Receptive and productive vocabulary learning through narrow reading and explicit instruction

In this multiple case study, I examined how Japanese EFL students developed their L2 receptive and productive vocabulary knowledge through narrow reading and two Generative Learning Theory tasks, sentence generation and summary writing. The eight participants used two of Wittrock’s (1974) Generative Learning Theory principles to increase vocabulary learning gains: (a) the generation of associations between the input and students’ knowledge and previous experiences, and (b) the generation of associations between different parts of the input. The counter-balanced design included eight adult participants with written receptive vocabulary sizes of 4,000–5,000 word families, who participated in four treatments. The target words were pseudowords whose original words were at the 4,000- and 5,000-word frequency levels in the BNC/COCA corpus. The vocabulary tasks were controlled for the number of encounters with target words in the narrow reading input and the number of retrievals of target words in the generative output tasks. The outcomes of the treatments were assessed using scoring rubrics for target-word knowledge in spelling, meaning, and part of speech. The participants’ contextual use of target words in three writing tasks were examined with (a) scoring rubrics for the correct use of the target word, including spelling, meaning, and part of speech, (b) Natural Language Processing (NLP) tools, and (c) a modified version of Joe’s (1995, 1998) Generativeness Scale. Two post-treatment semi-structured interviews, including Likert-scale questionnaire items and open-ended interview questions, were used to elicit the participants’ views of the vocabulary tasks. The research hypotheses were as follows. For research hypothesis 1, I examined the vocabulary learning gains of the four treatments, focusing on both total posttest scores and scores for each of the three aspects of target words: meaning, spelling, and part of speech. Research hypothesis 2 concerned contextual use of target words during the treatments focused on three writing tasks: higher-order thinking questions, sentence generation, and summary writing. In research hypothesis 2a, I examined the correct use of target word knowledge in contexts, focusing on the same three aspects of target words as in research hypothesis 1. Research hypothesis 2b concerned the complexity of target word use in contexts. In research hypothesis 3, I presented the participants’ responses to two Interview Guides, the first part on the results of the Likert-scale questionnaire and the second part on analysis of the interview transcripts. This study had three main findings. The efficacy of the treatments, based on the immediate and delayed posttest results, was sentence generation > narrow reading only > summary writing > comparison condition. For each of the three aspects of target word knowledge on the posttests, sentence generation showed the following results: meaning > part of speech > spelling. The other three treatments showed the same order: part of speech > meaning > spelling. For research hypothesis 2a, the order of correct use of target word knowledge in response to higher-order thinking questions was sentence generation > summary writing > narrow reading only > the comparison condition. When the two generative tasks were compared individually, sentence generation > summary writing. There were minor inaccuracies in spelling in narrow reading only, part of speech in sentence generation, meaning in higher-order thinking questions and summary writing, and to a larger degree in the comparison condition. For research hypothesis 2b, the comparison condition showed larger numerical values than narrow reading only for text length, average sentence length, text coverage with 2,000 high-frequency words and K3+ words, and the proportion of range 1 words. Narrow reading only had a higher proportion of Range 2 words. Including treatments with an additional generative writing task, both of which resulted in longer text, showed that summary writing produced the longest text among the four treatments. Sentence generation had the shortest average sentence length, while summary writing had the longest. Compared to the comparison condition and narrow reading only, the two generative tasks had lower coverage of 2,000 high-frequency words but higher coverage of K3+ words. Summary writing had the lowest text coverage with 2,000 high-frequency words but the highest with K3+ words. The proportion of range 1 words was sentence generation > comparison condition > narrow reading only > summary writing. The proportion of range 2 words was summary writing > sentence generation > narrow reading only > comparison condition. Regarding the total mean scores for levels of generativeness, the treatment order was sentence generation > narrow reading only > summary writing > comparison condition. For research hypothesis 3, the questionnaire results showed that the participants found sentence generation and summary writing effective for L2 productive vocabulary learning, but they preferred narrow reading and reading comprehension questions. Regarding task effectiveness for L2 vocabulary learning, the participants considered narrow reading more effective than the comparison condition. Seeing the target words in similar contexts was sometimes helpful for guessing their meaning, but these contexts did not provide a variety of clues. Reading different topics was considered effective because the participants could check their guesses in different contexts. The participants said that factual questions were ineffective for vocabulary learning because they only required extracting one sentence from the reading text with minor modifications, and because they could answer them without understanding target words. Conversely, higher-order thinking questions were considered more effective because many participants felt they needed to understand the target words to answer them. The participants were divided as to which generative task they considered more effective, sentence generation or summary writing. The vocabulary checklist with L1 translations during sentence generation enabled participants to complete the task with 100% understanding of the target words, and they focused on the target words when writing two related sentences. On the other hand, the participants sometimes completed summary writing without understanding the meanings of the target words. The participants were overloaded during summary writing because they had to pay attention to the content, organization, and placement of target words.