“Address me personally!”: On the role of language styles in a MOOC
Abstract
If learning materials are presented in either a conversational or formal language style, people will process them differently. This study reports an experiment in which 64 high school students watched an educational video in a massive open online course scenario in which the instructions were phrased either conversationally or formally. The video demonstrated and explained the psychological McGurk effect. After watching the video, participants performed a transfer task. Learning outcomes were assessed and answers were analysed by using Linguistic Inquiry and Word Count. Results showed better learning outcomes for participants in the conversational (vs. formal) condition. Participants in the formal (vs. conversational) condition referred more to others in their answers. Implications for the design of massive open online courses are discussed.
Lay Description
What is currently known about the subject matter:
- Language styles can affect individual dealing with learning materials.
- Addressing learners directly in the learning material can support transfer performance.
- Massive open online courses (MOOCs) have become the most prominent form of online learning in higher education.
- High school students (future students) are a new group of MOOC users.
What this paper adds:
- MOOCs contain a social dimension, which might be crucial for individual learning.
- Most MOOCs do not ask for transferring knowledge of just learned materials.
- Conversational language in the instruction led to better transfer performance and more self‐directedness.
- Instructional language styles affect individual dealing with learning materials in a MOOC scenario.
Implications of study findings for practitioners:
- MOOCs should apply conversational language style instructions for deeper learning.
- Conversational language style instructions and transfer tasks are easy to implement in already existing MOOCs.
- Being personally addressed in a MOOC might be a key for enhancing learners' involvement.
1 INTRODUCTION
In recent years, massive open online courses (MOOCs) have become the most prominent form of online learning in higher education. MOOCs present learning materials to large numbers of learners, that is, the individual learner is one of many other participants in such large online courses. Put differently, MOOCs include a social dimension that might affect individual learning. The number of other participants may be presented to the individual learner, for instance, to underline the success of a MOOC or the company providing the MOOC. Even if the individual learner does not know the exact number of participants, the mere label of MOOCs as “massive” implies that there are many other participants.
This study focuses on extended MOOCs (xMOOCs), which are the most common models of MOOCs (see Littlejohn & Hood, 2018; for the variety of MOOC models, see, e.g., Kaplan & Haenlein, 2016). In general, xMOOCs consist of thousands of participants and do often contain educational videos and multiple‐choice tests (see, e.g., Toven‐Lindsey, Rhoads, & Lozano, 2015; Xiong & Suen, 2018). Put differently, individual learning in xMOOCs is driven by a behaviouristic approach (see, e.g., Knox, 2018; cf. Cohen & Holstein, 2018), with rather external aspects of learning being predominant (e.g., video quality). This means that the individual learner might not be supported in an appropriate way when being part of an xMOOC. In line with this, findings of Margaryan, Bianco, and Littlejohn (2015) confirmed that most xMOOCs lack instructional quality, that is, potential needs of the individual learner are not taken into account. Moreover, and with regard to Margaryan et al. (2015), most xMOOCs also did not provide learning opportunities for in‐depth information processing of just learned materials, for example, transfer tasks.
Therefore, addressing participants personally in a rather impersonal online learning environment of an xMOOC might be an appropriate way to support individual learning. This article reports on a study of how far manipulating the instructional language style prior to an educational video in an xMOOC scenario affects participants' learning outcomes, especially in a transfer task.
