The first research question dealt with the effects of task complexity on the L2 written performance in the reading-writing continuation task, in specific, whether providing content support in the form of pictures would impact syntactic complexity,lexical complexity,accuracy and fluency of the written output. Results showed that the complex task group(-content support) outperformed the simple task group (+content support) in syntactic complexity. And there were no significant differences between the two groups in terms of lexical complexity,accuracy and fluency.

The findings for syntactic complexity and lexical complexity ran counter to the findings of previous studies(Ong&Zhang,2010; Kormos,2011;Revesz et al.,2016 that manipulated task complexity through +/-content support, which all found an enhancement in lexical complexity or syntactic complexity when content support provided. One possible reason may be the differences of writing tasks adopted in this study and task complexity manipulations in these studies. In specific, Ong & Zhang (2010) and Revesz et al. (2016)investigated the argumentative writing and altered the task complexity by providing writing topics and structure. Kormos(2011) adopted picture narratives in two groups, and the content support in his definition was whether there was clear logical relation among the pictures. Turning to the present study,the task was an under-researched kind of task that combined reading and writing, and the complexity was manipulated by whether providing pictures or not.

The findings did not lend support to Robinson's Cognition Hypothesis (2001a, 2001b,2003,2005a,2007a,2011). According to Robinson's claims, increasing task complexity along resource-dispersing dimension would lead to a decrease of linguistic complexity. The reverse effect on the syntactic complexity in this study may be due to,

as mentioned above,the characteristics of the continuation task and the form of content support. The reading-writing continuation task is a task in which the learners need to read a reading text before they write. Their writing should be based on their understanding of the reading text. That is, the learners have to refer to the reading text in order to finish their writing. In the complex task where no pictures were provided, the learners had to make up the subsequent plots of the story by themselves, which increased the load of their cognition. As the questionnaire showed, the mean ratings for the amount of time spent on planning the content were3.37(SD=1.115)  in the complex task and 2.70(SD=1.137)in the simple task, reaching a significant difference (p=0.035<0.05). According to Robinson's ((2001a,2001b,2003,2005a,2007a,2011) hypothesis, this high cognitive demand the task makes on the learner will deplete their attentional resources in language forming, thus leading to a decrease in linguistic complexity.However,this may not be the case in the present study.In this task,without the content support, the learners would to a greater extent resort to the preceding text for inspirations of story continuation. Repeated referring to the preceding text led to a closer interaction between the learners and the text. The close interaction with the preceding text enhanced their noticing of the language forms in the text, which then helped them construct their own language in more complex forms. Therefore,there is a great likelihood that the aid of the preceding text covered the impediment of the high cognitive load of the task. With sufficient linguistic resources available due to the presence of the reading text, it is not surprising that the learners could have a better performance in writing.What is more,the manipulation ofcontent support by providing pictures could also be a possible cause for the worse performance in the simple task.In the simple task, the content support was provided in the form of 20 sequential pictures. 

Although the pictures were proved quite easy to understand (The mean rating score of picture clarity was 4.63(SD=.629), which was close to the full score),it was also possible that some learners spent relatively more time on interpreting the pictures. The learners' attentional resources on language form might be dispersed by the need to interpret the pictures, hence resulting in a lower complex performance.

In terms of the lexical complexity, the reason why there was no difference may lie in the content of the continuation task. Both groups were asked to continue a story which was based on the same preceding text. The content of the story may have determined a rough range of lexicons. Without a large room for content creating, it was difficult for them to produce largely different range of words.

The results for accuracy were aligned with the findings ofKormos (2011), whose study was the only one that measured accuracy. However, it did not coincide with Robinson's (2001a, 2001b,2003, 2005a, 2007a,2011) convictions that accuracy will decrease when content support was not provided. A tentative explanation for the non-effect on accuracy could be that the freedom to write the content and the help of the preceding text mediated the cognitive complexity of having to create the content. In spite of the fact that learners in the complex task group have to create the content, they can avoid using the language that they did not master.Besides,the preceding text could serve as a guide and reviewer for learners to correct their language (Wang,2012).In this way, they could avoid a decrease in accuracy. It might also be due to the allocation of attentional resources. Writers might pay most of their attention to the meaning so they neglected the forms. Vanpatten (1990)once pointed out that when restricted by cognitive resources, learners will pay prior attention to meaning.

