#206 V̶+V
How does Grace’s direct speech differ from the narrator?
Grace is the narrator of Project Hail Mary, so you might expect his speech and narration to sound alike but there are differences primarily due to register.
Grace has 16,951 tokens of direct speech and 116,065 tokens of narration—nearly seven times as much narration as dialogue.
Words key to Grace’s speech (overused compared to the narrator):
| Speech | Narrator | LL | |
|---|---|---|---|
| you | 429 | 167 | 1,106.1 |
| yeah | 92 | 26 | 261.7 |
| your | 84 | 17 | 259.2 |
| yes | 101 | 40 | 258.9 |
| okay | 107 | 76 | 213.2 |
| oh | 53 | 28 | 121.6 |
| we | 174 | 433 | 107.6 |
| ’re | 109 | 196 | 104.9 |
| do | 308 | 1,039 | 103.9 |
Words key to the narrator (overused compared to Grace’s speech):
| Speech | Narrator | LL | |
|---|---|---|---|
| he | 13 | 1,117 | 216.2 |
| the | 568 | 6,799 | 191.8 |
| say | 21 | 966 | 146.6 |
| his | 2 | 449 | 105.0 |
| my | 77 | 1,335 | 83.5 |
| hand | 4 | 242 | 41.6 |
| she | 19 | 402 | 33.0 |
| head | 1 | 148 | 32.5 |
| screen | 2 | 165 | 31.6 |
Grace’s direct speech is dominated by interaction: you, yeah, okay, oh, we. These are the words of someone talking to another person (usually Rocky).
The narrator is full of third-person reference and physical description: he, his, say, hand, head, screen. Grace-as-narrator describes what other people do and what he sees and touches.
The most striking asymmetry: you appears 25 times per 1,000 tokens in Grace’s speech but only 1.4 per 1,000 in narration. When Grace talks, he’s engaged with someone. When he narrates, he’s alone with his observations.
None of this is particularly surprising but it highlights how word-choice varies by register even for a single speaker.
Log-likelihood (LL) measures how statistically surprising the difference in frequency is between two subcorpora. Higher LL = more distinctive. All values shown are significant at p < 0.001 (LL > 10.83).
Sometimes, I narrate things. It helps me think
07.001