Investigating the Relationships Between Vocabulary and Clause-internal Pauses
and its Development in L2 Speech
Waseda University, Japan/ University of Hawai'i at Mānoa, USA
The current study examines 1) relative proportions of pauses
related to lexical or other language problems, 2) their
longitudinal changes, and 3) actual examples of the pauses and
lexical items. To address these issues, in-class speech by three
undergraduate students in English-Medium Instruction (EMI)
were examined. It was found that 1) lexical pauses occurred
more in dialogues, 2) they profiled different paths and the
relationships varied over time, and 3) some categories of lexical
items were involved. The paper further discusses the nature of
L2 fluency in dialogue, and trajectories of the constructs
observed from the participants.
In the second language (L2) research tradition, fluency is one of
the dimensions of speech (i.e., the narrow sense of fluency;
Lennon, 1990), and becoming a fluent speaker entails a
multifaceted, non-linear development of attaining an optimal
rate of speech with fewer hesitations and repairs. Previous
research has been investigating how L1 and L2 fluency differs,
which aspects of fluency determine perceptions of fluency and
what are the causes of dysfluencies. While it is strongly
believed that the speed of speech delivery and pausing
phenomena are linked to the degree of fluent speech in general,
a recent body of research has suggested that not only these
aspects but location of pauses would be also crucial in L2
fluency (Foster & Skehan, 2012; Pawley & Syder, 1983;
Tavakoli, 2011; Witton-Davies, 2014).
Meanwhile, another line of research reports that pausing
phenomena in L2 speech are more closely related to vocabulary
than other linguistic problems, suggesting that lexical retrieval
can be one of the biggest obstacles in L2 speech fluency (de
Jong, 2016; Hilton, 2008; Witton-Davies, 2014). These studies
have generally employed cross-sectional design to find any
statistical relationships between the two constructs; however,
they do not provide how the two are related in long term.
Moreover, they do not examine the pausing phenomena in
relation to other constructs such as complexity in lexical and
syntactic domains, which are reported to affect the development
and performance in fluency domain. The current study therefore
delves into the relationships between pausing and vocabulary
items, and how the relationships change in closer detail. To
achieve this purpose, it employed a one-year longitudinal
design consisting of three EFL learners in a Japanese university.
To begin with, the following sections review issues related to
the nature of L2 fluency and its development in order: a) fluency
and L2 speech, b) determinants of L2 fluency, and c)
relationships between vocabulary and fluency.
2.1.!Fluency and L2 research
Traditionally, fluency is regarded as one of the three
multifaceted dimensions of second langauge (L2) speech
production with the other two dimensions of complexity and
accuracy (CAF; Ellis & Barkhuizen, 2005; Housen, Kuiken, &
Vedder, 2012; Skehan, 2009). Fluency generally concerns the
degree of automaticity to mobilize the L2; therefore, increased
fluency is often associated with faster speech rate (i.e., speed
fluency), fewer pauses (i.e., breakdown fluency), and/or fewer
repetitions or self-corrections (i.e., repair fluency) (Segalowitz,
2010; Skehan, 2009).
Research on L2 fluency can be classified into two types.
One strand of research, from Task-based Language Teaching
(TBLT) framework (see Robinson, 2011; Skehan, 2009), has
examined fluency as dependent variables with independent
effects of task manipulations in characteristics and/or
conditions, or investigated supportive or competitive
relationships among the different dimensions of CAF in
learning trajectories (Larsen-Freeman, 2006). One important
implication for this study is that fluency should be captured with
other constructs of speech measures at the same time.
Another line of research has been investigating the nature
of L2 fluency (see de Jong, 2016; Kahng, 2014). The latter
approach includes agendas such as temporal differences
between L1 and L2 speech (see de Jong, 2016; Tavakoli, 2011),
correlates with L2 proficiency (see de Jong, Steinel, Florijn,
Schoonen, & Hulstijn, 2013; Kahng, 2014; Witton-Davies,
2014), and possible causes of disfluencies (see Hilton, 2008).
