Video Self-Modeling (VSM) as a Strategy to Instruct CFL Students’ Sentence-Level Stress

Abstract

Sentence-level stress is one of the major means of expressing information focus in oral speaking, and it is of importance for Chinese as a foreign language (CFL) learners to accurately receive and send the right information in conversation. However, research related to teaching stress, especially sentence-level stress, is indeed scarce. In this study, we investigate whether video self-modeling (VSM) is applicable to improve CFL students’ sentence-level stress. VSM, as an innovative strategy, only shows the positive targeted behavior by using videos or audios of oneself, and aims to decrease students’ frustration and the negative influence caused by failed accomplishments. Twelve beginning-level CFL students, taken as the experimental group, received the edited perfect pronunciation audios with their own voice and used these own-voice audios to train their sentence-level stress. At the same time, another twelve advanced-level CFL students were taken as the control group, and received traditional instructional strategies from their class teacher. The whole training continued for ten sessions during a period of two and half months. Quantitative results show that with the help of VSM, CFL students’ sentence-level stress improved significantly as compared to the control group, with increased scores on the pronunciation of sentence-level stress words and increased scores in all three parameters: pitch, intensity, and duration. A post-training survey revealed that the participants’ preference for using their own voice as instructional material resulted in a feeling of success and satisfaction. The findings corroborate the importance of computer-assisted language learning in the second language (L2) field, and add solid evidence of using VSM in foreign-language training.

1. Introduction

In the linguistic field, the concept of prosody is raised within the framework of autosegmental-metrical (AM) phonology [1,2]. Prosody in Chinese is regarded as an important concept which involves different levels across an utterance, and the main prosodic features consist of tone, intonation, and stress [3]. Tones in Mandarin Chinese are categorized into four types plus a neutral tone, i.e., tone 1 (T1, high level), tone 2 (T2, mid-rising), tone 3 (T3, low-level or dipping), and tone 4 (T4, high falling), with the function of distinguishing lexical meanings [4]. Chinese tones are characterized by connecting to individual syllables, with a contour-based system with multiple contour tones, and multiple patterns of tone changes caused by adjacent tones or the prosodic utterance environment [5]. Intonation refers to the fundamental frequency (F0) patterns in speech that express information beyond lexical meanings, and usually occur at the phrasal and utterance level [6,7]. Tones and intonation are regarded as undivided aspects in Chinese, and interact with each other. Stress denotes that the stressed word or phrase in a sentence is more noticeable, so as to express the aims of the speaker and to attract the listener’s attention [8]. It is commonly agreed that there are three types of phonological stresses in Chinese, unstressed syllables, stressed syllables, and accented syllables [7,9,10]. Most studies only include the first two types of stress in the research on Mandarin prosody. Three parameters are used to differentiate stressed syllables from unstressed syllables both in words and in entire sentences [11], which are: pitch range (expanded/compressed), intensity range (increased/decreased), and duration range (long/short) [12]. Therefore, stressed syllables are assumed to have expanded pitch range, increased intensity, and longer duration. The vowels and consonants are likely to be fully pronounced.

Previous research has shown that studies on Chinese tone instructions have attracted a lot of attention from CFL researchers [4,13,14]. However, speakers do not utter single words in communication but rather express themselves in prosodic sentences; therefore, prosodic-level instruction research is indeed necessary for improving CFL students’ comprehensibility and intelligibility in communication [7,13,15]. There have been very few studies that explore explicit instructions on intonation and sentence-level stress in the CFL field as compared to the research on CFL students’ intonation and sentence-level stress acquisition [16]. Stress, being an integral part of Chinese prosody which is responsible for conveying the key information of a sentence, definitely deserves more attention from CFL researchers [10,12]. As such, in this paper, we aim to examine a strategy to develop CFL students’ sentence-level stress skills.

In recent years, computer-assisted language learning (CALL) has received growing attention for instructing Chinese pronunciation in the CFL field [4,17,18], whereas research on using CALL to instruct CFL students’ sentence-level stress is still limited [13]. Thus, we attempt to employ a strategy which is able to combine the advantages of CALL and Chinese sentence-level instruction. Research in the L2 field has shown the vast usage of visual–audio instruction for instructing L2 students’ Chinese language skills and the effectiveness of the employed visual–audio strategies [19,20,21]. However, in these previous studies, the good examples for students to imitate or study were all produced by others, not by the students themselves; therefore, students themselves were forced into a passive role in the interventions. Since the students are the focus in academic learning [22,23,24], especially in the L2 field, researchers are attempting to discover more effective strategies which can allow L2 students to take their own initiative and show motivation to learn the target language [25,26]. As such, a strategy that gives the students a core and active role in their own instruction by visually or audibly presenting them as successfully engaging in the target skill, is, ultimately, our preference for instructing CFL students sentence-level stress.

