Cronbach’s alpha & internal consistency
When you build a multi-item scale to measure something — anxiety, job satisfaction, self-efficacy — you’re assuming the items pull together toward one construct. Cronbach’s alpha tests that assumption: it’s the standard measure of internal-consistency reliability, scored 0 to 1. If alpha is low, summing the items into a scale score isn’t meaningful — they’re not measuring the same thing.
How to read it
| Alpha | Interpretation |
|---|---|
| ≥ 0.90 | Excellent (but check for redundancy) |
| 0.80 – 0.89 | Good |
| 0.70 – 0.79 | Acceptable |
| 0.60 – 0.69 | Questionable |
| < 0.60 | Poor — items likely aren’t one construct |
These are rules of thumb, not law — acceptable thresholds vary by field and by how high-stakes the decision is (clinical instruments expect ≥ 0.90).
Two things people get wrong
- Higher isn’t always better. Alpha above ~0.95 often signals redundancy — several items asking the same question in different words. That pads alpha without adding information; consider shortening the scale.
- Alpha rises with the number of items. A long scale can post a high alpha even with mediocre items, so report the item count alongside it and don’t read alpha as a pure quality score.
Reliability is not validity
This is the big one. Alpha tells you the scale is consistent, not that it measures the right thing. A scale can be perfectly reliable and still measure the wrong construct — like a precise scale that’s miscalibrated. Reliability is necessary but not sufficient; you need validity evidence separately. (Internal consistency is just one kind of reliability, alongside test–retest and inter-rater.)
Get the free Data Collection toolkit
A measurement-and-reliability worksheet, questionnaire-design checklist, and a data-management-plan starter — from Research Design Simplified. We’ll email you the download link.
Frequently asked questions
What is Cronbach’s alpha?
A 0–1 measure of internal-consistency reliability — how well the items of a scale measure the same construct.
What’s a good value?
Roughly: ≥0.70 acceptable, ≥0.80 good, ≥0.90 excellent; <0.60 poor. Field and stakes matter.
Can alpha be too high?
Yes — above ~0.95 suggests redundant items. It also rises just from having more items.
Does high alpha mean my scale is valid?
No — alpha is reliability (consistency), not validity (measuring the right thing). You need validity evidence separately.
The Statistical Test Selection Workbook
58 decision trees, assumption checklists and output interpretation templates.
Get the workbook