The Relationship between L1 Fluency and L2 Fluency in Writing: A Mixed 
Methods Study of the Writing Fluency of the Finnish Learners of English by 
Using Keystroke Logging Data 
 
 
 
 
 
 
 
 
 
 
 
 
   Annastiina Romo  
                         MA Thesis 
                                             English, Degree Programme for Language Teaching and Learning
                                  School of Languages and Translation Studies 
Faculty of Humanities 
University of Turku 
January 2020 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
The originality of this thesis has been checked in accordance with the University of Turku 
quality assurance system using the Turnitin OriginalityCheck service. 
  
UNIVERSITY OF TURKU 
School of Languages and Translation Studies / Faculty of Humanities 
 
ROMO, ANNASTIINA: The Relationship between L1 Fluency and L2 Fluency in 
Writing: A Mixed Methods Study of the Writing Fluency of the Finnish Learners of 
English by Using Keystroke Logging Data  
 
MA Thesis, 41 pages. 
English, Degree Programme for Language Teaching and Learning  
January 2020 
 
                                             
 
Second language (L2) fluency has been a widely researched topic in the field of SLA for 
many years. However, it is only recently that researchers have become interested in the 
possible effects of the first language (L1) on the L2 fluency, which means that these 
connections are still relatively poorly understood. Thus, the aim of this thesis was to 
explore the relationship between L1 and L2 fluency, focusing especially in writing 
fluency, which has been a neglected area of research compared to spoken fluency studies. 
The research employed a mixed-methods approach to investigate the issue, where 
the quantitative analysis of different fluency measures was complemented with a more 
in-depth qualitative analysis of the fluency features of the essays. The data consisted of 
essays written by 16 Finnish upper-secondary school students in their L1 (Finnish) and 
L2 (English) by means of a keystroke recording tool ScriptLog. The samples were 
examined with the help of various writing fluency measures, which were based on 
participants’ pausing and revision behaviour. The essays were then analysed qualitatively 
by using a writer profile categorization. The results revealed positive statistically 
significant correlations on three of the measures: the number of pauses, the mean length 
of pauses and P-Burst. The regression analysis further supported the findings by 
indicating that the L2 fluency measures could be predicted from the equivalent L1 
measures to some extent. The findings of the qualitative analysis showed that seven out 
of 16 participants shared similar composing processes in their L1 and L2. Thus, the 
findings together suggest that there is a link between L1 and L2 writing fluency to some 
degree, which should be acknowledged in future studies which focus on different factors 
affecting L2 fluency. In addition, the results of the present study can be used in teaching 
contexts by facilitating learners’ L2 writing and as a help of L2 fluency assessment. 
 
 
Keywords: writing fluency, ScriptLog, L2 fluency, L1 fluency, L2 English, writer profile
  
 
Table of Contents 
1 Introduction ........................................................................................................................ 1 
2 L2 Fluency .......................................................................................................................... 3 
2.1 L2 Writing Fluency .......................................................................................................... 5 
2.2 Writer Profiles .................................................................................................................. 9 
3 The Relationship between L1 and L2 Fluency ................................................................. 13 
4 Research methodology ..................................................................................................... 19 
4.1 The Aim of the Research ................................................................................................ 19 
4.2 Data Collection and Processing ...................................................................................... 20 
4.3 Methods of Analysis ....................................................................................................... 22 
5 Fluency in Finnish Students’ Writing .............................................................................. 25 
5.1 The Correlations between L1 and L2 Fluency Measures ............................................... 26 
5.2 Predicting L2 Fluency .................................................................................................... 28 
5.3 Qualitative Analysis: Individual Writer Profiles in L1 and L2 ...................................... 30 
6 Discussion ........................................................................................................................ 36 
7 Conclusion ........................................................................................................................ 40 
References ................................................................................................................................ 42 
Finnish summary ...................................................................................................................... 46 
 
 
List of Tables 
 
1 Means (and Standard Deviations) of the Fluency Measures……………………………….25  
2 Correlations between L1 and L2 Fluency Measures for the Whole Sample and for the Two 
Groups Separately……………………………………………………………………………27 
3 Results of the Simple Regression Models…………………………………………………29
 1 
 
1 Introduction  
The notion of fluency has been widely researched in the field of Second Language Acquisition 
(SLA) for many years. Many of the previous studies have focused on the effects of task 
characteristics (planning time, task difficulty, task repetition) on CAF measures, consisting of 
complexity, accuracy, and fluency (see e.g. Kim and Tracy-Ventura 2013; Thai and Boers 2016; 
Ong and Zhang 2010), L2 fluency development during residency abroad (see e.g. Hang 2013; 
Lennon 1990) and L2 fluency in relation to different proficiency levels (see e.g. Palviainen, 
Kalaja and Mäntylä 2012; Chenoweth 2001). The studies have generally examined fluency only 
from a foreign language learning perspective by using native speakers as a reference point. 
However, in recent years, the interest has shifted to the relationship between first language (L1) 
and second language (L2) fluency, since studies have shown that individuals vary also in their 
L1 fluency (see e.g. Maclay and Osgood 1959), which could indicate that factors affecting 
fluency in L1 could also be present in L2 performance. In other words, instead of being a 
language-specific state, L2 fluency could actually be more a trait-like feature which is affected 
by individuals’ L1 characteristics (see Derwing et al. 2009). 
Some studies have attempted to clarify the role of L1 fluency in L2 fluency (see e.g. 
Peltonen 2018; Derwing et al. 2009; De Jong et al. 2015; Huensch and Tracy-Ventura 2017). 
However, the results have not been able to provide a clear answer as to what extent fluency in 
L1 contributes to L2 fluency, since the relationship seems to depend on multiple different 
factors, such as the proficiency level of the speaker and cross-linguistic similarities and 
differences. For instance, although Derwing et al. (2009) and Huensch and Tracy-Ventura 
(2017) discovered positive correlations between fluency measurements in L1 and L2, they also 
found variation depending on the speaker’s L1 background. Thus, as Derwing et al. (2009, 555) 
state, the L1-L2 relation needs to be further examined, especially in language pairs which are 
not typologically that close to one another. Furthermore, as fluency is often considered to be a 
feature of spoken language, the studies have mostly focused on measuring oral fluency. This 
means that research on writing fluency has been left scarce, especially regarding the 
connections between L1 and L2 writing fluency. The earlier writing fluency studies have mainly 
focused on examining L2 fluency in relation to different proficiency levels (see e.g. Palviainen, 
Kalaja and Mäntylä 2012), or studied it from a developmental perspective (see e.g. Kowal 
2014). There is thus a strong need for writing fluency studies which incorporate L1 and L2 data 
and focus specifically on the effects of L1 fluency on L2 production.  
 2 
 
In order to fill in the research gap, the present study set out to investigate this matter by 
examining Finnish upper-secondary school learners of English (n=16) for their L1 and L2 
writing fluency. The data was collected in a form of essays, which participants composed by 
means of a keystroke recording tool ScriptLog. In order to examine the L1-L2 relationship 
profoundly, the research employed a mixed-methods approach where both, group and 
individual level analysis, were combined. Firstly, in the quantitative part of the study, the 
correlations between the L1 and L2 fluency measures were studied, after which regression 
analyses were conducted to examine how much of the variability in L2 measures could be 
predicted on the basis of L1 measures. The quantitative inspection of the data was then further 
complemented with a qualitative analysis, where the participants’ productions were examined 
individually for their fluency features with the help of the categorization of different writer 
profiles. Exploring the L1-L2 fluency relationship profoundly in writing is important as it has 
practical implications not only for L2 fluency assessment, but it could also help teachers to 
facilitate learners’ written text production. Furthermore, by using and comparing data from two 
morphologically and syntactically distinct languages, Finnish and English, the present study 
will provide more evidence on how fluency relations appear in language pairs in which is less 
likely that L1 knowledge and structures are directly transferred to L2.  
The study will begin by discussing the notion of L2 fluency and how it has been generally 
measured in learners’ L2 performance, especially in oral fluency studies. Then, it will move on 
to examine more closely L2 fluency in writing and different writer profiles, which is a related 
concept to writing processes and thus also employed by the present research. Theoretical 
framework will conclude by reviewing previous studies which have been conducted on the L1-
L2 fluency relationship. After that, the data collection methods will be presented in more detail 
in section 3. Sections 4 and 5 will focus on reporting and interpreting the results. The final 
section 6 will examine the reliability and validity of the study and present the final conclusions. 
 
 
 
 
 
