Confusion and Chinese Character Learning

This study explores the relationship between the experience of confusion and the outcomes of Chinese character learning by learners of Chinese as a heritage language (CHL). Based on the claim that impasses triggering confusion can lead to deeper learning of conceptually difficult material, the study employed three impasse-driven tasks. The tasks were designed to trigger a state of cognitive disequilibrium in the participants. After first encountering the tasks in Session 1, the 117 CHL learners were given 1 week to resolve their impasse-driven tasks in their own time for 30 min (limited by the computer system). Afterwards, we announced the correct answer for them to look at for 3 days, and then Session 2 was administered 1 week after the feedback. The results of this study showed considerable improvement in differentiating near-homographs and homophones presented in a sentence, and correcting wrong Chinese characters. The study's results suggest that incorporating well-designed confusing content into Chinese learning may help CHL learners deepen their learning of Chinese characters.

Keywords: Confusion; Chinese characters; impasse-driven learning; homophone; near-homograph

Confusion is a prevalent emotion during complex learning. Previous studies have pointed out the inextricable relationship between confusion and learning outcomes (D'Mello et al. [16]; Lehman, D'Mello and Graesser [43]; Richey et al. [61]). It is often assumed that confusion has a significant impact on learning when learners confront complex learning tasks. Learning Chinese characters is always one of the biggest challenges for learners of Chinese as a foreign language (DeFrancis [17]; Fan, Tong and Song [21]). Chinese language teachers frequently observe that learners become confused when studying Chinese characters (Shen [67]). In Chinese, over 80% of characters are phonograms, which consist of a semantic radical providing information relating to meaning, and a phonetic radical which provides information about pronunciation (Kim and Shin [38]; Zhou [87]). Radicals refer to the smallest stroke patterns that can be recursively used as a meaningful unit, providing meaning or pronunciation clues, to form a character (Hong et al. [31]). The rich number and novel composition of radical units creates a particularly confusing task for Chinese language learners, especially in establishing a stable association between Chinese characters' form, pronunciation, and meaning (Chang et al. [9]).

Chinese has a massive number of homophones (e.g. 找 (zhǎo, meaning 'find') and 沼 (zhǎo, meaning 'marsh')) and near-homographs (e.g. 我 (wǒ, meaning 'I') and 找 (zhǎo, meaning 'find')) at the morphological level that lead to fundamental problems in learning Chinese characters (Li et al. [45]). Wu et al. ([80]) note that learners are often confused by certain Chinese character pairs, such as near-homographic pairs and homophonic pairs. Korean, European, and American learners tend to make more errors in writing near-homograph characters, while Japanese learners made more errors in writing near-homophone characters (Kim, Christianson and Packard [39]; Wu et al. [80]). From the perspective of Chinese language processing, Yum et al. ([82]) observed that slow learners relied more on holistic pattern processing of Chinese characters, whereas fast learners kept track of a limited set of radicals and their relationships.

In relation to heritage language study, many studies have indicated that the home literacy background of Chinese heritage language (hereafter, CHL) learners has a significant effect on certain of their language skills such as their speaking, listening, grammar, and vocabulary breadth, but not on their knowledge of Chinese characters, including orthographic knowledge, semantic knowledge of radicals, and phonetic knowledge of radicals (Cheng [11]; Koda [40]; Loh, Liao and Leung [46]; Zhang and Koda [84]). In this study, the participants were CHL learners who had studied in the educational system of independent Chinese schools in Malaysia before coming to Taiwan. In the Malaysian system, the simplified Chinese writing system is adopted. It is reasonable to assume that learners' orthographic knowledge of simplified Chinese characters may facilitate recognition and production of traditional characters. However, Taiwanese Chinese language teachers teaching such students have observed that background knowledge may lead to adverse effects. These students tend to experience great confusion and make errors frequently when using traditional Chinese characters. Very limited research on this issue has so far been conducted, and it is not known whether CHL learners can benefit from their orthographic knowledge of simplified characters, or whether triggering the emotion of confusion can benefit them when learning traditional Chinese characters.

