The Likert scale: what it is and how to use it

would be skewing the results towards a positive opinion since the respondent would only have one option to express their disagreement with the service

Within the symmetry, the scales can be bipolar, that is, provide response options that represent two extreme concepts around neutrality (Confused / Clear) or unipolar, that is, have as extremes the absence and the maximum value of something (Nothing of agree / fully agree). The unipolar option tends to be preferred because it is easier for respondents to understand.

The Likert scale has to be balanced.

To ensure that we can correctly measure the opinion of the respondents, it is essential not only that the distance between the different answer options be the same but also that the respondents themselves perceive it as such.

For example, a scale with the following response options would raise questions

Terrible – Bad – Regular – Good – Optimum – Excellent

Is the difference between “fair” and “good” the same as between “optimal” and “excellent” or between “terrible” and “bad”?

The answer options must be descriptive and precise.

One of the most common mistakes we can make when designing a survey is asking questions or providing answer options that are unclear and that can be understood differently by each respondent.

This is a problem that can also affect our Likert scale, so we have to make sure we are as precise as possible and not give rise to subjective interpretations.

For example, how much is “somewhat in agreement”? Is it more than “fairly agree” or “slightly agree”?

 
It may or may not have a neutral value.

There are different opinions about whether a Likert scale must have a neutral central value (“indifferent”, “neither agree nor disagree”…) or, on the contrary, only positive or negative values.

Although it is true that respondents may not have a formed opinion on a certain topic, supporters of omitting the core value argue that it is too often used as a wildcard for those who do not want to spend too much time thinking about their response and that the presence of only positive or negative values allows forcing them to express their honest opinion.

You have to mix positive and negative statements.

Another very common bias observed when analyzing survey responses is that participants tend to agree with the statements presented or indicate a positive attitude (acquiescence or consent bias). To mitigate this effect, it is good practice to present Likert scale item statements mixing positive and negative statements.

The SUS questionnaire, used to measure the perceived usability of an interface or product, is designed in this way, asking participants to indicate their agreement or disagreement about statements such as:

“I felt very comfortable using this website” Positive

“I find this website very uncomfortable to use” Negative

With the same objective of avoiding consent bias, it is advisable to formulate the statements as questions rather than statements (to which the participants usually agree by default).

For example, ask:

To what extent are you satisfied with the care received? (From “very dissatisfied” to “very satisfied”)

It is less subject to consent bias than assertion

I am satisfied with the care received (from “Strongly disagree” to “Strongly agree”)

Items must be grouped consistently

Let’s avoid jumping from one topic to another within the same list of items. In order for respondents to be able to focus on their answers, the items on a Likert scale that are displayed together should deal with different aspects of the same topic.