Searching the scientific literature

PSYC 11: Laboratory in Psychological Science

Jeremy R. Manning
Dartmouth College
Spring 2026

Why search the literature?

  • You don't want to re-invent the wheel
  • Contextualizing your work makes it more impactful
  • Writing a good discussion section requires knowing what others have found

How to find relevant papers

  • Google Scholar: broad, includes citations and "cited by" links
  • Semantic Scholar: AI-powered recommendations, related papers
  • Reference chaining: find one good paper, then follow its references (and who cited it)
  • GenAI deep research: ask Claude, ChatGPT, Gemini, etc. for an in-depth literature search on your topic

Watch out: GenAI deep research has gotchas

  • Bibliographic errors: wrong authors, years, journals, titles, DOIs — sometimes for real papers
  • Hallucinated references: plausible-sounding citations that don't exist at all
  • Misleading summaries: the paper exists but the AI's description of its methods or findings is wrong
  • Bad taste: AI doesn't always know what's interesting, what's seminal, or which journals/preprints are reliable
  • Download the actual papers — don't trust a citation you've never opened
  • Verify bibliographic info against the PDF (or publisher page), not the AI's summary
  • Multi-pass workflow: pass 1 = research; pass 2 = check every claim against its source
  • For high-stakes work, do manual checks — the more it matters, the more you verify

Evaluating what you find

  • Recency: Is this finding current, or has it been superseded?
  • Source: Peer-reviewed journal? Preprint? Blog post?
  • Sample size and methods: Does the study actually support its claims?
  • Citations: How has the field responded to this work?

Think-pair-share: a meta question (3 min)

  • Find a real paper (using whatever approach you want)!
  • We'll spot check a few of them together

How deeply should you read?

  • Quick skim (30 sec – 2 min): title + abstract + key terms (10–50 papers)
  • Quick read (5 min): abstract + figures + captions + discussion (5–10 papers)
  • NotebookLM (notebooklm.google.com; 20ish mins): drop in PDFs and get summaries, mind maps, and an excellent audio overview. Sits between quick read and deep read
  • Deep read (hours): multiple passes, method-by-method, figure-by-figure (1–3 papers)

Audio overviews and AI summaries can have subtle inaccuracies. For deep understanding, go to the original paper. And practice quick skimming yourself — it's a skill that pays off forever.

The discussion section

  • Summarize what you did and what you found
  • Describe how your work fits in with the broader literature
  • Describe what you think the next steps are
  • Be honest about limitations

This week's lab: literature review

  • Think of an interesting question
  • Find a "template" paper and several related papers
  • Re-factor the template's discussion section, taking the other papers into account
  • Goal: practice synthesizing findings across multiple sources

Lab instructions may be found here

Questions? Want to chat more?

📧 Email me
💬 Join our Slack
💁 Come to office hours
  • Data sleuthing lab writeup deadline has been extended to Friday May 1 at 11:59pm. Note updated rubric!
  • Rest of today: Find your papers
  • Wednesday: Summarize and synthesize
  • Friday: Discuss and brainstorm