Phase 1: Prepare
Step 1 Warm-up recall
From M1.L2: how do you turn WHO-5 ratings into a 0 to 100 score? Write the rule in one line.
Private: nothing you type here is sent or stored.
Step 2 Learning goal
By the end of this lesson you can:
- define reliability and validity, and give one way to check each (Foundation);
- name four threats to self-reported well-being and one safeguard for each (Core);
- judge whether a stated claim is justified by the measure behind it (Core).
Phase 2: Teach
Step 3 A familiar example
A bathroom scale reads 2 kg too heavy every morning. It is perfectly consistent, and it is wrong. A second scale gives readings that jump around but average the true weight. The first is reliable and not valid. The second is closer to valid and not reliable. Scores on a well-being scale can fail in either way.
Step 4 Clear explanation
Reliability is consistency. Internal consistency asks whether the items of a scale agree with one another, often summarised by Cronbach’s alpha. Test-retest reliability asks whether the same person gets a similar score twice, a short time apart. A value of 0.70 or above is often treated as acceptable for alpha. That is a convention, not a law.
Validity is whether the scale measures what it claims to. Convergent validity: scores relate to measures of related ideas, and to how other people who know the person rate them. Discriminant validity: they do not simply repeat unrelated ideas. Reviews of life-satisfaction measures generally find moderate, not perfect, agreement with other people’s ratings.
Reliability is necessary but not sufficient for validity. A scale can be consistent and still miss the idea.
Four threats to self-report:
| Threat | Example | Safeguard |
|---|---|---|
| Mood and context at the time of answering | Surveying staff on the last day before a holiday | Vary timing, record context, repeat measurement |
| Social desirability | Answering the way one thinks looks good | Anonymous responses, neutral wording |
| Different use of the scale | One person’s 7 is another’s 5 | Compare changes within a person, use several items |
| Question wording and order | An earlier question about money changes the next answer | Fixed order, tested wording, pilot the survey |
Step 5 Worked example (4 marks)
A company surveys all staff on the last Friday before the year-end break, using one unnamed question, “Are you happy?”, and reports “90% of staff are happy”. Name four problems and one fix for each.
- Timing: the mood of the day inflates answers. Fix: ask at several ordinary times. (1 mark)
- One unvalidated item: nothing shows it is reliable or valid. Fix: use a published scale such as the SWLS. (1 mark)
- Unclear account: “happy” could mean feeling or judgement. Fix: state which account the question asks about. (1 mark)
- Social desirability if answers are not anonymous. Fix: anonymous responses. (1 mark)
Phase 3: Practise
Step 6 Learner practice (instant-marked)
1. Scale A gives tightly clustered scores but its average is far from the true value. Scale B gives scattered scores whose average is correct. Which statement is right?
Learners who swap them are treating clustering as accuracy. Clustering is consistency (reliability); being right on average is closer to validity.
2. A report says: “alpha = 0.55, so our scale is reliable.” Is the conclusion sound?
0.55 is below the usual 0.70 convention, so the items do not agree well. The report’s conclusion fails on its own number.
3. A report says: “scores two weeks apart correlated at 0.85, so the scale is valid.” Is the conclusion sound?
A high test-retest correlation shows reliability. It does not show that the scale measures the intended idea.
4. Respondents completed a survey that asked about their finances before asking about life satisfaction. Which threat is this?
Question order. Thinking about money first can shift the answer to a whole-life question.
Phase 4: Close
Step 7 Recap
- Reliability is consistency. Validity is measuring the intended idea.
- Reliability is necessary but not sufficient for validity.
- Four threats: context, social desirability, uneven scale use, question order.
- Each threat has a safeguard. None removes it entirely.
Step 8 Exit ticket (ungraded)
Think of a well-being survey you have answered. Which threat was most likely to affect your answers?
Private: nothing you type here is sent or stored.
Step 9 Go further (Extension)
Look up the four personal well-being questions asked in UK Office for National Statistics surveys (life satisfaction, worthwhile, happy yesterday, anxious yesterday). For each, say which of the three accounts from M1.L1 it mainly speaks to, and name one threat from this lesson that could affect the answers.
Common mistakes
| Mistake | Why it fails |
|---|---|
| Equating consistent with correct | Consistency is reliability. Correctness needs validity. |
| Treating 0.70 as a rule of nature | It is a convention, and acceptable values depend on purpose. |
| Dismissing self-report as worthless | It is imperfect but informative, and checkable. |
Limits. Reliability values depend on the sample tested. Validity is built from many studies, not one number.