emphasized that listening test developers should transform script-based item writing
to real listening-based item writing. If testers locate key points for test items only
based on reading the script, those key points for each item might differ from those
drawn by test takers when they are listening to the audiofile. Justice and fair play is
the gist of modern test and surely it is test developers’responsibility to follow those
ground rules.
Furthermore, the summary cloze test type is one of the research foci in the
current project, so implications centering around test targets or gap types are dis-
cussed as follows.
First, all the test-items should be material-bound and passage-dependent. The
number of items and distribution of items are test-designers’initial considerations.
Testers are supposed to avoid designing items that generate answers via the hints in
the given task or only via test-takers’background knowledge. If so, the construct of
the listening test is changed. In order to guarantee“material-boundedness”of test
items, we, testers need to examine carefully what idea units in the discourse ulti-
mately form listeners’discourse representation and to what extent lexical and
syntactic features of the input discourse can accommodate the “density and
embededness”of idea units (Field 2013: 122). Field (2013) has proposed that idea
units form discourse representation listeners build up while listening to the audio
input and the potential match to a test item. Afterward, to test-designers, it is
important to analyze the quantity of idea units as well as the distribution of them.
For the test type of a gap-filling task on the structured summary of a mini-lecture,
test-designers need tofirst make some initial decisions such as how many gaps
should be inserted in the structured summary and how the gaps should be dis-
tributed over the structured summary. Meanwhile, we have already discussed that
information is stored in the brain as idea units instead of individual words and
phonemes, and there is always a limit of people’s working memory. Therefore,
test-designers must be careful if they want to make test items on longer complex
sentences that contain a number of idea units. To test-designers, maybe most of the
idea units are important in building up discourse representation; however, from the
listener’s perspective, parsing those sentences and memorizing relevant idea units
are cognitively demanding. In the current project, participants have reported the
conflict between comprehension and note-taking. Even high achievers feel it is hard
to comprehend complex sentences that accommodate multiple idea units and
simultaneously write them down or memorize them for later retrieval. In this case,
though, in academic listening, efficiency of working memory determines L2 lis-
teners’information processing speed and their comprehension quality, I would like
to argue that writing items on adjacent information units is not advisable, simply
because test-takers can hardly parse multiple idea units at a time.
Second, details gaps should target lower level of cognitive processing.
Test-takers’lower–level processes such as decoding, lexical search, parsing are
targeted by details gaps. To design details items, a crucial point is that these items
should aim to elicit key detailed information that supports main points. In the
current project, according to the teachers’as well as students’perception based on
their respective questionnaire data, input decoding, such as decoding the acoustic
160 9 Conclusion and Recommendations