Burt Positions in Networks

Positions in Networks Author(s): Ronald S. Burt Source: Social Forces, Vol. 55, No. 1 (Sep., 1976), pp.
93-122 Published by: Oxford University Press Stable URL: http://www.jstor.org/stable/2577097 . Accessed: 14/11/2013 02:33
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to Social Forces.
http://www.jstor.org
This content downloaded from 128.12.255.4 on Thu, 14 Nov 2013 02:33:17 AM All use subject to JSTOR Terms and Conditions
Positions in Networks*
R O N A L D S. B U R T,
University of California, Berkeley
ABSTRACT The existence of an actor as a set of asymmetric relations to and from every actor in a network of relations is specified as the position of the actor in the network. Conditions of strong versus weak structuralequivalence of actor positions in a network are defined. Network structureis characterizedin terms of structurallynonequivalent,jointly occupied, network positions located in the observed network. The social distances of actors from networkpositions are specified as unobservedvariablesin structural equationmodels in order to extend the analysis of networksinto the etiology and consequences of networkstructure. We are each nested in a cacophony of relations with other actors in society. These relations serve to define our existence in society. We are who we are as a function of our relations to and from other actors in society. With the growth of technology and its concomitant division of labor, the determination of actors in society as a function of their relations with other actors is likely to increase rather than decrease. The problem for the social scientist then becomes one of conceptualizing the patterns of relations between an actor and the social system in which he exists in a manner optimally suited to explanation. Within the total set of all relations which link an actor to other actors in a social system, there are subsets of similar relations. There are economic relations linking the actor to specific other actors. There are relations of friendship, relations of kinship, and relations of status. There are political relations linking the actor to other actors. The list has no end. Each of these types of relations among actors in a social system serves to define a network of relations among the actors. This paper elaborates a conceptualization of networks of relations among actors in a system which simultaneously captures the basic characteristics of the structure in an observed network of relations and easily lends itself to the investigation of the etiology and consequences of that structure through the use of structural equation models. The central idea in the conceptualization is that of a position in a network, the specified set of relations to and from each actor in a system. The extent to which two actors jointly occupy the same network position is treated as the social distance between them and is specified as a scalar value. Two or more actors jointly occupy the same network position when they have similar relations to and from each actor in the network. Actors jointly occupying the same network position are discussed as
*Workon this paper was supportedby a grantfrom the National Science Foundation(GS-405-OOX),a grantfrom the Director's Fund at NORC and a fellowship from the National Instituteof Mental Health (3-5690-43-3453). Portionsof this paperwere presentedat the EighthWorldCongressof the International Sociological Association in the community research session. I appreciatecomments by J. S. Coleman, T. N. Clark, J. A. Davis, and E. 0. Laumannon previousdiscussions of the ideas presentedhere and the generosity of E. 0. Laumann and F. U. Pappi in making their data on elite actors in a German communityavailable to illustratethe ideas in the paper. 93
94 I Social Forces I vol. 55:1, september1976 being structurallyequivalent in Section 2 and definitions of strong versus weak equivalence are specified then illustrated.Based on the discussion in Sections 1 and 2, Section 3 proposes that the structurein an observed networkexists as a patternof relationsamong M + 1 sets of actors who togetherform the system of actorsbeing considered. The M sets of actors are each composed of multiple actors who jointly occupy structurallyequivalent positions in the network. Actors in different sets nonequivalentpositions. The M jointly occupied positions exist occupy structurally as four elementarytypes of networkpositions outlined in Section 3. The remaining set is a residual category of actors among whom there is no network position occupied by more than two actors. Sections 2 and 3 describe, primarily, the structurein an observed networkof relations. Section 4 moves ahead by specifying equation models as unobservedvariablesin orderto networkpositions in structural investigate the etiology and consequences of occupying differenttypes of positions in differenttypes of networks. IN NETWORKS AND ACTORS AS POSITIONS SOCIALDISTANCE Given a system composed of N actorsand one or more networksof relationsamong the actors, there are two perspectives from which the intensity of relationbetween two actors within a single network can be viewed: (1) from the perspective of the two actors as a dyad which is only secondarilyassociated with the overall network, or (2) from the perspectiveof the two actorsas elements of the overall network.The former can be discussed in terms of an asymmetric individual distance from one actorto another.The lattercan be discussed in terms of a symmetricsocial distance between the two actors in terms of theirrespective positions in the network. Let Idij be the individual distance from actor i to actor] where Idij can be operationalizedas any of a variety of measuresof the directed, asymmetricrelation from i towardj qua individual entities such as: (1) a measure of the difference of i (e.g., i's values, socioeconomic characterisbetween a profile of characteristics tics, or aspirations),and i's perceptionof actorj's profile on the same characteristics;' (2) a measureof the co-occurrenceof actorsi and] in archivalrecordswherej initiates action while i is the object of action;2 (3) the presence or absence of a sociometric choice link from actor i to actor j;3 (4) the minimum number of sociometric choice links required for actor i to reach actorj;4 (5) the minimum weighted sociometric choice link distance required for actor i to reach actorj;5 (6) the normalizedminimumnumberof sociometricchoice links requiredfor actori of individualdistance do not to reach actorj. 6 Although the six operationalizations measurethe same qualitativeidea, they all reflect an intensity of relationshipfrom one actor toward anotheras a dyad, a pair, of actors. The otherN - 2 actors in the which consideronly networkare ignored except in the last three operationalizations as many other actors as are necessary to complete a chain of sociometric choice links from actori to actorj.7 When two actors exist in a network with several additional actors, an
Positions in Networks I 95 integral aspect of the intensity of relationship between them consists of their differentrelationshipswith each of the other actors in the network(see researchon referencegroup influenceon individualperception,e.g., Tajfel;and researchon the motivation of actors to avoid dissonant situations by maintainingbalanced social relations, e.g., references in note 10). For example, if one actor is a member of a clique which excludes the other actor, then they will have a greatersocial distance between them than they would if both actors were members of the same cliqueeven if the individual distances between them were the same under both circumstances. Further,the distance between the two actors would be larger still if each actor were a memberof a clique which excluded the other actor. For each of the N(N - 1) asymmetricrelations between pairs of actors in the system, there is a value of Idij (assume Idii equals 0). An (N by N) matrix of individualdistances among the actors can thereforebe assembled. Let IDnnrefer to this matrix. Consider the ith row and column vectors of this (N by N) matrix. The row vector containsN elements which describe actor i's relationtowardevery actor in the network. The column vector conatinsN elements which describethe relations of every actor in the network toward actor i. Togetherthe ith row vector and ith column vector in IDnn completely describe the relationships of actor i with the actors in the network.If the two vectors are combinedinto a single (2N by 1) vector (N elementsfrom IDin and N elementsfrom IDni), the new vector, IDj*, can be discussed as the structurallocation of actor i in the network. IDj* is the "position" of actor i within the network being analyzed. IDj* defines actor i in terms of his relations to all other actors in the system as they are present in the networkbeing analyzed. Conceptualizing two actors solely as elements in a network, the social distance between them can be given as the distance between their respective geometrically as the networkpositions. Such a distance has a simple interpretation Euclidean distance between the actors i andj where actor i is defined by IDj* and actorj is defined by IDj*. The social distance between actors i and j can therefore be given as the square root of the sum of squareddifferencesbetweencorresponding elements of IDj* and IDj*:8
dij= dji = A
=
k=1(Idik
Id3k)2+
Nk=,1(Idki- Idkj)2 - Id2