This study seeks to better understand if the confusion approach can help learners to improve their retention of Chinese characters. The participants were CHL learners from Malaysia and the study set out to investigate whether the impasse-driven tasks developed for this study could bring about learning gains. According to Wu et al. ([80]), there are three types of character pairs that Chinese language learners are likely to find particularly confusing: namely (1) non-homograph and homophone (e.g. 小 (xiǎo, meaning 'small') and 筱 (xiǎo, meaning 'thin bamboo')), (2) near-homograph and homophone (e.g. 蛹 (yǒng, meaning 'pupa') and 俑 (yǒng, meaning 'tomb figure')), and (3) near-homograph and non-homophone (e.g. 膽 (dǎn, meaning 'guts') and 擔 (dān, meaning 'burden')). This study also considers the potential moderating impact of several cognitive variables, such as learning styles, language learning anxiety, and cognitive load.

Learning style is defined as the manner in which learners characteristically process information, knowledge, or experience. Keefe ([36]) suggested that learning styles serve as relatively stable indicators of how learners perceive, interact with, and respond to the learning environment. Some studies have used learning styles as an instrument to predict the learning outcomes of foreign language learners with various demographic backgrounds (Gohar and Sadeghi [23]). As observed by Yum et al. ([82]), Chinese language learners rely on different approaches to process Chinese characters, resulting in different learning effects. This study explored whether learning styles influenced the learning brought about by tasks requiring differentiation of near-homographs and homophones.

Foreign language learning anxiety is a distinct emotion of self-perceptions, beliefs, and feelings in a language learning setting (Teimouri, Goetze and Plonsky [74]). Anxiety has been found to interfere with the process of learning a language (Alias and Rashid [1]; Pekrun et al. [55]). Some studies conclude that language learning anxiety results in a large number of negative outcomes (e.g. Dewaele et al. [18]; Pekrun et al. [55]), while others assert that anxiety is positively correlated with learning (e.g. Brown [7]; Oxford and Ehrman [54]). This study examined the relationship between the impasse-driven tasks and anxiety in order to know whether the tasks trigger learners' anxiety resulting in low learning outcomes.

Cognitive load refers to the amount of information exceeding a person's cognitive capacity in working memory (Sweller, Ayres and Kalyuga [71]). Cognitive load is differentiated into extraneous, intrinsic, and germane cognitive load (Sweller [70]), in which the intrinsic cognitive load is negatively associated with cognitive performance and learning achievement (Hong et al. [32]; Schnotz and Kurschner [66]). This study investigated whether the impasse-driven tasks would trigger CHL learners' intrinsic load, thus affecting their learning outcomes.

In this study, we examined the relationship between tasks deliberately triggering confusion and impact on traditional Chinese character learning, taking learners' learning styles, language learning anxiety, experience of cognitive load and relative Chinese language proficiency into consideration.

To investigate how to improve efficacy in Chinese character learning, this study investigated three impasse-driven tasks designed to trigger the state of confusion in learners. The study was designed to test five hypotheses, which are discussed in the following subsections.

Confusion has been identified as one of the learning-centered emotions experienced by learners during educational activities (Calvo and D'Mello [8]; Rodrigo and Baker [64]). Confusion usually accompanies a state of cognitive disequilibrium that is triggered when there is an ongoing mismatch between incoming information and prior knowledge, which results in learners being unsure about how to process the information (D'Mello et al. [16]). Recent researchers have proposed that confusion, as one of the prevalent emotions during complex learning, can in fact create opportunities to develop a deeper understanding and build durable memories, achieving deeper learning (Arguel et al. [3]; Baker et al. [5]; Graesser et al. [26]; Graesser et al. [27]; Richey et al. [61]). Some studies assert that confusion is significantly and positively correlated with learning outcomes (Craig et al. [13]; D'Mello and Graesser [14]; Graesser et al. [25]). It should be clarified that confusion itself does not cause learning gain or lead to deeper learning. Rather, the cognitive activities accompanying the emotion of confusion, such as striving to restore cognitive disequilibrium, and solving impasses, have a direct relationship with learning gain (Graesser and Olde [24]; Stein, Hernandez and Trabasso [68]).

Impasse is considered one of the best means to trigger confusion (D'Mello et al. [16]; Sawyer [65]). Specifically, impasse describes a situation in which the learners cannot solve a given task due to a lack of necessary knowledge. Impasse-driven learning theory suggests that learning can be greatly facilitated by means of successful resolution of impasses (Blumberg, Rosenthal and Randall [6]; VanLehn et al. [76]). Indeed, according to VanLehn et al. ([76]) learning gain from complex learning is rare when learners do not reach an impasse. Once an impasse is detected and confusion is experienced, the individual needs to engage in effortful cognitive activities in order to resolve the confusion, which is what leads to deeper learning, more durable memory, and more successful retrieval (Heitzmann, Fischer and Fischer [30]; Reindl, Tulis and Dresel [60]).

