variation in pronunciation, intonation and accent, following the lecturer’s speed,
etc., is one of the important mini-lecture comprehension task demands. Verbal
protocols have also proved that unsuccessful decoding of words leads to failure in
lexical selection and meaning construction. Therefore, before choosing details for
slots in the gap-filling task, the test-designer must evaluatefirst, whether the details
targeted help support main points and second, whether decoding of these details
would entail discrepancy in test-takers’ performance. For example, the word
“huskiness”in TEM 8 2010 is a suitable example showing test-takers’different
decoding results such as“huspiness”, etc. which reflect phonological distinctions.
Some details gaps might also involve meaning construction process. Details gaps
might also aim at meaning building at the word level or lexical chunk level.
Third, key-point gaps should target higher level of cognitive processing. Based
on the currentfinding, the key point gaps do not target higher level of processing
because the gaps only target one or two content words neighbored by words directly
taken verbatim from the audio input. Those neighboring words actually provide
test-takers with clues. If test designers want to improve the difficulty level of
key-point gaps, there should be the relevant information larger than a lexical chunk.
That is to say, test-takers need to really comprehend the relevant part of information,
at least a group of sentences that accommodate supporting details, tofill in the gap.
Fourth, inference gaps should aim to elicit test-takers’inferences of the relevant
information so the words to befilled in the inference gaps might not be the verbatim
words from the recording. Self-evidently, this type of gaps should be calibrated
toward higher cognitive processes and considered cognitively demanding.
Comprehension of the relevant information paves the way to accurately answering
inference items. According to verbal protocols in the current project, test-takers
often infer answers assisted with their background knowledge. Inferences based on
common knowledge are considered construct-relevant only on the premise that the
background knowledge needed should be shared by all test-takers, otherwise
content bias would emerge.
Fifth, summary gaps are calibrated toward a block of information, i.e., a group of
sentences, instead of discrete lexical chunks or idea units. The presupposed target of
this gap type is to attain the test-takers’discourse construction process. Based on
verbal protocols, this type of gaps is very difficult to test-takers and also poses high
cognitive demand. To test-designers, the difficulty to design this type of gaps
mainly lies in three aspects: 1. how to select a sequence of idea units for a summary
gap; 2. how to determine the summary word/words generated from the sequence of
idea units; 3. whether to let the summary word/words appear in the recording.
Take the current project for example. Let’s look at the gap from 2010 TEM 8
Mini-lecture and Gap-filling Task: Proximity is person-, culture- and
_____-specific. The relevant sequence of idea units in the script is:“Once again, I’d
like to say, proximity is also both a matter of personal style and it’s also
culture-bound. So what may seem normal to a speaker from one culture may appear
unnecessarily close or distant to a speaker from another and standing close to
someone may be quite appropriate in somesituationssuch as an informal party but
completely out of place in othersituationssuch as a meeting with a superior.”If
9.4 Implications for Assessing Academic Lecture Comprehension 161