51918 skanowanie0034 (11)

6.4. The construction of tests

In this part generał procedures for test construction will be briefly laid down.

6.4.1.    Statement ofthe problem

The essential first step in testing is to make oneself perfectly elear about what it is we want to know and for what purpose. The following ąuestions have to be answered:

W What is the purpose of the test, i.e. what kind of test is it to be? (achievement, proficieney, diagnostic, or placement?)

—    What sort of learner will be taking the test?

—    What abilities/skills are to be tested?

—    How detailed must the results be?

6.4.2.    A solution ofthe problem

Once the problem is elear, then steps can be taken to solve it. The first form the solution takes is a set of specifications for the test. This will include information on: content, format and timing, criteria levels of performance, and scoring procedures.

Content

The fuller the information on content, the less important-should be the subseąuent decisions as to what to include in the writing of any .version of the test. The way in which content is described will vary with its naturę. The content of a grammar test, for example, may simply list all the relevant structures. The content of a test of a language skill, on the other hand, may be specified along a number of dimensions.

The test constructors first make an inventory of the discrete points and areas they want to test. This involves distinguishing broad objectives from morę specific ones. Test items are then developed for each objective.

Format and timing

This should specify test structure (including time allocated to components) and item types procedures, with examples. It should State what weighing is to be assigned to each component. It should also say how many passages would normally be presented (in the case of reading and listening) or reąuired (in the case of writing), how many items there will be in each component.

One basie issue of format is whether the test progresses to increasingly morę difficult items or whether easy and difficult items are interspersed. There are arguments on both sides. If items get increasingly morę difficult, the respondents may give up after a while and not attempt items after the first one they get stuck. Yet if respondents experience failure too freąuently at the outset of a test because of difficult items, they may be discouraged from attempting the remainder of the items in a section. Thus, there may be a psychological advantage to pacing the items so that they become progressively morę difficult. A compromise is to start the test with relatively easy items and then to start interspersing easy and difficult items.

Another issue of format relates to multiple-choice items. Such items lend themselves to guessing. Increasing the number of alternatives (from say, 3 to 4) decreases the likelihood of getting the item right by chance alone. There is a 33 % chance of getting a three-choice item right by guessing, and a 25 % chance of guessing correctly on a four-choice item. This of course assumes that all choices are eąually attractive to the respondent who does not know the answer to the item. This condition is not always met.

Instructions

. The instructions should be brief and yet explicit and unambiguous. The respondents should be informed as to whether guessing counts against them. They should also know the value of nich item and section of the test. Finally, the time allowed for each Kubtest and for the total test should be announced. If speed is a fuctor for a subtest, the respondents should be aware of this.

69