Session 2C

Classroom and out-of-classroom vocabulary learning: Are they related? Out-of-school exposure to English through TV, cinema, songs, computer games, and social media has become an important source of comprehensible input that can contribute to incidental learning of the language in general and vocabulary in particular (Nunan & Richards, 2015; Sundqvist & Sylvén, 2016)). Research shows that vocabulary can be acquired from audiovisual input (Puimège & Peters, 2019), from songs (Pavia, Webb & Faez, to appear) and from digital gaming (Sundqvist, 2019). To our knowledge, studies have not focused on individual vocabularies acquired by different learners through different language activities. Furthermore, it is not clear whether there is a relationship between how much has been learnt in school and how much can be acquired out of school. The present longitudinal study investigates individual out-of-classroom word learning from digital activities and the effect of classroom vocabulary knowledge on out-of-classroom learning success. We asked how many words (single words and multiple word units) were acquired from digital activities during one school year, and how the gains were affected by classroom vocabulary knowledge at the onset of the study. Fifty eight EFL 10th graders were pre-tested on classroom vocabulary by a self-designed meaning recall test of word samples from the national lexical syllabus. Based on the results, they were divided into three proficiency groups. During the school year, learners submitted weekly diaries where they recorded all the new words they encountered in digital activities. Based on the diaries, we constructed individual word lists for each learner. At the end of the study, each learner took a meaning recall test of his/her individual out-of-school vocabulary. The three proficiency groups were compared on word gains by ANOVA. In addition, initial vocabulary knowledge was correlated with out-of-school gains. We found that almost every student learned new words from digital activities, but vocabulary knowledge at the onset of the study exhibited a clear effect on gain patterns. Initial knowledge and out-of-school gains correlated moderately, and the three proficiency groups were significantly different from each other in how much they learned. We relate the results to incidental learning, classroom form-focused instruction, and to prerequisites for out-of-school learning. References Nunan, D, & Richards, J. (Eds.) (2015) Language Learning Beyond the Classroom. Taylor & Francis Sundqvist, P. (2019). Commercial-off-the-shelf games in the digital wild and L2 learner vocabulary. Language Learning & Technology, 23(1), 87- 113. Sundqvist, P., & Sylvén, L. K. (2016): Extramural English in Teaching and Learning: From Theory and Research to Practice. London: Palgrave Macmillan. Pavia, N., Webb, S. & Faez, F. (to appear). Incidental vocabulary learning from listening to L2 songs. Studies in Second Language Acquisition Puimège, E., & Peters, E. (2019). Learning L2 vocabulary from audiovisual input: an exploratory study into incidental learning of single words and formulaic sequences. Language Learning Journal, 47 (4), 424-438.

Research has indicated social benefits of drawing on students’ full language repertoires (Cummins, 2017), and lab-based studies have shown that background languages cannot be “turned off” during target language processing (Sunderman and Kroll, 2006). Furthermore, vocabulary experts recommend using first languages (L1s) to establish initial form-meaning mapping for new foreign/second language (L2) words (Schmitt, 2008). However, classroom-based research on which teaching/learning practices may be effective in multilingual classrooms is scarce. Studies on bilingual classrooms exist, and Laufer and Schmueli (1997) found that high school learners of English did better on a vocabulary test following a short learning-session when the meaning of English target words were in the L1 compared to L2 English glossings. In fact, Shin, Dixon and Choi (2019) report that intervention studies published in the last decade suggest that L2 learners bene?t more from L1 translations than from L2-only explanations. Yet, to the best of our knowledge, there are no studies comparing conditions involving more than two languages. In our talk, we present results from an intervention study comparing the effects on vocabulary learning of three week-long teaching/learning conditions in multilingual L2 English classrooms in Sweden. Multilingual classrooms were defined as classrooms where ? 5 learners have another L1 than the majority language (Swedish). The three learning conditions were: English Only (EO); English and Swedish (E&S); English, Swedish and any Other language(s) (E&S&O) known by the learners. Learners were instructed to follow the imposed condition that applied each week. Participants were learners (age 15-16) in six intact classes. At present, results have been analysed for three classes (N = 62), but in our presentation, results for all six participating classes (N = c. 120) will be presented. Teaching materials for each of the three treatment conditions comprised (1) a text, including 20 controlled and carefully piloted English target words in context, (2) vocabulary exercises, and (3) vocabulary lists covering the target words, the latter being either EO, E&S, or E&S&O. We used a counter-balanced, repeated-measures design, featuring a proficiency test – a pretest (60 words) – 3 x treatment – an immediate posttest (3 x 20 words) – a delayed posttest (60 words). The