forms that reduced structural patterns to almost geometric sets of relationships (Wolff 1950). Similar procedures are utilized by modem network theorists when conceptualizing connections between nodes in a social network. These uses are compatible with ours but at a more abstract level where virtually all behavioral content is removed and focus is on the formal or mathematical properties of the network. This content downloaded from 207.46.13.169 on Sat, 01 Oct 2016 05:29:04 UTC All use subject to http://about.jstor.org/terms 572 AMERICAN SOCIOLOGICAL REVIEW This points to the central and vital function of classification in the grand strategy of science. Its most important purpose is to allow us to generalize what we learn by the study of a few members of a category to other members of the same category. The ability to generalize includes the ability to predict. If we are able to say that something is a member of category and all Y's behave in such a fashion, then we can predict that the new object Y will behave in a similar fashion. But to make this procedure work optimally it must be grounded on valid structural models. These models represent how objects in nature are joined and how their parts operate in relationship to each other and, as a result, provide the basis for explanation and theory (Willer 1967; Bates and Harvey 1975). The purpose of classification, on the other hand, is to raise explanation to the level of a generalization to a class of similar objects. This is accomplished by a reiterative procedure in which a generalization in the form of a hypothesis is derived from a model and tested on a few selected members of a category. The ultimate success of the procedure depends upon preserving a clear distinction between concepts derived from classification, and those derived from modeling. If models are constructed using categories, the procedure becomes circular and selfconfirming. SUMMARY AND DISCUSSION The above argument is illustrated in Figure 1 which, from bottom to top, represents three levels of abstraction, each succeeding level moving further from the world of sense impressions. At the first level of abstraction, structural models of three separate groups are shown. Only three-person (position) groups are shown for the sake of simplicity. Let us assume they are all work groups of a similar type, for example, offices. Each group is represented symbolically by a set of geometric figures in a definite relationship to each other. The shape of the geometric figures represents the different positions and roles contained within each group. For example, the triangle represents the position of physician, the square the position of nurse, and the circle the position of office manager/secretary. Each geometric figure represents a unique position and is labeled by a subscript indicating, first, group location and, second, its identity in the group. The triangle or physician's position in group one is given the subscript 1,1 to indicate it is located in group 1 and is identified as position 1. In group 2, the physician's position is given the subscript 2,1, indicating it is position 1 in group 2, and so on. Inside each geometric figure the letters X, Y, and Z represent the actors occupying each position, and the subscript indicates group location. They could be read as Physician X1, Nurse Y1, Office manager Z1. The letters selected are arbitrary and they may represent a distinctive feature of the positional occupant such as sex, race, or other attribute. The double-ended arrows represent the unique relationships between actor-occupied positions and are designated by R with a triple subscript. The first number in the subscript indicates the group location, and the last two the positions in contact. For example, R112 indicates the relationship in group 1 between positions 1 and 2. It is important to note that each group is represented by a separate model involving particular people occupying particular social positions in association with each other in the context of different offices. These models would utilize concepts referring to unique social units and the relationships or connections that relate these units to each other to form an observable bounded whole. Although they are all physicians' offices, in the real world they exist as separately occurring behavioral networks. Again, the purpose of these structural models is to furnish an analyst with a symbolic representation of an entity, a social group in this case, which can be used to formulate theories about how entity elements function in relations to each other. To formulate theories for scientific testing, attributes of the parts of the model or of their relationships to each other are examined in terms of how changes in one are related to changes in the others. In the middle section of Figure 1 we show three classification schemes of positions, relationships, and actors or position occupants. For example, all triangles are classified together as physician positions and assigned the symbol A, squares are classified together as nurse positions and assigned the symbol B, and so on. Relationships and occupants have also been classified and symbolically designated. This content downloaded from 207.46.13.169 on Sat, 01 Oct 2016 05:29:04 UTC All use subject to http://about.jstor.org/terms CONCEPTUALIZING SOCIAL STRUCTURE 573
No takes yet. Share an insight, caveat, or question.
Bates et al. (1989) studied this question.
Synapse has enriched one closely related paper. Consider it for comparative context: