The effects of distribution, difficulty, and quantity of digital flashcard practice on language learning
- Autores: Jonathan Serfaty
- Directores de la Tesis: Raquel Serrano Serrano (dir. tes.)
- Lectura: En la Universitat de Barcelona ( España ) en 2023
- Idioma: inglés
- Tribunal Calificador de la Tesis: Carmen Muñoz Lahoz (presid.), Amanda Edmonds (secret.), Yuichi Suzuki (voc.)
- Programa de doctorado: Programa de Doctorado en Ciencia Cognitiva y Lenguaje por la Universidad Autónoma de Barcelona; la Universidad de Barcelona y la Universidad Rovira i Virgili
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
According to research into second language (L2) practice, learners should repeatedly engage in output activities with feedback in order to develop accuracy. Digital flashcards can be used to prompt L2 output and deliver feedback when teacher instruction or peer interaction are not available. Traditionally used for vocabulary, this tool could also be used for grammar learning by providing exemplars of target structures. The difficulty, distribution, and quantity of practice have been shown to affect learning and retention in other domains, but little is known about how these variables affect productive L2 grammar practice. Research into these areas could deepen our understanding of L2 learning processes while providing useful guides for enhancing L2 practice. This thesis includes four papers. Chapter 1 is an introduction to the topic, covering the importance and theories of L2 practice, a primer on digital flashcards, and the research gaps to be addressed. Chapters 2-5 are research papers. The first paper, published in System as Serfaty and Serrano (2020), investigated how flashcards could be used for grammar learning in an environment where other forms of learning are unavailable. Simple structures were studied by 31 low-proficiency learners, aged 9-17, in a rural setting in Cambodia over two weeks. They were tested after one day, two weeks, and four months. Results showed that participants made large gains in their grammatical accuracy and maintained these gains over time. Scores were equivalent for trained and untrained items, showing that the exemplars used in training provided rules that were generalized to novel sentences. The second paper, published in Applied Psycholinguistics as Serfaty and Serrano (2022), investigated how this type of learning might be affected by the distribution of sessions. Two complex structures were studied at intervals of one day or one week, tested after one week or one month. Participants (N = 117) came from an international school in Phnom Penh, aged 10-18. The optimal lag was predicted by individual differences. Participants with slower times and lower proficiency obtained higher scores from the shorter lag, whereas faster and more proficient learners benefited from a longer lag. Neither lag was better overall. The third paper, currently in review in Language Learning and Technology, repeated this methodology, using the same intervals, tools, and setting, with vocabulary items. Of the 96 participants analyzed in this study, 77 were also in the grammar analysis. This allowed for a comparison between grammar and vocabulary lag effects. This paper also aimed to ascertain whether a lag effect is found outside of lab conditions with secondary school students, which has not been found previously, and to explore whether lag effects are different for productive (form-recall) and receptive (meaning-recall) knowledge of learned words. Results showed a small but consistent advantage to the longer lag at both testing times, in contrast to grammar. The longer lag was particularly effective for retaining receptive knowledge at the 28-day posttest. The final paper, currently resubmitted to Language Learning after revisions, aimed to find the optimal amount of practice for the long-term retention of grammar knowledge. An artificial language was learned and then practiced on either one, two, three, or four relearning sessions on consecutive days, with 30 participants per condition (N = 129), aged 18-30. At a two-week posttest, it was found that average scores were significantly higher after a third relearning session. Accuracy during training peaked after the second relearning session, leading to a hypothesis that a threshold of knowledge is crossed after performing two sessions without errors. When re-coding the participants accordingly, this threshold was a stronger predictor of high posttest scores than the total number of sessions. The final chapter summarizes the findings from these papers, details implications from these findings on future theory, research methods, and pedagogy, and ends with some suggested future directions
