Likert Scales: How to Design, Score, and Analyze Them

If you have ever filled in a questionnaire that asked you to rate statements from “Strongly disagree” to “Strongly agree”, you have already used a Likert scale. It is the most common way researchers measure attitudes, opinions and perceptions — and yet it is also one of the most commonly mishandled parts of a student thesis. This guide covers the whole journey of Likert scale questionnaire analysis: how to design good items, score the responses correctly, and analyse the results in a way your supervisor will accept.

What Is a Likert Scale?

A Likert scale is a psychometric rating scale named after the psychologist Rensis Likert, who introduced it in 1932. Each item on the scale is a statement — for example, “I feel confident using statistical software” — and the respondent indicates how much they agree or disagree with it by choosing one point on a fixed set of ordered options.

A typical five-point scale looks like this:

• 1 — Strongly disagree
• 2 — Disagree
• 3 — Neither agree nor disagree
• 4 — Agree
• 5 — Strongly agree

Two things are important here. First, a single Likert item is one statement; a Likert scale is a set of several items that together measure one underlying concept, such as job satisfaction or research anxiety. Second, the numbers are labels for ordered categories: we know that “Agree” is more than “Neither agree nor disagree”, but we cannot be sure the gap between 1 and 2 is exactly the same size as the gap between 4 and 5. That detail matters when you get to the analysis stage, as we will see below.

When Should You Use a Likert Scale?

Likert scales are a good choice when you want to measure something that cannot be counted directly — an attitude, a perception, a feeling, or a judgement. They work well for student satisfaction surveys, attitudes toward online learning, self-reported confidence, perceived stress, and opinions about a new teaching method. They are a poor choice when you need an exact measurement (use a real number, such as hours or test scores, instead) or when a simple yes/no answer would do.

How to Design a Likert Scale: 5 Steps

Step 1: Define Exactly What You Are Measuring

Before you write a single item, write one sentence that defines your concept. “Student satisfaction with online learning” is a concept; “online learning” is not. A clear definition keeps your items focused and stops your scale from drifting into two or three different topics at once. If your definition contains the word “and” twice, you are probably measuring more than one thing — split it into separate scales.

Step 2: Write Clear, Single-Idea Statements

Each item should express one idea only. “My supervisor gives helpful feedback quickly” is two ideas (helpful, and quick), so a respondent who gets helpful feedback slowly cannot answer honestly. Split it into two items. Keep your language simple and concrete, avoid jargon, and avoid leading words such as “excellent” or “terrible” inside the statement itself. Finally, include a mix of positively and negatively worded items — for example, “I enjoy group work” alongside “I find group work stressful” — because a set of items that all point the same way invites careless, straight-line answering. Negatively worded items must be reverse-scored later (see the scoring section below).

Step 3: Choose Your Response Scale

The five-point scale is the standard choice for student research: it is easy for respondents, and it gives you enough variation to analyse. A seven-point scale gives finer detail and slightly more statistical power, but it can feel like overkill for a short classroom survey. Whatever you choose, label every point with words, not just numbers — “1” means nothing on its own, while “Strongly disagree” does. Keep the same scale for every item in the questionnaire, and put the positive end on the same side throughout so respondents do not have to re-read the options for each question.

Step 4: Decide About the Neutral Midpoint

A five-point scale has a middle option (“Neither agree nor disagree”). That is honest: some respondents genuinely have no opinion. Removing the midpoint forces people to take a side, which can distort your results. Keep the midpoint unless you have a specific reason to force a choice — and if you do remove it, say so in your methodology chapter and explain why.

Step 5: Pilot Test Your Questionnaire

Ask 5–10 people from your target group to complete the draft and tell you which items confused them. A pilot test almost always catches at least one ambiguous item, and it gives you a first check of your scale’s reliability before you collect your real data. Do not skip this step: it is far cheaper to fix an item now than to explain a broken scale to your examiner later.

How to Score a Likert Scale

Scoring turns words into numbers you can analyse. Follow these rules:

1. Assign numbers to the response options. On a five-point scale, Strongly disagree = 1 through to Strongly agree = 5.
2. Reverse-score the negatively worded items. If an item is worded negatively (“I find group work stressful”), a high score means the opposite of the other items, so flip it: 1 becomes 5, 2 becomes 4, 3 stays 3, 4 becomes 2, and 5 becomes 1. On a five-point scale the formula is: reversed score = 6 − original score. Forgetting this step is the single most common scoring error in student theses.
3. Combine the items into a scale score. Add up (or average) the items that measure the same concept. A total score and a mean score carry the same information; the mean is easier to interpret because it stays on the original 1–5 scale. Report both the mean and the standard deviation.
4. Keep single items and scale scores separate. You can report the percentage who agreed with an individual item, but do not run your main statistical tests on one item alone — single items are noisy, while a multi-item scale is far more reliable.
5. Check reliability before you analyse. Run Cronbach’s alpha on your items (most statistics software does this in a few clicks). A value of 0.70 or above is generally considered acceptable for student research; below that, look for a weak or misworded item and consider dropping it.

