This reflection paper challenges current test scoring practices on the grounds that most wrong-answer selections are thoughtful not random, presenting research supporting this proposition. An alternative test scoring system is presented, described and its outcomes discussed. This new scoring system increases the number of variables considered,…

The theory of score validity has undergone several revisions within the measurement community. The current consensus among professionals is a rejection of the trinitarian doctrine (J. P. Guion, 1980) of score validity and the recognition of a unified view that includes social consequences of test interpretation and use. While some aspects of the…

Both the attenuation paradox of traditional test theory and the assumption of local independence in person-item response theory have caused problems in interpretation. This paper demonstrates that the two are related concepts, and, through this demonstration, both are clarified. It is demonstrated that the breakdown of local independence leads to…

Presents a unified concept of test validity that integrates both the scientific and ethical considerations of test interpretation and use. Argues that the appropriateness, meaningfulness, and usefulness of score-based inferences are inseparable, and that this integration is based on construct validity. (FMW)

It is incorrect to say "the test is reliable" because reliability is a function not only of the test itself, but of many factors. The present paper explains how different factors affect classical reliability estimates such as test-retest, interrater, internal consistency, and equivalent forms coefficients. Furthermore, the limits of classical test…

Educational test theory consists of statistical and methodological tools to support inferences about examinees' knowledge, skills, and accomplishments. The evolution of test theory has been shaped by the nature of users' inferences which, until recently, have been framed almost exclusively in terms of trait and behavioral psychology. Progress in…

In almost all applications of measurement there is some sort of response by a human subject. Almost always, the response scale is ordinal, but almost always it is treated as if it were an interval measure. Methods for treating data ordinally are currently being developed in three areas: ordinal analysis for questionnaire responses, ordinal…

This paper points out that there are certain generalizations about directions for guessing and methods of scoring that require that data be derived from random groups design. It supports the viewpoint that it is neither sufficient nor appropriate to make such generalizations on the basis of an analysis of scores obtained from the answer sheets of…

A method is proposed for providing an absolute, in contrast to comparative, evaluation of how well two tests are equated by transforming their raw scores into a particular common scale. The method is direct, not requiring creation of a standard for comparison; expresses its results in scaled rather than raw scores, and allows examination of the…

An expanded framework for validating tests is needed to include the perspectives of teachers and students as well as of test makers and scientists. The development of educational assessments must take place within an understanding of how tests are used in context. (SLD)

A multi-faceted model for the selection of answers for multiple-choice tests was developed from the findings of a series of exploratory studies. This model implies that answer selection should be curvilinear. A series of models were tested for fit using the chi square procedure. Data were collected from 359 elementary school students ages 9-12.…

Current Scoring practices for multiple-choice tests are rooted in early Associationist Theory and are based on a two-step procedure: (1) right answers counted as ones and wrong answers are zeros, and (2) number of right answers form a total-correct score. The author contends that if either step is invalid, the use of the general linear model (GLM)…

A measurement approach called generalizability theory (G-theory) is an important alternative to the more familiar classical measurement theory that yields less useful coefficients such as alpha or the KR-20 coefficient. G-theory is a theory about the dependability of behavioral measurements that allows the simultaneous estimation of multiple…

To examine "bluffing"--ways in which conflicts in classrooms and evaluation procedures influence the styles of student writing and teachers' responses to different styles, a study analyzed the placement-test essays of 99 undergraduates entering Temple University (Pennsylvania) in the fall of 1982. Analysis of the texts was based on a…

Classical test theory, item response theory, and generalizability theory all treat the abilities to be measured as continuous variables, and the items of a test as independent probes of underlying continua. These models are well-suited to measuring the broad, diffuse traits of traditional differential psychology, but not for measuring the outcomes…