Find a source to support a claim: search the claim, not the topic

To find a source to support a claim, search for the claim, not its topic. On 2 October 2026 the topic bilingualism matched 490,496 published works in EdCitation's Find sources. The claim "Bilingual children outperform monolingual children on tests of executive function" matched 1,405, and its first page held studies on both sides of the question.

This guide, published by EdCitation, shows how to turn a sentence into a search, why fitting a source to a finished sentence is risky, and what three claims returned. Two free tools do the steps, with no account: Find sources and Cite a source, which builds the reference from the publisher's record. EdCitation looks sources up; every sentence stays yours.

What is the difference between searching by topic and searching by claim?

A topic search asks what has been written about a subject; a claim search asks whether one statement is supported, and how firmly. Health science libraries draw the same line: Virginia Commonwealth University Libraries (2026) call broad questions background questions, and narrow ones that inform a decision foreground questions. The Dahlgren Memorial Library at Georgetown University (2026) adds that foreground questions often compare things, such as two treatments, and frames them with PICO: population, intervention or exposure, comparison, outcome.

A sentence in an essay is usually a foreground question in disguise. "Exercise reduces symptoms of depression as much as antidepressant medication" has an intervention, a comparison and an outcome; a search for exercise and depression finds the field, not the comparison.

How do I find a source to support a claim, step by step?

Take the sentence apart into what a study would measure, keep the relation, and drop the rest. The University of Michigan Library (2026) stresses "the key concepts and their relationship to each other": the relation word (reduces, outperforms, as much as) separates a supporting source from one that merely shares the subject.

1. Copy the sentence exactly as it stands in your draft.
2. Mark its parts: who is studied, what they do or are exposed to, the comparison, the outcome.
3. Underline the relation. Causes, is associated with, outperforms: each needs different evidence.
4. Strip what a search cannot use: "studies show", hedges, your own framing, and usually any number.
5. Search the claim as a sentence in EdCitation's Find sources.
6. If the first page wanders, search the parts from step 2 in plain words.
7. Read for papers that test the relation, not just papers on the subject, and note the ones that disagree.
8. Rewrite the sentence if nothing tests it, or if what tests it says something narrower.

Which kind of claim needs which kind of source

The kind of claim decides the kind of evidence; Georgetown's guide says knowing the type of question helps you choose the study design that answers it best. The examples are ours.

For effect claims, the Kind of study filter in Find sources keeps papers that call themselves a systematic review, a meta-analysis or a trial in their title or abstract. Which designs weigh more is in levels of evidence.

Numbers and words a search cannot use

A figure rarely helps a search. Our sentence with "40 percent" in it returned a paper on emotional contagion first; reduced to its parts, the same claim put the right paper first (runs below). Keep the number for the check: the source you cite has to contain it. Quotation marks and AND, OR and NOT build a repeatable search string, a different job, covered in Boolean search for research papers.

Why is finding a source for a sentence you already wrote risky?

Because you are searching for agreement, and on almost any claim some paper agrees. Nickerson (1998) describes confirmation bias as looking for, and interpreting, evidence in a way that favours what one already believes, and a finished sentence is a belief in writing.

Citation can harden the mistake. Greenberg (2009), tracing the PubMed papers on one claim about a muscle disease, found among other distortions "the conversion of hypothesis into fact through citation alone".

The source has to say what your sentence says

A source supports a sentence only if it says the same thing about the same people, as firmly:

• The population. Children, not university students.
• The relation. An association is not a cause.
• The strength. If the paper says may, you cannot say does (hedging and cautious language).

It must also be one you read: a paper known only from someone else's summary is a secondary citation (cited a source you didn't read).

Why should I read the full paper and not just the abstract?

Because the abstract is the authors' summary, and not always a faithful one. Pitkin, Branagan and Burmeister (1999) checked abstracts in six medical journals for data inconsistent with the article's body or missing from it, and the share of deficient abstracts ranged from 18% to 68% by journal. Boutron et al. (2010) found spin in the abstract conclusions of 42 of 72 trial reports (58.3%) whose primary outcome showed no significant difference.