1.1. Video Self-Modeling (VSM)

VSM is one of the approaches that many language instruction specialists have used to improve communication skills and social and behavioral functioning in children with disabilities [27,28], as well as to improve students’ academic skills in classrooms [29,30,31]. VSM builds on the foundation of [32]’s social learning theory, by suggesting that individuals are able to learn by observing a behavior instead of experiencing it directly. VSM derives from video modeling, which uses video recordings of behavior to provide a visual representation of a target behavior [29,30]. VSM integrates one’s own behavior into the modeling, that is, it enables students to watch their own adaptive behavior to improve their performance and skills [33]. A video is recorded of a participant’s spontaneous behavior several times during each of the ten sessions, until the participant can achieve the targeted skills. Subsequently, the best video is edited by the instructor to include only the positive targeted behavior or skill. Negative behavior, errors, prompting, and irrelevant footages are all edited out [34]. VSM differs from traditional video modeling which produces videos with peers or other people. By contrast, VSM utilizes the videos of one’s own utterances and only shows the positive targeted behaviors, with the aim of decreasing students’ frustration from failed behavior [27,34,35].

1.2. Video Self-Modeling for Instructing Academic Skills

Various studies have been conducted to examine the effectiveness of VSM in improving the academic skills of different types of students of different ages and in a school-based setting [36,37,38]. Authors [38] reviewed the studies that implemented VSM for school-based academic abilities and attempted to summarize the effectiveness of VSM on students’ academic performance. VSM has proved to be an effective intervention technique for training the frequency, quality, and duration of target behavior and abilities [39]. Three different academic skills were included in the reviewed studies, concretely, reading, writing, and arithmetic. Students’ oral reading fluency [36,40,41], comprehension [36], and textual responses [42] were the main focus regarding academic reading capabilities. Only six studies were conducted, and all the participants were students who had disabilities but the potential to improve their targeted behaviors. All students who were involved in the VSM instructions achieved better scores and/or improved their performance. All six VSM intervention studies had very small groups of participants, less than five. More recently, [30] conducted a study on using VSM to train reading fluency with dual language learners with disabilities. The results showed that all three participants improved their reading fluency with familiar passages, and, although they had relatively lower scores with unfamiliar or untrained passages, an obvious reduction in making errors in reading was observed. Multiple baseline designs were used in all studies, and students were trained one to one. In addition to individual VSM interventions, group reading fluency is another direction for VSM researchers to tackle. Ref. [37] conducted a VSM intervention to train group oral reading fluency in three different groups (fast, medium, and slow), and increased speed was attained for the whole group and between groups. However, another VSM intervention on group oral reading fluency implemented by [31] arrived at a different result. No higher reading fluency was achieved by the four participants.

All students included in the previous VSM intervention studies on reading skills were focused on their native language (L1); hence, no research to date has focused specifically on VSM as an intervention for L2 students.

1.3. Research on Stress in the Field of CFL

Research on Chinese stress in the CFL field has mainly focused on finding the acquisition characteristics of CFL learners in pronouncing stress. Ref. [43] discovered that American CFL learners seemed to pronounce the syllables in sentences with longer durations. Ref. [44] implemented a read-aloud instruction to explore the duration of disyllabic stressed words and compared native speakers (NS) with non-native speakers (NNS). Generally, better scores were received by L2 learners in pronouncing the duration of disyllabic words in sentences. In addition, differences in the duration of dissyllabic words between NS and NNS were found in the durations of tone 1 and tone 2 of the first syllable in the sentences. Some researchers also attempted to distinguish the characteristics of CFL students in pronouncing sentence-level stress based on different language backgrounds. Ref. [45] examined syllable duration in sentences between NS and Korean CFL learners, and discovered that no significant differences between NS and Korean CFL learners at the intonation phrase boundary and prosodic phrase boundary. Korean CFL learners at different levels had different ways of pronouncing the duration of the syllables at the intonation phrase boundary. Several studies explored the characteristics of Thai CFL students’ sentence-level stress on duration and pitch levels by comparing them to native Chinese speakers [46,47,48]. Stress at the phrase level was more focused on by CFL researchers, including syllable duration and chunking abilities [49], as well as the relationships between syllable duration and tones [50,51,52].

