Observation, inference, hypothesis, and prediction
Separate what you noticed from what you think it means, then make a prediction that a test could challenge.
Field diagramA question becomes a test
Estimated reading
9 min
Estimated practice
15 min
Equipment
Paper or a text editor; no laboratory equipment required.
Progress stays in this page session. Allow lesson storage in Cookie settings to remember it; browser settings may also prevent saving.
Objective
By the end, you can:
Distinguish an observation from an inference.
Write a hypothesis as a provisional explanation.
Turn a hypothesis into a measurable, falsifiable prediction.
Start with the evidence
An observation is a record of something detected with a stated method: a color changed, a scale displayed 12.4 g, or a source reported a result. An inference is an interpretation built from observations: the color change may indicate a reaction, or the mass may have increased because water was absorbed. Keeping the two apart prevents a plausible story from becoming a claimed fact.
A hypothesis is a proposed explanation that could be wrong. It is more useful than a guess when it connects a cause to an outcome and names the conditions under which the connection should hold. A prediction is the specific observable result expected if the hypothesis is useful. A good prediction could fail, and its failure would teach you something about the explanation or the test.
Observation: after five minutes, the paper strip is darker than at the start.
Inference: moisture moved into the paper.
Hypothesis: higher humidity increases water uptake by this paper.
Prediction: strips exposed to higher measured humidity will show a larger mass increase over five minutes than strips exposed to lower humidity.
Boundaries matter
State what the observation does not establish. A darker strip does not by itself identify a chemical, prove causation, or generalize to every paper. A prediction should therefore name the material, exposure, outcome, time window, and comparison. If your measurement cannot distinguish a small change from ordinary variation, revise the question before collecting data.
Before building on a claim, make a quick source check: who made the claim, what was actually measured, whether the method fits your question, and what limitation or competing evidence is visible. A source can help you form a hypothesis without proving it. Record the source and its boundary alongside your observation so a later test does not turn an interesting report into an unsupported premise.
Use operational definitions: define “darker” with a color scale or instrument reading.
Name the comparison condition and keep the prediction directional only when evidence supports direction.
Treat an unexpected result as information; do not quietly rewrite the prediction afterward.
Use source quality as a reason to refine a question, not as permission to overstate a claim.
Worked example
From noticing to testing
A student notices that a paper towel feels heavier after being placed near a sink. The student avoids claiming that the sink caused the change and defines an exposure time and mass measurement.
01Observation: after 10 minutes near the sink, towel A reads 4.82 g; before exposure it read 4.20 g.
02Hypothesis: greater ambient moisture increases water uptake by the towel.
03Prediction: towels exposed in the more humid location will have a larger mass increase than towels in the drier comparison location.
04Limit: without humidity measurements, location remains a bundle of possible causes.
Materials
Set out what you need.
A supplied example dataset or a personal everyday observation
Notebook or worksheet
Safety & stop conditions
Pause if the work no longer fits the plan.
This paper exercise uses no hazardous materials. Do not turn it into a physical experiment involving people, animals, unknown chemicals, or environmental release without a separate risk review.
Method
Work through the steps.
01
Make two columns
Write only directly detectable statements in the first column and interpretations in the second. Label the method and time for each observation.
02
Draft the explanation
Write one sentence beginning “I propose that…” and include a cause, outcome, and context. Mark it as provisional.
03
Challenge it with a prediction
Specify what will be measured, compared, and recorded. Add a result that would count against the hypothesis.
Checkpoints
Observe, record, investigate.
Every observation names what was detected rather than why it happened.
The prediction contains a measurable outcome and comparison.
You can state one result that would weaken the hypothesis.
Common mistakes
Correct the process, preserve the record.
Writing “the sample reacted” as an observation.
Record the visible or instrument-detected change first; “reacted” is an inference unless reaction is established by a method.
Using a prediction that cannot fail, such as “something will happen.”
Specify a direction, threshold, category, or comparison that could be absent or reversed.
Treating one observation as a general rule.
State the population and conditions your proposed test actually addresses.
Practice
Rewrite a claim
Practice with this supplied scenario: two identical cups of water sit in different locations and one loses more mass overnight. The location, starting mass, final mass, temperature, and time are available; no cause is known.
Which statements are observations and which are inferences?
Write a hypothesis involving evaporation and a prediction with a comparison.
Name one alternative explanation that the first test cannot exclude.
01Is “the liquid became warmer because the reaction released heat” one observation?
Answer: No. A measured temperature change is an observation; “because the reaction released heat” is an inference that needs a defined comparison and evidence.
02Why must a prediction be capable of being wrong?
Answer: A risky prediction lets data discriminate between an explanation and alternatives. A statement that every possible result confirms the hypothesis cannot guide learning.
03Apply the distinction to a balance reading of 10.03 g.
Answer: The displayed value and the conditions of weighing are observations. Saying that the object contains exactly 10.03 g or gained mass because of a particular process requires assumptions about calibration, resolution, and alternatives.
