Abstract
The pervasive use of synchronous computer-mediated communication (SCMC) in second language (L2) learning has generated increasing interest among researchers in integrating SCMC with task-based language teaching (TBLT). This study examined the effects of task complexity on Chinese EFL learners’ language production in SCMC modality to develop optimal tasks that facilitate the learning of English in SCMC environments. Eighty-four intermediate Chinese EFL learners completed two interactive tasks (simple and complex) in dyads via text-based or video-based SCMC. Their English productions were transcribed and coded in terms of syntactic complexity, lexical complexity and accuracy for statistical analyses. The results indicated that increasing task complexity elicited significantly lower syntactic complexity in text-based SCMC, but without significant effects on syntactic complexity in video-based SCMC. Significantly higher lexical complexity and unaffected accuracy were observed in both SCMC modes as a result of an increase in task complexity. Regarding SCMC modality, text-based SCMC resulted in significantly lower syntactic complexity, but significantly higher lexical complexity and accuracy than video-based SCMC.
Plain language summary
The purpose of this study is to find out the effects that simple and complex tasks have on Chinese students’ English language production when conducted using synchronous computer-mediated communication (SCMC) platform that enables text-based and video-based interaction. Eighty-four intermediate Chinese EFL learners completed the simple and complex tasks in pairs via text-based or video-based SCMC. The students’ interactions were transcribed and analyzed for their complexity in terms of sentence structure and choice of words, and grammatical accuracy. The results revealed that increasing the complexity of tasks led to a significantly lower complexity of the students’ sentence structure in text-based SCMC but no significant effect in video-based SCMC. Furthermore, it also led to a significantly higher complexity of choice of words and no change in terms of grammatical accuracy for both the text-based and video-based SCMC. When comparing between text-based with video-based SCMC, it was found that text-based SCMC resulted in significantly lower complexity of sentence structure but significantly higher complexity on their choice of words and grammatical accuracy.
Introduction
Research on the impact of different forms of computer-mediated communication (CMC) on L2 learning is of great interest to applied linguists (Dao et al., 2023; Hung & Higgins, 2015; Namaziandost et al., 2021; Torres & Yanguas, 2021; Yanguas & Bergin, 2018). CMC includes interpersonal exchanges that occur with the assistance of computers (Herring, 1996) either asynchronously (ACMC) or synchronously (SCMC). While ACMC (e.g., interactions over discussion boards and emails) may occur with delayed interaction without requiring all participants to be online simultaneously, SCMC (e.g., text, audio, and video chat) takes place in real-time resembling face-to-face (FTF) communication (Yilmaz, 2011).
Peterson (2010) proposed examining task design in CMC contexts to maximize L2 learning opportunities provided by online interaction. To investigate task design in CMC settings, task complexity research inspired by Robinson’s (2001) Cognition Hypothesis has been extended to SCMC contexts. Robinson (2001) claims that varying task complexity along certain dimensions can lead L2 learners to focus on linguistic complexity and accuracy simultaneously, thus eliciting quality language production and facilitating language acquisition and development. His Cognition Hypothesis (Robinson, 2001) provides specific ways of task manipulation, and predicts the patterns of learners’ cognitive processing and language use such manipulations engage learners in.
Many studies motivated theoretically by Robinson’s (2001) Cognition Hypothesis have explored effective tasks that induced high quality L2 production, which made huge contributions to task design in FTF contexts (Abdi Tabari et al., 2023; Kim, 2020; Liang & Xie, 2023; Li et al., 2023; Luo, 2022; Mohazabieh et al., 2020; Rahimi, 2019). To date, only a handful of studies investigated Robinson’s model in SCMC most notably via text chat (e.g., Adams et al., 2015; Adams & Nik, 2014; Baralt, 2013; Nik, 2010; Nik et al., 2012). Unlike FTF task complexity studies which generally support the Cognition Hypothesis, these pioneer studies seemed to indicate that the Cognition Hypothesis may not be applicable to computer-mediated contexts, which suggests that the features of text chat (e.g., planning opportunities) may have been a contributing factor. These conclusions are however far from definitive because of the small number of these studies. It also is barely known how the Cognition Hypothesis will bear out in a different technological environment (e.g., video chat) other than text chat. Current positions in technology-mediated L2 learning indicate the need for more research in this area (Smith & González-Lloret, 2020).
Task modality (oral and written) is another critical task design factor that considerably influences L2 production (Gilabert et al., 2016). Research on task modality emphasizes the benefits of writing in L2 learning. The availability of output processing time, as well as the permanence and the visibility of written text consistently result in higher syntactic and lexical complexity and accuracy (Kormos & Trebits, 2012; Kuiken & Vedder, 2011; Son, 2022). Based on these previous findings, it can be assumed that the distinct features of different SCMC modes may also affect language production. For example, text chat exhibits features of writing, which allows more processing time to examine and to reflect on the linguistic form (Smith, 2003). In contrast, video chat demands more online oral output which reduces planning possibilities (Smith & González-Lloret, 2020). With extra processing time, text SCMC can be expected to have the advantage of eliciting more syntactically and lexically complex and more accurate output. Empirical evidence is however needed, given that despite the features text chat shares with writing, it also retains the conversational feel and flow of speaking (Smith, 2003).
In an attempt to fill the gaps, this study addresses two primary issues in task design: the effects of task complexity and the effects of SCMC modality on language production in terms of syntactic complexity, lexical complexity and accuracy. It seeks to provide insights into effective task design for SCMC settings by investigating how task complexity affects the performance of Chinese undergraduate EFL learners engaging in interactive tasks via text-based and video-based SCMC in a laboratory setting.
Literature Review
Task Complexity and Language Production in SCMC Context
Task complexity is defined as “the result of the attentional, memory, reasoning, and other information-processing demands imposed by the task structure on the language learner” (Robinson, 2001, p. 29). To guide the task design in research and pedagogy, Robinson (2001) established the Cognition Hypothesis in which two sets of dimensions of cognitive task complexity are distinguished, namely, the resource-dispersing dimension and the resource-directing dimension. According to Robinson, increasing task complexity along the resource-dispersing dimension (e.g., +-planning time; +-task structure; +-prior knowledge) diverts learners’ attention from linguistic forms to other information processing activities (e.g., organizing thoughts and planning what to say during real-time communication). As a result, the complexity and accuracy of learners’ L2 production would be negatively affected. On the other hand, increasing task complexity along the resource-directing dimension (e.g., +-here and now; +-reasoning demands; +-few elements) poses cognitive demands that directs learners’ attention to particular linguistic features necessary to meet task demands, hence resulting in the simultaneous growth in both complexity and accuracy. Robinson (2001) also assumed that cognitive task demands enhanced along the resource-directing dimension influenced the linguistic complexity of language production in interactive tasks differently. Specifically, the complexity of L2 production may be reduced in more complex interactive tasks in that multiple one-word/phrasal responses occur during interlocutors’ negotiations, resulting in lower syntactic complexity.
