You are on page 1of 33

Task complexity and linguistic and discourse features of narrative writing performance

Judit Kormos

Introduction

Although oral narratives are frequently researched in the field of second language acquisition,
studies on written narratives produced by foreign language learners are scarce despite the fact that
this genre is an important one both in language teaching pedagogy and in the assessment of foreign
language (FL) competence. Narratives are among the frequently taught written text types in general
foreign language courses starting from the beginning level, and they are also included in the
curriculum of composition classes for higher level college learners. A large number of high-stakes
language proficiency tests also employ narrative tasks to assess FL learners’ writing competence
(such as most levels of Cambridge ESOL proficiency exams). Moreover, narrative accounts of
events are important in authentic written communication contexts, and narratives can be embedded
in other text types such as supporting evidence in argumentative writing.
Another relatively under-researched area in the field of FL writing is the role of task
complexity in writing performance. Whereas the effect of task-complexity on oral language
production has been in the forefront of investigations in the past twenty years (for a recent summary
see Samuda and Bygate, 2008), considerably less research has been conducted on how different task
types and the complexity of tasks influence written output produced by FL learners (see e.g. Ellis &
Yuan, 2004; Ishikawa, 2006; Kuiken & Vedder, 2008; Ong & Zhang, 2010). Gaining an insight into
how different task characteristics are associated with the linguistic and discourse feature of LFL
written texts could assist both language teachers and testers in selecting tasks that have the potential
to elicit the targeted features of writing competence.
The current study contributes to examining interfaces between second language acquisition
research and the study of L2 writing by exploring the issue of task complexity in the domain of FL
1

writing. The research presented in this paper aimed to investigate the linguistic and discourse
characteristics of FL narratives produced by upper-intermediate foreign language learners in a
bilingual secondary school. In our analyses we used a variety of linguistic and discourse variables
and a recently developed computer tool (Coh-Metrix 2.0: McNamara, Louwerse, & Graesser, 2002)
to describe the characteristics of narrative texts. Although our study is based on a relatively small
sample of student performances and on data collected at one specific level of proficiency, the range
of variables used for the analysis is considerably wider than in previous research investigating the
linguistic features of FL writing performance. Our analyses also exemplify how variables used to
assess writing quality and measures of performance applied in task-based research can be combined
to provide an insight into the nature of FL writing. This micro-analytic perspective, which is
demonstrated on a small data-base in this study, can provide an example on the basis of which
further research using a larger corpus and a wider variety of text types can be conducted. As the
computer analytic tools in this study are freely available and easy to use, they can be applied by
writing teachers to diagnose areas of weaknesses and strengths in students’ linguistic repertoire and
use of discourse markers, by assessors of second language writing to generate automated measures
of lexical and syntactic complexity and cohesion, and last but not least by student writers to assess
the quality of their own writing.
As a background for comparison, we also collected data from a small group of L1 writers in
order to gain an insight into features of written task performance that is not hindered by difficulties
in accessing linguistic knowledge required to execute a task.

The comparison between the

characteristics of L1 and FL narrative performance can yield information on the lexical and
syntactic performance as well as on the frequency of cohesive markers that a particular narrative
writing task can be expected to elicit and can inform writing pedagogy and assessment about how
task demands shape language use.
The study also intended to answer the question how different types of narrative tasks that
make different cognitive demands on writers affect the linguistic and discourse features of output.

2

In this research project, two types of task were administered to students: a cartoon description task
in which students had to narrate a given content, and a picture story task, in which students had to
invent the plot of their narrative. Both in the assessment of FL writing and in writing pedagogy, it is
an important question to what extent the fact that students are allowed to generate the content of
their text and potentially tailor it to match their available linguistic resources influences
performance, especially the accuracy of writing. The fact that students can avoid certain syntactic
and lexical structures when they generate their own content might hinder linguistic development
and might provide only a partial view of students’ linguistic competence in writing. Specifying the
content of the written task, however, might necessitate the use of lexical and syntactic structures
targeted by teachers and language testers, and might make avoidance difficult. From this
perspective, tasks with a predetermined content might be more reliable measures of linguistic
competence in writing, but we have to note that the authenticity and motivating power of these tasks
might be questionable.
In this paper, first the theoretical background of the study is reviewed, which is followed by
the description of the study and the methods of analysis. Next, the results are presented and
discussed. The paper is concluded with a detailed discussion of the implications of the findings and
an outline for future research directions.

Background

One of the important albeit neglected questions in second language writing pedagogy is how
different task characteristics influence the quality of students’ writing.

Most models of the

processes involved in writing focus on mechanisms of planning, the linguistic formulation of
planned content and revision (e.g. Bereiter & Scardamalia, 1987; Flower & Hayes, 1980; Kellogg,
1996) and have been developed for L1 and not second language (L2) and FL writers. None of these
models discuss how the particular features of different writing tasks influence these processes and

3

29). which influence the complexity of the task. The cognitive complexity of L2 writing tasks that originates from the characteristics of the task itself can be hypothesized to be inherent in the formulation stage and can be determined by the demands tasks make on the planning of the content of the written text and/or on the linguistic encoding of the content. writing consists of three important interactive. Formulation involves planning the content of the writing and translating ideas into words. “task complexity is the result of attentional. syntactic encoding of clauses and sentences and expressing cohesive relationships in the text. Finally. monitoring ensures that the created text adequately expresses the writer’s intention. Robinson lists a number of task characteristics such as the number of elements to be considered in a decision-making task. In Robinson’s (2001) definition. In Kellogg’s (1996) model.and recursive processes: formulation. In the execution stage. task characteristics are mostly discussed with reference to task complexity. execution and monitoring. During planning. where the role of task demands has been thoroughly researched. writers retrieve ideas from their long term memory or from the input provided in the task rubrics and organize them in a coherent order. it is important to consider the demands that these tasks make at different stages of the writing process. Task complexity inherent in the task itself is generally considered to derive 4 . memory. it is not a simple enterprise to establish a rank order of complexity for different types of pedagogic tasks. and other information processing demands imposed by the structure of the task on the language learner” (p. the text is revised. writers use motor movements to create a hand-written or typed text. In order to define the construct of task complexity in writing tasks.how writers divide their attention among the various composing processes as a result of the different task demands. Although these task characteristics can be manipulated in task-design. availability of planning time and prior knowledge. In second language speech production research. In translating ideas into linguistic form three processes can be distinguished: retrieval of the lexical items. and if mismatches are found.

