Vignettes should be written so that people in different cultures
understand them as similarly as possible. Translation is of course
essential, as is cognitive debriefing during pretests.
We find that concrete vignettes that describe specific people
and situations are best able to provide constant anchors, although
this will not always apply.
Each set of vignettes corresponding to a single self-assessment
question should tap a single unidimensional concept. The process of
writing vignettes is like the process of testing a theory, in that
data (or the examples in the vignettes) tend to focus the mind. As
such, the process of writing vignettes tends to have important
effects on the concepts themselves. New dimensions are discovered,
and the features corresponding to them peeled off, making the
original set of vignettes more concrete. And of course sometimes
new vignettes and a new self-assessment question are added to
measure the new dimension.
Be careful of the details. Sex, age, and other variables can
enter the vignettes by something as simple as the name used or other
references. Ask whether these other variables are providing the
needed contextual detail for the respondent, in which case they
should be retained, or whether they are adding additional unintended
dimensions that could confuse the respondent or the analyst.
Ideally, only information that is an integral part of the
concept being measured should be part of the vignette description.
Everything else should be kept implicitly the same as the respondent
(so that DIF remains the same for the self-assessment and the
vignette questions).