These studies have made unique contributions to the
understanding of L2 fluency, providing a picture for what
characterizes L2 fluency (for a comprehensive review see
Segalowitz, 2010). The foci of this study are two-folds in that
while it examines the nature of relationships between pauses
(i.e., clause internal pauses as explained below) and vocabulary
in dialogic speech, it also seeks to provide a view of long term
changes of the relationships including other relevant
dimensions of speech production (i.e., syntactic and lexical
complexity). Now we turn to the issue of what characterizes L2
2.2.!Possible determinants of L2 fluency
Researchers normally study L2 oral fluency from three different
levels of phenomena. The first level concerns the cognitive
capacity or efficiency of a given learner for L2 processing (i.e.,
cognitive fluency). The second level, utterance fluency, refers
to the temporal manifestation of fluency in utterance resulting
from the interplay of contextual factors and cognitive operation
(e.g., words per minutes, number of pauses), consisting of the
three dimensions, i.e., speed, breakdown and repair fluency (see
above). The last, perceived fluency, is related to “the inferences
listeners make about a speaker’s cognitive fluency”
(Segalowitz, 2010, p. 48). Referring to this framework,
researchers have attempted to find a set of utterance fluency
measures that correlates with either perceived fluency or
general proficiency measures, which can be in turn used to infer
the cognitive fluency of the speaker (Segalowitz, 2010).
To date, speed and breakdown fluency are found to be the
important indices for perceived fluency or general proficiency
measures (Bosker, Pinget, Quene, Sanders, & de Jong, 2012; de
Jong et al., 2013; Derwing, Rossiter, Munro, & Thomson, 2004;
Kormos & Dénes, 2004; Rossiter, 2009). Kormos and Dénes
(2004) explored the correlates between fluency measures and
teacher judgments on speech samples (i.e., perceived fluency).
They found that speed fluency measures (e.g., speech rate, the
mean length of utterance) predicted fluency scores better than
breakdown measures (e.g., number of pauses). This view was
supported by more recent studies by de Jong et al. (2013), who
reported better association, though not strong, between
cognitive fluency and speed fluency than breakdown and repair
fluency. Although these studies do not provide favorable
evidence for breakdown fluency, other research does suggest
the relationships. For instance, Derwing et al. (2004) reported a
combined contribution of speed and breakdown fluency to
perceived fluency ratings in beginner learners’ speech.
Similarly, Rossiter (2009) concurred with Derwing et al. (2004)
by showing moderate correlations between both speed and
pause measures and fluency ratings. Furthermore, a recent study
on fluency by L2 Dutch speakers showed that breakdown
measures followed by speed measures were associated with
perceived fluency ratings, illuminating the possibility that
pauses are equally sensitive to the perception of fluency. In sum,
perceived L2 fluency can be largely influenced by speed and
breakdown, not by repair fluency, in speech production.
In fact, researchers have been considering that breakdown
fluency, especially pause location, is crucial in L2 speech
production for many years (Pawley & Syder, 1983; Tavakoli,
2011). These studies propose a view that L2 speakers are likely
to produce more pauses within meaningful boundaries (e.g.,
clause, constituents) than L1 speakers. On this particular issue,
de Jong (2016) has recently provided further evidence that the
temporal differences between L1 and L2 speakers can be
identified in frequency of pauses within utterance, but not
between utterances. This observation is further supported by
Kahng (2014), who reported a significant difference in silent
pause rate within a clause across various proficiency groups of
L2 speakers. Therefore, it is plausible to claim that location of
pauses plays a critical role in L2 oral fluency.
Although previous findings on possible determinants are
theoretically and empirically sound, we need to also consider
the effect of the nature of communication and nature of task
types on fluency (Derwing et al., 2004; Foster & Skehan, 1996;
Tavakoli, 2016). It is well suggested that dialogue can push
speakers to produce more fluent utterance than monologues
(Tavakoli, 2016; Witton-Davies, 2014). This theoretical
assumption is one of the motivations for this study, which
investigates the breakdown fluency in dialogic speech.
2.3.!Breakdown fluency and vocabulary
As mentioned above, several studies have investigated possible
causes of disfluency in L2 speech production (Hilton, 2008;
Witton-Davies, 2014). At the theoretical level, it is assumed that
breakdowns have to do with problems in lexical retrievals
(Skehan, 2009). Witton-Davies (2014) showed correlations
between breakdown fluency and vocabulary, suggesting an
empirical association between the two. In a more detailed study,
Hilton (2008) examined the link between L2 fluency and
vocabulary. He coded each breakdown according to possible
causes of the breakdown (e.g., lexical search, syntactic error,
morphological error), and found that 78% of the pauses were
classified into problems in lexical domain. This evidence shows
that breakdown fluency and vocabulary knowledge are closely
linked together in L2 speech production.
