Interpretation, not a second results section
The most common structural failure in a discussion is repetition: the same findings, in the same order, with the numbers copied across. Nothing is added, and the reader who has just read the results notices immediately. The discussion has a different job, which is to say what the findings mean and why they came out that way.
A quick diagnostic is to count statistics in your discussion. If test statistics, confidence intervals, and p-values appear frequently, you are probably restating rather than interpreting. Referring back to a result in words — “the improvement in the trained group” — is usually enough, since the numbers are one section away.
Weak: “Accuracy in the trained group was significantly higher than in the control group.” Stronger: “Training improved accuracy, and the size of that gain is consistent with an attentional account rather than a knowledge account, since the effect appeared on speeded trials and not on untimed ones.” The second sentence could not have been written from the results table alone.
The opening paragraph does most of the work
Discussions that begin with a summary of the study — “this study investigated the relationship between…” — waste their most valuable position. The reader already knows what you investigated. What they want, and what editors look for when skimming, is the answer.
A workable opening paragraph does three things in four or five sentences: it answers the research question, states the single most important finding in words, and signals how confident you are. Only then does it preview the structure of the discussion, and often that preview is unnecessary. Draft the paragraph, delete its first sentence, and check whether it improved — it usually has.
This paragraph is also where a mismatch with your introduction becomes obvious. If the question you answer here is not quite the one you posed at the start, one of the two needs rewriting. That mismatch is one of the more frequent reasons a paper reads as unfocused without any single sentence being wrong.
Placing your findings among other people’s
Comparison is what turns a result into a contribution. For each key finding, say whether it replicates, extends, qualifies, or contradicts what is already published, and name the specific studies involved. Vague alignment — “this is consistent with previous research” — with three citations attached tells a reader nothing about which previous research or in what respect.
Contradictions deserve more space than agreements, because they are where the interesting explanation lives. When your finding conflicts with a published one, look for the concrete difference first: population, measure, timing, analysis, or context. A specific account of the divergence is a genuine contribution; “further work is needed to reconcile these findings” is an abdication.
Every source you engage with here should already exist in the introduction or the literature review. New citations occasionally appear in a discussion for a reason — a paper published while yours was under review, or an unexpected finding that sends you to a literature you had no reason to read before. Flag that reason rather than letting the reference arrive unexplained.
The limitations paragraph that helps you
Limitations are often written defensively, as a list of apologies that reviewers can quote back. Written well, they do the opposite: they show you understand what your design can and cannot establish, and they pre-empt the criticism by handling it first. The reviewer who finds a limitation you already named and bounded has much less to object to.
The useful pattern is limitation, direction, consequence. “The sample was drawn from one institution, so the norms of that setting may have amplified the effect; the finding should be treated as applying to comparable institutional contexts until it is tested elsewhere.” That sentence tells a reader what to do with your result rather than merely conceding a flaw.
Leave out limitations that apply to every study ever run. Self-report is imperfect, samples are never infinite, and no design controls everything. Include only the limits that a reader would otherwise use to reach a different conclusion — and never use the paragraph to dismiss a real problem with a hedge.
Implications without overreach
The gap between what a study shows and what its authors claim it shows is where credibility is usually lost. Correlational designs do not license causal verbs; a single trial does not license a practice recommendation; a laboratory effect does not license a policy proposal. Match your verbs and modals to your design and the claim will survive review.
Practical implications are strongest when they are specific about who would do what differently. “These findings have implications for teaching practice” is unfalsifiable. “If the effect depends on same-session feedback, then feedback timetabled for the following week is unlikely to produce it, which matters for courses that batch marking” names an audience and an action.
Theoretical implications work the same way. Say which account your finding supports, which it strains, and what the strained account would need to add to accommodate it. That is a contribution a later paper can build on, and it is considerably more interesting than a claim of general importance.
