How do I write a discussion section without repeating the results?

The repetition problem has a mechanical cause. You wrote the results, then sat down to write the discussion, and the only material in your head was the results. What is missing is the second input: the literature, and specifically the expectations the literature set up.

So write the discussion with your review chapter open. Every finding either matched what the field expected, exceeded it, fell short of it, or pointed the other way, and each of those four is a different paragraph with a different job. That framing gives you something to say that is not a restatement.

The order that works is the reverse of the results section. Start with the most important finding rather than the first one, since the discussion is organized by significance rather than by the sequence of analysis. Then take the surprises. Then the null results, which deserve interpretation rather than an apology.

One paragraph is worth planning explicitly: what your findings mean for the question in the introduction. Many discussions interpret each result locally and never return to the question that motivated the whole thesis, which leaves the examiner to do the joining.

Hold on to one division of labor while drafting. The results section carries the underlying story of what the data show, and the discussion explains what those results imply (Zhang, 2014, pp. 1-2). The specific warning there is against restating what a self-contained figure or table already displays. So do not write that Figure 2 shows group A scoring above group B. Write what that gap does to your hypothesis and to the problem you set out to solve. When space is short, take the one or two findings that answer your research questions most directly, give them real detail, and move the rest to supporting material (Zhang, 2014, pp. 1-2).

Interpretation also means finishing the inference for the reader. Zhang's phrasing is that you should not make readers do the arithmetic or work the conclusion out of the data you presented (Zhang, 2014, p. 2). If your analysis rests on a hypergeometric enrichment test, repeating the p-value is not interpretation. Define the term for the audience you expect, then state the conclusion that follows from it. The same source sets the ceiling on that move: do not exaggerate and do not belittle the significance of your results (Zhang, 2014, p. 2). Keep what the data establish separate from what they only suggest.

How do I discuss findings that contradict the literature?

Lead with the contradiction rather than burying it, then work through the candidate explanations: differences in sample, measure, setting, or analysis. Say which explanation you find most likely and why. A contradiction handled openly is a contribution, and one handled quietly looks like an error you missed.

Check the mundane explanations first and say that you did. A different result can come from a different measure, a different population, a different time period, or an analysis choice, and ruling those out is what makes the interesting explanation credible.

Do not overclaim in either direction. Your study does not overturn a literature on its own, and it also does not have to defer to it. The honest position is usually that the evidence now points in two directions and here is what would resolve it.

Reviewers arrive with a short list of standing objections, and you can draft against them. Zhang names three that recur: the hypothesis was not adequately studied, an important earlier result was left out, or the new result is inconsistent with previous work (Zhang, 2014, p. 2). Turn each one into a question your paragraph has to answer. Does this finding address the hypothesis you stated? Does it agree with the prior work you cited? If it does not, what explanation does your own evidence actually permit?

How honest should the limitations section be?

Honest about anything that affects how your findings should be read, and specific about consequences. "The sample was drawn from one region, so the estimate may not transfer" is useful. Listing every imperfection is not, and it dilutes the limitations that genuinely matter.

Examiners read this section closely, and they are testing judgment rather than looking for weaknesses to punish. A candidate who names the three limitations that matter, in the right order of seriousness, demonstrates exactly the awareness a doctorate is meant to certify.

Pair each limitation with what it does and does not undermine. That turns a list of apologies into an argument about the scope of your claims, which is what the section is for.

Repetition is usually a structural problem rather than a sentence problem, so fix it by reordering rather than by cutting repeated lines at the end. Zhang argues for arranging studies and findings in a clear logical order, because an earlier result can establish the premise for a later argument, and planning that order stops the same issue from being discussed in several places (Zhang, 2014, p. 1). Then leave the chapter alone for a day or two and reread it as an impartial referee, checking significance, logic, conclusions, organization, and presentation (Zhang, 2014, p. 2). Before you submit, present the work in a lab meeting, a departmental seminar, or a thesis committee discussion (Zhang, 2014, p. 2).