The well-known problem of cumulating error probabilities is reconsidered from a general epistemological perspective, namely, the concepts of severity and of fairness of tests. It is shown that not only Type 1 but also Type 2 errors can cumulate. A new adjustment strategy is proposed and applied. (Author/JAZ)

Provides a brief, nontechnical overview of latent trait models and argues for the preferability of these models (particularly the Rasch logistic model) over classical test models. Offers an example application of the Rasch model and discusses implications for the use of latent trait models for communication educators. (AEW)

Conjunctive item response models are introduced such that: (1) sufficient statistics for latent traits are not necessarily additive in item scores; (2) items are not necessarily locally independent; and (3) existing compensatory (additive) item response models including the binomial, Rasch, logistic, and general locally independent model are…

Explored utility of disguise in structured self-report assessment of psychopathology, using university students. Data indicated that under normal test-taking circumstances, use of disguised test items was not advantageous. This relationship was moderated by several dimensional parameters. Results supported rational strategy of test construction,…

Discusses validation studies of three British English language proficiency tests--the English Proficiency Test Battery, the English Language Battery, and the English Language Testing Service. Concludes that valid language tests depend on test constructors' knowledge of language and on their judgment as to the parameters of language proficiency.…

In an attempt to restore the symmetry and balance between the study of person and item variation, this paper presents a novel methodology construct specification equations, which allows one to ascertain from the lawful behavior of items what an instrument is measuring. (Author/PN)

Estimation techniques for generating the covariance matrix for two new tests and an existing test without the necessity of any examinee having to take two complete tests is presented. An application of these techniques to linear, observed-score, test equating is presented. (Author/JKS)

Binomial models for domain-referenced testing are compared, emphasizing the assumptions underlying the beta-binomial model. Advantages and disadvantages are discussed. A proposed item sampling model is presented which takes the effect of guessing into account. (Author/CM)

To study sex differences in test performance, the performance of males and females on 16 General Certificate of Education exams was analyzed in England. Results show that males perform better on objective tests than females. (Author/JJD)

Reanalyzes data provided by Swanson (1976) and Straton and Catts (1980) to test claims of superiority for the three-alternative multiple-choice item test and to present possible oversights made by these researchers. Results suggest it is doubtful that three-alternative test items are better than four-alternative items. (PAS)

This review of issues about correcting for attenuation concludes that the basic difficulty lies in being able to identify and equate sources of variance in estimates of validity and reliability. Recommendations are proposed for cautious use of correction for attenuation. (Author/CM)

Critiques the Jackson Vocational Interest Survey. Found the randomized technique of matching statements into forced-choice item pairs is a major flaw in the instrument. Explores conceptual issues arising when item clustering is used with forced-choice statement matching. Introduces contingency validity in item construction. (R)

Errors in self-reports of three academic performance measures are analyzed. Empirical errors are shown to depart radically from both no-error and random-error assumptions. Self-reports by females depart farther from the no-error and random-error models for all three performance measures. (Author/BW)

Cliff's Index "c" derived from an item dominance matrix is utilized in a clustering approach, termed extracting Reliable Guttman Orders (ERGO), to isolate Guttman-type item hierarchies. A comparison of factor analysis to the ERGO is made on social distance data involving multiple ethnic groups. (Author/BW)

The paper considers the problem of selecting the t best of k normal populations and simultaneously determining whether the selected populations have a mean larger than a known standard. Illustrations are given for selecting the t best of k examinees when the binomial error model applies. (Author)