Glossary of reading & study terms

37 terms that come up across these guides, defined in one or two sentences each — the reading techniques, the memory research, the ways of using a source, and the text-analysis machinery behind automatic highlighting.

Reading techniques

Ways of getting through a text, and what each one is actually for.

Active reading: Reading while producing something — a question, a mark, an explanation — rather than letting your eyes pass over the words. The output is the point: it is what reveals the moment you stopped understanding. See active reading strategies.
Skimming: Reading selectively for the overall shape of a text you have not read before, typically headings, first and last sentences, and the conclusion. You know you are skipping most of the words. Distinct from scanning.
Scanning: Searching a text for one specific item — a date, a name, a figure — while deliberately ignoring everything else. Unlike skimming, you already know what you are looking for and roughly what it looks like.
SQ3Ralso: Survey Question Read Recite Review: A five-stage reading method: Survey, Question, Read, Recite, Review. The overhead is worth it on genuinely difficult material and excessive on anything you could follow in one pass.
Three-pass method: An approach to academic papers: a five-minute pass to decide relevance, an hour-long pass for content, and a deep pass only if you intend to build on the work. Most papers should stop after the first. See how to read a research paper.
Subvocalisationalso: subvocalization, inner voice while reading: The inner voice that sounds out words as you read. Commonly blamed for slow reading, but it is a by-product of language processing rather than the bottleneck, and suppressing it tends to reduce comprehension. See does speed reading work.
Saccade and fixation: Reading proceeds in jumps (saccades) between brief stops (fixations) of roughly 200–250 milliseconds, where the actual word recognition happens. Speed-reading systems that promise gains by widening the visual span are working against the narrow area of sharp vision available.
Regression: Moving your eyes back to re-read something. Deliberate regression on a difficult clause is useful; drifting back because attention lapsed is not, and only the second kind is worth training away.

Marking and note-taking

What the different kinds of marking up a text are, and how they differ.

Annotation: Marking a text and writing alongside it. The writing is what separates it from highlighting: a mark records that something caught your attention, a note records why. See how to annotate a text.
Marginalia: Notes written in the margins of a book. The term covers everything from a single question mark to a sustained argument with the author, and historically is often more valuable than the book it sits in.
Cornell methodalso: Cornell notes: A note page split into a wide right column for notes, a narrow left column for questions written afterwards, and a summary strip at the bottom. Its value is that covering the right column turns the page into a self-test. See how to take lecture notes.
Outline method: Notes taken as a nested hierarchy of headings and sub-points. A good default for well-structured material and a poor fit for lectures that jump around, because the structure has to be known as you write.
Highlighting budget: A limit set in advance on how much of a page you will mark, commonly around ten per cent. Without one, marking becomes a running "this seems important" judgement with nothing to compare against. See how to highlight textbooks.

Comprehension

The units of meaning you are trying to extract from a text.

Main idea: The claim a paragraph makes about its topic. Always expressible as a full sentence, frequently not the first sentence, and sometimes not written down at all. See how to find the main idea.
Topic: What a passage is about, usually a noun phrase with no verb — "volcanic eruptions". Distinct from the main idea, which is what the passage says about that topic.
Topic sentence: A sentence that states the paragraph’s main idea explicitly. A convention of instructional writing rather than a rule of prose; real paragraphs often place the claim at the end or imply it.
Supporting detail: Evidence supplied to back the main idea — examples, statistics, quotations. Usually the most memorable sentences in a paragraph, which is exactly why they get mistaken for the point.
Implied main idea: A main idea that no single sentence states, so the reader has to construct it from the details. Writing it out yourself in one sentence is the only reliable way to confirm you actually found it.
Signal phrasealso: transition marker: Wording that flags the structure of an argument — "however", "the key point is", "in contrast". Useful for locating claims quickly, and unreliable on its own, because emphatic phrasing frequently decorates minor points.

Memory and study

Why material does or does not stay, in the terms the research uses.

Retrieval practicealso: testing effect, active recall: Producing material from memory rather than reviewing it. The best-supported finding in the study literature, and reliably more effective than rereading. See how to remember what you read.
Spaced repetitionalso: distributed practice, spacing effect: Spreading review sessions across days or weeks instead of massing them together. The same total effort produces substantially more durable memory, because a slightly difficult recall strengthens memory more than an easy one.
Forgetting curve: The observation that memory for unreviewed material drops steeply at first and then flattens. The shape is robust; the exact percentages often quoted come from one 1880s study of nonsense syllables and should not be read literally.
Fluency illusionalso: illusion of competence: Mistaking the ease of reading familiar text for knowing it. The main reason rereading and heavy highlighting feel productive while changing very little. See does highlighting help you study.
Elaborationalso: elaborative interrogation: Connecting new material to things you already understand, typically by asking why it would be true. Gives you more than one route back to a fact, which is most of what makes it retrievable later.
Interleaving: Mixing different topics or problem types within a study session rather than blocking them. Feels worse and performs better, because it forces you to work out which approach applies instead of being told by the exercise.

Using sources

The three ways to put someone else’s work into your own, and the trap between them.

Paraphrase: Restating one specific passage in your own words at roughly its original length. Requires a citation; does not require quotation marks. See paraphrasing vs quoting vs summarising.
Summary: A compressed account of something substantially longer, keeping the main points and dropping the detail. Typically ten to twenty-five per cent of the original. See how to summarise an article.
Patchwriting: Keeping a source’s sentence structure and swapping individual words for synonyms. The most common way a properly cited piece of work still gets flagged as plagiarism, and it happens because the source was open while writing.
Direct quotation: Reproducing a source’s exact words inside quotation marks with attribution. Appropriate when the specific wording is itself the evidence — a statutory phrase, a disputed definition, a claim a reader would otherwise suspect you of distorting.

How the tool works

Terms from the text-analysis side, for anyone curious what is happening underneath.

Extractive summarisationalso: extractive summarization: Selecting and ranking sentences that already exist in a text. Because nothing is generated, it cannot state anything the source did not. This is what automatic highlighting is. See how automatic text highlighting works.
Abstractive summarisationalso: abstractive summarization: Generating new sentences that describe a text, which is what a language model does. More fluent than extraction and able to introduce claims the source never made — the errors fail in a less safe direction.
TF-IDFalso: term frequency inverse document frequency: A weighting that scores a word highly when it appears often in one document but rarely across documents generally. It is how a ranking system distinguishes a document’s subject matter from ordinary English.
TextRank: PageRank applied to sentences. Each sentence is connected to every other by how much vocabulary they share, and the sentences the rest of the document most supports rank highest.
Sentence segmentation: Splitting text into sentences — considerably harder than splitting on full stops, since abbreviations, decimals and initials all contain them. Everything downstream depends on getting it right.
Stop words: Extremely common words such as "the", "of" and "is", usually discarded before analysis because they appear everywhere and so distinguish nothing.
Stemming and lemmatisation: Reducing word forms to a common root so that "analyse", "analysed" and "analysis" count as the same term. Stemming chops mechanically; lemmatisation uses a dictionary and is slower but more accurate.
Redundancy filtering: Discarding a candidate sentence that largely repeats one already selected. Without it, the top-ranked sentences of a textbook section are often three restatements of a single point.