Hi Nawaz,

What you ask about is just one of the many controversies about Likert scales that remain debated to this day. Those of us who used these scales always have concern about exactly what a mid-point (e.g. "3" on a scale of 1 to 5) actually means. As you note, some participants may simply opt for the middle option to opt-out of participating.

That said, there is real evidence that people do adopt a decisively neutral sort of attitude about some things. Specifically, people seem to take less extreme, even largely neutral attitudes about topics or targets of little importance. People's attitudes seem to bifurcate into extremes (e.g. "strongly agree vs. strong disagree") was some topic becomes increasingly important. Bibb Latane and Andrzej Nowak discuss this in their chapter found within the 1994 text "Dynamical Systems in Social Psychology," for example.

Of course, for some studies there is a need to more clearly identify whether people agree or disagree with a given item. In this case, it's common practice to use a sort of Likert scale + forced-choice hybrid. For this type of scale, one simply uses a Likert scale with an even number of response items. For example:

1 = Strong disagree; 2= Disagree; 3 = Slightly disagree; 4 = Slightly agree; 5 = Agree; 6= Strongly agree

One concern with this sort of scale, though, is that participants must reveal agreement/disagreement, which could be a problem with controversial questions. There is a way around this. The survey could have an additional option for each question - "0 = Prefer to not respond". With this option, it is then very clear to the researcher that a participant did not wish to respond. That is much clearer than "neutral."

I hope this helps!

Hi Nawaz,

What you ask about is just one of the many controversies about Likert scales that remain debated to this day. Those of us who used these scales always have concern about exactly what a mid-point (e.g. "3" on a scale of 1 to 5) actually means. As you note, some participants may simply opt for the middle option to opt-out of participating.

That said, there is real evidence that people do adopt a decisively neutral sort of attitude about some things. Specifically, people seem to take less extreme, even largely neutral attitudes about topics or targets of little importance. People's attitudes seem to bifurcate into extremes (e.g. "strongly agree vs. strong disagree") was some topic becomes increasingly important. Bibb Latane and Andrzej Nowak discuss this in their chapter found within the 1994 text "Dynamical Systems in Social Psychology," for example.

Of course, for some studies there is a need to more clearly identify whether people agree or disagree with a given item. In this case, it's common practice to use a sort of Likert scale + forced-choice hybrid. For this type of scale, one simply uses a Likert scale with an even number of response items. For example:

1 = Strong disagree; 2= Disagree; 3 = Slightly disagree; 4 = Slightly agree; 5 = Agree; 6= Strongly agree

One concern with this sort of scale, though, is that participants must reveal agreement/disagreement, which could be a problem with controversial questions. There is a way around this. The survey could have an additional option for each question - "0 = Prefer to not respond". With this option, it is then very clear to the researcher that a participant did not wish to respond. That is much clearer than "neutral."

I hope this helps!

Dear Jay,

Your argument seems logical. In my opinion, if someone have 5 point likert scale then it suppose to be from -2 to +2. that is how neutral will have 0 score. It sounds better option as prescribed by William Zikmund. 

Hi Nawaz

Jay's arguments are very valid, but not always, because when the objective is to measure the neutral perception assessment takes much sense in the sense that through this ignorance can be determined. For example, if we measure the brand image attributes relative to Factor Analysis using principal components, with statements and through Likert scale 1-5 where 3 is indifferent or neutral, allowing the respondent to choose the midpoint when no know of any relationship between a brand and a particular attribute or construct. indeed, one could measure the degree of knowledge that exists on brands constructs cirtos or through this level Likert scale.

Hi Critobal,

Yes, Thats my argument that if a respondent does not know and/or marking Neutral, give him/her Zero score (following -2 to +2 scale)

Hi all,

Jay got it all down with his answer re the neutral option. It nicely captures the respondent's intention. However, to keep the ordinal scale level for the answers from those who do respond, do not code "Prefer to not respond" as 0.

Further, you may want to add a second neutral option "don't know" to capture another reason why some respondents may otherwise choose the middle of the scale.

For a reference re such issues in Web-based questionnaires, you may want to check out 

Reips, U.-D. (2010). Design and formatting in Internet-based research. In S. Gosling & J. Johnson (eds.), Advanced methods for conducting online behavioral research (pp. 29-43). Washington, DC: American Psychological Association.

available from http://www.uni-konstanz.de/iscience/reips/pubs/papers/Reips2010_Chapter_3_Design.pdf

In my opinion, when multiple questions measure one factor, the neutral option has no drawbacks. If a participant chooses to opt out of answering, you can detect a clear pattern of neutral answers and remove that data from your analysis; if a participant really is mostly neutral about something, multiple questions make sure that you get a reliable score.

So, basically, one should avoid using Likert scales for single questions.

"So, basically, one should avoid using Likert scales for single questions." (Mykolas)

Dear Mykolas,

Suggest then what to do

I would suggest, if possible, adding more questions to measure the same construct and use a 5-point the Likert scale. If that option is not possible, I would suggest using a 7-point scale for better data distribution. A large sample of participants would also help.

Why isn't anyone talking about semantic differential method here? Don't give any numbers but just two extreme ends (1) and (10) and two adjectives describing it and let them freely mark anywhere along the line to represent their feeling about it.

In surveys I have used a 5-point Likert scale with a neutral option, but also a separate "Not applicable/not sure" to distinguish those with a genuinely neutral opinion to those for whom there is no clear answer.