Sociometric triads as multi‐variate systems†

I wish to thank John Fry for dogged and sophisticated programming assistance. John Faris kindly made available illustrative sociograms. Research supported in part by NSF grant SOC 73–05489 to Carnegie‐Mellon University.

Binary choices (choose v. not choose) among three persons may be viewed as variables and the possible choices in a triad as a 64 celled contingency table, amenable to analysis in terms of log linear effects. Fifteen non‐redundant effects occur, of which four appear consistently in a data bank of 384 sociograms. Consequently, it is claimed one may reproduce the triad frequencies in most sociograms by the effects implied by four propositions: (A) P is more likely to choose O when O chooses P, net of all other variables; (B) P is more likely to choose O when X chooses O. This association is particularly strong when imitation produces transitivity; (C) when X chooses O (O chooses X), P tends to choose both or neither; and (D) P tends to choose an X who favors P's choices. P tends to not choose an X who favors P's non‐choices.

Notes

I wish to thank John Fry for dogged and sophisticated programming assistance. John Faris kindly made available illustrative sociograms. Research supported in part by NSF grant SOC 73–05489 to Carnegie‐Mellon University.