Briefly, this body of work suggests that confusion can be beneficial to learning if it is appropriately induced, regulated, and resolved by instructors or designed activities. Thus, this study hypothesises that learners can process learning materials more deeply through resolving their confusion; conversely, learning gain will be rare if learners do not experience confusion. The first hypothesis is based on the impasse-driven theory of learning, which suggests that experiencing confusion provides learners with access to deeper learning.

H1. Learning is significantly improved after experiencing confusion.

Learning style refers to the underlying approach employed by individuals to perceive and respond to new information in learning situations (Riding and Cheema [63]; Riding and Douglas [62]; Zarrabi [83]). MacKeracher ([48]: 71) defined learning styles as 'the characteristic cognitive, affective, social, and physiological behaviours that serve as relatively stable indicators of how learners perceive, interact with, and respond to the learning environment.' In other words, an individual's learning style is his or her preferred mode of learning in terms of the sensory modality by which the learner receives new information.

A popular categorisation of perceptual learning style preferences divides people into visual, auditory and kinesthetic learners (Gilakjani [22]). Visual learners typically prefer learning through visual information, such as images, maps and other non-verbal information. Auditory learners prefer acquiring knowledge through oral-aural formats, such as lectures and discussion. Kinesthetic learners prefer learning by doing, such as hands-on activities, field trips and exhibits. It was proposed that a positive relationship exists between visual perception skills and reading ability (Memiş and Sivri [50]). The current study's impasse-driven tasks were designed to have participants detect and correct errata in use of Chinese characters in a meaningful context. The study assumes that the ability to detect errata is directly correlated with participants' visual perception ability, leading to the second hypothesis.

H2. There is a significant difference between the Chinese character learning of visual and nonvisual learners.

Learner anxiety is one of the most important affective variables in learning a foreign language (Tallon [73]). Foreign language anxiety is a pervasive phenomenon associated with feelings of fear, which is defined as 'a distinct complex of self-perceptions, beliefs, feelings, and behaviours related to classroom language learning arising from the uniqueness of the language learning process' (Horwitz, Horwitz, and Cope [35], 128). The emotion of anxiety may negatively affect learners' self-esteem and self-confidence, influencing their learning outcomes (Elkhafaifi [19]; MacIntyre and Gardner [47]). For instance, Horwitz's ([34]) study revealed that students with high levels of anxiety received lower course grades than the students with lower levels of anxiety. Khajavy et al. ([37]) examined Iranian learners' willingness to speak in English, and indicated that learning English anxiety negatively affected learners' willingness to speak. Recently, an increasing number of studies have attempted to reduce learners' anxiety while learning a foreign language by integrating a variety of learning strategies (Hong et al. [33]; Verkijika and De Wet [77]).

The Chinese language is a relatively difficult foreign language for most learners due to its character-based writing system and tonal nature, leading to learners' potentially high levels of anxiety (Zhao, Guo and Dynia [86]). However, studies on Chinese language learners' foreign language anxiety are not plentiful, and certainly, little research has addressed the relationship between confusion and language learning anxiety. Therefore, this study proposes a third hypothesis.

H3. Learners' performance in confusion-based Chinese character learning activities differs significantly according to their degree of language learning anxiety.

In language learning settings, a state of confidence and certainty is generally preferable over a state of uncertainty and confusion. When learners experience confusion, teachers tend to try to alleviate the confusion by immediately providing scaffolds, especially for low-achieving learners. Lee, Shen and Tsai ([42]) observed that low-achieving students often behave passively in learning settings, and lack the confidence to carry out learning. It has been noted that letting low-achieving learners experience persistent failure or confusion does not encourage them to learn (Vygotsky [78]). Tomasello and Herron ([75]) also suggested that learners' learning efficacy can be enhanced when they receive immediate feedback.

In impasse-based activities which are designed to let the learners encounter confusion, it might be the case that only high-achieving learners will experience confusion, whereas low-achieving learners will simply fail to detect the impasse and thus will not experience confusion (VanLehn et al. [76]). Therefore, this study proposes a fourth hypothesis to explore whether learners with different levels of Chinese language proficiency will perform differently on the impasse-driven learning tasks.

