Scientific writing

Scientific Writing

When revising scientific writing, optimize for reader comprehension on first pass.

Focus on these principles:

1. Make the sentence easy to enter

Open with the topic of the sentence -- usually the thing the reader is already thinking about from the previous sentence.

Good openings often:

• continue the prior subject,
• name the object under discussion,
• or mark a clear contrast.

Do not begin with long setup unless the setup is itself the point.

2. Put the key point where it lands hardest

End the sentence with the new or most important information.

Readers naturally give extra weight to the end of the sentence. If the sentence has a punchline, result, contrast, limitation, or implication, try to place it there.

3. Put important actions in verbs

When a sentence feels heavy or vague, check whether the main action has been buried inside a noun phrase.

Prefer:

• "We compared the models." over:
• "We performed a comparison of the models."

Prefer:

• "This suggests selection bias." over:
• "This provides a suggestion of selection bias."

Do this only when it improves clarity.

4. Preserve the logical thread across sentences

Each sentence should connect backward to the one before it.

At the paragraph level:

• start from familiar context,
• move to the new point,
• then let the next sentence pick up that thread.

A paragraph should feel like linked steps, not isolated observations.

5. Surface the real claim

Do not let the main claim hide behind throat-clearing, vague abstraction, or stacked qualifiers.

Ask:

• What is this sentence actually claiming?
• What is this paragraph trying to establish?
• What should the reader now believe?

Revise so that claim is easy to locate.

6. Define terms before relying on them

Check that all important terms are defined before they are used heavily.

In particular:

• define new methods, models, estimands, assumptions, and acronyms,
• define specialized terms the first time they matter,
• do not assume the reader will infer your meaning from context if the term is central.

If a term is necessary but potentially unfamiliar, explain it briefly in plain language before using it as shorthand.

7. Minimize jargon and explain what must remain

Prefer plain language when it is equally precise.

Do not use jargon just to sound technical. If specialized terminology is necessary, make sure the reader can follow it without getting lost.

A good test:

• would an informed reader outside the narrow subfield know exactly what this means?
• if not, define it or rephrase it.

8. Preserve key details and citations that carry the argument

Do not remove concrete details or citations that help motivate, justify, or support the scientific point.

In particular, preserve:

• key statistics,
• effect sizes,
• sample sizes,
• numerical comparisons,
• dates, ranges, or rates,
• and the citations attached to those claims.

Do not replace informative details with vague summaries like:

• "many,"
• "substantial,"
• "large,"
• or "some evidence"

if the original version gave specific support.

When editing for concision, cut redundancy before cutting informative detail or its citation.

9. Keep scientific caution precise

Preserve uncertainty, but make it informative.

Prefer:

• "These results suggest..."
• "This pattern is consistent with..."
• "Under this model, we estimate..."

Avoid empty cushioning that adds no meaning.

10. Revise at the paragraph level, not just the sentence level

For each paragraph, identify:

• its role,
• its main claim,
• how the sentences are ordered,
• and whether the ending lands on the important point.

If the paragraph has no clear purpose, fix that before polishing sentences.

11. Make sure the reader does not get lost

Continuously check whether a careful reader can follow:

• what problem is being discussed,
• what object is under discussion,
• why a sentence appears where it does,
• and how each paragraph advances the argument.

When in doubt, add brief orientation:

• name the object again,
• state the purpose of the paragraph,
• or clarify the logical transition.

Do not force the reader to reconstruct the argument from scattered hints.

Revision workflow

When editing:

1. Identify the paragraph’s job.
2. Identify each sentence’s topic.
3. Check that all key terms are defined before they carry weight.
4. Remove or explain jargon.
5. Preserve key motivating and supporting details and the citations that substantiate them.
6. Move old/familiar information earlier.
7. Move new/important information later.
8. Convert buried actions into verbs when helpful.
9. Cut opening clutter that delays the point.
10. Preserve technical precision and scientific meaning.

For scientific/statistical writing specifically

Be especially clear about:

• what is being estimated, compared, or assumed,
• what was observed versus inferred,
• what the result is,
• and what its limitation is.

Define technical objects before discussing their properties. Do not simplify away technical distinctions.

Output behavior

When revising:

• prefer minimal edits if the prose is already strong,
• make structural improvements before stylistic ones,
• preserve the author’s meaning and level of formality,
• prioritize clarity of argument over "nicer sounding" prose.