-
Idji2],
(la) (lb)
V[(IDi*-
IDj*)' (IDi*- IDj*) -Idi2
Idji2].
In contrastto the operationalizations of individual distance, social distance is a symmetricmeasure (i.e., dij = dji).9 Further,althoughalternativeoperationalizations of individualdistancecan vary acrossresearchapplications,the specification of equation (1) is invariantsince social distance only has nominal meaning.'0 It is defined in terms of logic without reference to empirical informationand therefore generalizes to any empirical situation. Social distance derives empirical meaning from whicheveroperationalization of individualdistance is selected. (See Hempel's a, 654-66; b, 101-2, 197-213 discussion of empiricalversus nominalmeaningand Braithwaite's,50-2, 79-84 related discussion of direct versus indirect meaning).
96 I Social Forces I vol. 55:1, september1976 IN A NETWORK POSITIONS STRUCTURALLY EQUIVALENT Actorj's position in a networkis definedbyIDj*, the (2N by 1) vector of individual distances from actorj to each of the N actors and from each of the N actors to actor j. Given that actors have different relations with one anotheras a function of their interestsin each others' resources such as finances, prestige, and charm, it is to be expected that positions of actors in a network will be differentiallysimilar to one another.Two actorsoccupy the same position in a networkwhen they have the same relations to and from each actor in the network. Such a pair of actors can be discussed as occupying structurallyequivalent positions in the network. More generally, a set of actors can be discussed as occupying structurally equivalent networkpositions when their relations with all actors in the networkare identical so that:
IDi* = IDj* = IDk*, (2a)
which means that actors i, j and k occupy structurallyequivalent positions in a networkwhen the social distances among them equal zero:
dij = dik = d3k = 0.
(2b)
Since it requiresthat every elementof IDj* and IDj* be identical if actors i and j are to be discussed as structurally equivalent, equation (2) is a definition of strong structuralequivalence." When dealing with actual networks of relations, the strong definition of equivalence has little utility since there are likely to be minor differences between structurallyequivalent positions due to sampling variability,errorsof observation, and/or theoretically trivial differences between actors. For applied research therefore, it is convenient to relax equation (2) to a weak definition of structural equivalence. A weak definitionof structuralequivalenceof actor positions asks that dij, the social distance between actors i and j, be less than some criterion distance based on the distances among structurally nonequivalentpositions in order for actors i and j to be treated as occupying structurallyequivalentpositions. In other words, actors i andj occupy structurallyequivalentpositions in a networkunder a definitionof weak equivalence when: IDj* - IDj*, which can be stated as dij < a,
(3b)
(3a)
where alpha is a criterion distance based on the distances among the structurally nonequivalentpositions in a network. The social distances among actors in struc-
Positions in Networks I 97 turally equivalent positions under the definition of weak equivalence is less than some arbitrarycriterion, alpha 12 While the specification of a value of alpha for all networks is arbitrary, a useful approachto determiningalpha for particularnetworks is through a hierarchical clusteringalgorithm.A hierarchicalcluster analysis will presenta succession of alternative,increasing, values of alpha. The firstvalue of alphawill be zero. The actors who occupy structurallyequivalent positions when alpha equals zero are equivalent under the definition of strong equivalence given in equation (2). Subsequent values of alpha are then obtained by combining individual positions of actors into subsets of positions which are more similarto each other thanthey are to other actor positions. This process continues by gradually allowing greater and greater social distances between actors clustered together as being structurally equivalent. Under the final value of alphafor a network, all actors will be clustered together into a single network position-even though they are separatedby considerablesocial distances. Somewherebetween the extreme of a definitionof strong equivalence and the extreme of defining every position as equivalent, will be a reasonablepartitioningof the actor positions into M + 1 sets of positions;M structurally nonequivalentnetwork positions each jointly occupied by multiple, structurallyequivalent, actorsand one residualgroupof actorpositions each occupied by no more thantwo actors.Each of the M structurallynonequivalent networkpositions jointly occupied by multipleactors will appear in a hierarchical cluster analysis as a set of actor positions which are clustered together as equivalent when alpha is close to the definition of strong equivalence-i.e., demanding minimal difference between ID* vectors clustered together as equivalent-and which are not clustered with other positions of actors until alpha is extremelyweak-i.e., allowing almost any difference between ID* vectors. As an illustrationof the ideas of strong and weak structuralequivalence, Figures 1, 2 and 3 present analyses of the networks of economic exchange, social exchange and information-seeking among 45 community decision-makers (cf., Laumannand Pappi, a: Figures 3, 4 and 5 respectively). An (N by N) matrix of social distances as estimated from equation (1) using the operationalizationof individualdistance outlined in Appendix A was input into Johnson's connectedness method of hierarchicalcluster analysis which is presentedin part A of the figures and input into Roskam and Lingoes' smallest space analysis in two dimensions which is presentedin partB of the figures. The location of networkpositionsjointly occupied by multiple actors is based solely upon the hierarchicalcluster analysis. The smallest space analysis is presentedin order to give the reader a feel for the spatial distributionof actors in two dimensions and for comparisonpurposes with the original Laumannand Pappi discussion. Part B references actor positions by their sequentialnumberas is done in partA of the figuresas well as by the influence rankof the actor as presentedin the LaumannandPappi figures. Appendix B details the expected differences which will occur in a nonmetricanalysis of social distances as opposed to individualdistancesusing Figures 1, 2 and 3 versusFigures 3, 4 and 5 in Laumannand Pappi (a) as an example.
98 / Social Forces / vol. 55:1, september1976
0
P4 wow@WaB0 MCO 20 p44,
4)Ho
*MMMM 2 MREH:i g00001
WllU tol ,0~~~~~c 4) ~~~~~co
cN0.002 I~~~~~~~~~~.
.00
iI
\j H
CL
z
<
<
:s
@
m
#
V
?
t~
. . . . . . .4 .4 .R54g
. . . .
01 H
1
r r r
cvu\o I~
C\N
* ' * '
'
4)w
m ,
::s
w
Lu
(I)
04)~~~~~C
p4
4 4
4)
z 0
Cl) _
4o c
d WO (D4 U ) 4)0-
0.
0. crx
Q0
H C\i C\ m\j
A!"
. . .
ffff
. . .
<u
C1
Hco -H
+)J
o
_
z Co
43 co .r4U
~ ~ ~~4
r4Ico0
.OH'
o
U.
*
0 ).
n H co
.
Positions in Networks /99
~~~ -H
)CH
-~~~~~~~~~~\jN
A(-0CJ4
H 4-
~
0
~ ~ ~
~
('
~
\
~
..
~
-
~
4'
' 4' 0 0 0
0
A~~~~~~~~~~~~~~~~~4
U~ ~~ H 0
OO .r4~~~~ 4-H H I-I+0 O C)0 -p k
cc
40
-\
H L
...
0-
H 4 CH~~~~~~~~~~~~ 4 CY)
~4'C
~ ~ ~~ ~ ~ ~ ~ ~ ~ ~ ~ ~
>i
4)
4 *) H 4-)
4-40 H
co
to co
CY)
0C
~ ~~~~~~~_ ~02 C)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~4

)H HO *H-HPi04-H
to2
0
-
4 co 0
02 )4
4) 0
F0
4)04)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
0C) 40 qH
4) -pLLc
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~I~ ~ ~ ~ ~ ~ ~ ~~~~~~~~)
....
~~~0C
-H
'
~ ~~
to4
02
0
0
COIi
,.cs%-p
u
>
*HdH wC)co 4-)C)4C

CG4-3
O4
4)0C
0 4)
4 0 00~~~~~~~~~
C
od
4)
C.)
fr
1~~~0)~I
4-

0
0 4'
X~~~~~~~~~~~~~~~~~~~~~
.4~~~~~~~~CH
nHH 4-I
... p4 ~ ~~~~tCMJH *rl~~~~~~~~~~~~~~~~~~~~~~~~~r
U
0)
.4.4
~!qN;M muH
.
. .
.
.
Cc
,0
MC C
LU
: 0) C) . .C2 ('CM C'1H CM CO CO
.::.:.:.:.::.
. . .
. .
. .
. .
. .
. .
. .
0~~~~~
<
x
1 04' to :0) 0)C

H z0)
. :..
:::: .. .
UNI
r-I~~~~~~~~
a)
('J
ON
Z >< w .O H
0 VI
0 o h< > O
...
..
W O
0) 0M c
. .
rCO . .
.
.
.
.
.
.
. X.
... m .
.
HN
CO
P) 5
UC,M0
VVVVVVVVVVVVVVVVVVV.
-J
HO
U)Jt '0 0)
C
ciL
o:*
k. .
U 344
0H. 0 (f... AUUp4U2
0~
vvvv vvvv
Positions in Networks
101
Cd
Ul\
IV ri 'ri 0 -p -p 0 VI k 0 CH 0 4 Co r-I F4 0 0
0
C; co cl. r-4 Cd -H 0 Cc 0
0 10
CM CY) I.-I C\j (Y') U'\ C\j
C\j
cn
(X)
cu
m
p
ul
t)
4-3 Ea -ri
co
cn
CY') -p CH 0 0 p 0
co CY') \D
4J 0 4i 4-3 CY') co
44 4i rA
C) -p
0 .H Ea o A 0
pq
o rA 4) $4 ci 4)
0 >,
to A
4-') 0 4-p CH 0 co 0
0 10
CF\ C\j
04i rA -H 0r-4 a) 0 02 -p U 0
Cd CH
... .....
co (Y') C -p ri
02 EO 0 Cd 0 r-4 ri
Cd 0 44
RLrl\
0 -ri o EO 4-3 (d 4-4 0 to to .rq 0 "I -ri ri +-,
Lr
ON
OC)
0 ca 0 0
co C\j
F4 rq 0 -p -p CO Q) co Cd co r-4 $4 $4 .0
0 0
C/I'o
4)0
4-
I ITI, Cd r-i W A CH 0 m 0 P4
Cc
-p
.15
,zw
r-i -H
V) CH 0 9)
CY')
~~~~m
hr)
t.
.x. .
C\u
Cr.
t .. .
.
. 0
to 0 P4
w~~~~~~~~~\
CM)
IH
~~~~~~~~~LN %
N 0
...
00.
0 Z~1mI ~~~~~~z
L 0 4'J
0D
. . . . . . . || .SU
04,~~
I ~z
z
| CoaS m
( 0 0
|
0
r0
N2 0EC
> . . . .3 .300g ('Sol~co . . . . .

