Introduction

Complexity, accuracy and fluency (CAF) are effective and thorough measures of L2 writing performance (Housen and Kuiken, 2009). Complexity is often defined as the capacity to use a variety of sophisticated L2 structures and vocabulary, accuracy as the capacity to produce language that is similar to the target and free from errors (Housen et al., 2012), and fluency as writing the text in an easy-to-read manner without any elements that would cause the reader to pause while reading (Atasoy and Temizkan, 2016). The CAF have long been popular indexes for assessing writing performance, but the extant research focuses on the analysis of language production while ignoring language input.

Coupling language input with output, the xu-argument based continuation writing task has received a lot of attention. Wang (2012) formally proposed the continuation writing task, which requires L2 writers to read the incomplete input text and continue it creatively, coherently, and logically. This task has gained substantial interest from second language academics because it can provide the alignment effect and increase the effectiveness of learning a foreign language. Great achievements have been made in the research on verifying the learning promotion effect of the continuation tasks. In recent years, researchers have mainly explored the approaches to improving the learning promotion effect of continuations, such as genre (Zhang and Zhang, 2017), task complexity (Zhang and Jiang, 2020), input enhancement (Qiu and Wang, 2022), etc. However, related studies to date have merely dealt with a single factor, failing to investigate the effects and interaction of multiple factors on L2 learners’ performance in the continuation tasks.

Few studies have explored the effects of input enhancement and genre on the continuation writing task and the interaction between the two factors. Input enhancement refers to the visual highlighting of specific language forms in the form of textual enhancement, such as italics, boldface, underline, capital letters, etc., so as to arouse learners’ attention to these forms (Smith, 1991, 1993). Related studies in the continuation tasks have proved that input enhancement can draw students’ attention and improve the learning promotion effect (Xin, 2017; Wang, 2019; Qiu and Wang, 2022). As all related studies used the story continuation tasks, it is unclear whether different genres have an impact on how input enhancement promotes learning. Genre is a type of discourse designed to achieve a specific communicative purpose, with specific structures, contents, and language features (Qin, 1997). As the change in the genre of continuation writing tasks brings changes not only in writing style but also in the accompanying context and language imitation model, the learning promotion effect of input enhancement in the continuation tasks of different genres needs further exploration.

In light of these issues, the present study adopted a two-factor mixed experimental design to explore the effects of input enhancement and genre on Chinese EFL learners’ writing performance in the continuation writing task and the interaction between input enhancement and genre.

Literature review

Studies of CAF in L2 writing and the continuation writing task

L2 writing is a complex process and has long been the primary concern of second language acquisition research. As writing in a second language involves sophisticated cognitive processes (Xu and Qi, 2017), it is seen by L2 learners as being substantially more difficult to acquire than other language skills. In addition, learning to write in English could be challenging because learners need to consider task achievement, coherence and cohesion, lexicon, and grammatical range and accuracy (Fathi and Rahimi, 2020). Consequently, the subject of how to assess writing performance has always been crucial. According to Skehan (1998) and Ellis and Barkhuizen (2005), L2 writing proficiency is multi-componential and may be adequately and thoroughly assessed by the writing CAF. L2 researchers can, therefore, assess and analyze learners’ writing performance more precisely by taking into account the writing CAF.

Skehan (1996, 1998) introduced a proficiency model that brought complexity, accuracy, and fluency together for the first time. CAF have attracted wide attention in the exploration of written language features. Extant studies have used CAF to explore the development of L2 writing competence (Yi, 2016; Plakans et al., 2019; Yang and Kim, 2020; Liu et al., 2022). These studies demonstrate that CAF are a crucial gauge of L2 learners’ language proficiency underlying their performance and can be used to assess both the quality of language output (Yang and Kim, 2020) and the progress in language learning (Housen and Kuiken, 2009). However, relevant studies mostly focused on the analysis of language output, and the effects of input on production deserve further exploration.

The xu-argument based continuation writing task, proposed by Wang (2012), is a key method for increasing the effectiveness of learning foreign languages. It closely couples language input with output, language imitation with creation, and language learning with use. Existing studies have explored the influence of input complexity (Xin and Li, 2020; Wang and Cai, 2022), topic familiarity (Gu et al., 2022), iterative continuation task (Miao and Wang, 2022), input enhancement (Qiu and Wang, 2022; Zhang and Li, 2023), pre-task planning (Xie and Zhu, 2023) on at least one of the writing CAF of continuations. Among the above-mentioned studies, only Gu et al. (2022), Miao and Wang (2022), Qiu and Wang (2022), and Xie and Zhu (2023) explored all three writing CAF in continuations. As the majority of studies on writing CAF of continuations concentrate on a single factor, there is still a research gap regarding the impacts of multiple factors on writing CAF of continuations.

