A Likert scale which has an odd number includes a midpoint. Most Likert scales are 5-7. 7 is better than 5 and 9 is better than 7 to achieve measurement of a continuous variables. This is because the sensitivity of the measurement will be increased and be more confident of having a continuous rather than a categorical variable.

Dear Sandra. thanks for your suggestion. I have read some of research articles by Colman and Norris (1997), Corolyn et al, (1999) and Norman (2010). They sugested Likert sacle 7 is most appropriate to get high mean scores and participants easily can be answered.

I am completely right with your view to get continious varibales, likert scale 9 will be more appropriate. Can you please suggest any researches to justify this one.?

Try this reference:

Please vote for my answers if you think they are right.

http://link.springer.com/article/10.1007/s10459-010-9222-y#page-1

Although I would agree that 7 can be better than 5, I am not so sure about 9 being better than 7.

Once you move above 7 the differences between each point on the scale are more subtle and people have more difficulty in distinguishing between them (but this may depend on what your question is asking)

Hi Yogarajah,

I think that George has a good point.

Respondents can not diffentiate between the Likert scale points if they become spaced too close together. You have to differentiate between statistical and practical considerations, and try to find a balance.

Place yourself in the position of your respondent.

Imagine that a respondent in first instance is reluctant to fill out your questionaire, unless it is his/her hobby. Secondly, he or she doesn't want to give too extreme answers. Thirthly, do not understand the difference between the Likert points when they are indistinguishable from each other. Fourtly doesn't want to study your quenstions and answers to find subtile differences between the answer categories (factor time).

Maybe theoretically the more Likert scale points the better, but practice proves otherwise. I would chose for 7 or less scale points. Fewer points are better, because they force the respondent chose a well defined answer category.

I would label EACH answer category to help the repondent to make an informed choice, and I would add an extra answer for respondents who do not want to answer that paticular question (i.e. "no opinion" or "do not want to answer"). This saves you a lot of headace, because forcing a respondent to answer a question that he or she doesn't want to answer will pollute your analysis results.

And I would label EACH answer category with a clear and universal label. For example: Strongly agree, Agree, Neither, Disagree, Strongly disagree. You may notice, if you have problems in finding a suitable (logical) label for an answer category, you have too many answer categories.

I hope that this will help you.

Best,

Herman

I remember wanting to do 11 for my dissertation. Fortunately, my advisor talked me out of it. I used 5 instead for the reasons George and Herman articulated above, and I have had no difficulty interpreting the results, nor have any reviewers been at all surprised by the 5-point scale. Other things, yes, but not that.

Dear Clyde

It is very rare in practice that ordinal data are meeting parametric requirements. I feel if we go to atleast 7, then it will go slightly more like continious data. could you say that have you derived normal distribution with 5 Liker scale? please can you say your sample size?

Dear Yogarajah,

My sample was just short of 200. 197 or so. I recognize the issue you're talking about, but if you are using several items in each construct and taking the average score you wind up with a lot more than five possible values, which helps with the continous data issue. If you are using a single survey item when you operationalize your concept then you may have other difficulties. "How many new products did you introduce to the market in the last three years" you can ask once, but for managerial attitudes or perceptions you need several items.

Back to the 5 vs. 7 issue, I do think Geaorge and Herman make good points. Seven can make sense in a specific circumstance, especially if you can label EACH category as Herman suggests. If I were reviewing a paper for a journal and it used a 7-point scale instead of a 5-point one, I wouldn't have any problems with that. If I were on a dissertation committee, though, I would push for 5 because I think 5 gives you better odds of getting a large sample.

I hope this helps!

Dear clyde thx. As you meant several items for one constuct, it is fine we could add items then it might not arise much problems in getting normal distribution. But, other side, the researcher if try to establish the new concept, then the factor structure is unknown and exploratory in nature. In this case, he needs to do more advance statistical techniques such as EFA and SEM.

in this situation, If any one could you suggest, have you met the parametric requirments with your data set? What Likert scale supported well?

Hi there! Can anyone help me to identify the scale range for a 9 point likert scale? I have a questionnaire that uses a 9 point scale and I am in the midst of searching a research that states the range for it. Thank you in advance!