de Jong, T., Specht, M., & Koper, R. (2010). A Study of Contextualised Mobile Information Delivery for Language Learning. Educational Technology & Society, 13 (3), 110���125. 110 ISSN 1436-4522 (online) and 1176-3647 (print). �� International Forum of Educational Technology & Society (IFETS). The authors and the forum jointly retain the copyright of the articles. Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear the full citation on the first page. Copyrights for components of this work owned by others than IFETS must be honoured. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from the editors at kinshuk@ieee.org. A Study of Contextualised Mobile Information Delivery for Language Learning Tim de Jong, Marcus Specht and Rob Koper Centre for Learning Sciences and Technologies, Open University of the Netherlands, Heerlen, The Netherlands // tim.dejong@ou.nl, marcus.specht@ou.nl, rob.koper@ou.nl ABSTRACT Mobile devices offer unique opportunities to deliver learning content in authentic learning situations. Apart from being able to play various kinds of rich multimedia content, they offer new ways of tailoring information to the learner���s situation or context. This paper presents the results of a study of mobile media delivery for language learning, comparing two context filters and four selection methods for language content. Thirty-five people (18 male, 17 female M = 31.06 years, SD = 8.93) participated in this study, divided over seven treatments in total. The treatment groups were compared on knowledge gain, and the results indicated that the results differed significantly. The results found indicated an effect of both context filters as selection methods on the learner performance. In addition, the results indicated a cost/benefit trade-off that should be taken into account when developing contextualised media for learning. Keywords Contextualised language learning, Mobile learning, Mobile information delivery, Context filters, Empirical Study Introduction Undoubtedly, language is one of the most important of mankind���s abilities. As Pinker, S. (1994) puts it: ���For you and I belong to a species with a remarkable ability: we can shape events in each other���s brains with exquisite precision.��� The communication Pinker hints at is only possible if we are able to understand each other���s languages an increasingly important ability in a world that is rapidly internationalising, not in the least because modern-day technology allows us to communicate over large distances and across language boundaries. A perfect example of such technology is a mobile phone, which not only simplified and increased communication possibilities, but also led to communication virtually anywhere and anytime. In addition, these increasingly powerful handhelds, now often referred to as ���smart phones���, provide other types of connectivity next to voice communication, and are often used to access all sorts of information on the move. In this paper, we will explore mobile technology supporting second language learners to communicate in a non-native language. The importance of communication in a target language has been stressed by several theories of second language learning. While each of the theories has a different viewpoint on language learning, all of them see language learning as an essential social process. First, the input and interaction theories of second language learning emphasise the role of social interaction for target language input, output, and interaction. These theories have been based two hypotheses. On the one hand, the interaction hypothesis (Long, M. H., 1981, 1983, 1996) states the importance of language interaction to increase the comprehensibility and usefulness of language input for the individual language learner. Especially, the role of negotiation of meaning between a native and non-native speaker is an essential part of the research inspired by this hypothesis. On the other hand, the output hypothesis (Swain, M., 1985, 1995) states that certain aspects (syntax and morphology) of a second language are most effectively developed in second language production. According to Swain, language output raises consciousness of problems and gaps in current knowledge, can provide opportunities to tests hypotheses about the second language, and allows the language learner to reflect on the language explicitly. Second, the socio-cultural perspectives to second language learning are grounded in socio-cultural and activity theory (Vygotsky, L. S., 1962, 1978) in which language is seen as a tool for making meaning in the collaboration with target language speakers. Thus, the socio-cultural perspectives also consider language interaction but their emphasis is more on the social motive for second language learning. In this sense, the emphasis of these theories is on self-regulation through private speech to gain control over the language task (Frawley, W. & Lantolf, J., 1985), the influence of personal characteristics and interests on social interaction (Coughlan, P., & Duff, P. A., 1994 Roebuck, R., 2000), and language feedback of native speakers to scaffold a second language learner (Aljaafreh, A., & Lantolf, J. P., 1994 Nassaji, H., & Swain, M., 2000). Last, the sociolinguistic perspectives consider the second language learner as part of communities of practice and investigate

111 the role of the learner���s identity, emotions, and social position in a learner���s development of a second language (Bremer, K., Roberts, C., Vasseur, M.-T., Simonot, M., & Broeder, P., 1996 Heller, M., 1999 Norton, B., 2000 Ochs, E., & Schieffelin, B., 1995 Pierce, B. N., 1995 Wenger, E., & Lave, J., 1991). Moreover, the sociolinguist perspectives see language learning as a situated activity, in which the influence of the learning context on the learner is essential. Summarising, the second language theories mentioned here all emphasise the social aspect of language learning in which both language production as language input in real-world scenarios with target language speakers are important. Thus, the possibility to access information anywhere and anytime makes mobile devices a welcome tool to support a second language learner in real-world interactions with target language speakers. A variety of studies already investigated the opportunities of mobile devices for language learning. Kukulska-Hulme, A. and Shield, L. (2007) distinguish between using mobile devices in a more passive manner for learning content distribution and using them to encourage interaction of the second language learners in their target language environment. Most of the current mobile language learning studies aim at the former content distribution and offer vocabulary training in previously unused time slots, instant lookup of vocabulary anytime and anyplace, and repetition in the form of quizzes and surveys. For example, Levy, M., and Kennedy, C. (2005) describe learning Italian vocabulary via SMS messages that were sent at specific time intervals. Likewise, Fisher, T., Pemberton, R., Sharples, M., et al. (2009) provide an example of an extended e-book reader that allows the second language learner to instantly look up vocabulary and listen to a native pronunciation. Last, Thornton, P. and Houser, C. (2005) investigated the effects of e-mails with English vocabulary sent to mobile devices owned by Japanese students, and described the combination of textual information (explanations, quizzes) and video material for mobile language learning. In contrary to these more passive mobile