To summarize, research on L2 fluency generally provides a
view that speed and breakdowns influence the perception of L2
fluency, and breakdown fluency can be linked to vocabulary
retrieval processes during the speech production. However, few
research has been reported in relative frequency of pauses in
dialogic speech. Also they tend to be cross-sectional studies
looking at the rather stable picture of the relationships.
Therefore, more research will be necessary in terms of a) the
effect of dialogic speech on pausing phenomena, b) the long
terms changes of the breakdown fluency, and c) the qualitative
nature of the relationships between breakdowns and vocabulary
items in the speech production. In this study, therefore, three
research questions (RQs) below were investigated.
1)!In dialogic speech, which type of pauses occur more
than the other, pauses associated with lexical problems
or pauses with problems in other linguistic domains?
2)!How do the relationships among pauses and other
linguistic measures change overtime?
3)!What are typical examples of the relationships
between the lexical items and pauses?
The participants of this study were three undergraduate students
who majored in English language and literature at a university
in Japan. All the three students were from a target English
Medium Instruction (EMI) course, defined as a non-language
course taught in English (Hellekjær, 2010). They were
described as intermediate according to the TOEFL iBT score
(See table 1).
Table 1. The description of the participants
Haruka, a pseudonym, was a female senior student, who had
experience one-year study abroad one year prior to the data
collection. She took the EMI course offered by the same
instructor the previous year, so she was used to the instructional
style. Ken was also a senior student in the EMI, and had also
been in this EMI class from the previous year. Takahiro was a
male junior student in the EMI and new to the class taught by
the instructor. All the students received formal English
language teaching in Japanese junior/senior high schools, and
passed the entrance examination of the university.
3.2.!The EMI class
The EMI class consisted of various academic tasks, including
reading assignments (around 15 pages per weekly class),
weekly open-ended quizzes, two students’ presentations, and
discussions in small groups. The procedure of each class
obtained by researcher’s classroom observation is provided in
Figure 1. The basic structure of the class remained the same
throughout the year. Each week, students were divided
randomly by the instructor. They were usually supposed to form
4 small groups of from four to five people. This group was the
basis of group discussion, which was the data collection site of
These group discussions were prompted by the questions by
each student presenter on the issues in the assigned reading they
were supposed to read for the classroom (see Figure 1). Since
the participants read the readings, and answered to the quiz, they
were likely to have opportunities to think about the issues and
formulate their opinion beforehand.
The current study focused on the small group discussions in the
current EMI class. Upon informed consent by all the students in
the class, audio recorders were put on each table during the
classes. The data collection was conducted at three time points
of May (the second month of the first semester), July (the last
month of the first semester) and January (the last month of the
academic year). Audio data were later transcribed by the author
(around five hours of recordings).
Figure 1: The procedure of the target EMI class
To identify clause boundaries of the speech production,
Analysis of Speech Unit (ASU; Foster, Tonkyn, &
Wigglesworth, 2000) was used. Although the previous study
(Hilton, 2008) used T-unit as the sentential unit, this study
identified two advantages for using ASU. First, ASU can deal
with utterances that does not include finite verbs, which is
frequent in spoken language. Second, ASU is based on both
psycholinguistic and syntactic unit of speech production, which
enables us to identify clause boundaries. These strengths of
ASU allow us to classify pauses in a more fine-grained way.
The segmentation of ASU shows that the three speakers
produced a total of 532 ASU (the amount of ASU produced by
individual participants are shown in Table 4 with other
In order to identify pauses in the speech, the minimum length
of them must be determined first. Previous studies vary on this
issue depending on their purposes of research. Studies focusing
on task-based L2 production typically set minimum pause
lengths between 200ms to 400ms (see Kahng, 2014). The
decision is based on the finding that L1 speakers frequently
produce pauses shorter than 400ms (Deese, 1980). However,
the current study is not intended to measure pauses on such a
strict criteria, but seeks to find possible breakdowns in L2
speech as in Hilton (2008). Instead, to determine a valid pause
length, a unique aspect of this study must be considered. That
is, the present study investigates the dialogic speech production,
which may shorten the length of pauses in two possible ways.
First dialogue involves the intervention of the interlocutors at
certain duration of breakdowns (e.g., 2 sec.; Rieger, 2003).