Input enhancement and genre

Schmidt (1990) proposed that noticing is a prerequisite for acquisition, and conscious attention is the key to the conversion of input into intake. Long (2015) argued that attention to linguistic forms can be spontaneous, elicited, or even expressed. Input enhancement, the method of highlighting language forms from visual features such as different typography and fonts to attract learners’ attention (Smith, 1993), is one of the popular ways to improve their attention in writing. Related research has explored the effects of input enhancement on passive structures (Lee, 2007), subjunctive (Jabbarpoor and Tajeddin, 2013), relative structures (Ge, 2010), etc.

Strong interaction leads to strong alignment (Wang, 2018). Existing studies in the continuation writing task have explored the alignment effect of continuations through input enhancement. Xin (2017) carried out input enhancement in the form of task instruction and underlining of the subjunctive. It was found that such enhancement could significantly improve learners’ attention to the target forms, help them acquire the subjunctive, and promote learning. Wang (2019) highlighted the target structures through word-based print and found it can better align the continuations with the original text. Qiu and Wang (2022) processed the target structures of reading materials with boldface and underline, and found that input enhancement increased learners’ attention to language forms and enhanced the linguistic alignment effect. The above-mentioned studies mostly focus on the alignment effect of input enhancement on the continuation writing task. Only Qiu and Wang (2022) used the writing CAF to analyze the L2 learners’ performance in the continuation tasks.

Genre has always been an important factor in the research of L2 writing. Text genres differ from one another in terms of their communicative objectives, sociocultural practices, and roles (Yoon and Polio, 2017). Some linguistic features tend to vary depending on genre because of the unique ways that each genre develops ideas (Beers and Nagy, 2011). Research on reading and writing tasks shows that genre is one of the key factors affecting writing performance (Yu, 2009; Li, 2014). Regarding continuations in the continuation writing task, the shift in the genre of the reading materials impacts not only writing style but also the context and the language imitation model. As a result, genre is more likely to have an impact on the continuation tasks. Although a large number of existing studies have examined the effects of genre on writing (Yoon and Polio, 2017), there are only two studies relating to the effects of genre on the continuation writing task. Zhang and Zhang (2017) examined how the argumentation continuations and narration continuations differed in terms of alignment and linguistic error frequency. Zhang et al. (2023) explored the alignment effect and writing quality in the continuation tasks of narration, exposition, and argumentation. It can be seen that no study has explored the effects of genre on the writing CAF in the continuation tasks.

As mentioned above, extant studies have examined the impact of input enhancement and genre on language production, respectively. However, few studies have looked into how input enhancement and genre interact. As there are many variables influencing the learning promotion effect of the continuation writing task, it is crucial to investigate how these variables interact in order to make recommendations for creating continuation writing tasks. Although it has been demonstrated that input enhancement can enhance the learning promotion effect of continuation writing tasks in studies (Xin, 2017; Wang, 2019; Qiu and Wang, 2022) that used story continuation tasks, it is not clear whether different genres have an effect on how input enhancement promotes learning.

To sum up, few studies have investigated how input enhancement and genre affect L2 writing CAF in the continuations, and little is known about how different variables interact to affect the learning promotion effect of the continuation writing task. Therefore, taking the writing CAF as dependent variables, input enhancement as a between-subject variable, and genre as a within-subject variable, the present study adopted a two-factor mixed experimental design and attempted to explore the writing CAF produced in the continuation writing task and the interaction between input enhancement and genre. The following research questions were explored:

(1) What are the effects of input enhancement and genre on L2 learners’ performance of continuations in English continuation writing tasks?

(2) What is the effect of the interaction between input enhancement and genre on L2 learners’ performance of continuations in English continuation writing tasks?

Method

Participants

The participants in this study were 80 students from a senior high school in China. They were from two parallel classes taught by the same instructor. The participants were randomly divided into the enhanced group (30 males and 10 females) and the non-enhanced group (25 males and 15 females), and their ages ranged from 16 to 18. By the time of the study, all participants had been studying English for about 8 years and had never studied or lived in an English-speaking country. One week before the experiment, the participants took an English proficiency test. The test results showed that there was no significant difference between the two groups in terms of English proficiency (t = −0.058, df = 63.18, p = 0.954 > 0.05).

Materials and procedures

The research materials included one incomplete story and one incomplete argumentation. Considering the participants’ English proficiency, the length of each essay was controlled between 350–400 words, with 387 words for the story and 367 words for the argumentation. As learners’ familiarity with and interest in the writing subject greatly affects the quality of writing (Gu et al., 2022; Freedman, 1983), the material of the story continuation task was related to family life, and the material of the argumentative continuation task was tied to social media, both of which were all relevant to the daily life of the participants. The story tells about a mother who pawns her ring in order to buy pants for her child’s graduation ceremony, and the given first sentences provide important clues to the subsequent development. The argumentation continues by discussing the advantages and disadvantages of social media from three aspects, and the given first sentences lead the writer to analyze the advantages and disadvantages of social media from another perspective and draw a conclusion. In the continuation materials of the enhanced group, some language forms were marked with underlines, including vocabulary, lexical blocks, syntactic structures, etc. The continuation materials of story and argumentation in enhanced group both included 24 enhancements, while the continuation materials read by the non-enhanced group did not contain any form of marking. The participants were required to complete the writings in about 150 words.

Both groups first completed the story continuation task and then the argumentative continuation task a week later. The duration of each task was 50 min for both groups. The experiment required the participants to read the previous text carefully and fully understand it before writing, and the content of the writing must be consistent with the previous text. The use of dictionaries, computers, or any other assisting devices was not allowed during the writing process.

Data collection

After the continuation tasks were finished, the collected compositions were sorted out, excluding the data of the participants who participated in the continuation tasks only once. Finally, the data of 40 participants in the enhanced group and 47 participants in the non-enhanced group were available for analysis. Seven items of data from the non-enhanced group were arbitrarily eliminated in order to remove the impact of the difference in the number of participants on the experimental results. Finally, the data of 40 participants in each group were used for analysis. The researchers recorded the compositions of the two groups of participants into the computer, and built four small corpora, namely, enhanced story, enhanced argumentation, non-enhanced story, and non-enhanced argumentation. Each corpus contained 40 compositions.

Data analysis

Type Token Ratio (TTR), a calculation method of lexical diversity, is easily affected by text length, so it is not applicable to this study. In order to better measure lexical diversity, this study adopted the Uber index, which had been applied in many studies (Zheng, 2015; Zheng and Feng, 2017), to measure lexical diversity. The data was obtained by using Vocab Profile, an online analysis software, to obtain the number of types and tokens, and then substituted into the following formula for calculation:

$$U = \frac{{\left( {{{{\mathrm{logTokens}}}}} \right)^2}}{{{{{\mathrm{logTokens}}}} - {{{\mathrm{logTypes}}}}}}$$

As for lexical sophistication, online analysis software P_Lex was used to obtain the Lambda value, which measures the distribution of low-frequency words in the discourse. The Lambda value ranges from 0 to 4.5, and the larger the Lambda value is, the more infrequent words are used, and the higher the lexical sophistication is. Text size has little influence on Lambda value, so this study used Lambda value to measure lexical sophistication.

In addition, this study adopted C/S (clauses/sentences) and CN/T (complex nominals/T-units) in L2SCA developed by Lu (2010) to measure syntactic complexity. C/S was used to measure the general syntactic complexity, and CN/T was used to measure the phrasal complexity. TAASSC (Kyle, 2016), a syntactic complexity analysis tool, was used to obtain data of C/S and CN/T.

Lexical accuracy is the correct and appropriate use of words, usually measured by the occurrence of lexical errors in language production. The use of vocabulary relates to semantics, morphology, syntax, and pragmatics (Li, 2019). Therefore, the correct use of vocabulary does not mean the proper use. Factors such as collocation and context should also be considered. According to Engber (1995), lexical errors marked included errors relating to lexical choice and lexical form. Morphological verb errors involving past tense and third person singular were marked as errors of syntactic accuracy rather than errors of lexical accuracy. Lexical accuracy reflects the ratio of general lexical error per content word (LE/CW) (Li, 2019), which can be represented in the formula: 1-LE/CW (excluding names and numbers). Lexical errors were highlighted in Word 2016 by the researchers and double-checked through www.pigai.org, a writing assessment website. Content words were counted with the help of AntConc3.5.0, a corpus analysis tool, to retrieve the text tagged by TreeTagger3, an automatic part-of-speech tagging tool.

