Emotional Register: What Word-Level Affect Can and Cannot Tell You

Two readings describe the emotional weather of your word choices. Emotional Valence is how positive or negative the words are. Emotional Arousal is how activating they are: rage and serene can share a valence and sit at opposite ends of arousal.

Both come from Warriner et al. (2013), a set of 13,915 English words rated by thousands of people on a 1 to 9 scale. The engine looks up each content word you used, averages the ratings, and reports the result. No model reads your sentences.

It reads words, not meaning

This is the limitation that matters most, and it is not a small one. The engine averages word by word, so it cannot see negation, irony, or context.

She was not happy scores as more positive than it means, because happy is in the sentence and not carries no affect rating. What a wonderful mess this turned out to be reads as warm. A grief scene written in plain, unloaded language can score neutral while devastating the reader.

Treat the number as a description of your vocabulary, not of your scene.

Dialogue is excluded

The engine strips quoted speech before measuring. A character’s rage belongs to the character; the reading describes the narration around them. That is deliberate, and it means a dialogue-heavy scene is measured on comparatively little text.

Coverage tells you how much to trust it

Only words in the lexicon contribute. Below 35 percent coverage the reading is flagged as low confidence, which happens most often in secondary-world fiction where invented names and terms make up a large share of the content words. When you see that flag, the number is describing a small slice of your page.

Paragraphs where no word matches at all report nothing rather than zero. Zero would sit off the bottom of a 1 to 9 scale and drag the whole arc with it.

The arc is the more useful reading

A single valence number for a whole piece is close to meaningless. The per-paragraph series is where this earns its place: it shows whether your emotional register rises, falls, holds flat, or swings. Compare that shape against what you intended the scene to do.

A chapter you meant to build toward dread that reports a flat register is worth rereading. Not because the engine is right, but because it noticed something about your word choices that you were too close to see.

Running it

Both readings are part of the Pro metric set. Open a draft in the editor and run Grade this passage, or submit an exercise and read them on the results page alongside everything else the engine measured.