So far, there are only few task complexity studies conducted in SCMC contexts, solely in text-based SCMC (Adams et al., 2015; Adams & Nik, 2014; Nik, 2010). Different from FTF task complexity studies which lend general support to Robinson’s Cognition Hypothesis, the existing studies in text-based SCMC context provided findings that seriously doubt the application of his paradigm to this environment.
A series of studies (Adams et al., 2015; Adams & Nik, 2014; Nik, 2010) explored certain resource-dispersing variables in text-based SCMC. Nik (2010), for example, investigated two task implementation factors, +-task structure (TS) and +-language support (LS) on learner language production with 96 Malaysian undergraduates in text-based SCMC. The participants performed a simulation of a decision-making task on an engineering problem in teams of four under one of four experimental conditions based on (+-TS) and (+-LS). The results indicated that the -TS task (complex task) did not significantly affect syntactic or lexical complexity but induced lower accuracy, while the -LS task (complex task) led to higher syntactic and lexical complexity but lower accuracy. These results only partially supported the prediction of the Cognition Hypothesis (Robinson, 2007) that increasing task complexity along the resource-dispersing variables decreases the syntactic complexity, lexical complexity, and accuracy.
These results are largely supported by findings from Adams et al. (2015) which, using exactly the same tasks as Nik (2010), also investigated the effects of these two task implementation features (+-TS and +-LS). Results showed that - TS (complex task) hardly influenced syntactic or lexical complexity but produced significantly less accurate language. With regard to the impact of LS, -LS (complex task) significantly increased syntactic and lexical complexity, but decreased accuracy. Both the effects of +-TS and +-LS largely disconfirmed the Cognition Hypothesis (Robinson, 2007).
Adams and Nik (2014) studied the factor +-prior knowledge (a resource-dispersing factor) on L2 production in text-based SCMC with forty-eight undergraduates who completed a problem-solving task concerning software systems in teams of four. Participants with electrical engineering backgrounds who were familiar with this project formed a +prior knowledge (+PK) group, and those with chemical engineering backgrounds who were not familiar with this project formed a –prior knowledge (-PK) group. The +PK group would find the task easier compared with the -PK group. The findings indicated that the -PK group produced higher lexical complexity and global accuracy than the +PK group, but the syntactic complexity was at a similar level across the two groups. These findings denied the Cognition Hypothesis which claimed that the syntactic complexity, lexical complexity and accuracy of L2 production would decrease as a result of increasing task complexity along the resource-dispersing variables. Adams and Nik (2014) attributed the results to the unique features of text-based SCMC which provided online planning opportunities and thus allowed learners to compose and edit their own output before sending their messages, suggesting that the text-based SCMC mode mediated task complexity and ultimately, influenced L2 production. This interpretation was supported by Baralt (2013) who compared the effects of reasoning demand (a resource-directing variable) on the efficacy of recasts in traditional FTF and text-based SCMC environments. Her results indicated that the claims of the Cognition Hypothesis that increased task complexity would result in greater attention to recasts and ultimately more L2 learning were confirmed in the FTF context but not in the text-based SCMC context.
Together, these studies suggested that the Cognition Hypothesis might not accurately predict how task complexity affected task performance in text-based SCMC contexts. However, they are far from conclusive, as the number of these studies is so small. Undoubtedly, more research is needed. Besides, these studies only used text-based SCMC as the medium of interaction, neglecting other SCMC modes, for example, video chat. It would be intriguing to explore whether video SCMC which reduces the possibilities for planning would introduce a different dimension of “complexity” than that found in text SCMC or even in FTF communication, a gap also pointed out by Smith and González-Lloret (2020). To fill these gaps and provide insights into task design for technological environments, this study examines the validity of the Cognition Hypothesis in terms of to what extent increasing task complexity along the resource-directing factor +-few elements affects L2 production in terms of syntactic complexity, lexical complexity and accuracy in both text and video SCMC settings.
+-Few Elements and Language Production
There are several studies investigating how manipulating the resource-directing factor +-few elements impacts task complexity, and by extension, language production in FTF oral communication and in pen-and-paper writing (Abdi Tabari et al., 2023; Kim, 2020; Kuiken et al., 2005; Kuiken & Vedder, 2007, 2011; Rahimi, 2019; Révész, 2011; Robinson, 2001; Xu et al., 2023). Defined as “distinguishing between and selectively referring to one or more among many elements or objects” (Robinson, 2007, p. 194), the “number of elements” has been interpreted and operationalized in different ways.
Robinson (2001) investigated the effect of +-few elements on the performance of 44 Japanese undergraduates in two interactive city map tasks. The simple map task comprised few elements that took the form of landmarks that could be easily distinguished in a small area the participants were familiar with, while the complex task consisted of many elements of a large area unknown to the participants. The results showed that the complex task improved lexical complexity, whereas the syntactic complexity and accuracy of L2 production remained intact. Kuiken et al. (2005) and Kuiken and Vedder (2007) examined the written production of native Dutch speakers learning Italian and French. The “elements” were manipulated by varying the number of criteria to consider when deciding on the choice of holiday destinations, namely three in the simple task and six in the complex task. The studies showed that tasks with more elements had favorable effects on accuracy and lexical complexity, but without significant effects on syntactic complexity. Révész (2011) explored the effects of the number of elements and reasoning demands on L2 oral production in two interactive tasks where participants were asked to discuss how much money they would assign to a fund organization in groups of three or four. In the simple version, participants had to distribute $50,000.00 among three projects, while in the complex version, they had to allocate $100,000.00 to six programs. Results revealed that the complex task triggered greater accuracy and lexical complexity but lowered syntactic complexity.
Kim (2020) examined the effects of +-elements and +-reasoning on L2 oral production of intermediate-level Korean undergraduates. The participants narrated a series of pictures. The simple task consisted of two main characters; the complex task consisted of three main and two minor characters. The reason for the event’s problems in the simple task was relatively less complex than in the complex task. Increased task complexity resulted in significant improvement in syntactic complexity, but accuracy was barely affected. Xu et al. (2023) assessed the effects of +-few elements with intermediate Chinese EFL learners who completed two writing tasks on the topic of “choosing the best roommates.” Elements were operationalized as the number of candidates and the number of properties of the candidates. The simple task featured four options of candidates each of whom was presented with four properties, and two had to be assigned to a dormitory. The complex task had six candidates each of whom was marked by six properties, and two pairs had to be allocated to two dormitories. Results showed that more elements led to no significant differences in syntactic complexity and accuracy, but negatively affected lexical complexity.
Overall, the findings suggested that more elements induced certain improvements in L2 performance, aligning with the claims of the Cognition Hypothesis. However, inconsistent results have been yielded regarding linguistic complexity and accuracy. These mixed findings can be partially attributed to the lack of uniformity in operationalizing the variable under investigation (i.e., number of elements). In the present study, Kuiken and Vedder’s (2007) manipulation of “elements” as the number of criteria to consider in parallel when making a decision was employed, following the assumption that this kind of operationalization affects the cognitive processing engaged during the conceptualization stage, which should subsequently impact L2 production in the formulation of the verbal messages (Kormos & Trebits, 2012).
Previous Studies on Effects of SCMC Modality
The increasing prevalence of online learning has raised interest in exploring the potential of computer-mediated communication in affecting learners’ interaction and language production. While there are abundant empirical studies investigating the contrasting effects of FTF and text SCMC (Kessler et al., 2020; Kourtali, 2022; Qiu, 2022; Qiu & Bui, 2022; Torres & Vargas Fuentes, 2021), few studies have compared the effects of different SCMC modes (Dao et al., 2023; Hung & Higgins, 2015; Namaziandost et al., 2021; Torres & Yanguas, 2021; Yanguas & Bergin, 2018).
For example, Yanguas and Bergin (2018) examined the occurrence, nature and resolution of language-related episodes (LREs) across audio SCMC and video SCMC. The results showed that the quality of interaction was mediated by SCMC modality, with more unresolved LREs in the audio SCMC than in the video SCMC. More recently, Dao et al. (2023) compared how text SCMC and video SCMC would affect the frequency of peer feedback in L2 interaction. It was found that more feedback occurs in video SCMC than in text SCMC. The content analysis of the interview data revealed that learners attributed the differences in feedback occurrences to different features of the two SCMC modes. Torres and Yanguas (2021) examined whether different conditions (text, audio, and video) impact LRE engagement levels (i.e., elaborate, moderate, limited, non-interactive). The results indicated that the audio group engaged more in limited LREs than the text group, and the former also engaged more in elaborate LREs than the video group, suggesting that SCMC modality moderated how L2 learners addressed LREs due to the differential demands different SCMC conditions place on L2 learners. Apart from the research investigating SCMC modality effects on learners’ interaction features, there are also studies examining the role of SCMC modality as a mediator in learners’ L2 development. Namaziandost et al. (2021) investigated how different SCMC modes (text versus voice SCMC) influenced Iranian EFL learners’ oral proficiency and anxiety over six weeks, and found that both SCMC modes improved oral proficiency, but only text SCMC reduced anxiety levels. It was also found that learners in text SCMC and video SCMC used different communication strategies, resulting in varied learning opportunities provided by these two SCMC modes (Hung & Higgins, 2015).
The preceding literature confirmed the comparability of different SCMC conditions in that differences in outcomes across modes were detected. However, these studies that compared the effects of different SCMC modes only focused on L2 interaction patterns and L2 development. To the best of our knowledge, only one study to date (Tokutake et al., 2021) compared the effects of voice SCMC, video SCMC, and virtual reality (VR) on learners’ L2 production in terms of accuracy, lexical complexity and syntactic complexity. The results showed higher lexical complexity in VR but without significant differences in syntactic complexity or accuracy across the three modes. There is hardly any study investigating whether the distinct characteristics of text SCMC and video SCMC affect learners’ cognitive process, which consequently affects L2 productions in terms of syntactic complexity, lexical complexity and accuracy. The impact of SCMC modality on L2 production thus merits further exploration in terms of which SCMC mode benefits which dimension of language production. The present study addresses this issue, which could offer L2 teachers and practitioners some insights into the choice of technology mode as part of effective task design to ensure a more balanced L2 development.
Research Questions
The studies reviewed revealed the lack of research investigating the effects of task complexity in SCMC modality. No research explores the effects of SCMC modality on language production in terms of syntactic complexity, lexical complexity and accuracy. This study thus aims to fill the gaps by answering the following research questions:
RQ1: How does task complexity (+-few elements) affect L2 production of Chinese EFL learners in text and video-based SCMC in terms of:
a. syntactic complexity?
b. lexical complexity?
c. accuracy?
RQ2: How does SCMC modality (text and video) affect L2 production of Chinese EFL learners in terms of:
a. syntactic complexity?
b. lexical complexity?
c. accuracy?
Method
Participants
Eighty-four intermediate EFL learners were selected from approximately 330 third-year undergraduates majoring in English at a university in northwest China. Purposive sampling was used in this study because learners with lower levels of proficiency, according to Kuiken et al. (2005), may not have attained the proficiency level essential for devoting their attention to the increased task complexity, which was supported by Kormos and Trebits (2011) who proposed that it is necessary to reach a specific proficiency threshold for task effects to be detectable. Therefore, the third-year undergraduates were the targeted participants for the study. It was made clear that they could participate voluntarily and had the freedom to leave the research when they felt the need to. One-hundred and eight students were willing to join in the research. To choose participants with a maximally similar profile, a background questionnaire was disseminated to the potential participants to collect their personal information such as their English learning experiences and their experiences of computer-mediated communication, as these factors might influence their language performance in this study. Finally, 84 participants who had a homogeneous background in terms of the above-mentioned aspects were selected and they were randomly assigned to one of two groups to perform interactive tasks in dyads through the text-based SCMC (n = 42) or the video-based SCMC (n = 42).
