DEV Community

Cover image for Finding tense switches in a novel — and deciding which ones matter
Matthias Fahsold
Matthias Fahsold

Posted on Originally published at mfahsold.github.io

Finding tense switches in a novel — and deciding which ones matter

Finding tense switches in a novel — and deciding which ones matter

Something feels off in chapter twelve. You read a few pages and there it is: the scene slipped from past into present. Was that on purpose, or did it just happen? On page 40 you can tell. On page 400 you cannot.

A tense break is rarely a typo

The base tense of a novel is a choice. Past tense gives you a narrator looking back. Present tense puts the reader right inside the scene. The historical present adds pressure. A break between them is not a grammar mistake — it is a signal to the reader. The question is only whether you meant to send it, and whether it arrives at the right moment. Reading an entire manuscript just to find those moments is slow, which is why editors work with passages instead of impressions.

What Lixity shows you

Lixity reads your manuscript on your own computer and gives every paragraph a simple tense reading: present, past, mixed or neutral. Paragraphs that switch or noticeably mix tenses also get a small strength rating from 0 to 3.

Out of that come two lists, and they answer different questions:

  • Flagged passages (rating 2 or higher) are your short list: the paragraphs worth a second look.
  • The chapter line tells you the surrounding norm — the tense a reader experiences as the "base" of the scene. Without it, a single switch has nothing to stand out from.

One honest sentence: these readings come from verb forms and markers, not from a grammar parse. Dialogue, nested clauses and reported speech all shape the counts. That is exactly why the last step is always a person reading the passage — not a number.

The review loop, step by step

  1. Install and open. uv tool install --python 3.12 --managed-python lixity, then lixity service start --open and choose Open Project for your manuscript folder. (Windows, macOS and Linux are covered in the installation guide.) If you just want to see the loop first: the demo report opens in your browser, no install.

  2. Start with the review panel. Start your review offers Open passage (the first paragraph above your threshold, in reading order), Explore chapters and Open sentence rhythm.

  3. Work the flagged list. The Flagged passages panel lists every paragraph at or above your threshold, sorted by strength and line. Click a row and the paragraph opens, the page scrolls, the passage flashes — with the "flagged only" filter set, so the rest of the page gets out of your way.

  4. Read the context before you judge. Inside the opened paragraph, Values shows the source lines and the measurements. Then look at the scene: is the switch carried by a letter, a memory, an interior monologue? The chapter line in the header shows you the norm you are comparing against.

  5. Write your decision where you read it. The marker buttons (+ To-do, + Check, + Attention) leave a review note in the manuscript as a Markdown comment. It stays invisible while you read and shows up later in the Work markers panel.

When a switch is on purpose

A high rating measures a change, not a mistake. These cases are usually intentional:

  • Letters and documents. A quoted letter often shifts into present or into a different register; readers expect the frame to change.
  • Flashbacks and frame stories. Moving from present to past may be exactly what the structure asked for.
  • Free indirect speech. A character's inner voice can pull the tense along — effective, but easy to overuse.
  • A scene turn. One present-tense paragraph at the end of a past-tense chapter can be a strong, deliberate accent.

The flagged list decides none of this. It makes sure no unintended break hides in a long manuscript — and that the intended ones are visible enough to be a choice.

What this can and cannot do

Verb-form patterns miss irony, ellipsis and unusual syntax. Very short chapters give noisy estimates, which is why Lixity now marks them "not comparable" instead of ranking them. Dialogue-heavy scenes reflect speaker turns, not narrative tense. The flag_min_severity setting (1–3) controls how strict your short list is; start at 2. Lixity never rewrites anything — it is a flashlight, not a judge. The interpretation guide and the methods document explain the details.

Try it without installing

The demo report is made from invented text and opens in your browser. The Effi Briest and Pride and Prejudice examples are complete analyses of public-domain novels. For your own manuscript: install, open the project, and review one chapter — the loop is the same at any length.

Lixity home · Blog · Report a reproducible issue · Legal notice

Top comments (0)