The short version. You are not expected to read every assigned article front to back like a novel — nobody in the discipline does that. You are expected to extract the argument, the evidence, and the weakness. This page explains how, and how to judge whether a source is worth your time at all.
How to read an academic article
Never start at the beginning and read straight through. That is how you read a novel, and it is the slowest possible way to read research. An academic article is not written to be discovered in order — it is written so that a reader can find the argument quickly and then decide how much of the machinery they need.
So read in passes, starting with the parts that tell you the most per minute. Most articles need only the first two.
Pass one: five minutes — abstract, introduction, conclusion
Read the abstract, then the introduction, then the conclusion, then skim the section headings. Nothing else, and in that order. These three pieces are where authors state their argument most plainly, and the small differences between how the argument is stated in each are often revealing — the introduction promises, the conclusion delivers, and the gap between them tells you where the paper struggled.
By the end of five minutes you should be able to say what question the article asks, what it argues, and what kind of evidence it uses. That is frequently all you need. And if the article turns out not to be relevant to your paper, you have spent five minutes instead of an hour — which is the real reason to read this way.
Pass two: thirty to sixty minutes — the whole paper
Now read it through, but read knowing where it is going. Because you already have the argument, you can read each section asking what it contributes rather than trying to work out what is happening.
Look at the tables and figures properly — in quantitative work the tables usually carry the argument and the prose describes them, so a table you skipped is an argument you missed. Skim the technical apparatus unless you need it. Read the footnotes: in political science, the objections the author anticipated are very often buried there.
As you go, write down four things: the question, the argument, the evidence, and the weakness. Four lines, in your own words. If you cannot write the fourth line, you have not finished reading.
Pass three: as long as it takes — the machinery
Reserved for the handful of works you are building on directly — the ones you would replicate, extend, or argue against at length. Here you work through the method in detail, check whether the evidence actually supports the claims, and follow up the sources the author leans on hardest.
The anatomy of an empirical article
Most quantitative and mixed-method articles in political science follow the same structure. Knowing what each section is for tells you what to look for and what you can safely skim.
Abstract
What to get from it: the question, the finding, and the type of evidence, in about 150 words. What to watch: abstracts are written to sell the paper. The claim here is routinely stronger than what the results section supports. Treat it as a promise to be checked, not as a summary to be trusted.
Introduction
What to get from it: the puzzle, the argument, and what the author says is new. A well-built introduction also tells you which literature the author thinks they are in conversation with, which locates the paper for you instantly. What to watch: the last paragraph is usually a roadmap of the paper — read it and you have the structure.
Literature review
What to get from it: the state of the debate, and — if you are early in a research project — a reading list. The sources an author engages with most closely are the ones you should read next. What to watch: reviews are constructed to make the author's contribution look necessary. Notice which alternative explanations are described generously and which are dispatched in half a sentence; the second kind is often where the paper is vulnerable.
Theory and hypotheses
What to get from it: the causal mechanism — not just that X affects Y, but through what steps and actors. This is the section that most repays slow reading, because it is where the actual intellectual claim lives. What to watch: whether the hypotheses genuinely follow from the theory, and whether they are falsifiable. A hypothesis that no possible result could disconfirm is decoration.
Data and methods
What to get from it: what the evidence is, where it came from, how the key concepts were measured, and how the cases were chosen. What to watch: the gap between concept and measure — this is where most quantitative work is weakest and where the most interesting criticisms come from. Does "number of protest events" really capture political discontent? Also check the time period and the sample: what has been excluded, and would including it change the answer? You do not need to follow the estimation technique to ask these questions.
Results
What to get from it: what was found, and how strongly. Read the tables and figures directly rather than relying on the prose. What to watch: the difference between statistical significance and substantive size — an effect can be very precisely estimated and far too small to matter. Look at whether the key result survives the alternative specifications, and notice which specifications were not shown.
Discussion and conclusion
What to get from it: what the author thinks it means, the acknowledged limitations, and what comes next. What to watch: claims that have quietly expanded since the results section. Also read the limitations honestly — a paper that names a real limitation is usually more trustworthy than one that names none.
Appendices and supplementary material
What to get from it: robustness checks, full model results, question wording, coding rules. What to watch: this is increasingly where the interesting material is kept. If a result seems fragile, the appendix often tells you why.
Qualitative and theoretical articles do not follow this template — a comparative-historical piece or a theory paper may be organised by case, by period, or by concept. The same reading order still works: abstract, introduction, conclusion, then the whole thing.
Reading books efficiently
A scholarly book is not a long article and you are rarely expected to read one cover to cover for a course. Read the introduction, which contains the argument; the conclusion, which contains it again with the evidence attached; and then the one or two chapters that bear on your question. Chapter first paragraphs and last paragraphs will tell you which those are. If a book is central to your paper, go back and read more of it — but start this way.
What counts as a good source
There is a rough hierarchy of credibility. It is a starting point for judgment, not a substitute for it.
- Peer-reviewed journal articles. Scrutinised by specialists before publication. Not infallible, but the highest routine standard the discipline has.