1.1 Personalization
Being addressed personally can support the individual processing of learning materials. Mayer, Fennel, Farmer, and Campbell (2004; Experiment 1), for example, showed that addressing individual learners in the learning material—using a conversational language style (i.e., “your”) instead of a formal language style (i.e., “the”)—resulted in better transfer test performance. According to Mayer et al., the theoretical explanation for this phenomenon, also known as personalization effect, is as follows:
Using the self as a reference point increases the learner's interest, which in turn encourages the learner to use available cognitive capacity for active cognitive processing of the incoming information during learning. The deeper processing results in more meaningful learning as indicated by better transfer test performance. (p. 391)
In the experiment of Mayer et al. (2004), college students watched an animation about the function of the human respiratory system that was accompanied by a spoken text. The text was either presented in conversational or formal language style, for example: “During inhaling, the [your] diaphragm moves down creating more space for the [your] lungs” (Mayer et al., 2004, p. 391). After that, a retention test and a transfer test (five transfer tasks) were given to the students. Results only showed differences for transfer test performance, that is, students in the conversational (vs. formal) language condition found more acceptable answers across all transfer tasks when applying the learned information to a new problem. Indeed, a meta‐analysis by Ginns, Martin, and Marsh (2013) empirically supports the assumption that individual learners can benefit from learning materials presented in conversational rather than formal language style: the medium average effect size on learning outcomes in transfer tasks was d = 0.54 across all integrated studies. In line with this, Dutke, Grefe, and Leopold (2016) reported an increased transfer performance in German high school students learning a scientific text when it contained a conversational rather than a formal language style (cf. Reichelt, Kämmerer, Niegemann, & Zander, 2014).
MOOCs are used by all ages around the globe with diverse personal learning goals (Shah, 2017). In recent years, also high school students have become a further user group of xMOOCs (see, e.g., Brahimi & Sarirete, 2015). Therefore, this study assessed whether German high school students could also benefit from the personalization effect when it is not applied in the learning material itself but in the instructions prior to an xMOOC learning session.
1.2 Learning with videos in xMOOCs
Videos are widely used in higher education. They are presenting educational information on a variety of topics in an up‐to‐date way (e.g., Tapscott, 2008). The ease of sharing video lectures via the Internet in contrast to traditional lectures allows flexible access to learning contents (McGarr, 2009). This is also accompanied by a fundamental change in general learning habits: “Society has moved from a community of learners whose knowledge was closely tied to local community experiences into a world of globalized learners” (Edgar, 2012, p. 3).
In large online course formats such as xMOOCs, masses of individual learners can study diverse learning materials via short videos. The actual individual learning is generally organized by subsequent multiple‐choice tests (e.g., Toven‐Lindsey et al., 2015; Xiong & Suen, 2018). A general problem of this approach has already been pointed out by Margaryan et al. (2015): most xMOOCs provide no opportunity to apply newly learned information to everyday problems. In particular, they lack transfer tasks that would encourage a reflection on one's own learning process. In other words, encouraging students' use of metacognitive strategies (Entwistle & McCune, 2004) is not a regular component of learning in an xMOOC.
Therefore, this study investigated whether written responses to a transfer task would differ when high school students were instructed in a conversational (vs formal) language style prior to an educational video in an xMOOC.
1.3 Crowding and its effects on cognition
Being part of a large crowd can be challenging for the individual. Studies show that experiencing actual crowdedness can alter cognitive states and impair performance on complex cognitive tasks (Evans, 1979; Nagar & Pandey, 1987). For example, Maeng and Tanner (2013, Study 1) demonstrated that students' cognitive elaborations were lower when they were either in an actual crowded situation or merely asked to imagine being in a pictured crowd (vs. noncrowded) setting. For higher education, Hellmann, Adelt, and Jucks (2016) showed that being given information about the number of others influenced students' written responses on the word level in terms of how many references they made to themselves and to others.
This aspect also applies to the xMOOC learning situation in which individual learners are usually aware of working individually among many others. Therefore, this study also investigated whether personalized instruction might promote self‐directed learning of high school students when being presented with a large crowd of other learners.
1.4 Rationale
The language styles applied in learning materials can lead to different learning outcomes. Dutke et al. (2016) have shown that the level of personalization in learning materials (conversational vs. formal language style) can particularly affect high school students' learning measured as outcomes in a transfer task. In line with these previous findings, it was predicted that using a conversational (vs. formal) language style in the instructions for an xMOOC scenario would result in better learning outcomes in a transfer task (H1). Drawing on the aforementioned findings reported by Hellmann et al. (2016), it was also assessed whether the wording of high school students' written responses would differ between conditions, that is, whether they would make more references to themselves or to others in a transfer and evaluation task. Therefore, it was predicted that using a conversational language style in an xMOOC scenario would lead to more references to oneself, that is, a higher self‐directedness (H2a). In contrast, a formal language style would result in more references to others, that is, less self‐directedness (H2b). In a further exploratory analysis, it was also examined whether the number of idea units in written responses to the evaluation task would differ between conditions.