EFT/T (Error-free T-units) have been found to be a reliable measure of linguistic accuracy in L2 writing development (Li, 2019), so this study adopted it for measuring syntactic accuracy. Errors in syntactic accuracy should be distinguished from errors in lexical accuracy. The syntactic errors marked in this study mainly included word order, the absence and redundancy of major and minor constituents, morphological verb errors, and errors in combining clauses. Syntactic errors were highlighted in Word 2016 by the researchers and double-checked through www.pigai.org. T-units were labeled by the researchers with disagreements negotiated.

Writing fluency is divided into product-based writing fluency and process-based writing fluency (Latif, 2013). The present study focused on product-based fluency concentrating on written texts. The measure used in the study was the words per T-unit (W/T) (Larsen-Freeman, 2006). The data of words and T-units were respectively counted by Vocab Profile and manual marking.

This study adopted the two-factor mixed experimental design, which can be used to study the influence of two independent variables, one of which is the within-subject variable and the other is the between-subject variable, on the dependent variable at the same time. In this study, writing CAF were the dependent variables; input enhancement was treated as a between-subject variable with two levels: enhancement and non-enhancement; genre was treated as a within-subject variable also with two levels: story and argumentation. The participants were divided into an enhanced group and a non-enhanced group in accordance with the between-subject variable. The enhanced group wrote enhanced story continuations (SC) and argumentation continuations (AC), and the other wrote non-enhanced SC and AC. Extracted from the continuations, the data of each index of CAF were imported into SPSS 24.0. Repeated measures in the general linear model were used to test the between-subjects effects and within-subjects effects, and simple effect tests were conducted when there was a significant interaction between input enhancement and genre.

Results

Effects of input enhancement and genre on CAF

In order to have an overall view of language features, seven indexes were used to measure CAF in this study. The collected data was analyzed by SPSS 24.0. Table 1 presents the descriptive statistics of CAF of continuations with different input enhancement and genres. Table 2 reports the results of within-subjects effects and between-subjects effects.

Table 1 Descriptive statistics of CAF.
Full size table
Table 2 Tests of between-subjects effects and within-subjects effects.
Full size table

Lexical complexity was measured by Uber index and Lambda value, respectively for lexical diversity and lexical sophistication. As shown in Tables 1 and 2, the means of Uber index of enhanced SC and AC were higher than those of non-enhanced continuations, while the means of Lambda value of non-enhanced SC and AC were higher than those of enhanced continuations, but there were no significant differences in input enhancement (F = 0.407, p = 0.525 > 0.05, η²p = 0.005 < 0.01; F = 1.738, p = 0.191 > 0.05, η²p = 0.022 < 0.06). In addition, the means of Uber index of enhanced and non-enhanced SC were higher than those of AC, while the means of Lambda value of enhanced and non-enhanced AC were higher than those of SC, and the differences reached significant level (F = 9.664, p = 0.003 < 0.01, η²p = 0.11 > 0.06; F = 92.268, p = 0.000 < 0.01,η²p = 0.542 > 0.14), suggesting that the vocabulary used in SC was more diversified, while the vocabulary used in AC was more sophisticated. The results of lexical complexity in continuations with different input enhancement and genres can also be seen in Fig. 1.

Fig. 1: Means Plots of Uber Index and Lambda Value.
figure 1

Figure 1 takes input enhancement as the horizontal axis, Uber index and Lambda value as the vertical axis, drawing means plots, from which the lexical complexity of continuations with different input enhancement and genres can be observed more directly.

Full size image

Syntactic complexity was measured by C/S and CN/T, respectively for general syntactic complexity and phrasal complexity. As shown in Tables 1 and 2, the means of C/S and CN/T of enhanced writings of story and argumentation were higher than those of non-enhanced writings, but there were no significant differences in input enhancement (F = 3.254, p = 0.075 > 0.05, η²p = 0.04 < 0.06; F = 0.655, p = 0.421 > 0.05, η²p = 0.008 < 0.01). In addition, the means of C/S of enhanced and non-enhanced SC were higher than those of AC, while the means of CN/T of enhanced and non-enhanced AC were higher than those of SC, and the differences reached significant level (F = 24.013, p = 0.000 < 0.01, η²p = 0.235 > 0.14; F = 406.788, p = 0.000 < 0.01, η²p = 0.839 > 0.14), indicating that the syntactic structure in SC was more complex than that in AC, while AC used more complex nominals. The results of syntactic complexity in continuations with different input enhancement and genres can also be seen in Fig. 2.

