Likert Scales: Five Points or Seven? And Do You Need a Midpoint?
The number of points pulls precision and response rate in opposite directions. The right choice depends on whether you want to detect fine differences or make a decision.
The Likert scale is the familiar "strongly disagree … strongly agree." The perennial question: how many points?
Five versus seven
| 5 points | 7 points | |
|---|---|---|
| Response speed | Faster | Slower |
| Discrimination | Lower | Higher |
| On mobile | Comfortable | Crowded |
| Best for | General surveys, decision-making | Research, comparing similar groups |
A practical rule: if you want to make a decision ("Which department has a problem?"), five is enough. If you want to track small changes over time, seven gives you more sensitivity.
The midpoint: keep it or drop it?
The middle option ("no opinion" or "neutral") does two different jobs, and they shouldn't be mixed up:
- Genuine neutrality: someone whose view really is in the middle.
- Avoiding the question: someone who doesn't want to think about it or take a position.
If you remove the midpoint (an even-numbered scale), the respondent is forced to pick a side. That's good for decision-making but makes the data slightly artificial.
A better solution: keep an odd-numbered scale, but add a separate option for "don't know / haven't experienced this" that is excluded from the average. That way genuine neutrality is separated from lack of information.
Label every point
A scale that is labeled only at the ends ("1 = very poor … 5 = very good") with bare numbers in between pushes respondents toward the middle. When every point has a label, the distribution becomes more realistic.
Direction: negative to positive, or the reverse?
In Persian, which is written right to left, the scale should start from the right. The more important point: keep the direction consistent throughout the questionnaire. If "very good" is on the right in one question and on the left in the next, some respondents won't notice and your data gets inverted.
The straight-lining trap
When 20 items in a row share one scale, some people tick a single column all the way down. You can detect this in the report: a response with zero variance submitted in a very short time. Flag such responses and set them aside in the analysis, but don't delete them, because their rate itself tells you your questionnaire is tedious.