H4. There is significant difference between high- and low-achieving learners on the impasse-driven tasks.

Cognitive load is defined as 'the amount of mental effort required to locate specific information and to understand how this information is oriented within a larger information source' (Eveland and Dunwoody [20]: 57). Based on cognitive load theory, many studies on language acquisition have addressed the possibility that a heavy cognitive load has negative effects on language learning (Chen and Chang [10]; Clark, Nguyen and Sweller [12]; Wang [79]), and some studies have focused on how to minimise learners' cognitive load (Moreno [51], [52]).

Sweller ([69]) distinguished between three different types of cognitive load. Intrinsic cognitive load refers to the demand on working memory capacity imposed by the learning material. Extraneous cognitive load refers to the mental load associated with the manner in which the instruction is designed. Germane cognitive load is concerned with the acquisition of knowledge; it refers to the mental resources or effort devoted to acquiring schemata in long-term memory. Schnotz and Kurschner ([66]) indicated that intrinsic cognitive load is negatively associated with cognitive performance and learning achievement. In the current study, which measured only intrinsic cognitive load, the fifth hypothesis focuses on whether impasse-driven tasks increase learners' intrinsic cognitive load and result in low learning efficacy.

H5. There is significant difference between the performance of learners experiencing high- and low-cognitive load on the impasse-driven tasks.

The aim of the current study was to investigate whether learners would show learning gains in their knowledge of Chinese characters after engaging in impasse-driven tasks designed to trigger confusion. It also examined four learner variables – visual vs. non-visual learning style, language learning anxiety, Chinese language proficiency and experience of cognitive load – to see whether these would moderate any relationship between the pedagogic tasks and learning gains.

The participants were 117 CHL learners from Malaysia enrolled in a Chinese language programme in Taiwan. Their ages ranged from 17 to 19 years old, with a mean age of 18. Their Chinese language proficiency ranged from CEFR B1 to B2.

In communication with the participants, we referred to the impasse-driven tasks as 'character hunting' as the tasks involved detecting and correcting erroneous characters. The three tasks were designed to target near-homographs and/or homophones as shown in Table 1. In Task A, character pairs sharing the same pronunciation but differing in orthographic form, namely non-homograph and homophone pairs, were chosen as task items. e.g. 報 (bào, means 'report') and 抱 (bào, means 'hug'). In Task B, character pairs sharing the same pronunciation but slightly differing in orthographic form, namely near-homograph and homophone pairs, were chosen. e.g. 蛹 (yǒng, meaning 'pupa') and 俑 (yǒng, meaning 'tomb figure'). In Task C, the character pairs differing in pronunciation and slightly differing in orthographic form, namely near homograph and non-homophone pairs, were chosen, e.g. 膽 (dǎn, meaning 'guts') and 擔 (dān, meaning 'burden').

Table 1. Tasks.

As shown in Table 1, the correct target character was substituted by its incorrect 'mate' in a sentence. For instance, the character 蛹 (yǒng, meaning pupa) in Task B was substituted by its near-homograph and homophone mate 俑 (yǒng, meaning tomb figure), leading to a confusing sentence. The participants were required first to detect and circle the incorrect character helped by the clues provided by the meaning of the sentence, and then to write down the correct character, as shown in Figure 1. In the process of trying to understand the discourse fragment, and then detecting and correcting the target characters, the emotion of confusion was likely to have been triggered.

PHOTO (COLOR): Figure 1. Detecting and correcting a Chinese character error.

Each task included 30 items, which were chosen in consultation with two Chinese language instructors to be in accordance with the Band C level of the TOCFL.[1] All three tasks were loaded onto an online assessment platform – online assessment for science literacy (OASL) – which combines multimedia presentation and provides the function of 'Draw on a Picture' which allowed participants to circle an incorrect character and then to write down the correct character in the assigned square, as shown in Figure 1.