Fig. 2: Means plots of C/S and CN/T.
figure 2

Figure 2 takes input enhancement as the horizontal axis, C/S and CN/T as the vertical axis, drawing means plots, from which the syntactic complexity of continuations with different input enhancement and genres can be observed more directly.

Full size image

Accuracy was measured by 1-LE/CW and EFT/T, respectively for lexical accuracy and syntactic accuracy. As shown in Tables 1 and 2, the means of lexical accuracy of non-enhanced SC and AC were higher than those of enhanced continuations, and the differences reached significant level (F = 13.903, p = 0.000 < 0.01, η²p = 0.151 > 0.14). In addition, the means of lexical accuracy of enhanced and non-enhanced SC were both higher than those of AC, and the differences reached significant level (F = 42.392, p = 0.000 < 0.01, η²p = 0.352 > 0.14). In respect of syntactic accuracy, the main effect of input enhancement and genre were both significant (F = 6.068, p = 0.016 < 0.05, η²p = 0.072 > 0.06; F = 4.697, p = 0.033 < 0.05, η²p = 0.057 < 0.06), indicating that input enhancement and genre had a significant effect on syntactic accuracy of continuations. The results of accuracy in continuations with different input enhancement and genres can also be seen in Fig. 3.

Fig. 3: Means Plots of 1-LE/CW and EFT/T.
figure 3

Figure 3 takes input enhancement as the horizontal axis, 1-LE/CW and EFT/T as the vertical axis, drawing means plots, from which the syntactic accuracy of continuations with different input enhancement and genres can be observed more directly.

Full size image

Fluency was measured by W/T. As shown in Tables 1 and 2, the means of W/T of enhanced SC and AC were higher than those of non-enhanced continuations, but there were no significant differences in input enhancement (F = 1.227, p = 0.271 > 0.05, η²p = 0.015 < 0.06). In addition, the means of W/T of enhanced and non-enhanced AC were both higher than those of SC, and the differences reached a significant level (F = 23.534, p = 0.000 < 0.01, η²p = 0.232 > 0.14), indicating that AC were more fluent than SC. The results of fluency in continuations with different input enhancement and genres can also be seen in Fig. 4.

Fig. 4: Means Plot of W/T.
figure 4

Figure 4 takes input enhancement as the horizontal axis and W/T as the vertical axis, drawing a means plot, from which the fluency of continuations with different input enhancement and genres can be observed more directly.

Full size image

Interaction Between Input Enhancement and Genre

As shown in Table 2, the effect of the interaction between input enhancement and genre on lexical complexity was not significant (F = 0.194, p = 0.66 > 0.05, η²p = 0.002 < .01; F = 0.386, p = 0.536 > 0.05, η²p = 0.005 < 0.01). With respect to syntactic complexity, the effect of the interaction between input enhancement and genre was not significant (F = 1.308, p = 0.256 > 0.05,η²p = 0.016 < 0.06; F = 0.483, p = 0.489 > 0.05, η²p = 0.006 < 0.01). As for lexical accuracy, the effect of the interaction between input enhancement and genre was significant (F = 6.275, p = 0.014 < 0.05, η²p = 0.074 > 0.06). In respect of syntactic accuracy, the effect of the interaction between input enhancement and genre was extremely significant (F = 10.804, p = 0.002 < 0.01, η²p = 0.122 > 0.06). In terms of fluency, the effect of the interaction between input enhancement and genre was not significant (F = 0.025, p = 0.874 > 0.05, η²p = 0.000 < 0.01).

As the effects of the interaction between input enhancement and genre on lexical accuracy and syntactic accuracy were significant, simple effect tests were conducted through SPSS 24.0 and the results are reported in Tables 3 and 4, respectively.

Table 3 The interaction between input enhancement and genre in lexical accuracy.
Full size table
Table 4 The interaction between input enhancement and genre in syntactic accuracy.
Full size table

As shown in Table 3, the vocabulary used in SC was significantly more accurate than in AC of both enhanced and non-enhanced continuation tasks (p = 0.000 < 0.01; p = 0.006 < 0.01). As for story, there was no significant difference (p = 0.170 > 0.05) in lexical accuracy whether the reading materials enhanced or not. While the vocabulary used in non-enhanced AC were significantly more accurate than in enhanced AC (p = 0.000 < 0.01).

