Journal of Second Language Writing 30 (2015) 45–57

Contents lists available at ScienceDirect

Journal of Second Language Writing

The effects of cognitive task complexity on writing complexity

Mark Wain Frear *, John Bitchener
AUT University, Auckland, New Zealand

A R T I C L E I N F O A B S T R A C T

Article history: This study reports the findings of a within-subject experimental study that examined the
Received 19 January 2014 relationship between increases in cognitive task complexity and the writing of
Received in revised form 20 August 2015 intermediate L2 writers of English. Potential effects on lexical and syntactic complexity
Accepted 21 August 2015
were investigated. This article expands on past writing research using similar cognitive
Available online 15 September 2015
task complexity by adding a patently low complexity task to better track the effects of
complexity, and a subordination measure that investigates each dependent clause
Keywords:
separately. Thirty-four non-native speakers of English studying at language schools in
Cognitive task complexity
Second language writing New Zealand performed three letter-writing tasks of varying levels of task complexity. The
Complex output findings revealed a significant effect for task complexity on decreases in syntactic
complexity using a ratio of dependent clauses to T-units measure where independent
clauses were measured separately. Conversely, significant findings were found for
increases in lexical complexity, analysed as a mean segmental type-token ratio. The results
of this study are discussed in relation to the Cognition Hypothesis (Robinson, 2001a,
2001b, 2005, 2007, 2011).
ß 2015 Elsevier Inc. All rights reserved.

Introduction

Centrally, the present study contributes to on-going investigations into the suitability of writing as a beneficial medium
for promoting the development of complex language in L2 learners (Ishikawa, 2007; Kormos, 2011; Kuiken & Vedder, 2007,
2008, 2012; Ong & Zhang, 2010; Sercu, De Wachter, Peters, Kuiken, & Vedder, 2006). The emphasis is on whether the
manipulation of certain task-based variables, believed to affect a writer’s cognitive burden and thus attention, will have a
subsequent effect on their ability to produce attention-demanding elements in a written text.
Ostensibly, writing may not appear to be the best format for developing complex linguistic structures that require
attention, especially when they are not proceduralized or automatized. For example, the act of writing has been
characterized as a problem solving activity (Belcher & Hirvela, 2001) requiring a writer’s constant management of limited
attentional resources (Flower & Hayes, 1981). Thus, it appears counterintuitive that making writing more difficult by
increasing the burden on attentional resources has any beneficial effects. However, writing has also been ascribed intrinsic
characteristics such as recursion, planning time, and selective control (Ellis & Yuan, 2004; Grabe & Kaplan, 1996; Grabowski,
2007; Kormos & Trebits, 2012) that are potentially conducive to the production of complex linguistic structures that require
attention.

* Corresponding author.
E-mail address: mwfrear@[Link] (M.W. Frear).

[Link]
1060-3743/ß 2015 Elsevier Inc. All rights reserved.
46 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57

To date, task-based research with foci on the relationship between task complexity (as cognitive burden manipulation)
and writing is growing, but it still remains a lesser domain than the oral modality, which has historically been the main focus
of research (Carless, 2012). In the developing field of complexity and L2 writing, there have been a number of means by
which cognitive burden has been increased or decreased on writers, while a wide range of potential effects has also been
investigated. Variables have included the effects of planning time, writing assistance, and editing time on fluency, lexical
complexity, metacognitive process, and text quality (Ong, 2014; Ong & Zhang, 2010, 2013). Kormos (2011) tested the
addition and removal of narrative context on complexity, accuracy, and fluency in L1 (native language) and FL (foreign
language) writing. Ishikawa (2007) examined the manipulation of here-and-now variables (and the incidental addition of
planning time) on measures of accuracy, complexity, and fluency, and Kuiken and Vedder (2007, 2008, 2012) and Sercu et al.
(2006) investigated the manipulation of reasoning demands and number of elements on complexity, accuracy, and fluency.
Considering the variations in both independent and dependent variables across this relatively small pool of research,
there is a need for partial replication of some of these studies. Porte and Richards (2012) note the importance of partially
replicating studies in L2 writing (in which central elements of an original study remain the same, but non-major aspects are
varied between the past and present to facilitate comparisons). They also claim that replication or partial replication studies
are important ways to test the robustness of past research and address the disjointed and conflicting findings resulting from
the growing diversity of scope and topic in L2 writing research. We agree that sustained investigations may contribute to
fully understanding the relationship between one set of variables and any extra unaccounted for variables that might be
affecting results.
The present study is a partial replication study of Kuiken and Vedder (2007, 2008, 2012) and Sercu et al. (2006). Similar
measures are used to operationalize the independent variables (cognitive task complexity) and the dependent variables
(syntactic and lexical complexity). Additionally, similar frameworks have been utilized to make predictions about the
relationships between cognitive task complexity and linguistic complexity. These are the Cognition Hypothesis (Robinson,
2001a, 2001b, 2005, 2007, 2011) and the Trade-off Hypothesis (Skehan, 1998, 2003, 2014; Skehan & Foster, 1999, 2001).
However, in contrast to these previous studies, we have included variations in the operationalizing of the dependent
variables by adopting a more fine-grained approach to the measurement of syntactic complexity, with dependent clauses
being analysed separately (Wolfe-Quintero, Inagaki, & Kim, 1998) as well as in one inclusive group. We have also addressed
challenges related to the operationalizing of the independent variables. Regarding cognitive task complexity, Norris (2010)
and Révész (2014) suggest that more evidence is needed to investigate whether manipulating task complexity actually
affects the types of cognitive demands predicted by the Cognition Hypothesis (Robinson, 2001a, 2001b, 2005, 2007, 2011).
We considered an additional concern. Specifically, if modifying task complexity has an effect on cognitive burden, there is
currently no means by which these modifications can be accurately measured, and it is not clear how this might impact the
findings. As a result, we introduced a patently low complexity task to highlight the effects of task complexity, which can
otherwise be obscured by the inability to accurately gauge variations in task complexity between more complex tasks.
This paper starts with a brief review of the theoretical framework for cognitive task complexity. Subsequently, we discuss
the different predictions made by the Trade-off Hypothesis (Skehan, 1998, 2003, 2014; Skehan & Foster, 1999, 2001) and the
Cognition Hypothesis (Robinson, 2001a, 2001b, 2005, 2007, 2011) for increases in cognitive task complexity on output. In the
subsequent section, writing complexity is discussed, focusing on the lexical and syntactic measures utilized in the present
study. We also include rationales for using T-units and subordination. Finally, past studies on cognitive task complexity and
writing that utilized similar independent and dependent variables to this paper are reviewed.
We then report on the results of a within-subject experimental study in which second language learners (classified as
intermediate level with IELTS scores ranging from 4.5 to 5.5) were provided with writing tasks that had cognitive task
complexity manipulated in the task design. Tasks ranged from patently low complexity to two higher complexity tasks. The
results were analysed for evidence of changes in the linguistic features produced in the texts. Specifically, syntactic
complexity was evaluated as variations in T-unit length. This was operationalized by measuring the ratio of dependent
clauses to T-units across all dependent clauses, and the ratio of dependent clauses to T-units with each dependent clause
measured separately. Lexical complexity was measured by using a mean segmental type-token ratio. This is a measure of
lexical variety using the ratio of different lexical items to the total number of items used while accounting for text length.