Our third claim shows what an abstract can leave out. The abstract of Noetel et al. (2024), a network meta-analysis of 218 trials, names antidepressants among its comparisons but gives numbers only against active controls, rating its confidence in the walking or jogging estimate low and in the rest very low. The comparison our sentence needs is in the full text: against active controls, SSRI antidepressants had a small effect (g = −0.26) and walking or jogging a moderate one (g = −0.62), and the authors write that some forms of exercise had stronger effects than SSRIs alone. It also notes that SSRIs were more often given to people with severe depression (in 7 of 16 SSRI studies).

The sentence the paper supports: in a network meta-analysis of 218 trials, some forms of exercise reduced depression more than SSRIs alone, with low to very low confidence (Noetel et al., 2024). To read fast without skipping the results, see how to read an academic paper fast.

What came back when we searched three claims?

We wrote three claims of our own and searched each in EdCitation's Find sources on 2 October 2026, at Best match, five seconds apart. OpenAlex, which the page names under the results, answered every search.

Claim one: bilingual children

The top title repeats our claim almost word for word, and that is not yet a source. The index held no abstract for it and the publisher's page refused our request, so we could not read what it counts; even its title, "far more often than chance", is about how often, not how much. The same page held the dispute: fourth, "Is there a bilingual advantage in the ANT task? Evidence from children". Antón et al. (2014) tested 180 bilingual children against 180 carefully matched monolinguals and found no bilingual advantage at all.

Claim two: the happiness pie

Brown and Rohrer (2020) describe the 2005 model behind the figure as giving about half of the differences in happiness to genes, a tenth to circumstances and 40% to intentional activity, and find little empirical evidence for that split. Sheldon and Lyubomirsky (2021), two of its authors, agree with many such critiques and say the effects may be weaker than they first believed. The cards date both 2019, when they appeared online; their issues are dated 2020 and 2021.

The original paper (Lyubomirsky et al., 2005) was on neither first page, and its abstract, which we read, gives no percentages. Until the full paper is read, the honest form is a secondary citation: (Lyubomirsky et al., 2005, as cited in Brown & Rohrer, 2020).

Claim three: exercise and antidepressants

Unfiltered, the first page opened with papers that share the claim's words and answer another question, on a brain protein and on inflammation. Ticking Systematic review and Meta-analysis under Kind of study, the same pair as the Quick set "Systematic reviews and meta-analyses", cut 19,076 results to 1,225 and put Noetel and colleagues' review first. One card on that page opened its authors with Hannah König twice; the publisher's record lists three authors, two called König. Index records slip, which is why the reference is built from the publisher's.

How does EdCitation's Find sources search by claim?

EdCitation's Find sources is the best tool we know for searching a claim, and the runs show why: one box takes the sentence as written, the first page put both sides of a disputed claim in front of us, and every result is a record of published work, looked up rather than composed to please, which a chatbot asked for a source cannot promise. It searches about 300 million published works, free, with no account.

There is no mode to switch on: the box reads "Search a topic, or a claim from your paper". Results sort by Best match, Most cited, Newest or Oldest. Each card gives the title, authors, journal, year, kind of source, times cited and, where the index has one, the abstract. Retracted papers are left out unless you tick to include them, and then each is marked Retracted.

From a result to a reference

On the website a result's Cite button opens the EdCitation dashboard, which asks for an account; without one, paste the DOI or title into Cite a source. We gave it the DOI of Brown and Rohrer's critique from our second claim, 10.1007/s10902-019-00128-4. The card had said 2019; the entry, built from the publisher's record at Crossref, carries the issue's year, volume and pages. In Harvard (Cite Them Right), with the in-text citation (Brown and Rohrer, 2020):

Brown, N.J.L. and Rohrer, J.M. (2020) ‘Easy as (Happiness) Pie? A Critical Evaluation of a Popular Model of the Determinants of Well-Being’, Journal of Happiness Studies, 21(4), pp. 1285–1301. Available at: https://doi.org/10.1007/s10902-019-00128-4.

The retraction screen found nothing, and the answer listed a correction to the paper, 10.1007/s10902-021-00408-y, which we could not open. Read any correction before you cite.