. .
.
N'co
MgSSS||0X
.
.
.
.
.
.
.
.
0
.
| g | | | | 0 0 | | gl
Z O
C
N c ^|~~m . 1
0 0
< U2
4,
C'J (YLr (r)C J
Xo
02~
00 f4.qm
Z 0
n~
Z 0
.4 ~~~~~4
n Y
.......
..
0~
02 4 'H4
iL
m %vO .j
.H0
.I
H-
HtH
Cl)
4, 440
<
0)
H .C 4,
002k
~40.r4.r4 +
o
a.
4C'3~ 0
0~~~~~~~ 4
tr4)
Positions in Networks /103
m~~~~~~~~~~~~~~~~~~~~
00 0
m~~~~~~~~~~~~~~~~~~~
H
p44
OH
0~~~~~~~~~~~~~
0~~~~~~~~~~~~~~4
CH
0~~~~~~~~~~~~~~~~~~~~
a,.
44: 0
*r4'
d.I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~C
04'4 % * ~~...*4.
. .i. .
.04P2~~~~~~~~~~~~C V~~
~~~-4 0~~~~~~-4 oUN

0
*4
~~~~~~~~~~.4
co
0~~~~~~~~~~~~~~~~~
104 I Social Forces I vol. 55:1, september1976 Each row in the partA of Figures 1, 2 and 3 correspondsto a differentvalue of alpha given on the left-hand side of the figure. Positions of actors in a network which are clustered together as being equivalent at a given level of alpha are so indicated by Xs connecting them (e.g., the positions of actors 35 and 42 in the informationseeking network are to be considered equivalent when alpha is greater than 1.0). The firstvalue of alpha is zero. No positions in any of the threenetworksare equivalentunderthe definitionof strongequivalence. In each of the figures, the last and highest value of alpha allows all networkpositions to be treatedas equivalent. Following the above discussion, the M structurallynonequivalentpositions jointly occupied by multiple actors will appearin part A of the figures as mounds of Xs separatedby sharpdips or valleys between mounds. Such a condition will evidence a clustering together of actors when alpha is close to the definition of strong equivalence and the lack of additional actors positions entering the cluster until alpha is so large as to allow considerable social distances among actors clustered together as weakly equivalent. For example in the network of economic exchange relations (Figure 1) actors 2 and 39 are clustered together when alpha is 1.0 and other actors are clustered with them until alpha equals 1.5. After that, no actor is specified as structurally equivalentto the set composed of actors32, 6, 2, 39 and 44 until alphaequals 2.5-almost the largest value given to alphafor the network. The mound of Xs in the columns associated with these actors indicates that these five actorsjointly occupy a position in the networkwhich is structurally uniquefrom the positions of other actors and so has been identified as position E1 in the analysis. Actors jointly occupying the M structurallyunique positions in each network are indicatedbeneaththeir columns in the hierarchicalcluster analysis and are circled in the smallest space analysis (actors in residual category are unlabelled). The five structurallyunique positions in the network of economic exchange relations (i.e., M = 5) are referencedas E1, E2, E3, E4 and E5 (e.g., the actorsjointly occupying position E2 are 9, 12, 18, 20 and 41). The four structurally uniquepositions located in the network of social exchange relations are referenced as SI, S2, S3 and S4 (i.e., M = 4). The three structurallyunique positions located in the network of informationseeking relations are referencedas 1I, I2 and I3 (i.e., M = 3). THESTRUCTURE CHARACTERIZING NETWORK OF RELATIONS IN AN OBSERVED The observed position of an actor in a network exists as the patternof individual distances to and from actors in a system, i.e., as a particularform of interaction, given a single reason for interaction, i.e., given a particularcontent of interaction (see Simmel, 40-57, for an elaborationof the content-formdistinction).Empirically, structurein a network exists as N observed positions in the network; ID1, ID2, ID.. At a higher level of abstraction,idiographic positions (positions whose
Positions in Networks I 105

Table 1. STRUCTUREIN THE OBSERVED NETWORKSPORTRAYEDAS THE PATTERNOF RELATIONS SETS OF ACTORS* NONEQUIVALENT AMONG M+1 STRUCTURALLY
Economic
Exchange E 1
Network E 0 2 E 3 0
(Figure E 4
1) E 5 0 Residual 0 100% Total (15 (25 (21 (22 (18 (101 citations) citations) citations) citations) citations) citations)
87
13
E
E E E
0
12 0 5 22 21 Exchange S
0
4 0 0 0 1 Network S
0
56 0 5 17 18 (Figure S
0
8 90 55 39 42 2) S
0
0 0 13 0 3
0
20 10 22 22 16
100%
100% 100% 100% 100% 100%
(O0citations)
Residual Total Social
2
5
3
0 0
4
0 0 0
Residual 14 11 17 34 37 16 3) Residual 0 0 1 0 1 of citations of made by 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%
Total (21 (35 (18 citations) citations) citations)
S S S S
81 11 11 11 37 37 Seeking I
78 0 22 23 32
72 0 3 12 Network
33 0 3 (Figure
( 9 citations) (35 (118 citations) citations)
Residual Total Information
2
3
3
0 0
Total (31 (76 citations) citations)
I I I
97 33 72 0 76 *Cell entries
67 3 0 8 are
( 9 citations)
24 0 15 proportion Total
Residual Total
( 0 citations) (116 in citations) row in to ac-
actors
tors is at
in
column
position.
number
citations
made
by actors
row
right.
particularmixtureof form and content of interactionare of special significance for the system of actors being considered) and nomotheticpositions (positions defined solely as forms of interaction) present alternativeperspectives on structurein a network. As a first cut, those positions which are jointly occupied by multiple, structurallyequivalent actors are of special significance within a network. The definition of weak structuralequivalence provides a means of aggregating the N observed positions of actors in a system into M + 1 structurallynonequivalent subsets of actors. Idiographically,basic characteristicsof empirical structurein a networkare preservedas M specific patternsof relationsto and from each of the N actors (idiographicpositions) and a conglomerationof various patternsof relations associated with the actors comprisingthe residual category of actors. The decision
106 I Social Forces I vol. 55:1, september1976 to treatpositions in the residualcategory as idiographicwill be based on conceptual grounds since it has no clear resolution on empirical grounds. The structurein an observed network of relations can then be characterizedas the patternof relations among the M + 1 sets of actors. Such representationgof the structure in the networksanalyzed in Figures 1, 2, and 3 arepresentedin Table 1. Table 1 shows the distributionof sociometric choices made by actors in each subset over the M + 1 structurally nonequivalentsubsets of actors. For example, the networkof economic exchange relations is presentedin Table 1 as a patternof relationsamong six sets of actors-five sets of actors where each set jointly occupies a unique idiographic position in the networkand one residualset of actors. The actorsin position E1 give 87 percentof their sociometric choices to other actorsin the same networkposition. The remaining13 percentof their sociometricchoices are given to actorsoccupying of the structurein a networkis similar to the idea position E4. This characterization of a blockmodel of a network (see White et al.) with the exception that actors are not forced into one or another structurallyunique block. Each actor occupies his own networkposition. That position has finite social distances from each of the M structurally unique idiographicpositions in the network. Just as each of the M idiographicpositions in a network is reflected in the empiricalposition of an actor as a function of the actor's social distances to actors jointly occupying each idiographicposition, so the actor's empiricalpositionreflects pure forms of interaction, i.e., nomothetic positions. Viewing the M structurally nonequivalent, jointly occupied, idiographic positions as basic elements of the structurein a network, one can, as an alternativeto the characterization in Table 1, discuss the idiographicpositions in terms of the pure forms of interactionto which in a of empiricalstructure each is most similar.Nomothetically,basic characteristics networkcan then be preservedas M types of pure forms of interaction. Table 2 presents an exploratorytypology of four forms of interaction.Each cell of the typology representsa different nomothetic position in a network. The positions are classified accordingto the prominenceof actorswithin a network(zero versus non-zero proportionsof sociometric choices made by actors in the network which are given to actorsjointly occupying a position) and the tendency for actors to only initiateinteractionwith actorsto whom they are structurally equivalent(half or over versus under half of the choices made by actors jointly occupying an idiographic position are given to other actors occupying the same position). The columns distinguish between actors in a position who receive few or no choices from the system versus those that receive a nonnegligible proportion of those choices. The rows in Table2 distinguishbetween actorsin a position who give most of their citations to other actors in the position versus those who give most of their citations to actors occuping differentpositions from theirown. The four nomothetic positions in Table 2 are not intended to be exhaustive of the range of possible empiricalpositions, however, they do representbasic types of networkpositions as pure forms of interaction. The isolate position exists as a set of actors who give most of their citations
Positions in Networks I 107

