How to Draw Scientific Conclusions

Step 1: Revisit Your Hypothesis

Start by restating your original hypothesis clearly. Then lay out your key results alongside it. Does the data match what your hypothesis predicted? If your hypothesis predicted that increasing fertilizer concentration would increase plant growth, and your data shows a clear positive correlation between fertilizer amount and plant height with statistical significance, your hypothesis is supported. If plant height showed no relationship to fertilizer amount, your hypothesis is not supported.

Be precise in your language. Scientists rarely say a hypothesis is "proven." Instead, the data either supports or does not support the hypothesis. This distinction matters because future evidence might change the picture. A single experiment, no matter how well designed, can only provide evidence for or against a hypothesis. Proof, in the mathematical sense, is not something experimental science provides.

If your results partially support your hypothesis, say so. Perhaps the fertilizer increased growth at low concentrations but inhibited it at high concentrations. This is a more nuanced result than a simple yes or no, and it tells you something interesting about the relationship you are studying. Partial support often leads to refined hypotheses that are more accurate than the original.

Step 2: Evaluate the Strength of Your Evidence

Not all data is equally convincing. Evaluate the quality of your evidence by considering several factors. Was your sample size large enough to detect the effect you were looking for? Small samples produce unreliable results because a few unusual measurements can dominate the statistics. The larger your sample, the more confidence you can have in your conclusions.

Consider the magnitude of the effect you observed. A treatment that increases plant growth by 50% is more convincing than one that increases it by 2%, even if both are statistically significant. Large effects are harder to explain away as artifacts of measurement error or uncontrolled variables. Effect size matters as much as statistical significance when evaluating evidence.

Examine the consistency of your results. If every replicate in the treatment group showed greater growth than every replicate in the control group, that is stronger evidence than if the groups overlapped substantially with only a small difference in means. Look at the spread of your data, not just the averages. Consistent results across multiple trials, multiple conditions, or multiple measures of the same outcome provide stronger evidence than a single comparison.

Consider whether your methods were rigorous enough to produce reliable data. Were your instruments calibrated? Were your measurements taken consistently? Were your control groups properly maintained? Methodological weaknesses reduce the strength of any conclusion, no matter what the numbers show.

Step 3: Consider Alternative Explanations

This is perhaps the most important and most difficult step. Before concluding that your independent variable caused the observed changes, you must seriously consider whether something else could explain the results. Were there confounding variables that were not adequately controlled? Could the results be due to a placebo effect, observer bias, or regression to the mean?

Think about each potential alternative explanation and evaluate it against your experimental design. If you used random assignment, proper controls, and blinding, many alternative explanations can be ruled out. If your design lacked these features, you must acknowledge that your conclusion is less certain.

Consider whether the relationship you observed could be coincidental. Correlation does not imply causation. Two variables might move together because they are both affected by a third variable you did not measure. For example, ice cream sales and drowning rates both increase in summer, but ice cream does not cause drowning. Both are driven by warmer weather. Always ask: "Is there a plausible mechanism connecting my independent and dependent variables, or could this be a spurious correlation?"

Step 4: State Your Conclusion Clearly

Write your conclusion as a clear, specific statement that accurately reflects what your evidence supports. Avoid both overclaiming and underclaiming. An overclaimed conclusion goes beyond what the data shows. An underclaimed conclusion fails to report a genuine finding. Aim for a statement that a skeptical scientist would accept as a fair representation of your results.

Include appropriate qualifications. If your study was conducted only on a specific population, age group, or species, note that. If your results were statistically significant but the effect size was small, say so. If there were limitations in your methodology that reduce your confidence, acknowledge them. These qualifications are not weaknesses; they are evidence of scientific integrity.

A good conclusion typically follows this structure: a statement of what the data showed, a comparison to the original hypothesis, an acknowledgment of the strength and limitations of the evidence, and a connection to broader scientific understanding. For example: "The data showed that plants receiving 10g/L fertilizer grew an average of 4.2 cm taller than unfertilized controls (p = 0.003, d = 0.85), supporting the hypothesis that this fertilizer concentration promotes growth. However, this study tested only one plant species under greenhouse conditions, and results may differ for other species or outdoor environments."

Step 5: Identify Next Steps

Good science always points toward more science. What questions did your experiment raise? What would you do differently if you repeated the study? What additional experiments would strengthen your conclusion or explore its boundaries?

If your hypothesis was supported, the next step might be testing it under different conditions, with different populations, or at a larger scale. Replication by independent researchers is essential for establishing scientific confidence. If your hypothesis was not supported, the next step might be revising the hypothesis and designing a new experiment to test the revised version.

Unexpected findings deserve particular attention. If you noticed something in your data that you did not predict, that observation could be the seed of an important discovery. Document these unexpected findings clearly and suggest specific experiments that could investigate them. Some of the most significant advances in science have come from researchers who paid attention to results they did not expect and followed up on them systematically.