2 METHOD
The experiment was approved by the ethics committee of the University of Münster, Germany. All participants provided written informed consent.
2.1 Participants and design
Sixty‐four German high school students (45 female) with a mean age of 16.86 years (SD = 0.96) participated in an experiment hosted on the Questback EFS Survey platform. The experiment had a one‐factorial between‐subjects design with two experimental groups. The language style in the instructions was experimentally manipulated so that it was either conversational or formal. Participants were recruited during a university information day for high school students on the psychology campus of the University of Münster, Germany.
2.2 Procedure and materials
Data collection took place in a seminar room equipped with six laptops. At the beginning of the experiment, participants were greeted by the male experimenter and received detailed information on the experimental procedure and data usage.
2.2.1 Experimental manipulation
The instructions for the experiment were manipulated by using different language styles (conversational vs. formal). In the prior information about the study, participants were told about MOOCs and given two fictitious quotes to read in which students attending such courses reported their personal experiences. Participants also read that they were going to watch a short learning video from a MOOC and that they would subsequently answer some questions on this video. The next page in the experimental instructions introduced the short learning video in either a conversational or a formal language style (see Appendix A for all instructions). The video was the same for both conditions, had a length of 2 min 49 s, and contained both background information and a demonstration of the McGurk effect (McGurk & MacDonald, 1976). Briefly explained, the McGurk effect shows that hearing something diverging from what someone sees interferes with speech perception and brings it more into line with the visual information. The video was followed by a short knowledge test containing five multiple‐choice questions on the effect in the video (see Appendix B). Next, participants were asked to write about (a) a potential everyday application of the McGurk effect and (b) their opinion on learning with online videos. Questions were the same for both conditions. The two questions were presented one after the other on separate web pages. The last pages were used to collect additional variables and demographic data (see Section 2.3). After the experiment, participants were fully debriefed and given the opportunity to ask questions either straight away or later by email. Every participant was rewarded with a chocolate bar for participating in the experiment.
2.3 Dependent measures
2.3.1 Transfer task
The first open question examined whether the experimental manipulation of language styles influenced how far participants transferred information about what they had learned in the video to another context. In particular, it was assessed how far participants were able to apply a psychological effect they had learned about to an everyday problem. Therefore, participants were asked: “When watching a dubbed movie or TV series, one gains the impression that there is no perfect match between picture and sound. Nevertheless, this impression disappears over time. How might this be explained by the McGurk effect?” Potential response differences between groups were analysed with two measures: First, two independent coders, who were blind to the respective condition, analysed whether and to what extent responses contained a correct explanation of the learned effect and/or a clear application to the given problem. The correspondence of initial ratings between the coders was satisfactory, with a Cohen's (1960) κ of 0.708. Scores ranged between 5 (very detailed and correct explanation as well as clear application) and 0 (no correct explanation nor clear application). Second, all responses were scanned with the software program Linguistic Inquiry and Word Count (LIWC; Pennebaker, Francis, & Booth, 2007; German dictionary by Wolf et al., 2008). Here, the focus was on determining the number of first‐person pronouns and third‐person pronouns in all texts.
2.3.2 Evaluation task
The second question assessed participants' opinion about learning with online videos by asking, “To what extent is learning with videos useful?” Here as well, two measures were used to analyse potential group differences: First, the amount of first‐person and third‐person pronouns in all texts was assessed with LIWC; second, an explorative analysis was performed to assess the number of idea units (related utterances regarding one argument) in relation to text length in participants' responses. Here, it should be determined whether participants were more likely to generate idea units in an evaluation task when being instructed in a conversational in contrast to a formal language style (see also Mayer et al., 2004). One point per idea unit was assigned, and the amount of idea units per text was divided by its word count.
2.3.3 Short knowledge test
A multiple‐choice test was used to assess what participants had learned about the effect presented in the video. This contained five questions about the McGurk effect that each gave four alternative answers. Participants could score from 0 to 5 points.