 3 
 
2 L2 Fluency  
Although the notion of fluency is commonly used and studied in the field of SLA, there does 
not seem to exist a single definition of the term. Moreover, all the definitions seem to 
concentrate on characterizing spoken fluency instead of a written one. For instance, Skehan 
(2003, 510) considers fluency as one component of the framework CAF, which can be used to 
measure L2 performance. He views fluency as “the capacity to produce speech at a normal rate 
and without interruption” (ibid.). Fillmore, Kempler and Wang (1979), on the other hand, seem 
to have combined all these components into one. They define fluency as “an ability to talk with 
few pauses” and in a coherent and complex manner (Fillmore, Kempler and Wang 1979, 93). 
In addition, fluent speakers speak appropriately in different kinds of social contexts and 
situations and they are able to produce creative and imaginative language (ibid.). Lennon (1990, 
389390), in turn, uses the term in two different ways; in a broad sense to cover the speaker’s 
overall foreign language proficiency, and in a narrower sense to refer only to the rate and 
smoothness of the output. In this sense, fluent speakers are able to produce language at a native-
like pace without unnecessary pauses and hesitations, repetitions or false starts (Lennon 1990, 
390). In regard to the underlying cognitive processes, Schmidt (1992, 358359) views fluency 
as an “automatic procedural skill”, which develops as a result of proceduralized knowledge and 
thus, in the end, does not require much effort or attention on the speaker’s part. As can be seen 
from these examples, the common fluency features seem to be related to the effortlessness and 
smoothness of the L2 production. Additionally, the researchers seem to agree on the fact that 
fluency can be used as one indicator of L2 proficiency. However, the lack of a full consensus 
on the different features of fluency makes it a difficult research subject. Therefore, in order to 
measure fluency for research purposes, the researchers have had to find ways to operationalize 
the concept and define it more practically (Segalowitz 2010, 3; 5). 
In order to create more explicit measurements for fluency, Segalowitz (2010, 4849) has 
distinguished three senses of L2 fluency: cognitive, perceived and utterance fluency. The first 
one, cognitive fluency, refers to “the fluid (speed, efficiency) operation of the cognitive 
processes”, which enables learners to produce L2 utterances (Segalowitz 2010, 48; Segalowitz 
2016, 82). The second type of fluency, perceived fluency, relates to the subjective inferences 
on the speaker’s fluency, which listeners make based on their impressions from speech samples 
(Segalowitz 2010, 48). The third and final type, utterance fluency, refers to measurable features 
of the utterance, which include, for instance, pauses, hesitations and repairs (ibid.). Skehan 
(2003, 8) has divided utterance fluency further into separate measures of breakdown fluency, 
 4 
 
repair fluency, speech rate and automatization. Breakdown fluency is related to pausing 
phenomenon, while repair fluency covers reformulations, false starts and repetitions (ibid.). 
Speech rate can be measured by calculating e.g. words/syllables per minute and automatization 
via the length of run (ibid.). As all these above-mentioned models and definitions demonstrate, 
L2 fluency has been mainly explored in learners’ oral production. However, the same 
definitions can be applied to written fluency as well, although the two constructs somewhat 
differ from one another (see section 2.2.). Therefore, the present research will also base its 
definition of fluency on models of Segalowitz and Skehan by focusing on the utterance fluency. 
In other words, the L2 fluency in the present research refers to the temporal aspects of the 
writing process as well as revision behaviour. The diverse measures to analyse the fluency of 
the text will be examined more closely in the following section. 
According to Ellis and Barkhuizen (2005, 149), there are two different ways which 
researchers can use to assess the fluency of learner language. Firstly, they can concentrate on 
the perceived fluency by obtaining listener or reader ratings, which is a practical way to assess 
L2 performance (ibid.). The second option is to use various discourse-based measurements (cp. 
Skehan’s model), which provide precise numbers of different aspects of fluency (ibid.). Some 
studies have also combined the methods and compared the results (ibid.). Many SLA 
researchers have opted for the second type and they have generally based their 
operationalisations on Skehan’s model (Ellis and Barkhuizen 2005; 149; Olkkonen 2017, 15). 
In regard to the current study, the fluency of the texts was examined by means of different 
discourse-based measurements. Although listener/reader ratings were not used to assess the L2 
fluency in this study, it has to be noted that they might be reflected in the Common European 
Framework of Reference for Languages (CEFR) ratings as the following section will 
demonstrate. CEFR evaluations, on the other hand, are based on the fluency of the final product, 
which differs from the fluency of the production process, i.e. the focus of the present research. 
Ellis and Barkhuizen (2005, 156158) list the most common measures of fluency, which 
have been applied in the previous studies. They divide the measures into two different 
categories: temporal variables and hesitation phenomena (ibid.). Temporal variables are related 
to the speed of L2 performance, and they include variables such as speech/writing rate, the 
number of pauses, pause length and length of run (ibid.). Hesitation phenomenon, on the other 
hand, relates to dysfluency, and they include false starts, repetitions, reformulations and 
replacements (ibid.). In regard to the four hesitation variables, Ellis and Barkhuizen (2005, 158) 
argue that a single variable is valid to demonstrate this aspect of fluency, since previous studies 
have shown a strong relationship between them. Furthermore, when it comes to oral fluency, 
 5 
 
