Two records are kept apart: most adopted (what practice uses) and best evidence (what a study measured). An upper bound is what has been achieved; a lower bound is what has not been shown, or cannot be.
The two records
Question
Most adopted
Best evidence
Measured how
Q1 What a base unit holds, and how types derive
DITA topic: title, short description, prolog, body, related links, nested topics; concept, task, reference by specialization (OASIS 2015)
none measured; Kintsch 1978 gives the theory: a type is categories plus rules of order
specification read first-hand; theory
Q2 Does a typed structure help readers
topic types in DITA, Kubernetes 2026b, Red Hat 2026b
Kintsch 1982 (form plus cues helps global comprehension, not local; by abstract); Kools 2008 (headings help search); Jonassen 1980 (mapping helped retrieval, not recall); Horn 1969 (typed format against prose: no significant gain)
the committee's page; the status of DITA 2.0 was not confirmed
map types in Information Mapping, Horn 1969
8
Horn 1969 appendix; later lists differ (against.md)
What the base holds, compared
Base part
DITA topic (OASIS 2015)
Information map, 1969 (Horn 1969)
Text schema (Kintsch 1978)
Kubernetes page (Kubernetes 2026b)
a name or title
required
a title per map
not part of the schema
page title
a short statement of purpose
short description, optional
an introduction block on most concept maps
not part of the schema
overview paragraph
typed parts
body, with sections typed by the derived type
labelled blocks, typed by the map type
the categories of the text type
sections per content type
links to other units
related links
a connections block
not part of the schema
a list of next pages, up to five
units inside units
nested topics
not stated in the report
rules of order, sometimes recursive
none
derivation
specialization, declared in @class
a map type takes every block of another
none declared
none declared
Only DITA declares derivation so that a program can check it.
Information Mapping derives by content, in a chart written for writers.
The text schema is recursive in the reader's knowledge, not in markup.
Upper bounds (achieved)
A typed base with declared derivation and fallback processing is standardised and in industrial use (OASIS 2015; DITAWriter 2026).
Teaching readers several text types raises comprehension at once (Pyle 2017; Bogaerds-Hazenberg 2020) and transfers to untaught types (Hebert 2016).
Conventional form with cues helps global comprehension (topic, main points) and not local (cloze); the abstract gives no effect size (Kintsch 1982).
Typed boundaries help a model retrieve, in one domain (Jimeno Yepes 2024).
Lower bounds (not shown)
No study found compares the same content as untyped text, as DITA topics and as another typed template, on readers.
The one controlled test of a typed format against prose found no significant gain in learning (Horn 1969), and it has not been repeated in this search.
No study found tests the hypothesis that a reader who knows a base type reads a derived type faster or better.
In the 44-study review, taught students no longer outperformed controls at the delayed posttest; interventions with individual practice kept higher delayed effects on comprehension questions (Bogaerds-Hazenberg 2020).
A grammar of types cannot by itself tell a story from a non-story (Black 1979): form is not sufficient to define a kind.