TYPES OF NETWORKPOSITIONSAND THE PATTERNSOF RELATIONS Table 2. FOUR ELEMENTARY WHICHDEFINETHEM
Proportion choices
of given
network to the
position* Ratio choices the total choices the of position given number made positiont to over of by C SYCOPHANT D BROKER >.
ISOLATES
PRIMARY
POSITION
position
.5
ILLUSTRATIVE A 0? 0 0 B 0 high high
PATTERNS OF RELATIONSt C 0 0 low D 0 0 0 RESIDUAL 0 0 0?
ISOLATE PRIMARY SYCOPHANT
A B C
BROKER
D
RESIDUAL TOTAL
0
0 0 all
high
high high choices the made by actors the position. the
0
low low
low
high high in the being occupying over of to cells. been the
o?
0? 0? network the total which network number choices the aid of from column in the are given to be-
*Proportion the ing made actors tNumber classified by actors tCell cupying as tion cation is ?No of of the jointly of to entries
of
by actors position jointly position
occupying choices actors refer in Table are structure, of positions made the to 1. made which within
network
classified. position choices actors position descripclassifioc-
occupying row position
proportion are for given these not a network.
sociometric occupying Although used in this
actors
presented predictions network types
they
they in
have
exploratory
to actors within the set and receive no citations from actors in the system who are not in the set, i.e., who occupy nonequivalentpositions in the network. Such a set of actors could be a subsystem within the overall system of N actors or could be a group of actors each of whom is an isolate within the system (this latterpossibility assumes that at least one isolate made a sociometricchoice to himself or to an actor occupying the isolate position). Position E2 in Figure 1 and Table 1 is an example of an isolate position. A primaryposition in a network exists as a set of actors who give most of equivalentand their sociometric citations to actors with whom they are structurally who receive a nonnegligible proportionof the total numberof sociometriccitations made by the N actorsin the network. Using sociometriccitations as a measureof the interestof the citor towardthe citee, primarypositions are composed of actors who have greaterinterest in actors within their position than in actors occupying non-
108 I Social Forces I vol. 55:1, september1976 equivalentpositions although they are the object of nonnegligible interest from the overall system of N actors as representedin the network being analyzed. Several examples of primarypositions are illustratedin Table 1; E, E3, E4, SI, S2, S3, II and to some extent I2. The remaining two types of nomothetic positions distinguishedin Table 2 (cells C and D) are usually groupedtogetherwith primarypositions when a network is analyzed in terms of individual distances that are forced to be symmetric. These two types of positions are characteristically occupied by actorswho cite prestigious actors in primary positions but do not have their choices reciprocated. These positions, labelled "sycophant" and "broker" in Table2, are occupied by the usual hangers-onassociated with prestigious or wealthy actors.13 The term "sycophant" is perhapstoo harsh here, however, it emphasizes the idea of sociometric choices being given to actors outside the position and lack of choices being given to actors in the position by actors outside the position. A brokerposition in a network will also be jointly occupied by actors who give most of their citationsto nonequivalent actors, however, this position differs from the sycophantposition in that actors in the brokerposition receive a nonnegligibleproportionof citationsfrom actorsin the overall network. While actors in a brokerposition are the object of a nonnegligible proportionof the interestof actors in the system being considered, actorsin a sycophantposition are not.14 Although there are no examples of brokerpositions in the networkspresentedin Table 1, thereare two examples of sycophantpositions;E5 and I3. The actors in position E5 give 55 percentof their sociometric citationswithin the network of economic exchange relations to actors occupying position E4. Only 13 percentof theircitations are given to other actors in position E5 and no one from position E4 cites an actor in position E5. Within the overall system, only 3 percent of the available citations are given to actors in position E5. Actors in position I3 give 72 percent of their citations to actors in the primaryposition I1,while no one in positions I1, or I2 gives information-seeking sociometric citations to actors occupying I3. In summary,structure in a networkof observedrelationscan be characterized either as a pattern of relations among M structurallynonequivalent, idiographic positions and a residualcategory of nonequivalentempiricalpositions of actorsas is done in Table 1, or it can be characterizedin terms of the reflection of forms of interaction, nomothetic positions, in the idiographic positions.15 Under the latter characterization,the networks in Table 1 would be described as follows: the network of economic exchange relations consists of three primarypositions, one isolate position, one sycophant position and a residual category of actors, the networkof social exchangerelationsconsists of threeprimarypositions, a sycophant position and a residual category of actors, and the network of information-seeking relations consists of two primary positions, one sycophant position and a single in terms of idiographicpositions as given in Table 1 is isolate. The characterization the more accurate, however, the characterization in terms of nomotheticpositions emphasizesthe forms of interactionin a network.The choice between the alternative
Positions in Networks I 109 characterizationscan only be made based on the purposes of a given research project. VARIABLES IN NETWORKS AS UNOBSERVED POSITIONS EQUATIONS IN STRUCTURAL More importantthan their utility as a means of characterizingthe structurein an observed network of relations, the differentiation of distinct types of network positions sets the stage for making statementsabout the expected antecedentsand consequences of actors occupying various network positions. To the extent that a particularnetwork plays a significantpart in some social phenomenon, occupancy of differenttypes of positions within the networkshouldbe associatedwith different antecedents and consequences. Related to the questions why particularnetwork structuresshould be observed given specific conditions in a system of actors is the problemof measurementerrorin relationaldata. (See Holland and Leinhardt,c, for a detailed discussion of potential consequences for a network analysis of errorin relational data.) All three problems (the etiology of network structure, the consequences of network structureand the measurementerrorusually associated with data) can be rigorously investigated using the ideas presented in the previous
sections.
Let Dnnbe the (N by N) symmetricmatrix of social distances among the N actors in a network. Columnj of the matrix, Dj, is a (N by 1) vector of the social distances of each actor from the networkposition occupied by actorj. The set of N to differentcolumn vectors in Dnncan be partitionedinto M + 1 sets corresponding the M + 1 structurallynonequivalent sets of actors located in the hierarchical cluster analysis. Each of the vectors of social distancesto actorswithin one of the M sets of structurallyequivalent actors can now serve as an indicatorof the "true" vector of social distances separatingeach actor from the jointly occupied position. The idea here is that there is an unobserved, true vector of social distances separatingeach actor from a jointly occupied position in the network and that this unobservedvector is responsible for the observed variationin each of the indicator vectors of social distances. Just as socioeconomic status can be conceptualizedas an unobserved variable with level of completed education and dollars of income serving as indicator variables, the sycophant position in the network of social exchange relations, S4 in Table 1, can be conceptualizedas an unobservedvariable with the positions of actors 11, 21 and45 serving as indicatorvariables. In algebraic form, this suggests three epistemic statementslinking the unobservedvariableto its indicators(where all variablesare expressed as deviation scores):
D
D21=
(6 I 1,S4)S4 + Wll, (621,S4)S4
(4a)
+ W21,
(4b)
D4=
(645,S,)S4 + W45,
(4c)
110 I Social Forces I vol. 55:1, september1976

W21and W45 are vectors of error components in the observed social where WI11, distancesfrom actors 11, 21 and45 respectively, (OiLS4) is a covariancebetween the unobservedposition S4 and the observed position of actor i, and S4 iS the vector of "true" social distances separating each of the N actors from the unobserved sycophantposition. A (3 by 3) observed variance-covariancematrix, S, can be estimated from the three observed vectors: D1I, D21 and D45. The matrix S can then be approximatrix, X, which is based on six unknown matedby a predictedvariance-covariance associated with equation (4) (the variance of S4 is assumed standardized parameters in equation 5 in orderto make the equationidentified):
0
X
=821 L545J [02S4](611621645)
W21
?
021
(5a)
+ L0
or, expressed in matrixnotation:

S =
X
= zA[p] Al + 02,
(Sb)
where the subscriptsto the delta coefficients, 6, have been abbreviatedto include only the indicator number since they all contain S4, 'F2S4 iS the variance-here standardized-of the vector of "true" social distances separating actors in the networkfrom position S4, and 02wll, 02w21 02,45 are variancesof the errorscores in the vectors WI,, W21, W45. Equation (5) is the basic factor analysis model for which a variety of estimationproceduresare easily available (see e.g., Mulaik).16 factoranalyticmodel has the same form as the unrestricted The confirmatory in Av,0 or ? can be thatspecific unknownparameters in except (5), equation model forced to equal a priori values. (See e.g., Joreskog, a; Lawley and Maxwell; Mulaik.) Joreskog (b) extends the multiple factor, confirmatoryfactor analytic model such that causal inferences can be drawn among unobservedvariablesbased on the observed covariances among the indicatorvariables. (See Joreskog and van Thillo for the computerprogramassociatedwith the model and Burt, c; Wertset al.; Alwin and Tessler for applied discussion of the model.) Assuming a population multivariatenormal distributionon the indicators, maximum-likelihoodestimates are obtained for unknown parameters in the model and a chi-square statistic is routinely available which assesses the significance of the difference between S and with degrees of freedom equal to the numberof overidentifying X and is distributed restrictionson the model being estimated.17 As an illustrationof the investigation of the antecedentsand consequences equationmodels, consider the diagram of aspects of networkstructurevia structural in Figure 4. In Figure 4, squares enclose unobserved variables representingidiographic positions (labeled with the nomothetic position to which each is most similar) and circles enclose the observed positions of actors. E5 is the sycophant position in the observed network of economic exchange relations. S4 is the syco-
Positions in Networks / 1]]
U)
?~~~co
0XeNLS
-X
LO --
Cd~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\
)D AI
>Cd
CX
co
O L 00 NX@
D-
co
CY
<
1 L-18
Cd C)
E 30 alO
" dCo
112 / Social Forces / vol. 55:1, september1976 phantposition in the observednetworkof social exchange relations. I1, I2 and 13 are the M structurallyunique, jointly occupied network positions in the network of informationseeking relations. The observed positions of actors 3, 19 and 38 in the network of economic exchange relations are specified as indicators of E5 (see Figure 1). The observed positions of actors 11, 21 and 45 in the network of social exchange relations are specified as indicators of S4 (see Figure 2). Similarly, observed positions in the network of informationseeking relations are specified as indicatorsof their respective networkpositions (see Figure 3). Figure 4 illustrates the idea of specifying jointly occupied network positions as unobserved variables in structuralequation models. Its generalizabilityis
negligible since the structural equations determining S4, I,
12
and
13
have no
theoretical rationale,'8 and the model (a restrictedfactor analysis with the factor dispersion matrix unconstrained) has to be further restricted in order to insure unique unstandardizedparameterestimates of other than the error variances (the standardized estimates are unique). As a way of thinking, however, Figure 4 deals with an interesting phenomenon and one that is difficult to analyze using more traditional conceptualizations. It is hypothesized that being a sycophant in the network of economic exchange relations leads an actor to occupy a sycophant position in the network of social exchange relations. Occupying a sycophant position in either the network of social or the network of economic exchange relations then affects the type of position the actor occupies in the network of informationseeking relations. Table 3 presents the correlations and standarddeviations from which the in Figure 4 have been estimated. Table4 presentsthe standardunknownparameters ized and unstandardizedparameterestimates for cP and A.19 The chi-square approximationindicates the adequatefit of the hypothesizedmodel to these data. The unobservednetworkpositions in Figure 4 measurethe social distance of an actor from each position in the metric of social circles where a distance of one social circle from actori is defined as the numberof sociometricchoices away from actori actorj must be in orderfor the probabilityof actori not initiatinginteraction withj to equal 1.0 (see Appendix A). Accordingly,the path coefficient, P,s leading from unobserved position E to unobserved position S measures the expected numberof social circles of change in an actor's social distance from S which will occur as a result of an increaseof one social circle in the social distanceof the actor from position E. For example, if an actor in the network of economic exchange relations manipulatedhis exchange relations such that his position in the network moved one social circle away from the sycophant position, E5, then other things equal, proceeding as if the coefficients in Figure 4 are unique, he would simultaneously move .37 social circles away from the position jointly occupied by
sycophants in the network of social exchange relations
(Ps4ee
=
.37). In the in-
formationseeking network, he would move .28 social circles toward the primary position I, (Pile5 = -.28) he would remain relatively the same social distance from position '2 (Pi2e. = .05) and he would move .38 social circles away from the
Positions in Networks / 113