speech rate and silent pause phenomena have been associated with L2 proficiency more often 
than some other measures (Segalowitz 2010, 39). Despite the fact that some variables seem to 
be more suitable than others to assess fluency, choosing and applying valid measurements is 
not entirely unproblematic. According to Segalowitz (2010, 39; 41) and Ellis and Barkhuizen 
(2005, 163), researchers have not been able to fully discover which measures constitute the 
most valid and reliable markers of fluency. Thus, researchers have ended up using different 
measures and even different operationalisations, which makes comparing results across studies 
difficult (ibid.). In addition, Segalowitz (2010, 39; 41) points out that differing data collection 
methods can also affect the research outcome, as some studies have used interviews and others, 
for instance, retelling stories to elicit speech samples. Although the authors refer mostly to 
problems regarding measuring oral fluency, the same issues concern measuring writing fluency. 
In order to avoid these problems and obtain comparable results, the present research will 
employ similar methods and operationalisations used by other studies. By using the same 
measures, operationalisations and data collection methods which most of the previous studies 
have used, the results across diverse studies can be compared more validly. However, given the 
fact that all the possible limitations cannot be avoided in the research process, the results of the 
present study also need to be regarded with caution.   
In this section, I have covered the multiple definitions of fluency and discussed the 
various ways it can be measured in L2 performance. Although the measures introduced in this 
section have been mostly used to assess spoken fluency, similar measurements can be applied 
to analyse written productions. However, it is good to bear in mind that writing fluency 
measures need to be operationalised differently and that individual’s oral and writing fluency 
cannot be directly compared to one another due to the divergent conditions of producing written 
vs. spoken texts (Ellis and Barkhuizen 2005, 145; Kowal 2014, 232). Thus, before investigating 
writing fluency, we need to understand the cognitive processes which underlie the writing 
process. The following section will provide an overview of these issues by focusing on the L2 
writing fluency phenomenon and introducing different means on how to measure it validly. 
2.1 L2 Writing Fluency 
Although the actual definition of fluency is rather the same for spoken and written 
communication, the cognitive processes of producing written and spoken language differ from 
one another. Torrance and Nottbusch (2012, 404406) list three main differences between 
written and spoken language processing, which need to be taken into consideration before we 
 6 
 
can validly measure L2 writing fluency. Firstly, they point out that in speaking, more 
concentration on fluency is needed compared to writing, since speakers are faced with a 
pressure of producing a fluent utterance in real-time, whereas writers can pause and modify the 
text as much as they choose, without the receiver ever finding out (ibid.). The second difference 
in language processing is related to orthographic processing, which is only present in written 
production and needs thus additional attention from writers (ibid.). Thirdly, the authors note 
that the feedback affordance can be slower in written mode compared to speaking, because 
writers are able to see the produced text and, therefore, both error monitoring and correction 
can be postponed (ibid.). Given these divergent characteristics of producing written and spoken 
language, it is clear that writing fluency studies need to employ and operationalize fluency 
measurements differently. Additionally, as Kowal (2014, 232) notes, the conditions of 
producing text and speech differ from one another. While oral text emerges and disappears fast 
from the speaker’s mind, the written text remains visible even after the writer has moved on to 
another topic (ibid.). Therefore, it is possible that there is a divergence between individuals’ 
fluency in writing and speaking, which means that the results of speech and written fluency 
studies cannot be directly compared to one another (ibid.). 
When it comes to measuring writing fluency, there are two different approaches: product- 
and process-oriented approaches. Product-oriented measures focus on the fluency of the end 
product, whereas process-oriented measures take into account the process of writing 
(Palviainen, Kalaja and Mäntylä 2012, 48). In the end product, fluency is related to readers’ 
perceptions as to what constitutes as fluent text (ibid.), and much like in speech fluency studies 
which use listener ratings, it can be studied by obtaining reader ratings. However, as readers 
are unable to see how fluently the writer produces the text, they will have to base their 
perception of fluency on linguistic features of the text, such as text structure, coherence and 
cohesion. This, in turn, is linked to the overall writing proficiency, which Wolfe-Quintero, 
Inagaki and Kim (1998, 2) define as “an ability to produce a variety of genres and rhetorical 
features”, as well as “the use of a range of vocabulary and syntactic structures”. The writing 
proficiency can be assessed by using various developmental measures, such as the age of the 
writer or the number of years a writer has studied the language (Palviainen, Kalaja and Mäntylä 
2012, 48). Moreover, writing proficiency can be measured by judging the text quality on the 
basis of pre-set criteria (ibid.), such as different language proficiency scales. 
As mentioned in the previous section, the present research did not use reader perceptions 
as a direct method to assess the fluency of the texts, but rather as a measurement of the overall 
quality of the texts and thus providing information on the proficiency level of the participants. 
 7 
 