language learning approaches, mobile learning solutions supporting target language interaction are largely left unconsidered (Petersen, S. A., & Divitini, M., 2005). To address this lack of solutions Petersen, S. A., and Divitini, M. (2005) provide two scenarios for community-based mobile language learning, one of which focuses at interaction between students in a native and students in a non- native environment. Similarly, Kukulska-Hulme in her review of MALL also emphasises the importance of real- world interaction, and stresses the lack of mobile language learning solutions for speaking and listening (Kukulska- Hulme, A., & Shield, L., 2007). An interesting example of a context-aware mobile language learning system aimed at real-world interaction is JAPELAS (Ogata, H., & Yano, Y., 2004) that provides the learner with the correct Japanese politeness expressions based on a learner profile, location, and the person addressed. What���s more, Ogata, H. and Yano, Y. (2004) present TANGO, a mobile learning system that uses RFID-tagged real-world objects to teach vocabulary. Another example of mobile support for language interaction is the LOCH system that supports second language learners to carry out tasks in a Japanese target language environment (Paredes, R. G. J., Ogata, H., Saito, N. A., et al., 2005 Ogata, H., Yin, C., Paredes R. G., et al., 2006). In addition, the tasks carried out with LOCH were all focused on communication in the target language and were supported by a teacher that could view the GPS location of the students to give location-specific feedback. While the research mentioned above, considers language interaction at both the object and location level, it did not explore the effects of the learner context on the interactions in the target language. Thus, the influence of using different context granularities (object-based vs. location-based) to provide second language support at varying levels of specificity is not clear. A critical question that remains unanswered is whether there are differences between the efficiency of learning support provided by object-based and location-based information delivery. Moreover, if there are any differences are there any circumstances in which either of these granularities proves more efficient. Related to that, the context filters available can result in different forms of user interactions that may also influence the learner performance. In the study presented here, we aim to address part of these questions and present an evaluation of a language-learning tool that focuses on interaction support for second language learning. More specifically, we compare the effectiveness of object-based filters against location-based filters, and investigate the effects of several levels of mobile user interaction ranging from the users providing all context information themselves to the system automatically detecting the user���s context. It is expected that the more specific object-based filter leads to a more specific interaction with the learning content, and therefore a better learner performance. In addition, we expect that the automatic context detection will prove less of a burden on the learner and will prove the more efficient. The evaluation was carried out in a lab setting, where a number of rooms were equipped with objects according to a certain theme (market, restaurant, etcetera). In this paper, we adapt a framework for evaluating mobile learning from a technological (desirability, usability) and an educational perspective (effectiveness) that was proposed in (Sharples, M., 2009). The results of the evaluation with this framework will be presented in this paper.

112 Method Designs This study used a between-groups design, with two independent variables: the context filter (with two levels: room filter and object filter) and the selection method used (with four levels: semacode-based, number-based, list-based, and location-based). The dependent variable was the immediate knowledge gain calculated from the number of correct answers given in the pre-test questionnaire and the post-test questionnaire. The context filter independent variable was based on the context dimension of the reference model presented in (De Jong, T., Specht, M., & Koper, R., 2008). The room filter delivers the learning content based on location context, i.e. the room the learner is located in. The object filter delivers learning content based on identity context, more specifically the object the learner is currently interacting with. In this respect, the location-based filter provides learning content for a more general context than the object filter. The selection method independent variable specifies the variations of user interfaces that were used during the experiment, each with a different form of user control. Each variation acquired the context information via a different selection method: either directly from the learner (number-based or list-based), semi-automatically (semacodes), or automatically (location-based). In the number-based and list-based variations, the learner provides the context information respectively by (1) entering an object or room number in a text field, or (2) by choosing a room or object from a list with all possibilities available. The semi-automatic variations identified the object or room context by the semacode they were tagged with, and finally, the automatic variation detects the learner���s room location using a location tracking system. Each treatment variation in the study employed another combination of the selection method and the context filter, all of which are given by table 1. Because a location-based object filter was not available seven instead of eight treatments were tested. Table 1: Overview of the seven treatment groups that were used Selection method Context Filter Semacode-based Number-based List-based Location-based Room Filter SRF NRF LRF LORF Object Filter SOF NOF LOF x The dependent variable, the knowledge gain (KG), was calculated with the following formula: KG = (���KQposti - ���KQprei) / i, where i = 25. (1) Equation 1 calculates the knowledge gain, as a ratio, by subtracting the total number of correct answers of the pre- test (���KQpre) from the number of correct answers of the post-test (���KQpost) for all participants, and dividing the results by the total number of questions in the tests i. The minimum knowledge gain is therefore 0, the maximum knowledge gain equals 1. The manipulated variables led to the formulation of the following hypotheses: ��� Hypothesis 1: learners using an object filter (SOF, NOF, LOF) will have a higher knowledge gain (KG) than those using a room filter. We expect the specificity of the context information to influence the learning experience. In particular, we think that learning content filtered with more specific object context information, will lead to more specific interaction with the objects, and therefore will lead to better learning performance. ��� Hypothesis 2: learners using a selection method that requires fewer actions (SRF, SOF, LORF) to access content will have a higher knowledge gain (KG) than those requiring more actions. We expect the interaction with the learning content in the mobile software will also influence the learner performance. A more efficient user interface that requires fewer actions from the learner, in our case the semi-automatic semacode-based (SRF, SOF) and automatic location-based (LORF) selection methods, will lead to more efficient information access and a better learner performance.