As shown in Table 4, the syntactic structures used in enhanced AC was more accurate than in enhanced SC (p = 0.000 < 0.01), while non-enhanced continuations had no significant difference in syntactic accuracy (p = 0.431 > 0.05). As for story, there was no significant difference in syntactic accuracy (p = 0.734 > 0.05), while the syntactic structures used in enhanced AC were more accurate than in non-enhanced AC (p = 0.000 < 0.01).

Discussion

The present study examined the effects of input enhancement and genre on the writing CAF of continuations and the interaction between input enhancement and genre. We addressed two research questions and the findings concerning each question will be discussed in turn.

Effects of input enhancement and genre on CAF

With respect to complexity, there was no significant difference in lexical complexity and syntactic complexity between enhanced continuations and non-enhanced continuations, which is consistent with the findings of Qiu and Wang (2022). The enhanced group did not show a significant advantage in lexical complexity, which may be attributed to three factors. The first one is the restriction of the scope of language use in the reading materials. The material of story continuation task used in this study restricted the use of words to family life and the material of argumentative continuation task to social media. Secondly, the number of enhanced words in the reading material was limited. The two genres only had 24 enhancements of words, lexical chunks, and syntactic structures. Therefore, the participants had to rely more on their accumulated linguistic expressions. In addition, learners might subconsciously avoid using words or structures that they have not mastered. In L2 English writing, learners tend to avoid using complex words that they can understand when reading but cannot express their thoughts when writing (Xiang, 2012). Therefore, although participants pay attention to the enforced structures when writing, they are more likely to choose simpler words and sentences. In order to better play the learning promotion effect of the continuation writing task, teachers should control the difficulty level of the reading materials in line with students’ language ability. As for syntactic complexity, although the means of C/S and CN/T of enhanced continuations were higher than those of non-enhanced continuations, the difference in syntactic complexity between enhanced and non-enhanced writings did not reach a significant level. The syntactic complexity of written output is affected by task complexity (Zhang and Jiang, 2020), task instruction and working memory (Yi and Ni, 2015) and other factors. In this study, the two groups were given the same reading materials and task instructions with no requirements on the accuracy or complexity of the writing, so there was no difference in task difficulty. In addition, both groups of participants could read back the previous text in the process of writing, so they were not restricted by working memory. Therefore, there was no significant difference in syntactic complexity between enhanced and non-enhanced continuations in this study.

Although there were no significant differences in complexity on input enhancement, significant differences were found in complexity between SC and AC. The words used in SC were significantly more diverse than those in AC, while the words used in AC were significantly more sophisticated than those in SC, which is consistent with the findings of Yoon and Polio (2017). Zhang and Jiang (2020) found that complexity is easily affected by genre, and different genres have a certain impact on lexicon and syntax. In terms of lexical diversity, story writing stresses the vividness and coherence of the plot and concentrates on the description of the event progress, which is conducive to stimulating the use of verbs and adverbs, so story continuation task had a significant advantage in lexical diversity. The low variety in AC may be attributed to participants’ use of formulaic phrases regular in argumentation writing, such as “in my opinion”, “for example”, “in conclusion”, etc. The significant differences in lexical sophistication can be explained by the alignment effect of the continuation writing task. Wang and Wang (2015) found that learners’ continuations were aligned with the previous reading materials, and many words in the original were used in continuations. The reading material in the story continuation task had quite a number of daily dialogues in relatively simple words, while the reading material in AC involved logical reasoning and relatively complex vocabulary use. The words used in AC were more sophisticated than in SC, in line with the finding by Yoon and Polio (2017) that L2 learners tend to use more sophisticated language to complete a task demanding causal inference than a task requiring simple fact transmission. As for syntactic complexity, C/S and CN/T data of SC and AC presented opposite results. The general syntactic complexity of SC was significantly higher than that of AC, while the phrases used in AC were significantly more complex than those in SC. The main reason is that the story focuses on describing the process of events, and the sentences are too scattered, which leads to the increase of clauses in the sentences, resulting in high C/S. The language in story was relatively simple, using less complex nominals, resulting in a low CN/T. In AC, students’ general syntactic complexity (C/S) was low, which is because argumentation has fewer run-on sentences than in story. The AC used more complex nominals (CN/T), which can be explained by the finding in Zhang and Zhang (2017) that the phrasal alignment in AC is stronger than in SC. The genre differences determine that more phrases can be imitated in argumentation than in story.