We developed three research instruments in order to measure our participants' learning styles, anxiety and cognitive load. The learning styles scale focused on visual perceptual preference, and was based on Reid's ([59]) perceptual learning style preference questionnaire (PLSPQ). The PLSPQ consists of 30 items which were adapted and translated into Chinese. The language learning anxiety scale was based on Horwitz et al.'s ([35]) foreign language classroom anxiety scale (FLCAS); eight items were chosen from the 33 items of the Chinese version FLCAS and the description of items was modified to meet the topic of this study (i.e. 當我在玩「尋字遊戲」時, 我感到很緊張和困惑。 'I get nervous and confused when I am playing the character hunting game). To measure intrinsic cognitive load, the study employed participative self-reported perception with eight items from the Chinese version of the NASA task load index (Hart and Staveland [29]), in which the description of items was modified to refer to the confusion-based activities (i.e. 玩尋字遊戲耗費我很多心理和知覺上的心力 'The character hunting game required a lot of mental and perceptual activity to finish it.') All the instruments used 5-point Likert-type scales scored on a continuum ranging from 'strongly disagree' to 'strongly agree.'

The learning style scale was administered before Session 1, while the language learning anxiety and the intrinsic cognitive load scales, which related to the impasse-driven tasks, were administered after Session 2.

To assess the participants' existing language proficiency, their summative assessment grades in their Chinese language course were used to evaluate their language proficiency. This assessment was designed based on the content learned in the course during one semester.

This experiment was administered in three Chinese language classes over a 2-week period. All participants were informed of the rules and purpose of the study. Each participant completed the same task (Task A, B or C) twice via the online platform described above, once in Session 1 and once in Session 2. The task had to be completed within one hour. After first encountering the task in Session 1, the participants were then given one week to resolve the impasses experienced in their own time via a review of the same task they had looked at in Session 1, limited to 30 min' access. The review was designed to remain confusion and to trigger active resolution of impasses. Afterwards, the feedback, that is the tasks with the correct answers, was supplied to look at for 3 days online, aiming to help participants restore their mental disequilibrium. The participants had to log into the online platform and check the answer in their own time. Then, Session 2 was administered one week after this feedback was provided.

The repeated measures design may lead to the possibility of a practice effect, while the exposure to the task does not guarantee the learning gain for two reasons. First, according to the power law of practice, skill or performance may increase rapidly at first, but later, even minor improvement takes considerable effort, especially for complex tasks (Newell and Rosenbloom [53]). Second, the impasse-based tasks involved two cognitive processes: detection and correction. In Session 2, the performance of detection may have benefitted from a practice effect due to exposure to text including these characters, but the performance of correction is unlikely to have been influenced due to the complex and confused orthographic knowledge of traditional Chinese characters.

This study used SPSS 22 software to perform analysis of variance. Independent sample t-tests and paired-samples t-tests were conducted to measure participants' performance.

In the beginning, there were 150 students participating in this study. They were equally divided into three groups. Each group was assigned to complete one of the three impasse-driven tasks. However, a number of participants were excluded due to failing to complete the instruments, skipping the review process, ignoring the online feedback, or absence in Session 2. Thus, data from only 117 participants were useful for the data analysis, of which 46 participants completed Task A, 38 completed Task B, and 33 completed Task C. Table 2 shows the Session 1 mean scores of the three tasks: Task A (M = 79); Task B (M = 79.14), and Task C (M = 70.90). The three tasks, Tasks A, B, and C, were tested to ensure that they were of the same degree of difficulty. According to the testing scores, there were no significant differences among the three tasks.

Table 2. Performance on the Session 1 and Session 2.

Table 2 shows the scores of the first and second time taking the same test. Table 3 shows that the scores of sessions 1 and 2 had no significant differences through statistical verification (p >.05). The overall mean score for Session 2 (M = 95.64) was higher than that for Session 1 (M = 76.35). Divided into Tasks A, B and C, the mean of A is from 79 to 95.54, B is from 79.14 to 95.79, and C is from 70.90 to 95.60. These means indicate that all three groups made considerable progress in the learning of Chinese characters.

Table 3. Task mean score differences.

To examine whether differences in learning styles, namely, visual versus nonvisual learning preference, influenced the participants' performance on the tasks, an independent samples t-test was conducted. Using the results of the learning styles questionnaire, visual and non-visual learners were identified as those whose ratings were in the highest 27% (visual learners) and the lowest 27% (non-visual learners). Differences between these participants' respective performances were then evaluated. The results are presented in Table 4. The independent samples t-tests showed no significant differences in the scores of the visual learners and non-visual learners on any of the tasks.