Their English proficiency was tested by the V_YesNo (Meara & Miralpeix, 2016) test online. V_YesNo test is a quick vocabulary test which despite its simplicity can efficiently describe L2 proficiency. Instead of providing multiple-choice prompts, it asks whether subjects know the meaning of an item. Vocabulary tests have been proven to have a high correlation with proficiency (Milton, 2009). Speaking proficiency was found to be significantly correlated with V_YesNo test scores by Uchihara and Clenton (2020). The V_YesNo test was thus considered an adequate indicator of L2 proficiency for this study. The test manual suggests that a score of 3,500 to 6,000 indicates an intermediate level (Meara & Miralpeix, 2015). The mean scores of the two groups (text-based SCMC: M = 4614.43, SD = 449.83; video-based SCMC: M = 4591.88, SD = 372.39) classified the participants as intermediate learners. To further ensure that the two groups were at the same level of English language proficiency, an independent t-test was conducted. The results indicated there were no statistically significant variations in V-YesNo results across the groups (t(82) = 0.25, p = .80, d = 0.05), suggesting the homogeneity of the groups in English language proficiency.
Based on the background questionnaire, the subjects were adult males (n = 8) and females (n = 76) with their ages ranging from 20 to 23. The notable gender imbalance in this study could be ascribed to the fact that the number of female English majors was disproportionately larger than that of male English majors in the research context. However, the gender distribution of the two groups was comparable (text group: 4 males and 38 females; video group: 4 males and 38 females). All of the participants had Chinese as their native language. None had ever been to English-speaking countries. Since Chinese students normally started learning English as a compulsory subject from grade 3 in primary school, they had been learning English for approximately 12 years by the time the study was conducted. Besides, they all utilized computers extensively for a range of activities including gaming, online-learning, searching for information they needed on the internet, and chatting socially. Therefore, they were very comfortable with text and video communication via computers.
Like most of the higher learning institutes in China where English was learned as a foreign language, the research context offered compulsory English language programs focusing on reading, writing, listening and speaking to English majors. However, most often, grammar was taught with a greater emphasis than communicative language use. Furthermore, as English was learned as a foreign language, it was only used in English classrooms but not in daily life or at work.
Design and Procedures
This study followed a 2 (task complexity) × 2 (different SCMC modality) repeated measures factorial design with task complexity as the within-subjects factor and SCMC modality as the between-subjects factor. The two SCMC groups (text vs. video) engaged in the tasks in different language laboratories with each monitored by the researcher and an instructor. They were instructed to perform two interactive decision-making tasks (one simple and one complex) in dyads via text chat or video chat of the video-conferencing platform WeMeet (similar to Webex or Zoom) which was widely used for online teaching and learning during the COVID-19 pandemic in China. Both the researcher and the instructor had taught the course of English listening and speaking in the computer laboratory and thus were familiar with the laboratory. They were also experienced in using WeMeet.
The file containing the directions for the two tasks was uploaded to WeMeet using its feature of “document file sharing” which enabled the participants to open and read the file simply by double-clicking it. They could also access the file during task performance. The task instructions were given in Chinese language, to prevent the participants from using the words in the prompts. The tasks were counterbalanced to avoid the carry-over and practice effects. Following Cho (2018), the participants were given a brief pre-task planning time (3 min) to read the instructions and make their choice. The time allotments (21 min per task for the text group and 5 min per task for the video group) for task performance were determined according to the time each group spent on each task in the pilot study. Immediately after completing each task, they were asked to complete the self-rating questionnaire in the Chinese language administered to them to measure their perceived difficulty of the task. Figure 1 illustrates the study procedure.
Study procedure.
The text-based SCMC group kept their audio and video off, and communicated by typing their messages in the chat box. They were not allowed to talk to their interlocutor directly if they happened to sit close to each other, and were required to use proper English words and sentence structures. Each dyad was asked to save their chat log in a document file. The video-based SCMC group was instructed to turn on the video and record their performance using the “recording” feature; they were not allowed to type in the chat box during the task performance. To avoid noise from other participants engaged in oral communication at the same time and ensure the quality of the recording, participants in this group were asked to wear earpieces connected to microphones throughout the experiment.
Tasks
Two interactive decision-making tasks on two similar topics were used, with “elements” manipulated as criteria to consider when making a choice. The simple task had fewer elements than the complex task (Robinson, 2001). Adapted from the tasks in Kuiken et al. (2005) and Mahpul and Oliver (2018), participants were instructed to discuss with their partners and defend one option out of five as to which hotel to choose or which apartment to rent. When making the choice, several requirements (three in the simple task versus six in the complex task) had to be considered in parallel. The participants had acquired the prior knowledge of the vocabulary and grammar required to complete the tasks. Their textbooks contained units about holidays and traveling which involved hotel booking and purchasing a house from a real estate agent.
According to previous research, it has been argued that validity evidence for the operationalization of task conditions should be provided in task complexity studies (Norris & Ortega, 2009; Rahimi, 2019). That is, the complex task should be perceived as more cognitively taxing than the simple task as assumed. The self-rating questionnaire has been widely used as an effective measure of task complexity in earlier research (Cho, 2018; Mahpul & Oliver, 2018; Rahimi, 2019). It has been considered the most cost-effective method to discriminate between simple and complex tasks when compared to others (see Révész et al., 2016 for more information on other methods). Using Likert-type scales, this technique asks students to rate the perceived cognitive load of the tasks.
In this study, task difficulty was measured by a 6-point Likert scale self-rating questionnaire adapted from Robinson’s (2001) original questionnaire, with higher values on the scale indicating greater task difficulty. The reliability of the self-rating questionnaire was tested by Cronbach’s alpha which revealed a very high level of internal consistency with a value of .839. As each participant completed two self-rating questionnaires (one for the simple task and one for the complex task), a total of 84 valid questionnaires from the text-based SCMC group (n = 42) and the same number of valid questionnaires from the video-based SCMC group (n = 42) were collected. We found that the text SCMC group perceived the complex task (M = 3.15, SD = 0.57) as more cognitively taxing than the simple task (M = 2.34, SD = 0.64), t(41) = 6.88, p < .001, d = 1.06. The same was also true of the video SCMC group whose ratings for the complex task (M = 3.36, SD = 0.65) were higher than those for the simple task (M = 2.95, SD = 0.37), t(41) = 3.41, p = 0.001, d = 0.53. This confirmed the validity of the task manipulations.
Measures of L2 Production
The participants’ language productions in the two tasks via the two SCMC modes were assessed on three dimensions, namely, syntactic complexity, lexical complexity and accuracy. Measures that constitute valid descriptors of each dimension were used.
Analysis of Speech Units (AS-unit)
The present study employed the AS-unit as a fundamental unit of analysis. Defined as: “a single speaker’s utterance consisting of an independent clause, or sub-clausal unit, together with any subordinate clause(s) associated with either” (Foster et al., 2000, pp. 365–366), AS-units are considered the most appropriate for simultaneous interaction full of language fragments. Previous studies examining the effects of task complexity on L2 oral production (e.g., Mahpul & Oliver, 2018; Santos, 2018; Vasylets et al., 2017) and on L2 interaction via text SCMC also adopted AS-unit as the basic unit of analysis (e.g., Adams et al., 2015; Adams & Nik, 2014; Nik, 2010).
Measures of Syntactic Complexity
Syntactic complexity refers to the “range of forms that surface in language production and the degree of sophistication of such forms” (Ortega, 2003, p. 492). Norris and Ortega (2009) suggested that the sub-dimensions of syntactic complexity such as overall complexity (mean length of AS-unit), complexity by subordination (mean number of clauses per AS-unit), and complexity by subclausal or phrasal elaboration (mean length of clause or mean number of modifiers per noun-phrase) should be measured at minimum. According to them, the subordination measures ought to be the most predicative source of complexification at intermediate levels, while phrasal-level measures could be expected as powerful measures of complexification at advanced stages of development. The participants in this study were intermediate learners, so complexity by subclausal or phrasal elaboration which is more suitable for the advanced learners was not measured. Three measures concerning overall syntactic complexity and complexity via subordination were tapped, including the mean length of AS-unit, the number of clauses per AS-unit and the ratio of subordinate clauses to the total number of clauses.
Measures of Lexical Complexity
Lexical variation and lexical sophistication were used to measure lexical complexity. Lexical variation refers to the range of different words employed in learners’ production with a larger range denoting a higher degree of diversity (McCarthy & Jarvis, 2010). Following the previous similar studies (e.g., Adams et al., 2015; Nik, 2010; Révész, 2011), Guiraud’s index was employed as the measure for lexical diversity. It was calculated by dividing the number of types by the square root of the number of tokens. Lexical sophistication measures the percentage of unusual or advanced words in learners’ language production. This study employed Lexical Frequency Profile (LFP; see Laufer & Nation, 1995 for more information about LFP) to measure lexical sophistication as in other similar studies. Participants’ transcripts were uploaded to the Range program (Nation & Heatley 2002) which matched the words in the transcripts with the 14,000 most frequently used words in the British National Corpus (BNC) word frequency list. Following Adams and Nik (2014), the percentage of words used beyond the first 1,000 words of the BNC wordlists was calculated.
Measures of Accuracy
The degree to which the L2 output corresponds to the norm of the target language is referred to as accuracy (Housen & Kuiken, 2009). This study used the number of errors per 100 words as a global measure of accuracy because it was objective and claimed to be more valid than unit-based measures of accuracy (Inoue, 2016). This approach has also been adopted by Ruiz-Funes (2015) and Awwad (2017). Polio and Shea’s (2014) guidelines for error coding were used to code the errors. In addition, one specific accuracy measure based on the correct use of verbs (target-like use of verbs) was also used. Verbs were selected because Chinese learners of English tended to struggle with the accurate use of tense, aspect, modality and subject-verb agreement due to the absence of these morphological variations in Chinese.
Reliability of Data Coding
The written interactions from the text group (21 dyads) were directly copied into Word documents from the chat box of WeMeet, while the oral interactions of the video group (21 dyads) were transcribed by the researcher. Altogether 42 valid interaction transcripts (21 from the simple task performance and 21 from the complex task performance) of the text group and the same number of valid interaction transcripts (21 from the simple task performance and 21 from the complex task performance) of the video group were obtained. Except for the word count which was automatically counted by the Range program (Nation & Heatley, 2002), these transcripts were coded and rated manually in terms of syntactic complexity, lexical complexity and accuracy by the researcher and a second rater with a PhD in Applied Linguistics. At the training session, the researcher and the second rater studied the instructions on the coding system first and then completed the coding of six randomly selected transcriptions together. The rationales for the coding decisions were discussed until the agreement was reached. The researcher then coded the rest of the transcripts independently and a randomly selected sample of 40% (as in Rahimi & Zhang, 2017) of the transcripts was recoded by the second rater to check for the interrater reliability. Cohen’s (1992) kappa values indicated high intercoder agreement for syntactic complexity (mean length of AS-unit = 90%; the number of clauses per AS-unit = 92%; the ratio of the subordinate clause to the total number of clauses = 94%), lexical complexity (Guiraud’s index = 100%; LFP = 99%), and accuracy (the number of errors per 100 words = 94%; target-like use of verbs = 92%) which were well above the acceptable level.
Statistical Analyses
Paired samples t-tests were administered for RQ 1, while independent t-tests or Mann–Whitney U tests were conducted for RQ 2 depending on the normality of the distributions checked by the Shapiro-Wilk test. Analyses were carried out using SPSS version 26 with the alpha level set at 0.05 for all tests. Cohen’s d and r were employed to measure effect sizes for the t-tests and the Mann–Whitney U tests. Following Cohen (1992), d values of 0.20, 0.50, and 0.80 and r values of .10, .30, and .50 were considered small, medium, and large, respectively. Statistical power analysis for all tests using GPower 3.1 (Faul et al., 2009) was also performed to ascertain whether the sample size was adequate to reliably show a non-significant result without risk of Type II error (i.e., false negative where significant differences are not detected). Based on the results, 34 participants per group, 64 participants per group, and 68 participants per group were needed for paired samples t-tests, independent t-tests and Mann–Whitney U tests respectively to detect medium effect sizes with an α = .05 and power = .80. The sample size (n = 42 per group) was found to be sufficient to obtain reliable findings for RQ 1 (which explored the effects of task complexity) but inadequate for RQ 2 (which explored the effects of SCMC modality). As it was difficult to recruit as many participants as 68 for each group, we carried on the study with the designed sample size. Despite the insufficient sample size for RQ 2, the significant findings for this research question are reliable because the insufficient sample size does not affect Type I error (i.e., false positive where insignificant differences are found to be significant). However, the insignificant findings for RQ 2 should be taken as highly experimental as they might be false negative (as mentioned earlier) due to the insufficient sample size.
Results
Research Question 1: The Role of Task Complexity in Text SCMC and Video SCMC
Descriptive statistics for all the measures of language production across the simple and complex tasks in the text-based SCMC and the video-based SCMC modes are displayed in Table 1. The findings indicate that increasing task complexity decreased syntactic complexity and lexical diversity (Guiraud’s index), but improved the lexical sophistication (LFP) and accuracy for the text-based SCMC group. The descriptive statistics (Table 1) however, hardly reveal any obvious difference in performance across the two tasks for the video-based SCMC group, with the exception of LFP which displays an increased performance in the complex task.
