# The optimal number of response categories in the Likert scales

**URL:** <https://discourse.datamethods.org/t/the-optimal-number-of-response-categories-in-the-likert-scales/5546>\
**Category:** general\
**Tags:** categorization, measurement, response-variable\
**Created:** [May 2, 2022, 2:51pm UTC](https://discourse.datamethods.org/t/the-optimal-number-of-response-categories-in-the-likert-scales/5546 "2022-05-02T14:51:23Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![Leni](https://discourse.datamethods.org/letter_avatar_proxy/v4/letter/l/839c29/32.png) [@Leni](https://discourse.datamethods.org/u/Leni)\
**Post date:** [May 2, 2022, 2:51pm UTC](https://discourse.datamethods.org/t/the-optimal-number-of-response-categories-in-the-likert-scales/5546/1 "2022-05-02T14:51:23Z")

</div>

Hi, everyone.  
I want some advice on using the Likert scale. I have some questions that need your help in answering.

1. Is neutral choice (middle point) necessary on the Likert scale? Why? Because I found that some researchers used odd points while the others used even points on the Likert scale.
2. I know that the Likert scale has variations in response categories from 2-point to 10-point scales. What is the optimal number of response categories that is preferable to use? Why?

Thank you.

---

<div class="post-metadata">

**Author:** ![f2harrell](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/f2harrell/32/165_2.png) [@f2harrell](https://discourse.datamethods.org/u/f2harrell)\
**Post date:** [May 2, 2022, 4:25pm UTC](https://discourse.datamethods.org/t/the-optimal-number-of-response-categories-in-the-likert-scales/5546/2 "2022-05-02T16:25:59Z")

</div>

From a general statistics standpoint, the more the better. I likek using slider buttons on web forms that provide a continuous output (actually internally coded as 0-100 but the users don’t see the numbers).

---

<div class="post-metadata">

**Author:** ![R\_cubed](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/r_cubed/32/1518_2.png) [@R\_cubed](https://discourse.datamethods.org/u/R_cubed)\
**Post date:** [May 2, 2022, 10:38pm UTC](https://discourse.datamethods.org/t/the-optimal-number-of-response-categories-in-the-likert-scales/5546/3 "2022-05-02T22:38:54Z")

</div>

Came across this interesting paper on this very topic. I’m not so confident Rasch analysis can produce a metric scale from ordinal data, but it provides a decent introduction to the theory behind it.

> **[The importance of rating scales in measuring patient-reported outcomes -...](https://link.springer.com/article/10.1186/1477-7525-10-80)**
>
> Background A critical component that influences the measurement properties of a patient-reported outcome (PRO) instrument is the rating scale. Yet, there is a lack of general consensus regarding optimal rating scale format, including aspects of...

---

<div class="post-metadata">

**Author:** ![f2harrell](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/f2harrell/32/165_2.png) [@f2harrell](https://discourse.datamethods.org/u/f2harrell)\
**Post date:** [May 3, 2022, 12:01pm UTC](https://discourse.datamethods.org/t/the-optimal-number-of-response-categories-in-the-likert-scales/5546/4 "2022-05-03T12:01:18Z")

</div>

I’m not sure that their Rasch analysis is the right basis for favoring simple scales with low resolution. I would base the decision on the ability of point-in-time scales to predict ultimate outcomes, and on statistical power.

---

<div class="post-metadata">

**Author:** ![Leni](https://discourse.datamethods.org/letter_avatar_proxy/v4/letter/l/839c29/32.png) [@Leni](https://discourse.datamethods.org/u/Leni)\
**Post date:** [May 3, 2022, 6:24pm UTC](https://discourse.datamethods.org/t/the-optimal-number-of-response-categories-in-the-likert-scales/5546/5 "2022-05-03T18:24:42Z")

</div>

Thank you for answering my question @f2harrell & @R_cubed. It’s helped me a lot.