Cognitive task complexity

In the study of second language writing, complexity can be applied to two terms: Cognitive task complexity refers to the
modifications made during task design that make a task difficult to complete, whereas writing complexity refers to the
language in written output that can be considered varied and elaborate (Ellis, 2003). This study utilized both aspects of
complexity by investigating the effects of cognitive task complexity on the texts produced by English L2 writers.
Broadly, cognitive task complexity can comprise the interaction of two elements manipulated in the design of
pedagogical tasks (Ellis, 2003). These elements are types of information and amounts of information (Brown, Anderson,
Shilcock, & Yule, 1984). These features are theorized to affect the cognitive burden a student experiences during task
performance by placing varying demands on learners’ cognitive resources. The effects of modifying these elements are
frequently analysed using two frameworks: These are the Trade-off Hypothesis (Skehan, 1998, 2003, 2014; Skehan & Foster,
1999, 2001) and the Cognition Hypothesis (Robinson, 2001a, 2001b, 2005, 2007, 2011).
 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57 47

The Trade-off Hypothesis (Skehan, 1998, 2003, 2014; Skehan & Foster, 1999, 2001) assumes that cognitively demanding
tasks require trade-offs due to limited attentional resources. In terms of complex and accurate performance, both are viewed
as separate and competing dimensions that compete for limited attention, with one dimension likely receiving less attention
than the other. Though trade-offs in performance will not always be apparent due to the influence of other factors (Skehan &
Foster, 2012), the need for trade-offs in performance can be viewed as the default position. When resource limits are reached,
learners will focus on meaning, which Skehan (1996) claims they are naturally and unavoidably predisposed towards. The
effects of trade-offs on writing complexity may be the production of less complex language as a simpler way of expressing
meaning when attention limits are overtaxed.
Alternatively, Robinson (2001a, 2001b, 2005, 2007, 2011) takes the position that attentional resources are more flexible
than the Trade-off Hypothesis suggests, and that increases in cognitive task complexity (under the correct conditions) could
lead to dual increases in accurate and complex language production (Robinson, 2001b). In this study, we do not make any
claims about the potential for cognitive task complexity to affect both complexity and accuracy. We focus only on the
potential effects of increases in cognitive task complexity on complex output as proposed in Robinson’s Cognition
Hypothesis (2001a, 2001b, 2005, 2007, 2011).
Robinson claims there are developmental parallels between adults learning a second language and children’s
development of their first language (Slobin, 1993), and the developmental process might be emulated in adults by increasing
cognitive task complexity. For example, pedagogic tasks could be sequenced using increases in cognitive task complexity,
and this sequencing might be similar to the way children meet increasingly complex demands when learning a first language
(Robinson, 2001a, 2005). Thus, increases in conceptual development, triggered by increases in cognitive task complexity,
should lead L2 learners to produce the language that expresses those concepts.
To operationalize the Cognition Hypothesis, Robinson developed the Triadic Componential Framework (Robinson, 2001a,
2001b, 2005, 2007, 2011) that details the means by which complexity might be manipulated between tasks and across
syllabi. Within the Triadic Componential Framework, a distinction is made between resource-directing and resource-
dispersing elements. Increasing the complexity of resource-directing elements is expected to increase cognitive/conceptual
demands on a learner. Robinson also claims that increases in resource-directing elements may guide learner attention in a
way that facilitates noticing. Noticing is the conscious attention to language forms, and it is considered a prerequisite of
learning (Schmidt, 2001). However, resource-dispersing elements have effects closer to those predicted by the Trade-off
Hypothesis (Skehan, 1998, 2003, 2014; Skehan & Foster, 1999, 2001) when complexity is increased on limited attentional
resources. When increased in complexity, resource-dispersing elements do not direct learners’ attention to the language
required to meet the demands of a complex tasks. Instead, attention is dispersed, making the completion of a task more
difficult. For example, less planning time is considered a resource-dispersing element.
The cognitive task complexity elements utilized in the current study (reasoning demands and number of elements) are
described as resource-directing (Robinson, 2001a, 2001b, 2005, 2007, 2011). Given the appropriate conditions, Robinson
believes that increases in these dimensions may facilitate the production of complex language output as learners strive to
find language to express the increases in conceptual complexity.
Finally, questions have been raised about the operationalization of cognitive task complexity. Révész (2014) and Norris
(2010) stated the need for more evidence about the causal relationship between cognitive task complexity and the types of
cognitive burden they are theorized to trigger in learners. A related issue that should be investigated is the current inability
to accurately manipulate degrees of cognitive task complexity between tasks and how this might affect cognitive burden and
performance. If the levels of applied cognitive task complexity are too high, differing degrees of cognitive burden might not
trigger different outputs but may instead trigger attentional overload. This might lead to trade-offs in production similar to
predictions made by the Trade-off Hypothesis (Skehan, 1998, 2003, 2014; Skehan & Foster, 1999, 2001). However, below a
certain level of task complexity, results may favour the Cognition Hypothesis (2001a, 2001b, 2005, 2007, 2011) because
attention is not overloaded. The strategy we have used to address this methodological issue is to add a task that is patently
lower in cognitive task complexity. As a result, a clearer illustration of whether attentional overload is affecting the results
between the more complex tasks might be revealed.

Writing complexity

In this paper, writing complexity refers to complex written language whose production is deemed to be difficult,
imposing a certain burden on the writers. This is especially the case when they are pushed to the edge of their current ability.
The importance of investigating complex language is partially based on the following presumptions: Complex language
production may represent pushed output in which learners create more elaboration and structure in their developing
language ability, use language more efficiently, align the language being learnt with target language use, express complex
ideas efficiently and precisely, and potentially signal acquisition (Cheng, 1985; Cook, 1994; McLaughlin, 1990; Skehan, 1996;
Swain, 1985).
The two types of complex language investigated in the present study are subordination (using a ratio of dependent
clauses to T-units as a means to determine the mean length of T-units) and lexical variety (mean segmental type-token ratio).
There were a number of reasons for choosing these specific measures. First, Ellis and Barkhuizen (2005) stressed that
language learners may use different means to realize complexity in their writing. Thus, both lexical and syntactic complexity
48 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57