As for accuracy, the present study adds to the existing studies. As there was interaction effect in accuracy between input enhancement and genre, the effects of input enhancement of different genres were different. As for story, there was no significant difference in accuracy whether the reading materials were enhanced or not, which is inconsistent with the finding of Qiu and Wang (2022). The reason may be that the indexes used are different in the two studies. Qiu and Wang (2022) found that the alignment effect can significantly affect the accuracy of language form rather than the language meaning, while the index of lexical accuracy in the present study not only focused on the correct use of the lexicon but also the proper use relating to meaning. As for syntactic accuracy, the continuations in our study were short with 150 words on average, and the use of enhanced structures was limited, so the differences were not significant. More importantly, the language in story is generally easier than that in argumentation. The participants can produce the correct forms depending on their accumulation of linguistic expressions even with no attention drawn to enhanced structures. But the language of argumentation is relatively difficult with reasoning examples. Therefore, AC had significant differences in lexical accuracy and syntactic accuracy. But the two measures behaved in opposite ways. The words used in non-enhanced continuations were more accurate than in enhanced continuations, while the sentences used in enhanced continuations were more accurate than in non-enhanced continuations. The latter can be explained by the finding in Qiu and Wang (2022) that input enhancement can improve the accuracy of language form. The finding of lexical accuracy in our study adds that even though learners tend to imitate the use of the enhanced structures, they may not use them correctly without systematic learning. For example, the enhanced word “contact” was used 8 times in enhanced AC, twice as many as in non-enhanced AC, but used correctly once only, while the rest were mismatched with “with somebody”. Therefore, if students do not have a basic understanding of the enhanced structures, the learning promotion effect of the continuation writing task may not be achieved.

Under the condition of input enhancement, the words used in SC were more accurate than in AC whether enhanced or not. Skehan and Foster (2001) found that higher accuracy is associated with lower complexity. As the lexical sophistication of AC was significantly higher than that of SC in this study, and the increase in complexity may result in new errors, the words used in AC were less accurate than those in SC. As for syntactic accuracy, the syntactic structures in AC were more accurate than those in SC in enhanced continuation tasks, while there was no significant difference in syntactic accuracy in non-enhanced continuation tasks. As there is competition in attentional resources, more attention to content will result in less attention devoted to form (Skehan and Foster, 2001). Thus the focus on linguistic form will be reduced since learners tend to pay more attention to content rather than form in SC. As the first paragraph given in AC provided a definite statement, students have to read back to the previous text and use the enhanced structures. Therefore, the syntactic structures in enhanced continuation tasks were more accurate in AC than those in SC.

As for fluency, there was no significant difference between enhanced continuations and non-enhanced continuations, which is consistent with the findings of Qiu and Wang (2022). Skehan and Foster (2001) believed that due to the limited attentional capacity, there would be competition among the fluency, accuracy, and complexity of language output. As students tend to pay more attention to accuracy and complexity when writing, the continuations have no significant advantage in fluency. In addition, fluency is correlated with the automatization in language use (Skehan, 1998), which is the result of a lot of practice over a long period of time and can only be changed after repeated training. Therefore, the enforced structures can only improve the attention to language forms, but can hardly make a significant change in fluency. The present study also found that AC were significantly more fluent than SC, which can be explained by their genre differences. SC enable learners to describe situations and express feelings more freely with simple and short sentences. While argumentation involves reasoning examples, the structures used in argumentation are usually more complicated than those in story. For example, a sentence in AC written by a participant has almost 30 words: “Most people trust that, social media will make the relationship of people be better and we will be able to share the things we like or want wherever we stay”. Such long sentences result in a high W/T.

