it's for?