Establishing Cutting Scores for Army Language Proficiency Tests
Abstract
Based on two prototypes, 33 additional language proficiency tests were constructed. An attempt was made to adhere to prototype item content and composition to a sufficient extent to insure relative comparability in validity and difficulty for all of the new tests. The present Research Memorandum reports the results of the validation and statistical analysis undertaken for three tests not previously covered -- French, German, and Polish. Validity coefficients obtained were of sufficient magnitude to indicate that the three tests are highly efficient measures of language proficiency. Coefficients for these tests (.66 to .87) closely approximated those for the Chinese-Mandarin and Russian prototypes (.68 to .86). It was concluded that the tests were fairly comparable with respect to validity. Levels of fluency were designated 'Good', 'Fair', and 'Poor' and cutting points were computed at these levels by use of equal percentiles on the criterion and on the predictor measures. A common set of cutting scores was established for both parts of the tests and for all three languages. Amounts of misclassification estimated to result from this procedure varied from 0% to 13% of the total cases. It was also concluded that validity and comparability assumptions had been met and that common cutting scores could be generalized to all tests of foreign language proficiency.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1961
- Accession Number
- ADA079214
Entities
People
- J. J. Mellinger
- K. Cook
- Rudolph G. Berkhouse
Organizations
- Adjutant General's Corps