The results for fluency were in accordance with the findings of Ong & Zhang (2010) and Revesz et al.(2016) that it appeared unaffected by task complexity. One possible reason might lie in the restriction of the writing time. Though the simple task group were provided with 20 pictures that could inspire a very detailed and long narration, the learners would adjust the concreteness of their description in order to make sure that theycould finish it in the limited time. While for the complex task group, though they were encouraged to write as long as they could, the extension of their imagination was also restricted by the limited time.

Taken together, we could conclude that the support of content in the form of pictures did not seem to result in substantial changes in the linguistic performance, except for a decrease of one index of syntactic complexity. These findings offered us insights into the manipulation of task complexity. It seems to suggest that task complexity operationalized with respect to +/-content support in the form of providing or not providing pictures could not lead to so much impacts on linguistic performance as other variables, such as +/-reasoning demands,+/-few elements,+/-here-and-now, etc., could(e.g,Kuiken et al,2005;Kuiken& Vedder,2006,2007,2008;Ishikawa, 2007).A tentative speculation might be that it was not difficult forlearners to invent an

ending for the story,and the content of the pictures was easy to imagine.Therefore, the relatively small distinction between providing pictures and not providing pictures was less likely to pose much impacts on learners'linguistic performance. Moreover, the manipulation of +/-content support could also be a potential reason.On the one hand, the time given for writing the continuation texts may have allowed the learners to capitalize their attentional resources efficiently no matterwhether they had to create the

content of the story or not.On the other hand, the demand of having tocreate thecontent and the demand of encoding given content with the available linguistic resources may have eliminated each other's effect.

It may also be related to the individual differences. Robinson (2001a,2001b,2003, 2005a,2007a,2011) mentioned the role of learner factors in the task performance,such as motivation, learning style, working memory capacity,learning aptitude, etc.In the reading-writing continuation task,in addition to the writing motivation, there is another learner factor that is crucial to the writing—imagination,since it is a story continuation task which needs learners to create a story with their imagination. Though these participants had similar language proficiency and writing expertise, there might be differences in their imagination. Those with good imagination could perform better in the-content support task, because they can conjure up a good ending to the story without any difficulty.Giving pictures may hit their imagination and affect their interest in writing, leading to a worse production than not giving pictures. In contrast, those with poor imagination would perform better in the +content support task.

These speculations could be testified by the questionnaires. Some participants in the complex task group said that they did not even know how to start writing without any hint and it cost them much time to conjure up a good story. They stated that they might have written longer if they knew how the story developed. A few participants in the simple task, however, mentioned that they would rather there was not any content support since they preferred to continue the story as they like. They said that their thoughts were stuck when they thought of a good story but found that they had to write as the pictures showed. It showed that the imagination might be a major reason that the effects of the two tasks on the linguistic performance were not that obvious.

3. Discussion for research question 2

The second research question asked whether task complexity manipulated through +/-content support affected alignment in the reading-writing continuation task。as measured by lexical, phrasal and clausal alignment. The results revealed that either at lexical, phrasal and clausal level, those in the -content support task group tended to align more with the original text than those in the +content support task group. It can be argued that increasing task complexity with respect to -content support leads to a stronger alignment effect.

In the complex task (-content support) where learners had to create the content by themselves, their attentional and memory resources were dispersed. The demand that they had to distribute their attentional resources effectively to both form and content of the output was relatively high. Since the continuation of the story was based on the preceding text, learners needed to understand the text fully before they continued. Under the condition where no content support was provided, the preceding text was the only support that they could resort to. Therefore, they had to resort to the preceding text so as to lower their cognitive load. They would read the text repeatedly in order to get more inspirations for their continuation, which resulted in a more intensive interaction between the learners and the preceding text. And the repeated reading and understanding of the text might enhance the learners'noticing of the language forms in the text. So they would adopt some expressions from the text in their own writings. 

