Likert Scale

Home Forums Methodspace discussion Likert Scale

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
  • #602
    Godwin James

    I am using a questionnaire which was devised out of my interview data. I modified a questionnaire from an online source it has: strongly agree 1, agree 2, disagree 3, strongly disagree 4, don’t know 5. Most of the Likert scales I have seen generally have strongly agree as 5. Is there any problem with assigning the values as I have them on my questionnaire?

    Stephen Gorard

    It’s your instrument and you have to defend what you do.

    Some people would argue that not having a clear view on something is a valid response. So on a 5 points scale the middle one is used to represent ‘neither agree nor disagree, or don’t know’. Putting it in the middle shows the reader that it lies between slightly agree and slightly disagree. 

    Others would argue that too many people would use this mid point, and so they try to force an opinion. The scale is really 4 points, with a don’t know at the end. This has a few problems. People might misread as the order is no longer so intuitive. Denying people their most likely option sounds like the opposite of research. And the gap between the points becomes even more unbalanced (the distance from agree to disagree is enormous), making it harder to claim these represent numbers of any kind. 

    Yee Pek Har

    Hi Stephen, 

    May I know how do I to control for social desirability on questionnaires? Can I have two questions and see if they correlate?

    Thank you very much.  

    Dave Collingridge

    As far as I can tell you have a 4-point scale with the option to opt out (don’t know). A 4-point scale tends to really restrict the range of possible values to use in your statistics. Most tests look at variability between responses. If the possible variability is too small you may have a difficult time finding significant differences between groups if that is something you are interested in. If group comparisons are likely then consider a 5-point scale. If you are merely after descriptive statistics then a 4-point scale is probably fine.

    If this is an opinion survey then you might want to reconsider having a “don’t know” option. If people truly don’t know then you might ask why they are taking the survey in the first place. Or you might consider narrowing your inclusion criteria to people who truly have an opinion. Or you might consider dropping a question if there is a potential for a lot of “don’t know” responses. The purpose of an opinion survey is to get people’s opinions.

    I agree with Stephen Gorard that something in the middle of a Likert scale usually represents something between the SA and SD sides of the scale. In my opinion this should be “neutral” or “neither agree nor disagree”. Note that “neutral” is different than “don’t know”. I would not put don’t know in the middle because it basically represents opting out from answering the question. – 2 cents.


    One of the assumptions when giving a questionnaire is that your respondents have the information so, twhen the word don’t know should not appear anywhere. If you have a problem with the wording just use strongly agree on one end and strongly disagree on the other end. If you leave it as it is many pple will opt for don’t know

    Stephen Gorard

    Ignore this. It is clearty not true. Not knowing something is a legitimate response by anyone in any situation.  

Viewing 6 posts - 1 through 6 (of 6 total)
  • You must be logged in to reply to this topic.