Likert scales, commonly known as rating scales, are a popular research tool used to measure attitudes, opinions, and feelings. They can have varying degrees of scale points, such as 5, 7, or 10, and measure constructs such as agreement, importance, likelihood, and familiarity. Many factors go into creating Likert scales, and how you set up your scale and your battery can impact the insights you obtain. In this series, we’ll look at several commonly debated attributes of Likert scales, including:

• Number of scale points
• Labeling response options
• Number of attributes
• Formatting your battery

In this post, we’ll focus on the number of scale points. To determine the size of your scale, you first need to decide whether or not to include a mid-point (i.e., neutral option). Some researchers prefer to use even-numbered scales so that respondents are forced to choose a side.

Others prefer an odd-numbered scale to give respondents the ability to express a neutral view.

Which is correct? In our opinion, forcing respondents to pick a side when they genuinely feel neutral can result in skewed data, so for that reason, I prefer to use odd-numbered scales unless there’s a specific need not to have a mid-point.

Once you’ve decided on the mid-point issue, the next step is to determine how large the scale should be. Likert scales will typically range from 4 to 11 points, maintaining a balancing act between the level of differentiation required and ease of use. For example, a 9-point satisfaction scale will allow you to get very granular with the data but may be difficult for respondents to use because there are so many options.

On the other hand, a 5-point scale will be easier to answer, but might not show the level of differentiation needed to distinguish between attributes.

I use 5-point and 7-point scales on most rating questions and batteries. To determine which to use, we must look at: 

• Our client’s goals – as the scale must match the level of granularity they require.
• The degree to which we expect respondents to have strong opinions.
• Historical data significance – edit questions from old surveys with extreme caution, as changing the scale will make it difficult to draw comparisons.
• Whether the scale is uni-polar or bi-polar (more on that below)

When using a uni-polar scale, that is, one which goes from none to positive, a 5-point range typically works best. Importance is an example of a uni-polar scale.

Bi-polar scales, ones that range from positive to negative extremes, can be 5-point or 7-point. For example, a 5-point agreement scale allows the respondent to distinguish between somewhat and strongly, whereas a 7-point scale would have a little more granularity.

Scales with more than 7-points tends to be cumbersome, so we reserve those scales for Net Promoter questions (which are always 11-point scales) and historical questions where changing the scale is not allowed.

Check back soon for a discussion on labeling!