Table 4. Session 1 Performance of visual and non-visual learners.

These results suggest that having a visual or a non-visual learning style preference did not influence participants' performance in the impasse-driven Chinese character learning tasks. As can be seen from Table 4, the non-visual learners in fact showed slightly greater improvement than the visual learners on the confusion-based tasks (18% vs. 21% improvement), but this difference was not statistically significant.

To explore whether the learning gains made by participants from the two different learning styles on the different tasks were significant, a paired-samples t-test was conducted. As shown in Table 5, the visual learners demonstrated statistically significant improvement in Session 2 on Task A (p =.011) and Task C (p =.007) but not on Task B (p =.064). The non-visual learners performed significantly better on all three tasks.

Table 5. Learning Gains of Visual and Nonvisual Learners.

1 * p <.05, ** p <.01.

Task A and Task B were designed to trigger confusion regarding the form of the characters, while Task C was designed to trigger confusion regarding the sound of the characters. The study therefore assumed that visual learners who prefer to process information in a visual way would perform better than nonvisual learners on Task A and Task B; however our results did not provide evidence to support this.

It has been noted that language learning anxiety is negatively correlated with language learning gain (Elkhafaifi [19]). To explore whether the 'character-hunting' tasks would trigger learners' anxiety and potentially negatively influence learning gains, we compared the performance of participants whose ratings on the language learning anxiety scale were in the highest 27% (high anxiety learners) and in the lowest 27% (low anxiety learners). Table 6 shows the results from an independent samples t-test used to identify any significant difference in mean scores of the high- vs. low-anxiety learners in each of the three tasks at Session 1. None of the tasks showed a statistically significant difference between the two groups of learners These results suggest that participants' language learning anxiety had no association with their performance on the impasse-driven tasks at Session 1.

Table 6. Session 1 performance of high- and low-anxiety learners.

To examine the hypothesis that levels of language learning anxiety may affect learners' gains from the impasse-driven tasks, a paired-samples t-test was conducted to compare the learning gains of high-anxiety learners with those of low-anxiety learners. As shown in Table 7, both the high-anxiety and the low-anxiety learners showed significant improvement in their Session 2 scores on all three tasks.

Table 7. Learning gains of high- and low-anxiety learners.

2 * p <.05, ** p <.01.

For the high-anxiety learners, the overall mean score increased by 20% (18.95 points from M = 76.78 at Session 1 to M = 95.73 at Session 2). A similar overall increase was seen among the low-anxiety learners with an increase of 22% (20.98 points, from M = 74.33 at Session 1 to M = 95.32 at Session 2).

These results suggest that the impasse-driven tasks could be helpful for both high- and low-anxiety learners; in this study, anxiety level did not appear to affect learners' ability to make gains in Chinese character learning after experiencing confusion.

Traditionally, persistent confusion is expected to accompany poor learning (D'Mello and Graesser [15]). It has been suggested that it is not reasonable to let low-achieving learners experience persistent failure or confusion (Vygotsky [78]). This study thus hypothesised that higher-achieving learners might perform better than lower-achieving learners. To test the hypothesis, an independent samples t-test was conducted to compare the scores of the highest and lowest achieving participants. Based on the summative assessment from their Chinese language course, results from the highest achieving 27% and the lowest achieving 27% were compared.

As shown in Table 8, the high-achieving learners performed significantly better on Task A and B than the low-achieving learners (respectively, t = 3.738, p =.001 and, t = 4.353, p =.001). However, in Task C, there was no significant difference between the high-achieving and the low-achieving learners (t = 373, p =.713). To further explore whether level of Chinese language ability influenced learning gains made by participants through the impasse-driven tasks, a paired-samples t-test was conducted to compare mean scores on Session 1 and Session 2 for both groups. The results are presented in Table 9. Both high-achieving and low-achieving participants for all three tasks had significantly higher mean scores at Session 2 compared with Session 1. Overall, the low-achieving participants increased their scores by more than the higher-achievers: by 30% (28.34 points from M = 66.70 to M = 95.43) compared with 15% (14.77 points from M = 82.98 to M = 97.75). These results suggest that although Chinese language ability may have influenced the performance of participants in Session 1, it did not appear to impact on learning gains in Chinese character learning through impasse-driven tasks measured at Session 2.

Table 8. Session 1 performance of high- and low-achieving learners.

3 * p <.05, ** p <.01.

