Skip to content

Annotating

Principle · All principles
Evidence · 25 claims (9 for, 13 mixed, 3 against) · 33 studies (13 causal, 7 review, 6 quant-synthesis, 3 associational, 2 design, 1 qualitative, 1 theoretical), q1–q4 · 7 of 33 report an effect size · 16 claims rest on one study

Conditional relationship

For a reader working through a text they will need to use later (an expository, disciplinary or argumentative passage, read alone or before discussion or writing), annotating means marking the text while reading: highlighting or underlining, margin codes, notes, questions, summaries. The relationship this page states has two parts, and the evidence for them is very different.

  1. Marking (highlighting, underlining). Marking directs attention to the marked material. Later performance on test items about that material may improve. Items about unmarked material do not, and neither does comprehension of the passage as a whole. The wiki's evidence here is for immediate or short-delay tests (immediate-performance, some delayed-retention), mostly with university students (adult). What matters is which content gets marked, more than who marks it.
  2. Elaborative annotation (notes that paraphrase, justify, question or summarize). The expectation is that such notes support comprehension (conceptual-understanding) and later use in discussion or writing, because they require the reader to select, explain or question. No claim in the wiki tests learner-written margin notes against reading or against highlighting. The nearest evidence is for neighbouring activities that are not annotation: trained question generation during reading, and self-explanation. This part of the relationship is a transported proposal, not an observed effect.

Learner state conditions both parts. A reader with no criteria for what is relevant (novice for this text and task, whatever their general reading level) may mark the wrong material, and marking then shifts attention toward it. The relationship is stated for a learner-made annotation with a follow-up use. Instructor-supplied highlighting is a different intervention and is noted under the evidence. Learners annotating together in a shared text is a configuration of this principle, described at Social Annotation. The element is Text Underlining and Annotating. No converted pattern sits beside this principle.

Fitting the design to a situation

Ask for three facts when the brief omits them: what the reading is for (recall of specific content, comprehension of the whole text, or a later discussion or essay); who the readers are (school or university, and whether they know enough about the topic to judge what matters); and who will mark the text (the learner, or the instructor in prepared materials).

If the brief says… Then change… Basis
The goal is recall of specific content from a text, tested soon after Learner highlighting can stay, but tell readers what the reading is for and check one marked page early; do not expect it to help with understanding the whole text Learner highlighting gives a small memory benefit but no reliable comprehension benefit [~M] (meta-analysis: 0.36 on memory, 0.20 on comprehension)
The goal is understanding the whole text, or making inferences across it Replace free highlighting with one written question or one-sentence summary per section, and teach question generation first (model it, then practise it) Highlighting shows low utility [-M] (one primary study found underlining hurt inference questions); Self-questioning improves comprehension [+M] (taught question generation; median 0.36 on standardized tests)
Learners are school students (pre-k-12) Do not assign unguided highlighting. Supply the key sentences already marked, and add a task that uses them (answer, explain, or ask a question about each) Learner highlighting helped college students but not school students; instructor-supplied highlighting improved memory and comprehension [~M] (0.24 for school students; 0.44 for instructor highlighting)
Learners are novice on the topic and have no criteria for what matters Give the reading purpose and three to five margin codes (?, claim, evidence), mark the first paragraph together, and check that the marks fall on the main claims before readers go on alone Test performance follows whatever is underlined, relevant or not [~M] (67 college freshmen, immediate test); the codes and the check are untested proposals
The instructor can mark the text for learners (prepared readings, slides, a pre-highlighted edition) Mark only the content the outcomes and the test address; add headings and an overview; leave incidental detail unmarked Student-generated underlining was not significantly better than experimenter underlining [~M]; Signaling essential content improves transfer from multimedia lessons [+M] (6 of 6 experiments, median d = 0.52)
Learners are struggling readers in middle or secondary school Use a close-reading routine across several weeks: annotate a short text, reread it, answer text-dependent questions, discuss Scaffolded close reading improves comprehension for struggling readers [+M] (a package; annotation is one part and its share of the effect is not known)
Readers are asked to explain the text in their own words Prompt a short explanation at each section ("why does this follow?", "how does this connect to the last section?") written in the margin, not a copied phrase Prompting learners to self-explain improves understanding [+M] (g = .55, immediate tests; prompted explanation, not margin notes as such)
The reading feeds an essay written from several sources Do not require a summary of each source as the annotation; let readers reread, and if they do summarise, check the quality of the summaries before the essay Summarizing each source did not beat rereading twice for integrated essays [~M] (295 adults; better summaries went with better essays)
Learners annotate or take notes on laptops or tablets Ask for paraphrase, questions and links rather than copied text; do not ban devices on the strength of this evidence Laptop note-takers transcribed more verbatim; a replication found no consistent difference [~M] (lecture notes, not text annotation)
Setting is online-self-paced with a digital text Put a question box after each section of the reader instead of relying on a highlighter, and show the learner's own questions back to them before the end-of-unit check untested proposal
A seminar, discussion or writing task follows the reading Ask for two annotated passages to bring as evidence, and begin the discussion from them; check afterwards whether learners used their notes untested proposal; the wiki has no claim on whether annotations used in a later discussion or essay improve it
Only one session Give one reading purpose, have learners mark and annotate a short passage, then use the marks at once (pairs compare and answer two questions); do not promise lasting gains untested proposal

