What is Inter‑item Correlation?
Inter-item correlation measures how closely responses to two survey items move together — it shows whether items intended to measure the same idea are related. It helps diagnose whether items belong together, are redundant, or aren’t measuring the same thing.
Inter-item correlation is the statistical association between scores on two items within a questionnaire or scale. For each pair of items you calculate a correlation coefficient (commonly Pearson’s r for continuous items or polychoric correlation for ordinal/Likert items). High positive correlations mean respondents tend to answer the two items similarly; low or near-zero correlations mean the items are unrelated; strong negative correlations indicate items are moving in opposite directions (often a sign of reverse-wording or a problem). Practically, researchers look at the pattern of pairwise correlations and the average inter-item correlation to judge internal consistency, redundancy, and whether items form a coherent scale.
Usage example
You create a 5‑question ‘service satisfaction’ scale. After collecting responses, you compute pairwise inter-item correlations and find most item correlations around 0.4–0.6, one item correlates at 0.08 with the others, and two items correlate at 0.92 with each other. The low-correlation item may not measure satisfaction and should be revised or removed; the very high pair suggests those two items are redundant and could be consolidated.
Practical application
Inter-item correlations matter because they are a simple, practical check on the quality of a multi-item measure. They help you: identify items that don’t belong in a scale (low correlation), spot nearly duplicate items you can remove to shorten a survey (very high correlation), inform factor analysis and Cronbach’s alpha interpretation, and improve response quality by ensuring questions are consistent and clear. For multilingual surveys, low inter-item correlations can also reveal translation or cultural differences; using automatic translation with a feedback loop (as Hearo does) helps identify and fix wording that harms item relationships.
FAQ
How is inter-item correlation different from Cronbach’s alpha?
Inter-item correlations are the pairwise correlations between items. Cronbach’s alpha summarizes internal consistency across all items and depends on both the average inter-item correlation and the number of items. Alpha can be high either because items correlate well or because you have many items; looking at inter-item correlations helps you spot problematic items that alpha alone can hide.
What correlation values should I expect or aim for?
There’s no single correct value, but common guidance is: average inter-item correlations between about 0.15 and 0.50 indicate reasonable homogeneity without redundancy. Pairwise correlations around 0.3–0.7 are often desirable. Values near zero suggest unrelated items; values above ~0.8–0.9 suggest redundancy.
Can I get negative inter-item correlations?
Yes. Negative correlations occur when items are oppositely worded or measure different constructs. If items are intended to be aligned, negative correlations often signal reverse-worded items haven’t been correctly recoded or that an item is problematic and needs rewriting or removal.
What should I do if correlations are low or inconsistent across language versions?
First check for coding errors and proper reverse‑scoring. If that’s not the issue, review item wording — poor translation or cultural mismatch can reduce correlations. Pilot the items, gather participant feedback about phrasing (flagging), and revise translations. In Hearo’s workflow, participant feedback and admin review make it easier to locate and fix wording that breaks item relationships across languages.