Split-half reliability calculator: Spearman-Brown and Guttman
The split-half method divides the scale into two parts, adds up the scores of each and checks how similar they are. Since each half has half the items, the correlation is corrected with the Spearman-Brown formula to estimate the reliability of the full scale. Paste the responses from Excel: one row per person and one column per item.
The calculation runs in your browser. We don’t store what you type.
How it’s calculated
Items are split into two halves: odd versus even (most recommended) or first half versus second half (SPSS’s default).
Each person’s score on each half is added up and Pearson’s correlation between the two sums (r) is calculated.
Spearman-Brown: 2r / (1 + r). If the halves have different numbers of items, the unequal-length version is used, like SPSS.
Guttman: 2 × [1 − (variance of half 1 + variance of half 2) / total variance]. It doesn’t assume both halves have the same variance.
How to read the result
- Read it like Cronbach’s alpha: 0.70 or more is acceptable, 0.80 or more is good and 0.90 or more is excellent.
- If Spearman-Brown and Guttman differ a lot, the two halves have very different variances: report Guttman.
- The result changes with how you split the items. Odd versus even spreads easy and hard items, and end-of-test fatigue, more evenly.
- Calculate one reliability per dimension, not one for the whole questionnaire.
Frequently asked questions
- Spearman-Brown or Cronbach’s alpha?
- Cronbach’s alpha is the average of all possible split halves, so it is more stable and the most requested. Use Spearman-Brown if your advisor or university asks for it, or as extra information.
- Why does the result change if I split differently?
- Because each split groups different items. That is why odd versus even is preferred, and why you should state how you split.
- What if I have an odd number of items?
- The halves have different lengths. The calculator shows the unequal-length Spearman-Brown, which is the right one in that case.
- Does it match SPSS?
- Yes. In SPSS, Reliability Analysis with the "Split-half" model splits into first and second half. Choose that option here to compare.
Know what you want to ask?
Build the survey in SurvIQ, share it by link, WhatsApp or QR code, and watch results come in live. The Free plan is free forever.
Create a free survey