were used in this study to identify whether the participants’ responded to variations in cognitive task complexity by writing
different types of complexity.
That this paper is a partial replication study also influenced the choice of measures. We utilized measures similar to those
used by Kuiken and Vedder (2007, 2008), but they were expanded upon with a more fine-grained approach to investigating
the ratio of dependent clauses.
Beyond the requirements of using the same target measures for a partial replication study, some justification should also
be provided for the measures. T-units provide a readily accessible, quantifiable unit for measuring segments of written
language (Ellis & Barkhuizen, 2005). Additionally, T-units are considered one type of indicator of a learners’ development in
writing ability because the average length of T-unit increases with a writer’s maturity (Cooper, 1976; Hudson, 2009; Hunt,
1965). However, longer T-units can also be associated with flawed writing (Crowhurst, 1983; Hake & Williams, 1979). The
present study does not take a default position that longer T-units are inherently better than shorter T-units. Rather we
recognize that short T-units may, in the hands of a good writer, be used to express meaning in a concise way that in some
cases is more appropriate than long T-units. Nonetheless, T-unit length remains a good indicator of writing complexity and
therefore was used in this study.
Subordination is considered a good indicator of complex output (Norris & Ortega, 2009) for intermediate learners, making
it suitable for this study’s participants. However, measurements of T-units with subordination usually cluster subordinate
clauses into one group, potentially obscuring the detailed effects of task complexity. To elaborate, subordination is viewed as
inherently difficult (Mendhalson, 1983) with associated cognitive burdens during production (Cheung & Kemper, 1992;
Lord, 2002). Therefore, when subordination is produced as a result of attention-taxing task complexity, it may be difficult for
intermediate learners, who could be additionally susceptible to the degree to which each dependent clause is automatized.
As a result, analysing dependent clauses separately may provide a clearer picture of the effects of cognitive task complexity
on subordination because when each clause has a different position on the developmental continuum, it may be more or less
attention demanding to access. By analysing dependent clauses as one group and separately, any potential differences
between the two approaches should be highlighted.
The second target measure was a mean segmental type-token ratio, which is a measure of lexical variety. Wolfe-Quintero
et al. (1998) state that measures of lexical variation appear to be related to the development of second language, particularly
those measures that also account for text length. The type-token ratio measure of lexical variety used in this study is a
measure of lexical variety that accounts for variations in text length.

Writing research utilizing reasoning demands and number of elements

Of the relatively small group of studies focusing on cognitive task complexity and writing, the work of Kuiken and Vedder
(2007, 2008, 2012) and Sercu et al. (2006) have utilized similar independent and dependent variables to this study.
Sercu et al. (2006) used letter-writing tasks in which the participants wrote to a friend about their choice of holiday
destination. Cognitive task complexity was manipulated by increasing the number of elements and reasoning demands
between tasks. Syntactic complexity was investigated by analysing the number of clauses per T-unit and number of
dependent clauses per clause. Lexical variation was analysed using a type-token ratio and a type-token ratio correcting for
text length (the number of word types per square root of two times the total number of word tokens).
Kuiken and Vedder (2007, 2008, 2012) utilized similar letter-writing tasks and variations in number of elements and
reasoning demands. Kuiken and Vedder (2007) analysed syntactic complexity, including number of clauses per T-unit and
the number of dependent clauses per T-unit. Lexical variation (the occurrence of frequently and in frequently used words)
was investigated with a type-token ratio that corrected for text length. Kuiken and Vedder (2008) analysed syntactic
complexity as the number of clauses per T-unit, and a dependent clause ratio that examined the degree of syntactic
embedding. Lexical variation was analysed using two type-token ratios, one that corrected for text length (Wolfe-Quintero
et al., 1998), and one that did not. Finally, Kuiken and Vedder (2012) used a combination of results from past work that
included the results from Kuiken and Vedder (2007, 2008).
In terms of lexical variation, these studies showed that increases in cognitive task complexity resulted in increases in
lexical variation (Kuiken & Vedder, 2007) and type-token ratio (Sercu et al., 2006). However, the type-token ratio used by
Sercu et al. (2006) did not account for text length, and the findings for increases in lexical variation (Kuiken & Vedder, 2007)
were inconsistent between groups (Kuiken & Vedder, 2012). Kuiken and Vedder (2008) found no significant increases in
lexical variation as a result of increases in cognitive task complexity.
When analysing syntactic complexity, Kuiken and Vedder (2007, 2008, 2012) and Sercu et al. (2006) examined the
number of clauses per T-unit and the number of dependent clauses per clause (with dependent clauses analysed as one
group). None of the findings confirmed any effect for cognitive task complexity on syntactic complexity.
No strong support for predictions by the Cognition Hypothesis (Robinson, 2001a, 2001b, 2005, 2007, 2011) about
complex language production can be inferred from these results. The effects of cognitive task complexity on lexical
complexity were only noticeable in type-token ratio measures that did not account for text length, and the findings for
lexical variation were inconsistent between groups. Findings also provided no clear support for the Cognition Hypothesis’
predictions for syntactic complexity.
Though statistically non-significant, we felt the past findings for syntactic complexity bore further consideration, as the
absence of any significant effects may have been attributable to issues not addressed in the research. First, the inability to
 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57 49

accurately and precisely vary the amounts of cognitive duress applied to learners when independent variables are being
manipulated makes it difficult to know what effects are being generated between complex tasks. In such cases, the addition
of a patently low complexity task may provide clearer evidence whether any effect exists between lower and higher degrees
of cognitive task complexity. Second, the blunt nature of the ratio of dependent clauses to T-units measured across all
dependent clauses may have influenced past findings. Measures of subordination, as utilized in studies by Kuiken and
Vedder (2007, 2008, 2012) and Sercu et al. (2006), are more complex than they appear (Bulté & Housen, 2012). We believe
that for intermediate learners, for whom each dependent clause may be automatized to a different extent, the effects of
cognitive task complexity may be more clearly observed on each individual dependent clause.

The present study

This paper reports the findings of a within-subject experimental study designed to assess the effects of cognitive task
complexity on lexical and syntactic complexity, with dependent clauses measured both as one group and separately.
Cognitive task complexity was manipulated by varying the amount of reasoning demands and numbers of elements in
the task instructions of three letter-writing tasks. One task was designed to be patently low complexity with the
expectation that it might illustrate any effects between the independent and dependent variables otherwise obscured
by the inability to accurately gauge variations in cognitive task complexity between more complex tasks. The cognitive
task complexity variables used in the present study are considered resource-directing variables (Robinson, 2001a,
2001b, 2005, 2007, 2011) with the potential for positive effects (Robinson, 2001b) on complex production. In this study,
the positive effects are viewed as increases in complex language production consistent with the participants’
developmental level.
In light of the issues mentioned above, the following questions were formulated:

1. What are the effects of cognitive task complexity on syntactic complexity when measured (a) as the ratio of dependent
clauses to T-units across all dependent clauses, and (b) as the ratio of dependent clauses to T-units when dependent
clauses are measured separately?
2. What are the effects of increases of cognitive task complexity on lexical variety?
3. Does using patently low complexity tasks contribute to the tracking of cognitive task complexity on complex writing?

Method

Participants and context

Thirty-four non-native speakers of English studying in New Zealand volunteered to participate in this study. The various
nationalities of the group were Japanese (2), Chinese (12), Korean (11), Thai (3), Russian (2), Vietnamese (2), and French (2).
There were 22 female and 12 male participants, and though the ages ranged from 18 to 41, most participants (26 out of 34)
were in their twenties. They were all students for whom English was a second language, and they were all identified as
intermediate level language learners by the school where they studied. Their IELTS levels ranged between 4.5 and 5.5. The
learning context was an Auckland language school, which ran classes focused on raising students overall proficiency in
reading, writing, speaking, and listening.

Target measures

The target measures used in this study were a mean segmental type-token ratio, which is a measure of lexical variety; and
the ratio of dependent clauses to T-units, a measure of syntactic complexity. The ratio of dependent clauses to T-units was
analysed across all dependent clauses and also with each dependent clause analysed separately (Wolfe-Quintero et al.,
1998).

