Likert Scale: Definition, Examples, 5 vs 7 Points and Analysis

What is a Likert scale?

The Likert scale goes back to the psychologist Rensis Likert, who introduced it in 1932 to measure attitudes. Respondents rate their agreement with a series of statements (items) on an ordered response scale. The scores of all items that measure the same attitude are summed or averaged into a scale score.

In everyday use, any agreement rating is called a “Likert scale”. Methodologists distinguish: a single question with agreement options is a Likert-type item; the combination of several such items into one measure is the Likert scale. The distinction matters for analysis.

Response formats: 5 vs 7 points and other choices

There is no single correct number of points, but some conventions are well established.

• Label all points, not just the ends, for clearer interpretation.
• Midpoint or not: keep a neutral midpoint if “neither” is a genuine attitude; offer a separate “don’t know” option for people without an opinion.
• Stay consistent: use the same format and direction for all items in a scale.
• If you use an established scale, keep its original format so you can compare results.

Other types of response scales

Agreement is not the only dimension. Depending on the item, frequency, importance or satisfaction may fit better. Keep the stem and the response scale consistent: “How often …?” needs frequency labels, not agreement labels.

Writing good Likert items

Each item should be a clear statement expressing a position, not a question. Use several items per concept (typically 3–10) so that the scale score is reliable.

• One idea per statement (no “and”).
• Avoid extreme wording (“always”, “never”) that almost nobody agrees with.
• Mix in some reverse-coded items (“I rarely feel motivated to study”) to detect careless responding, but not so many that they confuse respondents.
• Avoid negations inside statements where possible; “not unhappy” is hard to rate.

How to analyze Likert data

The analysis depends on whether you work with single items or scale scores.

Single items are ordinal. Report frequencies and percentages, the median and the interquartile range, and use non-parametric tests such as the Mann–Whitney U test or Spearman’s correlation. A diverging stacked bar chart shows the distribution well.

Scale scores (mean or sum of several items) have many possible values and are, by broad convention, treated as approximately interval. You can report means and standard deviations and use t-tests, ANOVA, Pearson correlations and regression. Methodologists still debate this, so state your choice and, if in doubt, check whether parametric and non-parametric tests agree.

Reliability: Cronbach’s alpha

Before you combine items, check whether they measure the same thing. Cronbach’s alpha is the standard index of internal consistency. Values of .70 or higher are acceptable, .80 or higher good. If alpha is low, inspect the corrected item-total correlations: items below about .30 may not fit the scale. Do not delete items just to push alpha up without a content reason, and report any changes.

Common mistakes

These errors often cost marks:

• Forgetting to recode reverse-coded items before averaging.
• Reporting the mean of a single Likert item as if it were precise interval data, without comment.
• Mixing 5- and 7-point formats within one scale.
• Creating a “scale” from items that do not belong together and never checking alpha.
• Labeling only the endpoints, which makes the midpoints ambiguous.

Tools and institutional requirements

At /umfragen you can build Likert scales with labeled points, mark reverse-coded items and get Cronbach’s alpha per scale automatically. The reliability calculator works with your own data too. Departments differ on whether they accept parametric tests for Likert data; your university’s own guidelines take precedence. Upload them to Professor Digital and the analysis suggestions are tailored to them.