in which students have to narrate their own story ─ we can see that different aspects of task complexity might have a differential effect on the writing process. therefore. Nevertheless. which is also relevant in studying second language writing. which results in increased conceptualization effort during planning and a potentially reduced load in the transcribing process.g. which increases the processing load in the linguistic encoding (transcribing) phase. and hence cognitively complex tasks are complex both in terms of planning and linguistic formulation (see e. students do not need to conceptualize the content of the story. however. This view is based on the assumption that complex concepts require the use of complex syntactic structures. but they can tailor it to match their linguistic resources. 2005) Cognition Hypothesis. In a story narration task. There are two influential models of task complexity in this field. they suggest that increasing task complexity reduces the 5 . 2003. A key factor in task-based language learning. These models make contrasting predictions as to the effect of increasing task complexity along various dimensions on L2 performance.from the conceptual demands a task makes in the planning stage. 1996). that is. If we consider two writing tasks such as the ones used in the current study ─ a cartoon description task. possible that tasks make separate complexity demands on the planning and formulation stages. students have to express the content prescribed by the task in whatever linguistic resources they have available in the L2. Robinson’s (2001. which have attempted to answer the question how attentional resources can be used. 2005) Cognition Hypothesis). Skehan and Foster’s (2001) Limited Attentional Capacity Model views attention and memory as limited in capacity. and a picture narration task. in selecting and ordering the relevant ideas (Kellogg. 2003. coordinated and directed to different aspects of language production during task completion: Skehan and Foster’s (2001) Limited Attentional Capacity Model and Robinson’s (2001. in which the story-line is given. is that students need to the allocation of their attention in order to successfully meet the linguistic demands of the task. however. In a cartoon description task. learners need to design their own story. It is. which eases the complexity demands in the phase of planning.

which is probably due to the fact that in some tasks. certain characteristics make resource-dispersing attentional demands on learners. The resource-directing dimensions of task performance call learners’ attention to the linguistic features which are needed to meet task demands (e. depletes learners’ attention without having the beneficial effect of directing it to any specific linguistic aspect of L2 production. two sets of dimensions of cognitive task complexity are distinguished: resource-directing and resourcedispersing dimensions. 2009). Skehan and Foster also claim that cognitively more demanding tasks draw learners’ attention away from linguistic forms so that enough attention can be paid to the content of the message (for a recent account of the model see Skehan.g. reducing the pre-task planning time). to events taking place then and there).. reference to events happening here and now vs.. In the Cognition Hypothesis. Few studies have examined systematically the combined effects of resource-directing and resource-dispersing variables on performance (but see Gilabert. however. Révész. The hypotheses put forward by the Limited Attentional Capacity model and the Cognition Hypothesis have received mixed support. Robinson proposes that increasing task complexity along resource-directing dimensions can lead to greater accuracy and grammatical complexity of L2 output because such demands can direct learners’ attention to how the concepts and functions required by the task have to be grammaticized using specific linguistic forms. whereas others simultaneously draw their attention to certain linguistic aspects of performance..pool of available attention and memory resources.g. some aspects of performance will be attended to while others will not. The Cognition Hypothesis states that sequencing tasks from cognitively simple to complex allows students to progress towards successfully performing real-world target tasks. 2005 Iwashita et al. As a result. 2001. Cognition Hypothesis differs from the Limited Attentional Capacity Model in that it assumes that attention is subject to voluntary regulation. Increasing complexity along resource-dispersing dimensions. although in 6 . 2009). whereas the resource-dispersing dimensions of the task act as attentional limitations in determining what aspect of the task can be heeded (e.

the resource-dispersing dimension of task complexity might play a different role than hypothesized for speaking tasks. These results were also replicated in a recent study by Ong and Zhang (2010) with regard to the fluency and lexical variety of argumentative tasks. Their results 7 . 1996). students do not need to simultaneously plan and linguistically encode their message. L2 writers might need to pay more attention to lexical and syntactic encoding in the composing process. Nevertheless. despite the fact that in writing the monitoring of the output does not need to be carried out parallel with producing the output. Moreover. which was not found to have an effect on the fluency and lexical variety of writing. Ong and Zhang also studied two further dimensions of task complexity particularly relevant to the writing process: the availability of previous drafts. and therefore L2 writers can focus on one stage of processing at a time.real-world communicative tasks these two dimensions simultaneously affect performance (for a recent discussion of this issue see Pallotti. Kuiken and Vedder’s (2008) research addressed the question how the number of elements students had to take into account in a letter writing task impacts on the linguistic characteristics of written tasks. which influenced the lexical variety of students’ output. and consequently might not be able to devote sufficient resources to the global organization of the text (McCutchen. Unlike in speech. Ellis and Yuan (2004) found that students in the pre-task planning condition wrote longer narrative texts and used a wider variety of syntactic structures as opposed to learners who were not allowed to make use of pre-task planning time. the limitation of attentional resources might influence writing processes especially in an L2. and the provision of ideas and macrostructure. Research on the role of task-types and cognitive complexity in L2 and FL written task performance has been scarce. (2009)). L2 writers might experience difficulties in heeding both linguistic accuracy and the discourse structure in the monitoring process. accuracy and lexical and syntactic complexity) of writing. It also needs to be considered that due to the fact that writing is often a less time-constrained activity than speaking. One line of investigations in this field has concentrated on how the provision of planning time affects the linguistic characteristics (fluency.

In the field of the assessment of second language writing. 1990) investigated the effect of outlining on writing fluency and quality among L1 writers and found that planning influenced specific qualities of the text but played a limited role in writing fluency. it is also important to consider how L1 writers perform under various task conditions as it gives an insight into which characteristics of the writing tasks are inherent in the task itself and which ones cause variation only among writers whose language proficiency is not fully developed. Ishikawa (2006) studied another dimension of task-complexity in narrative writing: the +/. which highlights the importance of examining particular features of the task that might contribute to variance in writing performance. they all seem to suggest that more complex writing tasks might have beneficial effects on writing quality. Schoonen’s (2005) study indicated that the type of text students had to produce in a test of writing had a greater effect on scores than the learners’ writing ability. 8 . students achieved higher scores than in tasks which were deemed easier. researchers have also investigated how different types of task affect test scores. In order to understand how particular writing tasks affect the composing processes and trigger particular language use. They argued that cognitively complex writing tasks might motivate learners to produce better performance than cognitively less complex tasks.Here and Now dimension. Although these studies analyzed different aspects of task complexity and used different measures to assess the linguistic quality of students. In another study. but in terms of the syntactic complexity and lexical variety no differences were observed. and thus their results are often contradictory. In a series of studies Kellogg (1988. Students who had to write a narrative in the past tense (-Here and Now condition) were found to be more accurate and fluent and displayed higher syntactic complexity than those whose essays had to be produced in the present tense (+Here and Now condition).indicated that students were more accurate in the cognitively more complex task in which they had to consider a higher number of elements. Hamp-Lyons and Mathias (1994) found that in tasks which were judged as more difficult by expert raters.

whose attentional capacities are often depleted by struggling with the linguistic encoding of their message. A number of indices of syntactic complexity. Kellogg (2001) pointed out that specific components of working memory are involved in nearly every stage of writing from planning to editing. As Schoonen et al. working memory limitations still play an important role in the various stages of writing. 2002).. but also in the quality of the writing produced (for a comprehensive review see Leki. Although the overall frequency of cohesive devices might not be a reliable indicator of the general quality of the text (Connor. The differences between L1 and L2 writers do not only manifest themselves in the psycholinguistic processes of composing. Cumming & Silva. below a certain threshold of FL linguistic knowledge. adequate L2 competence is an important prerequisite for the improvement of higher level writing skills such as planning.Another important issue to consider when comparing task effects in L1 and L2 is the difference between L1 and L2 composing processes.g. except for the most advanced ones. inhibiting writing processes such as planning or monitoring” (p. Working memory limitations and the lack of automaticity in lower level writing skills are particularly relevant for L2 writers. Despite the fact that writing is less constrained by time than speaking. 2008). 2009). Studies examining the development of writing skills among children also show that lower level writing skills such as hand-writing and spelling need to sufficiently automatic for children to be able to coherent texts (e. The cohesion of L1 and L2 writing was also found to differ across a number of dimensions. the frequency of sub-ordinate clauses. the writer will be fully absorbed in struggling with the language.81). In other words. the under. 1984). (2009) argue. such as the length of T-units.and 9 . Previous research also shows that texts produced by L1 writers are characterized by higher lexical variation and sophistication than essays produced by ESL learners (for a recent study see Crossley & McNamara. noun-modification. indicated that L2 texts tend to be syntactically less complex than L1 texts (for a comprehensive study see Hinkel. For most L2 students. the execution and monitoring stage of writing requires considerably more effort than for L1 writers. Bourdin & Fayol 1996). formulation and revision (Manchón et al. 2009).

The aims of the research The study reported in this paper intended to fill the apparent research niche concerning the linguistic and discourse structure of narratives in L1 and FL writing. How does the performance of L1 and FL writers differ in terms of lexical. syntactic and cohesive differences between two types of written narratives tasks: one in which the content of the narrative is given and another in which students need to plan the content of their stories? The following subsidiary research question was also posed: 2..over-use of specific cohesive markers by L2 writers might suggest that cohesive relations are not as effectively expressed by L2 learners as by L1 writers (for recent study on narratives see Kang. What are the lexical. The research presented in this paper is also novel in its selection of participants as it focuses on an under-represented group of FL writers: secondary school students (see Leki et al. 2008). Another aim of the research was to examine the role of task-complexity in second language writing. The main research question of the study was the following: 1. The Hungarian participants were students (N=44) in the second academic year of a bilingual education program. syntactic and cohesive characteristics in two narrative tasks which make different cognitive complexity demands on the stages of writing? Method Participants The study was conducted at two research cites: in a Hungarian-English bilingual secondary school in Budapest. In this project I collected narrative task data from 44 upper-intermediate Hungarian learners of English as well as from 10 young native speakers of British English. which consists of a so-called 10 . Hungary and at a university in the UK. 2005).

At the time of the research. The participants’ level of proficiency was rated as slightly above intermediate (corresponding to B1/B2 in the Common European Framework of Reference (Council of Europe. they had just begun the second academic year of their studies. The cartoon told a story of a car which broke down in the middle of 11 . which was the beginning of the autumn term of the first year of their studies. The input to the task was provided visually. with instructions in the first language of the participants. The participants in the UK were 10 under-graduate students (3 male and 7 female).zero year and four years of bilingual secondary education. they had limited experiences concerning academic writing at university level. who had just started their university studies at undergraduate level. They were between 18 and 20 years old and were studying Linguistics. which had to be included in the story. This population of students was selected to serve as a benchmark for writing performance expected from native English speakers who. The pictures were presented in the correct order and formed a coherent story line. Instruments The cartoon description task involved the description of a comic strip consisting of six pictures. During the zero-year the students took part in an intensive English language training program. The teaching method used was predominantly communicative combined with focus-on-form instruction. The participants of our study completed the zero-year program in June 2006 before starting their secondary studies in September 2006. The students all passed an A-level exam in English Language. which aimed to prepare them for studying several school subjects in English in the following four years of secondary school. 2001) by their teachers on the basis of the language tests that they administer to the students. after finishing their secondary education. and their age was between 17 and 18 years (27 female and 17 male). pursue studies at university level. however. At the time of the data-collection for the present research.

This task. which all had to be included in the narrative. could potentially elicit the use of certain low-frequency words such as carriage and wizard. however. the participants not only had to rely on their language skills. a storm on sea. They wrote their text in handwriting under the supervision of their classroom teachers. but they also had to use their imagination and find a way to relate the pictures to one another and invent a story around them. In order to successfully complete this task. and proved to be solvable within the available time-frame. in which these tasks were administered orally to a similar target population.the desert. The picture clues consisted of a book. The students could decide on the order in 12 . an island. and the stories produced did not reveal any misunderstandings of the plot either. This type of task did not require the conceptualization of the plot. this task can be characterized as cognitively more complex than the cartoon description task. The tasks were adapted from Albert’s (2007) dissertation. In terms of conceptualization. a house in a town. and an open door. Participants in Hungary completed two written narrative tasks one after the other during a regular English class. The car was transformed into a carriage without a horse by a wizard instead of being repaired. They were supervised by the author of this article and were not allowed to use any assistive tools or devices. The students were asked to work individually and were not allowed to use any dictionaries or thesauruses. The participants in both settings had 30 minutes to complete the two tasks. Students in the UK wrote the two narratives in a separate room at the university. and was consequently considered to place a relatively low cognitive load on the participants in terms of conceptualizing their message. They were instructed to write a minimum of 150 words for each task. In the pilot study students did not report any difficulties with understanding the cartoons. The picture narration task required students to tell a story based on six unrelated pictures. a boat. Procedures The tasks were previously piloted on a similar sample.

psy. 1997). This program identifies what percentage of the words students used in their narratives belongs to the most frequent 1000 words (Base-list 1) and 2000 words (Base-list 2) based on “A General Service List of English Words” by West (1953). Alternative assessments of lexical diversity also show very high correlation with D-value (McCarthy & Jarvis. McCarthy & Jarvis. Silverman.g. The calculation of the D-value is based on a mathematical probabilistic model as operationalized in the VOCD software of the CHILDES database (http//childes. Harris Wright. This measure has been used in a number of studies (e. we applied Malvern and Richards' (1997) D-formula. Malvern et al..g. it is still one of the most reliable measures of lexical diversity that can be calculated by available software programs. and 45% of the students chose to write the story narration task first. Analysis A large number of variables were used to assess various aspects of written performance (for a summary see Table 1). The program also assesses what percentage of words is frequent in upper secondary school and university texts from a wide range of subjects (Base-list 3). 2007.edu). 2002. 2003. Vocabulary range was measured with the help of Nation’s Range program (Heatley.which they completed the tasks. & Coxhead. 2002). Owen & Leonard. Nation. and although some researchers argue that the D-value is dependent on text-length (e. 2000). 2010).cmu. 13 . Lexical competence In order to measure lexical diversity. 2004. & Newhoff. This third frequency band is based on “The Academic Word List” by Coxhead (1998. which uses a random selection of tokens in plotting the curve of TTR against increasing token size for the text to be investigated (Malvern and Richards. 2010). Silverman & Bernstein Ratner. Approximately 55% of the students started with the cartoon description task. 2000).

temporal.0 program. which identifies the frequency of different types of connectives and provides spatial. Just & Carpenter. which reflects clausal complexity. 1976). 1985. Concreteness values in this program can range from 100 to 700. Syntactic complexity and accuracy Syntactic complexity was operationalized as the ratio of subordinate clauses. 1981). and which gives information about the complexity of noun-phrases.0 program. Connectives were selected because they are important devices to signal relations in text and as such contribute to coherent discourse (Halliday & Hasan. The skilful use of connectives can provide an insight into the textual quality of writing (Cameron et al. From among the measures of the frequency of content words. CohMetrix 2. which was calculated as the number of words within a clause (Norris & Ortega. An additional measure of syntactic complexity was the mean number of modifiers per noun-phrase. which was calculated using the Coh-Metrix 2. Cohesion Cohesion indices were also obtained with the use of the Coh-Metrix 2. the log frequency of content words was selected because it was found to be a more reliable indicator of lexical complexity than the raw frequency of content words (Haberlandt & Graesser. CohMetrix provides concreteness ratings based on the MRC Psycholinguistics Database (Coltheart. 14 .0 also provides a measure of the mean number of words before the main verb. 1980).The Coh-Metrix 2. In the case of second language writers. which was expressed relative to the total number of clauses and as the length of clauses.0 program was used to analyze the concreteness of content words. CohMetrix 2. which was calculated relative to the total number of clauses. 1995). where a high value indicates that the words used tend to be highly concrete and a low value that the words are abstract.0 also calculates the frequency of content words. 2009).. accuracy of performance was measured with the ratio of error-free clauses. intentional and causal cohesion indices.

when they extend the information provided in the text (e. Causal cohesion is a measure of the extent to which explicit causal links between sentences are expressed and is particularly relevant in narrating events.Connectives can be characterized as positive. 1978) and latent semantic analysis (Deerwester. intentionality. Zwaan and Radvansky (1998) identified five situational dimensions of cohesion: causation. van Dijk & Kintsch. In preparation for the computer analysis. but due to the fact that the intentional cohesion measures were 0 for the majority of participants.g. Therefore I decided to apply an analysis of the situational dimension of cohesion (Kintsch. after. causal and logical (Halliday & Hasan. 1983). and. 1998. Coh-Metrix 2. Connectives can be also classified as additive. 1990. which reflects the extent to which verb tenses and aspects assist in the formation of a cohesive text. however. when they restrict and cease to elaborate information (e. until). nouns and verbs. temporal and spatial cohesion. the incidence of positive and negative additive (e. therefore. Sanders. Spooren & Noordman. unless) was measured using the Coh-Metrix 2. Spatial cohesion. after). 1997). particles. which is derived from the the incidence of location and prepositions. Although there are other ways of analyzing cohesion such as coreference and argument overlap (Kintsch & Van Dijk. the figures for intentional cohesion are not reported in this study. because. also). 1992). which is a measure of how spatial content and the sentences are related by spatial particles or relations. and as negative. These dimensions of cohesion are expressed with the help of connectives. but. temporal. Furnas.0 program. temporal (when. Landauer & Dumais. until.g. these are difficult to measure reliably for short texts.0 can provide indices for causal. intentional. although) and logical connectives (or. Temporal cohesion. Landuaer & Harshman. location nouns and motion verbs. but. 2002. The index of causal cohesion is calculated by dividing the total number of causal verbs with causal particles. and. although) (Louwerse.g. Based on this classification system. all texts were parsed by the author for ambiguous sentence structure and punctuation that might have potentially affected the interpretation of the text 15 . is calculated by averaging the location and motion ratio scores. if. Dumais. causal (because. is operationalized as the averaged repetition score for tense and aspect. space and protagonists. 1976). time.

The picture narration task elicited more abstract words in general as indicated by the concreteness index F (52) =20. both effect sizes (eta squared) and statistical power (P) were used. The other task effect was found in the frequency of positive temporal connectives F (52) =5. The Cohen’s kappa for inter-rater agreement was 0. A small selection of texts was resubmitted for computer analysis to check if the same values for the linguistic and cohesive values were obtained. 1988).30.85. Insert Table 1 around here Results The factorial analysis of variance detected only a small number of task effects (see Tables 2 and 3).67. eta squared =. No instances of ambiguity were found and occasionally missing punctuation marks were manually inserted.0 (Statistical Package for Social Sciences) was used. which shows that the difference between these tasks with regard to concreteness is indeed important. P =. Due to the fact that the analyzed variables were normally distributed in both groups of participants. In all cases. For the statistical analysis SPSS 13. The effect size can be regarded as large (Cohen. and the statistical power was also lower than 16 .54. Main effects as well as interactions of these variables were computed. however only in the medium range. In order to ensure the reliability of the coding of the variables that were not rated by computer analysis.by Coh-Metrix. another rater was asked to code the data. In order to assess the importance of differences caused by the task and native status variables. and the statistical power is also high. The effect size for this difference was.64). the computer analysis produced the same result. parametric statistics were used despite the differing number of students in the groups. P =. eta squared =. The main statistical analysis used was the factorial analysis of variance with the task constituting the within subjects variable and native/non-native status the between subjects variable.99).10.

eta squared =. FL writers seemed to have used significantly less varied vocabulary in both tasks (F (52) =20.90. ------------------------------------Insert Tables 2 and 3 here ------------------------------------- The results reveal significant differences between L1 and FL writers in terms of lexical variety and complexity (see Tables 2 and 3). which suggests that the task condition had a limited effect on the use of temporal connectives.48). we can conclude that the need to conceptualize the story-line did not seem to result in major linguistic and cohesive changes.75. The eta squared values indicate a large effect size and the statistical power for these differences is also high. Similarly. Non-native writers also produced significantly shorter clauses than the native speaking participants (F (52) =4. 1983). P =1.09. P =.07. eta squared =.expected. but the effect size values and the statistical power analysis suggests that this difference is not of high importance.58). Discussion As regards the effect of task demands on narrative performance. the effect of native speaker status on causal cohesion can also be regarded as of medium size (F (52) =3. P =. eta squared =. eta squared =. Genre is one of the important factors that affects the use of cohesive devices in writing (Smith & Frawley.31. and it also influences the 17 .24.43.0) than their native counterparts.63.99) and significantly more words from the first 1000 word-range (F (52) =36. P =. It can be argued that the lack of substantial difference between the two types of tasks is due to the fact that both tasks required learners to write in the same genre.

the competition of attentional resources in writing tasks should also be considered with regard to cohesion. accuracy and complexity. writing fluency. it can be inferred that the differences in language use were caused by the task demands. accuracy and complexity. A tentative explanation might be that the fact that the participants had to plan the content of the story might have taken away from the attentional resources available for the explicit signaling of cohesive relations.lexical and syntactic range of expression as well the use of connectives (for a review see Biber & Conrad. however. The task in which students had to narrate predetermined content. These findings are particularly interesting because in the lack of significant interaction between task and native-speaker status. Consequently. Therefore. the results can be taken to suggest that the fact that students had to encode given content might have promoted the use of more abstract vocabulary in the cartoon description task irrespective of whether students wrote in their L1 or in a FL. A possible implication of this both for second language acquisition research in general and for the study of writing in particular might be that in addition to the traditional three aspects of task-based performance that are juxtaposed task based research: fluency. Both L1 and FL participants also employed temporal connectives to indicate the sequence of events in the cartoon description task more frequently than in the picture narration task. 2009). seems to have promoted the use of more abstract vocabulary and triggered the use of more explicit signaling of temporal relations. The results concerning accuracy and syntactic complexity indices reveal that at the upperintermediate level writers in this study did not avoid the use of complex syntactic structures and constructions that they have not yet fully mastered when given the opportunity to tailor a narrative text to match their own linguistic resources. If we examine the overall frequency of connectives in the two tasks we can notice that L1 and FL writers used slightly fewer connectives overall in the picture narration task. it might be assumed that both types of tasks provide similar opportunities for learners to display their linguistic competence in writing. 18 .

as measured by type-taken ratio adjusted for text length. it is also possible that certain differences in cognitive complexity demands such as the one in the current study have a limited effect on the overall quality of writing. 2005) Cognition Hypothesis. but they did not need to plan the story. are not in line with Skehan and Foster’s (2001) Limited Attentional Capacity Model either. in these tasks the two types of complexity demands: having to plan the content vs. as the two tasks displayed highly similar accuracy values. In the cartoon description task the FL writers were required to express the storyline with whatever linguistic knowledge they had available. however. however. The picture narration task can be considered the opposite. On the other hand. and the difference in lexical complexity is in the opposite direction predicted by the Cognition Hypothesis. complexity factors and analytic tools employed in them. as here students had to devise a plot. The results. If the task in which students need to plan the content of their narrative is considered to be cognitively more complex. The comparison of our study with previous investigations on written task complexity is difficult due to the different types of tasks. the time given for producing the narrative texts in this research might have allowed students to utilize their attentional resources efficiently regardless of the fact whether they had to plan the content of the story or not. but they could potentially avoid the use of linguistic constructions which are beyond their competence.As for the competing theories of the role of attention in task performance. which in the past have mainly been investigated in the context of L2 speech production. our findings with regard to FL writing suggest that the most important resource-directing characteristic of the cartoon description task is pushing learners to use more sophisticated vocabulary and thereby increasing lexical complexity. which can be considered the most similar to the present research. was 19 . The results concerning the accuracy and syntactic measures. In the study. Ong and Zhang (2010) found that lexical complexity. encoding given content with the available linguistic resources might have extinguished each other’s effect. as except for one measure of lexical complexity. students’ performance in the two tasks was similar. On the one hand. 2003. suggest that this was not the case. the findings do not provide support for Robinson’s (2001. Nevertheless.

Leki et al. Connor. but the cohesive features of the narratives show similarities in the two groups of participants. also found that essays produced by highly proficient students did not display significantly different spatial. Our findings in this respect seem to be different as the task with given content elicited higher lexical complexity as measured by the frequency of abstract words. It is important to note. whereas in our project we investigated narratives. The findings concerning the lack of significant differences in cohesion seem to be contradictory to the results concerning expository and argumentative genres (e. but these differences only have a medium effect size. The findings indicate significant differences between L1 and FL narratives in terms of lexical variety. and no differences between tasks were found with regard to D-value. causal and temporal cohesive markers than those written at lower level of FL competence. in which Coh-Metrix was used to compare the cohesion of high and low rated FL essays. Interestingly. Tapper. that both Crossley and McNamara’s 20 . 2005). complexity and syntactic complexity. 1984.. The reasons for the similar frequency of co-ordinating conjunctions in this study might be related to the characteristics of the narrative genre and to the fact that no differences between in L1 and FL students in the ratio of sub-ordinate clauses was found either. 2008. which is the variable that can be regarded similar to the type-token ratio used in Ong and Zhang’s study. The reason for this difference might be related to the fact that in Ong and Zhang’s research participants were required to produce an argumentative text. in which FL learners were generally found to use a higher frequency of co-ordinating conjunctions than L1 writers.g. The only difference that might indicate higher level of cohesion in L1 writing can be observed in the figures of causal cohesion. the values of which are almost twice as high in the L1 narratives than in FL texts. however. It seems that the upper-intermediate learners participating in this study were aware of the syntactic and cohesive demands of narratives in terms of co-ordination and conjunction and had the linguistic means to express these relations. Crossley and McNamara’s (in press) recent study.higher in the task condition where students were provided with the least amount of help concerning the content of essay than in the two other conditions in which less support was given with regard to content.

. in press. 1988.g. it 21 . This finding adds another piece of evidence for the conclusion of other studies in second language writing and assessment (e.(2010) and our analysis only investigated the frequency of explicit markers of cohesion and was not suitable for detecting qualitative differences in the use of cohesive markers. 2009. these results show that in several areas of lexical competence (Meara. Reid. Salsbury. McNamara & Jarvis. 1978. 1992. Silva. The differences between previous investigations and our study might be explained with the discourse characteristics of narratives. namely to name. in which FL writers produced longer clauses (Gates. Therefore. 1993). FL writers were found to use more words from the first most frequent 1000 words. Our findings reveal that FL writers seem to use lexically less varied language. Silva. 2003.. Crossley. to set up expectations about narrative entities and events and to sum up past or upcoming events (Dasinger & Toupin. Meara & Bell. The fact that the frequency of cohesive links as assessed by the Coh-Metrix program is similar does not mean that L1 and FL writers used these markers with similar efficiency and variety. These results seem to be different from the findings of studies on expository texts. and their writing is lexically less dense as indicated by the frequency of content words. In addition. but in terms of the mean number of words before main verbs and subordination complexity no differences between L1 and FL narratives was observed. in press. density and range. 1982. Taken together. Park. FL writers were found to display lower levels of clausal complexity as indicated by the variables of clause length. to present main characters. enable and continue narrative actions. Jarvis et al. 1993) and used fewer sub-ordinate clauses (Hu et al. 2002) such as lexical variety. which reflects smaller vocabulary range. as shown by the D-value. Sub-ordinate clauses in general and relative clauses in particular have several important discourse and linguistic functions in narratives. Crossley & McNamara. situate and identify old and new referents in the story. 1994). 2001) that lexical profiency is an important component of writing competence and an area FL writers need to focus on if they want to improve their composition skills. FL writers at upperintermediate level show a significantly different profile from native speakers. to motivate.

Moreover. clausal complexity and causal cohesion proved to be the most useful in describing task and proficiency-related variation. I argued that in addition to assuming that cognitively complex concepts increase task complexity both in the planning and formulation stage of writing. these variables have the potential to distinguish lower and higher levels of language competence. The study also explored how a wide range of linguistic variables and automated analyses can be used in investigations of writing performance. the indices measuring lexical competence. tasks can also pose independent complexity demands in separate stages of the writing process depending on whether they require the expression of pre-determined content. the differences between L1 and FL writers can be assumed to be caused by the fact that L1 writers used a considerably higher number of prepositional phrases and adverbs to add detail and emphasis to their stories. which gives learners diagnostic information with regard to their strengths and weaknesses and can foster learner autonomy in the acquisition of FL writing skills. From a theoretical perspective.is probably due to these unavoidable functions of subordinate clauses that both L1 and FL writers used them with similar frequency in their narratives. the findings suggest that in automated assessment. In this paper I proposed new ways in which the cognitive complexity of tasks can be understood and analyzed in FL writing with reference to the different stages of the composing process. several of the performance measures employed in this 22 . The automated analysis exemplified in this study might not only be used in high-stakes testing contexts but also for self-assessment. With regard to writing assessment. Conclusion This paper demonstrated how issues traditionally raised in task-based research with reference to speaking tasks can be investigated in the case of FL writing and how methods of analyses from the fields of writing research and psycholinguistic investigations of speech production can be combined to gain an insight into the characteristics of FL writing processes. With regard to the length of main clauses.

Our study has also shown that the major difference between L1 and FL writers can be found with relation to lexical variety. Further research might be necessary to gain more insight into how different task-complexity demands influence the linguistic quality of FL writing in other types of tasks and in additional genres. which had previously only been used in the investigations of FL writing. Furthermore.study. FL writers should be encouraged to use a wider range of complex words productively in their writing through lexically focussed pre-writing and revision tasks. It would also be important to investigate the interaction of FL proficiency with complexity demands in different types of tasks. our research provides further evidence that even at higher levels of L2 competence. Further qualitative analyses of cohesive markers and of task-specific linguistic 23 . I explored how specific linguistic and discourse characteristics of narrative tasks are influenced by aspects of task-complexity in L1 and FL writing. This indicates that allowing students to generate their own content in narrative writing tasks does not lead to the avoidance of linguistic constructions that the students do not fully master yet. but should also include a lexical development component. writing assessment and pedagogy might give preference to these types of tasks over those in which students have to narrate predetermined content. Therefore. writing classes would not only need to focus on the improvement of text-level composition skills. sophistication and range. The variable whether students had to narrate a story with a given content or whether they were free to plan the plot of the story exerted a major impact only on one measure of lexical sophistication and had a minor effect on the overt expression of temporal cohesion. could also be applied in the study of L2 speech and yield new insights into how task demands affect cohesion and lexical variety and complexity in speech production. In the research reported here. A larger and longer sample of writing might also allow for a more detailed analysis of coherence relations in texts and for more sophisticated statistical analysis of the data. Given that tasks in which learners have freedom over content might promote the development of textual organization skills and might be more motivating and engaging.

& Linton. K.. M. 245-264. S.. Cohen. 18. Computational assessment of lexical differences in L1 and L2 writing.). 15. M. A. B. M. C. Cambridge: Cambridge University Press. S. Journal of Second Language Writing. S... Text cohesion in children’s narrative writing. U. (1984). 119-135. Webster.constructions that examine variety and accuracy of use could also provide more insight into taskand proficiency-related variation in FL writing performance. (1996). 18. Á.. 17. A. 301–316. Biber. Quarterly Journal of Experimental Psychology. Lee. A. D. Bereiter. Bourdin. Common European Framework of Reference for Languages: Learning. 213-238. 33. TESOL Quarterly. Connor. Unpublished PhD dissertation. Hunt. Cameron. 497–505. (2007). J. D.. Hillsdale. NJ: Lawrence Earlbaum. ELI Occasional Publications. Hillsdale. 16. C. NJ: Erlbaum. The psychology of written composition. Crossley. Statistical power analysis for the behavioral sciences (2nd ed. (1988). Vol. and style. Budapest: Eötvös Loránd University. Council of Europe (2001). (2009). & McNamara. A new academic word list. Cambridge: Cambridge University Press. 34. Mode effects in a sentence production span task. (1987). assessment.. Register. References Albert. Coltheart. M. genre. 257–269. Applied Psycholinguistics. (1981). & Scardamalia. 24 .. Coxhead. 2009. The MRC psycholinguistic database. Coxhead. Cahiers De Psychologie Cognitive – Current Psychology of Cognition. (2000). Victoria University of Wellington: Wellington. Creativity and task performance. A. (1995). A study of cohesion and coherence in English as a second language students’ writing. teaching. & Fayol. Papers in Linguistics: International Journal of Human Communication. K. An Academic Word List.. Munro. S. (1998). & Conrad.

2026A Gilabert. & Jarvis. NJ: Lawrence Erlbaum. and American. P. 39 (4)..A. Journal of Research in Reading. D.S. Native speakers and task performance: Comparing effects on complexity. Gregg & E. Hillsdale. MA: MIT Press.. (2005). & Tavakoli. J. and lexical difficulty. R.Crossley. Language Learning. D. The effects of planning on fluency. Landauer.. R. W. (1978). & McNamara. Lancaster. (1980). 31-50). & Hayes. Task complexity and L2 narrative oral production. 866-896. Flower. Paper presented at the 3rd Biennial Conference of Task-based Language Teaching. F. (2009). Ellis. Crossley. & McNamara. R..). L. 391-407. (in press). Steinberg (Eds. Unpublished doctoral thesis. Relating events in a narrative: a crosslinguistic developmental study (pp. fluency and lexical diversity. Task complexity: models. The development of relative clause functions in narrative.S.). T.. 59–84 Fellbaum. Cambridge. S. readability. D. NJ: Lawrence Erlbaum. Cognitive processes in writing (pp. Dasinger. 41. S. complexity and accuracy in second language narrative writing. Gates. C. Journal of the American Society For Information Science. UK. WordNet: An electronic lexical database. 452-514). R. Predicting second language writing proficiency: The role of cohesion.. Dissertation Abstracts International. University of Barcelona. 59. Deerwester. Predicting lexical proficiency in language learners using computational indices. Furnas. T. (2004).. An analysis of grammatical structures in compositions written by four groups of students at Southern Technical Institute: Indo-Iranian. P. In L.. & Harshman. (2009). C. Latin-Romance. G. & Toupin. T. & Yuan. Sino-Tibetan. and the need for a paradigm shift. Studies in Second Language Acquisition. Hillsdale. S. Ellis. K.. (1998). predictions. S. Salsbury. (1994). Dumais. (in press). Foster. I. (1990). Language Testing. 26. Indexing by latent semantic analysis.A. Slobin (Eds.. 25 . L. Berman & D. S. In R.. The dynamics of composing: Making plans and juggling constraints.

T.S.aspx Hinkel. Language Learning. Journal of Second Language Writing. In C. A. Heatley. L.. T.. Clevedon. Halliday. 114..Garcia-Mayo (Ed. The effect of manipulating task complexity along the (_Here-and_Now) dimension on L2 written narrative discourse. (2008). Hamp-Lyons.. Language Learning and Communication. 3.. Brown. RANGE and FREQUENCY programs. I.ac. (1982). 357–374. I. 39-49. K.). & Hasan. J. A. 259-279. 14. & Elder. & Carpenter. Psychological Review. Silverman. & Silva. S. A. Z. P. W. Hu.. Aphasiology. D.. (1976). 26 . & Mathias. M. NJ: Lawrence Erlbaum. Cohesion in English. http://www. A theory of reading: From eye fixations to comprehension. (2002). M. (1985).. Mahwah. Component processes in text comprehension and some of their interactions. Examining expert judgments of task difficulty on essay tests.. London: Longman. (1994).. Harris Wright. (1980). A. R. 51. Investigating tasks in formal language learning (pp. Journal of Experimental Psychology: General. London: Routledge. & Graesser. 1. A synthesis of research on second language writing. E. Iwashita. K. M. & Coxhead.nz/lals/resources/range. Can we predict task difficulty in an oral proficiency test? Exploring the potential of an information-processing approach to task design.. C. Ishikawa. P. 443-452. Written narratives as an index of L2 competence in Korean EFL learners. Journal of Second Language Writing. 49–68. (2002). (2005). & Newhoff.. Measures of lexical diversity in aphasia. Just. & Brown. 401-436.Haberlandt. Cumming. (2001). 136–156). (2006). 17. A. A.P. McNamara. 329–354.. F. Some linguistic differences in the written English of Chinese and Australian students. Leki. Kang. C. T.. S. L. Y. N. UK: Multilingual Matters. M. 87. H. (2003). Nation. Second language writers’ text: Linguistic and rhetorical features.victoria. A.

R. Cognitive task complexity and written output in Italian and French as a foreign language. W. In R. Mahwah. 57-71).) The science of writing: Theories. Louwerse.. R. R.Kellogg. UK: Multilingual Matters.). individual differences and applications (pp. and research (pp. 14. A model of working memory in writing. (1996). Landauer. 27 . R. 21. Writing in foreign language contexts: Learning. Kellogg. M. An analytic and cognitive parameterization of coherence relations.M. L. I. . Effectiveness of prewriting strategies as a function of task demands. 211–240. (2009).. 48-60. Psychological Review. Memory and Cognition. M. teaching. and representation of knowledge. (2001).American Journal of Psychology. (2002). & Vedder. J. Kintsch. American Journal of Psychology. induction. Levy and S. Clevedon. Competition for working memory among writing processes. 102-129). & Dumais. M. 363–394. Journal of Experimental Psychology: Learning. 355-365. S. Kellogg. (1998). 15-35. & Murphy. Attentional overload and writing performance: Effects of rough draft and outline strategies. Psychological Review. Comprehension: A paradigm for cognition. T. T. Manchón (Ed. T. 103. 114. NJ: Lawrence Erlbaum. Roca de Larios. T. 327-342. Toward a model of text comprehension and production. (1988). Kintsch. Cambridge: Cambridge University Press. In C. Cognitive Linguistics. The temporal dimension and problemsolving nature of foreign language composing processes: Implications for theory. & Van Dijk. methods. A solution to Plato’s problem: The latent semantic analysis theory of the acquisition... F. 104. Ransdell (Eds. Kuiken. 175-191. (1990). Manchón. W. (1997). K. 85. M. R. 17. (2008). (1978). A. Journal of Second Language Writing. Kellogg.

Malvern. M. (1996). 299-325. J. M.. 555-578. R. Meara. P.. McCutchen. Coh-MetrixCoh-Metrix: Automated cohesion and coherence scores to predict text readability and facilitate comprehension. D. L. Beckwith. M. A theoretical and empirical evaluation of vocd. 459-488. J. A new measure of lexical diversity. & Ortega. Meara. Richards. D. P. (1997). Ryan and A.. M. Cognitive Science Laboratory. (2010). Norris. S. B. 18. Gross. 28 . Effects of task complexity on the fluency and lexical complexity in EFL students’ argumentative writing. NH: Palgrave Macmillan.. D. D. McCarthy.. & Jarvis. Prospect. G. 24. & Miller. Ong.. M. & Jarvis. P. Behaviour Research Methods. B.A. and HD-D: A validation study of sophisticated approaches to lexical diversity assessment. Second Language Research. & Graesser. Evolving models of language. M. A. L. vocd-D. Journal of Second Language Writing. P. (1990). J. (2007). J. Princeton University . H. 24. A capacity theory of writing: Working memory in composition. 219-233. D. SA. In A. and Richards. 19. 42. University of Memphis. D. & Jarvis.. K. & Bell. 30. 393–407. 8. Language Testing. (2002). Chipere. N. Miller. & Durán. (2004). 5-19. S. P. No. Wray (Eds. Memphis. 459-488. Clevedon: Multilingual Matters. Lexical diversity and language development: Quantification and assessment. Houndmills.McNamara. (2001). (2009). Malvern. S. Fellbaum. 381-392.. P. MTLD. The rediscovery of vocabulary. D. J. Language Testing. 43. P_Lex: a simple and effective way of describing the lexical characteristics of short L2 texts. Louwerse. A theoretical and empirical evaluation of vocd. 16... TN McCarthy. Applied Linguistics. Educational Psychology Review. Five papers on WordNet. & Zhang..). (2010). Institute for Intelligent Systems.. Towards an organic approach to investigating CAF in instructed SLA: The case of complexity. (2002). C. C. McCarthy. (2007).

Applied Linguistics. refining and differentiating constructs. M. Towards a blueprint of the foreign language writer: The linguistic and cognitive demands of foreign language writing. Robinson. Stevenson. Snellings. 1-32. (1992). P. J. G. CAF: Defining. 15.. 31. 1-35. M. 21. Toward a taxonomy of coherence relations. task difficulty. Academic and ethnic background as factors affecting writing performance. Tasks in second language learning. Generalizability of writing scores. In R. In A. 22. teaching and research (pp. 590-601. van Gelderen. Second Language Studies. W. (2009). M. P.. Manchon (Ed. Reid. 1. (2005). 29 . J. Park. CA: Sage. The Cognition Hypothesis of adult. Language Testing. & Noordman. P. P. Writing across languages and cultures: Issues in contrastive rhetoric (pp. Y (1988). (1992). Learning. and second language development. M. 43.Pallotti. P. V. (2009) Task complexity. Bristol: Multilingual Matters. 261272). A. M. R. Journal of Second Language Writing. Révész.. Task complexity. 30. (2003). 27-57. Robinson. 45–107. (2009). M. 22. G. Purves (Ed. Schoonen.). (2001).. and task production: Exploring interactions in a componential framework. Sanders. Robinson. International Review of Applied Linguistics. Cognitive complexity and task sequencing: Studies in a componential framework for second language task design. Newbury Park. (2008). 77-101).. Applied Linguistics.). Writing in foreign language contexts. Spooren. & Bygate. A. An application of structural equation modeling. Basingstoke: Palgrave. T. Discourse Processes. (2005). Schoonen. L. 79–107. 437-470. focus on form. A computer text analysis of four cohesion devices in English discourse by native and non-native writers. R. Samuda. Studies in Second Language Acquisition. task-based language learning. 1-30.

Longman. L. 347-373. R. & Kintsch. J. 1363-1377.. & Bernstein Ratner. Connectives in advanced Swedish EFL learners' written English . T.. 183-205). 48. New York : Academic Press. 5. S. Eisenberg. J.. 510-532. Skehan. P. 116-144. The development of expressive elaboration in fictional narratives. and lexis. 45-72. Language. accuracy. Kaderavek. M. W. A. B. TESOL Quarterly. Cognition and tasks. W. Modelling second language performance: Integrating complexity. Applied Linguistics. & Foster. (2005). T. (2005). Gillam. Conjuctive cohesion in four different English genres. (2009). Mathematical Scientist. & Frawley. London Zwaan. Tapper. & Radvansky. (1983). R. (1953). Justice. Smith. Harm. Skehan. P. A. West. W.11. P. (2001). R. 123. Van Dijk. Silverman. Journal of Speech. Cognition and second language instruction (pp. Robinson (Ed. 30.657–675. M. Strategies of discourse comprehension. Situation models in language comprehension and memory. Summary of task performance measures used in the study Lexical variety D-value Lexical complexity Lexical range Log. The Department of English: working papers in English linguistics. N.). frequency of content words Concreteness Syntactic complexity Clause length Ratio of subordinate clauses Words before Accuracy Ratio of error-free clauses 30 Cohesion Causal Temporal Spatial Frequency of connectives Positive and negative additive Positive and negative causal Positive and . Text. T..preliminary results. (1998). and Hearing Research. H. Word frequency distributions and type–token characteristics. G. M.A.. L. In P.. (2000). N.Silva. (1993). N. 162-185. Cambridge: Cambridge University Press. 27. Toward an understanding of the distinct nature of L2 writing: The ESL research and its implications. (1983).. 3.. Table 1. A.. Ukrainetz. M. fluency. A general service list of English words. S. Psychological Bulletin..

main verbs Modifiers per NP 31 negative temporal Positive and negative logical .

90 (26.18) 0.86 1.02) 58.30) 20.82 (0.14) 10.66) 84.81 (0.91) 416.12) .96) 2.34 (.84) 6.24** 0.48 (.25 (3.33) 82.92 b (.34) 420.43 (1.29 (.91) 6.54** .65 (1.63 (2.82 (.30) 3.01) 0.01 12.78 (6.02 Ratio of subordinate clauses .79) 2.93) 398.47 (.76 (.12) .95) 3.15) 2. frequency content words 2.03 0.34 (.32) 3.13 Words before main verb 2.01) .90* 3.21) 2.01 Ratio of errorfree clauses 0.22) .05 level.5 (20. Denotes significant effect at the p <.12 1.56 1.44 (1.53 Lexical variety D value F task F nativeness F interaction Accuracy 0.5 (16.98) 6.99 (4.18) 84.04 (1.33 Syntactic complexity * Denotes significant effect at the p <.23 (.34 Log.04 Concreteness 394.28) 0.34 (17.44 1.15 b (2.71 (.51) 4.94) 2.08 (20.05 2.42 Lexical complexity Range Wordlist 1 88.12) 2. ** 32 .02 Range Wordlist 3 .71) 0.80 (.73 (.28) 3.12) 0.47 (3.13** 0.00) 74.06 Range Wordlist 2 6.8) 0.56) 89.27 0.28 4.12) .35 (4.04 0.75** .10) .21 (.97 Modifiers per NP .42 (16.09) .75 (2.79 (3.02 20.48) 10.03 2.69 (.Table 2 Descriptive statistics of the linguistic variables Non-native speakers Mean (SD) Cartoon Picture description narration Native speakers Mean (SD) Cartoon Picture description narration 62.01 level.98 (21.99) 9.44 (1.54 0.79) 0.41 (17.01 36.9) Clause length 5.97 3.46 (2.35 (.83 (4.25 (.

88 (11.66 (13.04 (8.62 (10.16) .89 .27) 9.12 (2.80 a (14.41) .48 .77) .43) 27.70) (.13 (9.06 .86) 11.73 a (11.14 (.46 (8.17) .17) 16.12) (.03) .54) 23.53 (7.07) 8.85 .80 .65) Denotes significant difference between cartoon description and picture narration in the NNS group at the p <.33) 9.69) .44) .49 (8.55) 34.35) 9.16) 19.46 (.42 (1.93 (7.92) 1.23) 21.64) .40) (.38 (13.21) Native speakers Cartoon Picture description narration Mean Mean (SD) (SD) .01 level 33 .50) 24.18) 37.06 36.89) 0 (0) 24.43 (1.Table 3 The descriptive statistics for the different cohesion indices Variables causal cohesion temporal cohesion spatial cohesion Frequency of connectives (per 1000 words) positive additive negative additive positive causal negative causal positive temporal negative temporal positive logical negative logical a Non-native speakers Cartoon Picture description narration Mean Mean (SD) (SD) .59 (.28) 15.86) 11.27 (9.64 (8.19) 23.90 (.69 (14.23 a (14.08) (.11) (.47 (9.51) .35 (5.83 (.42 .73 (.10) .42 (1.16 (1.35 (13.02 (7.06) 13.41.08) 35.53 (16.14) 8.95 (12.45 .28) 9.85 a (10.43 . (.54) 19.95 (6.28) 9.