Cronbach's Alpha Calculator

StatisticsLast updated: August 17, 2026

Cronbach's alpha (α) is the coefficient of internal consistency reliability: it shows how consistently the items of a scale measure the same construct, and it is reported in almost every thesis and paper that uses a questionnaire. Alpha normally runs between 0 and 1; the closer it is to 1, the more the items agree with one another and the more consistent the measurement is. Findings obtained with a scale of low reliability cannot be interpreted soundly, however large the sample.

Paste in a score matrix with one row per respondent and one column per item (copy-paste from Excel or SPSS works): the calculator returns the alpha coefficient, its interpretation against the George and Mallery cutoffs, and an item analysis table with the mean, standard deviation, corrected item-total correlation, and "alpha if item deleted" for each item. Always reverse code your reverse-worded items before entering the data.

Separate the scores with spaces, commas, or tabs; you can paste straight from Excel. At least 3 respondents and 2 items are required. Reverse code reverse-worded items first.

Cronbach's Alpha Reliability Cutoffs

The alpha coefficient is interpreted against George and Mallery's widely cited classification; the generally accepted lower bound for research use is 0.70. The cutoffs below are the boundaries this calculator uses to produce its "reliability level" label. In item analysis, a corrected item-total correlation of at least 0.30 is also expected.

Excellent
α ≥ 0.90
The level expected in clinical and selection measures where decisions are made about individuals; values above 0.95 can point to item redundancy
Good
0.80 – 0.90
The band most often reported in theses and papers
Acceptable
0.70 – 0.80
The generally accepted lower bound for research use
Questionable
0.60 – 0.70
Accepted only in exploratory studies, and only with caution
Poor
0.50 – 0.60
The scale needs to be reviewed through an item analysis
Unacceptable
α < 0.50
Negative values almost always come from reverse-worded items that were not reverse coded

The calculator labels the alpha coefficient it computes against these cutoffsKaynak: George, D. and Mallery, P. (2003) — SPSS for Windows Step by Step

Formula

α = (k / (k − 1)) · (1 − Σs²ᵢ / s²ₜ)
k: number of items
Σs²ᵢ: sum of the item variances
s²ₜ: variance of the total score

If the items agree with one another, the variance of the total score is far larger than the sum of the item variances and alpha approaches 1; if the items are unrelated, the two variances converge and alpha drops toward zero. The variances use the sample formula (with the n − 1 denominator), so the results match SPSS exactly.

How to Calculate

  1. Paste your data into the box with one row per respondent and one column per item; the scores can be separated by spaces, commas, or tabs.
  2. If your scale has reverse-worded items, reverse code them before entering the scores (for a 5-point Likert item: new score = 6 − old score).
  3. Read the alpha coefficient and its interpretation against the George-Mallery cutoffs in the results section.
  4. In the item analysis table, check the items whose item-total correlation falls below 0.30 and look at the 'alpha if deleted' column.
  5. Base the decision to drop an item not on the statistics alone, but on content validity and a theoretical justification as well.

Worked Examples

A 5-item scale with 10 respondents

For the 5-item scores of 10 respondents, α = 0.940 — excellent internal consistency. All item-total correlations are high (0.942 for Item 1, for example) and deleting any item lowers alpha: dropping Item 1 brings it down to 0.907.

Cronbach's alpha (α): 0.940 · Reliability level: mükemmel · Number of items (k): 5

An item that was not reverse coded (error example)

In this 6-respondent data set, item 2 is reverse worded and was entered without being reverse coded: alpha collapses to the meaningless value of -1.176. The item-total correlation of Item 2 is -0.966, and deleting that item raises alpha to 0.943 — the classic signature of a reverse-coding mistake.

Cronbach's alpha (α): -1.176 · Reliability level: kabul edilemez · Number of items (k): 4

Frequently Asked Questions

What is a good Cronbach's alpha?
The general convention is that a scale should reach an alpha of at least 0.70 for research use. In the George and Mallery classification, 0.80–0.89 is good and 0.90 or above is excellent. Values between 0.60 and 0.69 are questionable and are accepted only cautiously, in exploratory work. For clinical measures used to make decisions about individuals, 0.90 and above is expected.
What does a negative alpha mean?
A negative alpha means the average covariance between the items is negative, and it almost always signals a data coding problem. The most common cause is reverse-worded items entered without being reverse coded. Recode those items and compute again; if the problem persists, your items most likely do not measure the same construct.
Is an alpha above 0.95 a problem?
A very high alpha is not always good news. Values above 0.95 can indicate that the items are near duplicates of one another (item redundancy). The scale is then longer than it needs to be, and dropping some of the overlapping items reduces the burden on respondents. Take the number of items into account as well: 0.95 is unremarkable in a 40-item scale but striking in a 5-item one.
How many respondents does a reliability analysis need?
The common recommendation in the literature is at least 30 respondents for a stable estimate of alpha; scale development studies suggest 5–10 respondents per item and a total of at least 100–300 people. This calculator works from 3 respondents upward, but remember that in small samples the coefficient can vary noticeably from sample to sample.
How do I interpret the item-total correlation?
The corrected item-total correlation is the Pearson correlation between an item's score and the sum of the remaining items. The usual criterion is at least 0.30: below that, the item is not measuring the same construct as the rest of the scale, and values near zero or negative indicate an item unrelated to the scale or a reverse-coding error.
How is alpha reported for a scale with subscales?
If the scale consists of several subscales, alpha should be computed and reported separately for each subscale rather than for the scale as a whole. Pooling items that measure different constructs can inflate or deflate alpha artificially. Common practice is to present the alpha of each subscale in a table and, if a total score is also used, to give its alpha separately.