How to Interpret the Results Section of a Research Paper
Many new readers find the results section intimidating because of the statistical notation, dense tables, and technical figures. The key insight is that you do not need to understand every number. Focus on the findings that relate to the main research question, and use the steps below to navigate the section systematically.
Step 1: Start with the Figures
Figures are the most accessible entry point to the results. Bar charts, scatter plots, line graphs, box plots, and heat maps all present data visually, making patterns, trends, and differences easier to spot than in a table of numbers. Before reading any text, look at each figure and ask yourself: What is this showing? What pattern do I see? Do the groups look different from each other?
Forming your own visual impression before reading the authors' description is important because it gives you an independent check on their interpretation. If a bar chart shows two groups with nearly identical heights but the text calls the difference "dramatic," that mismatch is a red flag worth investigating.
Pay attention to the y-axis scale. A common visual trick (sometimes intentional, sometimes not) is to zoom in on the y-axis so that a small difference looks large. A bar chart where the y-axis starts at 95 instead of 0 can make a 2% difference look enormous. Always check the axis labels and scale before drawing conclusions from the visual impression.
Step 2: Read the Figure Captions
Figure captions (also called legends) are dense with information. A good caption explains what the figure shows, what the axes represent, what each symbol or color means, the sample size, and any statistical annotations. Some captions include abbreviated results, such as "* p = 0.003" or "Error bars represent standard error of the mean."
Captions should make figures self-contained, meaning you can understand the figure without reading the surrounding text. If a caption is missing critical information, like what the error bars represent, the figure is harder to evaluate. Error bars might show standard deviation, standard error, or 95% confidence intervals, and each tells a different story about the data's variability and precision.
Step 3: Navigate the Tables
Tables present precise numerical data that figures cannot capture. Start by reading the table title, which tells you what data the table contains. Then look at the column headers and row labels to understand the table's structure. Columns typically represent different measurements, groups, or statistical tests, while rows represent individual variables, time points, or subgroups.
Find the numbers that correspond to the main research question. You do not need to understand every cell in a complex table on your first reading. If the paper investigated whether a new drug lowers blood pressure, find the blood pressure column and compare the values between treatment and control groups. The footnotes, often marked with symbols or letters, explain abbreviations and statistical conventions used in the table.
Table 1 in clinical papers is almost always a demographics table, describing the characteristics of the study participants. This table helps you assess whether the groups were comparable at the start of the study and whether the sample is representative of the population you care about.
Step 4: Understand the Statistical Reporting
Results sections report statistics using a standard notation that becomes familiar with practice. The most common elements are:
P-values tell you the probability that the observed result would occur by chance if there were no real effect. Values below 0.05 are conventionally considered statistically significant, but this threshold is arbitrary and does not indicate practical importance. A p-value of 0.049 and 0.051 are essentially identical in meaning, despite falling on different sides of the threshold.
Confidence intervals provide a range of plausible values for the true effect. A 95% CI of [2.1, 8.7] for a mean difference means the researchers are reasonably confident the true difference lies between 2.1 and 8.7. Narrow intervals indicate precise estimates, wide intervals indicate uncertainty. If a confidence interval for a difference crosses zero, the result is not statistically significant.
Effect sizes quantify how large the observed difference or relationship is. Cohen's d measures standardized mean differences (0.2 is small, 0.5 is medium, 0.8 is large). Correlation coefficients (r) range from -1 to 1. Odds ratios and relative risks compare the likelihood of outcomes between groups. Effect sizes tell you whether a finding is practically meaningful, not just statistically detectable.
Step 5: Compare Text to Data
After examining the figures, tables, and statistics, read the narrative text of the results section. The text should accurately describe what the data show. Check for consistency: does the text say "a significant increase" when the table shows p = 0.47? Does it describe a "large effect" when the effect size is tiny? Does it emphasize one finding while glossing over others that contradict the hypothesis?
Also note what the results section does not report. If the study measured five outcomes but only reports three, the other two may have shown unfavorable results. Selective reporting is a form of bias that inflates the apparent success of the study. Good results sections report all pre-specified outcomes, regardless of whether they reached significance.
Common Pitfalls When Reading Results
Confusing statistical significance with practical importance is the most common error. A study with 100,000 participants might find that a supplement raises test scores by 0.3 points on a 100-point scale, with p = 0.001. That result is highly statistically significant but practically meaningless.
Another pitfall is being impressed by many statistically significant results without considering the multiple comparisons problem. If a study tests 20 different comparisons, one of them is expected to be "significant" at p = 0.05 purely by chance. Look for whether the authors adjusted for multiple comparisons using methods like Bonferroni correction or false discovery rate control.
Finally, do not mistake absence of evidence for evidence of absence. A non-significant result does not mean there is no effect. It means the study did not detect one, possibly because the sample was too small or the measurement was too imprecise. The distinction between "no evidence of effect" and "evidence of no effect" is crucial.
When Results Seem Contradictory
You will sometimes encounter results sections where different analyses within the same paper appear to tell different stories. The primary outcome may be non-significant while secondary outcomes show strong effects, or the main analysis and sensitivity analysis may disagree. When this happens, prioritize the pre-specified primary outcome over secondary analyses, and trust the main analysis over sensitivity analyses unless there is a clear methodological reason to prefer the alternative approach. Authors sometimes emphasize the most favorable results in their discussion while downplaying less favorable ones, so reading the results section carefully prevents you from absorbing a biased interpretation.
Across different papers, contradictory results are common and expected. Differences in study populations, measurement methods, sample sizes, and analysis choices all contribute to variation in findings. Rather than asking which single study is correct, consider what the full body of evidence suggests when taken together. Systematic reviews and meta-analyses exist specifically to synthesize contradictory results into a coherent overall estimate.
Always look at the data yourself before reading the authors' interpretation. Start with figures, then tables, then statistics, and check whether the narrative accurately represents the evidence.