Second, dialogic speech allows speakers to circumvent
breakdowns caused by their limited linguistic resources, using
paraphrases or other strategies when compared to narrative
tasks (Crowther, Trofimovich, Isaacs, & Saito, 2015). Due to its
exploratory nature of this study, pause length was determined
based on both the research findings mentioned above and actual
data that the current study worked on (see Table 2).
Table 2. The raw frequency of pauses by pause lengths
0.3 sec.< 1sec.
> 1 sec.
> 2 sec.
> 3 sec.
Based on the existing literature and the data, pause length
for the current analysis was set as one second for RQ 1 and 2,
and two seconds for RQ3. The rationale behind this decision is
the fact that setting criteria as two seconds (Rieger, 2003) does
not provide a sufficient number of pauses in RQ1 and 2.
Once ASU and locations of pauses were identified, all pauses
within clauses were selected and manually coded according to
a coding scheme adapted from Hilton (2008). This coding
scheme enables us to classify the possible domains of linguistic
problems that may cause dysfluency (see Table 3). In the
process of coding, another possible code emerged due to the fact
that participants produced pauses which were not associated
with any of the categories by Hilton (2008). These pauses were
considered as “No error,” which did not seem to involve any
lexical retrievals nor erroneous utterance.
Table 3. The coding scheme adapted from Hilton (2008)
3.4.4.!Other linguistic measures
In order to answer RQ2, two complexity measures (i.e.,
syntactic and lexical) were included in the analysis. This was
because previous studies suggested that research took into
account the possible interactions of the dimension of speech
production (see above). The current study included “mean
length of ASU” as global syntactic complexity measure (Norris
& Ortega, 2009), and “SUBTLEXus Range CW Log” as a
A pause involves error in
A pause involves direct
request for words
A pause involves (infrequent)
words or phrases that might
cause retrieval problems.
A pause involves errors in
morphology (e.g., -ed, -s, -ing)
A pause involves errors in
A pause involves errors in
word order, or syntax
A pause involves syntactic
reformulations and repetitions
No Error (NE)
A pause involves no error nor
none of the above
lexical complexity (i.e., sophistication; both quality and
quantity of vocabulary use) measure calculated with Tool for
the Automatic Analysis of LExical Sophistication (TAALES)
(TAALES; Kyle & Crossley, 2015). “SUBTLEXus Range CW
Log” is one of the five measures in the result of the stepwise
multiple regression analysis for speaking scores in Kyle and
Crossley (2015), which can reflect the sophistication of
vocabulary use in speaking. Specifically, this index is a
logarithmic transformation of a range index (i.e., the number of
different text files which include a given word in a given corpus,
which is SUBTLEXus corpus). Thus, the smaller value in this
index means that the text contains more sophisticated
vocabulary in the speech, since vocabulary used in fewer text
files in corpus can be considered more domain specific (Kyle &
Crossley, 2015). Note that, for the sake of simplicity in
presentation, this index will be inverted in order to compare it
with other indices.
To answer RQ1, the proportion of pauses (> 1 second) were
compared between pauses associated with lexis and ones with
the rest of linguistic domains. Since the data did not satisfy the
assumption of independent samples (i.e., pauses are produced
by three participants), the data were not statistically compared.
Instead, the current analysis focused on the comparison of the
proportions of pauses in the two domains by the participants
across time points. To address RQ2, indices of frequency of
pauses in both domains, and complexity measures in lexis and
syntax were plotted in line plots on individual bases. To enable
the comparison between different values, the current analysis
required transformations of the values into z-scores. This way,
we were able to look at the general tendencies of the
relationships between them over time. As for the third RQ, the
current study provides the readers with typical cases of pauses
(> 2 seconds) that are associated with lexical pauses.
As for RQ1, lexical pauses were more frequent than pauses that
was related to other linguistic domains on average (61.23%
versus 38.77%), suggesting that pauses related to vocabulary
may still occur more in dialogic speech. However, Table 4 gives
us the results from a different perspective. It revealed that only
one of our participants, Haruka, constantly produced more
pauses related to lexical problems than ones in other domains.
The other two participants, Ken and Takahiro, rather produced
more pauses in linguistic domains in May and January.
Therefore, considering the relatively small variation, the
conclusion should not be drawn as to the proportion of pauses;
rather it revealed more interesting variation within and between
participants in this study.
