Cognitive Dissonance about Likert Scale

[fpetillo]

I am wondering if the typical format of the Likert Scale creates cognitive dissonance for survey respondents. The Likert Scale is bipolar (e.g. Strongly Disagree to Strongly Agree) and typically has number coding (e.g. 1, 2, 3, 4, 5). Presumably the middle point (if there is one) indicates a neuteral point between the extremes (e.g. neither agree nor disagree). But if this is true, shouldn't the coding scale run -2, -1, 0, +1, +2 ? From an analysis POV, any number range will do, but I'm wondering if respondents cognitively process different numerals differently. The simplest solution, I think, is to leave out the numerical coding from a survey form and just use the descriptive categories. But where numbers are used, do you think it creates any dissonance for the respondent to see a "3" where there really ought to be a "0"?

[Jake]

But if this is true, shouldn't the coding scale run -2, -1, 0, +1, +2 ?

Sometimes people do it this way. There is some work examining differences in self reports as a function of these kinds of labeling conventions. E.g., below:

Schwartz, N. (1999). Self-Reports: How the Questions Shape the Answers. American Psychologist, 54, 93-105.

[Jake]

BTW: I copied/pasted that reference, but I notice now that it says "Schwartz" when it should be "Schwarz"

[fpetillo]

Got it. Thanks!

[SmoothJohn]

You are assuming that the categories correspond to equal-size intervals. This is contested.

Unfortunately, the interpretation of the relative sizes of the intervals is an empirical matter. How can we get at that?

[fpetillo]

That is a great debate! But I'm not sure that it is relevant here. Regardless of what you think of the intervals, does a code of, say, "3" for the middle category inadvertently signal to the respondent a quantity of something that isn't there if the middle category really means "neither this nor that?"

You are assuming that the categories correspond to equal-size intervals. This is contested.

Unfortunately, the interpretation of the relative sizes of the intervals is an empirical matter. How can we get at that?

[CowboyBear]

You are assuming that the categories correspond to equal-size intervals. This is contested.

Unfortunately, the interpretation of the relative sizes of the intervals is an empirical matter. How can we get at that?

Some studies have actually asked participants to place particular responses options along a line indicating the attribute of interest. You can then see if the gaps between categories are all about the same. Example below for a scale that uses "faces" as the response options instead of the usual Likert format, but you get the idea:

Bieri, D., Reeve, R. A., Champion, G. D., Addicoat, L., & Ziegler, J. B. (1990). The faces pain scale for the self-assessment of the severity of pain experienced by children: Development, initial validation, and preliminary investigation for ratio scale properties. Pain, 41(2), 139–150. doi:10.1016/0304-3959(90)90018-9

Would be interesting to hear what people think of this. One complication is figuring out what exactly we'd want the results to look like before we'd say the scale has "interval" properties. All participants placing all responses exactly equidistantly apart? All responses equidistantly apart on average? No significant differences between average distances between options?

[CowboyBear]

Also, just in terms of the negative numbers - I think there's some kind of idea out there that the average populace finds negative numbers confusing and scary. Hence development of standard-score alternatives to z-scores like McCall's t (M = 50, SD = 10) and IQ (M = 100, SD = 15) which help to avoid negative-number standard scores.

[Dason]

Also, just in terms of the negative numbers - I think there's some kind of idea out there that the average populace finds negative numbers confusing and scary.

For good reason too. If you add a negative number to a positive number then the negative number kidnaps the positive number and neither are ever seen again - almost like neither ever existed. It's terribly tragic.

[CowboyBear]

For good reason too. If you add a negative number to a positive number then the negative number kidnaps the positive number and neither are ever seen again - almost like neither ever existed. It's terribly tragic.

Terribly so. Not to mention the vile sorcery that's rumoured to go on when two negative numbers are multiplied together... a negative times a negative equals a positive?!! Mathematicians, they so crazy.

[fpetillo]

These pale to what happens when you take the square root of a negative number. An imaginary number? Really? I thought they all were imaginary... not withstanding Plato.

Perhaps the way to resolve my unease about the scale numbers is to eliminate the scale. I could be a traditionalist about Likert and just use the categorical descriptors so far as my survey form is concerned. What I do in my private data lair stays in my private data lair.

Thanks to all.

Terribly so. Not to mention the vile sorcery that's rumoured to go on when two negative numbers are multiplied together... a negative times a negative equals a positive?!! Mathematicians, they so crazy.