How Do You Separate Numerical Results From the Authors’ Narrative?

Are you reading the result or the authors’ description of it?

Quantitative papers rarely present numbers without commentary. A difference may be described as “substantial,” an association as “strong,” a pattern as “promising,” or a non-significant result as evidence that groups did not differ.

Those words are not the numerical result itself. They are a narrative layer placed around it.

Critical reading therefore requires a small but consequential separation: first establish what the numerical analysis actually produced, then examine whether the language surrounding that result is proportionate to the evidence.

Extract the numbers before accepting the narrative

How to read quantitative findings underneath the prose

Find the estimate the sentence is describing

When an author makes a numerical claim, locate the statistic underneath it. Depending on the study, that might be a mean difference, risk ratio, odds ratio, correlation, regression coefficient, standardized effect size, hazard ratio, prevalence estimate, proportion, or another quantity.

For comparative studies, also identify the values being compared. A statement that one group “performed better,” for example, becomes more informative when you know that the groups averaged 82.1 and 80.9 on a 100-point scale.

The purpose is not to strip the study of context. It is to establish a numerical anchor against which the prose can be evaluated.

Ask how large the result actually is

A result can be statistically detectable without being large. Conversely, an estimated effect can be potentially important while remaining too imprecise for a confident conclusion.

This is why quantitative reporting standards emphasize effect estimates rather than relying only on significance tests. CONSORT 2025, for example, recommends reporting the result in each trial group together with the estimated effect size and its precision, such as a 95% confidence interval. APA reporting guidance likewise emphasizes effect sizes and confidence intervals alongside statistical significance.

So when a paper says an intervention “improved performance,” ask: improved by how much?

Read the confidence interval with the point estimate

A point estimate gives you one estimate from the observed data. It does not, by itself, communicate how precisely that quantity has been estimated.

A confidence interval helps represent that uncertainty under the statistical model and assumptions used. CONSORT 2025 recommends confidence intervals around estimated treatment effects and notes their value even when conventional statistical significance is not reached.

Suppose an estimated difference is 4.0 points. A confidence interval from 3.2 to 4.8 conveys a different degree of precision from an interval running from -1.5 to 9.5. The same point estimate sits at the center, but the evidence surrounding it is very different.

Do not translate p-values into effect magnitude

A small p-value does not mean the effect is large, important, or practically consequential. Likewise, a p-value above a conventional threshold does not establish that the effect is exactly zero.

The p-value addresses a statistical question under specified assumptions. It does not replace the effect estimate, its uncertainty, or substantive judgment.

This becomes particularly important when authors treat statistical significance as practical importance. A result can cross a significance threshold while remaining trivial for the decision the research is supposed to inform.

Check denominators and absolute quantities

Relative measures can sound dramatic when presented without absolute quantities.

Imagine an outcome occurring in 2 of every 1,000 participants in one group and 1 of every 1,000 in another. Depending on the effect measure, the relative difference can appear large, while the absolute difference is one event per 1,000 participants.

Neither representation is automatically the “correct” one. They answer different questions. For binary outcomes, CONSORT 2025 recommends reporting both absolute and relative effect sizes where appropriate because each contributes useful information.

Whenever percentages are reported, identify the denominator. “50% more” tells you much less than “an increase from 2% to 3%.”

Distinguish descriptive statistics from inferential estimates

A difference between two sample means is not automatically the same quantity as an adjusted effect estimate from a regression model. Likewise, an observed percentage is not necessarily the same as a model-adjusted predicted probability.

When authors move between these quantities in prose, keep track of which number supports which statement.

Check whether the estimate is adjusted

Many papers report both unadjusted and adjusted results. An adjusted estimate is produced by a model that accounts for specified variables or features of the design.

Do not assume that “adjusted” automatically means “better” or “more correct.” The usefulness of an adjustment depends on why particular variables were included, how the model was specified, whether assumptions are reasonable, and the study design.

If the narrative emphasizes an adjusted estimate, identify what adjustment was made and whether that analysis was prespecified.

Separate primary results from favorable secondary findings

One of the most important narrative choices authors make is deciding which numbers receive attention.

A study may have a weak or inconclusive primary outcome while producing an apparently favorable secondary outcome or subgroup analysis. If the prose concentrates heavily on the favorable result, a casual reader may come away with a stronger impression than the primary analysis warrants.

Check which outcome was designated as primary, which analyses were prespecified, and whether the narrative has shifted attention toward a subgroup rather than the primary result.

Tables can reveal what the prose de-emphasizes

Numerical tables are especially useful because they often place several outcomes, groups, estimates, and intervals next to one another. CONSORT 2025 notes that trial results are often more clearly displayed in tables than in text.

Compare the prose with the table rather than reading either in isolation. Is the largest amount of narrative attention given to the primary outcome? Are null or unfavorable findings reported with comparable clarity? Are confidence intervals omitted from the prose even though they appear in the table?

This comparison helps you see not only what the study reported but also what the authors chose to foreground.

Adjectives are claims too

Words such as “strong,” “substantial,” “minimal,” “dramatic,” “meaningful,” and “negligible” contain judgments about magnitude or importance. They may be entirely reasonable, but they should not pass unnoticed simply because they are adjectives rather than statistical statements.

Try temporarily deleting the adjective.

“A substantial improvement of 2.1 points” becomes “an improvement of 2.1 points.” You can then ask whether 2.1 points is substantial on that measure, according to what criterion, and for which practical purpose.

Numerical results still require context

Numbers do not interpret themselves. An effect size depends on the measurement scale, research design, population, comparison, analytical assumptions, and substantive context.

Separating the number from the narrative is therefore an analytical step, not the final judgment. Once you know what the study actually found before interpretation, you can decide whether the authors’ description fairly represents it.

Compare the statistical result with the sentence built around it

The neutral reconstruction does not prove that the authors’ adjective is wrong. It reveals what must be justified. Perhaps a two-point improvement has important consequences in this setting. If so, the paper should provide a defensible basis for that interpretation.

Common mistakes when reading numerical findings

Build a numerical summary before a verbal one

When a quantitative paper matters to your research, create a short numerical reconstruction of its central result before adopting the authors’ wording.

You can then move from the numerical result to the broader question of what the data show versus what the authors conclude.

Check the numbers behind the narrative

Questions about reading numerical research results

Let the numbers constrain the story

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.