Bottom-up or top-down processing as a discriminator of L2 listening performance

Studies in reading comprehension and listening comprehension have investigated the processing skills used by skilled and less-skilled readers/listeners. Some of these studies observe that skilled readers/listeners are those who are better able to engage in top-down processing whereas others maintain that they are better able to engage in bottom-up processing. This study investigates the kind of processing skill that is more important in discriminating the performance of L2 learners on listening test items in large-scale public examinations in Hong Kong over a period of seven years. Two variables were investigated: the schema type of the aural text and the question type. Two types of schema were identified: the first type is referred to as the 'non-matching' schema type in which the schema activated by the initial linguistic input is not congruent with the subsequent linguistic input. Candidates need to be able to process the incoming linguistic cues rapidly and accurately, and to revise their developing schema accordingly in order to get the correct answer. The second type is referred to as the 'matching' schema type in which the schema activated by the initial linguistic input is congruent with the subsequent linguistic input. Hence candidates could rely on top-down processing to get the correct answer. Two types of questions were also identified. The first type is 'global' questions which required candidates to understand the text as a whole and to draw conclusions or inferences. Hence, candidates need to process all the key linguistic cues rapidly and accurately in order to get the correct answer. The second type is 'local' questions which only required candidates to pick out specific details. Hence, even if candidates failed to process some linguistic cues, they would still be able to get the correct answer. The mean criterion scores for the correct answers, which indicate the performance of the candidates who chose these answers in the entire paper, were compared for schema type and question type using a series of paired t-tests. The results showed that 'non-matching' items yield significantly higher mean criterion scores than 'matching' items irrespective of question type, suggesting that bottom-up processing was more important than top-down processing in discriminating the listening performance of L2 learners on test items.link_to_subscribed_fulltex