vocabulary test format targeted meaning recall (expressed in any of the participants’ languages or even drawings). ANOVAs of gain scores comparing pretests with immediate and delayed posttests showed that all three classes performed the best in the E&S condition, irrespective which week this treatment/condition took place. Thus, the condition in which English vocabulary was presented with Swedish translation equivalents (E&S) yielded higher gain scores. Also, strong correlations (at .70 - .80) between English proficiency scores and gain scores were observed, indicating that gain scores increased as a function of proficiency. In our presentation, in addition to gain scores, the potential influence of factors such as age of onset of acquisition, language dominance, preferred learning approach and perceived learning will be incorporated as covariates in the analysis. References Cummins, J. (2017). Flerspråkiga elever: Effektiv undervisning i en utmanande tid [Multilingual students: Effective teaching in challenging times]. Stockholm: Natur och kultur. Laufer, B., & Schmueli, K. (1997). Memorizing new words – Does teaching have anything to do with it? RELC Journal, 28(1), 89–108. Schmitt, N. (2008). Review article: Instructed Second Vocabulary Learning. Language Teaching Research, 12, 329–363. Shin, J., Dixon, L. Q., & Choi, Y. (2019). An updated review on use of L1 in foreign language classrooms. Journal of Multilingual and Multicultural Development. First View. Sunderman, G., & Kroll, J. F. (2006). First language activation during second language lexical processing: An investigation of lexical form, meaning, and grammatical class. Studies in Second Language Acquisition, 28(3), 387–422.

In SLA studies, it is important for researchers to document learners' proficiency in experiments aiming to explain the linguistic knowledge or behavior of these learners. In many cases, in lieu of proper general proficiency test suites, researchers use a vocabulary test as a quick-and-dirty proxy for general language proficiency, as strong correlations have been found between vocabulary scores and scores from general language proficiency tests (Alderson, 2005). These vocabulary tests - either levels or size tests - must in themselves be valid measures of the intended construct. A central issue in this regard is item sampling, as the words for these tests are sampled from 1,000-word frequency bands. These samples are intended to be representative of the 1,000 word populations from which they are sampled. Existing vocabulary size tests feature between five and 200 item samples per 1,000-word band, and the only evidence of the representativeness of the number of items used is that they were randomly selected. A further concern is that if sample-yielded vocabulary knowledge estimates differ greatly from the target construct of how many words out of the target 1,000-words are known by a learner, then this also questions the construct validity of the data collected. Further, if for example two sets of 30 items completed by the same learner yielded vocabulary knowledge estimates that differ greatly, this will question the reliability of data that can be yielded from 30 items. This is because reliability is the consistency of measurement (Bachman & Palmer, 1996), and the 30 items used to represent the 1000 item target construct is expected to be as representative as another 30 items sample. In a study by Gyllstad, Vilkait? & Schmitt (2015), the authors investigated what is an appropriate sampling ratio to yield an accurate estimate of a 1,000-word band. They demonstrated that a sample ratio of 10:1000 failed to yield an accurate estimate, and that a ratio of 30:1000 provided less inaccurate estimates. However, their study compared test-takers’ meaning-recognition scores from 10-30 item samples with meaning-recall scores from a 100-word sample, and not the full 1,000-words from which samples were taken. A further limitation of their study is that it used only a single set of 10, 15, 20, 25 and 30 item samples when investigating the representativeness and accuracy of estimates yielded form these samples sizes. In this presentation, we discuss and illustrate the effect that an increase of sample size has on confidence intervals of individual learner vocabulary knowledge estimates, i.e., the accuracy of vocabulary knowledge estimates, and so the safety with which inferences can be made from test scores. Japanese university students (N=103) of a range of English proficiencies completed two English vocabulary tests. Both tests, one meaning-recall, and another meaning-recognition, measured knowledge of the same 1,000 words from the 3rd 1,000-word frequency band. Participants’ scores were the higher on the meaning-recognition format (M=785.05, SD=127.96, ?=.994), dropping on the meaning-recall (M=423.50, SD=189.89, ?=.997). Using a powerful bootstrapping approach, and drawing on the unique data set at hand, responses for the 1,000 items for each test format were repeatedly sampled to create thousands of simulated tests of different item lengths (5-200). Results show that the fewer items used in the test, the wider the distribution of size estimates, with 20-item tests giving a range of 595-955 words, 50-item tests 664-886 words, and 100-item tests 696-854 words. The results imply that SLA researchers using vocabulary tests as proxies for general proficiency tests must to a greater extent be aware of these issues, and carefully decide how much trust to put in scores from various test lengths for their own particular purposes.

