Hello Yusuf,

You may add the neutral option, but doing so may change some of the characteristics of the resultant scores. You could certainly run a quick pilot study to determine empirically whether this might be the case with your target population. (The reason hinges mostly on the debate between having the options being "forced choice"--no neutral option, vs. that of whether the neutral category captures only instances of genuine ambivalence vs. some other respondent attribute, such as not bothering to consider the stimulus.)

Reversing the order of the scale should have no effect on its utility, so long as the layout makes it perfectly clear to the respondents as to the direction of the scale. In countries where text is read from left to right, the usual practice is low end of the scale on left hand side, high(er) end of the scale on the right-hand side.

Good luck with your work.

Hello Yusuf,

You may add the neutral option, but doing so may change some of the characteristics of the resultant scores. You could certainly run a quick pilot study to determine empirically whether this might be the case with your target population. (The reason hinges mostly on the debate between having the options being "forced choice"--no neutral option, vs. that of whether the neutral category captures only instances of genuine ambivalence vs. some other respondent attribute, such as not bothering to consider the stimulus.)

Reversing the order of the scale should have no effect on its utility, so long as the layout makes it perfectly clear to the respondents as to the direction of the scale. In countries where text is read from left to right, the usual practice is low end of the scale on left hand side, high(er) end of the scale on the right-hand side.

Good luck with your work.

David Morse Thank you so much, Sir. I really appreciate it.

And if you don't mind, another question in the same vein would be: What about changing a 7-point Likert-style scale into a 5-point (with a neutral option) or even 4-point one (without a neutral option) before implementing it? Since I'm working with secondary and high school students, I tend to think that the lesser the number of responses on the scale is, the less confused students will be.

I would definitely add a neutral scale point. Categorical scales are always ordered and if you want to defend using them for calculating or estimating something, at least the assumption of a kind of "semi-interval" is essential. Avoiding the neutral position IMHO makes that assumption impossible as the difference between "certaintly true" to "true" is surely somewhat smaller than from true to not true.

Likewise, an important question is whether on the relevant latent response dimension, there is a true neutral stance. This is a theoretical question.

All the best,

--Holger

Yusuf Polat, you have received some thoughtful advice from both Holger Steinmetz and David Morse. Permit me to throw some additional considerations into the pot, however:

I'm not sure what the option "neutral" might mean, and I suspect it could mean different things for different people, and different things for different items. It could mean that the respondent has a genuinely "impartial" attitude to an item. However, it could mean other things such as attitudes lying quite strongly on both sides of the issue, but, because of that, the respondent takes the "average", which comes out as choosing neutral. Or it could mean "Don't know" (particularly if a don't know option isn't provided), or it could mean "Not applicable" (particularly if a not applicable isn't provided). Or it could mean simply undecided. Or perhaps "I'm not interested in this item".

So, in essence, I don't know what "neutral" really means. And that comes down to the validity of the scale you're using.

Yusuf Polat

Besides the insightful inputs in the previous comments, if planning to include a middle anchor on your scale, making it a five-point Likert-type scale, you might consider how the middle value may affect your calculation, and interpretation, of the scale items. In other words, is it the mean or the median you'll consider when looking into the central tendency?

Well, since Likert scales yield ordinal data, and your middle value might have a meaning, I would suggest you calculate a median, rather than a mean on your motivation scale.

Best,

Mohialdeen Alotumi, please feel free to disagree with me, but I think it's important to distinguish between items and scales when referring to Likert-type scales. Usually the latter are discussed in relation to a combination of individual items, and scores on scales are obtained by combining responses to individual items.

There is some (sometimes raging) disagreement about the nature of the data for both items and scales, but the most common view seems to be that items yield data at the ordinal level (even though many researchers ignore that by providing means and SDs to summarize item-level responses), and that scores on scales can be regarded as at the interval level.

If scores on scales (or subscales) are regarded as being at the interval level of measurement, I think it's quite OK to calculate such things as means and SDs for them rather than nonparametric descriptive statistics such as the median and interquartile range.

I hope what I've written is helpful.

Robert Trevethan, No disagreement on the obvious distinction between items and scale, and on what you've mentioned, "scores on scales are obtained by combining responses to individual items".

To be clear, since the would-be included midpoint might be construed as a non-neutral value, piloting the instrument, as suggested by David Morse, could enhance the validity and reliability of the survey. I do agree there is a debate on the matter among scholars whether to include a midpoint/not, and whether to consider the median or the mean. A good study that renders evidence-based suggestions is Chyung et al. (2017), which also highlights different views on the matter. Here I quote from p. 17:

"The presence of a midpoint makes a difference in treating the scale as an interval scale only if survey respondents use the Likert scale midpoint for a true neutral meaning, as intended."

I would recommend Yusuf Polat go through the following study, which has evidence-based recommendations when/when not to use a midpoint (see p.21-22).

Chyung, S. Y. Y., Roberts, K., Swanson, I., & Hankinson, A. (2017). Evidence-based survey design: The use of a midpoint on the likert scale. Performance Improvement, 56(10), 15–23. https://doi.org/10.1002/pfi.21727

Cheers,

Mohialdeen Alotumi, thank you for your response and for the citation to the article you've referred to. That kind of sharing is RG at its best. :-)

Dear Robert Trevethan, you're most welcome! I highly appreciate your enlightening insights on this thread and others.

4 point Likert scale is basically a forced Likert scale. The reason it is named as such is that the user is forced to form an opinion. There is no safe 'neutral' option. Ideally a good scale for market researchers, they make use of the 4 point scale to get specific responses. https://www.formpl.us/blog/point-likert-scale

Check https://www.limesurvey.org/blog/20-blog/108-likert-scale-how-to-properly-scale-your-survey-responses

Also check http://ankitmarketing.blogspot.com/2012/10/the-4-point-likert-scale.html?m=1