To address RQ2, frequency of pauses in both categories
were plotted in line plots individually with other constructs of
lexical and syntactic complexity measures. Figures 2 to 4 shows
the relationships among the constructs over time. As noted
above, the indices of pauses, and complexity measures in lexis
and syntax are plotted in these figures to see general trajectories
of our participants (see Table 4 for raw scores). The two lines
show pauses in either domain, and two dashed lines present the
complexity measures in syntax and lexis. Greater z-scores mean
more complex or fluent in each index.
The results also showed the individual variability over time.
Although Haruka showed relatively stable variations compared
to the other two participants, her performance still fluctuated
across the occasions. We can see the inverse relationships
between pauses related to vocabulary and lexical sophistication
value of vocabulary she produced. The pauses in other domains
and syntactic complexity showed competitive relationships; one
cannot increase if the other scores higher.
Table 4: The descriptive statistics
Figure 2: The individual trajectory by Haruka
Figure 3: The individual trajectory by Ken
Figure 4: The individual trajectory by Takahiro
May July Ja n
Haruka (Developmental trajectory)
May July Ja n
Ken (Developmental trajectory)
May July Ja n
Takahiro (Developmental trajectory)
The results by Ken and Takahiro showed more variability
across time points. It seemed that lexical sophistication and
pauses related to vocabulary had competitive relationships,
whereas syntactic complexity and pauses related to other
linguistic domains could be supportive. Takahiro presented a
more difficult picture in the relationships. Seemingly, he
developed both his lexical sophistication and breakdown
fluency linked to vocabulary over time without any tradeoffs,
and the relationship related to grammar seems to be supportive
as well. In sum, the current study gives us a view that there are
some distinct patterns for the developmental pathways by the
participants in terms of the constructs investigated.
For RQ3, the pauses ( > 2 sec.) was examined using
AntConc (Anthony, 2014), and concordance line was
qualitatively examined to find possible relationships between
the pauses and lexical items. Three emerging relationships were
observed, namely, frequent words, infrequent words, and
formulaic sequences (FS). Table 5 shows examples of each
category. Frequent vocabulary tends to be general words (below
3000 word-level in BNC-COCA) such as “basic,” “method,”
“problem,” etc. Infrequent words include “valid,” “tolerant,”
etc. Formulaic Sequences (FS) such as “it depends on the” tend
to be involved before the pauses. Other examples of vocabulary
items related to pauses are technical terms that is covered in the
textbooks, such as “communicative approach,” “universal
Table 5: Illustration of the vocabulary items that
Environment is the (2.15 sec) basic.
But, some learners tend to confuse or mix this
(3.08 sec) image about the production and
English introducing class (1.05 sec) the
giving model is (2.03 sec) valid to CD teach
So it depends on the (2.37 sec) students'
personality or characteristics.
Motivated by the accumulated studies on L2 fluency (Bosker et
al., 2012; de Jong, 2016; Hilton, 2008; Tavakoli, 2011, 2016;
Witton-Davies, 2014), this study made an attempt to provide
further account for the nature of breakdown fluency and the
development of those features. The RQs asked in this study
were 1) the relative proportion of pauses related to vocabulary
versus other domains, 2) the longitudinal relationships between
the two types of pauses and other dimensions, and 3) the
instances which can be attributed as pauses in vocabulary
The result of the first RQ found that the proportion of pauses
related to vocabulary were larger than the pauses related to
other domains, which supported the previous studies (Hilton,
2008; Witton-Davies, 2014). However, the proportion varied
depending on the participants on different occasions. This is
partly attributed to the context of the current study, where
pauses are examined in free speech rather than input based tasks
such as video or picture narrations (Crowther et al., 2015;
Hilton, 2008). Another factor we need to take into account is
the dialogic nature of this study, which may produce shorter and
less frequent pauses than monologic tasks (Tavakoli, 2016). As
predicted from these previous studies, these two factors might
function to reduce the lengths and proportion of pauses in
general, which may contribute to the current results.