Only very few studies attempted to implement interventions for instructing CFL students’ sentence-level stress. Ref. [53] explored the effectiveness of visual-aid instruction on improving Korean CFL students’ sentence-level stress perception and production ability. Students in the experimental group watched a series of slides which included 24 sentences from the pre-tests and two additional example sentences. Each slide presented one sentence together with the sound waveform of each character at the bottom of the slide and yellow-circled characters on top of the slide, representing the pitch level, intensity and duration. Results showed that students in the experimental group achieved better scores both on sentence-level stress perception and production post-tests. Test raters evaluated students’ post-tests using their own personal judgement. Ref. [54] investigated the effectiveness of speech imitation and speech shadowing on training CFL students’ sentence-level stress. Results showed that both strategies were helpful for improving students’ sentence-level stress. However, neither intervention was a long-term strategy, which means that the students only received the training once.

Thus far, no research has been conducted to examine all three parameters of the sentence-level stress of CFL students, and very little research on sentence-level stress instruction has been conducted using technological media, such as smartphones, computers, tablets and similar devices. Therefore, there is a need to increase the diversity of using technologies in sentence-level training and provide more evidence for the effectiveness of call on Chinese sentence-level stress training, so as to enable CFL teachers to have more choices when selecting appropriate strategies in their teaching.

1.4. The Present Study

Given the scare literature on VSM with L2 language training and on Chinese sentence-level in examining all three parameters (pitch range, intensity range, and duration range), we conducted a VSM intervention to instruct CFL students’ sentence-level stress. Thus, this study will focus on the following questions:

• Is VSM effective at improving CFL students’ pronunciation of sentence-level stress?
• What do interviews with CFL students reveal about VSM instruction and its impact on their learning of Chinese sentence-level stress?

In this study, students’ own adaptive-behavior recordings were implemented as the training material in the teaching process. Research has shown that one appreciates signal change more in one’s own voice, as compared to hearing a familiar person’s voice [55]). When producing a pitch to match a target stimulus, one monitors the pitch of their voice and compares it to the perceived pitch of the target [56]. Ref. [57] uncovered that individuals match the pitch of their own voice more accurately as compared to the voice of somebody else. The participants’ mechanism enables them to accurately match pitch, regardless of the stimulus type. These findings supported the validity of using VSM for instructing Chinese sentence-level stress. By listening to the positive targeted pronunciation, students will only focus on the correct pronunciation uttered by themselves [29].

The mentor theory [58] argues that speech perception is carried out by analyzing the signal based on the innate knowledge of the articulatory production of the particular sound. In addition, when learning a tone-based language, for L2 learners, the areas involved in the brain are not only the expansion of pre-existing language-related areas, but also the stimulation of additional cortical areas specialized in functions needed for learning the new language [59]. This means that Chinese language learners need to have drill practices when learning sentences, so as to let the brain become used to linking pronunciation to the meaning of words and to activate new areas of their brains.

Praat [60], as a measuring software to examine the features of pronunciation, was used for measuring the students of the two groups in the intervention. Praat has been broadly used in L2 computer-assisted language instruction, especially for pronunciation and speaking training. One of the reasons is that it can recognize, label, and analyze the value of speech signals, and the second reason is that language researchers can obtain data (such as pitch, duration, intensity, etc.) using calculations. Thus, in this study, we will use Praat to collect the data of the two groups of students, so as to analyze the effectiveness of VSM on instructing students’ sentence-level stress.

2. Methodology

2.1. Participants

In total, 24 students were involved in a 10-week intervention study. All participants were enrolled as bachelor students in the “Chinese Language and Culture” program in a public university in Belgium. All students spoke Dutch as their native language and had the same age range. They studied Chinese as a Foreign Language (CFL) as a compulsory part of their program. Considering the small number of students in this program, the assignment of participants to the experimental and control group was constrained due to sample sizes. The participants of the experimental group consisted of twelve first-year CFL university students (ten female and two male), who were at the beginners’ level of learning Chinese, based on their performance of Hanyu Shuiping Kaoshi (HSK). The participants of the control group consisted of twelve second-year CFL university students (ten female and two male), who were at the intermediate level of the HSK (Hanyu Shuiping Kaoshi (HSK) comprises of written tests and speaking tests (HSKK). For the written test, it consists of six bands (levels), namely HSK (Level I), HSK (Level II), HSK (Level III), HSK (Level IV), HSK (Level V), and HSK (Level VI). The HSKK has three levels (i.e., Beginning, Intermediate, and Advanced). HSK (Level I) and HSK (Level II) correspond to HSKK beginning level. HSK (Level III) and HSK (Level IV) match HSKK intermediate level). Being aware of the expected differences in initial CFL proficiency level between both research samples, we invested in a stringent measurement of the pre-intervention proficiency of all participants on the base of a pre-test. In terms of the learning environment, all students used the same text- and workbooks in their first year, were taught by the same first-year teachers, and experienced the same instructional strategies and evaluation approaches in the first year.

Informed consent was obtained from all participating students prior to the start of the intervention.

2.2. Trainers

One of the authors conducted the video-preparation, data-collection, recording, data-analysis, and intervention phases of the research process. The instructor received training in video modeling beforehand, and has experience and knowledge in single-subject design, Chinese as a foreign language teaching, linguistics, and the use of the mobile phone for different teaching purposes. All the authors designed the training plans and test materials together.

2.3. Setting and Materials

The study was set up at the start of the second semester for both the first- (experimental group) and second-year (control group) students. Starting in the second semester was needed to guarantee that the novice students had at least acquired basic Chinese language knowledge at word level and pronunciation level, to be able to be involved in the more advanced sentence-level lessons in the second semester. All intervention sessions were organized once a week in a one-to-one setting. Ten sessions were organized and each session took around 20–30 min for each student. The intervention was affected by the COVID19-pandemic, so the last six intervention sessions of the semester were delivered through Zoom. Prior to the ten sessions, a pronunciation pre-test was administered face to face. After the ten sessions, a pronunciation post-test was administered, in addition to a post-intervention interview via Zoom.

The ten one-to-one VSM instruction sessions were recorded with a Canon Powershot G3 X BK EU 23 camera (Tokyo, Japan). A De Zoom H4n Pro Handy recorder (Hauppauge, NY, USA) was used for the pre-test and students’ own smartphones were used for the online post-tests. The recording environment was confirmed to be quiet to ensure high-quality recordings. Students were allowed to use their own mobile phones, tablets or laptops to record their after-class activities. Their reading materials were selected from their regular workbook (Integrate Chinese I) [61], guaranteeing correspondence with their regular Chinese class content.

The validity of the Intervention materials was checked by two independent CFL teachers (including the grammars and available literatures on Chinese sentence types) to confirm whether they comprehensively covered all the important varieties in Chinese sentences (such as yes-no questions, wh-interrogatives, exclamatives, declaratives, and imperatives).

2.4. Video Self-Modeling: Independent Variable

The study was integrated into the regular Modern Chinese course schedule of the two groups of participants. Figure 1 illustrates the entire scenario of the intervention design for each participant within each group and each session, including what the instructor, the course teachers, and the students did during the intervention, how long an instruction session lasted, what feedback the instructor provided, and in what way the instructor trained students’ sentence-level stress with VSM. Two independent CFL teachers and one university educator were responsible for checking the intervention material, in order to validate that its content closely followed the nature and orientation of the VSM theory. The control group teaching activities were also incorporated in the material.

Figure 2 illustrates the after-class activity schedule for the students of each experimental group to follow. Those instructions were explained at the beginning of the intervention, but the instructions were also sent together with the recordings by email to each student after each session. The instructor did not interfere with the training strategies employed in the control group. The course teacher of the control group used her own regular training strategies to instruct students’ stress.

2.5. Data Collection: The Pre- and Post-Pronunciation Test

The data collection was based on the administration of the pre- and post-pronunciation test, and of a post-intervention interview.

The pre- and post-pronunciation test for measuring students’ sentence-level stress consisted of 16 sentences (see Appendix A) which had to be pronounced aloud by each individual student. The 16 sentences were randomly arranged. Three parameters of the sentence-level stress (pitch, intensity, and duration) were subsequently coded by the researchers. The coding reflected the quality level of the specific pronunciation parameters and was, therefore, turned into a score of 0 or 1 (in two cases, two disyllables had to be scored). This implies that students could achieve a maximum ratio level score of 18. The coding was carried out by two independent CFL teachers. This allowed for the calculation of the Cohen’s κ inter-rater reliability index, which reflected a high reliability of 0.952 (p < 0.001). Complete agreement could be then achieved through discussion, in case initial disagreement arose.

The three parameters for this study (pitch range, intensity range and duration range) were to determine the effect of VSM on the production of CFL students’ sentence-level stress. Refs. [7,12] explicitly explained that sentence-level stress in Chinese is characterised by longer duration, expanded pitch range, and louder intensity [62]. Praat software [60] is widely used in ESL and the CFL field to measure language intonation and stress, visually presenting the pitch contour, pitch level, intensity level, and duration of certain vowels or prosodic features [13,14,63,64].

The codes used to judge the quality of the three parameters were developed on the basis of a preliminary study, since no related benchmarks were available in the literature. In this pilot, we involved five female and five male CFL teachers to read the same materials of the pre- and post-pronunciation tests. All ten CFL teachers were in-service teachers who attained Grade B of Putonghua Level 1 and their ages ranged between 25–44. Their pronunciation was recorded in view of analysis with the Praat software. The minimum value and maximum value of all five female CFL teachers’ pitch, intensity, and duration of the stressed words were collected. The same procedure was applied in relation to the five male CFL teachers’ output. From the data of the five female CFL teachers, we derived minimum and maximum values to be used as the benchmark to evaluate the pitch, intensity and duration values of the female students’ pronunciation in the intervention. The same procedure was followed to obtain the benchmarks from the male teachers to evaluate male students’ pronunciation in the intervention.

The coding referred to above consisted of evaluating whether students’ pronunciation did (score 1) or did not (score 0) fall within the minimum and maximum benchmarks. These procedures were also applied to evaluate the intensity range and the duration range. Thus, each student could reach a maximum score of 18 points for each dependent variable, and a minimum of 0 points.

2.6. Data Collection: The Post-Intervention Interviews

Post-intervention interviews with each individual student of the experimental group were conducted to gather their perceptions of the VSM instructional approach. The interviews consisted of three questions: (a) “Do you think the video self-modeling which was used in this research will contribute to the repertoire of the sentence-level skills of your Chinese learning?”; (b) “Are you satisfied to participate in the sentence-level stress training program by using video self-modeling?”; and (c) “What comments do you want to add concerning the conducted training?” The same two CFL researchers checked the post-survey questions to ensure the survey collected detailed explanations and thoughts from CFL students about the VSM training. Next, students’ answers were transcribed, and included in a content-analysis procedure to extract main themes. The interviews allowed us to understand students’ views about the VSM training format, how it affected their learning of sentence-level stress, and their feelings about this innovative approach.

2.7. Data Analysis

SPSS (version 28, IBM, Armonk, NY, USA) was used to analyze the quantitative data for the student outcomes of the two groups. Wilcoxon matched pairs tests were carried out to explore whether participants in each group improved their sentence-level stress with the instruction. Mann–Whitney U tests were also carried out between these two groups, to observe which group made more progress.

As to students’ perceptions of the VSM instructional format, all interview transcriptions were analysed following an open-ended content-analysis approach. This allowed the extraction of main themes and sub-themes.

3. Results

3.1. VSM Test Scores

Based on the prosodic features of Chinese sentence-level stress, we attempted to explain students’ results in three parameters: pitch, intensity, and duration.

Table 1. Descriptive data of pitch, intensity, and duration parameters in the pre- and post-tests.

Group	Test	N	Pitch		Intensity		Duration
			M	SD	M	SD	M	SD
Experimental	Pre	12	10.42	3.55	3.83	3.01	7.67	3.63
	Post	12	14.00	2.41	9.08	3.77	10.50	3.03
Control	Pre	12	9.00	1.95	6.17	4.09	9.83	2.17
	Post	12	10.75	2.70	5.33	5.80	7.92	2.57

shows the descriptive data of pitch, intensity, and duration parameters of the pre- and post-tests.

Wilcoxon matched pairs tests showed that the experimental group improved students’ sentence-level stress in the parameters of pitch, intensity, and duration after instruction, indicating that an overall training benefit was achieved after receiving VSM training. In terms of the experimental group, post-test scores were significantly higher than pre-test scores, respectively, for pitch (Z = 2.946, p < 0.001, r = 0.850), intensity (Z = 2.956, p < 0.001, r = 0.853), and duration (Z = 3.006, p < 0.001, r = 0.868). However, different results were presented by the control group. Post-test scores were not significantly higher than pre-test scores regarding the three parameters for pitch (Z = 1.852, p = 0.064, r = 0.535), intensity (Z = 0.482, p = 0.630, r = 0.139), and duration (Z = 0.671, p = 0.502, r = 0.194).