However, as the texts were graded by using The Common European Framework of Reference 
for Languages (CEFR), which, in fact, mentions fluency in all of its six levels A1-C2, the grade 
could also be regarded as an indicator of the fluency of the end product. The CEFR is a guideline 
used for foreign language assessment across Europe, which defines different levels of 
proficiency in order to measure language learners’ progress at each stage of learning (CEFR 
2001, 1). In addition to general rating scales, CEFR also provides a more detailed set of level 
descriptions for different sub-skills of language learning, namely speaking, reading, writing and 
listening. In the level descriptions for writing, fluency is referred to as one of the characteristics 
of the written production. For instance, at the level C2, the writer should be able produce “clear, 
smoothly flowing, complex texts, in an appropriate and effective style and a logical structure, 
which helps the reader to find significant points”, while at the level B1, in regard to the thematic 
development, the language user is supposed to “be able to reasonably fluently relate a 
straightforward narrative or description as a linear sequence of points” (CEFR 2001, 61; 125). 
As a result, it seems that the more fluent the writing process is, the better the final text is in 
terms of its cohesion, complexity and overall quality. The CEFR evaluations can be regarded 
as useful indicators of the fluency of the end-product, and many studies have in fact studied the 
possible relationship between the quality of the end-product and the fluency of writing process 
(see e.g. Lindgren, Spelman Miller and Sullivan 2008; Spelman Miller, Lindgren and Sullivan 
2008; Palviainen, Kalaja and Mäntylä 2012; Stevenson 2005), albeit the results of these studies 
have been somewhat contradictory with one another. Nevertheless, in the present research, the 
role of CEFR evaluations was mainly to provide information on the proficiency level of the 
participants. The proficiency level, in turn, was used in the analysis in order to determine 
whether the degree of correlation between L1-L2 fluency measurements would be dependent 
on the L2 proficiency, as some studies have demonstrated (e.g. Huensch and Tracy-Ventura 
2017; Peltonen 2018).  
While product-oriented studies disregard the writer and focus more on the writing 
outcome, in the process-oriented studies, the writer becomes the subject of the research. 
According to Hayes and Flower’s cognitive model of the writing process, writing consists of 
three major cognitive processes: planning, translating and reviewing (Hayes and Flower 1980, 
12). These are then further divided into different subprocesses, which include for instance 
generating relevant information for the writing task from the long-term memory, setting goals, 
and reading and editing of the text (ibid.). In process-oriented studies, writing is considered to 
be fluent when transitions between these cognitive processes are effortless and smooth 
(Palviainen, Kalaja and Mäntylä 2012, 4950). L2 writers are likely to experience difficulties 
 8 
 
in fluent retrieval of words and grammatical structures due to their limited proficiency in L2, 
which will thus burden their working memory more (Lindgren, Miller and Sullivan 2008, 134; 
Schoonen et al. 2003, 171). This, in turn, could lead to more frequent or different pausing 
behaviour and more frequent revisions compared to L1 writers (Schoonen et al. 2003, 171; Hall 
1990; Spelman Miller 2000). In addition, as their working memory is occupied by lower-level 
linguistic demands, they might be less focused on the quality of the text, for instance in terms 
of organizing the text properly (Schoonen et al. 2003, 171). Besides differences in pausing and 
formulation behaviour, previous studies have shown that L2 writers produce shorter texts, lower 
number of words between the pauses, devote more time in solving formulation problems and 
display generally a slower rate of writing compared to L1 writers (see Silva, 1993 for a review 
of studies; Hall 1990; Chenoweth and Hayes 2001; Roca de Larios, Manchón and Murphy 
2006; Van Waes and Leijten 2015).  
In regard to measuring the fluency of the writing process, studies have generally used 
measures which relate to speed, length and time of production (Palviainen, Kalaja and Mäntylä 
2012, 50). Thus, the traditional way to measure fluency has been to divide the total number of 
words by the total writing time (ibid.), which provides the writing rate. Written fluency is, in 
fact, regarded mainly as a temporal phenomenon in many of the previous studies, and some 
studies, like Chenoweth and Hayes’ (2001, 81), have even defined it as “the rate of production 
of text”. Similarly, Wolfe-Quintero et al. (1998) viewed speed and length as the most crucial 
parameters of fluency while they reviewed 39 studies in an attempt to find out the most valid 
measures of fluency connected to language proficiency. In their analysis, they concluded the 
best measures to be the number of words per T-unit, a T-unit being an independent clause with 
possible subordinate clauses attached to it, the number of words per error-free T-unit and the 
number of words per clause (Wolfe-Quintero et al. 1998, 29). However, as Kowal (2014, 231) 
and Chenoweth and Hayes (2001, 82) point out, Wolfe-Quintero et al. (1998) did not include 
in their research any studies which also take into account the possible disturbances in fluency, 
such as pauses and false starts, and thus these measures do not really explicate the production 
processes which enable a writer to compose text. Instead, researchers have started to use Burst 
Size, defined by Chenoweth and Hayes (2001, 88) as the number of words produced between 
pauses or revisions, as a better indicator of the fluency of the writing process. Kaufer, Hayes 
and Flower (1986, 126127) found early on that writers produce texts in parts rather than full 
sentences. These sentence parts, i.e. bursts, were identified by pauses of two seconds or more, 
and the average length of them seemed to be connected to the writer's proficiency level (ibid.). 
Based on the results, the authors suggested that the average length of the bursts increases when 
 9 
 