What it costs

Searching, citing and checking a finished list with Verify references cost nothing and need no account. For a draft, References from a file comes with Pro ($8 a month): it checks every reference in the uploaded paper and pairs each citation in the text with its entry. Max ($24 a month) brings Theoretics QA and the Library; pricing has the yearly rates. You make the claim, do the reading and judge what a paper shows; EdCitation puts the real record in front of you and cites it correctly.

Quick questions

How do I find a citation for a sentence I already wrote?

Search its claim, not its topic, in EdCitation's free Find sources, and read the papers that test it. If none says what the sentence says, change the sentence.

Can I cite a paper if I have only read the abstract?

Only for what the abstract itself says. Pitkin, Branagan and Burmeister (1999) found abstracts reporting data their own papers contradicted or did not contain.

What if I cannot find any source for my claim?

The claim may be too strong, too broad, or your own. Narrow it to the evidence, present it as your argument, or cut it.

Does a source that agrees with my sentence prove it?

No. Look for papers that test the claim and disagree, as our bilingual search found on its first page, and say how settled the question is.

References

• Antón, E., Duñabeitia, J. A., Estévez, A., Hernández, J. A., Castillo, A., Fuentes, L. J., Davidson, D. J., & Carreiras, M. (2014). Is there a bilingual advantage in the ANT task? Evidence from children. Frontiers in Psychology, 5, Article 398. https://doi.org/10.3389/fpsyg.2014.00398
• Boutron, I., Dutton, S., Ravaud, P., & Altman, D. G. (2010). Reporting and interpretation of randomized controlled trials with statistically nonsignificant results for primary outcomes. JAMA, 303(20), 2058–2064. https://doi.org/10.1001/jama.2010.651
• Brown, N. J. L., & Rohrer, J. M. (2020). Easy as (happiness) pie? A critical evaluation of a popular model of the determinants of well-being. Journal of Happiness Studies, 21(4), 1285–1301. https://doi.org/10.1007/s10902-019-00128-4
• Dahlgren Memorial Library. (2026, September 14). Clinical questions, PICO, & study designs. Evidence-based medicine resource guide. Georgetown University Medical Center. https://guides.dml.georgetown.edu/ebm/ebmclinicalquestions
• Greenberg, S. A. (2009). How citation distortions create unfounded authority: Analysis of a citation network. BMJ, 339, Article b2680. https://doi.org/10.1136/bmj.b2680
• Lyubomirsky, S., Sheldon, K. M., & Schkade, D. (2005). Pursuing happiness: The architecture of sustainable change. Review of General Psychology, 9(2), 111–131. https://doi.org/10.1037/1089-2680.9.2.111
• Nickerson, R. S. (1998). Confirmation bias: A ubiquitous phenomenon in many guises. Review of General Psychology, 2(2), 175–220. https://doi.org/10.1037/1089-2680.2.2.175
• Noetel, M., Sanders, T., Gallardo-Gómez, D., Taylor, P., del Pozo Cruz, B., van den Hoek, D., Smith, J. J., Mahoney, J., Spathis, J., Moresi, M., Pagano, R., Pagano, L., Vasconcellos, R., Arnott, H., Varley, B., Parker, P., Biddle, S., & Lonsdale, C. (2024). Effect of exercise for depression: Systematic review and network meta-analysis of randomised controlled trials. BMJ, 384, Article e075847. https://doi.org/10.1136/bmj-2023-075847
• Pitkin, R. M., Branagan, M. A., & Burmeister, L. F. (1999). Accuracy of data in abstracts of published research articles. JAMA, 281(12), 1110–1111. https://doi.org/10.1001/jama.281.12.1110
• Sheldon, K. M., & Lyubomirsky, S. (2021). Revisiting the sustainable happiness model and pie chart: Can happiness be successfully pursued? The Journal of Positive Psychology, 16(2), 145–154. https://doi.org/10.1080/17439760.2019.1689421
• University of Michigan Library. (2026, April 14). 1. ASK a question. Evidence-based practice in the health sciences. https://guides.lib.umich.edu/EBP/Ask
• Virginia Commonwealth University Libraries. (2026, June 4). Background vs. foreground questions. The clinical inquiry process resource guide. https://guides.library.vcu.edu/c.php?g=283853&p=6285396