Herewith Pat Doreian and I announce that the second special issue of the
Journal of Mathematical Sociology is available (Frans N. Stokman and
Patrick Doreian (eds), 2001, Evolution of Social Networks. Part II.
Special Issue of Journal of Mathematical Sociology, 25, Number 1, 138

Please, find below the short introduction where Pat and I introduce the


Frans N. Stokman and Patrick Doreian

The Journal of Mathematical Sociology (JMS) published a special issue on
Evolution of Social Networks in 1996 under the responsibility of the two
present guest editors. Gordon and Breach published in 1997 a volume that
contained the original articles of the special issue together with three
new articles (Doreian and Stokman, editors. Evolution of Social
Networks, New York: Gordon and Breach, 1997). In the first JMS special
issue and subsequent volume, a distinction was made between network
dynamics and network evolution. We were primarily interested in studies
that concentrated on the underlying mechanisms that induce network
change. In other words, we were interested in network evolution and not
only with network change. Partly as a consequence of this emphasis, most
contributions in the first volume focussed on theory, methods, and
simulation. In this respect, the first volume did not bridge the gap
between theory and empirical testing. This induced our desire to edit a
second volume on Evolution of Social Networks where modeling and
empirical analyses are integrated or at least combined. The present
special issue contains four of such contributions. A number of others
will follow next year in the third special issue on the topic of
Evolution of Social Networks.

The article of Robins and Pattison provides a powerful statistical
method for social network analysis by generalizing the p* model to
networks over time. Notwithstanding the complexity of the statistical
model, they successfully embed their article in a broader theoretical
context and provide arguments for why their method is of significance
for the study of network evolution. The article is therefore also
interesting for social network researchers who are less familiar with
statistical methods. One of these broader perspectives is related to the
question at which level of the network evolution takes place. They
emphasize that the individual can be seen as the only intentional unit
in the system, unless an outside authority can design the network. This
implies that global systematic evolution properties have to be generated
locally. Consequently, 'such a systematic property reflects shared norms
or behaviors across actors, norms or behaviors that could be construed
as inherent in the particular social relation for this group of people'.
More implicit in their reasoning is a second implication, namely the
importance of local structure for generating the global network. The
latter justifies the orientation of their method to local social
neighborhoods. Another interesting thought is related to their constant
tie assumption: the stronger the tendency in evolution, the less change
is to be expected over time. This makes empirical network evolution
studies not only a tedious effort (for the researcher as well as for the
subjects) but also a risky one: do we find sufficient change to explain
network evolution? Also for this reason, the two illustrations in the
article are well chosen: evolution of friendship ties in a
well-established group versus the evolution of friendship and trust in a
training group during a four days training course.

The fact that systematic network evolution is generated locally
justifies the shift in emphasis of Doreian and Krackhardt from global
balance in their 1996 contribution to triplets in the present
contribution where they again use the well-known Newcomb
(pseudo-dormitory) data.  The concept of balance as applied to 'group
level' networks was based on a globalization of the concept of cognitive
balance of Heider. However, Heider's concept is completely focused on
one subject under study and the perceptions of that subject. These
perceptions concern his or her relation to another subject, his or her
relation to a third subject or object, and his or her perception of the
relation of the other subject to the third subject or object. Under the
assumption that the relation of the second subject to the third is
perceived by the first subject as it is reported by the second, Doreian
and Krackhardt focus on a count of triplets. If all three relations
exists, the triplet can be categorized into eight categories, based on
the sequence of  'like' or 'L' and 'dislike' or 'D' relations, starting
from the subject under study i. All triples with an uneven number of D
relationships are unbalanced. For the evaluation of the frequencies,
they relate the frequency count to the expected value under a null model
where the observed numbers of like and dislike relationships are
distributed among the pairs under the uniform distribution.  For all
triplets starting with the L relation, the basic structural balance
hypothesis is supported. However, the frequencies of the triplets
starting with a dislike relationship deviate from the predictions based
on Heider's theory. We would expect more balanced DLD and DDL triplets
than expected, whereas they find substantially less. In contrast, the
unbalanced triplets DLL and DDD occur much more frequent than expected.
The authors give different alternative explanations for this finding. In
one of the alternative explanations it is stressed that different
network mechanisms operate simultaneously. In the Newcomb dormitory,
large differences in popularity and unpopularity emerged that could have
had strong effects on the frequencies of the triplet types, whereas the
null model does not take these differences into account. It is one of
the nice properties of the Robins and Pattison approach that effects can
be estimated under control of others.

The Lazer study analyzes selection and contagion in an American
bureaucratic organization, the Office of Information and Regulatory
Affairs. Lazer's study is a nice example of a study of the co-evolution
of individual characteristics and network structure. For the internal
network in the office, Lazer is confronted with the problem of network
stability that Robins and Pattison expect under strong systematic
network evolution. The network was largely institutionally determined
and showed less evolution than networks in an informal setting like that
of Newcomb's dormitory. Lazer introduces the concept network elasticity
for this phenomenon. He was fortunate that the external network was
characterized by a much larger elasticity than the internal one. In the
contagion part of his study, he is in a similar way confronted with the
stability of individual characteristics, denoted the amount of
individual plasticity. As the individuals who left the organization were
unaffected by the change in milieu, Lazer questions the influence of the
organizational network on the individual attitudes in the context of his
study. In our opinion, the empirical investigation of selection and
contagion effects is one of the most important topics in social network
evolution. Lazer's study shows that they can only be studied in contexts
where substantial network change is combined with substantial shifts in
individual characteristics. We need theories and evidence in which
domains elasticity and plasticity are to be expected. Lazer's study
contribution does that.  There is one parallel with the results of
Doreian and Krackhardt: the identity of individuals who emerge as
popular or unpopular is something that co-evolves with network

Wittek also confronted the problem of large network inelasticity in the
organization he analyzes. Nevertheless, the few relational changes have
serious impacts on the positions of certain actors in the network. His
contribution is particularly important for two reasons. First, he is
dissatisfied with network theories that stress the importance of
similarity of individual characteristics for network selection without a
theory from which the relevant characteristics can be deduced. He does
precisely that for trust relationships in a competitive environment.
Second, he shows that just a few changes in relationships can have large
effects on the position of individuals in the global network structure
and on the global network structure itself.  Wittek uses data from an
organization for three time points. The management of the organization,
in effect, created an informal experiment in which an individual present
at the first time point, was removed for the second time point and
returned for the third time point.  Wittek's paper suggests that more
stable settings are also interesting for network evolution studies.

We are confident that the four studies will inspire further social
network evolution studies. We strongly believe that further progress in
social network theories heavily depends on our creativity to studying
and explaining both social network evolution and change.

Frans N. Stokman
Professor of Social Science Research Methodology
University of Groningen
Grote Rozenstraat 31
9712 TG Groningen
The Netherlands
Phone work      +31.50.3636259
Fax work        +31.50.3636226
Private phone   +31.50.5350040
Email:          [log in to unmask]