writers’ proficiency develops (ibid.). The idea of the burst has since been developed further and 
many studies have added a measurement of Fluency during Burst (see e.g. Spelman Miller, 
Lindgren and Sullivan 2008), referring to the speed of writing during the burst. Given the fact 
that the unit of a word is problematic to define and use especially in computer-based studies, 
researchers have started to use characters instead of words as a measurement unit (see e.g. 
Palviainen, Kalaja and Mäntylä 2012; Lindgren, Spelman Miller and Sullivan 2008). Moreover, 
by focusing on characters instead of words, researchers are also able to compare the fluency of 
morphologically and syntactically distinct languages (Palviainen, Kalaja and Mäntylä 2012, 
51). As a result, the present research also used characters as a measurement unit in the analysis, 
for the length of words could vary significantly between Finnish and English. 
Based on previous research, it thus seems that investigating pauses, revisions and bursts, 
as well as text length and production rate, could potentially provide insight into the L2 writing 
fluency. A closely related concept to writing processes is also writer profiles, which have been 
created based on individuals’ pausing and revision behaviour and the duration of their writing 
process. As they are created on the basis of many of the fluency measurements listed in this 
section, they could be useful in analysing and comparing L1 and L2 fluency behaviour 
individually. The different writer profiles will be presented next, as they will also be used in 
the qualitative analysis of the present research to help to categorize the participants. 
2.2 Writer Profiles 
As mentioned in the previous section, the writing process consists of several cognitive 
processes, namely planning, formulation and revising, as well as their subprocesses. These 
processes are not independent of one another, but rather dynamic and overlapping at times 
(Manchón 2013, 103). In other words, in the course of composing a text, writers activate some 
processes while deactivating others, and move on from one process to another in a cyclical 
manner rather than linearly (ibid.) Given this dynamic and recursive nature of the writing 
process, it is unlikely that there exists a one single writing process but instead many possible 
ways for a writer to compose the text.  
Previous studies have demonstrated that individuals tend to adopt different writing 
profiles in their composing process. In one of the earliest studies on writing processes, Hayes 
and Flower (1980) were able to identify four different writing profiles based on their study, 
where they used think-aloud methods to examine college students’ pen and paper writing tasks. 
These profiles included the following types: 1) Depth-first, in which the writer moves on to the 
 10 
 
next sentence only after finishing the first one, meaning that planning, translating and reviewing 
of each sentence is completed before moving on to another one; 2) Postponed review (term used 
by Van Waes and Schellens 2003, 831), in which all the possible thoughts are written down 
first, and only later reviewed, 3) Perfect first draft, in which the writer’s aim is to produce a 
perfect first draft, and 4) Breadth-first, in which “a draft is planned and then written out in full 
before any review takes place” (Hayes and Flower 1980, 19-20). Writer profiles have since 
been investigated more for instance by Schwartz (1983), who created a similar typology to 
Hayes and Flower’s (1980) but consisting of nine writer profiles. The problem of both of these 
studies, as Van Waes and Schellens (2003, 831832) note, is that they describe the profiles in 
a rather general manner and the categories are not mutually exclusive, meaning that it is possible 
to categorize writers to several profiles instead of only one. 
In a more recent study, Van Waes and Schellens (2003) developed a new writing typology 
by examining and comparing the effect of the writing mode (computer/pen&paper) on writing 
behaviour in L1 and by adding a novel quantitative approach to the definition of writer profiles. 
In their research, Van Waes and Schellens (2003) used a computer program Keytrap, which is 
a tool designed to record different variables related to participants’ pausing and revision 
behaviour, as well as the duration of their writing process. In the present research, a similar 
keystroke logging tool (ScriptLog) was used, which also provides information regarding the 
writing process, such as the number and length of pauses and the total duration of the writing 
process. Therefore, the typology proposed by Van Waes and Schellens (2003) was adopted for 
analysis of the present research. 
Van Waes and Schellens (2003, 836837) base their typology of writer profiles on three 
different aspects of the writing process. Firstly, they investigated the aspect of the time taken 
and the final product, which included variables of total duration of the writing process, duration 
of each phase (phase 1: from initiation to the completion of first draft, phase 2: from the 
completion of first draft to the completion of the final version), ratio of time spent pausing to 
time spent actively writing and number of words in the final text (ibid.). Secondly, the 
participants’ pausing behaviour was examined by means of the number, duration and type of 
pauses, as well as the linguistic and temporal location of pauses (ibid.). Finally, the revision 
behaviour was studied by examining the number, level, purpose, location, remoteness and 
temporal location of revisions (ibid.). Based on the findings, Van Waes and Schellens (2003, 
844845) were able to establish five different profiles: 
 11 
 