Table 3. CORRELATIONS AND STANDARD DEVIATIONS FOR THE OBSERVED POSITIONS IN THE EXCHANGENETWORKSAND INFORMATION SEEKINGNETWORK WHICHARE SPECIFIEDAS INDICATOR VARIABLESIN FIGURE4*
D 3 D 3 D 19 D 38 D 11 D D 45 D4 D6 D35 D 42 D D 30 standard deviations 1.00 .76 .83 .31 .30 .35 -.34 -.30 .09 .06 .47 .31 .83
D 19
D 38
D 11
D 21
D 45
D 4
D 6
D 35
D 42
D 8
D 30
1.00 .82 .27 .28 .29 -.33 -.30 .05 .02 .34 .24 .79 D3, D network,
1.00 .33 .32 .39 -.36 -.30 .08 .06 .39 .26 .74 and D38 are 1.00 .97 .90 -.41 -.31 .09 .08 .40 .32 .89
1.00 .92 -.44 -.33 .10 .08 .43 .35 .90
1.00 -.48 -.36 .14 .13 .45 .37 .91 economic D 823 1.00 .89 -.56 -.54 -.87 -.77 1.16 exchange
1.00 -.49 -.46 -.72 -.62 .77 network,
1.00 .96 .35 .29 .76 D1,
1.00 .34 .28 .79
1.00 .86 .90
1.00 .76 from the network.
*Positions social exchange
from the D35,
D21 and D45 are seeking
and D4, D6,
D42,
and D30 are
from the
information
sycophant position 13 (Pi3e5 = .38).2 In a similar manner, the path coefficients leading from the sycophantposition in the network of social exchange relationsto the threejointly occupied positions I1, I2 and 13 can be interpreted in terms of social circles. COMMENTS Using the idea of a position in a network, I have sought to extend the analysis of into areas of investigationwhich heretoforehave been difficult to network structure handle with more traditional perspectives. Several advantages derive from the perspectiveoutlinedin the previousdiscussions. First, multidimensional asymmetric relational measures are expressed in a symmetric scalar of social distance which incorporatesrelations toward and from actors so that easily available cluster and factor analytic algorithmscan be used to analyze asymmetricrelationaldata (Section 1). Second, the concept of structural equivalence is weakened in orderto allow for the possibility of measurementerror in the observed positions of structurally equivalent actors (Section 2). Third, network structureis characterizedby the relations among M + 1 structurallynonequivalent sets of actors in a networkwhere only M sets of actorsjointly occupy networkpositions-so that each actor is discussed in terms of his social distance from the M jointly occupied network positions ratherthanforcing him to occupy one of them, i.e., ratherthanpartitioning

Table 4. VARIANCE-COVARIANCE MATRIX,D, AND FACTORLOADINGMATRIX, A, FOR THE MODEL IN FIGURE4*
E5 1.04 .39 ) = -.43 .08 .53 L .72 .67 .69 .0 .0

A=
S4 (.34) 1.22 -.54 .11 .57 (.88) (.87) (.94) .0 .0 .0 .79 .81 .76 .0 .0 .0 .0 .0 .0 estimates are given (.97) (1.0) (.92)
1 (-.38) (-.44) 1.21 -.63 -1.14 .0 .0 .0 .0 .0 .0 1.06 .62 .0 .0 .0 .0 in parentheses. (1.0) (.89)
2 (.08) (.10) (-.56) 1.03 .43 .0 .0 .0 .0 .0 .0 .0 .0 .75 .75 .0 .0 (.99) (.97)
3 (.44) (.43) (-.88) (.36) 1.40 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .75 .56 (.99) (.87)
.0 .0 .0 .0 .0 .0
.0
*Standardized
the network into nonoverlappingcliques (Section 3). Fourth, network structure is characterizedby the elementary types of jointly occupied positions of which it is composed so that the etiology and consequences of occupying different types of positions in networks on concomitant social phenomena can be investigated (Section 3). And finally, networkpositions are specified as unobservedvariablesin structuralequation models so that the flexibility, mathematical simplicity and precision of structuralequation models can be broughtto bear on questions of the etiology and consequences of network structure,to analyze the applicabilityof the usual no errors-in-variablesassumption in network analysis, and to assess the adequacyof alternativehypothesizednetworkstructures as descriptionsof observed networksof relations (Section 4). In order to simplify the exposition here, a single network of relations has been assumed within a system of actors. Multiple network systems of actors are addressed elsewhere as a simple expansion of the conceptualizationgiven here (Burt, 1976f). A computerprogram, STRUCTURE, is available in FORTRANIV for the IBM 370 that inputs either a sociometric choice matrix as analyzed here or multiple networks of relations defined on any other grounds and outputs social distances, a hierarchical cluster analysis of empirical positions of actors in a networkor across multiple networks, and various sociometricindices specified here in footnote fifteen and in Lin. Cardimages are outputso as to be used in analyses of antecedentsand consequencesof an actor's empiricalposition in a system of actors.

NOTES 1. Let Gik be a vector of actor i's scores on K variables which characterizesome domain of attributes over which actor i is to be comparedto otheractors. Let GiZk be actori's perceptionof actorj's scores on the same K variables. Then a composite score of the differencebetween actori and actorj-as perceived by i-can be given as the Euclidean distance between the vectors Gik and Gijk (See Cronbachand Gleser's discussion of weighting profile differences). 2. Burt (b) and Burt andLin discuss the use of contentanalysis of archivalrecordsas a means of generating relationaldata over time which is based on the idea that the relationsbetween actors in a system are reflected in the joint occurrenceof actors in descriptionsof events occurringin the system. 3. This is the standardsociometric data generatedby questions such as: "Who are your best friends?", "With whom do you most often discuss communityaffairs?", "From whom do you most often purchase goods?", etc. 4. Stemmingfrom the work of Festinger,Luce and Luce and Perry,matrixmultiplicationis often used to determinethe smallest numberof sociometric choice links requiredby actor i to reach actor]. If Z is an (N by N) matrixof binarysociometric choice data (from footnote 3), then the matrixof shortestpossible choice link distances among actors can be found throughthe sum Z* = Z + Z2 + . . . + ZN 1, where nonzero elements in the ith power of the original matrixare set equal to i, each successive power of the original matrixhas deleted from it any elements which were nonzero in lower powers of the matrixand zii* always remainsequal to zero. The longest chain of choice links connecting two actorsin the network equals q choice links when Zq+l has no nonzero elements which were nonzero in lower powers of the matrix. 5. In regard to the shortest numberof sociometric choice links connecting actor i with actorj (z*ij in footnote 4), Katz (b) suggests that as choice links become furtherremoved from actori, they should have less significance for him than do choice links close to him. Katz suggests insertinga scale factor in the computationof Z* which decreases as Z is raised to increasing powers. Hubbell outlines a weighting scheme which weighs choices from prestigious actors more heavily than choices from relatively insignificantactors. 6. See for example, the measureof individualdistanceoutlined in AppendixA as the probabilityof actor is the one which is used in the forthcoming i not initiatinginteractionwith actorj. This operationalization numericalillustrations. 7. Even when Idij is analyzed by a nonmetricclusteringalgorithmfor bending datato a specified number of dimensions (e.g., Lingoes; Roskam and Lingoes; Young) the analysis will overlook the structural aspect of social distance due to the reliance by these methods on the criterionof monotonicitywhich is only concerned with the distance between pairs of actors-even though the entire (N by N) matrix is being analyzed. The inability of methods of analysis based on the criterionof monotonicityto consider indirectlinkages among actors is succinctly discussed by Coleman. 8. Equation(1) makes two major assumptionsabout the relations in a network. First, it is assumed that the individual distance of an actor from himself is zero, (Idii = 0). When this is not the case, then equation (1) can be given as: d*u = V[(IDi* - IDj*)'(IDi* - IDj*)], which is quite different from equation (1). In orderfor actors i andj to have zero social distance in a networkwhere {Idii} 4 0, they must have zero differences between their respective relations with each other actor in the network and they must relate to each other in the same mannerthat they relate to themselves. Equation(1) is a less general statementof the same idea. If it is assumedthe {Idii1= 0, then dij will only equal zero if actorsi andj have zero individualdistancesbetween them and identicalrelationsto other actorsin the network. I have subtractedID2j and ID2j from the equation in (1) since these terms appear twice in d*ij when {Idii} = 0. A second assumption made in equation (1) is that all relations among actors should be weighted equally. This assumptioncan be deleted if the equationis complicated still further.Let W be a = the weight of the importanceof actor i relative (2N by 2N) diagonal matrixwhere {wii} = {Wi+NAi+NI to the other actors in the network (e.g., prestige scores, power scores, etc.). The generalized form of equation(1) given above can be furthergeneralizedto take into accountthe differentialimportanceof the actorsin the networkas d*i*j = V [(IDi* - IDj*)' (W) (IDi* - IDj*)]. In orderto weight differenttypes of relationsmore heavily than others (e.g., to give null relationsmore importanceover nonzerorelations) all 2N diagonal elements can vary according to the difference term they weight. For example if zero individual distances were to be weighted as twice as important as nonzero relations, the diagonal elementsin W could be defined as {Wkkl = 2 if Idik or IdJk, Idki or Idkjequal zero, otherwiseit equals 1. 9. Katz (a), in a discussion of the general utility of treatingsociometricchoice data via matrixmethods, suggests the cosine of the angle between the choice vectors of two actorsas a measureof their similarity. This measureis the correlationbetween the rows in ID,, when individualdistance is operationalizedas

