Hypothesis Quality Checklist (Save This!)
A good hypothesis is like a good map: it tells you where to look, what to measure, and what you expect to find. Use this checklist anytime you want to turn a curiosity (“I wonder if…”) into something you can actually test.
1) Falsifiability & Testability: “Could this be proven wrong?”
A hypothesis should be risky in a good way—meaning there’s a possible result that would clearly contradict it.
✅ Look for
- A real-world test you could run or observe.
- A clear outcome that would make you say: “Welp, guess my idea was wrong.”
🚩 Warning signs
- “It’s always true” claims.
- Claims that dodge evidence (e.g., “It works, but only when you can’t measure it”).
Example
- Strong: “If I water basil daily, it will grow faster than basil watered every 3 days.”
- Weak: “Basil likes attention.” (What counts as attention? How would it fail?)
2) Operational Definitions: “What exactly do my words mean?”
Big ideas need measurable meanings. An operational definition explains how you’ll measure a concept in the real world.
✅ Look for
- Vague terms replaced with specific measures.
- A measurement method you could repeat.
Helpful swaps
- “Healthier” → resting heart rate, blood pressure, VO₂ max estimate, doctor-approved metric
- “More focused” → minutes on-task, number of interruptions, quiz score
- “Better sleep” → total sleep time, wake-ups, sleep latency (time to fall asleep)
Example
- “More energetic” (vague) → “Rate energy from 1–10 at 3 pm daily” (operational)
3) Prediction Quality: Measurable and Directional
A hypothesis often comes with predictions: specific outcomes you expect.
✅ A high-quality prediction is…
Measurable
It includes something you can count, time, rate, score, or categorize.
Directional
It says which way the change should go.
- Increase/decrease
- Higher/lower
- Faster/slower
Upgrading predictions
- Not great: “Music practice improves skill.”
- Better: “After 2 weeks of daily 20-minute practice, I will play the piece with fewer mistakes than before.”
Tip: If your prediction could be “better OR worse OR the same,” it’s not directional yet.
4) Correlation vs. Causation: Spot the Sneaky Traps
Sometimes two things move together—but that doesn’t prove one caused the other.
✅ Correlation (moves together)
- “When I drink more coffee, I feel more alert.” (They’re linked in your data.)
✅ Causation (one makes the other happen)
To claim causation, you need extra care: control, random assignment, ruling out other explanations.
🚩 Warning signs you’re sliding from correlation to causation
- Third-variable problem: Something else causes both.
- Example: “People who exercise more sleep better.” (Maybe less stress causes both.)
- Reverse causality: The arrow might point the other way.
- Example: “More sleep causes more exercise.” Or: “More exercise causes more sleep.”
- Selection effects: The groups were different from the start.
- Example: “This supplement works—my friend tried it!” (Your friend may also have changed diet, routine, etc.)
Checklist question: If I changed only X, would Y still change? How do I know?
5) Avoid These Pitfalls (They Wreck Good Hypotheses)
Pitfall A: Vague terms
- “Better,” “more,” “effective,” “healthy,” “improves”
Fix: Define the measurement.
- “Better” → “fewer errors,” “faster time,” “higher score,” “lower resting heart rate”
Pitfall B: Moving goalposts
Changing what “counts” after you see results.
- “I said it would help… just not in that way.”
Fix: Write success criteria before testing.
- “Success = at least 10% improvement in ___ compared to ___.”
Pitfall C: Post-hoc stories (aka “I’ll explain it later”)
Making up a neat explanation after the fact can be tempting.
Fix: Separate:
- What happened (data) from
- Why it might have happened (possible explanations)
It’s totally okay to brainstorm explanations—just don’t pretend you predicted them.
Quick “Hypothesis Quality” Self-Score (0–2 each)
Use this as a tiny rubric when you draft:
- Falsifiable/testable: 0 = no, 1 = sort of, 2 = clearly testable
- Operational definitions: 0 = vague, 1 = partial, 2 = fully defined measures
- Predictions measurable + directional: 0 = missing, 1 = one of the two, 2 = both
- Correlation/causation awareness: 0 = claims causation carelessly, 1 = mentions limits, 2 = designs/phrases carefully
- Pitfalls avoided (vague terms, moving goalposts, post-hoc stories): 0 = yes, 1 = somewhat, 2 = clean and pre-committed
Exit Ticket (Quick Draft + Self-Score)
Pick an interest area (coffee, fitness, plants, music practice, or space).
- Draft one hypothesis.
- Write two predictions (measurable + directional).
- Self-score using the checklist above (0–2 each), and revise your draft if anything feels fuzzy.
Takeaway
A strong hypothesis isn’t fancy—it’s clear. If you can define your terms, make directional predictions, and explain how your idea could be wrong, you’re doing real science thinking (even if you’re testing coffee and sleep).