Observation, state and explanation

An annotated page is an observation: what was marked, how much, in what form, with what written content. Comprehension, monitoring and "deep processing" are inferred states, and the page itself does not show them. A densely marked page can come from careful selection or from marking almost everything. A sparse page can come from confident understanding or from disengagement. Record the task purpose given, the prompts or codes supplied, whether annotation was modelled, the reader's prior knowledge of the topic, and the follow-up task. Then interpret the annotations against a separate probe of comprehension.

Response Plausible explanations Observation that could change the interpretation
Most of the text is highlighted; few or no written notes No criteria for relevance; text too difficult to judge importance; marking used as a sign of effort to be seen Ask the reader to name the three most important ideas and say why. Compare with a passage on a familiar topic. Ask what they thought the marking was for.
Marks fall on peripheral detail; main claims unmarked Misreads the text's structure; purpose of the reading unclear; prior knowledge directs attention to familiar detail Give the reading purpose explicitly and re-observe on a matched passage. Probe structure separately (e.g. "what is the author's main claim?").
Margin notes copy phrases from the text Does not know what an elaborative note looks like; text beyond current comprehension; time pressure Show one modelled paraphrase or question and observe the next section. Check comprehension orally without the notes. Vary the time allowed.
Rich notes and questions, but weak performance on a later inference or writing task Local understanding not integrated across the text; notes never used afterwards; the task demands something the notes did not target Compare a section-level probe with a whole-text inference item. Check whether the reader returned to the notes during the follow-up task.
Annotates well with prompts or codes, little without them Prompt-dependent strategy; prompts supplied the relevance judgement; no reason to annotate when nothing follows Fade the prompts over matched texts and record what changes. Ask whether the reader annotates for their own purposes outside class.

These rows are proposals. None has been tested with learners, and a single probe can shift confidence between explanations but does not identify a cause. A probe is itself an exposure that can change the reader's state, so record every one.

Evidence and qualifications

  • Learner highlighting of text gives a small memory benefit but no reliable comprehension benefit, and evidence on other forms of annotation is not yet recorded here [~M]: the claim's title is broader than its evidence. Both entries concern highlighting or underlining, not written annotation. A meta-analysis (q4, 36 articles, 85 effect sizes, college and K-12 students, read from the abstract only) compared reading with and without being asked to highlight. Learner highlighting improved memory for the text (0.36) but not comprehension (0.20), and helped college students (0.39) but not school students (0.24). Highlighting supplied by the instructor improved both memory and comprehension (0.44 each). The second entry, a review, rates highlighting low utility. The claim does not test margin notes, questions or summaries, and its own Discussion says so. Not yet checked against its sources.
  • Highlighting shows low utility for improving learning outcomes [-M]: a narrative review (q3, review · r3, no pooled effect size) of highlighting and underlining across Air Force trainees, children and undergraduates, with immediate and delayed tests (up to two months). Most of its primary studies found no benefit over simply reading. One found that underlining hurt performance on inference questions requiring connections across the text and left factual recall unaffected. This limits part 1 to marked content and recall, and is a reason to expect that unguided marking will not improve comprehension. Not settled: the abstract available could not confirm the entry.
  • Relevancy of emphasized text directs attention and influences test performance [~M]: one experiment with 67 college freshmen, who read passages with underlining on high-relevant or low-relevant content. Test scores rose on whichever content was underlined (F(1, 49) = 13.56; F(1, 40) = 6.98), regardless of its instructional importance. Testing was immediate. The i2 code is not backed by a printed effect size, only F and p. This supports part 1's mechanism, and also its cost: marking the wrong content shifts performance toward it. Not yet checked against its sources.
  • In one experiment, students who generated their own underlining did not score significantly higher on tests than students given experimenter-generated underlining [~M]: in the same study, students who did their own underlining did not perform significantly better than students given the experimenter's underlining. The claim's title says "as effective as", but a non-significant difference does not establish equivalence, and this entry reports no n. Read with the relevancy claim, the evidence suggests that which content is marked matters more than the act of marking. It does not test elaborative annotation. Not yet checked against its sources.
  • Self-questioning improves comprehension [+M]: the nearest evidence for part 2, though the activity is not annotation. Students in a review of intervention studies were taught to generate their own questions during reading. Comprehension on new material at the end of training improved, with a median effect of 0.36 on standardized tests and 0.86 on experimenter-developed tests (abstract only). A second review of 35 experiments with K-12 students, with and without disabilities, reports the strategy effective, with no effect size given. It supports question annotations only if they are taught and practised as question generation. It does not show that writing questions in the margin, untaught, has the same effect. Not yet checked against its sources.