2.3.4 Additional variables
To control for unintended effects, the following ratings on 7‐point scales were collected: individual response behaviour (“How important was it for you to complete all the tasks in this study properly?” ranging from 1 [not at all important] to 7 [very important]), individual motivation (“How motivated were you when working on the task in this study?” ranging from 1 [not at all motivated] to 7 [very motivated]), perceived impact of individual contribution (“How relevant to science is your contribution to this study?” ranging from 1 [not at all relevant] to 7 [highly relevant]), and individual mood (“How is your current mood?” ranging from 1 [very bad] to 7 [very good]). Need for cognition was also assessed with an adapted short scale that contains four items (see Beißert, Köhler, Rempel, & Beierlein, 2014). Furthermore, participants were asked whether they had any previous knowledge of the McGurk effect. Moreover, it was assessed whether participants had taken online courses before, and if so, to rate their experiences with online learning on a 5‐point scale from 1 (very negative) to 5 (very positive). Finally, a suspicion check was included asking participants to say what they thought was the true purpose of the experiment.
3 RESULTS
Participants took a mean time of approximately 26 min to complete the experiment, with no difference between conditions, t(62) = −0.895, p = .375, ns. Regarding the amount of words in the written responses, there was no difference between conditions, either for transfer, t(62) = −1.179, p = .867, ns, or for the evaluation task, t(62) = −0.168, p = .243, ns. No participant guessed the experimental purpose in a suspicion check at the end of the experiment. As hypotheses are unidirectional, all p values referring to them are reported one‐tailed.
3.1 Transfer task (learning outcomes)
With regard to H1, an independent‐samples t test revealed a significant difference between conditions for instructional language style: Participants in the conversational language condition demonstrated a higher quality of response than those in the formal language condition, t(62) = −1.718, p = .045, d = 0.43 (see Table 1 for means and standard deviations).
| Tasks | xMOOCs | |
|---|---|---|
| Conversational language style | Formal language style | |
| Transfer task (learning outcomes) | 1.25 (1.41) | 0.75 (0.84) |
| Transfer task (LIWC) | ||
| First‐person pronouns | 0.04 (0.24) | 0.08 (0.44) |
| Evaluation task (LIWC) | ||
| First‐person pronouns | 0.36 (0.70) | 0.43 (0.95) |
| Third‐person pronouns | 2.62 (2.12) | 1.83 (1.81) |
| Idea units (exploratory) | 0.05 (0.01) | 0.05 (0.02) |
- Note. LIWC = Linguistic Inquiry and Word Count; xMOOCs = extended massive open online courses.
3.2 Transfer task (LIWC)
An independent‐samples t test indicated no difference between the conversational and formal language condition for first‐person pronouns, that is, self‐directedness, t(62) = 0.388, p = .699, ns (H2a; see Table 1 for means and standard deviations). In contrast, a Mann–Whitney U test revealed a significant difference between conditions for third‐person pronouns, indicating that participants in the formal language condition (Mdn = 2.17) referred more to others than those in the conversational language condition (Mdn = 0.00), U = 328, p = .003, r = .34 (H2b).
3.3 Evaluation task (LIWC) and exploratory analysis of idea units
An independent‐samples t test revealed no difference between the conversational and the formal language condition for first‐person pronouns, t(62) = 0.333, p = .741, ns (H2a). Another independent‐samples t test revealed no difference between conditions for third‐person pronouns, t(62) = −1.597, p = .115, ns (H2b). A further independent‐samples t test showed no difference between conditions in the number of idea units in relation to text length, t(62) = 0.647, p = .520, ns (see Table 1 for means and standard deviations).
3.4 Short knowledge test
An independent‐samples t test indicated no difference in the total score of multiple‐choice responses between the conversational and formal language style condition, t(62) = −0.973, p = .334, ns. Furthermore, participants' overall score was very high (M = 4.05, SD = 0.89; see Table 2 for means and standard deviations).
| Conditions | xMOOCs | |
|---|---|---|
| Conversational language style | Formal language style | |
| Short knowledge test | 4.16 (0.95) | 3.94 (0.84) |
| Individual response behaviour | 5.78 (1.16) | 5.75 (0.92) |
| Individual motivation | 5.56 (1.10) | 5.59 (1.13) |
| Perceived impact of individual contribution | 4.50 (1.32) | 4.19 (1.22) |
| Individual mood | 5.47 (1.02) | 5.47 (1.02) |
| Need for cognition | 4.96 (0.81) | 4.53 (1.00) |
- Note. xMOOCs = extended massive open online courses.