1. Initial planners made rather few revisions, especially during the second phase of 
the writing process, and they tended to pause the longest during the initial 
planning phase. Their average pause length and the total time spent pausing were 
also above the average. 
2. Average writers showed values which were the closest to the average values of 
the whole group in each of the variables. 
3. Fragmentary Stage I writers (or Fragmentary first-phase writers, as used by 
Mutta, forthcoming) tended to revise more during the first phase of the 
composition process compared to the second phase, and the total number of 
revisions was higher compared to other profiles. They spent little time on initial 
planning and paused relatively often during writing, although the average duration 
of the pauses was short. Therefore, their writing process could be considered to 
be highly fragmented.  
4. Stage II writers (or Second-phase writers as used by Mutta, forthcoming) made 
most of their revisions during the second phase and above the word level. They 
devoted a rather long time to initial planning, but once they started to write, they 
paused relatively little, albeit those pauses were relatively long. 
5. Non-stop writers revised the least of all the groups and paused less often than 
others. They did not spend a long time on initial planning and completed the task 
more quickly than other groups 
Van Waes and Schellens’ taxonomy was used by Mutta (forthcoming) in her research on 
writing processes and writer profiles. The aim of the study was to examine the relationship 
between L2 fluency measures and writer profiles, and how they, in turn, are related to the quality 
of the end product (ibid.). The online data were collected by means of the recording tool 
ScriptLog and it consisted of texts produced by 11 Finnish university students of French (ibid.). 
The fluency of the texts was examined by using both, the process-oriented measurements (e.g. 
pause length, median transition times, number of editions and deletions, length of initial writing 
phase) and end-product values (e.g. the number of words, type/token ratio, evaluation scores) 
(ibid.). The process-oriented variables differed slightly from those used by Van Waes and 
Schellens (e.g. the use of paragraphs instead of phases), which Mutta explained to be due to the 
difficulty of distinguishing different writing phases from one another (ibid.). In the end, Mutta 
was able to classify the participants into four different writer profiles (leaving out the average 
writers), although she points out that some of the writers seemed to possess characteristics of 
 12 
 
several profiles and, therefore, the categorization was mostly done on the basis of prevailing 
features of the writers (ibid.). The results of the study showed that there was no linear 
relationship between the process-oriented fluency measures and the quality of the end product 
(ibid). Consequently, it could be assumed that the different writer profiles were not connected 
to the success of the end product (ibid.). Writer profiles are also discussed in Mutta’s (2007) 
doctoral dissertation, where she examined writer profiles as part of the research related to L1 
and L2 writing processes. By examining the individual writer profiles, Mutta was able to find 
similarities between individuals’ L1 and L2 writing processes (Mutta 2007, 223). Although the 
sample size in this case was small (N=5), the findings suggest that writers’ L1 and L2 profiles 
could be related to one another, thus indicating also a relation between L1 and L2 fluency. The 
present research was set out to investigate this matter. However, before moving to the research 
methodology, which will present the aim and the methods of the study in more detail, I will 
review some of the previous studies which have been conducted on the L1-L2 fluency 
relationship. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 13 
 
3 The Relationship between L1 and L2 Fluency 
The last two sections have provided insight into L2 writing processes and different fluency 
measures. They have not, nevertheless, answered the question as to why there exists variation 
among L2 users in terms of their fluency. Studies suggest that at least some of the variance 
could be explained by learners’ fluency in L1. In other words, individuals who are fluent in 
their L1 could also possess similar characteristics in their L2 performance. However, the 
relationship between L1 and L2 fluency is not entirely straightforward, as many of the earlier 
studies have demonstrated. Next, some of these studies and their conclusions are presented. 
First, this section will provide an overview of the previous research which has had the same 
goal as the present study. However, it has to be noted that as there does not seem to exist 
previous studies which have focused on L1-L2 fluency connections specifically in writing 
contexts, most of the studies presented here have been conducted by using oral data. Thus, their 
results might not be entirely comparable. They provide, nevertheless, an overview of the 
possible outcome also for the current research. After this, this section will also present some of 
the earlier writing fluency studies with a different focus compared to the present study, but 
which have, nevertheless, referred to L1-L2 connections briefly in their results. Thus, their 
findings are useful for the current study as well. 
As mentioned, many of the previous studies which have examined the use of L1 in L2 
productions have mainly focused on measuring speaking or reading skills. The studies have 
found that there does, in fact, exist a relationship between L1 and L2 skills at least to some 
extent, although the relationship seems to depend on multiple different factors, such as L2 
proficiency level and crosslinguistic similarities and differences. One of the first comprehensive 
studies examining the topic was conducted by Derwing et al. (2009), who compared L1 and L2 
oral productions of 16 Slavic and 16 Mandarin beginner learners in their longitudinal study. 
They found statistically significant correlations between L1 and L2 fluency at the initial stages 
of L2 exposure for both groups, although the correlation was found to be stronger for the Slavic 
group (Derwing et al. 2009, 552553). At the later stages of data collection, the findings 
revealed a correlation only for the Slavic group (ibid.). The authors explain that the variance 
between the two groups might be due to the typological closeness of Slavic speakers’ L1 to 
English, which might have given them an advantage over the Mandarin speakers especially at 
the beginning of learning (ibid.). In addition, they argue that some of the individuals in the 
Mandarin group might not have developed during the study, which could explain the lack of 
correlation at the later stages of data collection (ibid.). The results of the study are thus 
 14 
 