given in footnote 3. If the vector is extended to be defined in terms of relationsfrom other actorsas well as to other actors, then Katz's idea could be applied here as the cosine between IDi* and IDj* as a symmetricmeasure of the social distance between actors i andj. I have instead selected the Euclidean distance in equation (1) because it does not force the linearityassumptionon the quantification of social distance. 10. In a similar sense the idea of individual distances being transitive or balanced is invariant over researchapplications(see e.g., Davis, a, b; Hallinan;Heider; Holland and Leinhardt,a, b). The social distance between actors i andj can be seen as a measure of the dissonance or imbalance in the overall relationbetween actors i andj within a specific network. dij will increase as a function of the difference in relationsbetween actors i andj and all actorsin the networkbecome different. Unfortunately, much of the work on transitivitywithin a triad does not consider the three person group in terms of the actors linked to each of the persons in a triad(see Flament;and Hallinanand McFarlandfor exceptions to this statement).In relationto the discussions of transitivity,equation(1) considers all N persons in a network in order to expand the distance between persons who are in intransitivetriads or tetrads of N-person groups. 11. The definitionof strong equivalence is taken from the basic concept of equivalence in typology (see e.g., Kelley, 9-10) discussed in sociology by Holland and Leinhardt(b) and Lorrainand White. Lorrain and White (63) suggest: "Objects a, b of a categoryC are structurally equivalentif, for any morphismM and any object x of C, aMx if and only if bMx, and xMa if and only if xMb. In other words, a is structurallyequivalent to b if a relates to every object x of C in exactly the same way as b does." A morphismis any direct or indirectrelation between two objects or actors. A category is a set of objects and a set of morphisms. In equation(2) the category over which two actorscan be structurally equivalent consists of the set of N actors as representedin networks of some type of relations and the morphism of individual distance (aMx = Idax, bmx = IdbX, xMa = Idxa, xMb = IdXb). In a related perspective, Holland and Leinhardt (b, 110-1) specify structuralequivalence of actors in an M-clique when the M-clique is situated within a larger network of actors. The utility of stating structuralequivalence in terms of the scalar in equation(2) is twofold: (1) its greatersimplicity, and (2) its facile generalizationto a weak criterionof structural equivalence as given in equation (3). 12. Reference the differencebetween two actors' relationsto other actorsin the networkas "structural" distance. This stands in contrastto individual distances between the two actors themselves. Admitting the possibility of weak structuralequivalence brings in problems of interpretingthe conditions of distance underwhich actors will be deemed equivalent . The only time that the individualand structural values of individualand structural distanceresponsiblefor an observed social distancecan be determined from the observed social distance with certaintyis when dj equals zero. In such a case, all component distances have to be zero. In equation (3), however, an individual only knows that if dj is less than the criterion distance then structuraland individual distances are also less than the criterion. Isolates will have zero structuraldistance between them but individual distances of 1.0 separatingthem so that the social distances among isolates will equal 1.41. This means that the isolates in a networkwill be weakly structurally equivalentwhenever the criterionalpha is greaterthan 1.41 (e.g., in the economic exchange network in Figure 3, the isolates are simultaneously clustered together as position E2 as soon as the criterionincreases from less than 1.4 to less than 1.5). In short, special attentionneeds to be paid to the degree to which a chosen nonmetric solution has collapsed otherwise meaningful distances among actors. This caution extends to the clustering algorithms mentioned by White et al. in order to locate "blocks" of actors in a network-particularly in the case of solutions which are a "lean" fit (See Breigeret al.). 13. The idea of sets of actorsbeing defined by their common referenceto actorsor events externalto the set-as are the actors jointly occupying sycophant or broker positions as discussed here-has been discussed under various labels such as; "quasi-groups" (Ginsberg, 40), "latent groups" (Olson, 50), "levelling coalitions" (Thoden van Velzen, 241-2) and "latent corporateactors" (Burt, b). 14. Boissevain (147-69) provides an interesting discussion of some of the expected behaviors and characteristicsof actors occupying what is here describedas a brokerposition in a network. 15. This line of reasoning can be continued to express the extent to which the empirical position of an actor reflects each of the four nomotheticpositions in Table 2. Define two scalars describingaspects of actor j's empirical position in a network: Aj = the prominence of actor j within the network = with actorj; (N - >N J1 Id/) IN = 1.0 when all actorshave maximumprobabilityof initiatinginteraction equivalentto himself = Bj = the tendency for actorj to initiate interactiononly with actors structurally = 1.0 when actorj only has maximum probabilityof initiating interaction [ Ni=1IdjidjJ / [Ni=1dji] with those actors who have zero social distance from him, where Idij and dij are respectively individual