Table 9. Learning gains of high- and low-achieving learners.

4 * p <.05, ** p <.01.

In order to check for any differences in the performance of learners who might have experienced the impasse-driven tasks as representing a high intrinsic cognitive load compared with those experiencing a low intrinsic cognitive load, an independent samples t-test was conducted first on Session 1 scores, as shown in Table 10. Results showed that for all three tasks, there were no significant differences between the two types of participants in their performance at Session 1. However, we note that the mean scores were higher for the low cognitive load participants on Tasks A and B, while the contrary was true for Task C with those experiencing high cognitive load achieving a slightly higher – though not significantly so – mean score.

Table 10. Session 1 performance of high and low cognitive load learners.

To check whether experience of cognitive load played a role in influencing learning gains from the impasse-driven tasks, a paired-samples t-test was conducted to compare the scores at Session 1 and Session 2 of the high cognitive load participants and the low cognitive load participants. As shown in Table 11, both groups showed significantly higher scores at Session 2 compared with Session 1 for all tasks. Comparing the overall mean scores from the two sessions, the higher cognitive load learners increased their mean scores by 15% (i.e. 21.23 points, from M = 73.81 to M = 95.04) while the lower cognitive load learners increased by 30% (18.45 points, from M = 78.50 to M = 96.95). Although individual learners experienced the impasse-driven tasks as representing varying levels of intrinsic cognitive load, this factor did not appear to work significantly against learning gains.

Table 11. Learning gains of high and low cognitive load learners.

5 * p <.05, ** p <.01.

For CHL learners, the logographic nature of the Chinese characters is often the most difficult part of mastering Chinese. Homographic and homophonic characters are particularly confusing. Instead of providing immediate feedback on learners' errors in tasks involving such characters, the current study offered delayed feedback with the goal of letting learners experience some confusion, which they then had to try to resolve. Based on the ideas of Baker et al. ([5]) and Richey et al. ([61]), this study holds the view that confusion can create opportunities for achieving deeper learning in which learners develop a deeper understanding and build durable memories in their learning of traditional Chinese characters.

Our first research focus concerned participants' performance on Session 1 and learning gains made by Session 2 after participating in the impasse-driven activities. Session 1 results showed that participants completing Task A (non-homograph and homophone; M = 79.00) and those completing Task B (near-homograph and homophone; M = 79.14) achieved slightly higher scores (though statistically significantly) than participants completing Task C (near-homograph and non-homophone; M = 70.90).

This result could suggest that errors with homophonic character pairs would slightly easier for the participants to detect and correct than errors with non-homophonic character pairs. This finding would not support the concept of a 'homophone effect' (Pexman et al. [57]; Pexman and Lupker [56]). Based on findings of longer latencies for homophones than for non-homophones in a lexical decision task, the concept of a homophone effect suggests that when a language user encounters a word with a homophone (e.g. maid), the phonological representation will activate both the correct orthographic representation (e.g. maid) and the orthographic presentation of the homophone mate (e.g. made) (Pexman et al. [58]). The homophone effect is evidence that phonology mediates access to word meanings. In this study, homophonic and non-homophonic characters were represented in a meaningful context so that information in the discourse could help participants narrow the possible meanings of the target characters. As a result, the phonological representation did not appear to interfere with the participants' performance of errata detection and correction. Instead, the phonological representation may have helped participants activate the correct orthographic representation in the discourse.

Regarding learning gains, there was significant improvement on all tasks at Session 2. The improvement was greatest for Task C participants, who had the lowest scores at Session 1. These results support the argument that confusion can be associated positively with learning gain (Arguel et al. [4]; Arguel et al. [3]; Craig et al. [13]; D'Mello and Graesser [14]; Graesser et al. [25]). After being confused by homophonic mates or near-homographic mates of target characters and then resolving their confusion, the participants were arguably able to consolidate their memory retention and deepen their perception and understanding of the target characters.

The impasse-driven tasks were based on detecting and correcting errata. The process of detection is directly related to visual perceptual skills, so Hypothesis 2 assumed that different learning styles (i.e. visual vs. non-visual) would significantly influence the participants' performance and learning gain on the impasse-driven tasks. However, there were no significant difference between visual learners and non-visual learners at Session 1, while both groups showed significant improvement after completing the impasse-driven tasks at Session 2. Some previous studies have found a positive relationship between visual perception skills and reading achievement (e.g. Hsu, Hwang and Chang 2013; Memiş and Sivri [50]), while others claim that no association exists between these two factors (e.g. Kollöffel [41]; McBride-Chang and Ho [49]). The findings of this study show no evidence of a link between visual perception skills and performance in detecting or correcting errata of characters in a meaningful context.