Descriptive Statistics for All Measures Under Simple and Complex Performance.
The paired samples t-tests for the text SCMC group in Table 2 indicate that these changes are only significant for syntactic complexity on the subordination measure (t(41) = 2.02, p = 0.05) and for lexical complexity indexed by LFP (t(41) = 4.34, p < .001). The effect size on the subordination measure was very weak (d = 0.31), but the effect of size on LFP fell into the medium range (d = 0.67).
Results of Paired Samples T-Tests for the Text-Based SCMC Group (N = 42).
Note. M = mean; SD = standard deviation; Text = text SCMC group; Video = video SCMC group S = simple task C = complex task; MLAS = mean length of AS unit; CAS = clauses per AS unit; RSC = ratio of subordination clauses to the total number of clauses; E/100W = errors per 100 words; TLV = target-like verbs.
p < .05. **p < .001.
The paired samples t-tests for the video-based SCMC group (Table 3) suggest no statistically significant effect of task complexity on syntactic complexity or accuracy. Lexical complexity indexed by LFP, however, was significantly affected with a large effect size (t(41) = 5.39, p < .001, d = 0.83). These results suggest a marginal role of task complexity in language production in both of the two SCMC modes.
Results of Paired Samples T-Tests for the Video-Based SCMC Group (N = 42).
Note. S = simple task C = complex task; MLAS = mean length of AS unit; CAS = clauses per AS unit; RSC = ratio of subordination clauses to the total number of clauses; E/100W = errors per 100 words; TLV = target-like verbs.
p < .001.
Research Question 2: The Role of SCMC Modality in Simple and Complex Tasks
Tables 4 and 5 summarize the results of inferential statistics obtained for all the measures of language production across the SCMC modes under the simple task and the complex task respectively. Tables 1 and 4 demonstrate that under the simple task condition, lexical complexity was significantly higher on the measure of Giraud’s index in text-based SCMC with a large effect size (t(82) = 4.73, p < .001, d = 1.04). Tables 1 and 5 show significantly lower syntactic complexity, but higher lexical complexity and higher accuracy in text-based SCMC as compared with the video-based SCMC under the complex task condition in that text-based SCMC produced shorter AS unit (t(82) = 2.53, p < .05) with a medium effect size (d = 0.55) but higher Giraud’s index (t (82) = 4.44, p < .001) and fewer errors per 100 words (z = 2.49, p < .05) with large (d = 0.98) and small (r = 0.27) effect sizes respectively, suggesting a more prominent role of the SCMC modality when the task imposed higher cognitive demands. These results suggest that the video-based SCMC mode elicited more syntactically complex output, while the text-based SCMC mode led to more lexically diverse and more accurate language.
Comparisons Between Text-Based SCMC and Video-Based SCMC Under Simple Task.
Note. MLAS = mean length of AS unit; CAS = clauses per AS unit; RSC = ratio of subordination clauses to the total number of clauses; E/100W = errors per 100 words; TLV = target-like verbs; Sig = significance; Effect = effect size.
p < .001.
Comparisons Between Text-Based SCMC and Video-Based SCMC Under Complex Task.
Note. MLAS = mean length of AS unit; CAS = clauses per AS unit; RSC = ratio of subordination clauses to the total number of clauses; E/100W = errors per 100 words; TLV = target-like verbs; Sig = significance; Effect = effect size.
p < .05. **p < .001.
Discussion
Role of Task Complexity
The first research question focused on the effects of task complexity manipulated by the resource-directing factor +-few elements on language production in terms of syntactic complexity, lexical complexity and accuracy in text-based and video-based SCMC contexts. The results indicated that this manipulation of task complexity affected the two SCMC groups differently on syntactic complexity, but identical effects were detected for the text-based SCMC group and the video-based SCMC group in terms of accuracy and lexical complexity. More precisely, (a) increasing task complexity elicited significantly lower syntactic complexity by the subordination measure for the text-based SCMC group while no significant changes were found in this dimension for the video-based SCMC group; (b) both text-based and video-based SCMC groups produced significantly higher lexical complexity by the measure of LFP in the complex task performance as compared to the simple task performance. Meanwhile, increasing task complexity failed to generate detectable effects on accuracy neither for the text-based SCMC group nor for the video-based SCMC group. The subsequent paragraphs discuss our findings for each aspect of production.
Regarding syntactic complexity, the Cognition Hypothesis (Robinson, 2001) predicts that increased task demands by the resource-directing factors in interactive tasks decrease syntactic complexity because of the multiple one-word or phrasal responses during interlocutors’ negotiations. The results of the text-based SCMC group lend partial support to this prediction as a lower ratio of subordination was found under the complex task condition. Comparisons of the current findings with those of previous research examining the factor +-few elements revealed that unaffected syntactic complexity in written production was reported in Kuiken et al. (2005), Kuiken and Vedder (2007), Zalbidea (2017), and Xu et al. (2023). The discrepancy in these findings might result from the fact that these previous counterpart studies used monologic tasks while the present study used interactive tasks. According to Robinson (2001), more interaction and turn-taking in the complex task (as opposed to the simple task) were likely to “mitigate learners’ attempts to produce complex syntax and subordination” (p. 36). However, the effect size for this measure was rather small (d = 0.31), suggesting that task complexity only accounted for 31% of the causes for the difference. In other words, task complexity had very slight effects on syntactic complexity in text-based SCMC. It could be noticed that the subordination ratios in both tasks were small (simple task: M = 0.22, SD = 0.11; complex task: M = 0.19, SD = 0.10). This could be due to the specific feature of text-based SCMC which, as Smith (2003) described, is characteristic of shorter sentences, abbreviations, and simplified syntax.
The results of the video-based SCMC group, however, contradicted the predictions of the Cognition Hypothesis (Robinson, 2001), because no significant effects were found on syntactic complexity. These findings aligned with those of oral production in Zalbidea (2017). A possible explanation for the findings of the video-based SCMC group regarding syntactic complexity might have to do with the time pressure the participants faced in the video-based SCMC. Due to the real-time nature of the video-based SCMC, learners might be obliged to give prompt responses to their interlocuters to keep the flow of the conversation. Consequently, they may have used the simplest syntactic structures in both the simple and the complex tasks to ensure a quick delivery of the intended message.
