Based on the identified themes, we developed a series of Likert scale and multiple-option items reflecting the most common reading and typing tasks reported on a cell phone (Table 1).
The verbal analogies test contained 41 four-option multiple choice items.
The most thorough review to date of empirical studies into the optimal number of options in multiple choice items in all domains was the meta-analysis carried out by Rodriguez ([2005]).
Overall therefore the findings of this study suggest that despite the considerable theoretical and research interest which has been devoted to the issue of the optimal number of options in multiple choice items, for the measurement of language structure and reading ability, the issue may be of minor importance in considering the performance of the multiple choice format.
This paper also reports on how those changes were affected by the number of options in the multiple choice items as well as comparing the effect on validity and performance noted among the multiple choice items with differing numbers of options, as against the effects noted between items in multiple choice and the constructed response, short answer format.
The study looked particularly at differences between the performance of stem-equivalent items in the constructed response, short answer format and multiple choice items with 3- 4- or 5-options.
Knowledge will be measured using seven multiple choice items, with two to four answer options.
The study also considered how the effectiveness of the distractors included in the multiple choice items was affected by the numbers of options in the items, as well as examining whether any effects noted differed between structure and reading items.
The researchers conclude that option number format is of little importance in the measurement of language ability compared with the much greater construct variation apparently induced by the use of multiple choice items as against constructed response formats.
Overall however it is notable that the analysis of change and consistency between the two tests produced three highly consistent data sets across the three different option number format groups, suggesting once again that the number of options in a multiple choice item is a matter of little significance in the measurement of language based constructs.
It contained both three and four option items.
