Social Networks 8 (1986) 205-211
Ronald S. BURT
At once charming and dunning, Lin Freeman has asked me to offer
some remarks on a paper by Faust and Romney (1985) published in a
recent issue of Social Neworh.
This is not a commission I relish. I
have a great respect for other work by the authors and in this paper
they are not at their best. Still, Lin insists on the value of public
discussion and there is some value in providing a cautionary note to
casual readers who might be misled by the Faust and Romney paper.
In order to cut away the fat surrounding the main proposal in Faust
and Romney’s article, let me separate two papers within
There is a three-page paper on pages 81 through the top of 84 in which
Faust and Romney propose replacing Euclidean distances with correla-
tions in order to detect structurally equivalent
individuals within a
network. This does not seem wise as a general practice for network
analysis but its value is an empirical question certainly open to debate
on a case by case basis. The remaining twenty-four pages of comment
to demonstrate that Burt and Bittner provide a nonsense
interpretation of a study population by analyzing Euclidean distances
instead of correlations. This second paper is an ill conceived adventure,
ultimately an analytical
failure, criticizing misrepresented evidence
from the original paper and eventually arriving at precisely the conclu-
sion reached by Burt and Bittner.
The three-page paper begins with a reminder (repeated throughout
the article) that Euclidean distance, the widely used measure for
detecting similar profiles of data, is an aggregate of at least three
components; differences between the mean values across variables in
each profile, differences between variances within each profile, and the
correlation across variables in each profile. Faust and Romney propose
’ Department of Sociology. New York, NY 10027, U.S.A.