Quality & Quantity 33: 361–380, 1999.
© 1999 Kluwer Academic Publishers. Printed in the Netherlands.
Structured, Qualitative Comparison
Between Singularity and Single-Dimensionality
KARL HENRIK SIVESIND
Department of Sociology and Human Geography, University of Oslo, PO Box 1096 Blindern,
N-0317 Oslo, Norway
Abstract. This article deﬁnes qualitative data as representations of human acts and utterances,
conventionally analysed in the form of long, coherent texts. The purpose of qualitative research is to
interpret the actors’ understandings and intentions.
On the background of this deﬁnition, I make a distinction between variable-coding and
theme-coding and between code-oriented and content-oriented analysis. Variable-coding of the
text-content should be avoided, since this implies reducing multi-dimensional qualitative data to
single-dimensional data. However, variable-coding of background-information can be used to-
gether with theme-coding of the text-content without corrupting the qualitative data. I call this
approach ‘structured, qualitative comparison’. In that way it is possible to retain the many levels
of meaning of qualitative data throughout the research-process, while focusing on the importance of
macro-variables or scope-conditions, which indicate the potential area of validity of the ﬁndings.
By examples from a comparison of Norwegian and German factories, I demonstrate the use-
fulness of such a content-oriented analysis-style for qualitative research, as well as for combining
qualitative and quantitative data. In this way, the basic characteristics of qualitative data are preserved
throughout the research-process.
Key words: qualitative research, sorting qualitative data, combining qualitative and quantitative
Qualitative data can be deﬁned as representations of human acts and utterances,
such as interview-transcripts, observation-notes, documents and other forms of
texts. The aim of the analysis is to interpret the understandings of the actors. In the
methods literature, the approaches to sorting and analysing this kind of data often
encounter one of the following problems, depending on the level of abstraction.
Some researchers analyse qualitative data by looking for patterns of co-
occurring codes or the lack of co-occurrence. This presupposes that codes are
attached to pieces of the qualitative data-material according to variation in cer-
tain clearly deﬁned features. I call this variable-coding, since such codes have
the structure of a variable. This approach has the negative consequence that the
ambiguity, subtleness and context-dependence of natural language is broken down
to single-dimensional codes. The richness of the material is replaced by informa-