The SLPT is adaptive. Each answer, whether correct or incorrect, affects the progression of the test, so that the respondent’s proficiency level is determined quickly and efficiently.
The platform can also operate in non-adaptive mode, so that items can be tested and calibrated.
The administration of an adaptive test is non-linear. The way in which the learner answers the first questions, whether correctly or incorrectly, determines the progression of the test. As the test is administered, it dynamically selects, from a bank of approved and calibrated items, the item most likely to assess the respondent’s skill level in light of his or her preceding answers. The answers are analyzed as the respondent gives them, and tasks are selected progressively according to the answers; the assessment of the respondent’s proficiency level is thus refined quickly and efficiently. The test ends when the proficiency level is determined. An adaptive test, in comparison with a non-adaptive test, makes it possible to reduce the number of items presented, decrease the test administration time and increase the reliability of skill assessment.
The results for the adaptive test vary between -3 and +3. For purposes of interpretation, these results (called “thetas”) are matched with the language proficiency scale levels. In the Cut scores section, the proficiency levels are identified and the thetas corresponding to each level are defined.
The test is generally used in non-adaptive mode so that a representative sample of respondents can go through a representative number of approved items (say, about a hundred). The items can then be calibrated, i.e., a difficulty value on a scale of -3 to +3 can be obtained. Once the test includes about a hundred calibrated items, it can be used in adaptive mode. An automatic calibration mechanism for new items is included in adaptive mode: each respondent is given a certain number of non-calibrated items in each category (for example, two per category), in addition to the items chosen to measure his or her language skill. The results for these items will not be used in the assessment of skill. Once a certain number of respondents (say, 150) have answered a given non-calibrated item, the administrator will calibrate the item, by clicking on the Calibrate button in the Calibration section.
The administration of the non-adaptive test is linear, with a predetermined number of questions to which the learner must respond. The test ends when the bank of questions is used up.
The test in non-adaptive mode randomly presents all items created and approved in the editing module. The non-adaptive mode is used essentially to calibrate items, i.e., to determine their level of difficulty by having a representative sample of respondents take the test. In this mode, the respondent’s results are not calculated in thetas, but according to the number of correct answers out of the number of items presented.