3.5 Additional variables
Independent‐samples t tests revealed no differences between conditions for individual response behaviour, individual motivation, perceived impact of individual contribution, and individual mood (ts < 1.000, ps > .343, ns). Another independent‐samples t test indicated no difference between conditions for need for cognition, t(62) = 1.912, p = .060, ns. For this scale, Cronbach's alpha was .528 (this rather low reliability score can be explained by the low connectedness of the items of this short scale; see Beißert et al., 2014). Moreover, 90.6% of all participants reported that they had no prior knowledge of the McGurk effect. Finally, about one half (53.1%) of the participants had already used online learning services and rated their experiences with them as being mostly positive (M = 4.21, SD = 0.73) with no difference between conditions, t(32) = −0.803, p = .428, ns (see Table 2 for means and standard deviations).
4 DISCUSSION
This study highlights that personalization effects are not limited to learning materials. In particular, results showed that individual learners can benefit from instructional texts that address them personally immediately before being presented with a short learning video. As pointed out in the introduction, xMOOCs generally lack transfer tasks even though these can be crucial for individual learning and knowledge gain (Margaryan et al., 2015).
The experiment revealed two main findings in an xMOOC scenario. First, participants showed better learning outcomes in a transfer task when instructional texts were presented in a conversational (vs. formal) language style. Second, participants demonstrated a higher focus on others in a transfer task when instructions were presented in a formal (vs. conversational) language style. Otherwise, references to oneself in the transfer task did not differ between conditions. The findings on the focus on others could be interpreted as an indication of being less self‐directed. Put differently, being addressed personally in an xMOOC scenario potentially promotes self‐directed learning and transfer task performance.
There were no differences between conditions in word use or the amount of idea units in the task of evaluating learning with videos. This could be because the evaluation task was presented after the transfer task, which might have further weakened the effect of the different instructional language styles. Moreover, the manipulation before the learning material was probably not salient enough. In this vein, effects might well have been stronger if the language styles of the transfer and evaluation tasks were also manipulated linguistically. In addition, there was no difference between conditions for the short knowledge test.
These findings can be evaluated in terms of differences in task complexity. The evaluation task and short knowledge test were less appropriate for eliciting a personalization effect than the transfer task. To quote Mayer et al. (2004, pp. 393–394): “if personalized versions encourage additional constructive processing—such as organizing and integrating the material—then personalized versions should create their greatest effects on measures of transfer.”
In sum, the findings of this study are in line with Dutke et al. (2016) and can be seen as an extension to the existing literature on the personalization effect: The findings demonstrate that it is even sufficient to address learners personally prior to learning materials to create an effect on their transfer performance. From the findings of this study, the personalization effect has to be seen in a wider scope of instructional design strategies (i.e., using conversational language in the instructions prior to a learning session), especially when applied in large online courses like xMOOCs.
4.1 Limitations and future research
Here, a scenario study was conducted, that is, specific assumptions were tested in an experimental setting and not in a real xMOOC environment. That is, a rather artificial scenario was used. The experimental design allowed us to compare controlled conditions. However, this affects ecological validity. In addition, a seminar room was used for realizing the online experiment via laptops. Therefore, it might be possible that the environment interfered with the experimental setting. Furthermore, the data of the study were cross‐sectional, which does not allow for suggestions regarding long‐term user experience. Furthermore, the sample of this study had a rather narrow set of participants: Only high school students, who share a similar age and educational background, participated in the study. Learners in xMOOCs can be of many different ages and educational backgrounds, therefore, various characteristics of instructional texts might be relevant for different groups of learners. This issue should be addressed in future research. Moreover, it would be crucial to test the specific assumptions of this study in a real xMOOC setting. This should require relatively low effort because only changes in instructional texts and not within already existing learning materials (e.g., videos) are necessary. Finally, it remains an outstanding issue which degree of personalization might be optimal to promote individual learning in large online courses like xMOOCs.