- Scholarly books from university presses. Also peer reviewed, and better for sustained arguments, deep historical work, and theory.
- Edited volume chapters. Variable — review standards differ a great deal across volumes.
- Government, IGO, and national statistical agency data. Statistics Canada, Eurostat, the World Bank, the OECD, UN agencies. Usually reliable as data; read the methodology, and remember that what a government chooses to measure is itself a political decision.
- Working papers and preprints. Current, but not yet vetted. Usable with care, and say in your paper that they are unreviewed.
- Think tank and NGO reports. Quality ranges from excellent to advocacy in a suit. Ask who funds the organisation and whether it exists to find things out or to argue a position.
- Quality journalism. Excellent for events, chronology, and what people said. Not a source for theoretical claims.
- Everything else — blogs, opinion columns, encyclopaedias, social media. Fine as leads and starting points. Not evidence in a paper.
Two clarifications students ask about. Wikipedia is genuinely useful for orienting yourself in an unfamiliar area and for finding the real sources in the footnotes — use it that way, and cite what it cites, not the article. And a source's age is not a defect: a 1965 article that founded a debate may matter far more than something from last year. Ask what the source is doing in your paper, not just when it appeared.
Questions to ask of any source
- Who produced it, and what is their expertise? A physicist writing on epidemiology is not an expert on epidemiology.
- Who paid for it? Funding does not automatically discredit research, but undisclosed funding from an interested party is a serious warning sign.
- What is the evidence, and can you see it? Are data, code, and methods available? Transparency is the single best proxy for good faith.
- Do the conclusions exceed the evidence? An experiment on 200 undergraduates does not establish how electorates behave.
- Has it been engaged with? Check who cites it and what they say. A well-cited piece that everyone cites in order to disagree with it is telling you something.
- Does it acknowledge limitations? Honest research says what it cannot show. Research that admits nothing is either trivial or overselling.
Reading statistics and polls without being fooled
You will encounter numbers everywhere in political science, and most of the ways they mislead are simple enough to spot.
- Who was asked, and how were they chosen? A random probability sample and an online opt-in panel are not the same instrument. A poll of people who chose to answer tells you about people who choose to answer.
- Margin of error and sample size. A 3-point margin means a 2-point "lead" is not a lead. Subgroup results — "among young men in Quebec" — carry much larger margins than the headline number, and are routinely reported as though they do not.
- Question wording and order. Small changes move results substantially. Always find the actual questionnaire; reputable pollsters publish it.
- Percentages of what? A 100% increase from 2 cases to 4 cases is not news. Watch for percentage change reported without the base.
- Compared to what, and over what period? Start dates get chosen to produce the desired trend. Ask what the series looks like if you start five years earlier.
- Does the measure capture the concept? Does "number of protests" measure political discontent, or does it measure how well protests were recorded that year?
- Who is missing? Non-response is rarely random. The people hardest to reach are usually different in ways that matter.
Evaluating news and online sources
Two techniques do most of the work. Read laterally: instead of studying a suspicious site to judge it, open new tabs and find out what other sources say about that site and its claims. This is what professional fact-checkers do, and it is far faster and more reliable than staring at the page in front of you. And trace claims upstream: a story cites a report, which cites a study. Go to the study. Very often the study says something narrower than the report said it said, which is narrower again than the headline.
Warning signs worth taking seriously: no named author; no date; no links to primary sources; a striking statistic with no traceable origin; a headline that fits your priors a little too comfortably; an organisation whose funding is not disclosed anywhere on the site; and a journal you have never heard of that publishes across every field at once and charges to publish — which is what a predatory journal looks like.
Finding sources
- Start with your library, not with Google. The Concordia Library and the George W. Wallace Library at Champlain give you access to databases you would otherwise be paying for, and subject librarians will help you build a search strategy — an underused free service that will save you hours.
- Key databases for our field: JSTOR, Scopus, Web of Science, ProQuest Political Science, and the International Political Science Abstracts.
- Google Scholar is good for coverage and for its "cited by" feature. Connect it to your library in the settings so full-text links appear. Be aware that it indexes indiscriminately, so it will happily show you predatory journals alongside good ones.
- Snowball. Find one excellent recent article on your question and mine its reference list backwards; then use "cited by" to work forwards to newer work. Two or three iterations of this will map a literature faster than any keyword search.
- Read review articles first when entering a new area. The Annual Review of Political Science exists precisely to give you the shape of a debate in thirty pages.
Keeping track of what you read
Install Zotero now, not in your final year. It is free, it captures citations from your browser in one click, and it generates bibliographies in whatever style you need. Every hour you spend setting it up saves several later, and it eliminates an entire category of accidental plagiarism — the kind where you can no longer tell which notes were your paraphrase and which were the author's words.
Whatever system you use, follow one rule when taking notes: mark quotations as quotations, immediately, with the page number. Most plagiarism I encounter is not dishonesty. It is a student who copied a sentence into their notes in September and could not tell in November that it was not theirs.
If you cannot find sources on your question, the problem is often the question rather than the search. Come to office hours — usually twenty minutes of reframing solves it.