๐บ Survey: what the genre owes its reader¶
Assumes the core, ยง1โ9 of WRITING-STANDARDS.md. This file holds the writing rules that apply to a literature survey, a background section, a related-work section or a "state of the art" alone: what the genre is for, how rare its figures and numbered equations are, and what its reread looks for. The element rules are in figures.md, mathematics.md and tables.md; this file repeats none of them.
๐ฏ What a survey owes its reader¶
A survey's job is to organize a field and locate the gaps in it: not to teach the topic and not to advocate a position. Name which reader first (WRITING-STANDARDS.md ยง1): a thesis reader, a grant reviewer, a paper's related-work section. There are two failure directions:
- Drifting into a textbook chapter. You start explaining the concepts for a reader who doesn't know them. A survey's reader is a researcher entering the area; they need the map, not the lesson. If the user actually wants the lesson, the textbook-chapter writer exists: say so.
- Drifting into an argument. You start defending one approach as correct. Weighing alternatives is the deliverable here; picking a winner is the thesis chapter's job, and doing it in a survey hides the disagreement the reader came for.
Unlike the tutorial genre, alternatives and options are the point. Never collapse them for the sake of a cleaner narrative.
A survey's unit for multi-source synthesis is the paragraph (WRITING-STANDARDS.md ยง11): a body paragraph closes on two or more citekeys wherever the evidence allows.
๐ Figures are the rarest of any genre¶
The comparison table already carries a survey's structural work, so a figure earns its place only for a relationship the table cannot express, such as a taxonomy no comparison row shows, and sits beside the relationship it captures. Absence is the default, not an oversight. When a survey does draw one, the whole of figures.md ยง10 applies to it.
This is also the genre most likely to want a citekey inside a figure,
because a taxonomy naturally attributes each branch. That is exactly the
wrong place for it: the citation gate reads the draft and does not
follow \input, so a key in a node label evades the one check this
pipeline exists for. Attribute in the prose or the comparison table,
where the gate can see the key. A taxonomy redrawn from a source
survey's own figure is the same violation as copying the figure, in
different pixels, whichever notation it is drawn in.
๐ข Mathematics: rare, and mostly unnumbered¶
A survey rarely has much mathematics, but a threshold or an error bound
quoted from a source is a quantity, and
mathematics.md ยง12 applies to it like any other: a
bare `k = 4` renders as typewriter text.
Most of that mathematics stays unnumbered. ยง12 gives a displayed equation a number only if it is standalone, the final step of a derivation, or reused later, and a survey, which reports results rather than proving them, rarely has a derivation at all. Number one when a comparison across sources genuinely needs it: restating a formula two papers define differently before contrasting them, say. The numbered form then has a sentence that refers to it and explains what it shows, like a table or a figure.
๐ The reread¶
Read the draft once as its reader before presenting it (WRITING-STANDARDS.md ยง6). In a survey, check specifically for:
- a term used before it is defined;
- a theme heading that doesn't match what the subsection actually argues;
- a comparison-table row that repeats prose already above it;
- any paragraph whose first sentence doesn't carry its point.