Skip to: Main Content / Navigation

  • Facebook
  • Twitter
  • LinkedIn
  • Add This

All Forums > Survey Development

Four point scale

I know there has been some discussion about a 5 point sclae vs. a 7 point sclae vs. a 10 point sale.

But, I recently heard someone say they like to use a 4 point scale in order to force people to choose either a positive or negative preference. Their belief is that many people have opinions but choose the middle option because they are "fence sitters".

What are your thoughts?

Likert Scales

While I am an advocate of even scales, IMHO a 4-point is much too narrow and want capture variability well enough. My preferred would be at least a 6-point minimum.

Six point scale is better than 4!

I completely agree that if you are going to use even scales you would need to do at least 6 points...otherwise you are almost forcing people to be too negative or positive.

Often when I participate in a study I choose the neutral option (out of laziness) when I really should be choosing a "doesn't apply" or "never used" option.

I could see where an even scale with those options would get the most accurate results.

range of scales

My two cents:

The range of the scale should make sense given the question that is being asked. I have been at project planning meetings where people with no research experience have said things like, "make all the scales five points" and "a seven point scale is always better--let's do that for all the questions." Neither comment was logical under the circumstances. If the question is mundane like, "How much do you like chocolate ice cream?" providing a 10-point scale is illogical. It only implies a superior level of precision and the reliability of an item like that is suspect. (If the respondent picks a 6 today, will they pick a 6 tomorrow?) Likewise, if the question is amenable to a particularly precise response, you don't want to force the respondent into a falsely negative or positive category (as can occur with a four point scale).

Four point scale

Several great points have been made regarding the value of a larger scale in order to have better discrimination of results. But, to get back to your original question which I understood to be more of a question about odd versus even numbered scales, I see nothing wrong with an odd numbered scale (e.g., 5 or 7 values). If someone is "fence-sitting" then they are doing it for a reason; they clearly do not feel strongly about the issue. If you force them into taking a position, then I would suggest you are getting a false read from your respondents. For instance, in a low involvement product (let's say a pad of paper), why would you put a gun to a respondent's head and say choose Yes or No, when the real answer might be "I just don't care"? Sometimes the number of fence-sitters is very valuable information.

For what it's worth,
Dave Bernstein