Worked Example: Scoring a 5-Item Job Satisfaction Scale

Imagine a short job satisfaction scale with five items. Item 3 is negatively worded, so it must be reverse-scored. Here are the responses from six employees:

*Item 3 is negatively worded (“I often think about leaving this job”). Respondent A’s original answer of 2 reverses to 4 (6 − 2 = 4), so A’s total is 4 + 5 + 4 + 4 + 5 = 22, and the mean is 22 ÷ 5 = 4.4. Across all six employees the group mean is 3.5, which sits just above the neutral midpoint of 3 — mild overall satisfaction, with a spread from 2.0 to 4.6. Notice how different the picture would be if Item 3 had not been reversed: respondent D, the least satisfied employee, would have looked like one of the most satisfied.

How to Analyse Likert Scale Data

The honest answer is that researchers disagree about this — and your examiner will expect you to know that. Strictly speaking, Likert responses are ordinal data: the order is meaningful but the gaps between points are not guaranteed to be equal, so purists recommend non-parametric tests such as the Mann-Whitney U test or Spearman’s correlation. In practice, however, most published research treats summed scores from multi-item scales as approximately interval data and uses t-tests, ANOVA and Pearson’s correlation — and reviewers generally accept this when the scale has at least five items, the sample is reasonably large, and the data are roughly normally distributed.

A safe, defensible approach for a thesis is:

• Start with descriptives: report the mean and standard deviation for each scale, plus the percentage of respondents who agreed or disagreed with each individual item.
• Check reliability first: report Cronbach’s alpha for every scale before presenting any group comparisons.
• Match the test to the data: use t-tests or ANOVA for comparing groups on a multi-item scale score when the data look roughly normal; use Mann-Whitney U or Kruskal-Wallis when the data are clearly skewed or the sample is small.
• Report effect sizes: a p-value tells you whether a difference is likely real, but an effect size (such as Cohen’s d) tells you whether it matters. Report both.
• State your choice in the methodology: write one sentence explaining that you treated summed Likert scores as interval data, or that you used non-parametric tests because the data were ordinal. Either choice is acceptable if you justify it.

6 Common Mistakes to Avoid

1. Forgetting to reverse-score negatively worded items — this silently corrupts your total scores, as the worked example shows.
2. Writing double-barrelled items — one idea per statement, always.
3. Mixing different response scales in one questionnaire — keep the same 1–5 options for every item.
4. Analysing one item as if it were a scale — single items are unreliable; your conclusions should rest on the combined scale score.
5. Reporting only the mean — always add the standard deviation, so readers can see how spread out the responses were.
6. Skipping the pilot test — an untested questionnaire is the most expensive mistake in survey research, because you only discover the problem after the data are collected.

Frequently Asked Questions

How many items should a Likert scale have?

Most well-designed scales have between 4 and 10 items per concept. Fewer than 4 items makes reliability hard to establish; more than 10 risks boring your respondents into careless answering. Five to seven items is a sensible target for a student questionnaire.

Should I use a 5-point or 7-point Likert scale?

Use 5 points for most student research — it is simple, familiar, and gives enough variation. Choose 7 points only when you need finer distinctions and your respondents are motivated enough to use them, for example in a study of expert judgements. Do not mix the two in one questionnaire.

Can I use a t-test on Likert scale data?

Yes, in most cases — provided you are testing a summed score from a multi-item scale (not a single item), your sample is reasonably large, and the scores are roughly normally distributed. If any of those conditions fail, use the non-parametric equivalent (Mann-Whitney U instead of an independent-samples t-test) and say why.

What is a good Cronbach’s alpha for a Likert scale?

A value of 0.70 or above is generally considered acceptable, 0.80 or above is good, and 0.90 or above is excellent. Below 0.70, check whether one item is dragging the value down — dropping a weak item often raises alpha noticeably. Report the alpha value for every scale in your results chapter.

Key Takeaways

A good Likert scale starts with a clear definition, uses simple single-idea items, keeps one consistent response format, and is pilot tested before data collection. Score it by reversing negatively worded items and combining the items into a mean score, check its reliability with Cronbach’s alpha, and then choose your statistical test to match the shape of your data. Get those steps right, and Likert scale questionnaire analysis becomes one of the most straightforward parts of your thesis — not the part your examiner questions.