(AppendixA) and social (equation 1) distancefrom actori to actorj. The extent to which actorj occupies a primary position within the network can then be given as; (A) (Bj). The extent to which actor j occupies a sycophantposition in the networkcan be given as; (1 - Aj) (1 - Bj). The extentto which actorj occupies an isolate position is given as; (1 - Aj) (Bj). Finally, the extent to which actorj occupies a brokerposition is given as; (Aj) (1 - Bj). Each of the indices equals 1.0 at its maximum and zero at its minimum. Since they are expressed without reference to particularrelations between particularactors, these indices can be used as crude indicatorsof structuralforms reflected in the empirical positions of actorsin separatenetworksand separatesystems. 16. MacRae and Bonacich discuss cliques in networks in terms of a factor analytic model. The present situation differs from theirs in two respects. First, use of social distances, equation (1), allows asymmetricrelational data to be expressed in the form of a symmetric scalar so that easily available in a network. Second, cluster analytic and factor analytic algorithmscan be used to analyze the structure while both MacRae and Bonacich explore possible methods of applying the factor analytic model to locate cliques, the present approachis using a nonmetric, nonlinear cluster analytic model to locate structurallynonequivalentpositions and then using a confirmatoryfactor analytic model to assess the adequacy of the hypothesized structure. This second difference provides a method of assessing the to an or nonequivalentpositional structure significanceof the lack of fit of a hypothesizedclique structure observednetworkas well as enabling an investigatorto specify antecedentsandconsequencesof network structurein structural equationmodels (see Figure 4). 17. The chi-squarestatistic is given as a function of the discrepancybetween S and l;
X2=
(N -1) [lIn Il; I + trace (z-1 S)- In I51- r] equationmodel. where r equals the total numberof observed variablesin a hypothesizedstructural
Since the social distance function estimates dj by taking into account otherobservations,however, there will be fewer thanN independentobservationsunderlyingthe chi-squarestatisticand the estimationof S. This, in combinationwith the usual small N in networkanalysis, means that the chi-squarestatisticis a very rough approximationand should only be taken as an indicationof substantivesignificancewhen its value has a high probabilityof occurringor an extremely low probabilityof occurring. 18. I have in mind here the specification of the determinantsof change in patternsof relations among actors in a network. How do jointly occupied, structurallynonequivalentpositions come to exist in a network? Explanations along this line have been idiographic due to the emphasis in past network analyses on description. However, this promises to be a most fruitful area for investigation as several recent works emphasize. Jointly occupied positions could arise as a function of causal relationsamong networksbased on differenttypes of relations (e.g., Laumannet al.) as a functionof social backgrounds (e.g., Bartonet al.; Gans; Laumann;Laumannand Pappi,b), as a function of changes in environmental conditions which affect the system of actors (e.g., Allen; Burt, d, b, e; Stanworthand Giddens) or as a function of actors purposively exploiting their positions in exchange networks in orderto realize goals (e.g., Boorman;Burt, d, e; Granovetter, a, b). 19. The structuralcoefficients among the unobserved variables in Figure 4 are regression coefficients which can be estimated from the variance-covariance matrix (D using ordinaryleast squares since the model specifies no paths equal to zero. These estimates will be equivalentto those in Figure 4 which are maximum likelihood estimates generated from a confirmatoryfactor analytic model (programmedas LISREL, Joreskogand van Thillo). Since the emphasis in the presentdiscussion is on network structure rather than structuralequation models, I have avoided the statistical and algebraic complications involved in the general estimationand specificationof coefficients among unobservednetworkpositions beyond the bivariate covariances in equation (5). The reader interested in more detailed treatmentis referredto Joreskog (b) for a general discussion or to Burt (a, c), Wertset al., and Alwin and Tessler for applieddiscussion. 20. I have ignored the second-ordereffects of the compoundpath coefficients measuringchange in posinetworkresultingfrom change in the social exchange network. Changein tion in the information-seeking E5 leads to change in S4 which in turn leads to change in II, I2 and I3. For example, an increase of one social circle of distance in E5 would lead to an increaseof (Pi3e5 + Ps4e. Pi34) = .38 + .13 = .51 social circles of distance away from position I3 instead of the .38 reportedin the text.
118 / Social Forces / vol. 55:1, september 1976 APPENDIXA: AN OPERATIONALIZATION OF INDIVIDUAL DISTANCE AS THE OF ACTORI NOT INITIA PROBABILITY WITH TINGINTERACTION ACTORJ In the interestof statingindividualdistance in a metric which is interpretable across different types of networks (e.g., networks of friendshiprelations, of economic exchange relations, etc.) it would be convenient to have a measure of the probabilityof actor i not initiating of Idij to be substitutedin equation(1) of the interactionwith actorj as an operationalization text. In other words, if Idij equals 1.0 in a networkof economic exchange relations, it shows that actor i will not seek economic goods from actorj as the networkis currentlyarranged. The operationalization given here is based on the minimum numberof choice links required by actor i to reach actorj (zij* in note 4). It has the advantageof a metric interpretation in terms of the circle of actorslinked to actori within the networkand has thereforebeen used in the text illustrations. The operationalizationof Idij as the above described probabilisticmeasure requires three pieces of information:(1) conditions leading Idij to equal its minimum, (2) conditions leading Idij to equal its maximum, and (3) a functionrelatingchange in Idij to change in zi. Actor i will have a circle of actors with whom he has a nonzero probabilityof initiating interaction(cf. Kadushin'sdiscussion of the concept of an actor's social circle). Within this social circle, actori has the highest probabilityof interactingwith himself. Idii = 0. There is a close to zero probabilityof interactionbeing initiated by actor i toward actors separated from him by an infinite zw* (e.g., isolates in the network or persons in separatesystems of actors). Idij can be treatedas having reachedits maximumwhen zu* is largerthanthe largest finite zij* for actor i. Let jjbe the largest finite zij* linking actori to any actorin the network. Since values of zij* which are larger than jj imply a zero probabilityof actor i initiating interactionwith actorj, j provides a boundaryfor the social circle aroundactori. The value of 4w can be discussed as the range of actor i's social circle. Actor i would not be expected to initiate interactionwith actors outside his social circle unless he needs unusualresources not readilyavailable from actorswithin his circle (e.g., see Granovetter, a; Lee; Milgram;White, for discussion of the small world phenomenonand the use of weak links to realize unusual interests). Given the above anchoringsfor the maximumand minimumvalues of Idjjrelative to zw*the rate of change in Idij in relation to change in zi.* can be drawn from researchon human perception. Findings in psychophysics demonstratethe limited ability of persons to retain informationpast a boundarylevel of stimulation and inverse ability to discriminate differences in levels of stimulation as the overall level of stimulation approaches the boundarylevel (see e.g., Haber;Simon). The level of stimulationhere is the proportionof actors in his social circle with whom actor i will initiate interaction.It is to be expected that actor i would have little difference in Idij for the few actors who are close to him relative to the other actors in the network, that the change in Idij would increasewith change in z* as a functionof the numberof actorsat a distancez*, and that actori would have little difference in Idij for the actors who are near the boundaryof his social circle and beyond. Let the N values of zw*be rankorderedfrom lowest to highest for actor i: Changein Idij would then be a function of the cumulativenumberof actorsin i's social circle who are less than or equal to Zi.*choice links away from i: , if i =j, cumulativefrequencyof zIj* f * Id13 = [(numberof actorsin social circle of actor i) + 1]' ifz*1 1, if z* > (U.
IdWwill be low (i.e., reflect a high probability of i initiating interaction with j) when Z*ijis low relative to the other Zik* linking actor i to actors in the network. The important characteristic of the individualdistancebetween actor i andj-in terms of the probabilityof i initiatinginteractionwithj-is not the absolute numberof choice links requiredby i to reach
Positions in Networks / 119 j, but ratherthe location of actorj in actor i's social circle. The decision to interactwith actor j can be seen as a marginal consumptiondecision. The commodity being consumed here is the initiationof interactionwith a proportionof the actors in actor i's social circle. From an initial point of Idij = .5, increases in zw*result in smaller and smaller increases in Id1juntil Idij reaches its maximumof 1.0. From the same point, decreases in zw*result in smaller and smaller decreases in Idij until Idij reaches its minimum at 0. What this says is that as Zw* approachesits limits at either the boundaryor center of actor i's social circle, the marginal propensityof actor i to initiate interactionwith actorj declines. APPENDIXB: CLUSTERS OF ACTORSIN NONMETRIC CLUSTERING ALGORITHMS BASED ON INDIVIDUALVERSUSSOCIALDISTANCES It is common for individual distances among actors to be forced to be symmetricand then to be subjectedto a nonmetricclusteringalgorithm-often a smallest space analysis (see review by Bailey). Given this frequentpracticeby sociologists, an importantquestionin referenceto the location of the M structurally nonequivalentpositions in Section 3 concerns the different inferences which would result from a nonmetric cluster analysis of symmetrizedindividual distances instead of the social distances used to compareactors as networkpositions. A comparisonof the smallest space analyses in Figures 1, 2 and 3 with Figures 3, 4 and 5 in Laumannand Pappi (a) will illustratethe potentiallydifferentconclusions which can be gleaned from analyses based on individual versus social distances between actors. As a general rule of interpretation,an analysis based on individual distances (as exemplified by Laumannand Pappi's careful analysis) will conclude that two actors occupy similarpositions in a networkif they interactfrequentlyor if they have homophily in terms of some profile of characteristics-i.e., if they have small individualdistance between them. An analysis based on social distances (as illustratedhere in Figures 1, 2 and 3) will conclude that two actors occupy similar positions if they have the above characteristicof small individual distance between themselves as well as having similar individual distances to and from the other actors in the network-i.e., if they have small social distance. Analyses of individual distancesemphasize linkages between pairsof actors. Analyses of social distancesemphasize similaritiesbetween patternsof relationshipswith the N actors in a system. There are two general characteristicsof a network analysis in terms of which an analysis of individual distances can be comparedwith one based on social distances: (1) the expressed structurally nonequivalentpositions in a network, and (2) the distributionof actors in reference to those nonequivalentpositions. Concerningthe first characteristic,notice that Figures 1, 2 and 3, through the use of social distances, emphasize the similarity between actors in structurallyequivalent positions and emphasize the differences between actors in nonequivalentpositions (underthe weak criterion). The structurallynonequivalentpositions in a network are thereby made more obvious and easier to detect in an analysis. This emphasis on structureis most vividly demonstrated in Figure 3 where the previous scatterof actors throughthe smallest space analysis of individual distances (Figure 5 in Laumannand Pappi) is here reduced to three nonequivalentnetwork positions. These structurallynonequivalentpositions provide a set of referencepoints from which each actor in a networkcan be located and discussed in contrastto the idea of comparingactors in terms of their relative distance from the "center" of a smallest space. Concerningthe second characteristicof a networkanalysis given above, actorsare distributedin space with referenceto nonequivalent positions accordingto the principle of interpretingan analysis based on individualdistances versus one based on social distances which was emphasizedabove. For example, the isolates in the economic exchange network are occupying a joint position in the network instead of being scattered around the periphery of the network as they would be in an analysis of individual distances. Similarly, the six most influential actors in the community jointly occupy a position in the information-seekingnetwork (position I, in Figure 3) because they all have the characteristic of receiving many citations from other actors in the networkwhich

they do not reciprocate.Their position in the networkis placed on the peripheryof the social space as representedin the smallest space analysis since they occupy a position which is very different from the positions of most actors in the network. In an analysis of individual distances these actors would be near the center of the space since so many actors claim to interactwith them (this characteristic is elaboratedby Laumannand Pappi, a, as a measureof the integrativecentralityof the actor). In an analysis of social distances, on the otherhand, an actor is close to the center of the social space when his position is equally dissimilarto the positions of the other actors in the network. The more that the actors jointly occupying a networkposition receive a disproportionate numberof sociometric choices-more or lessthen the more peripheralthe position will be in the social space as representedin a smallest space analysis. REFERENCES Elite Cooptation:InterlockingCorAllen, M. P. 1974. "The Structureof Interorganizational porateDirectorates."AmericanSociological Review 39:393-406. Alwin, D. F., and R. C. Tessler. 1974. "Causal Models, Unobserved Variablesand Experimental Data." AmericanJournal of Sociology 80:58-86. Bailey, K. D. 1974. "Cluster Analysis." In David R. Heise (ed.) Sociological Methodology 1975. San Francisco:Jossey-Bass. Barton, A. H., B. Denitch, and C. Kadushin. 1973. Opinion Making Elites in Yugoslavia. New York:Praeger. Boissevain, J. 1974. Friends of Friends: Networks, Manipulators and Coalitions. Oxford: Blackwell. Bonacich, P. 1972. "Factoring and Weighting Approaches to Status Scores and Clique Identification."Journal of MathematicalSociology 2:113-20. Boorman, S. A. 1975. "A CombinatorialOptimizationModel for Transmissionof Job InformationThroughContactNetworks." The Bell Journal of Economics 6:216-49. Braithwaite,R. B. 1953. Scientific Explanation.New York:Harper. Breiger,R. L., S. A. Boorman, andP. Arabie. 1974. "An Algorithmfor Blocking Relational Data, with Applications to Social Network Analysis and Comparisonwith Multidimensional Scaling." TechnicalReport 244, Institutefor MathematicalStudies in the Social Sciences, StanfordUniversity. Burt, R. S. a: 1973. "ConfirmatoryFactor Analytic Structuresand the Theory Construction Process." Sociological Methods and Research 2:131-90.
. b.: 1975. "Corporate Society: A Time Series Analysis of Network Structure." Social
Science Research 4:271-328.

. c: 1976. "Interpretational Confounding of Unobserved Variables in Structural Equa-
tion Models." Sociological Methods and Research 4: in press.

. d: 1976 "Power in a Social Topology." In R. J. Liebert (ed.), Power Paradigms
and Community Research. London:Sage.