Vocabulary knowledge is a complex construct that involves mastery of several aspects, such as form, word parts, collocation and meaning (Nation, 2013). The majority of vocabulary studies have focused on measuring form-meaning mapping (Milton & Fitzpatrick, 2014), and thus, our understanding of the acquisition of other word knowledge aspects is still limited. While recent years have experienced a growing interest in the study of collocation (Boers & Webb, 2018) and word parts (Mizumoto et al., 2019), there are relatively few studies that have examined the acquisition of multiple meanings and how the different meanings that a word conveys affect learning. This is surprising because the majority of words have more than one meaning (Hoshino, 2018) and one of the challenges of lexical development is learning the different meanings that words convey. Indeed, the limited available research suggests that L2 learners struggle with learning the different meanings of words (Crossley et al., 2010; Schmitt, 1998), and that it is one of the most difficult aspects of knowledge to be learnt (González-Fernández & Schmitt, 2019; Wolter, 2009). Moreover, while there are some studies (Hoshino & Shimizu, 2018; Verspoor & Lowie, 2003) that have focused on the study of polysemy (i.e., various related meanings derived from the same origin), it is still unclear how homonymous (i.e., multiple unrelated meanings derived from different origins) meanings of words are learnt by L2 learners compared to polysemous meanings. In addition, it is unclear whether the acquisition of secondary meanings of familiar words differs from the acquisition of main meanings of unfamiliar words.
The present study addresses these gaps by comparing the deliberate learning of homonymous and polysemous meanings. Following an experimental design, 66 Spanish-speaking EFL learners were divided into a control group (n = 30) and an experimental group (n = 36). The participants in the experimental group were taught, via flashcards, new, secondary meanings of 10 polysemous (e.g., chair) and 10 homonymous (e.g., toast) words (controlled for length, PoS, cognateness, and frequency) for which they already knew the most common meaning (e.g., 'a seat for one person' for chair, and 'bread that has been heated to make it brown' for toast), as evidenced by pre-tests. 10 primary meanings of unknown words (e.g., rake as 'a garden tool' and flesh as 'the soft tissue of the body') with the same variable manipulation were also included for comparison. Meaning recall and meaning recognition knowledge was tested immediately after the treatment and again one week later. Mixed-effect modelling analyses were used to compare the effect of word type (polysemous, homonymous, and primary meaning), as well as to explore the effects of participants' proficiency level, and intralexical word factors on learning. The results indicate that there were similarities in the acquisition of polysemous and homonymous meanings, as well as between these types of multi-meaning words in relation to learning the primary meanings of the unknown words. Moreover, the experimental group demonstrated major gains in both polysemous and homonymous meanings, although performance varied over time. Finally, the findings suggest that the etymological origin of the various meanings of words might not be the main determinant factor in the acquisition of multiple meanings. The implications of the findings will be discussed in detail.