The result of the second question (RQ2) showed the
dynamic (competitive or supportive) relationships between the
variables (Larsen-Freeman, 2006). Also, it may be possible to
suggest that there are some individual patterns of the
relationships. For instance, Haruka and Ken, on the one hand,
showed competitive relationships between the frequency of
pauses in vocabulary domains and a range index of lexical use
in their speech over time. This can be interpreted as diachronic
tradeoff between a part of breakdown fluency and lexical
sophistication (Housen et al., 2012). Although the source of the
fluctuation cannot be observed in this study, they seem to
prioritize, consciously or unconsciously, one dimension of the
production over the other. On the other hand, Takahiro rather
showed supportive relationships between those indices (i.e.,
frequency of pauses was reduced at the same time as he
produced more sophisticated vocabulary over time). One thing
that can be assumed from the finding is that Takahiro was the
one who experienced the EMI for the first time, suggesting that
he was gradually building up the fluency towards the end of the
course (see Figure 5). The decrease in fluency and complexity
in other linguistic domains from July to January and the overall
high score for lexical sophistication may also indicate that he
was relatively oriented towards lexically complex production.
It is claimed from the overall results that constructs that were
developed and increased at one time can be suppressed because
of the development in other constructs (see the relationships
between fluency and complexity in Figure 2 – 4), which support
the view of diachronic tradeoff in L2 development (Housen et
As for RQ3, potential causes of the pauses were found in
frequent words, infrequent words and FSs. First and most
counterintuitively, frequent words (e.g., method, basic, image)
were associated with the relatively longer pauses (>2 sec). This
is assumed to be executions of strategic competence by the
participants, where they compensate for an imminent
breakdown by using generic words that can be easily accessed
by the participants rather than sticking to the specific,
elaborated words that takes time to retrieve (Segalowitz, 2010).
Infrequent words also occur after the pauses, which can be
attributed to the lexical retrieval processing of those words
(Skehan, 2009). Third, FSs sometimes involve pauses after
them. Generally, FSs are considered as fluency builders to
enable speakers to buy time to conceptualize and formulate the
following utterances (Serrano, Stengers, & Housen, 2015;
Wray, 2002). It might be suggestive from the result of the close
scrutiny, however, that FSs cannot always be fully incorporated
in the utterance as fluency builder. Although they may facilitate
the natural flow of the speech within those sequences, the
retrieval process of the following core words may be another
issue. Learners will need to learn how to integrate the FS in a
way that s/he can facilitate the retrieval of the single core
vocabulary that fills the slots in the FSs.
This study investigated the relationships between the
breakdown fluency within clauses and vocabulary use in
dialogic L2 speech. It found that 1) proportion of lexical pauses
are reduced in dialogic speech, 2) dynamic relationships
between the variables on individual bases, and 3) frequent
words and FSs can be associated with pauses within a clause
(see above). However, several limitations should be considered.
First, relationships observed here are totally exploratory. Since
the data does not satisfy the independent assumption of Chi-
square test, the frequency comparison was not performed.
Second, although this study sought to investigate the
developmental paths experienced by three participants, the
points of data collection was rather limited, which may limit the
scope of this study to the assumption of general tendencies.
Further research should examine the trajectory in more fin-
grained way. Nevertheless, this study can one of the few
attempts to document potential dynamic relationships among
variables of interests (Larsen-Freeman & Cameron, 2008;
Verspoor, De Bot, & Lowie, 2011), which can be further
developed to particular theorizations in L2 development
modulated by contextual factors.
Anthony, L. (2014). AntConc (Version 3.4.3). Tokyo, Japan:
Waseda University. Retrieved from
Bosker, H. R., Pinget, A. F., Quene, H., Sanders, T., & de Jong,
N. H. (2012). What makes speech sound fluent? The
contributions of pauses, speed and repairs. Language
Testing, 30(2), 159-175. doi:10.1177/0265532212455394
Crowther, D., Trofimovich, P., Isaacs, T., & Saito, K. (2015).
Does a Speaking Task Affect Second Language
Comprehensibility? The Modern Language Journal,
99(1), 80-95. doi:10.1111/modl.12185
de Jong, N. H. (2016). Predicting pauses in L1 and L2 speech:
the effects of utterance boundaries and word frequency.
International Review of Applied Linguistics in Language
Teaching, 54(2). doi:10.1515/iral-2016-9993
de Jong, N. H., Steinel, M. P., Florijn, A. F., Schoonen, R., &
Hulstijn, J. H. (2013). Linguistic skills and speaking
fluency in a second language. Applied Psycholinguistics,
34(05), 893-916. doi:10.1017/s0142716412000069
Deese, J. (1980). Pauses, prosody, and the demands of
production in language. In H. W. Dechert & M. Raupach
(Eds.), Temporal variables in speech: Studies in honour of
Frieda Goldman-Eisler (pp. 69-84). The Hague: Mouton
Derwing, T. M., Rossiter, M. J., Munro, M. J., & Thomson, R.
