How to Handle Contradictory Findings in a Literature Review

Move one: verify the contradiction is real

A large share of apparent contradictions dissolve under inspection. The studies measured different outcomes (self-reported mood versus diagnosed episodes), different populations (college students versus adolescents), different exposure windows (current use versus lifetime use), or different points on the same dose curve. Before refereeing anything, put the two studies side by side on one line each: population, outcome measure, exposure definition, period, design. If any row differs materially, the contradiction may be an artifact, and the honest sentence is that the findings are not directly comparable — which is itself a finding the review can use.

The verification pass also protects you from the subtler version: studies that agree on the direction but differ on magnitude. A 12% effect and a 3% effect are not a contradiction; they are a range, and the review's job is to explain the spread. Reserve the contradiction machinery — the full four-move treatment — for findings that genuinely point opposite ways on the same measurable question. Graders can tell the difference, and a review that treats every variance as a contradiction reads as one that cannot read effect sizes.

The numbers at a glance

The four moves for handling contradictory findings — the range, and where each row actually shows up in coursework. Skim it first; the sections below explain why the range exists and which row applies to you.

Move two: find the variable that explains it

Real contradictions usually have a hidden moderator — the variable whose levels separate the studies. The standard suspects, in the order to check them: population (age, baseline health, setting), period (before and after a policy or technology change), measure (self-report versus objective marker, screener versus clinical interview), method (cross-sectional versus longitudinal, sample size, follow-up length), and dose or exposure definition. The explanation sentence is the review's most valuable product: the two findings reconcile once delivery format is separated — in-person cohorts show the effect, remote cohorts do not.

Finding the moderator takes table work, not inspiration. Add the candidate variables as columns to the verification sheet from move one and look for the column where the studies differ while the others match. When two variables differ at once — population and period — the honest sentence names both and flags the confound, and the review's resolution shifts to the evidence needed to separate them. A contradiction you cannot explain is still reportable; the resolution then becomes what the field needs to test the moderator, which is a legitimate and respectable gap statement.

Move three: present both findings at full strength

The presentation rule is symmetry: each side of the contradiction gets its finding stated as a result, its design named, its sample sized, and its limitations acknowledged in the same breath. Three trials with 2,000 participants found the effect; the single study that did not ran 90 participants self-selected from one clinic. The reader can weigh that — but only if you show the weights. The failure mode is asymmetry dressed as balance: the favored finding gets its numbers, the disfavored one gets a dismissal. Graders mark the pattern instantly, and it reads as advocacy rather than synthesis.

Symmetry extends to language. Both findings are reported, not admitted; the disfavored study is well-designed but limited to, not flawed. The review's authority comes from the reader's trust that the refereeing is fair, and fairness is visible in the evenness of the prose. A practical check: read your two presentation sentences aloud without the study names — could a stranger tell which side you will resolve toward? If yes, the sentences are loaded, and the fix is to move the evidence-quality judgments into explicit statements where they can be examined rather than smuggled into adjectives.

Move four: make the referee call explicit

A review must land somewhere, and the resolution sentence is where. The pattern: weigh, call, warrant. Given the design asymmetry, the current evidence favors the effect at in-person delivery, though the remote result keeps the boundary unsettled. Three parts — the basis (design asymmetry), the call (evidence favors), and the honest boundary (unsettled). What the resolution never does is erase the loser: the call coexists with the contradiction, and the boundary sentence tells the reader exactly how much the call is worth. Reviews that end in both sides have merit are not resolving; they are declining the job.

The resolution sentence is also where the review connects to your own paper. If your thesis builds on the favored side, the resolution is its foundation; if your thesis is precisely the unsettled boundary — the moderator nobody has tested — the contradiction is your gap statement's evidence. Either way the call is load-bearing, which is why it earns its own sentence rather than hiding in a paragraph's final clause. Write it, read it to a classmate, and ask what the paper is claiming; a correct answer means the referee call did its work.

The sentence patterns referees actually use

Six patterns carry nearly all referee work. Verification: the two findings are not directly comparable — the first measured diagnosed episodes, the second self-reports. Reconciliation: the discrepancy reconciles once X is separated. Weighing: the larger trial carries more weight here because its blinding survived the follow-up. Boundary: the effect holds at in-person delivery; remote settings remain untested. Confession of confound: population and period differ together, so neither explains the split alone. Agenda: testing the moderator directly is the field's clearest next step, and the present study takes it up.

Notice what the patterns share: each one states a relationship between studies, which is the sentence type the mechanical test from the literature-review page counts. A review's synthesis score is largely its referee-sentence density — one per contradiction at minimum, two or three where the split is the paper's business. Students underuse these patterns mostly because drafts get written source-by-source, and no single source hands you a relationship sentence. Write them in a separate editing pass, with the two studies' one-liners side by side on screen; the sentences come easily when the materials are arranged.

A five-minute audit of your own source list

Before submitting, run the one-sentence test on every reference: can you state, without opening the PDF, what this source argues and where it supports your paper? Sources that fail the test are collected, not used — cut them or promote them by actually integrating their evidence. Count what remains; that number, not the folder in your reference manager, is what satisfies the requirement.

Two composition checks finish the audit. First, no more than a third of your list should be non-scholarly — news pages, organizational posts, encyclopedias. Second, your three or four load-bearing sources should appear in the text two or more times each; if every source appears exactly once, your paper is a relay of summaries, and the fastest fix is deepening the analysis around your best two sources. The citation generator below formats whatever survives the audit in APA, MLA and Chicago.