--- name: literature-search title: Find and follow scholarly works with literature_search and literature_lookup description: How to search scholarly sources well with CLIO's literature tools - keyword queries, field and kind, reading merged results, following references and citations, telling preprints from published versions, and reaching readable text. keywords: - level:L0 - literature --- `literature_search` asks several scholarly sources at once and merges the works they return. `literature_lookup` opens one work: its metadata, where to read it, what it cites, what cites it, and its other versions. Read a work's text with `web_fetch` on its `readable_url`. ## Search - Write keywords, not questions: the main terms of the topic and their common synonyms. Run separate searches for distinct vocabularies rather than one long query. - `field="any"` (the default) asks every broad source: OpenAlex, Crossref, arXiv, Europe PMC, OSTI.gov and DataCite. Set `field` only to narrow a search to one area's sources. - A source missing from the results is named in `notes` with what happened and what to do; the other sources' works are complete. - Set `kind` to `preprint`, `review` or `dataset` when that is what you need; `after`/`before` bound the publication date. - `found_in` lists the sources that returned a work. Several sources agreeing is a sign of a well-indexed work, not of quality. - `payload_file` holds every work found with full abstracts; read it before searching again with a similar query. ## Follow - Use `literature_lookup` with a DOI, arXiv id, `pmid:` or OpenAlex `W` id. - Follow references backward for the foundations of a result and `cited_by` forward for replications, corrections and later work. A highly cited paper's recent citers often say whether its claim still holds. - `versions` links a preprint to its published version. Cite the version you actually read; when both exist, check whether the published one changed the claim. ## Read - An arXiv `readable_url` is the abstract page; the full text is usually at `https://arxiv.org/html/`, and the PDF at `/pdf/`. - When the readable URL is the DOI resolver and the publisher page is paywalled, look for an open copy: the preprint, a repository version, or the author's page. - Record what you actually read for each work: full text, abstract only, or metadata only. An abstract supports a claim only as far as it states it.