Across these, outcomes are mostly immediate or short-delay tests of the passage read, and the populations are mostly university students. Do not read a highlighting null as evidence against written notes, or a question-generation gain as evidence for margin notes. They are different activities. The wiki has no claim that compares annotation formats directly, or that reports whether annotations used in a later discussion or essay improve it.

The evidence by question

Each question below links a general claim that gathers the studies behind it into one table, one row per distinct study, with each study's design and effect as recorded and no averaged verdict (eval/synthesis/annotating.json, scripts/synthesize_claim.py). The count is of distinct studies across the claim and the narrower claims that bear on it.

Further evidence, not yet read against this model

Claims this page cited before it was rewritten as a conditional model. They are evidence about the relationship, but none has yet been re-read against the model above; each says how far its own sources have been checked.

Objective and learner-valued goal

Ask what the reader wants from the text and from annotating it, separately from the designer's objective. The two may agree, diverge, or be unknown; record which. Compliance with an annotation routine is not evidence that the reader values it. Then choose the outcome the annotation is meant to serve and the instrument that shows it: recall of marked content, comprehension of the whole passage, an inference across the text, or the quality of a later discussion contribution or essay. These need different evidence, and the annotated page is not that evidence.

For example, a nursing student may value getting through a long clinical guideline quickly and marking "what will be on the exam". The designer wants the student to be able to explain why one recommendation applies to one patient group and not another. Heavy highlighting serves the first aim and, on the evidence above, probably not the second. A design can keep both: let the student mark, then ask for one margin question per section about when the recommendation applies, and check that capability with a case item, not a recall item.

What would revise this model?

Part 1 should weaken if well-selected learner highlighting fails to improve performance on the marked content in comparable readers, or if it improves comprehension of the whole passage, against the meta-analytic and review evidence above. Part 2 is the larger gap. It would gain support from a comparison of learner-written elaborative annotations against reading, or against highlighting, with comprehension or transfer outcomes and a stated delay. It would lose support if such a comparison found no difference once reading time is matched, or if trained question generation helped while margin annotation did not. If prompts or modelling explain the whole gain, the active ingredient is the prompt, not the annotation, and the model should say so. Do not protect it by attributing every null to poor annotation quality after the fact.

Claims

  • Note-taking — annotation is note-taking anchored directly to a source text
  • Evaluating Sources — annotation can focus attention on evidence, credibility, and bias
  • Inquiry-based Learning — annotations help learners generate and refine questions from reading
  • Purposeful Reflection — reviewing one's own annotations supports metacognitive reflection after reading

Examples

  • Social Annotation — learners annotate together inside a shared digital text and respond to one another's questions and interpretations
  • Assigned Readings with instructor-provided annotation prompts before seminar or discussion
  • Pre-Seminar Preparation — annotations can serve as evidence banks for later dialogue
  • Margin codes such as ?, !, claim, evidence, and connection can provide lightweight structure for novices

Key Sources

  • Pressley, M., & Afflerbach, P. (1995). Verbal protocols of reading: The nature of constructively responsive reading. Routledge.
  • Kintsch, W. (1998). Comprehension: A paradigm for cognition. Cambridge University Press.
  • Fisher, D., & Frey, N. (2014). Close reading and writing from sources. International Reading Association.
  • Liu, J. (2006). Assessing student critical thinking through annotation. [Cited via secondary source without full publication details]
  • Arduini-Van Hoose, N. (2020). Critical thinking. In Educational psychology. Retrieved from https://edpsych.pressbooks.sunycreate.cloud. CC BY-NC-SA 4.0.
  • Oster, L. (2001). Using the think-aloud for reading instruction. The Reading Teacher, 55(1), 64-69.
Page metadata
Field Value
type principle
id annotating
title Annotating
description For readers of a demanding text, marking it may improve later performance on what was marked, while gains in comprehension are expected only from annotations that make the reader select, explain or question, and remain largely untested.
status review
canonical True
generated by claude/unspecified on 2026-10-02
trust unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.