With respect to between-group comparisons, Mann–Whitney U tests of mean ranks instead of medians were carried out.

Table 2. Rank data in Mann–Whitney U tests for pitch, intensity, and duration scores in the pre- and post-tests.

Test	Group	N	Pitch		Intensity		Duration
			Mean Rank	Sig.	Mean Rank	Sig.	Mean Rank	Sig.
Pre-test	Experimental	12	14.54	0.152	9.79	0.059	10.42	0.146
	Control	12	10.46		15.21		14.58
Post-test	Experimental	12	16.46	0.006	15.38	0.044	15.79	0.022
	Control	12	8.54		9.63		9.21

presents the descriptive data of mean ranks concerning the pitch, intensity, and duration variables of pre- and posttests. No significant differences were found between the two groups’ pre-tests regarding the pitch parameter (U = 47.50, p = 0.152, r = 0.23), intensity parameter (U = 39.50, p = 0.059, r = 0.385), and duration parameter (U = 47.00, p = 0.146, r = 0.297). This indicated that there were no significant differences at the sentence level at the starting point of the intervention between the two groups. However, the running data shows very different results between the two groups concerning the post-tests. Much higher scores were achieved by the experimental group on all three parameters, respectively, pitch (U = 24.50, p = 0.006, r = 0.565), intensity (U = 37.50, p = 0.044, r = 0.411), and duration (U = 32.50, p = 0.022, r = 0.467), which indicated that VSM was more effective than the other regular strategies used for training Chinese sentence-level stress.

3.2. Post-Training Interview of VSM Intervention

In addition to clarifying the interpretation of the quantitative data, as discussed above, post-training interviews with the experimental group of students were also conducted to acquire a comprehensive overview of the VSM intervention. The data collected was freeform text, and was largely unprompted, being elicited by the three questions mentioned above.

Concerning the first question “Do you think the video self-modeling which was used in this research will contribute to the repertoire of the sentence-level skills of your Chinese learning?”, all the participants reported that VSM was helpful for improving their sentence-level stress skills.

The instruction was very helpful and well-organized. I like to have this one-to-one VSM instruction and I start to get used to having it every week.

Hearing their own perfect pronunciation was emphasized by almost all students, and this feeling of success helped to increase their motivation to train their speaking abilities:

I like to hear my own good pronunciation and use it to train my speaking. I feel it is easier to imitate and find the right flow of the sentences.

Awareness raising on sentence-level stress was also mentioned by most of the students and they emphasized the value of the VSM intervention:

I did not have any idea of sentence-level stress before the training, but I start paying attention on sentence-level stress now.

By using video self-modeling instruction, I feel my pronunciation is better than before, and I start paying attention to the stress and try to emphasize the stresses in speaking.

As to the second question “Are you satisfied to participate in the sentence-level stress training program by using video self-modeling?”, half of the students expressed their satisfaction with participating in the VSM intervention:

I am happy to participate in this training because I love to speak Chinese and want to improve my Chinese speaking.

Two students mentioned that the after-class activities took quite a long time to finish, and three students mentioned their dissatisfaction with their performance during the intervention.

Several comments were provided by the interviewed students concerning the third question “What comments do you want to add concerning the conducted training?” Six students pointed out that they would like to continue using this strategy, if possible, and three students shared their worries about the difficulties of using this strategy as it needed the participation of the teacher, the production of recordings and the editing of audios.