For lexical complexity, similar patterns of task complexity effects were observed for the text-based SCMC group and the video-based SCMC group in that both SCMC groups employed significantly more lexically sophisticated language (LFP) in the complex task, but no significant effects were found for lexical diversity (Guiraud’s index). These findings partially supported the predictions of the Cognition Hypothesis that increasing task complexity along the resource-directing factor +-few elements improves lexical complexity. Similar to our results, Kuiken et al. (2005) and Zalbidea (2017) also reported insignificant effects of task complexity on lexical diversity. Concerning the comparisons of findings about LFP, Kuiken and Vedder (2007) also found significantly more infrequent words in the complex task for the learners of French. Our interpretation of the current findings might be related to the nature of the experimental tasks. Based on Kuiken and Vedder (2007), enhancing task complexity by the factor ± elements would automatically induce a higher amount of reasoning. In the complex task, our participants used more sophisticated lexical items perhaps to strengthen the defense of their choice. It was plausible that, in their perception, using a variety of words would not achieve this communicative goal. This assumption was supported by Vasylets (2017) which also found higher lexical sophistication but unaffected lexical diversity as a result of increased task complexity manipulated by the resource-directing factor “+- reasoning demands”.
For accuracy, the Cognition Hypothesis (Robinson, 2001) predicts that increased task complexity along the resource-directing factor +-few elements induces higher accuracy because more elements in the complex task enhance cognitive load, which directs participants’ attention to language forms to meet the demands of the task. This prediction was refuted by our study which hardly found any significant effects on accuracy for either the text-based SCMC group or the video-based SCMC group. These findings are in line with Kim (2020) and Xu et al. (2023). Nonetheless, enhanced levels of accuracy in the complex task were found in Kuiken and Vedder (2011) for both written and oral productions and in Révész (2011) for oral production.
These divergent findings might be attributable to the differences in the characteristics of the participants in terms of first language (L1) and second language (L2) as well as L2 proficiency. In Kuiken and Vedder (2011), the participants were native Dutch who learned Italian with intermediate L2 proficiency, whereas the participants in the present study were intermediate English learners with Chinese as their native language. An alternative interpretation involved the inadequate and insensitive measures of accuracy used in the present study when compared with Kuiken and Vedder (2011), given that only one type of error (verb errors) was measured in this study while a distinction was made between four types of error (including grammatical, lexical, orthographic and pragmatic errors) in Kuiken and Vedder (2011).
The comparison between the present study and the study of Révész (2011) was particularly interesting. Although Révész (2011) also used interactive tasks, as mentioned earlier, it employed different tasks that operationalized “elements” differently in comparison to the present study. In Révész (2011), the task complexity was designed along two resource-directing factors simultaneously, namely +-reasoning and +-few elements. Therefore, it was difficult to determine whether the higher accuracy in the complex task was the result of more reasoning, or more elements, or the combined effects of both.
Additionally, in light of the previous task complexity studies in SCMC contexts (Adams et al., 2015; Adams & Nik, 2014; Nik, 2010) (all of which examined the resource-dispersing factors), the medium of interaction via which the participants performed the tasks might also contribute to the current findings regarding accuracy, albeit along different lines of reasoning for the two SCMC modes. For the text-based SCMC group, learners were given additional processing time due to the slow typing speed. Moreover, the saliency of output was also afforded. It could thus be suggested that these features of text-based SCMC may have anchored learners’ attention to form in both the simple and complex tasks, hence the similar L2 performance in terms of accuracy. The video-based SCMC, on the contrary, was fast and evanescent. Contrary to the slower text-based SCMC, the video-based SCMC participants hardly had time to plan, formulate and monitor their output. It was plausible that even if more elements in the complex task drew learners’ attention to language form, the effects might not be strong enough to overcome the online pressure of oral production (Vasylets, 2017).
In summary, we concur with the findings of Adams and Nik (2014) that the Cognition Hypothesis (Robinson, 2001) may not accurately predict the effects of task complexity in online communication, at least not in the two SCMC modes (text and video) that have thus far been investigated. It seems that task complexity does not operate alone. Instead, it is the interplay between task complexity and other factors such as the nature of the tasks and the SCMC mode via which the participants completed the tasks that shaped language productions. Although the effects of more “elements” on syntactic complexity for the two groups pointed to different directions (significantly lower for the text-based SCMC group and no significant changes for the video-based SCMC group), as discussed earlier, the effect size for the text-based SCMC group was marginal. This means that, the effects of task complexity on L2 production were not substantially constrained by the SCMC modality. However, these conclusions were far from definitive. It should be borne in mind that the previous studies have operationalized “elements” in a variety of ways. Different operationalizations of elements may yield discrepant findings. More research with different ways of operationalizing “elements” (e.g., Révész’s (2011) or Robinson’s (2001) operationalizations as opposed to Kuiken et al.’s (2005) operationalization employed in the present study) is therefore warranted to elucidate this issue. Following this, we would support the call of Michel et al. (2012) for a more nuanced and more specific manipulation of elements as a task complexity variable.
Role of SCMC Modality
The second research question concerned the role of SCMC modality in terms of the three dimensions of language production (syntactic complexity, lexical complexity and accuracy). The results showed that compared to the video-based SCMC, the text-based SCMC produced significantly shorter AS units in both the simple and complex tasks but greater lexical diversity and accuracy in the complex task only.
The effects of SCMC modality on syntactic complexity were somewhat unexpected in that the text-based SCMC, similar to writing, provides learners with planning opportunities to attend to the content and form of the message (Smith, 2003) which had been consistently found to improve the complexity and accuracy of L2 output (Yuan & Ellis, 2003). This may not necessarily happen in SCMC settings. A possible reason is that the time-consuming demands of typing and the retained “feel” of real-time interaction via text-based SCMC (Smith, 2003) may have put the text-based SCMC group under more pressure than the video-based SCMC group, which may have predisposed them to use simpler syntax. In contrast, video SCMC was faster with less physical effort. Therefore, participants may have produced more linguistic material, such as longer units. These findings ran parallel with those of Pyun (2003) which compared text-based SCMC with FTF discourse in dyadic discussions. However, they ran counter to those of Sauro (2012) which reported no significant difference in syntactic complexity between text-based SCMC and spoken discourse. As Sauro (2012) used monologic tasks and the present study used interactive tasks, the different task conditions might constitute a variable that influences learners’ performance. Specifically, it was possible that in monologic tasks, the learners in both text-based SCMC and spoken discourse could afford to pause and attend to linguistic forms without the pressure imposed by the quick information exchange in interactive tasks, hence the similar performance across the two modes investigated in Sauro (2012).
