digest
Short Summary
Main recommendation: Read papers in unfamiliar fields in layers: first select carefully, then skim (title, abstract, figures), then read deeply with outside resources, and finally read critically across multiple papers. Papers are iterative hypotheses rather than settled truth, so selection and critical reading matter as much as comprehension. For busy reading in a field you already know, switch to a faster red-flag screening approach.
Key Points
- Paper selection comes first: ask a professor for their top 10 papers in a field, check course resources (e.g., Coursera), use NeuroMatch for neuroscience recommendations, or find a recent good review.
- Read in stages for unfamiliar topics: title, abstract, and figures first to confirm relevance, then introduction, conclusion, and figures, highlighting unfamiliar terms to look up.
- For deep understanding, combine the paper with talks by the first or last author, related code (Papers with Code or "Code Availability" sections), blog posts, and discussion with peers.
- Test understanding by explaining the results from the figures alone, then critique the paper by identifying the follow-up study that would make it more convincing.
- She personally likes to compare about five papers on the same topic at once, since authors can report identical results yet interpret them differently.
- To keep up with a heavy literature flow, screen abstracts for red flags (small samples, finicky statistics, objectionable terminology), then check whether results match methods and discussion.
Details Worth Noting
- Concrete instructions:
- Selection: Ask your professor for their top 10 papers in a new field, or look for a review from the last one to two years (five at most) if no recommendations are available.
- First pass: Read the title, then the abstract, then the figures. If the abstract is about something different from the title, drop the paper from your list.
- Unfamiliar field: Read the introduction, conclusion, and figures, highlight unknown terms, look them up afterwards, and keep referring back to the figures while reading.
- Deep understanding: Find the first or last author's conference talks by searching their name; check Papers with Code or the paper's "Code Availability" section; reimplement parts of the method if time allows; search for blog posts on older papers; discuss with a peer.
- Self-test: Look only at the figures and explain the results aloud or to a colleague without the paper.
- Critique: Decide what the next research question should be (for example a larger sample size, a different statistical test, or a more sound interpretation).
- Fast reading for known fields: Abstract first, drop it on red flags, then figures, then methods versus results, then discussion; check that the discussion matches the results.
- Literature note-taking: Keep an Excel sheet with one row per paper and columns for notable elements, themes, and the follow-up question the paper should address.
- Reading habit: Set a dedicated daily time (for example, one or two papers in the library), make reading enjoyable (e.g., with coffee), and consider a reading group for discussion and accountability.
- Caveats: The speaker says the discussion sometimes does not match the results, and cites "trending toward significance" as a big red flag, noting this is a pitfall especially in neuroscience and sometimes psychology.
- Pop science reading: The speaker recommends pop science books for a less dry way of talking about science, while noting they are not always accurate.
- Rate of publication: The speaker estimates about 10 to 15 papers per day appear in their field, managed with Google Scholar research alerts.
Bottom Line
Reading research well is a skill built on careful selection, staged reading, outside resources, and critical comparison across papers, not just finishing the text.
