Tekstisisesed seosed eesti keele kui teise keele lugemistestis: mõõtmisvahendi valiidsus
Testing understanding text cohesion and coherence in Estonian as a second language: The validity of different task types

Summary/Abstract: The article discusses a study that was conducted to establish whether gapped text tasks are more valid measures of learners’ ability to identify and use cohesive devices and coherence relations in texts than the paragraph sequencing task. Two tasks were created based on the same narrative text that had previously been used in Estonian as a second language B2-level examination as a paragraph sequencing task. #is task was analysed by Tiina Kikerpill (2010) and several drawbacks were pointed out. One of the tasks created was a gapped sentence task and the other a gapped paragraph task. Both tasks were administered to 17 first-year students of the University of Tartu whose first language is Russian and who had all successfully passsed the school-leaving examination in Estonian as a second language and whose Estonian, therefore, is expected to be at least at level B2. Basic statistical data were calculated. Four of the students in each group were asked to report verbally (both concurrently and retrospectively) on their task completion. Their reports were analysed. The statistical analyses showed that the gapped paragraph task was the easiest of the three (76.5% of the informants completed it without any mistakes). In order to complete the tasks, the informants made use of such co-reference devices as repeated forms and pronominal forms as well as bridging and elaborative inferences. However, in the gapped paragraph task the lexical repeated forms were used more often than in the gapped sentence task, where inferencing dominated. This was most likely due to the fact that coherence in these two tasks was indeed dependent on di&erent devices. The study demonstrated that the informants focused on the text that preceeded the gap and only rarely checked that the text after the gap followed smoothly. This may of course be due to the linear nature of reading or due to the nature of narratives, but it would be worth investigating whether it is possible to design such tasks where test-takers are forced to pay attention to the text after the gap. Another aspect worth investigating is whether a different text type (such as expository, for instance) would allow to measure the knowledge and use of coherence relations other than the ones used in the current study.

• Details
• Contents