The impasse-driven tasks were designed to trigger the emotion of confusion, while learners' learning anxiety may be triggered simultaneously. Many studies have shown that anxiety has a negative effect on learning a second language (e.g. Elkhafaifi [19]; Pekrun et al. [55]; Tallon [73]). Therefore, this study hypothesised that language learning anxiety would negatively influence participants' learning gains in the impasse-driven tasks. The results of Session 1 showed no significant difference between high- and low-anxiety learners in Tasks A, B and C. This may give some indication that the anxiety triggered by the impasse-driven tasks did not influence the participants' performance. The high- and low-anxiety learners made considerable progress in the study.

Knowledge of Chinese characters is fundamental to mastering the Chinese language. This study assumed that participants' Chinese language ability would influence their performance on the impasse-driven Chinese character tasks. According to the results of Session 1, the high-achieving participants performed significantly better than the low-achieving participants on Task A (non-homograph and homophone) and Task B (near-homograph and homophone); we interpret this as suggesting that the high-achieving participants were able, at the outset, to make better use of phonological information as a clue to detect errata than the low-achieving participants. However, by Session 2, both low-achieving and high-achieving participants showed significant improvement, with the low-achieving participants showing greater gains than the high-achieving participants. These results do not support the argument that letting low-achieving learners experience confusion leads to poor learning outcomes (VanLehn et al. [76]).

This study further hypothesised that the participants who experienced the impasse-driven tasks as imposing a low cognitive load would perform better than the participants who experienced the tasks as imposing a high cognitive load. The results of Session 1 showed no significant difference between these two groups of participants, although, it is interesting to note that those who experienced a high cognitive load performed slightly better than those who experienced a low cognitive load on Task C which appeared to be the most difficult. This result may hint at the fact that a reduction in cognitive load is not always beneficial for learning (Leppink and van den Heuvel [44]). An appropriate cognitive load can maximise an individual's long-term performance and lengthen their memory retention (Zhang et al. [85]). At Session 2, all participants, regardless of whether they felt the tasks imposed a high or low cognitive load, showed significant improvement.

This study introduced a novel pedagogic approach by making use of tasks that aimed to trigger confusion to help to CHL learners distinguish similar Chinese characters. Many studies have suggested that confusion should be resolved immediately in a learning setting by providing instant feedback. The pedagogic implication of this study is to highlight the potentially positive impact of confusion on learning gain, and to show that impasse-driven tasks and delayed feedback may well be beneficial for learning Chinese character. Further, in contrast to approaches focusing on the internal structure of a Chinese character, this study suggests that the clues from a meaningful context can help learners to infer the meaning and the usage of a character in context. Our results also confirmed that the impact of the impasse-driven tasks did not appear to be not significantly influenced by individual learner variables such as visual vs. non-visual learning style, anxiety, experience of cognitive load or existing Chinese language proficiency.

Regarding the limitations of this study, the results are necessarily limited to CHL learners in the Asian context who had prior knowledge of simplified Chinese characters. It has been noted that learners with a high level of prior knowledge are more likely to persist with an impasse-based task than those with a low level (Arguel et al. [2]). Graesser ([28]) proposed the 'zone of optimal confusion' and suggested that it is important not to have too little or too much difficulty. This study did not consider learners with a low level of Chinese orthographic knowledge. Thus, future research is needed in order to compare samples from learners from different contexts and/or with different levels of prior knowledge.

While designing the confusion impasse-driven tasks, this study focused on triggering confusion through the form of Chinese characters, but did not take the radical position within the Chinese characters into consideration. Radical position regularity is a type of orthographic knowledge that encompasses the rules of the specific positions of radicals in Chinese characters (Xu et al. [81]). It has been noted that radical position regularity is important to the Chinese orthographic knowledge of CSL learners (Taft, Zhu, and Peng [72]). Accordingly, future research should consider including radical position as a parameter to examine any relationship with confusion and Chinese character learning.

No potential conflict of interest was reported by the author(s).