On the contrary, when learners were provided with pictures that contained the content of the story in the simple task, there seemed to be less need for them to resor to the preceding text. They would, after understanding the preceding text, pay the most of their attentional resources to the description of the pictures. The rich information contained in the pictures provided learners inspirations and enabled them to write a suitable continuation. As can be seen from the questionnaire,learners stated that the pictures to a large extent assisted their writing, with a mean rating of4.07 (SD=.917). They did not need to create the content by themselves,so they did not need to refer much to the preceding text. The weaker interaction between the learners and the preceding text resulted in weaker alignment effect. In contrast, they would align with the pictures since they interact more with each other.It can be seen from the words and phrases that they used: police,dogs, have a picnic,on the rock, food, hug, etc.,which were what the pictures contained. These findings were suggestive that the content provision in the reading-writing continuation task, though hindered the alignment with the preceding text, could induce strong alignment with the given content in that it elicited learners to produce linguistic expressions highly in line with the content of the pictures.

All things considered, these findings provided support for Wang's (2016) claims in the Extension Hypothesis that the alignment strength was closely related to the interaction intensity between the learner and the reading text. Under the condition where no content support was provided, the learners interacted more with the preceding text, thus increasing the strength of alignment.

4.  Discussion for research question 3

The third question intends to reveal whether there are any links between linguistic alignment and linguistic performance. The results showed that there were significant positive correlations between alignment and syntactic and lexical complexity when regarding the participants as one group. When comparisons were made between the two groups, results revealed significant positive correlations between alignment and complexity,accuracy,and fluency in the complex task group. There were no significant correlations between alignment and any of the measures of linguistic performance in the simple task group.

The findings about positive correlations between alignment and accuracy and fluency in the complex task group were in line with the findings of prior studies (Xue, 2013;Xiao,2013;Yuan,2013; Peng,2015),which found positive correlations between

linguistic alignment and language accuracy or fluency. However, these significant correlations were only foundin the complex task group, but not in the simple task group or two groups together. These findings could provide supporting evidence for Wang's (2015)conviction that the alignment effect is the mechanism underlying the facilitating role ofthe reading-writing continuation task in L2 learning. Wang (2015)contends that as learners align with the text,the high-quality linguistic expressions from the given text could compensate their own linguistic competence, thus leading to a better linguistic performance,such as few errors,more sophisticated words and more complex sentences.However,the findings could only partially support the contention. It may be due to the patently low frequency ofalignedexpressions in the simple task group, which made it difficult to detect any correlations between the two aspects. While in the complex task group, the stronger alignment made the correlations between the two aspects more noticeable. It conveyed that stronger alignment could lead to better linguistic performance at all the three levels (complexity,accuracy and fluency).These results were conceivable.With stronger alignment, the learners incorporated the high-quality language from the preceding text into their own linguistic output, thus making their language more complex. And the correct language in the preceding text could serve as a monitor or reviewer for the learners, with which they could make few errors. The cognitive load lowered by the availability of language assistance enabled them to produce more ideas, thus writing longer texts.

However, things were different when all the participants were regarded as one group, where significant correlations were only found between alignment and syntactic complexity and lexical complexity. It seemed to indicate that the impact alignment could exert on linguistic performance was more outstanding in terms of linguistic complexity. To put it another way, the most noticeable effect of alignment on L2 linguistic performance was that stronger alignment could bring about more complex language. These results might be attributed to the characteristics of the linguistic elements they aligned with the preceding text. Taking a deeperlook at the participants' continuations, it could be found that some sentences were much more frequently aligned. The most magnificent features of the most frequently aligned sentences were that there were multiple clauses or verb phrases in one T-unit. Moreover, these sentences tended to entail variant and sophisticated words describing the pebble. Therefore, the alignment of these linguistic elements consequently facilitated the syntactic complexity and lexical complexity of participants'written performance.

5.Summary
 

For question one, the findings of effects of task complexity on L2 linguistic performance were not in line with Robinson'sclaims in the Cognition Hypothesis. It may be due to the different task types adopted and different manipulations of task complexity in the present study.Regarding question 2, the stronger alignment effects in the complex task could be explained by more intense interaction of the learners with the preceding text, which was caused by the high cognitive load posed on their attentional resources. Instead, in the simple task group the provision of writing support reduced the learners'necessity of interacting with the preceding text, thus leading to weaker alignment effects. With respect to question 3, significant positive correlations were detected between alignment and some indices of L2 linguistic performance, though not in the simple task group. The findings were in line with previous studies and partially corroborated Wang's(2015)conviction that it is the alignment effect in the reading-writing continuation task that facilitates L2 learning.