Interaction between input enhancement and genre

In this study, there was a significant effect of the interaction between input enhancement and genre in accuracy. As discussed above, the effects on CAF were different in different input enhancement and genres. SC had no significant difference in lexical accuracy and syntactic accuracy whether the reading materials enhanced or not. In AC, the words used in non-enhanced continuations were more accurate than in enhanced continuations, while the sentences used in enhanced continuations were more accurate than in non-enhanced continuations. As input enhancement can improve the accuracy of language form, it is recommended to enhance the target structures using underlines, boldfaces, and other ways when designing the continuation writing task. But the target structures should be carefully chosen. Xin and Li (2020) stated that the best learning efficiency can be achieved when the active alignment occurs in the zone of proximal development combining input and output with the help of context. Therefore, in order to achieve the best learning promotion effect, teachers should design the continuation writing task in accordance with students’ current level of acquisition, choosing those recently learned structures that have not been totally mastered as enhanced input. Under the condition of input enhancement, the words used in SC were more accurate than in AC whether enhanced or not. As for syntactic accuracy, the syntactic structures in AC were more accurate than those in SC in enhanced continuation tasks, while there was no significant difference between SC and AC in non-enhanced continuation tasks. Therefore, different genres of the continuation writing task can be adopted for different purposes. Story continuation tasks can be used to train students’ ability to express their ideas, while argumentative continuation tasks can be used to help master complex phrases and syntactic structures.

However, the effects of the interaction between input enhancement and genre in complexity and fluency were not significant. There may be two reasons: first, between-subject variable input enhancement had no significant effect on complexity and fluency of continuations, so SC and AC were almost not affected at the two levels of between-subject variable. Second, there were only two levels of within-subject variable, story and argumentation. If other genres such as practical writing and exposition are considered, the within-subject variable is increased to three or four levels. Then under different between-subject variable levels, the results of different levels of within-subject variable may be different.

Although input enhancement and genre had no interaction effect on language complexity and fluency, they can be combined in continuation writing tasks to improve learning efficiency. It was found that input enhancement can improve attention to language forms (Xin, 2017; Qiu and Wang, 2022). The descriptive statistics of complexity and fluency in our study also found that, except for Lambda value, the average of Uber index, C/S, CN/T and W/T for enhanced SC and AC were all higher than those for non-enhanced continuations, indicating that input enhancement still has a certain effect on the complexity and fluency of written language. In terms of genre, SC and AC have their own advantages. Statistics show that the average length of 80 SC was 172.25 words (SD = 40.41), and the average length of 80 AC was 148.70 words (SD = 25.44). The result of paired samples t-test shows that there was a significant difference in the number of words between SC and AC (p = 0.00 < 0.01), indicating the length of continuations is affected by genre with SC producing longer texts. Therefore, coupling with the findings in this study, story continuation tasks can stimulate students’ imagination, and produce diverse content and complex syntactic structures. Argumentative continuation tasks can provide complex language input, and students can produce complex words and phrases through the alignment effect with the continuation materials. Therefore, input enhancement can give full play to the learning promotion advantages of the two genres by increasing the attention to the target forms and enhancing the language alignment.

Conclusion

This study explored the effects of input enhancement and different genres on the writing CAF in continuation writing tasks. It was found that genre significantly impacted complexity, accuracy, and fluency, while input enhancement had a significant effect on accuracy, but not on complexity and fluency. In addition, the results showed that there was a significant effect of the interaction between input enhancement and genre in accuracy, but not in complexity and fluency.

In order to improve the learning promotion effect of the continuation writing task, three suggestions are put forward according to the above research findings: First of all, teachers should select interesting continuation materials with topics closely related to students’ life and learning. Target structures should be in line with the student’s zone of proximal development as enhanced structures, which can help them gradually master relevant expressions through creative imitation. Secondly, students should be given different genres to continue, and not restricted to stories, as this study showed that story continuation tasks and argumentative continuation tasks have their own advantages respectively, which can be adopted alternately to help students express themselves freely and master complex structures in English. Finally, students should pay attention to the accumulation of vocabulary, phrases, collocations, etc. A variety of language expressions and a large vocabulary are the basis for a good composition, while continuation materials cannot immediately improve the complexity and fluency of writing output. Thus the most important is to rely on students’ accumulation of linguistic expressions.

There are several shortcomings in this study: First, this study only investigated the effects of input enhancement and genres on the writing CAF in the continuation writing task. Future studies can use the two-factor mixed experimental design to investigate the effects of other factors on the writing CAF. Second, the within-subject variable genre only had two levels, story and argumentation. Future research can expand to include practical writing, exposition and other genres. Third, this study only examined short-term effects, and future longitudinal studies can be conducted to explore the development features of different genres or enhanced continuations as well as the dynamic interactions among complexity, accuracy, and fluency.