Instruments

Three instruments, Task 1 (low complexity), Task 2 (medium complexity), and Task 3 (high complexity), were used for
data collection. These tasks were devised to elicit language samples reflecting the complex language competency of the
participants when placed under increasing cognitive duress.
Each instrument was designed to initiate different levels of attention demanding activity through the manipulation of
cognitive task complexity in the task instructions. Task 1 (low complexity) was intended to require the least amount of
cognitive duress to affect task completion, whereas Task 2 (medium complexity) and Task 3 (high complexity) were both
designed to necessitate higher levels of cognitive duress to complete.
Task 1 (see Appendix A) was the low complexity task. It was patently low complexity, meaning it was designed to ensure
that the levels of cognitive burden being applied were low enough (in comparison to Tasks 2 and 3) that the inability to
accurately measure cognitive task complexity did not skew the results when compared against Task 2 and Task 3.
50 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57

Task 1 consisted of a written handout that provided a situation and a set of instructions. The situation involved an English-
speaking friend who was coming to New Zealand. The instructions directed the participant to write to this friend and tell him
about New Zealand. The task instructions were devised to be easy to understand while simultaneously avoiding phrases that
instructed the writer to form any opinions or state any reasons why New Zealand might be worthwhile moving to.
Furthermore, Task 1 did not have additional information for the participants to use in the writing of the letter, unlike Tasks
2 and 3. Task 1 participants were expected to rely on their own resources.
Task 2 (see Appendix B) was the medium complexity task. It was made more complex than Task 1 by manipulating the
reasoning demands and number of elements in the task instructions. This was expected to stimulate higher amounts of
attention demanding activity (Brown et al., 1984; Ellis, 2003; Robinson, 2005, 2007) in order to complete the task. Reasoning
demands were incorporated into Task 2 by designing instructions that directed the participants to write to a friend who was
coming to New Zealand and inform that friend which of the two restaurants they would visit upon arrival. Information
pertaining to the restaurants was provided with the task instructions. Furthermore, the choice of restaurant needed to be
justified based on the information about each restaurant as well as the visiting friends’ information. The second cognitive
task complexity dimension was number of elements. This dimension was added by supplying more choice of elements
(needed for the completion of the task) in the task instructions than were used in Task 1 (low complexity).
Task 3 was the high complexity task. It was similar to Task 2 (medium complexity); however, there were greater amounts
of cognitive task complexity. The number of elements that needed to be considered was increased, thus augmenting the
reasoning demands. There were three restaurants to choose from rather than two. Additionally, the participants were also
expected to include the information of two more friends who would be visiting the restaurant as well as the preferences of
the person receiving the letter.

Procedure

The data were collected in two stages. In stage one, participants were approached at their schools and invited to
participate. Those who were interested attended a meeting, which also served as stage 1 of the data collecting process.
Initially, the participants completed a consent form then the date for stage 2 of the data collecting process was negotiated. In
all cases, stage 2 was performed between three to five days after stage 1. After the date for stage 2 was set, the participants
were given Task 1 (low complexity). The participants had 2 min to view the task and ask questions about any aspects of the
instructions. 30 min was provided to complete the task. No dictionaries or smart phones were allowed.
Stage 2 involved the completion of two consecutive tasks of differing levels of cognitive task complexity (Task 2, medium
complexity; and Task 3, high complexity). First, the schedule for stage two was explained, and then the participants were
given the first task as well as a 2-min comprehension check. Clarifications of the instructions were provided on request.
30 min was provided to complete the task. Upon completion, there was a 5-min break (in class) where the participants could
relax but were discouraged from discussing the tasks. Second, after the break, the second task was distributed. As with the
first task, there was a comprehension check followed by a 30-min time limit for completion.
Task 2 and Task 3 were counterbalanced to account for participant fatigue, with random selection of task order made by
picking a number (2 or 3) from a container. Task 1 (low complexity) was performed at a time prior to Tasks 2 and 3 and was
not part of the counterbalancing process. Although this exclusion may have had some effect on the results, Task 1 was
excluded from counterbalancing for a number of reasons. First, time constraints imposed by student availability required
that Task 1 be performed at a separate time. Second, the data collected in this study was one of three groups participating in a
larger study; and as such, it was believed that Task 1 would be an additional indicator of participant skill level (in conjunction
with the IELTS scores) between the group used in the present study, and the two other groups that were not included. Task
1 was designed to induce participants to use their own resources. Thus, they were expected to produce language at a level
they were comfortable with as opposed to being pushed to synthesize the extra information supplied in the more complex
tasks. This was expected to provide an indication of skill level for all participants separate from task complexity or the
additional variables used in the two groups not included in this study.
Table 1 illustrates the way in which Task 2 and Task 3 were counterbalanced during the second stage of the data collecting
process.

Analyses

The different analyses of ratio of dependent clauses to T-units required that all dependent clause types within the T-units
be recorded separately during coding. This allowed for the analysis of dependent clauses as one group and also as separate
items (see Table 2). To determine the ratio of dependent clauses to T-units across all dependent clauses in any one text, the

Table 1
Counterbalancing Task 2 and Task 3 during stage 2.

Participants 1st task assigned to participants 2nd task assigned to participants

34 participants 17 participants Task 2 (medium complexity) Task 3 (high complexity)

17 participants Task3 (high complexity) Task 2 (medium complexity)
 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57 51

Table 2
Two ways to analyse the ratio of dependent clauses to T-units.
Data sample:
I have been here for three weeks. I think this city is the best one in New Zealand. I was so happy when I heard that you were coming to New Zealand.
1. Dependent clauses analysed as one group. 3 T-units and 3 dependent clauses
2. Dependent clauses analysed separately. Adjective clauses
3 T-units and 0 dependent clauses

Nominal clauses
3 T-units and 2 dependent clauses

Adverbial clauses
3 T-units and 1 dependent clause

Table 3
Performance comparisons between tasks.

Mean and standard variation

Dependent Task 1 Task 2 Task 3

Variables M SD M SD M SD

Syntactic
DC per T-unit 1.41 .20 1.40 .15 1.35 .14
ADJ-DC per T-unit 1.05 .06 1.07 .08 1.07 .08
NOM-DC per T-unit 1.18 .10 1.20 .10 1.17 .08
ADV-DC per T-unit 1.16 .11 1.10 .08 1.09 .08
Lexical
MSTTR 82.00 3.53 82.82 4.00 83.81 3.52

number of dependent clauses and T-units in that text were added then divided by the number of T-units. To ascertain the
ratio of dependent clauses to T-units with dependent clauses separated in the same text, the number of T-units was added to
the number of whichever dependent clause was being analysed.
A mean segmental type-token ratio was used in the analysis of lexical variation because it is a measure that accounts for
text length by dividing each text into segments of equal word length. To avoid, as much as possible, wasting data, the texts
from the letter-writing tasks were divided into segments of 40 words. This number was based on the size of the work
provided by the students.
Syntactic complexity was analysed using a repeated measures ANOVA with a confidence level of .05. This provided a score
representing the mean length of T-units for each task (see Table 3). Lexical complexity was also analysed using a repeated
measures ANOVA with the same confidence level. This provided a score representing the mean number of varied lexical
items for each task (see Table 3). For both syntactic and lexical complexity, a Bonferroni corrected post hoc pairwise
comparison with a confidence level of .017 was used as needed. Before proceeding with the analyses, the data were checked
for normality. The histograms revealed slight positive skews in the distribution; however, a Friedman Test confirmed that
this skew did not affect the validity of the ANOVA results.

Results

Table 3 shows the mean score and standard deviation of dependent clauses per T-unit for all three tasks, with dependent
clauses analysed as one group (DC per T-unit). The result was not statistically significant, F(2,32) = 3.02, p = .06. There was no
significant mean scores variation among Task 1, Task 2, and Task 3 when testing for increases in task complexity on the
number of dependent clauses per T-units, across all dependent clause types. Table 3 also illustrates the mean scores and
standard deviation for dependent clauses per T-unit where adjectival (ADJ-DC per T-unit), nominal (NOM-DC per T-unit),
and adverbial (ADV-DC per T-unit) dependent clauses were measured separately. Repeated measures ANOVAs established
no statistical differences in syntactic complexity between the three tasks for adjectival dependent clauses, F(2,32) = .70,
p = .50 or for nominal dependent clauses, F(2,32) = 2.21, p = 0.12. However, a statistically significant result was found for
adverbial dependent clauses, F(2,32) = 4.72, p = .01. The effect size, calculated using partial eta-squared (h2p ), was .228, which
is considered large.
A Bonferroni corrected post hoc pairwise comparison with an alpha level of .017 was performed on the adverbial
dependent clauses. This revealed statistically significant mean scores variation (p = .02) between Task 1 (low complexity)
and Task 2 (medium complexity) and a significant mean scores variation (p = .02) between Task 1 (low complexity) and Task
3 (high complexity). This shows that Task 1 has a significantly higher mean score for adverbial dependent clause per T-unit
than Task 2 and Task 3. Additionally, no significant change in the mean number of adverbial dependent clauses per T-unit,
attributable to variations in task complexity between Task 2 and Task 3, was established.
52 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57

The mean scores and standard deviations of the mean segmental type-token ratios (MSTTR) for all three tasks are also
illustrated in Table 3. A repeated measures ANOVA yielded a statistically significant result, F(2,32) = 3.40, p = .04 among Task
1, Task 2, and Task 3. The effect size, calculated using partial eta-squared (h2p ), was .176, which is considered large. Though
lexical variety incrementally increased between each task, not each increase was statistically significant. A Bonferroni
corrected post hoc pairwise comparison with an alpha level of .017 was performed on the mean segmental type-token ratios.
This revealed that the only statistically significant mean score variation (p = .03) was between Task 1 (low complexity) and
Task 3 (high complexity).

Discussion

This study had four questions related to the central objective of the effects of increases in cognitive task complexity
(reasoning demands and number of elements) on writing complex output (ratio of dependent clauses to T-units and mean
segmental type-token ratio).

1. What are the effects of cognitive task complexity on syntactic complexity when measured (a) as the ratio of dependent clauses
to T-units across all dependent clauses, and (b) as the ratio of dependent clauses to T-units when dependent clauses are
measured separately?

The first objective was comparing the effects of increases in cognitive task complexity on two measures of ratio of
dependent clauses to T-units, and to analyse the findings at individual dependent clause level. The results for ratio of
dependent clauses to T-units across all dependent clauses appeared to show that increases in cognitive task complexity
effected no statistically significant change (p = .06) in the mean length of T-units when dependent clauses were analysed as a
single entity. This result is consistent with previous findings (Kuiken and Vedder, 2007, 2008, 2012; Sercu et al., 2006).
Characteristics inherent in the writing process such as planning time, selective control, and recursion (Ellis & Yuan, 2004;
Grabe & Kaplan, 1996; Grabowski, 2007; Kormos & Trebits, 2012) were predicted to facilitate the production of complex
linguistic structures that require attention. These characteristics were expected to aid processes predicted by the Cognition
Hypothesis (Robinson, 2001a, 2001b, 2005, 2007, 2011) in which increases in cognitive task complexity would create
increases in conceptual development, leading to the production of complex language needed to express those concepts.
Contrary to expectations, increases in cognitive task complexity did not result in an increase in syntactic complexity as
measured.
Nevertheless, the non-significant results of this study may be an indicator that applied levels of cognitive task complexity
were incorrectly aligned with the degree to which the participants’ ability to use subordination was automatized, thus
overtaxing limited attentional resources. Potentially, the findings represent incorrect variable alignment as opposed to a
refutation of the Cognition Hypothesis.
Whereas the ratio of dependent clauses to T-units across all dependent clauses showed no significant results, the ratio of
dependent clauses to T-units analysing each dependent clause separately revealed more detailed variations, which included
one significant result. The results showed statistically significant findings for decreases in adverbial dependent clauses, non-
significant increases and decreases in nominal dependent clauses, and almost no variation in adjectival dependent clauses
across all three tasks. Because these varied results for each individual dependent clause resulted from the application of the
same amount of cognitive task complexity, it appears that other elements have contributed to these variations.
Adverbial dependent clauses exhibited statistically significant variations with Task 1 (low complexity) having a higher
incidence of adverbial clauses than Task 2 (medium complexity) and Task 3 (high complexity). One possible explanation for
these findings is that compared to Task 1, both Task 2 and Task 3 may have been too complex. In other words, beyond a
certain level of cognitive task complexity, no differences were found in the linguistic complexity of the output because
writer’s attention was overtaxed. This overtaxation, in turn, may be a function of learners’ proficiency, in particular, the
extent to which they may have automatized the production of certain language structures. We suggest that assessing such
threshold language proficiency effects on the relationsip between task complexity and output complexity is an important
area for future research.
Other prospective causes for these findings include the pragmatic requirements of the tasks (Bygate, 1999; Ryshina-
Pankova, in press; Ryshina-Pankova & Byrnes, 2013) and personal choice (Pallotti, 2009). In a study that utilized argument
tasks with similar complexity demands (giving reasons and justifications) as the current study, Bygate (1999) found that
different task types led to different frequencies of subordination. Ryshina-Pankova (in press) also makes the argument that
form and function should not be separated in the study of complexity. Ryshina-Pankova and Byrnes (2013) noted that
increases in complexity, operationalized as the occurrence of nominalization, were associated with increased mastery of the
academic register. It is possible, therefore, that the differences in findings between different subordinate clauses in our study
were also a by-product of the task type in that particular dependent clauses were not frequently needed as a means of
fulfilling the pragmatic requirements of the task. With regard to personal preference, Pallotti (2009) suggested that
differences in accurate, complex, and fluent output should not only be attributed to psycholinguistic factors (such as
memory) because sometimes the elements of production may just be a matter of choice. However, we suggest that
psycholinguistic factors like memory are central to making language choices. Furthermore, the notion that the participants
 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57 53

in this study just simultaneously choose to use the same language seems less likely than their language choices being a
response to the task-related factors to which they were all exposed.

2. What are the effects of increases of cognitive task complexity on lexical variety?

The second objective was to investigate the effects of increased cognitive task complexity on lexical variety. Unlike the
decreases that mostly characterized the findings for syntactic complexity, there were increases in lexical complexity,
specifically between Task 1 (low complexity) and Task 3 (high complexity).
The findings seemed to provide some evidence that increases along resource-directing dimensions (Robinson, 2001a,
2001b, 2005, 2007, 2011) could trigger the production of more complex language. However, it is possible that the increases
in lexical complexity were the result of the increasing amount of lexical items supplied with each increasingly complex set of
instructions because writers can produce texts using information taken from task instructions or memory (Kormos, 2011).
Therefore, the findings may not be the result of pushed complex output due to increased complexity. Rather, the retrieval of
lexical item from the instructions may be a confounding variable.
Ellis and Yuan (2004) claim that, under some conditions, when learners formulate what they will write, they may
prioritize the search for lexical items over the generation of grammar, similar to Levelts’ (1989) psycholinguistic speaking
model. This prioritizing may also explain the increases in lexical variety while syntactic complexity remains constant or,
in the case of adverbial dependent clauses, decreases. Participants may have used the limited cognitive resources available
to them to focus on the easier lexical means of meeting the pragmatic requirements of the tasks. If true, then the findings
for lexical complexity might only ostensibly support Robinsons’ Cognition Hypothesis, with trade-offs between syntactic
and lexical means of expression having more in common with the Trade-off Hypothesis (Skehan, 1998, 2003, 2014; Skehan
& Foster, 1999, 2001). The pressure brought to bear on limited attentional resources by cognitive task complexity as well
as the cognitive pressure of producing language that was not fully automatized may have required a trade-off. In this
instance, the trade-off was between more or less demanding types of complex expression.

3. Does using patently low complexity tasks contribute to the tracking of cognitive task complexity on complex writing?

The third objective was to examine whether patently low complexity tasks contributed to the tracking of cognitive task
complexity on complex writing because cognitive task complexity increases or decreases cannot be accurately controlled
and measured between tasks. As a result, findings might be obscured if tasks were either too complex or not complex
enough. The findings in this study indicate that the patently lower complexity task clarified the effects of cognitive task
complexity on syntactic complexity.
In the case of adverbial dependent clause length and mean segmental type-token ratio, the patently low complexity task
(Task 1) appeared to provide clearer indications of the effects of increases in cognitive task complexity. Statistically
significant results were found between Task 1 (low complexity) and one or both of the other more complex tasks.
However, two points need to be made about these results. First, the fact that the patently low complexity task (Task 1)
was, by necessity, performed separately and not included in the counterbalancing, has to be considered as a potential factor
affecting these results. Second, a qualification needs to be made regarding the low complexity task. To create a task of
patently less complexity, Task 1 was designed to encourage the participants to only utilize their own resources (as opposed
to the synthesizing of extra information supplied in the task instruction for Tasks 2 and 3). Additionally, Task 1 instructions
avoided utilizing words or phrases that might elicit any opinions or reasoning demands. Problematically, the potential exists
for the low complexity task to have different pragmatic requirements. This may be a contributing factor affecting
subordinate output in conjunction with variations in cognitive task complexity. In future research, this potential mitigating
factor needs to be addressed.
This study was framed as a partial replication of the work of Kuiken and Vedder (2007, 2008, 2012) and Sercu et al. (2006),
studies that used similar independent and dependent variables. Considering that the significant variations in the current
study were all observed between the patently low complex task (something not used in past research) and the more complex
tasks, this might have highlighted an unaccounted for variable affecting past findings, notably the inability to accurately
measure cognitive task complexity between complex tasks. It is worth noting that the significant results were seemingly
generated by the combination of both the patently lower complexity task and the more detailed measure of subordination,
elements that were both absent in previous studies. This highlights the possibility that the results of previous work might
have been affected by the omission of these variables.
As mentioned earlier, partial replication studies may contribute to fully understanding the relationship between one set
of variables and any extra unaccounted for variables that might be affecting the results. This paper may have illuminated
two such factors, specifically, the added effects of using patently low complexity tasks and a more detailed analysis of
subordination on studies in complexity and writing.

Limitations and future research

As with most studies, there remain a number of limitations with this research. First, the potential difference in pragmatic
task requirements between the patently lower complexity task and the two higher complexity tasks is a possible limitation.
54 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57

This may need to be addressed in future research to provide a stronger case for differences in performances being related to
variations in task complexity.
Second, there may be limitations related to our analyses of structural complexity, in which subordination might not have
accurately captured the means by which intermediate learners express written complexity. Contrary to Norris and Ortega’s
(2009) proposed stages of syntactic complexity development, in which subordination is viewed as an appropriate expression
of complexity for intermediate writers, findings by Bulté and Housen (2014) and Crossley and McNamara (2014) point to
different stages of development. In a study that targeted intermediate and advanced writers, Bulté and Housen (2014) found
that over time, increases in complex writing, operationalized as clausal coordination and phrasal elaboration, were evident;
however, this was not the case with subordination. Future research on writing complexity may need to consider additional
measures in addition to subordination when investigating syntactic complexity.
Third, more investigations are needed in the field of cognitive task complexity to ensure that cognitive burdens, of the
type stipulated by Robinson (2001a, 2001b, 2005, 2007, 2011), are actually being triggered by the independent variables
(Norris, 2010; Révész, 2014). A number of ways that may contribute to a better understanding of cognitive burden have been
suggested. For example, individual differences aside, a combination of subjective self-ratings, secondary task methodology,
subjective time estimations, and some psychophysiological techniques may be a good way to estimate the cognitive load
purported to be created by tasks complexity (Révész, 2014). Another method may be to test cognitively complex tasks, like
those used in this study, on language users who have attained a high level of proficiency. This may provide a clearer
indication of how these tasks affect learners under optimal processing conditions and thus provide a clearer insight as to
what might be expected in comparison to learners situated at different stages of the language learning continuum.
Fourth, the sample size of this research could be another issue of concern. The present study has a sample size of
34 participants, which could be considered relatively small and thus affecting statistical power. If power is low due to the
sample size, only large effects are likely to have been detected. This may have impacted the main effects and the significance
of the pairwise comparisons. Future research on the same variables may benefit from larger samples.
Fifth, it could be argued that the difference between the current findings and the previous work (Kuiken & Vedder, 2007,
2008, 2012; Sercu et al., 2006) might be a matter of the degrees of cognitive task complexity applied between tasks. Had the
cognitive task complexity levels used in this study been further increased between tasks, the non-significant findings may
have resulted in a statistically significant decrease in dependent clause production. Given the lack of a clear means by which
degrees of cognitive task complexity can be accurately and incrementally varied, it is difficult to know what degree of
variations would have elicited a statistically significant difference. Focusing on means by which cognitive task complexity
might be more accurately measured and applied may be a central concern for future research in this area.
Research might also want to focus specifically on the effects of cognitive task complexity on acquisition, or the
introduction of pre-task planning time as an extra variable. It might be worthwhile to ascertain whether the extra processing
time afforded by planning time makes a difference to the type of cognitively stressful performances that were investigated in
the current research. Additionally, assessing the thresholds where task complexity overtaxes attentional resourse on
language that is not fully automatized is also an important area for future research.
Finally, future research should take into consideration the social impacts on cognitive task complexity and production.
Given the predicted effects of motivation on attention levels, elements like motivation or the effects of task motivation might
also be considered.

Concluding remarks

This study investigated the effects of variations in cognitive task complexity (reasoning demands and number of
elements) on writing complexity. For the most part, the results from this research appear to have met the main objectives of
this study, with the use of a patently non-complex task and a more detailed analysis of subordinate clauses appearing to
contributing to significant finding. Robinson’s Cognition Hypothesis, a theory originally framed around oral production, was
tested in the written modality, with the belief that aspects of the writing process might facilitate complex output under
cognitive pressure. Though lexical complexity appears to have increased as a result of increases in cognitive task complexity,
this was not the case for subordination. However, we do not rule out the potential for cognitive task complexity to facilitate
increases in subordination under certain conditions, such as when the application of appropriate measures of cognitive task
complexity align with the appropriate degree to which a language structure is automatized.
 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57 55

Appendix A. Task 1 (low complexity)

Letter one. Name

This activity is about writing a letter to a friend. Read the information below then
write a letter based on the situation and following the instructions.

Situation:
1. You have a close English-speaking friend called Peter.
2. Peter is thinking about coming to New Zealand.

Instructions:
1. You have to write a letter to your friend of about 200 to 250 words.
2. In this letter, you should tell Peter about New Zealand.
3. Start the letter below.

Dear Peter,

Appendix B. Task 2 (medium complexity)

Letter two. Name

This activity is about writing a letter to a friend. Read the information below then
write a letter based on the situation and following the instructions.

Situation:
1. Your friend John is coming to New Zealand for one weekend, and there are two
restaurants he really wants to try.
2. There is only time to go to one restaurant. As a result, John wants you to choose
one restaurant.
3. Neither of the restaurants you have checked are perfect for John and your
requirements.

Instructions:
1. Look at John’s requirements in list A.
2. Look at the restaurant information in list B.
3. Consider your own personal preferences.
4. Using the information from lists A and B and your own preferences, write John a
letter of between 200 and 250 words telling him which restaurant you have chosen
and why you choose it.
5. Start the letter below.

Dear John,

Task 2: Supplementary information lists A and B

List A. Information regarding you and John.

John’s information:
1. He is arriving on Saturday morning and leaving on the following Monday
afternoon.
2. Seafood and Pork are his favourite food.
3. He generally eats a lot.
4. He doesn’t particularly love sweet food, but enjoys dessert some times.
5. He likes to drink a glass of wine with dinner.
6. He only speaks English.
7. He will be staying with you during his time here, so transportation will be your
responsibility.

Your information:
1. When you are considering the restaurant, consider your actual personal preferences.
56 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57

List B: Restaurant information.

Restaurant 1:
Opening times: 4:00 pm to 8:00 pm Monday to Saturday.
Prices: Main courses (main meal) cost around $20.
Availability: Usually the restaurant is very busy and bookings (reserve a table) are
necessary to get a table.
Critic’s review of food quality:
The seafood selection (what is available) is good, and is considered very high quality.
The beef is average quality.
The pork is average quality.
There are no desserts (ice cream etc.) at this restaurant.
The portions (size of meal) are average size.
Drink: The beer and wine is expensive. There is no BYO (bring your own drinks).
Staff: Some staff speak English, some staff only speak Japanese.
Service: The service is quick, but the staff do not appear friendly.
Entertainment: Karaoke after 7:00pm.
Location: In Auckland’s’ central city.
Parking: Restaurant supplies no parking.

Restaurant 2:
Opening times: 4:00pm to 9:00 pm Tuesday to Sunday.
Prices: Main courses cost around $28.
Availability: Quiet during the week, sometimes busy on the weekend. No booking is
necessary.
Critic’s review of food quality:
The seafood selection is good and the quality is good.
The pork is good quality.
The beef is high quality.
The desserts are very high quality.
The chicken is average quality.
Portions are larger than average size.
Drink: Beer and wine are very expensive, however, you can BYO.
Staff: Some staff members speak English, and some staff members speak Chinese.
Service: The service is efficient, and the members of staff are helpful.
Entertainment: There is no entertainment supplied by the restaurant.
Location: In Auckland’s central city.
Parking: Restaurant supplies a small amount of parking for customers, though much
less than the restaurant requires

References

Belcher, D., & Hirvela, A. (2001). Introduction. In D. Belcher & A. Hirvela (Eds.), Linking literacies: Perspectives on L2 reading–writing connections. (pp. 1–14).
Ann Arbor: University of Michigan Press.
Brown, G., Anderson, A., Shilcock, R., & Yule, G. (1984). Teaching talk: Strategies for production and assessment. Cambridge: Cambridge University Press.
Bulté, B., & Housen, A. (2012). Defining and operationalising L2 complexity. In A. Housen, F. Kuiken, & I. Vedder (Eds.), Dimensions of L2 performance and
proficiency: Complexity, accuracy and fluency in SLA (pp. 21–46). Amsterdam: John Benjamins.
Bulté, B., & Housen, A. (2014). Conceptualizing and measuring short-term changes in L2 writing complexity. Journal of Second Language Writing, 26(42–65).
Bygate, M. (1999). Task as context for the framing, reframing and unframing of language. System, 27, 33–48.
Carless, D. (2012). TBLT in EFL settings: Looking back and moving forward. In A. Shehadeh & C. A. Coombe (Eds.), Task-based language teaching in foreign
language contexts: Research and implementation (pp. 345–358). Amsterdam: John Benjamins.
Cheng, P. W. (1985). Restructuring versus automaticity: Alternative accounts of skill acquisition. Psychological Review, 92, 414–423.
Cheung, H., & Kemper, S. (1992). Competing complexity metrics and adults’ production of complex sentences. Applied Psycholinguistics, 13, 53–76.
Cook, V. J. (1994). Linguistics and second language acquisition. London: Macmillan.
Cooper, T. C. (1976). Measuring written syntactic patterns of second language learners of German. Journal of Educational Research, 69, 176–183.
Crossley, S. A., & McNamara, D. S. (2014). Does writing development equal writing quality? A computational investigation of syntactic complexity in L2
learners. Journal of Second Language Writing, 26, 66–79.
Crowhurst, M. (1983). Syntactic complexity and writing quality: A review. Canadian Journal of Education, 8, 1–16.
Ellis, R. (2003). Task-based language learning and teaching. Oxford: Oxford University Press.
Ellis, R., & Barkhuizen, G. (2005). Analyzing learner language. Oxford: Oxford University Press.
Ellis, R., & Yuan, F. (2004). The effects of planning on fluency, complexity, and accuracy in second language narrative writing. Studied in Second Language
Acquisition, 26, 59–84.
Flower, L. S., & Hayes, J. R. (1981). A cognitive process theory of writing. College Composition and Communication, 32, 365–387.
Grabe, W., & Kaplan, R. B. (1996). Theory and practice of writing: An applied linguistic perspective. London: Longman.
Grabowski, J. (2007). The writing superiority effect in the verbal recall of knowledge: Source and determinants. In M. Torrace, L. van Waes, & D. Galbraith
(Eds.), Writing and cognition: Research and applications (pp. 165–179). Bingley, UK: Emerald Group.
Hake, R., & Williams, J. M. (1979). Sentence expanding: Not can, or how, but when. In D. Daiker, A. Kerek, & M. Morenberg (Eds.), Sentence combining and
the teaching of writing. Akron, OH: L & S Books.
Hudson, R. (2009). Measuring maturity. In R. Beard, D. Myhill, J. Riley, & M. Nystrand (Eds.), The SAGE handbook of writing development (pp. 349–362).
London: Sage Publications.
Hunt, K. W. (1965). Grammatical structures written at three grade levels. Urbana, IL: The National Council of Teachers of English.
 M.W. Frear, J. Bitchener / Journal of Second Language Writing 30 (2015) 45–57 57

Ishikawa, T. (2007). The effect of manipulating task complexity along the [here-and-now] dimension on L2 written narrative discourse. In M. del Pilar
Garcı́a Mayo (Ed.), Investigating tasks in formal language learning (pp. 136–156). Clevedon: Multilingual Matters.
Kormos, J. (2011). Task complexity and linguistic and discourse features of narrative writing performance. Journal of Second Language Writing, 20, 148–161.
Kormos, J., & Trebits, A. (2012). The role of task complexity, modality, and aptitude in narrative task performance. Language Learning, 62, 439–472.
Kuiken, F., & Vedder, I. (2007). Task complexity and measures of linguistic performance in L2 writing. International Review of Applied Linguistics in Language
Teaching, 45, 261–284.
Kuiken, F., & Vedder, I. (2008). Cognitive task complexity and written output in Italian and French as a foreign language. Journal of Second Language Writing,
17, 48–60.
Kuiken, F., & Vedder, I. (2012). Syntactic complexity, lexical variation and accuracy as a function of task complexity and proficiency level in L2 writing and
speaking. In A. Housen, F. Kuiken, & I. Vedder (Eds.), Dimensions of L2 performance and proficiency: Complexity, accuracy and fluency in SLA (pp. 143–170).
Amsterdam: John Benjamins.
Levelt, W. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT Press.
Lord, C. (2002). Are subordinate clauses more difficult? In J. L. Beebe & M. Noonan (Eds.), Complex sentences in grammar and discourse: Essays in honour of
Sandra A Thompson (pp. 223–233). Amsterdam: John Benjamins.
McLaughlin, B. (1990). Restructuring. Applied Linguistics, 11, 113–128.
Mendhalson, D. J. (1983). The case for considering syntactic maturity in ESL and EFL. International Review of Applied Linguistics in Language Teaching, 21,
299–311.
Norris, J. M. (2010). Understanding instructed SLA: Constructs, contexts, and consequences. Plenary address delivered at the annual conference of the European
Second Language Association (EUROSLA).
Norris, J. M., & Ortega, L. (2009). Towards an organic approach to investigating CAF in instructed SLA: The case of complexity. Applied Linguistics, 30,
555–578.
Ong, J. (2014). How do planning time and task conditions affect metacognitive processes of L2 writers? Journal of Second Language Writing, 23, 17–30.
Ong, J., & Zhang, L. J. (2010). Effects of task complexity on the fluency and lexical complexity in EFL students’ argumentative writing. Journal of Second
Language Writing, 19, 218–233.
Ong, J., & Zhang, L. J. (2013). Effects of the manipulation of cognitive processes on EFL writers’ text quality. TESOL Quarterly, 47, 375–398.
Pallotti, G. (2009). CAF: Defining, refining and differentiating constructs. Applied Linguistics, 30, 590–601.
Porte, R., & Richards, K. (2012). Focus article: Replication in second language writing research. Journal of Second Language Research, 21, 284–293.
Révész, A. (2014). Towards a fuller assessment of cognitive models of task-based learning: Investigating task-generated cognitive demands and processes.
Applied Linguistics, 35, 87–92.
Robinson, P. (2001a). Task complexity, cognitive resources, and syllabus design: A triadic framework for examining task influences on SLA. In P. Robinson
(Ed.), Cognition and second language instruction (pp. 287–318). Cambridge: Cambridge University Press.
Robinson, P. (2001b). Task complexity, task difficulty, and task production: Exploring interactions in a componential framework. Applied Linguistics, 22,
27–57.
Robinson, P. (2005). Cognitive complexity and task sequencing: Studies in A componential framework for second language task design. International Review
of Applied Linguistics in Language Teaching, 43, 1–32.
Robinson, P. (2007). Criteria for classifying and sequencing pedagogic tasks. In M. del Pilar Garcia Mayo (Ed.), Investigating tasks in formal language learning
(pp. 7–26). Clevedon: Multilingual Matters.
Robinson, P. (2011). Second language task complexity, the Cognition Hypothesis, language learning, and performance. In P. Robinson (Ed.), Second language
task complexity: Researching the Cognition Hypothesis of language learning and performance (pp. 3–37). Amsterdam: John Benjamins.
Ryshina-Pankova, M. (2015). A meaning-based approach to complexity in L2 writing: The case of grammatical metaphor. Journal of Second Language
Writing, 29, 51–63.
Ryshina-Pankova, M., & Byrnes, H. (2013). Writing as learning to know: Tracing knowledge construction in L2 German compositions. Journal of Second
Language Writing, 22, 179–197.
Schmidt, R. (2001). Attention. In P. Robinson (Ed.), Cognition and second language instruction (pp. 3–32). Cambridge: Cambridge University Press.
Sercu, L., De Wachter, L., Peters, E., Kuiken, F., & Vedder, I. (2006). The effect of task complexity and task conditions on foreign language development and
performance: Three empirical studies. International Journal of Applied Linguistics, 152, 55–84.
Skehan, P. (1996). A framework for the implementation of task based instruction. Applied Linguistics, 17, 38–62.
Skehan, P. (1998). A cognitive approach to language learning. Oxford: Oxford University Press.
Skehan, P. (2003). Task-based instruction. Language Teaching, 36, 1–14.
Skehan, P. (2014). Limited attentional capacity, second language performance, and task-based pedagogy. In P. Skehan (Ed.), Processing perspectives on task
performance (task-based language teaching) (pp. 211–260). Amsterdam: John Benjamins.
Skehan, P., & Foster, P. (1999). The influence of task structure and processing conditions on narrative retellings. Language Learning, 49, 93–120.
Skehan, P., & Foster, P. (2001). Cognition and tasks. In P. Robinson (Ed.), Cognition and second language instruction (pp. 183–205). Cambridge: Cambridge
University.
Skehan, P., & Foster, P. (2012). Complexity, accuracy, fluency and lexis in task-based performance: A synthesis of the Ealing research. In A. Housen, F.
Kuiken, & I. Vedder (Eds.), Dimensions of L2 performance and proficiency: Complexity, accuracy and fluency in SLA (pp. 120–199). Amsterdam: John
Benjamins.
Slobin, D. (1993). Adult language acquisition: A view from child language study. In C. Perdue (Ed.), Adult language acquisition: Cross-linguistic perspectives:
Vol. 2. The results (pp. 239–252). Cambridge: Cambridge University Press.
Swain, M. (1985). Communicative competence: Some roles of comprehensible input and comprehensible output in its development. In S. Gass & C. Madden
(Eds.), Input in second language acquisition (pp. 235–253). New York: Newbury House.
Wolfe-Quintero, K., Inagaki, S., & Kim, H. Y. (1998). Second language development in writing: Measures of fluency, accuracy, and complexity. Honolulu:
University of Hawaii Press.

Doctor Mark Frear is currently teaching academic writing at Qatar University in Qatar. He is currently researching the various effects of cognitive task
complexity, pre-task planning time, and motivation on the written complexity of second language writers.

Professor John Bitchener is currently supervising 11 doctoral students in areas of L2 teaching and learning at AUT University, Auckland, NZ, and has published
more than 50 articles in top Applied Linguistics journals. He is author of one book on thesis writing and another on Written Correction for Second Language
Acquisition and Writing (with Ferris). Two new books (one on written CF for L2 development, with Storch, and Addressing the writing issues of L2 writers) are due
2015.