CHI-WEB Archives

ACM SIGCHI WWW Human Factors (Open Discussion)


Options: Use Forum View

Use Monospaced Font
Show Text Part by Default
Condense Mail Headers

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

Print Reply
"ACM SIGCHI WWW Human Factors (Open Discussion)" <[log in to unmask]>
Felix Ding <[log in to unmask]>
Sat, 10 Jun 2006 16:21:55 +0200
Gael Laurans <[log in to unmask]>
Gael Laurans <[log in to unmask]>
text/plain; charset=ISO-8859-1; format=flowed
text/plain (77 lines)
On 6/10/06, Felix Ding <[log in to unmask]> wrote:
> Hi all:
> I have several questions about "Semantic Differential" technique
> developed by Osgood et al:
> 1, I don't understand how the adjective pairs are worked out. After
> reading some documents and asking for some advices from other people,
> I found that the method they use is quite subjective, normally by
> using "brain storm", or gathering those pairs from documents/
> magazines/books then filtering them all by themselves. Well, thus the
> method is so subjective that the research conclusions are totally
> affected, isn't it?

I don't remember exactly how Osgood selected his items, but I don't think it
matters so much. If you want to find out, you should look up his book "The
measurement of meaning".

One of the main result of semantic differential studies is that factor
analysis of this kind of ratings consistently reveals three main dimensions
: evaluation (good-bad), potency (strong-weak) and activity
(active-passive). This structure was already suggested by Wilhelm Wundt in
the 19th century and has been found to fit many different data (word
connotation like Osgood, similarity ratings between adjectives, emotion
expressed by pictures, self-report of current mood, etc.)

If what you have in mind is how does one build a (semantic differential)
scale to measure some construct like user satisfaction, then the general
process is generally the following :
1. Generate items (pairs of adjectives for the semantic differential,
statements for a Likert scale) by brainstorming, asking experts, reviewing
the literature, etc.
2. Find some stimulus to test the scale, for example two websites, one that
is expected to generate high satisfaction (based on informal comments from
experts and users) and another associated with low satisfaction.
3. Select the best items, i.e. those that discriminate best between your two
websites. If your construct includes several subdimensions, you probably
want to select items that load high on one of these dimensions.

2, How a user selects his preferred scale in one adjective pair? For
> instance, in this adjective pair(Pearson, 1983):
> Accuracy: The correctness of the output information.
>                         accurate   3 2 1 0 -1 -2 -3   inaccurate
>                                 high   3 2 1 0 -1 -2 -3   low
>                       consistent   3 2 1 0 -1 -2 -3   inconsistent
>                         sufficient   3 2 1 0 -1 -2 -3   insufficient
> If I'm a user, of course I would prefer the adjectives on the left by
> selecting the positive scales, who wants to get INACCURATE/
> INSUFFICIENT information!

Not sure I understand this question. To my knowledge, semantic differential
scales are used mainly to express a rating of some stimulus (a picture, a
product) or of the person's current state. You would not typically present
such a scale to someone and ask "what would you want to see in this

It is however well known that people react differently to - even slightly -
different questions, including of course negatively vs. positively phrased
statements. That's why building a scale involves more than just putting
together some questions.

GaŽl Laurans

PS: if you want specific comments on a particular tool, it could be helpful
to provide full references.

    Tip of the Day: Suspend your subscription if using auto replies
     CHI-WEB: POSTINGS: mailto:[log in to unmask]
              MODERATORS: mailto:[log in to unmask]