Applying Co‐occurrence Text Analysis with ALCESTE to Studies of Impression Management

AuthorMartin W. Bauer,Laura Illia,Karan Sonpar
Date01 April 2014
DOIhttp://doi.org/10.1111/j.1467-8551.2012.00842.x
Published date01 April 2014
Methodology Corner
Applying Co-occurrence Text Analysis
with ALCESTE to Studies of
Impression Management
Laura Illia, Karan Sonpar1and Martin W. Bauer2
IE School of Communication, IE University, Spain, 1School of Business, University College Dublin, Ireland,
and 2Department of Methodology and Institute of Social Psychology, London School of Economics, UK
Corresponding author email: Laura.Illia@ie.edu
This paper reviews the potential role of co-occurrence text analysis using ALCESTE, a
computerized text analysis program. Using an illustrative case study from the biometric
industry, we demonstrate that this method offers a number of advantageous features,
including the provision of visual outputs which are useful for interpreting results, the
ability to study longitudinally the effectiveness of impression management at the inter-
organizational level of analysis and the possibility of studying large textual data sets
without using predefined dictionaries. Meanwhile, key limitations of the method include
its limited versatility, its tedious data-cleaning process and its ineffectiveness in identi-
fying the centrality or tonality of the discourse. Our overall conclusion is that the
introduction and more widespread use of this method in management is timely, particu-
larly for scholars interested in studying narrative fidelity and frame amplification.
Introduction
Although discourse is central to shaping meanings
and managing impressions, the use of computer-
ized packages has been relatively limited in efforts
to study language and discover both the latent and
the hidden meanings in discourse (notable excep-
tions include Hodgkinson, Maule and Bown, 2004;
Jones and Tarandach, 2008; Kabanoff, Waldersee
and Cohen, 1995; Ocasio and Joseph, 2005; Short
and Palmer, 2008; Wolfe, Gephart and Johnson,
1993). The use of computerized packages enables
researchers to overcome several limitations that
are likely to occur in the manual coding of text.
These limitations include the time-consuming
nature of manual coding, the potential for rater
bias (this is more likely during the manual identi-
fication of codes because such techniques might
not be representative of or might over-emphasize
certain areas), concerns regarding the reliability
and accuracy of coding as a result of researcher
fatigue, practical difficulties in coding large data
sets, and the lack of development of clear protocols
and means of comparing and contrasting different
data sets (Lu and Shulman, 2008; Short and
Palmer, 2008; Wolfe, Gephart and Johnson, 1993).
In the light of these limitations of manual coding, a
number of studies have urged the adoption of
computerized textual analysis software.
Despite some impressive advances in the com-
puterized coding of text, a review of some of the
leading computerized applications for textual and
The authors are grateful for the cooperation and finan-
cial support provided for the field study by the Swiss
National Science Foundation. Comments by Candace
Jones, Magdalena Wojcieszak, Joep Cornelissen, Craig
Carroll, Stelios Zyglidopoulos, Johan van Rekom,
Nicole Kronberger, Aude Bicquelet and Kevin Corley to
earlier drafts were helpful for revising the paper. A final
thanks goes also to the reviewers from the Academy of
Management Conference, EGOS Colloquium and par-
ticipants at the research seminars at Brunel Business
School and London School of Economics and Political
Science whose comments were very constructive and
helpful in shaping this paper.
bs_bs_banner
British Journal of Management, Vol. 25, 352–372 (2014)
DOI: 10.1111/j.1467-8551.2012.00842.x
© 2012 The Author(s)
British Journal of Management © 2012 British Academy of Management. Published by John Wiley & Sons Ltd,
9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA, 02148, USA.
content analysis within the management literature
reveals a few weaknesses. First, several computer-
ized applications impose predetermined diction-
aries upon the data set rather than inductively
developing dictionaries through the data.
Although this feature may be advantageous as it
permits us to study vocabulary and word choice,
only dictionaries that are tailored to a text are
sufficiently flexible to identify specific language
that is not included a priori in the dictionary (e.g.
unique and industry-specific jargon, the name of
the company, the name of a product or project)
(Short and Palmer, 2008).
Second, several current studies in management
analyse the language of companies and their inter-
locutors separately and operate under the
assumption that words are being used with the
same semantic value at each occurrence. Thus,
the conjoint analysis of text from two different
sources has been underdeveloped. For example,
Fiss and Hirsch (2005) and Kennedy (2008)
compare company and media content by analys-
ing texts from these two sources separately. This
type of analysis has the advantage of highlighting
the main themes debated by different groups, but
it does not indicate whether words are used in
similar or different semantic contexts. Addition-
ally, although there has been a significant focus on
the use of keyword frequency counts to measure
the attention paid to a given issue (e.g. Tuggle
et al., 2010), less emphasis has been placed upon
the idea that communication occurs in a context
and that the context in which certain keywords
are used is important (Sonpar and Golden-Biddle,
2008). This limitation is significant, because it
means that one infers only the meanings of words
without considering the context in which words
are used. Admittedly, some studies have high-
lighted the need to study the context in which
discourse is made (Sonpar and Golden-Biddle,
2008), such as keyword-in-context (KWIC) in
content analysis (e.g. Weber, 1990) and expansion
analysis in the textual approach (e.g. Gephart,
1993), but protocols are often ignored or con-
ducted on a post hoc basis.
Finally, and at the most general level, despite the
advances in text analysis arising from the greater
use of technology over the last two decades, a
recent review of the impression management (IM)
literature by Bolino et al. (2008) highlights a few
methodological directions for future research
including a need for new ‘data-analytic methods’
from related disciplines to analyse text (p. 1098),
study of the effectiveness of discourse through
the adoption of multi-level research designs, and
development of frameworks that can ‘compare
and contrast efforts of companies who are
attempting to create the same image’ (p. 1098).
This paper introduces a computer-aided text
analysis program, namely the co-occurrence text
analysis methodology with ALCESTE (Reinert,
1987). We critically evaluate the advantages of
this methodology in addressing some of the above
limitations and highlight some of its constraints
compared with other leading textual analysis soft-
ware programs. To ensure clarity, we show how
co-occurrence methodology may lead to elabo-
rated theories of discourse and IM (e.g. Bolino
et al., 2008; Elsbach, 1994), and use an example
from the biometrics industry to show how an
analysis of multiple organizations and across mul-
tiple levels is feasible. The IM literature primarily
focuses on studying the reactive responses of cor-
porations and their value in addressing episodic
attacks on legitimacy (e.g. Elsbach, 1994). Little is
known, however, about the various ways in which
legitimacy can be pre-emptively managed by
organizations through communications that
prevent the escalation of long-term criticism, even
though the literature has identified a role for both
defensive and proactive IM (see Bolino et al.,
2008). In order to make this kind of theoretical
advancement, a methodology that compares cor-
porate language over time is necessary.
The paper is structured as follows. First, we
provide an overview of the various types of
computerized text analysis software. Second, we
introduce ALCESTE and the co-occurrence
methodology, highlighting both its advantages
and its limitations. Third, we introduce a theoreti-
cal problem and an empirical setting as a pri-
mer for how the co-occurrence analysis using
ALCESTE might be used. Fourth, we provide an
empirical demonstration. Finally, we conclude.
Overview of various types of
computerized text analysis
At the most elementary level, the three types of
computerized text analysis are representational
text analysis, positioning text analysis and inferen-
tial text analysis (see Table 1, adapted from
Co-occurrence Text Analysis with ALCESTE 353
© 2012 The Author(s)
British Journal of Management © 2012 British Academy of Management.

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT