Language Learning ISSN 0023-8333

Native Speakers and Task Performance:

Comparing Effects on Complexity, Fluency,
and Lexical Diversity
Pauline Foster
St. Mary’s University College

Parvaneh Tavakoli
London Metropolitan University

This article argues that a native-speaker baseline is a neglected dimension of studies into
second language (L2) performance. If we investigate how learners perform language
tasks, we should distinguish what performance features are due to their processing an L2
and which are due to their performing a particular task. Having defined what we mean
by “native speaker,” we present the background to a research study into task features
on nonnative task performance, designed to include native-speaker data as a baseline
for interpreting nonnative-speaker performance. The nonnative results, published in this
journal (Tavakoli & Foster, 2008) are recapitulated and then the native-speaker results
are presented and discussed in the light of them. The study is guided by the assumption
that limited attentional resources impact on L2 performance and explores how narrative
design features—namely complexity of storyline and tightness of narrative structure—
affect complexity, fluency, accuracy, and lexical diversity in language. The results show
that both native and nonnative speakers are prompted by storyline complexity to use
more subordinated language, but narrative structure had different effects on native and
nonnative fluency. The learners, who were based in either London or Tehran, did not
differ in their performance when compared to each other, except in lexical diversity,
where the learners in London were close to native-speaker levels. The implications of
the results for the applicability of Levelt’s model of speaking to an L2 are discussed, as
is the potential for further L2 research using native speakers as a baseline.

The study was funded by the Economic and Social Research Council of the United Kingdom, award
reference RES-000-22-1155, and was the subject of a presentation by Pauline Foster at PacSLRF
2006 in Brisbane.
Correspondence concerning this article should be addressed to Pauline Foster, St. Mary’s
University College, Waldegrave Road, Twickenham, Middlesex, TW1 4SX, UK. Internet:
[email protected]

Language Learning 59:4, December 2009, pp. 866–896 866


C 2009 Language Learning Research Club, University of Michigan
Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

Keywords storyline complexity; native speaker performance; task effects; narrative

structure; processing capacity

Naturally enough, to observe how features of a second language (L2) might be

affected by different instructional or methodological factors, empirical studies
in second language acquisition (SLA) employ participants who are L2 users. In
the past 15 years or so there has been an abundance of such studies published,
exploring a variety of SLA research areas, among which are testing (Elder,
Iwashita, & McNamara, 2002; Iwashita, Elder, & McNamara, 2001), interac-
tion patterns (Swain & Lapkin, 2001), and classroom task types (Ellis, 1987;
Samuda, 2001; Skehan & Foster, 1997). Native speakers might appear in these
studies as task interlocutors or raters for nonnative performance, but beyond
this, it is rather unusual for their own performance to be subject to scrutiny. Yet,
for some research questions, such as those that compare interactional behavior,
or the effects of task type, a clearer assessment of the way L2 speakers behave
would involve reference to a native-speaker benchmark. It is well known that
native-speaker language performance varies consistently according to the style
of language deemed appropriate for a task, the time available to perform it,
and the pressure of the context (Givon, 1985; Ochs, 1979; Sacks, Schegloff, &
Jefferson, 1974). It is only reasonable then that the performance of language
learners should be influenced not just by the extent of their L2 competence but
also by the same emotional, pragmatic, and information-processing constraints
that operate on all humans. If there is a native-speaker baseline, any discerned
differences between native and nonnative performance could more confidently
be ascribed to the additional processing demands of a second and imperfectly
known language.
There are some studies that include a native-speaker dimension. For testing
in particular, native speaker/nonnative speaker comparisons have been im-
portant in showing whether proficiency tests reliably distinguish native and
nonnative performance while not distinguishing among native speakers them-
selves (Alderson, 1980). Several studies into native-speaker performance on
English proficiency tests showed that particular tests did do this (e.g., Clark,
1977; Johnson, 1977; Stricker, 2002), but a few (e.g., Oller & Conrad, 1971;
Whiteson, 1979) showed they did not, reporting that the native-speaker partic-
ipants could not be distinguished by the test because they sometimes scored
worse than the nonnative speakers. Opinions might differ as to whether this is
because very proficient nonnative speakers are in fact indistinguishable from
native speakers, or because native-speaker proficiency cannot be assumed to be
uniformly high, or because results can be influenced by degree of rehearsal and

867 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

familiarity with language tests (which a native speaker may lack completely).
Added to this is an unease with using the “native speaker” as the gold standard
for language proficiency tests, especially for a language like English, which has
a multitude of native varieties. Davies (2003) showed how the definition of the
term native speaker involves a complex of sociolinguistic and psycholinguistic
attributes—some inherited, others acquired—and Rampton (1990) advocated
the use of the term language expert over native speaker, as that would allow
highly proficient L2 users to be counted as equal to first language (L1) users.1
For L2 research that uses L2 performance to illuminate developmental or
instructional factors, these concerns are perhaps less of an issue. This research
is usually undertaken not to take a measure of proficiency but to see how dimen-
sions of L2 performance are affected by variables such as type of input, type
of interactional pattern, type of task, the presence or absence of planning time,
or the presence or absence of performance time pressure. Overwhelmingly,
such studies involve participants who are not “expert users” of the L2 but are
learners of low to high intermediate level, who are very obviously distinguish-
able from native speakers. What variety of English they are learning is also
not a matter for concern, as long as they are all learning the same one (which,
normally, they are). For the native-speaker element, it is sufficient that these
participants be first language users of whatever variety of the target language
the L2 participants are learning. One can exclude expert L2 learners from this
category on the grounds that we cannot be sure they are indistinguishable from
L1 users, so it is best to avoid the possible confound. One could add that as it
would be much harder to locate a group of very highly proficient L2 users than
a group of L1 users, it is best to proceed with the latter.
As we argued earlier, using a native-speaker baseline in research into L2
performance gives greater validity to any claims that this is affected by the
independent variable(s) under scrutiny. If L1 performance is shown to be af-
fected in the same way, then something other than L2 processing is the cause.
However, to date, very little data of this type has been collected. Foster (2001)
illustrated that it is worth doing. Using native-speaker participants, the study
found that, broadly, planning time before a task conveyed similar performance
benefits on native speakers, as was shown in a parallel study on nonnative
speakers (Foster & Skehan, 1996). This indicates that the fluency and complex-
ity of language generated in task performance depends on more than ease of
language processing and is affected by the design of the task itself. The field
is wide open for replications of, for example, the pretask and online planning
literature (see Ellis, 2005), which would add native-speaker baseline data and
set up comparisons of these with data from advanced and intermediate learners.

Language Learning 59:4, December 2009, pp. 866–896 868

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

In this article, we present the analyses of native-speaker data gathered in a

larger study of the effects of task design on learner output. We briefly review the
research literature on narrative task performance in an L2 and lay out our wider
research project into the effects of narrative task design. The nonnative-speaker
results (Tavakoli & Foster, 2008) are presented in synopsis before we move on
to describe in detail the native-speaker dimension of the project, discussing
these findings in relation to the nonnative-speaker results.

Background to the Study

Currently Levelt’s model (1989, 1993) is the most satisfactory account of the
processes of L1 speaking and is widely accepted. It describes three parts in
speech production: Conceptualization is the preverbal stage in which the ideas
to be expressed are selected and organized; formulation encodes these ideas
into linguistic form through the retrieval of lemmas and syntactic structures;
articulation translates these forms into a phonetic plan that is executed by the
neuromuscular system. Obviously, there is a degree of parallel processing to
this. Conceptualizing a prelinguistic idea can proceed as the previous one is
being articulated, and when a word is selected for use, its collocations and
colligations are activated along with it (Hoey, 2005), whereas the fine motor
control needed for speech articulation has a high degree of automatization. If
this were not so, speech would be characterized by very frequent pauses as
articulation waited for the next concept to be hatched and delivered for formu-
lation, and conceptualization of the next idea would be waiting for articulation
of the previous one to be complete. Of course, there are always some pauses
in speech: deliberate ones that punctuate the stream and help the listener to
decode it, as well as other dysfluencies that arise out of a longer search to
find the right formulation for a concept, or indeed to conceive an idea in the
first place. Additionally, there is the self-monitoring and editing that cause a
speaker to change tack midstream and start afresh. These may be more frequent
if challenging subject matter, or external distractions, absorb too much of the
speaker’s attention.
This is an important consideration. In the context of the psychology of
learning, human attention is a limited resource. Where the need for attention
is greater than the resources available, as in attending to two things at once,2
the mind will prioritize one, with the result that the other will suffer, or it will
split attention to both tasks, with the results that they will be performed slowly
or poorly. New skills, like typing or driving, are very attention-demanding
at the start, but through repeated practice and the process of automatization

869 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

(Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977), they gradually de-
mand less. Performance speeds up, accuracy increases, and other things that
need attention, like carrying on a conversation, can be undertaken at the same
time. Skehan (1998) applied this model to SLA; an imperfectly learned L2
imposes a large burden on the learner’s attention because it involves, among
other things, meaning and form, fluency and accuracy. A learner is someone
who has not yet developed fully automatized control over L2 morphosyntax or
articulation and so makes choices as to where to allocate attention, prioritizing
one aspect of performance, such as being syntactically accurate, over another,
such as being phonologically accurate.3
The concept of competing claims on limited attention lies behind much
research into task-based L2 performance (Ellis, 1987; VanPatten, 1990). The
thrust of the research has been to determine how language learners are likely to
allocate attention and what task features can be said to influence this. Planning
studies (Crookes, 1989; Foster & Skehan, 1996; Ellis & Yuan, 2005; Ortega,
1999) show a consistent effect for planning time, in terms of increased com-
plexity and fluency in performance. Some show increases in accuracy (Foster
& Skehan; Mehnert, 1998; Wigglesworth, 2001) although Foster and Skehan
suggested that the results indicate a trade-off between accuracy and complex-
ity, as competing demands on limited resources mean that some things get
higher priority. Task design has also figured in this line of research. Tasks that
are deemed to be cognitively less demanding allow greater attention to lan-
guage form because they take less for language meaning. Research suggests
that features affecting the cognitive complexity of a task include such things
as the task-taker’s familiarity with the content and/or format, the clarity of the
information given, the amount of computation involved, the number of other
interlocutors involved, and time allowed for completion (see Robinson, 2001b,
and Skehan & Foster, 2001, for a fuller exploration of this).
In the study reported in Tavakoli and Foster (2008), we explored how
the demands of the task being carried out affect the allocation of a language
learner’s limited attentional resources and impact on language performance,
as displayed by its fluency, accuracy, complexity, and lexical diversity. Our
particular interest was in two narrative task characteristics that previous studies
have suggested affect performance: storyline complexity and inherent structure.
Storyline complexity refers to whether a narrative has background as well as
foreground events, with a narrative with background events classified as more
complex than one with only foreground events. Thus, in the study we report
here, the Walkman task (see the Appendix) is categorized as a narrative with
storyline complexity because although the foreground action is about a man

Language Learning 59:4, December 2009, pp. 866–896 870

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

taking a walk about town, there are things going on simultaneously—a car
crash, a bank robbery, an escaped tiger—that the narrator must include. The
Journey task, on the other hand, is categorized as without storyline complexity,
because beyond the foreground action of a couple cycling from place to place,
there are no other actions to include. Narrative structure refers to how the
episodes within the narrative connect to make the whole. A narrative with an
inherently loose structure, such as Walkman, can have its episodes reordered
without loss of coherence. (It does not matter if the bank robbery comes before
or after the escaped tiger.) A narrative with an inherently tight structure cannot
be reordered in this way. In the Football task, each frame must stay where it is
for the narrative to make sense.
Our line of enquiry was prompted by a post hoc finding in previous research.
Tavakoli and Skehan (2005) concluded that the unexpectedly high syntactic
complexity scores in two of the narrative tasks they used could be ascribed
to the narrative structure. An inspection of the data transcripts pinpointed
increased syntactic complexity at moments when the participants were trying
to connect foreground and background events. As Matthiessen and Thompson
(1988) and Harris and Bates (2002) both noted, in English the conditions,
reasons, and purposes of an event are signaled through syntactic subordination,
so it is to be expected that anyone connecting background to foreground events
in a narrative will be prompted to do so through subordinate clauses.
Two other studies (Foster & Skehan, 1996; Skehan & Foster, 1997) explored
the extent to which familiar and unfamiliar task content would impact on L2
task performance. In general, the results went as predicted: that talking in
an L2 about well-known information resulted in a more fluent and accurate
performance, whereas talking about new and unfamiliar information entailed
a performance that was less fluent and less accurate (but more complex).
Intriguingly, a post hoc analysis of the data from the two studies together
showed that a variable other than familiarity of content might be involved. The
most fluent performances across the two studies were observed in tasks that did
not have the same degree of familiar information. With the addition of planning
time (another variable under investigation), these two tasks returned the most
accurate performance. The interpretation offered was that the key variable was
structure: Both of these tasks dealt with events that had a clear time sequence
from beginning to middle and end and this had eased the L2 processing burden
and enabled greater attention to accuracy and fluency in performance. This was
tested out by Skehan and Foster (1999) and upheld. A narrative task with a
tightly structured storyline was associated with greater fluency and accuracy
than a loosely structured narrative. If a narrative is tightly sequenced from

871 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

beginning to end, the processing burden of telling it in an L2 is eased, but this

does not operate in a narrative in which the episodes are loosely ordered.
Apart from Foster (2001),4 none of the research into the impact of task
design or implementation conditions has a native-speaker dimension, although
clearly Foster’s results suggest that it is interesting to attempt it. Consequently,
the study reported by Tavakoli and Foster (2008) was designed to encompass
baseline data from native speakers in its exploration of how language perfor-
mance is influenced by narrative structure and storyline complexity.

The Nonnative-Speaker Study: Tavakoli and Foster (2008)

This part of the larger study examined how narrative structure and storyline
complexity might influence the performance of L2 users. The participants were
100 learners of English, 60 based in Tehran and 40 based in London, allowing
comparisons to be made between learners within the target language community
and learners outside it. As Figure 1 shows, each participant performed two of
four narrative tasks that had different degrees of narrative structure (loose
or tight) and storyline complexity (with or without background events). The
narratives with one storyline (Football and Journey) contained only foreground
information, whereas narratives with two storylines (Picnic and Walkman)
contained both foreground and background information. In the two loosely
structured narratives (Journey and Walkman) the events could be reordered
without compromising the main story, and in two tightly structured narratives
(Football and Picnic) this was impossible. The cartoon strips for all tasks are
reproduced in the Appendix.

i n he r e n t n a r r a t i ve s t r u c t u r e

Storyline complexity loose tight

+ foreground Journey Football

− background n = 50 n = 50

+ foreground Walkman Picnic

+ background n = 50 n = 50

Total n = 100. Participants did either Journey and Football tasks, or Walkman and Picnic
tasks. The task order was counterbalanced. The native speaker study reported here used
exactly the same design, with n = 20 in each cell, and total N = 40.

Figure 1 Research design for Tavakoli and Foster (2008).

Language Learning 59:4, December 2009, pp. 866–896 872

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

The learners in London were mostly female, between 19 and 47 years of

age. They came from a wide variety of mother tongue backgrounds. In an
Oxford Placement test they had all been scored as “intermediate” (band 4).
The learners in Tehran were all female, between 19 and 45 years of age. All
had Persian as a mother tongue. In the locally administered placement test of
English proficiency they had achieved scores that put them in an “intermediate”
category. A Pearson product-moment correlation was run to compare the local
test with the Oxford Placement test. The coefficient correlation (r = .56, p <
.01) was considered large enough to make the two tests comparable. All learners
in both venues were also given oral interviews to confirm their proficiency and
correct class placement.
Quantitative measures5 of lexical diversity, fluency, accuracy, and complex-
ity of the transcribed data were taken and analyzed through t tests and analyses
of variance. The results largely confirmed findings of previous research. Nar-
rative tasks with a background element gave rise to significantly more complex
language than those without. Narrative tasks with a tight narrative structure
gave rise to significantly more accurate language than those with a loose struc-
ture. There was some evidence that a tightly structured storyline was associated
with greater fluency than was a loosely structured storyline. Task design had
no discernible effect of lexical diversity. However, although the comparison
between learners in Tehran and London showed little difference in syntactic
complexity, accuracy, and fluency, the two groups of learners differed in their
use of vocabulary. Learners in London used vocabulary that was significantly
more diverse than that used by the learners in Iran.

The Native-Speaker Study

The aim of this part of the research project study was to discover whether
narrative task design influences native-speaker language in similar ways (i.e.,
in terms of syntactic complexity, fluency, and lexical diversity). Accuracy was
not investigated, however. Although native speakers make occasional slips of
the tongue and suffer a few lapses in concentration that lead them into syntactic
dead-ends, by and large their knowledge of L1 grammar is secure (Foster, 2000).
The tasks used were identical to the ones in the nonnative-speaker study.

Hypotheses
Essentially the same hypotheses were used as in the nonnative-speaker part of
the study.

873 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

H1: Compared to narratives with only foreground events, those in which

both foreground and background information need to be described will
be associated with native speakers producing more complex language
(Harris & Bates, 2002; Matthiessen & Thompson, 1988).
H2: Compared to narratives with a loose structure, narratives with a tight
structure will be associated with native speakers producing more fluent
language (Skehan & Foster, 1999; see H4).
H3: Compared to narratives with only foreground events, narratives with
both foreground and background events will not be associated with
native speakers using language with greater lexical diversity. (We take
the null hypothesis in the absence of previous research on this question.)
One further hypothesis was formulated for the comparison of the native-
and nonnative-speaker results:
H4: Task effects on fluency, complexity, and lexical diversity will be the
same for both native and nonnative speakers. (Foster, 2001, has shown
that native and nonnative task performance is affected in similar ways
by the variables of planning time and task design.)

Design
This replicates the design used in the nonnative-speaker phase of the research
(Figure 1), with the number of participants in this phase being the only dif-
ference. It has a 2 × 2 factorial design with storyline complexity and inher-
ent narrative structure as the two independent variables. Storyline complexity
was operationalized as number of storylines; the narratives with one storyline
(Football and Journey) contained only foreground information, whereas nar-
ratives with two storylines (Picnic and Walkman) contained both foreground
and background information. Inherent narrative structure was defined in terms
of whether the narratives were loose or tight. In the two loosely structured
narratives (Journey and Walkman) the events could be reordered without com-
promising the main story, and in two tightly structured narratives (Football
and Picnic) this was impossible. The participants told either Journey and Foot-
ball or Walkman and Picnic. This design allowed for a within-participants
comparison for inherent narrative structure and a between-participants com-
parison for the presence or absence of background events (see Figure 1). These
tasks are from Heaton (1966), Jones (1980) and Swan and Walter (1990) (see
Appendix).
Participants were 40 native speakers of English,6 all first-year students
studying Literature or Psychology at a university in London. They were between

Language Learning 59:4, December 2009, pp. 866–896 874

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

18 and 60 years of age. The four narratives were typical of picture stories
used in English as a foreign language (EFL) classrooms. Each comprised six
cartoon frames. Previous trialing had shown the narrative in each was engaging,
clear, and unlikely to baffle. Each participant met individually with one of the
researchers in a quiet room. They were asked to look over the first picture story
for up to 3 min in order to understand the narrative and then they were asked
to tell it, as if to someone who could not see the pictures. Their performance
was recorded. When the first task was completed, the same procedures were
followed for a second story. To control for any practice effect, the order of
the stories was counterbalanced. Inspections of the transcripts showed that all
of the participants accomplished the tasks as expected, telling the stories as
presented in the pictures. No one did the tasks in an idiosyncratic way or came
up with an anomalous story.

Transcripts, Measures, and Coding

The recorded performances were transcribed and coded for the following de-
pendent variables:
• Repair Fluency: false starts, reformulations, word repetition, and replace-
ment.
• Breakdown Fluency: number of midclause and end-clause pauses; total
amount of midclause and end-clause silence.
• Syntactic Complexity: AS units, clauses,7 and mean length of utterance
(MLU).8
• Lexical diversity: corrected measure of diversity (D) (Malvern & Richards,
2002).9
Using Soundscriber, the recorded data were first transcribed into Word
documents and then transformed to CHAT format files. GoldWave Digital
Editor was used to measure all of the pauses greater than 0.4 s occurring
between and at clause boundaries. CLAN software of the CHILDES program
was employed to calculate MLU and D. A sample of 10% of the total transcripts
was checked by a second rater and an interrater reliability score of greater than
95% was obtained on each measure.

Analyses and Results

Initially, a repeated-measures MANOVA was performed to investigate the ef-
fects of narrative task design on the native speakers’ performance. This deter-
mined whether further detailed analyses were allowed. The results show that

875 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

the effect of inherent structure on the native speakers’ performance did not
reach significance (F = 2.10, ns, η2 = .149). Neither was there a significant
difference in the interaction of inherent structure and storyline complexity (F =
2.24, ns, η2 = .158). The only difference to reach statistical significance was
the effect of storyline complexity (F = 3.25, p < .05, η2 = .213), indicating
that the native speakers’ performance on foreground and background narratives
is significantly different from that on tasks with foreground events only. The
effect size for this difference indicates a noticeable effect (Tabachnic & Fidell,
1996) of storyline complexity variable on the native speakers’ performance. To
test Hypotheses 1 to 4, t-tests were run to compare the mean scores on measures
of fluency, complexity and lexical diversity.

H1: Effects of Storyline Complexity

Hypothesis 1 predicted that complexity of the language produced by the native
speakers will be affected by the number of storylines in a narrative task. Com-
plexity was measured through ratio of syntactic clauses to AS units and also
through MLU. The results are shown in Tables 1 and 2, respectively.
Table 1 shows significant differences for the complexity of the native speak-
ers’ performance on both pairs of narratives. For the tightly and loosely struc-
tured narratives, the presence of two storylines is associated with a significantly
higher ratio of clauses to AS units: 1.71 vs. 1.50 for Picnic/Football (F = 2.56,
p < .01) and 1.86 vs. 1.54 for Walkman/Journey (F = 2.62, p < .01). Table 2
shows similar results for MLU. Statistical significance is reached for the com-
parison between the two tightly structured narratives—0.81 vs. 9.09 for Pic-
nic/Football (F = 2.64, p < .01)—although not for the loosely structured
narratives—10.73 vs. 9.33 for Walkman/Journey (F = 1.87, ns). Nevertheless,
the trend here is for narratives with two storylines to be associated with a

Table 1 Mean number of clauses per AS unit

M SD

Tightly structured narratives

Picnic (+ background) 1.71 0.22
Football (− background) 1.50 0.30
∗
F = 2.56, p < .01 , η2 = .150
Loosely structured narratives
Walkman (+ background) 1.86 0.39
Journey (− background) 1.54 0.37
∗
F = 2.62, p < .01 , η2 = .150

Language Learning 59:4, December 2009, pp. 866–896 876

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

Table 2 Mean length of utterance

M SD

Tightly structured narratives

Picnic (+ background) 10.81 1.83
Football (− background) 9.09 2.28
∗
F = 2.64, p < .01 , η2 = .160
Loosely structured narratives
Walkman (+ background) 10.73 2.67
Journey (− background) 9.33 2.00
F = 1.87, p < .07 ns, η2 = .08

longer MLU, with the native speakers consistently producing longer syntactic
runs when narrating Walkman and Picnic. The results from these two measures
of complexity lend support for H1.
Hypothesis 4, which predicts no difference in the effects of task design on
native and nonnative speakers, is also supported here. As reported by Tavakoli
and Foster (2008), nonnative performance on these same narrative tasks was
affected in the same way by storyline complexity. The narrative tasks where
the storyline contained background as well as foreground information resulted
in learners attempting significantly more complex language to describe them,
both in Tehran (Picnic vs. Football: F = 4.13, p < .001; Walkman vs. Journey:
F = 2.02, p < .05) and in London (Picnic vs. Football: F = 4.30, p < .001;
Walkman vs. Journey: F = 2.85, p < .01).
In terms of MLU there was also an effect for storyline complexity; in tasks
with background events, learners produced significantly longer MLUs, both in
Tehran (Picnic vs. Football: F = 4.63, p < .001; Walkman vs. Journey: F =
2.87; p < .01) and in London (Picnic vs. Football: F = 3.78, p < .001; Walkman
vs. Journey: F = 4.82, p < .001). The effect of storyline complexity is the same
for both learners and native speakers. Figures 2 and 3 illustrate how the data for
all three groups (native speakers, learners in London, and learners in Tehran)
follows the same general pattern across tasks, although it also indicates that,
for the Picnic task, the nonnative-speaker data from London converges with
the native-speaker data.

H2: Effects of Task Structure

Hypothesis 2 predicted that fluency of the language produced by the native
speakers will be greater for narratives with a tight structure. Because of the
multifaceted nature of fluency, we looked at a wide range of measures of

877 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

2.00
1.90
Syntactic Subordination

1.80
1.70
1.60 NNS Tehran
1.50 NNS London
1.40 NS
1.30
1.20
1.10
1.00
Journey Football Walkman Picnic
Narrative Tasks

Figure 2 Patterns of syntactic subordination across the three groups.

11

10

9 NNS Tehran
MLU

NNS London
8 NS

6
Journey Football Walkman Picnic
Narrative Tasks

Figure 3 Patterns of MLU across the three groups.

both repair fluency (reformulation, false start, replacement, and repetition) and
breakdown fluency (number of pauses occurring in the middle and at the end
of clauses, total amount of silence in the middle and at the end of clauses).
Table 3 shows the results of the t tests comparing measures of repair fluency
and Table 4 shows the results of t tests comparing breakdown fluency.
The figures in Table 3 are all very low, indicating that native speakers, by and
large, do not need to repair their speech. The scores are, with one exception,
higher for the loosely structured narratives (Journey and Walkman) than on

Language Learning 59:4, December 2009, pp. 866–896 878

879

Foster and Tavakoli

Table 3 Mean values for measures of repair fluency

Football Journey F p Picnic Walkman F p

Effects on Complexity, Fluency, and Lexical Diversity

Language Learning 59:4, December 2009, pp. 866–896

Reformulations 0.2 (0.52) 0.35 (0.67) 0.90 .370 ns 0.30 (.80) 0.50 (.51) 0.94 .359 ns
False starts 1.00 (2.00) 1.45 (2.00) 0.96 .345 ns 0.80 (.89) 0.85 (1.00) 0.20 .847 ns
Replacements 0.65 (1.2) 0.85 (1.38) 0.94 .359 ns 0.45 (.51) 0.25 (.55) 1.28 .214 ns
Repetitions 0.4 (1.00) 0.55 (1.23) 0.82 .419 ns 0.25 (.55) 0.50 (.88) 0.142 .171 ns
Note. Standard deviations are given in parentheses.
Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli

Table 4 Mean values for measures of breakdown fluency

Football Journey F p Picnic Walkman F p

Effects on Complexity, Fluency, and Lexical Diversity

Mid-clause pauses 2.53 (2.05) 3.65 (3.22) 1.64 .116 ns 2.70 (3.61) 3.15 (2.30) 0.63 .536 ns
End-clause pauses 9.19 (4.41) 9.50 (3.09) 0.34 .734 ns 7.82 (2.44) 8.00 (2.97) 0.41 .685 ns
Mid-clause silence 1.89 (2.92) 2.92 (2.73) 1.88 .080 ns 2.06 (2.78) 2.58 (2.20) 1.19 .249 ns
End-clause silence 6.22 (7.00) 7.00 (3.27) 1.21 .241 ns 5.85 (2.78) 6.26 (3.01) 0.69 .496 ns
Note. Standard deviations are given in parentheses.
880
Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

tight ones (Picnic and Football), but there is no significant difference for any
measure. We can conclude that the inherent structure of a narrative does not
affect any measures of repair fluency in native speakers’ performance.
Table 4 shows that though the native speakers consistently produce fewer
pauses and a smaller amount of silence on tightly structured narratives (Football
and Picnic) than on loosely structured narratives (Journey and Walkman), these
differences are far from significant. Taken together these results do not support
H2 and suggest that native-speaker fluency is not affected by inherent narrative
structure.
For H4, concerning the comparison of effects on non-native speaker perfor-
mance, we use the results reported in Tavakoli and Foster (2008). These showed
that inherent task structure had a very limited effect. The only measure of repair
fluency to reach statistical significance was false starts in the Tehran data for
the Walkman/Picnic comparison (3.63 vs. 6.23, F = 3.49, p < .01). This is as
predicted, but in the absence of any other significant results for repair fluency,
it cannot be taken as strong support for it. For the learners in London, no sig-
nificant results were reached for any measure of repair fluency. On measures
of breakdown fluency, performances were generally more fluent for narratives
with tight structure, but only two measures reached significance: These were
number of pauses midclause for the learners in Tehran for the Football/Journey
comparison (9.05 vs. 11.46, F = 3.38, p < .01) and total amount of silence mid-
clause for the learners in London for the Walkman/Picnic comparison (8.13 vs.
9.89, F = 2.67, p < .05). Taken together, the nonnative and native results show
no effect on native speakers for inherent narrative structure, whereas nonnative
performance, especially in breakdown measures such as pausing, is affected
to some small degree. H4, then, receives weak support.10 Figure 4 shows the
pattern of results for all three groups of participants. As reported earlier, a task
effect is discernible but weak, with a little less pausing on tasks with inherent
structure. What is most noticeable is that the native speakers pause very much
less midclause than their nonnative counterparts. Figure 5 shows an opposite
pattern, with end-of-clause pauses very much common in the nonnative data,
although with the learners in London converging on the native speaker levels
for the Picnic task.

H3: Lexical Diversity

Hypothesis 3 was a null hypothesis, predicting that storyline complexity would
not be associated with greater lexical diversity. The results are shown in Table 5.
The results in Table 5 show that although the presence of two storylines in
Picnic did not give any greater lexical diversity compared to presence of a single

881 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

14

12
No. of mid clause pauses

10
NNS Tehran
8
NNS London
6
NS
4

0
Journey Football Walkman Picnic
Narrative Tasks

Figure 4 Patterns of midclause pausing in the three participant groups.

10

9
No. of end-clause pauses

8
NNS Tehran
7
NNS London
6
NS
5

3
Journey Football Walkman Picnic
Narrative Tasks

Figure 5 Patterns of end-clause pausing in the three participant groups.

storyline in Football, the Walkman/Journey comparison indicates that the lexi-

cal diversity of performance on Walkman was significantly different. However,
as we acknowledge in the parallel nonnative-speakers study, because of the dif-
ferent stimuli prompting vocabulary choices, such cross-task comparisons are
not properly valid and must be treated with caution.11 H4 predicted no differ-
ence in the way storyline complexity would affect lexical diversity across native
and nonnative speakers’ performance. This was measured through within-task
comparisons, so the validity problem we note earlier does not apply. Figure 6

Language Learning 59:4, December 2009, pp. 866–896 882

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

Table 5 Mean values of D

M SD

Tightly structured narratives

Picnic (+ background) 39.90 8.88
Football (− background) 40.21 7.89
F = 0.120, p < .905 ns
Loosely structured narratives
Walkman (+ background) 45.67 10.15
Journey (− background) 38.75 9.77
∗
F = 2.22, p < .03 , η2 = .115

50.00

45.00
D: Lexical Diversity

40.00
NNS Tehran
35.00 NNS London
NS
30.00

25.00

20.00
Journey Football Walkman Picnic
Narrative Tasks

Figure 6 Patterns of lexical diversity across the three groups.

shows how similar the patterns of results are. In both the London and the
Tehran data, the Picnic task, which does have background events, gives rise to
less lexical diversity than the Football task, which does not. However, Walkman,
the other task with background events, gives rise to greater lexical diversity
than Journey, which has no background. As there was no significant effect for
storyline complexity in either the London or the Tehran data, we can conclude
that this narrative feature does not influence lexical diversity in either native or
nonnative performance.
We summarize here that fluency and lexical diversity in native-speaker
narratives are not affected by narrative design and that native speakers maintain
fluency regardless of whether they are narrating a story with a tight narrative
structure or a loose one. It is also notable that incidences of repair fluency were

883 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

uniformly few (Table 2) and that when native speakers did pause, it was far
more likely to be on a syntactic boundary than between them (Table 3). Lexical
diversity in native speakers was unaffected by task design with similar scores
for D reported for single- and dual-storyline tasks. There was, as predicted,
an effect on syntactic complexity for storyline complexity, with the ratio of
clauses to AS units significantly higher for the narratives with dual storylines
and MLU significantly higher also for dual narratives with an inherently tight
structure.
When comparing task effects on native and nonnative speakers, it is clear
that storyline complexity operates in the same way on syntactic complexity
regardless of whether the speaker is using English as an L2 or a native language.
However, whereas a tight inherent narrative structure exerts no influence on
native-speaker fluency, it does seem that it helped nonnative speakers to perform
more fluently in the L2. In neither group did narrative structure have a significant
effect on lexical diversity.
Interestingly the data presented earlier indicate that the variance among
native and nonnative speakers’ performance was not greatly different. The
standard deviation figures show that there is no particular difference between
the two groups in the variance in scores for fluency, length of utterance, and
lexical diversity.

Discussion
The study reported here adds a native-speaker dimension to the nonnative
speaker study reported by Tavakoli and Foster (2008) in order to see whether
effects that might be ascribed to the demands of processing an imperfectly
known language are better ascribed to the demands of processing a task. The re-
sults show quite clearly that native speakers, whose L1 skills rest on a secure
and implicit knowledge of its systems, maintain their fluency regardless of the
structure or complexity of the narrative they are telling and that when they
pause (as they must, for a seamless performance would be highly unnatural
and very hard on the listener), they are much more likely to do so at syn-
tactic boundaries, indicating that they plan and deliver their stream of speech
in clause-length runs (Pawley & Syder, 1983). Our data show that pausing in
nonnative speakers’ performance occurs more often in the middle of clauses.
They are less secure in planning and delivering a whole clause of the L2 at a
time, but, as Tavakoli and Foster (2008) showed, the processing burden can be
eased by a narrative such as Picnic and Football, in which the events follow on
from each other in a predictable way. The native-speaker data presented here

Language Learning 59:4, December 2009, pp. 866–896 884

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

shows that L1 fluency is something that does not rest on this aspect of task
design, whereas L2 fluency does, at least in terms of degree and positioning of
pausing.
For complexity, the conclusions are different. Tavakoli and Foster (2008)
showed that for nonnative speakers, greater storyline complexity resulted in
greater syntactic complexity. This pattern of results was obtained here too with
the native speakers and indicates that increased syntactic complexity is, at least
partly, a task effect and not wholly a language processing effect. Dual storylines
provoke greater use of subordinated clauses in English; single storylines do
not. A native speaker selects subordination to weave two storylines together,
not because a secure knowledge of English subordination allows it but because
the nature of the narrative like Picnic and Walkman requires it. Thus, a learner
with an insecure control of English subordination (such as relative clauses,
if/when clauses, unless/although/in case clauses) has a harder task telling such
a narrative and may be less accurate or less fluent than on a narrative with only
a single storyline.
In Tavakoli and Foster (2008), we suggested that, for oral testing purposes,
care had to be taken when choosing narratives for an exam. The native-speaker
data presented in this part of the study reinforce such a conclusion.
The lexical diversity results are most interesting. The comparison of the
mean scores of the three groups of participants’ lexical diversity of performance
shows that, unlike on any other measure, the learners in London are far closer
to the native speakers than they are to the learners in Tehran. This was not
predicted by any of the original hypotheses of the study, so it does invite a post
hoc analysis of variance to compare the three groups across the four tasks. This
is shown in Table 6.
We see from Table 6 that across all four tasks the learners in London are
producing language that is significantly more lexically diverse than that of the
learners in Tehran and at the same time is not significantly less diverse than
that of the native speakers. The first part of this result was already clear in
Tavakoli and Foster (2008) where the t test to compare D scores from the
London and Tehran learners on all tasks reported significant results in favor of
greater diversity in the London data (Football p < .01; Journey p < .02; Picnic
p < .003; Walkman p < .006). The ANOVA in Table 6 shows the remarkable
extent of this difference. The learners in London use a diversity of vocabulary
that is closing in on the native speakers. When we consider ANOVAs for all
other dimensions of performance, we see this result on only one other variable:
total seconds silence at end of clause, where across all tasks the London data
are significantly different from the Tehran data, and not significantly different

885 Language Learning 59:4, December 2009, pp. 866–896

Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli

Table 6 Mean values of D for learners in London (L), learners in Tehran (T), and for native speakers (NS)

Tehran (n = 60) London (n = 40) NS (n = 40) F p Significant differences η2

Effects on Complexity, Fluency, and Lexical Diversity

∗
Football 28.75 (11.20) 38.37 (11.18) 40.21 (7.89) 9.02 .001 T vs. L & NS .212
∗
Journey 25.82 (9.49) 36.11 (11.01) 38.75 (9.77) 11.84 .001 T vs. L & NS .261
∗
Picnic 27.76 (5.89) 36.59 (9.46) 39.90 (8.88) 15.87 .001 T vs. L & NS .322
∗
Walkman 33.62 (6.40) 43.37 (12.43) 45.67 (10.15) 11.61 .001 T vs. L & NS .258
Note. Standard deviations are given in parentheses.
∗
Significant differences are reached across tasks.
886
Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

from the native-speaker data (Football p < .001, η2 = .322; Journey p < .02,
η2 = .335; Picnic p < .001, η2 = .282; Walkman p < .002, η2 = .175). We can
see these results as linked in some way. Knowledge of lexis, especially the
kind that is gained implicitly through frequent exposure, is not just knowledge
of individual words but of chunks of words that occur regularly in the same
patterns.12 A wider knowledge of words in phrases, such as would be gained
through living within the target culture and hearing the language on a daily
basis, would increase the learner’s ability to plan and execute phrase by phrase,
not word by word, which a lesser exposure, say only through classroom contact,
might allow. Hence, the learners in London have acquired lexical knowledge,
which means they pause more naturally at clause boundaries than within them.
Again, the native-speaker data shows this aspect of L2 performance is closing
in on native-speaker patterns of fluency. The results suggest that living inside
the target language community conveys a considerable lexical advantage, with
learners in London able to draw upon an enriched and networked lexicon.
A further qualitative analysis of the narrative transcripts undertaken of each
narrative, frame by frame, examined the lexical choices of the learners and the
native speakers in the face of the same picture prompts and showed that the
learners in London do indeed have a greater familiarity with nativelike lexical
selections (i.e., they were much more likely than the learners in Tehran to make
lexical choices similar to the native speakers), and that they use many more
lexicalized phrases13 than the learners in Tehran. The London data contained
twice as many of these as the Tehran data, although only half as much as
the native speaker data (see Foster, 2009). Crucially, however, there is no
grammatical advantage. As shown by Tavakoli and Foster (2008), in terms of
accuracy of performance the two groups of learners are not distinct.
Such results pose interesting questions for how a model of L2 speaking
fits Levelt’s model of L1 speaking, discussed earlier. We could perhaps safely
assume that in the prelinguistic conceptualization phase, native and nonnative
speakers are on a level playing field, sharing whatever conceptual demands
the speaking task imposes. However, in the formulation stage, during which
linguistic coding of the message takes place, the playing field is not even. Native
speakers have greater knowledge of and faster access to the linguistic code and
can formulate information even as conceptualization feeds it in. Nonnative
speakers with less knowledge (of say, lexical relationships or relative clauses)
and slower retrieval may have to conceptualize and formulate utterances in
alternation rather than in tandem due to insufficient attentional capacity and may
also be going through an intermediary stage whereby formulation occurs first in
the structures of their L1 and then translated before it can move to articulation.

887 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

Both such delays would account for the higher incidence of midclause pausing
that characterizes L2 performance (and, of course, may be exacerbated by a
less than automatic neuromuscular control of L2 phonemes). However, greater
use of nativelike lexical phrases in the London learner data (Foster, 2009)
would also account for the different pausing patterns (Figures 4 and 5) between
learners in London and Tehran. The learners in London, with a greater store
of chunked phrases, were able to plan and execute more clause-length runs
because their L2 formulator was taking the same processing shortcuts as native
speakers. We have reported here that whereas the tightness or looseness of
narrative structure made no difference to native-speaker fluency, it had affected
nonnative-speaker performance, with tight narrative structure associated with
increased fluency and accuracy. We can see this task feature in terms of easing
attentional demands on the conceptualizer, because the direction of a tight
narrative requires little thought or interpretation, and so for the L2 formulator
and articulator, there is more attention available. For native speakers, who
make few errors beyond an occasional slip of the tongue or loss of syntactic
cohesion, neither the formulator nor the articulator require much attentional
capacity to start with, so the extra space afforded by a tight narrative has no
discernible effect.
Although one might claim that with a few modifications to account for depth
of knowledge and speed of retrieval, the Levelt model is directly applicable to
L2 performance (see Kormos, 2006), research evidence is needed to put some
flesh on these bones. This article has gone some way to providing a small
amount, and clearly there is scope for a great deal more work. Especially,
it would be interesting to compare native-speaker performance to different
levels of nonnative-speaker performance. It would be very useful to be able to
characterize nonnative-speaker performance at a level approaching nativelike
to explore in which dimensions their performance comes to resemble that of
native speakers and, ultimately, to contrast this with high- and low-intermediate
and beginning learners. Our data suggest that intermediate learners with greater
exposure to the target language might be relying more on lexical processing
than grammatical processing. One question that presents itself is whether it
would be possible to see at what point in their development, learners integrate
lexical and syntactic processes as they approach higher levels of proficiency.
Coupled to this would be explorations of task variables, such as the design
of content or implementation conditions, to illuminate further which variables
do or do not affect native-speaker performance and at what levels of nonnative
proficiency we see a convergence with native-speaker results. Although most of
this research is envisaged to be based on comparisons of groups of learners and

Language Learning 59:4, December 2009, pp. 866–896 888

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

native speakers, there is possibly much to be learned from a within-participants

design where the same speaker’s performance is compared across the L1 and
the L2. The scope is wide and the potential great. Fortunately, native speakers
of the languages typically researched in SLA are in plentiful supply and not
hard to find.
Revised version accepted 7 December 2008

Notes
1 See Skehan and Foster (2009) for a full discussion of the definition of native and
nonnative speakers.
2 Such as simultaneously rotating your left foot clockwise and your left hand
anticlockwise or speaking on a mobile phone while driving through heavy traffic.
3 Robinson (2001a, 2001b) used work in psychology by Navon (1989) and Neuman
(1996) to challenge limited-capacity processing, proposing a different model of
attention in which there are multiple attentional “pools.” For him, depletion of
attention in one pool has no effect on the amount remaining in another, so language
learners do not need to prioritize one dimension of performance over any other. In
his revised Triadic Framework (Robinson & Gilabert, 2007), accuracy and
syntactic complexity depend on whether the cognitive complexity of a task is
driven by resource-directing or resource-dispersing variables (pp. 166–167).
4 As we note earlier, this study is a replication of the study by Foster and Skehan
(1996) using native rather than nonnative speakers.
5 Full details of these, and the statistical procedures used, are given in this article and
are compared to the parallel native-speaker results.
6 All had learned English from early childhood; none were proficient in any other
language.
7 AS units and clauses divide the data into complete and dependent syntactic units
and enable a syntactic complexity measure to be calculated as a ratio of clauses to
AS units. Foster, Tonkyn, and Wigglesworth (2000) give a full and detailed
definition of how these are identified.
8 MLU was expressed as mean number of words per AS unit.
9 Unlike a straightforward type-token ratio, the measure D corrects for length of text
and is therefore more valid.
10 The very different patterns of results in London and Tehran are discussed further
below.
11 We thank Jeanine Treffers-Daller for this observation.
12 See, for example, Weinert (1995).
13 These are defined here as either fully or partially prefabricated chunks of language
that constitute single choices for retrieval by the formulator.

889 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

References
Alderson, C. (1980). Native and non-native speaker performance on cloze test.
Language Learning, 30, 59–76.
Clark, J. (1977). The performance of native speakers of English on the Test of English
as a Foreign Language (TOEFL Report No 1). Princeton, NJ: Educational Testing
Service.
Crookes, G. (1989). Planning and interlanguage variation. Studies in Second Language
Acquisition, 11, 367–383.
Davies, A. (2003). The native speaker: Myth and reality. Clevedon, UK: Multilingual
Matters.
Elder, C., Iwashita, N., & McNamara, T. (2002). Estimating the difficulty of oral
proficiency tasks: What does the test-taker have to offer. Language Testing, 19(4),
343–368.
Ellis, R. (1987). Interlanguage variability in narrative discourse: Style shifting in the
use of the past tense. Studies in Second Language Acquisition, 9, 12–20.
Ellis, R. (2005). Planning and task performance in a second language. Amsterdam:
Benjamins.
Ellis, R., & Yuan, F. (2005). The effects of careful within-task planning on oral and
written task performance. In R. Ellis (Ed.), Planning and task performance in a
second language (pp. 167–193). Amsterdam: Benjamins.
Foster, P. (2000). Attending to message and medium: The effects of planning time on
the task-based language performance of native and non-native speakers.
Unpublished doctoral dissertation, King’s College, London.
Foster, P. (2001). Rules and routines: a consideration of their role in the task-based
language production of native and non-native speakers. In M. Bygate, P. Skehan, &
M. Swain (Eds.), Researching pedagogic tasks: Second language teaching, learning
and testing (pp. 75–95). London: Longman.
Foster, P. (2009). Lexical diversity and native-like selection: The bonus of studying
abroad. In H. Daller, D. Malvern, P. Meara, J. Milton, B. Richards, and J.
Treffers-Daller (Eds.), Vocabulary studies in first and second language acquisition:
The interface between theory and application (pp. 91–106). Basingstoke, UK:
Palgrave-Macmillan.
Foster, P., & Skehan, P. (1996). The influence of planning and task type on second
language performance. Studies in Second Language Acquisition, 18, 299–323.
Foster, P., Tonkyn, A., & Wigglesworth, G. (2000). Measuring spoken language.
Applied Linguistics, 21(3), 354–375.
Givon, T. (1985). Function, structure, and language acquisition. In D. Slobin (Ed.), The
crosslinguistic study of language acquisition (Vol. 1, pp. 1008–1025). Hillsdale, NJ:
Lawrence Erlbaum.
Harris, C., & Bates, E. (2002). Clausal background and pronominal reference: A
functionalist approach to c-command. Language and Cognitive Processes, 17(3),
237–269.

Language Learning 59:4, December 2009, pp. 866–896 890

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

Heaton, J. B. (1966). Composition through pictures. Essex, UK: Longman.

Hoey, M. (2005). Lexical Priming: A new theory of words and language. London:
Routledge.
Iwashita, N., McNamara, T., & Elder, C. (2001). Can we predict task difficulty in an
oral proficiency test? Exploring the potential of an information-processing approach
to task design. Language Learning, 51(3), 401–436.
Johnson, D. (1977). The TOEFL and domestic students: Conclusively inappropriate.
TESOL Quarterly, 11, 79–86.
Jones, L. (1980). Notions in English. Cambridge: Cambridge University Press.
Kormos, J. (2006). Speech production and second language acquisition. London:
Routledge.
Levelt, W. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT
Press.
Levelt, W. (1993). Language use in normal speakers and its order. In G. Blanken, H.
Dittman, H. Grimm, J. Marshal, & C. Wallesch (Eds.), Linguistic disorders and
pathologies (pp. 1–15). Berlin: de Gruyter.
Malvern, D., & Richards, B. (2002). Investigating accommodation in language
proficiency interviews using a new measure of lexical diversity. Language Testing,
19(1), 85–104.
Matthiessen, C., & Thompson, S. (1988). The structure of discourse and
subordination. In J. Haiman & S. Thompson (Eds.), Clause combining in grammar
and discourse (pp. 95–120). Philadelphia: Benjamins.
Mehnert, U. (1998). The effects of different length of time for planning on second
language performance. Studies in Second Language Acquisition, 20(1),
83–108.
Navon, D. (1989). The importance of being visible: On the role of attention in a mind
viewed as an anarchic intelligence system. European Journal of Cognitive
Psychology, 1, 191–238.
Neumann, O. (1996). Theories of attention. In O. Neumann & A. Sanders (Eds.),
Handbook of perception and action: Vol. 3. Attention (pp. 389–446). San Diego,
CA: Academic Press.
Ochs, E. (1979). Planned and unplanned discourse. In T. Givon (Ed.), Syntax and
semantics (pp. 12–31). New York: Academic Press.
Oller, J., & Conrad, C. (1971). The cloze technique and ESL proficiency. Language
Learning, 21, 183–195.
Ortega, L. (1999). Planning and focus on form in L2 oral performance. Studies in
Second Language Acquisition, 21, 109–148.
Pawley, A., & Syder, F. (1983). Two puzzles for linguistic theory: Native-like selection
and native-like fluency. In J. Richard & R. Schmidt (Eds.), Language and
communication (pp. 191–226). London: Longman.
Rampton, B. (1990). Displacing the “native speaker”: Expertise, affiliation, and
inheritance. ELT Journal, 44(2), 97–101.

891 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

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(1),
27–55.
Robinson, P., & Gilabert, R. (2007). Task complexity, the Cognition Hypothesis and
second language learning and performance. International Review of Applied
Linguistics, 45, 161–76.
Sacks, H., Schegloff, E., & Jefferson, G. (1974). A simplest systematics for the
organization of turn-taking in conversation. Language, 50(4), 696–734.
Samuda, V. (2001). Guiding relationships between form and meaning during task
performance: The role of the teacher. In M. Bygate, P. Skehan, & M. Swain (Eds.),
Researching pedagogic tasks second language learning, teaching and testing (pp.
119–141). Harlow, UK: Longman.
Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human
information processing: I. Detection, search and attention. Psychological Bulletin,
84, 1–66.
Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human information
processing: II. Perceptual learning, automatic attending, and a general theory.
Psychological Review, 84, 127–190.
Skehan, P. (1998). A cognitive approach to language learning. Oxford: Oxford
University Press.
Skehan, P., & Foster, P. (1997). Task type and task processing conditions as
influences on foreign language performance. Language Teaching Research, 1(3),
185–212.
Skehan, P., & Foster, P. (1999). The influence of task structure and processing
conditions on narrative retelling. Language Learning, 49(1), 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
Press.
Skehan, P., & Foster, P. (2009). Second language learning, acquisition and
performance: A task-based perspective. Unpublished manuscript.
Swain, M., & Lapkin, S. (2001). Focus on form through collaborative dialogue:
Exploring task effects. In M. Bygate, P. Skehan, & M. Swain (Eds.), Researching
pedagogic tasks: Second language learning, teaching and assessment. London:
Pearson Education.
Swan, M., & Walter, C. (1990). The new Cambridge English Course, Practice 2.
Cambridge: Cambridge University Press.
Stricker, L. J. (2002). The performance of native speakers of English and ESL speakers
on the computer-based TOEFL and GRE general test. Princeton, NJ: ETS.

Language Learning 59:4, December 2009, pp. 866–896 892

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

Tabachnic, B. G., & Fidell, L. S. (1996). Using multivariate statistics. New York:
Harper Collins.
Tavakoli, P., & Foster, P. (2008). Task design and second language performance:
The effect of narrative type on learner output. Language Learning, 58(2),
439–473.
Tavakoli, P., & Skehan, P. (2005). Strategic planning, task structure and performance
testing. In R. Ellis (Ed.), Planning and task performance in a second language (pp.
239–277). Amsterdam: Benjamins.
VanPatten, B. (1990). Attending to form and content in the input: An experiment in
consciousness. Studies in Second Language Acquisition, 12, 287–301.
Weinert, R. (1995). The role of formulaic language in second language acquisition: A
review. Applied Linguistics, 16(2), 181–205.
Whiteson, V. (1979). A study of proficiency and linguistic variety of non-native EFL
teachers. Unpublished doctoral dissertation, Indiana University, Bloomington.
Wigglesworth, G. (2001). Influences on performance in task-based oral assessments.
In M. Bygate, P. Skehan, & M. Swain (Eds.), Researching pedagogic tasks: Second
language teaching, learning and testing (pp. 186–209). London: Longman.

893 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

Appendix

Figure A1 Journey task (Jones, 1980). Reprinted with permission.

Language Learning 59:4, December 2009, pp. 866–896 894

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

Figure A2 Picnic task (Heaton, 1966). Reprinted with permission.

Figure A3 Walkman task (Swan & Walter, 1990). Reprinted with permission.

895 Language Learning 59:4, December 2009, pp. 866–896

Foster and Tavakoli Effects on Complexity, Fluency, and Lexical Diversity

Figure A4 Football task (Heaton, 1966). Reprinted with permission.

Language Learning 59:4, December 2009, pp. 866–896 896