CHI-WEB Archives

ACM SIGCHI WWW Human Factors (Open Discussion)


Options: Use Forum View

Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Nitesh Goyal <[log in to unmask]>
Reply To:
Nitesh Goyal <[log in to unmask]>
Fri, 23 Jan 2009 11:49:58 +0100
text/plain (111 lines)

This seems like an interesting situation. I agree with William that the
dimensions should be the same. However, in questions that can offer a large
number of possibilities as an answer, it might be difficult to suffice all
the possible alternatives. So, I would suggest thinking about the following.

1. How do we create questions, when we do not know the user terminology?

First, I would suggest,research for questions available online or other
sources in that domain. This, would help improve the knowledge of the kind
of appropriate words used for a particular domain. However, like in the
original posting, these might not reflect the best possible alternatives.
So, caution should be exercised.

2. Ask the user for the terminology?

Create a smaller user study and ask them to suggest words that reflects
their stand on the question. This might be expensive and more time
consuming. But, worth it, if the labels on the Likert Scale matter hugely.

3. Present the users with a blank, to fill in, along with the Likert Scale.

At the end of the Likert Scale, provide an empty blank for users to fill in
their choice of word, if not available on the scale.  This might break the
utility of the Likert Scale. But then, it would help in 2 ways: preventing
erroneous data from creeping in, and giving the freedom to the designer of
the scale to interpret the answer closest to the scale according to him.
Since, his mental model relates better to the scale.

Hopefully, this helps!

Thanks and Regards,
Nitesh Goyal
RWTH, Aachen

On Fri, Jan 23, 2009 at 11:11 AM, William Hudson <
[log in to unmask]> wrote:

> How to use Likert scales seem to be a pretty touchy subject (from my
> past experience), but I just had to share this example of how (I think)
> not to do it.
> This is a ten-point scale from a film rental site I use:
> Painful
> Poor
> Dull
> Disappointing
> Just OK
> Average
> Good
> Very good
> Great
> I love it
> One issue that struck me about this is that some of these words are just
> not on the same dimension of meaning as the others. For example, the
> film I wanted to rate I thought was just bad, but I found that the word
> corresponding to my strength of feeling on the 10-point scale was
> 'dull'. Dull it wasn't - it was full of action and killing, but still
> bad.
> My own approach is to label just the end points, so in this example,
> 'very bad' at one end and 'very good' at the other with numbered points
> in between. However, I have found that this does confuse respondents
> very occasionally ('What do the points in between correspond to?' is the
> occasional question).
> However, if you like labelling each intermediate point with words, I
> think they need to be in what I would call a single dimension, like
> those shown here:
> <
> and%20Validity/Likert.html<>
> >
> What do list members think? What do you use when conducting surveys?
> Regards,
> William Hudson
> Syntagm Ltd
> Design for Usability
> US Toll Free 1-866-SYNTAGM
> UK 01235-522859
> World +44-1235-522859
> mailto:[log in to unmask]
> Syntagm is a limited company registered in England and Wales. Registered
> number: 1895345. Registered office: 10 Oxford Road, Abingdon OX14 2DS.
>    --------------------------------------------------------------
>    Tip of the Day: Quote only what you need from earlier postings
>     CHI-WEB: POSTINGS: mailto:[log in to unmask]
>              MODERATORS: mailto:[log in to unmask]
>    --------------------------------------------------------------

    Tip of the Day: Quote only what you need from earlier postings
     CHI-WEB: POSTINGS: mailto:[log in to unmask]
              MODERATORS: mailto:[log in to unmask]