How Do You Write About Conflicting Evidence Without Simply Listing Both Sides?

What Should You Say When the Studies Do Not Agree?

Some studies report a benefit. Others report no clear difference. A few suggest harm. The easiest literature-review structure is to create two camps: studies supporting the effect and studies opposing it.

That approach acknowledges disagreement, but it rarely explains it. It may also manufacture a debate that the evidence does not actually contain. A statistically non-significant result, for example, is not automatically evidence for the opposite conclusion. Two studies may appear contradictory while examining different populations, outcomes, interventions, or time points.

The real synthesis question is not simply, “Which side has more studies?” It is, “What exactly conflicts, how comparable is the evidence, what might account for the difference, and what can we reasonably conclude despite the inconsistency?”

Analyze the Disagreement Instead of Creating Two Lists

Not Every Difference Between Studies Is a Genuine Contradiction

Define the proposition that supposedly conflicts

Before describing evidence as conflicting, write down the claim on which the studies supposedly disagree.

“The findings are mixed” is too vague. Do studies disagree about whether an intervention improves examination scores? About how large the effect is? About whether benefits persist? About which population benefits? About how participants experience the intervention?

Once the proposition is explicit, you can determine whether the studies actually provide evidence about the same thing.

Check whether the studies are sufficiently comparable

Two findings can point in different directions without contradicting each other if they concern different conditions.

An intervention may improve achievement among novice learners but not advanced learners. A six-week intervention may produce an immediate effect that disappears at one-year follow-up. One study may examine satisfaction while another measures performance.

These findings describe variation. Calling them contradictory may obscure the very condition that explains the difference.

Do not classify studies by statistical significance

One study reports a statistically significant benefit. Another estimates a similar benefit but has a wider confidence interval crossing the null. It would be incorrect to describe the first as “positive” and the second as “showing no effect” and then declare the studies contradictory.

Their effect estimates may actually be quite compatible. Differences in precision can produce different significance classifications even when estimated effects are similar.

Conflict should therefore be evaluated from the estimates, uncertainty, study design, and substantive findings rather than from whether individual p-values fall on opposite sides of a conventional threshold.

Compare direction, magnitude, and uncertainty

Quantitative findings can disagree in several ways. Effects may point in opposite directions. They may point in the same direction but differ substantially in magnitude. Confidence intervals may overlap extensively or barely at all. Some estimates may be too imprecise to distinguish meaningful benefit from harm.

Cochrane recommends assessing the presence and extent of between-study variation in meta-analysis and cautions against simple thresholds for heterogeneity statistics. Interpretation depends partly on the magnitude and direction of effects and the strength of evidence for heterogeneity.

Ask whether the disagreement follows a pattern

Once genuine variation is identified, compare the characteristics of studies producing different findings.

Do positive findings occur primarily in one population? Are negative findings concentrated among longer follow-up periods? Do studies with intensive interventions differ from those with minimal implementation? Does one measurement approach produce systematically different results?

If the differences correspond with meaningful study characteristics, the literature may support a conditional conclusion rather than a verdict that the evidence is simply “mixed.”

Distinguish observed patterns from explanations

Suppose studies with trained instructors report larger benefits than studies without formal instructor training. You can report that pattern if the evidence supports it.

Concluding that instructor training caused the difference is a stronger claim. Perhaps the trained-instructor studies also used different populations or more intensive interventions.

Cochrane cautions that subgroup analyses and meta-regressions have considerable pitfalls. Post-hoc investigations of heterogeneity can generate hypotheses, but explanations developed after inspecting study results require particularly cautious interpretation.

Evidence strength matters when studies disagree

Conflicting studies should not automatically receive equal interpretive weight. One group may contain studies at substantial risk of bias, indirect evidence, or very imprecise estimates. Another may provide more trustworthy and directly relevant evidence.

This does not mean inconvenient studies should disappear. It means the synthesis should explain why some findings warrant more confidence than others.

In formal certainty assessment, inconsistency is considered alongside other limitations such as risk of bias, indirectness, and imprecision. A conflict among credible and comparable studies may reduce confidence more substantially than disagreement driven mainly by studies with serious limitations.

Do not turn the literature into “for” and “against” camps too quickly

A binary structure can make research disagreement resemble a debate between positions rather than a set of estimates, observations, and interpretations produced under different conditions.

Instead of writing:

“Several studies support X. However, several studies do not support X.”

try to characterize the evidence more precisely:

“Positive effects are reported primarily in short-duration interventions with structured facilitation, whereas studies of less structured implementation generally report smaller or uncertain effects. Whether facilitation accounts for this difference remains unclear because intervention intensity and population also vary across the two groups.”

The second version preserves the disagreement while beginning to explain its structure.

Conflicting qualitative evidence also requires analysis

Conflict is not restricted to quantitative findings. Participants in some qualitative studies may experience an intervention as empowering while others describe it as burdensome or intrusive.

Those accounts may reflect different contexts, participant positions, implementation conditions, or genuine diversity of experience. A synthesis should resist replacing them with an artificial average sentiment.

Sometimes the appropriate conclusion is that the phenomenon is experienced differently by different groups and that the variation itself requires explanation.

Do not force consensus when the evidence remains inconsistent

After examining comparability, methods, populations, outcomes, bias, and plausible moderators, disagreement may remain.

That is a legitimate conclusion. Cochrane notes that considerable variation in results, particularly variation in the direction of effects, may make an average intervention effect misleading.

A literature review can therefore conclude that current evidence does not support a stable general effect, that effects may vary across settings for reasons not yet established, or that the evidence remains insufficient to distinguish competing interpretations.

Turning “Some Studies Say Yes, Others Say No” Into a Synthesis

The two-sides version

“Some studies show that laptop use improves learning, while other studies show no effect or negative effects. Therefore, the literature is mixed.”

This is accurate at a superficial level, but it stops exactly where synthesis should begin.

The synthesis-driven version

“Across the hypothetical studies, the relationship between laptop use and learning appears to vary with how the technology is incorporated into classroom activity. Performance benefits are reported when laptop use is structured around instructional tasks, whereas unrestricted use produces either uncertain outcomes or poorer performance when off-task behavior is common. Evidence that outcomes differ according to self-regulation further suggests that the consequences of laptop access may depend on both instructional design and student behavior. These studies therefore do not support a simple conclusion that laptops are inherently beneficial or harmful.”

Common Mistakes When Writing About Conflicting Evidence

Write the Conflict as a Problem to Analyze

When you encounter conflicting evidence, resist drafting separate paragraphs for supportive and unsupportive studies immediately. First build a comparison table or synthesis matrix showing the finding alongside characteristics that might matter: population, intervention, outcome, context, design, measurement, follow-up, precision, and risk of bias where relevant.

Then ask whether the disagreement remains after like is compared with like.

A strong paragraph therefore does more than balance two sides. It tells the reader what conflicts, how consequential the disagreement is, what might account for it, how confident those explanations are, and what conclusion survives after the conflict has been considered.

Before Describing Evidence as Conflicting, Check:

Frequently Asked Questions About Conflicting Evidence

Your Job Is to Characterize the Conflict, Not Make It Disappear

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.