. e: 1976. "The Rise and Fall of Jointly Occupied Positions in Social Structure."
Paperpresentedat the annualmeeting of the Public Choice Society. of Conceptualization . f: 1976. "Positions in MultipleNetworkSystems: A Structural Role-Set, Role Distance, and Status." Chicago: NORC, University of Chicago (unpublished). Burt, R. S., and N. Lin. 1975. "Network Time Series from Archival Records." Paper presented at the annualmeeting of the American Sociological Association. Coleman, J. S. 1970. "Clustering in N Dimensions by Use of a System of Forces." Journal of MathematicalSociology 1:1-47. Cronbach, B. E., and G. C. Gleser. 1953. "Assessing Similarity Between Profiles." Psychological Bulletin 50:456-73.
Positions in Networks / 121 Davis, J. A. a: 1967. "Clustering and StructuralBalance in Graphs." Human Relations 20:181-7. . b: 1968. "Social Structureand Cognitive Structures."In R. Abelson et al. (eds.), Theories of Cognitive Consistency:A Sourcebook. Chicago: Rand McNally. Festinger,L. 1949. "The Analysis of Sociograms Using MatrixAlgebra." HumanRelations 2:153-58. Flament, C. 1963. Applications of Graph Theory to Group Structure. Englewood Cliffs: Prentice-Hall. Gans, H. 1962. The Urban Villagers: Groupand Class in the Life of Italian-Americans.New York:Free Press. Ginsberg, M. 1934. Sociology. London: Butterworth. Granovetter,M. S. a: 1973. "The Strengthof WeakTies." AmericanJournal of Sociology 78:1360-80. . b: 1974. Getting a Job: A Studyof Contacts and Careers. Cambridge: Harvard University Press. Haber, R. N. 1968. ContemporaryTheory and Research in Visual Perception. New York: Holt, Rinehart& Winston. Hallinan, M. T. 1974. The Structureof Positive Sentiment. New York:Elsevier. Hallinan, M. T., and D. D. McFarland. 1974. "Higher Order Stability Conditions in MathematicalModels of Sociometric or Cognitive Structure."Journal of Mathematical Sociology 4:131-48. Heider, F. 1958. The Psychology of InterpersonalRelations. New York:Wiley. Hempel, C. G. a: 1952. "Concept Formation." In 0. Neurath, R. Carnap, and C. Morris (eds.), Foundations of the Unity of Science. Chicago: University of Chicago Press. . b: 1965. The Structureof ScientificExplanation. New York: Free Press. Holland, P. W., and S. Leinhardt.a: 1970. "A Method for Detecting Structurein Sociometric Data." AmericanJournal of Sociology 76:492-513. . b: 1971. "Transitivity in Structural Models of Small Groups." ComparativeGroup Studies 2:107-24.
. c: 1973. "The Structural Implications of Measurement Error in Sociometry."
Journal of MathematicalSociology 2:1-27. Hubbell, C. H. 1965. "An Input-OutputApproach to Clique Identification."Sociometry 28:377-99 Johnson, S. C. 1967. "HierarchicalClusteringSchemes." Psychometrika32:241-54. MaximumLikelihood FacJoreskog, K. G. a: 1969. "A General Approachto Confirmatory tor Analysis." Psychometrika34:183-202.
. b: 1973. "A General Method of Estimating a Linear Structural Equation System."
In A. S. Goldbergerand 0. D. Duncan (eds.) StructuralEquationModels in the Social Sciences. New York:Seminar. Joreskog, K. G., and M. van Thillo. 1972. "LISREL, A GeneralComputerProgramfor Estimating a LinearStructural EquationSystem Involving Multiple Indicatorsof Unmeasured Variables."Research Bulletin RB-72-56. Princeton:EducationalTestingService. Kadushin,C. 1968. "Power, Influenceand Social Circles: A New Methodologyfor Studying Opinion Makers."AmericanSociological Review 33:685-99. Katz, L. a: 1947. "On the MatrixAnalysis of Sociometric Data." Sociometry 10:233-41. . b: 1953. "A New Status Index Derived from Sociometric Analysis." Psychometrika 18:39-43. Kelley, J. L. 1955. General Topology. New York:Springer-Verlag. Laumann,E. 0. 1973. Bonds of Pluralism. New York:Wiley-Interscience. Laumann, E. 0. and F. U. Pappi. a: 1973. "New Directions in the Study of Community Elites." AmericanSociological Review 38:212-30. . b: 1976. Networks of Collective Action: A Perspective on Community Influence Systems. New York:Academic.

Laumann,E. O., L. M. Verbrugge,and F. U. Pappi. 1974. "A Causal Modelling Approach to the Study of a CommunityElite InfluenceStructure."AmericanSociological Review 39: 162-74. Lawley, D. N., and A. E. Maxwell. 1971. Factor Analysis as a Statistical Method. New York:Elsevier. Lee, N. H. 1969. The Searchfor an Abortionist. Chicago: University of Chicago Press. Lin, N. 1968. "Innovative Methods for Studying Innovationin Education."InResearch Implicationsfor EducationalDiffusion. East Lansing:Departmentof Education,Michigan State University. Lingoes, J. C. 1972. "A General Survey of the Guttman-Lingoes Nonmetric Program Series." In R. N. Shepard, A. K. Romney, and S. G. Nerlove (eds.), Multidimensional Scaling. Vol. 1. New York:Seminar. Lorrain, F. P., and H. C. White. 1971. "StructuralEquivalence of Individuals in Social Networks." Journal of MathematicalSociology 1:49-80. Luce, R. D. 1950. "Connectivity and GeneralizedCliques in SociometricGroupStructure." Psychometrika15:169-90. Luce, R. D., and A. D. Perry. 1949. "A Method of Matrix Analysis of Group Structure." Psychometrika14:95-116. MacRae, D. 1960. "Direct FactorAnalysis of Sociometric Data." Sociometry23:360-71. Milgram, S. 1967. "The Small WorldProblem" Psychology Today22:61-7. Mulaik, S. A. 1972. The Foundations of Factor Analysis. New York:McGraw-Hill. Olson, M. 1965. The Logic of Collective Action. Cambridge:HarvardUniversity Press. Roskam, E., andJ. C. Lingoes. 1970. "MINISSA-1: a FORTRANIV Programfor the Smallest Space Analysis of SquareSymmetricMatrices."Behavioral Science 15: 204-20. Simmel, G. 1950. The Sociology of Georg Simmel. (K. H. Wolf, trans.) New York: Free Press. Simon, H. A. 1974. "How Big is a Chunk?"Science 183:482-8. Stanworth, P., and A. Giddens. 1975. "The Modern CorporateEconomy: InterlockingDirectorshipsin Britain 1906-1970." Sociological Review 23:5-28. Tajfel, H. 1969. "Social and CulturalFactors in Perception." In G. Linzey and E. Aronson (eds.), The Handbookof Social Psychology. Reading, Mass.: Addison-Wesley. Thoden van Velzen, H. U. E. 1973. "Coalitions and Network Analysis." In J. Boissevain and J. C. Mitchell (eds.), Network Analysis Studies in Human Interaction. Paris: Mouton. Werts, C. E., K. G. Joreskog, and R. L. Linn. 1973. "Identificationand Estimationin Path
Analysis with Unmeasured Variables." American Journal of Sociology 78:1469-84.
White, H. C. 1970. "Search Parameters for the Small World Problem" Social Forces 49(December):259-64. White, H. C., S. A. Boorman, and R. L. Breiger. 1976. "Social Structurefrom Multiple Networks." AmericanJournal of Sociology 81 :in press. Young, F. W. 1972. "A Model of Polynomial Conjoint Analysis Algorithms." In R. N. Shepard, A. K. Romney and S. B. Nerlove (eds.), MultidimensionalScaling. Vol. 1. New York:Seminar

Burt Positions in Networks

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Burt Positions in Networks

Uploaded by

Copyright:

Available Formats

Positions in Networks Author(s): Ronald S. Burt Source: Social Forces, Vol. 55, No. 1 (Sep., 1976), pp.

University of California, Berkeley

Nk=,1(Idki- Idkj)2 - Id2

IDj*)' (IDi*- IDj*) -Idi2

98 / Social Forces / vol. 55:1, september1976

*MMMM 2 MREH:i g00001

WllU tol ,0~~~~~c 4) ~~~~~co

Positions in Networks /99

OO .r4~~~~ 4-H H I-I+0 O C)0 -p k

~ ~~~~~~~_ ~02 C)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~4

*HdH wC)co 4-)C)4C

100 / Social Forces / vol. 55:1, september1976

... p4 ~ ~~~~tCMJH *rl~~~~~~~~~~~~~~~~~~~~~~~~~r

1 04' to :0) 0)C

0H. 0 (f... AUUp4U2

CM CY) I.-I C\j (Y') U'\ C\j

102 / Social Forces / vol. 55:1, september1976

> . . . .3 .300g ('Sol~co . . . . .

C'J (YLr (r)C J

Positions in Networks /103

~~~-4 0~~~~~~-4 oUN

Positions in Networks I 105

Residual Total Social

Total (21 (35 (18 citations) citations) citations)

( 9 citations) (35 (118 citations) citations)

Residual Total Information

Total (31 (76 citations) citations)

( 0 citations) (116 in citations) row in to ac-

Positions in Networks I 107

ILLUSTRATIVE A 0? 0 0 B 0 high high

PATTERNS OF RELATIONSt C 0 0 low D 0 0 0 RESIDUAL 0 0 0?

ISOLATE PRIMARY SYCOPHANT

by actors position jointly position

classified. position choices actors position descripclassifioc-

occupying row position

proportion are for given these not a network.

sociometric occupying Although used in this

presented predictions network types

(6 I 1,S4)S4 + Wll, (621,S4)S4

110 I Social Forces I vol. 55:1, september1976

=821 L545J [02S4](611621645)

or, expressed in matrixnotation:

Positions in Networks / 1]]

.37). In the in-

Positions in Networks / 113

1.00 .92 -.44 -.33 .10 .08 .43 .35 .90

1.00 -.49 -.46 -.72 -.62 .77 network,

1.00 .96 .35 .29 .76 D1,

1.00 .34 .28 .79

1.00 .86 .90

1.00 .76 from the network.

*Positions social exchange

from the D35,

D21 and D45 are seeking

and D4, D6,

and D30 are

114 I Social Forces I vol. 55:1, september1976

E5 1.04 .39 ) = -.43 .08 .53 L .72 .67 .69 .0 .0

2 (.08) (.10) (-.56) 1.03 .43 .0 .0 .0 .0 .0 .0 .0 .0 .75 .75 .0 .0 (.99) (.97)

3 (.44) (.43) (-.88) (.36) 1.40 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .75 .56 (.99) (.87)

Positions in Networks / 115

116 / Social Forces / vol. 55:1, september1976

Positions in Networks / 117

120 / Social Forces / vol. 55:1, september1976

IDj)' (IDi- IDj*) -Idi2

WllU tol ,0~c 4) ~co

... p4 ~ tCMJH *rl~~~~~~~~~~~~~~~~~~~~~r

~-4 0~~~~-4 oUN