The findings about enhanced lexical complexity in text SCMC were inconsistent with those of Sauro (2012) which found insignificant changes in lexical complexity between text-based SCMC and spoken discourse. As discussed previously, the divergent findings might be attributed to the different task conditions of Sauro (2012) (which used monologic tasks) and the present study (which used interactive tasks). Turning to accuracy, the current findings were unsurprising. These findings supported Pyun (2003) and Kim (2017), both of which also observed higher accuracy in text-based SCMC than in FTF communication. The findings in terms of lexical complexity and accuracy in text-based SCMC could be interpreted with reference to the availability of more processing time in this mode. As for lexical complexity, this extended processing time in text-based SCMC allowed participants to search their language resources more exhaustively and to monitor the word selection more carefully, thus benefiting lexical diversity. In contrast, the video-based SCMC required the participants to make their lexical choices very quickly and also limited the monitoring process, resulting in a certain amount of undetected word repetition, hence the lower level of lexical diversity. By the same token, more processing opportunities in the text-based SCMC allowed for closer monitoring and editing of linguistic form, thus facilitating improvement in accuracy. However, the effect size for accuracy was very weak (r = 0.27), which necessitated confirmatory empirical evidence of this finding from future research.
Finally, it is worth noting that the SCMC modality seemed to play a more robust role when the task was more cognitively demanding in that the significantly higher syntactic complexity in the video-based SCMC and the greater accuracy in the text-based SCMC were only observed in the complex task characterized by more elements. This could be explained by the insufficient sample size. As mentioned in the earlier section, the number of participants was inadequate for RQ 2, which means the findings were liable to violating Type error II (i.e., false negative). Therefore, the insignificant differences in syntactic complexity and accuracy of the two SCMC groups in the simple task performance were likely to be false negative. More research is warranted to elucidate this issue.
Conclusion
The present study aimed to explore effective task design for technology-mediated settings by examining the effects of task complexity on Chinese EFL learners’ L2 production in terms of syntactic complexity, lexical complexity and accuracy in both text and video-based SCMC contexts. It also investigated how SCMC modality is related to language production across tasks of differing cognitive complexity. Findings for RQ 1 revealed that the Cognition Hypothesis did not bear out well in the text-based SCMC or the video-based SCMC. Another major finding is that the effects of task complexity on L2 production were not substantially constrained by the SCMC mode in that similarly slight effects were found as a result of increases in cognitive task demands for both groups. Findings for RQ2 indicated that text-based SCMC elicited higher lexical diversity and accuracy while video-based SCMC seemed to be more favorable for syntactic complexity.
These findings suggested that the Cognition Hypothesis seemed to be inapplicable in the new turf of SCMC modes. Furthermore, the comparison between text SCMC and video SCMC also revealed a slight deviation from the previously well-established benefits of writing over speaking in terms of syntactic complexity. This study posed several pedagogical implications. Given the findings that increasing task complexity along the task complexity variable +-few elements only had identically minor effects on L2 production in text and video-based SCMC modes, we intend to suggest that it is unquestionably inadvisable to design tasks for the two SCMC modes simply by increasing the number of elements in tasks and expect improvement in learners’ L2 production. On one hand, the planning opportunities and visibility of L2 output in text-based SCMC may mitigate the differences in cognitive load between simple and complex tasks, resulting in similar performances under the two task conditions. On the other hand, the hard-pressed nature of video-based SCMC is likely to nullify the effects of enhanced task complexity, leading to no significant changes in language productions across the two tasks. In line with the findings regarding the SCMC modality effects, we would suggest that different SCMC modes which provide different learning opportunities with their distinct features, could serve as a more significant part of task design. To be more precise, when the goal of teaching is to improve learners’ ability to use accurate and lexically complex language, text-based SCMC might be the appropriate choice of communication medium because its inherent nature, such as the availability of planning opportunities and the visibility of output, could facilitate deeper linguistic processing, whereas the video-based SCMC is the medium of interaction to choose if the aim is to encourage learners to produce syntactically complex language. What is more, blended use of both text and video-based SCMC is also recommended to diversify the application of SCMC mode in L2 learning as well as to enhance balanced L2 development.
It should be noted that the generalizability of these findings might be limited, given that the study was conducted in an EFL context, where learners have few opportunities to use English. In addition, all the participants shared the same L1 with homogeneity in English proficiency level. Learners’ language production may have differed when communicating in a multilingual setting or groups with diverse English proficiency levels. Future studies may experiment with participants from a variety of L1 backgrounds and different L2 proficiency levels to assess the generalizability of the current findings. Secondly, the sample size (n = 42) per group was relatively small for RQ 2. As mentioned earlier, the results of power analysis required 68 participants per group to obtain reliable findings for RQ 2 (which explored the effects of SCMC modality) with an α = .05 and power = 0.80. To confirm the reliability of the current findings regarding the effects of SCMC modality, future studies might need to increase the sample size. Thirdly, as the English majors were overrepresented by females in the research context, the number of male and female participants in this study was notably imbalanced. Future research is recommended to recruit a similar number of male and female participants to avoid the potential influence of gender imbalance on the generalizability of the results. Fourthly, as discussed earlier, this factor has been operationalized in different ways which can be employed in future studies to explore if different results will be obtained. Finally, future studies may add one more group, audio-based SCMC, and look into how this mode affects task complexity and ultimately, L2 production.
Footnotes
Acknowledgements
Special thanks go to Paul Meara for his kind recommendation of V-YesNo test after he learned about the study. We are grateful to all our participants for taking part in this study. The following devoted instructors were exceptionally helpful and cooperative throughout data collection: Xiaocong, Xue; Yanfeng, Liu; Xiuping, Li; Dongyan, Li; Zhongjiang, Tang; Qian Hao; Qi Cao; Fei Huang and Yi Su. We highly appreciate their support.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Ethical Approval
The ethical permission was not applied for because we consider it unnecessary for the following reasons: a) we obtained the permission of Yulin University to approach and contact the participants; b) we had distributed consent form to the participants which described the aim and procedure of the study as well as the voluntary nature of participation; the consent form also stated how the data would be used and how the participants’ anonymity would be protected.
Consent
We had obtained the informed written consent from the participants before the experiment was conducted.
Data Availability Statement
The data from this study are available upon request.