I. (2004). Second Language Fluency: Judgments on
Different Tasks. Language Learning, 54(4), 655-679.
Ellis, R., & Barkhuizen, G. P. (2005). Analysing learner
language: Oxford University Press, USA.
Foster, P., & Skehan, P. (1996). The influence of planning and
task type on second language performance. Studies in
Second Language Acquisition, 18(03), 299-323.
Foster, P., & Skehan, 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. 199-220).
Amsterdam ; Philadelphia: John Benjamins.
Foster, P., Tonkyn, A., & Wigglesworth, G. (2000). Measuring
spoken language: a unit for all reasons. Applied
Linguistics, 21(3), 354-375. doi:10.1093/applin/21.3.354
Hellekjær, G. O. (2010). Lecture comprehension in english-
medium higher education. Hermes, 45, 11-34.
Hilton, H. (2008). The link between vocabulary knowledge and
spoken L2 fluency. Language Learning Journal, 36(2),
Housen, A., Kuiken, F., & Vedder, I. (2012). Dimensions of L2
performance and proficiency : complexity, accuracy and
fluency in SLA. Amsterdam ; Philadelphia: John
Kahng, J. (2014). Exploring Utterance and Cognitive Fluency
of L1 and L2 English Speakers: Temporal Measures and
Stimulated Recall. Language Learning, 64(4), 809-854.
Kormos, J., & Dénes, M. (2004). Exploring measures and
perceptions of fluency in the speech of second language
learners. System, 32(2), 145-164.
Kyle, K., & Crossley, S. A. (2015). Automatically Assessing
Lexical Sophistication: Indices, Tools, Findings, and
Application. TESOL Quarterly, n/a-n/a.
Larsen-Freeman, D. (2006). The Emergence of Complexity,
Fluency, and Accuracy in the Oral and Written Production
of Five Chinese Learners of English. Applied Linguistics,
27(4), 590-619. doi:10.1093/applin/aml029
Larsen-Freeman, D., & Cameron, L. (2008). Complex systems
and applied linguistics: Oxford University Press.
Norris, J. M., & Ortega, L. (2009). Towards an Organic
Approach to Investigating CAF in Instructed SLA: The
Case of Complexity. Applied Linguistics, 30(4), 555-578.
Pawley, A., & Syder, F. (1983). Two puzzles for linguistic
theory: Nativelike selection and netivelike fluency. In J. C.
Richards & R. W. Schmidt (Eds.), Language and
communication (pp. 191-226). London: Longman.
Rieger, C., L. (2003). Disfluencies and hesitation strategies in
oral L2 tests. Paper presented at the DiSS’03, Disfluency
in Spontaneous Speech Workshop, Göteborg University,
Robinson, P. (2011). Second language task complexity :
researching the cognition hypothesis of language learning
and performance. Amsterdam ; Philadelphia: John
Benjamins Pub. Co.
Rossiter, M. J. (2009). Perceptions of L2 Fluency by Native and
Non-native Speakers of English. Canadian Modern
Language Review, 65(3), 395-412.
Segalowitz, N. (2010). Cognitive bases of second language
fluency. New York: Routledge.
Serrano, R., Stengers, H., & Housen, A. (2015). Acquisition of
formulaic sequences in intensive and regular EFL
programmes. Language Teaching Research, 19(1), 89-
Skehan, P. (2009). Modelling Second Language Performance:
Integrating Complexity, Accuracy, Fluency, and Lexis.
Applied Linguistics, 30(4), 510-532.
Tavakoli, P. (2011). Pausing patterns: differences between L2
learners and native speakers. ELT Journal, 65(1), 71-79.
Tavakoli, P. (2016). Fluency in monologic and dialogic task
performance: Challenges in defining and measuring L2
fluency. International Review of Applied Linguistics in
Language Teaching, 54(2). doi:10.1515/iral-2016-9994
Verspoor, M., De Bot, K., & Lowie, W. (2011). A dynamic
approach to second language development: Methods and
techniques (Vol. 29): John Benjamins Publishing.
Witton-Davies, G. (2014). The Study of Fluency and its
Development in Monologue and Dialogue. Unpublished
doctoral dissertation. Lancaster University. Lancaster,
UK. Retrieved from
Wray, A. (2002). Formulaic language and the lexicon.
Cambridge ; New York: Cambridge University Press.