4. Discussion

4.1. Increased Scores for Chinese Sentence-Level Stress by Using VSM

This study examined the effectiveness of the VSM strategy and students’ perceptions of it, with a specific focus on Chinese sentence-level stress production. One of the main findings is that VSM allowed the participants in the experimental group to significantly improve their skills of pronouncing sentence-level stress. The improved production using students’ self-produced audio recordings in the current study adhere to previous findings [37,38] concerning the effects of VSM. In addition, this study on VSM enriches the literature by having a relatively large number of participants as compared to the other VSM studies, in addition to using a control group. Focusing on sentence-level stress also expands VSM instruction from ASD or general-discipline fields to foreign-language skill training. Previous research on CFL pronunciation instruction endorsed the use of technologies to enhance teaching Chinese pronunciation [65,66,67,68,69], but the absence of using students’ own good pronunciation as a training method constitutes a gap in the literature that the present study addresses. By applying VSM, students are able to listen to their own good pronunciation, and, as a result, it became easier for them to find the correct pitch level and imitate it. Listening to or watching one’s own perfect performance also contributed to increasing students’ confidence and motivation for learning this language skill [70,71,72,73]. This aspect was also emphasized by several participants from the experimental group in the post-training interviews. As such, the combination of the use of technology (e.g., smartphones, or computers) and self-voice audio input has proved to provide an exceptional experience for students in the context of their language training. Future research could further develop VSM instruction based on specific contexts and language skills.

4.2. Increased Scores for Chinese Sentence-Level Stress in All Three Parameters (Pitch, Intensity and Duration)

After ten sessions of training, the experimental group’s results in all three parameters suggest that VSM instruction had the general effect of improving all the prosodic features of Chinese sentence-level stress [13], as compared to the results from the control group. Students in the control group did show a slight improvement in the pitch pronunciation parameter, though less as compared to the experimental group. However, they did not improve in the other two sentence-level stress parameters of intensity and duration. This outcome is in-line with the results from previous studies focusing on CFL students’ stress development [14]. The latter authors explain this by looking at the weaker instructional strategies being adopted by the CFL teachers. The authors stress that CFL teachers lack specific pedagogical content knowledge about pronunciation teaching. They are also reported to master only limited knowledge of linguistic concepts concerning sentence-level stress. According to other authors, the results can also be explained by the adoption of inappropriate instructional pronunciation strategies that, in the end, have a negative influence on students’ speaking outcomes [74]. During their first year, students in our control group did receive extensive pronunciation training, but no longer during their second year. Though they still might be aware of pronunciation parameters—as reflected in their initial pre-test scores—they received less feedback and specific individual directions to further improve pronunciation. The curriculum for these second-year CFL students pays more attention to syntax, morphology, and pragmatics [75] and CFL teachers switch more to CFL communication-focused activities, such as presentations, group discussions, debates, drama performances, and narrative expression [76,77]. This is at the expense of correcting pronunciation mistakes. This result is also consistent with previous studies that oral speaking fluency is playing a more important role than accuracy in intermediate- and advanced-level students’ Chinese learning and teaching [13,76]. The results stress, at the same time, that pronunciation instruction should not be dropped during subsequent phases of CFL instruction.

Concerning both groups of students’ performance on the three parameters, as pitch level is the main focus of CFL teachers for instructing Chinese tones and prosody, the other two parameters are easily neglected by CFL teachers who lack specific educational training regarding Chinese prosody. Intensity and duration have been researched in previous studies, as perceptual cues for CFL students to locate the focus of the sentences in speaking, even when CFL learners have not detected certain prominent prosodies in the spoken sentences [78]. However, when CFL students are uncertain about pronouncing the appropriate intensity and duration of the stressed words, perceptual cues will certainly be affected. This also corresponds to the research results that students’ L2 intelligibility and comprehensibility are closely related to stress and other prosodic feature learning [79,80]. Combined with the experimental-group students’ interview answers, similar problems were also stressed by a few students, who stated that they were not taught relevant knowledge with regard to sentence-level stress. After the training, students could determine the prosodic features of sentence-level stress and were clear about what students need to pay attention to in their speaking or reading if they want to emphasize certain information. Therefore, there is a need for CFL language instructors to acquire enough knowledge on Chinese stress or other prosodic features and be aware of prosody training in the future.

5. Theoretical and Pedagogical Implications

As the present study and previous literature imply, intervention on sentence-level stress is indeed necessary for CFL teachers, researchers, and curriculum designers. This study extends the literature by involving a large number of students and examining the effect of VSM in the context of L2 pronunciation intervention for the first time. Significant improvement by comparing the pre-post scores within the experiment group and between the experimental group and control group strengthens the value of VSM intervention and expands the instruction strategies for training L2 students’ language skills. Being the first study using CFL students’ adaptive perfect pronunciation recordings to train their sentence-level stress to achieve better scores, it confirms previous research results that one can recognize one’s own recorded voice more accurately [81]. Therefore, using recordings of one’s own good pronunciation could certainly be a recommendable choice for L2 instructors or teachers to consider in instructing pronunciation and speaking skills. Researchers are encouraged to explore VSM interventions as a means of effectively helping L2 students’ communication skills, not only in the context of CFL, but also for other L2 contexts.

