The Meaning of "Validity" and "Reliability" as Applied to Sociometric Tests

Abstract

IT IS the purpose of this paper to attempt a clarification of the meaning of the terms &dquo;validity&dquo; and &dquo;reliability,&dquo; as they may be applied appropriately in an evaluation of sociometric tests. These instruments are being employed increasingly in educational and psychological research, having been in wide use by sociologists for a number of years. It is their intention to focus on an analysis of the individual as a member of a group, rather than on the individual as an entity in isolation.It seems pertinent, in view of the unique possibilities which sociometry affords through its use as a tool of psychological research, to try to define as clearly as possible its nature in relation to the methods more familiar to psychological workers. A psychometric orientation to the use of paper-and-pencil tests implies a frame of reference in their evaluation which is not immediately applicable to sociometry.'fhe Concept of &dquo;Validity&dquo; An operational definition of a psychological test would involve a statement to the effect that it is one in which an individual responds to a number of items, these responses being taken to be indirect evidence of the possession by that individual of certain characteristics to a certain degree. A sociometric test, on the other hand, requires the selection by each individual in a specified group of one or more other individuals in that group on the basis of stipulated criteria of choice. Expressed simply, in the former case, the individual &dquo;reports on&dquo; himself; in the latter, he &dquo;reports on&dquo; and is &dquo;reported on&dquo; by others.1 The writer wishes to thank Harold B. Pepinsky and John B. Holland for their careful reading of this paper and for their constructive criticism.

‘Do you still like to play with him?'

“…However, in 1949, Pepinsky already stated that the dynamic nature of social relations causes fluctuations in sociometric data over any period of time. Such fluctuations reflect actual changes in behaviour and group relations rather than the unreliability of the measure (Pepinsky, 1949;Ramsey, 1995). A similar idea is advocated by a group of scholars working in dynamic systems theory (see for instance Fogel, 1993;Thelen & Smith, 1994;Van Geert, 1994;and Van Geert & Steenbeek, 2005, for an overview).…”

“…In spite of its importance, very little research has been done about the stability of given ratings of children (Pepinsky, 1949;Bukowksi & Newcomb, 1984;Terry, 2000). Duncan and Cohen (1995) found that in grade 1, 3, 6 and 8, children of the same gender rated each other higher than children of the opposite gender and popular children gave higher ratings than children of other sociometric groups.…”

“…By looking at fluctuations in a developmental process, one gets a better idea about how development originates (Van Geert & Van Dijk, 2002;Pepinsky, 1949;Fogel, 1994;Newcomb & Bukowski, 1993;Wu et al, 2001). Thus, instead of focusing uniquely on stability (as the underlying, real construct that awaits discovery), the alternative approach presented in this article is based on the idea that variability is an essential feature of development.…”

Some Methods of Comparing Sociometric Matrices

“…h performs the same function for the square matrix of sociometric ranks. This anpears to meet Pepinsly's demand for a measure of reliability of sociometrics , in the sense of consistency over judges (7). No splitting of the group into chance halves is required.…”

Sociometry in School‐‐i. Sociometric Techniques

“…The problem of splitting a sociometric test into two equivalent halves rules out the use of split-half reliability. A discussion at greater length of these points may be found in Pepinsky (1949) and Gronlund (1959). The most important questions seem to be concerned with ways of eliciting true choices at any time and with the stability of choices at different ages and under varying conditions.…”