How to Interpret Cronbach's Alpha Output in SPSS (and Report It in APA 7)

SPSS's Reliability Analysis gives you Cronbach's alpha in its second table, and most people stop there. The more useful table is the last one, Item-Total Statistics, which shows whether each item belongs in the scale and catches the most common mistake in scale scoring: a reverse-worded item that was never recoded. This guide reads every table column by column with a worked example that starts with exactly that mistake, lists the misreadings to avoid, and gives the APA 7 write-up. The reporting format alone is in the Cronbach's alpha APA guide.

What SPSS gives you

Analyze > Scale > Reliability Analysis, with the items in the Items box, Model: Alpha, and Item, Scale and Scale if item deleted ticked under Statistics > Descriptives for, produces:

1. Case Processing Summary: how many respondents were used.
2. Reliability Statistics: Cronbach's alpha. This is the number you report.
3. Item Statistics: mean and SD of each item.
4. Item-Total Statistics: how each item relates to the rest of the scale.
5. Scale Statistics: mean, variance and SD of the total score.

The worked example

Sixty students answered a six-item study-engagement scale on a 1 to 5 agreement scale. Item 4 is worded the other way round ("I find it hard to stay focused when I study"), so agreeing with it means less engagement. All numbers below come from one dataset.

First run: item 4 not yet recoded

The first analysis used the items exactly as entered.

An alpha of .360 for six items written to measure one thing is a warning. The Item-Total Statistics table shows why:

Item 4's Corrected Item-Total Correlation is −.700: the more engaged a student is on the other five items, the lower they score on item 4. And Cronbach's Alpha if Item Deleted for item 4 is .771, far above .360. That pattern, a strongly negative item-total correlation on an item you know is reverse-worded, means it was not recoded. It is not a reason to delete the item.

The fix: Transform > Recode into Different Variables, recode item 4 as 6 minus its value (1→5, 2→4, 3→3, 4→2, 5→1; on a 1 to k scale, k + 1 minus the value), and run the analysis again with the recoded item. If alpha itself comes out negative, SPSS adds a footnote saying the value is negative because of a negative average covariance among items and suggests checking item codings: the same problem, more severe.

Second run: Case Processing Summary

Footnote a reads "Listwise deletion based on all variables in the procedure." Anyone who skipped even one item is excluded from every statistic. Here all 60 were complete. If many are excluded, say so, and consider how the scale will be scored for people with missing items.

Second run: Reliability Statistics

• Cronbach's Alpha = .823: the internal consistency of the six items. This is the value you report, as α = .82.
• Cronbach's Alpha Based on Standardized Items = .819: the alpha you would get if every item were first converted to z scores, computed from the average inter-item correlation (.43 here). It matters only if the items are on different scales and you plan to standardize them before summing. Otherwise report the first column.
• N of Items = 6.

Second run: Item Statistics

Each item's mean and SD, after recoding. Look for items with very little variance (everyone answered the same way) or a mean at the end of the scale; such items cannot correlate well with anything. Item 4's mean (2.97) is now on the same footing as the others.

Second run: Item-Total Statistics

• Scale Mean if Item Deleted: the average total of the other five items.
• Scale Variance if Item Deleted: the variance of that five-item total.
• Corrected Item-Total Correlation: the correlation between the item and the total of the other items. "Corrected" means the item is left out of the total, so it is not correlated with itself. Values above about .30 are usually taken as acceptable. Items 1 to 5 are between .59 and .70; item 6 is weaker at .34.
• Squared Multiple Correlation: the R² from predicting the item from all the other items. It is another view of how much the item shares with the rest; not every SPSS setup prints it.
• Cronbach's Alpha if Item Deleted: what alpha would be without that item. For items 1 to 5 it is lower than .823, so each one helps. For item 6 it is .840, a little higher.

Should item 6 go? Not on this evidence alone. Dropping it would raise alpha from .823 to .840, a trivial gain, and might remove part of what the scale is meant to measure. Decide on content: if item 6 covers something the construct needs, keep it. Deleting items to push alpha up, especially in your own sample, overfits the scale to your data.

Second run: Scale Statistics

The mean and SD of the six-item total score (18.13 and 4.51), useful for the descriptives table if you analyse the summed score.

Misreadings that cost marks

• Reporting alpha from a run with an unrecoded reverse-worded item. A negative Corrected Item-Total Correlation is the giveaway. Recode and re-run.
• Deleting an item because "Alpha if Item Deleted" is higher. A small gain is not a reason. Item content decides.
• Reporting the standardized alpha by default. Report the Cronbach's Alpha column unless the scale is scored from standardized items.
• Reading alpha as evidence of one dimension. A high alpha can come from a long scale with several related factors. To show the scale measures one thing, use factor analysis.
• Reporting alpha without the number of items and the sample. Alpha rises with the number of items, so readers need both.
• Calling .70 a pass mark that applies everywhere. It is a convention. Many fields expect .80 or more for instruments used to make decisions about individuals.
• Using a leading zero. Alpha cannot exceed 1: write α = .82, not 0.82.

The APA 7 format template

Internal consistency of the [number]-item [scale name] was [acceptable/good/excellent] in this sample, Cronbach's α = .XX, 95% CI [.XX, .XX].

The write-up for this example

Item 4 was reverse-scored before analysis. Internal consistency of the six-item study engagement scale was good, Cronbach's α = .82, 95% CI [.74, .88] (N = 60). Corrected item-total correlations ranged from .34 to .70.

SPSS's Reliability Analysis does not print a CI for alpha by default. The interval above uses Feldt's method; in SPSS you can get the same kind of interval through the intraclass correlation coefficient option in the Statistics dialog (a two-way mixed, consistency ICC gives an average-measures value equal to alpha). If you report alpha without a CI, the sentence still works.

Let KyroStat do the write-up for you

Checking each item, recoding reverse-worded ones, and keeping the CI straight is where scale reporting goes wrong. KyroStat scores the scale (with reverse-scored items flagged), computes alpha with its confidence interval and the item-total statistics, and writes the APA 7 sentence with the code behind every number.

Also useful for the write-up (affiliate links): Proofreading for your thesis or paper (20% off with code PAP20): Paperpal

Frequently asked questions

Which number in the SPSS output is Cronbach's alpha? The Cronbach's Alpha column of the Reliability Statistics table. Report it with two decimals and no leading zero, together with the number of items.

What does a negative Corrected Item-Total Correlation mean? The item runs opposite to the rest of the scale. Usually it is a reverse-worded item that was not recoded. Recode it (for a 1 to 5 scale, 6 minus the value) and run the analysis again.

Should I delete an item if Cronbach's Alpha if Item Deleted is higher? Not automatically. A small increase is not worth losing content. Delete only if the item is also weak on content grounds, and say so in the method section.

Which alpha do I report, raw or standardized? The raw Cronbach's Alpha, unless your scale score is built from standardized items.

What is a good Cronbach's alpha? Conventions call .70 acceptable, .80 good and .90 excellent, but very high values (above about .95) can mean the items are redundant.

Which related guides might I need? Often these: Cronbach's alpha APA format, the intraclass correlation and descriptive statistics.