5 CONCLUSION
In xMOOCs in which the individual learner might gain the impression to be one among many others, applying a conversational language style in the instructional texts before the videos might especially support self‐directed learning and transfer task performance. Therefore, xMOOCs might benefit from applying conversational style instructions and transfer tasks in order to support deeper individual learning and to promote an active understanding of contents.
APPENDIX A: LANGUAGE STYLES: FORMAL VERSUS CONVERSATIONAL
| Formal: Text passages read by participants in the condition with nonpersonalized texts (translated from German). | Conversational: Text passages read by participants in the condition with personalized texts (translated from German). |
|---|---|
| Introduction to the experiment: | Introduction to the experiment: |
| The Internet provides the opportunity to learn a lot of things online. Online courses with very high numbers of users are in keeping with current trends. Usually, several thousand persons take part in these online courses, also called massive open online courses (MOOCs). | In the Internet, you have the opportunity to learn a lot of things online. Online courses with very high numbers of users are in keeping with current trends. Usually, several thousand persons take part in these online courses, also called massive open online courses (MOOCs). |
| Researchers collected the direct quotes of students on what it feels like to take an online course with many other learners. Examples are: | We collected direct quotes of students for you, on what it feels like to take an online course with many other learners. Examples are: |
| “Participation in an international MOOC with approximately 12,000 other people from every part of the world is quite interesting.” | “Participating in an international MOOC with approximately 12,000 other people from every part of the world is quite interesting for you.” |
| “There is the knowledge of being together with so many others who are unknown, but also want to learn the same things as oneself.” | “You have the knowledge of being together with so many others who are unknown to you, but also want to learn the same things as you.” |
| “When misunderstandings regarding aspects of their content arise, it is impossible to easily ask a neighbouring person as in a seminar at the university. In MOOCs, there are many others, but to which of them should one turn?” | “If you misunderstand aspects of content, it is impossible for you to easily ask the person sitting next to you as in a seminar at the university. In MOOCs, there are many others, but who should you turn to?” |
| How is it possible to improve learning in such MOOCs? Educational videos are an important part of online learning. A short educational video of a MOOC will now be presented. Subsequently a few questions will be asked. | How is it possible to improve learning in such MOOCs? Educational videos are an important part of online learning. We shall now show you a short educational video of a MOOC. Subsequently we shall ask you a few questions. |
| Introduction to video: | Introduction to video: |
| Please wait … the learning video is loading and will begin automatically. After the video was seen, “Next” can be used. | Please wait … your learning video is loading and will begin automatically. After you have seen the video, you can use “Next.” |
APPENDIX B: SHORT KNOWLEDGE TEST
All participants received the same five multiple‐choice questions on the short online video with four alternative choices (translated from German).
-
The described effect is named …
(a) MacDonald effect; (b) McGurk effect; (c) McKurg effect; (d) McShamrock effect
-
The described effect is primarily known in the field of …
(a) Perceptive psychology; (b) Clinical psychology; (c) Developmental psychology; (d) Social psychology
-
The described effect is considered to be proof of …
(a) The influence of speech on a person's evaluation; (b) The influence of seeing on hearing; (c) The influence of speech on the perception of hearing; (d) The influence of hearing on the perception of seeing
-
What was the name of the study that was reported in the video?
(a) Seeing Lips and Hearing Voices; (b) Hearing Voices and Seeing Lips; (c) Seeing Voices and Hearing Lips; (d) Hearing Lips and Seeing Voices
-
In which year was the article published?
(a) 1972; (b) 1967; (c) 1976; (d) 1970
Biographies
Jens Riehemann is a postdoctoral researcher and lecturer at the University of Münster, Germany. He received his PhD in psychology in 2017. His research addresses digitization in higher education with a special focus on supporting individual learners in massive open online courses.
Regina Jucks is a full professor at the University of Münster, Germany. She received her PhD in psychology in 2001 and her habilitation in 2005. Since 2011, she has headed the local Center of Teaching in Higher Education. Her research addresses various fields of learning through communication with a special focus on digitized interactions.
