Open AccessTHE TOEFL COMPUTERIZED PLACEMENT TEST: ADAPTIVE CONVENTIONAL MEASUREMENT
Author(s) -
Hicks Marilyn M.
Publication year - 1989
Publication title -
ets research report series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.235
H-Index - 5
ISSN - 2330-8516
DOI - 10.1002/j.2330-8516.1989.tb00338.x
Subject(s) - computerized adaptive testing , equating , test of english as a foreign language , test (biology) , computer science , test score , backtracking , scale (ratio) , statistics , algorithm , data mining , mathematics , psychometrics , standardized test , rasch model , language assessment , paleontology , physics , mathematics education , quantum mechanics , biology
This study focused on methods of computerized adaptive testing using conventional scoring methods in order to develop a TOEFL computerized placement test. Some advantages of adaptive conventional measurement were illustrated, such as providing the user with an alternative, efficient test with most of the same properties of the conventional test. Test specifications and structure very closely parallel the full‐length test, yielding scores on the (familiar) reported score scale, thus providing the user with comparable interpretation within the limits of differences in score precision. As a consequence of simulation studies conducted during the first phase of this study, the multilevel testing paradigm was adopted for this development. Its implementation produced three test levels varying in difficulty, each approximately half the length of the regular TOEFL. The basic testing algorithm routed examinees through item blocks or testlets that permitted backtracking in order to review answers and change them, an option not easily implemented in standard computer adaptive testing. The ability to control important facets of test construction, as well the degree of measurement effectiveness, using this testing method was illustrated. Resulting test levels were equated to the established scale via IRT true score equating, and some desirable properties of the obtained score scales were described, namely, overlapping scales at the boundaries of the test levels and limits on the scores obtainable on each of the levels. Data from a preliminary validation study were presented that indicated the test was functioning satisfactorily. Responses to a questionnaire administered to the sample of examinees revealed that 72% of the group had never been exposed to a computer before this testing experience. Nonetheless, 59% indicated that they preferred the computerized test to the paper‐and‐pencil version or were neutral. The apparent lack of computer experience among ESL students will need to be considered in any computerized test development for this group of examinees.