The literature on sentence-level stress instruction is also enriched in the CFL context by examining all three parameters. Post-test scores on the three parameters decreased in the sequence of pitch range, intensity range, and duration range. As pitch level is the most familiar prosodic item for CFL students in their pronunciation training, it is easier for them to reach the adequate pitch range. CFL students lack awareness of intensity, and this might relate to their weakness in identifying the stressed words. Therefore, CFL teachers should help students to develop their awareness of increasing the loudness of the stressed syllables at the beginning of their studies. The lowest scores on duration are consistent with previous research on CFL students’ stress: that CFL students tend to have longer duration than native speakers. Reasons for that could be that CFL learners need more time to process and organize the utterance [44], and the length of the stress in the sentence also influences CFL leaners’ duration when pronouncing it [43]. Similar results are especially emphasized in previous CFL pronunciation research: that stress instruction was widely ignored in the teaching [13,14,82,83,84]. Therefore, stress instruction and training should attract CFL teachers’ attention and be included in their teaching curriculum. Moreover, the use of technological media in the intervention and the significant improvement by the experimental group also strengthen the effectiveness of CALL in pronunciation instruction [7].

6. Limitations of This Study and Directions for Future Research

Although our findings are promising, some limitations should also be considered. First, the scope of the training group is relatively small. Although the effect size of the intervention is <0.05 for all three parameters, effect size differs within three parameters. The different background knowledge of the three parameters in both groups could have been more homogenous if participants in the two groups had been taught Chinese stress-related knowledge before the training. Second, the two groups in the study are not on the same language level and the number of male and female students is unbalanced. Previous studies have shown that CFL students are weak in intonation and stress across all language levels [14,63], meaning that the language-level difference did not have an additional impact on the intervention effect. However, different language levels might influence the intervention result on training other Chinese pronunciation items (such as tones or segments) or other language skills [85]. Therefore, it is recommended that future studies choose two groups of students from the same language levels to assess L2 language skills and balance the number of female and male students. In addition, there are notable concerns about the instructional reading materials. While the post-tests from the experimental group resulted in an improvement as compared to the control group, the materials chosen for training and testing were from the textbook and workbook used by the experimental-group students and the training sessions were inconsistent with the experimental-group students’ daily learning progress. Therefore, students from the experimental group were relatively more familiar with the sentences, and this might cause potential bias. It is uncertain if the effect of VSM intervention can also achieve success when using unfamiliar reading materials for both groups with regard to their knowledge of vocabulary and grammar. CFL researchers could attempt to tackle this issue in the future. Third, although positive feedback was received from the students, there is a lack of formal assessments of students’ competence of, and motivation for, using VSM. Ref. [32]’s social learning theory and [31] provide a theoretical basis that intervention studies should examine student’s level of confidence in his/her own competence [86] and their intrinsic motivation [87] for using the implemented strategy, as confidence and motivation are regarded as the two crucial factors of students’ development and success [88]. This issue could also be explored through explicit self-efficacy and self-determination assessment studies, to discover CFL students’ competence and motivation when using VSM. Fourth, in the present study, while only audio recordings were used in the VSM intervention, integrating audio and visual resources in language instruction has attracted more attention and proved to be effective in promoting learning [89]. Future research is recommended to present corrected tone contours on the screen together with audios by Praat [17,90], so that students can receive both auditory and visual training at the same time.

7. Conclusions

This paper set out to investigate the performance of CFL students exposed to VSM, by comparing the results of quantitative and qualitative data analysis with those of a control group. As shown in the discussion session, the results obtained in this study proved VSM to be beneficial to CFL learners’ Chinese sentence-level stress learning. It was found that the participants in the experimental group performed significantly better than the control group, with increased scores by the group using VSM across all three parameters. The positive outcome achieved using students’ own adaptive accurate audio recordings indicated an alternative and promising strategy for L2 students’ pronunciation learning. It sheds light on future approaches for implementing VSM in the instruction of other pronunciation items or other language skills in both CFL and L2 contexts. The positive results from the VSM intervention provide a basis for measuring all three parameters of Chinese sentence-level stress in the context of CFL. This study also demonstrates that CFL teachers should involve more comprehensive stress-related knowledge teaching in the curriculum.