Quantifying multi-track novels

I’ve been obsessed for years with novels that alternate between two (or more) time periods of the same story. Without seeking them out intentionally—I like not knowing much about a fiction book before reading it—almost everything I read last summer had that structure:

• Now Is Not the Time to Panic: The main timeline tells a story from when the narrator was a teenager, with flash-forwards to her being confronted about it by a journalist as an adult.

• The Brief Wondrous Life of Oscar Wao: Stories by or about Oscar and other family members that move back and forth across generations.

• The Night Circus: Each chapter opens with a specific date, helping the reader track a beginning, middle, and end of the story as they’re told out of order.

• Americanah: Two stories of immigrants settling in America and the UK interleaved with substantial chapters flashing back to their time together in Nigeria.

• A Visit from the Goon Squad: A book of short stories that jump around different times in the same characters’ lives.

• The Overstory: Several characters’ stories are told in alternating chapters until they all come together. (Admittedly this one is more about different strands in space than in time, but it has a similar effect.)

When I first wrote about multi-track stories, I hypothesized that they’re popular because all the linear stories have already been told: playing with time increases possibilities exponentially, allowing an author to create a new shape. But that relied on my anecdotal observation that multi-track stories are a recent invention. Is that true?

I can now answer that question without reading hundreds of books myself—just ask large language models to research a representative sample of novels and quantify which ones jump around in time.

I gave ChatGPT, Claude, and Gemini two lists: the New York Times’ 100 Best Books of the 21st Century (removing 30 non-fiction books) and Penguin’s 100 Must-Read Classic Books. I gave each the same prompt explaining what I meant by multi-track novels, asked them to classify each book from both lists, and followed up with clarifications to my criteria if they were wrong about any of the books I’d read.¹

According to their analysis, multi-track novels are more common today, but not dramatically so:

• ChatGPT says 14% of the modern list fits, and 7% of the 20th century list

• Claude (24% vs 19%) and Gemini (26% vs 15%) are looser on the classification overall, but the relative change is in the same ballpark

The full book-by-book results are here (tell me if you see anything that looks wrong). I expected the multi-track novels to be much rarer historically, and perhaps a bit more common in modern times.

1. The original prompt:

Call a book MTN = YES if it meets all of these:

1. Braided structure (interleaving): The book alternates among two or more strands throughout (A/B/A/B...), not just “Part I then Part II,” and not just a few flashbacks.

2. Not the same story from multiple POVs (low overlap): The strands do not mostly cover the same events/scenes from different perspectives.

   e.g., if it’s t=1, character=A, then t=2, character=B, then t=3, character=A, etc. then it doesn’t count

   but if it’s t=1, character=A, then t=10, character=B, then t=2, character=A, etc. then it counts

3. Each strand has real weight (no token frame): No strand is just a thin frame or occasional inserts.

   Rule of thumb: no single strand > ~90% of the narrative.

4. Time relationship: Reinforcing from above, strands may be from the same story if they are non-linear in time.

Quick summary

• YES: multiple distinct stories (or one story from different points in time), braided together.

• NO: one story told linearly via multiple perspectives, or mostly linear with very rare flashbacks (~10% or less), or a token frame.

Full conversations: ChatGPT (5.2 thinking), Claude (4.5 Sonnet), Gemini (3 Thinking)