inconclusive as to what extent the L1 fluency influences the L2 fluency, since the correlations 
were not consistent and significant at all the stages of the data collection process. Therefore, 
the authors emphasize the need for further longitudinal investigations, involving also a wider 
variety of L1s in order to examine fluency in different language pairings (Derwing et al. 2009, 
555). 
In a more recent study, De Jong et al. (2015) investigated two different L1 language 
groups (n=53), English and Turkish, and compared their speech samples in L2 Dutch. Besides 
examining the cross-linguistic differences of the two L1 groups, their aim was to investigate 
whether measures of L2 fluency which were corrected for L1 fluency behaviour could predict 
the overall L2 proficiency (measured by a productive vocabulary task) better than uncorrected 
L2 measures (De Jong et al. 2015, 226). The results revealed statistically significant correlations 
between all the L1 and L2 fluency measures, the strength of the correlation ranging from r = 
.37 for mean syllable duration to r = .76 for mean silent pause duration between ASU (Analysis 
of Speech Units) (De Jong et al. 2015, 233). In addition, the findings of the regression analyses 
showed that all the L2 fluency measures could be predicted from the L1 measures, the success 
of prediction varying from 21 % for mean syllable duration to 57 % for mean length of pauses 
between ASU (De Jong et al. 2015, 234). Moreover, the results indicated that both, uncorrected 
and corrected measures, predict L2 proficiency equally well, except for mean syllable duration, 
for which the corrected measure was a stronger predictor than the uncorrected one (De Jong et 
al. 2015, 237). Regarding the question of cross-linguistic differences, the results showed that 
the two language groups differed in their L1 fluency in syllable duration, the number of silent 
pauses and repetitions, but no similar differences were found in their L2 (De Jong et al. 2015, 
231232). Therefore, the authors concluded that the typological closeness of L1 did not affect 
the predictive value of L2 measures (ibid.). 
Huensch and Tracy-Ventura (2017) studied longitudinally the connections between L1 
(English) and L2 (Spanish and French) (n=49) oral fluency by employing similar research 
design and fluency measures as Jong et al. (2015). By taking into account not only L1 fluency, 
but also cross-linguistic differences and the L2 proficiency (measured by Elicited Imitation 
Test), they wanted to explore the relative contribution of each variable in predicting the L2 
fluency (Huensch and Tracy-Ventura 2017, 761). Overall, the results demonstrated weak to 
strong positive correlations between the L1 and the L2 fluency measures both before and after 
5 months’ residence abroad, although only a subset of the correlations was statistically 
significant (Huensch and Tracy-Ventura 2017, 773). Regression analyses showed that before 
the residence abroad, only three of the measures (mean syllable duration, the number of silent 
 15 
 
pauses per second, mean silent pause duration within ASU) could be predicted based on L1 
measures, L2 group status and proficiency; L1 fluency only contributing to the first two of these 
(Huensch and Tracy-Ventura 2017, 774; 779780). In contrast, after the residence, L1 fluency 
contributed significantly to all the L2 measures, while the L2 group contributed to measures 
where cross-linguistic differences had been found between the native speakers as a result of a 
separate analysis comparing native English, Spanish and French speakers (ibid.). L2 
proficiency, on the other hand, did not contribute any of the measures (ibid.). Based on these 
findings, the authors conclude that the role of L1 fluency and cross-linguistics differences seem 
to change as learners improve in L2 fluency; the relationship being weaker at the initial stages 
of learning and explaining variance in L2 fluency more as learners develop (Huensch and 
Tracy-Ventura 2017, 780). 
The connections between L1 and L2 oral fluency have also been examined in the Finnish 
context by Peltonen (2018), who studied Finnish learners of English (n=42) by using a mixed-
methods approach. The school level of the participants, ninth grade and second-year upper 
secondary school, was taken into account in the analysis in order to determine the L2 
proficiency level of the learners (Peltonen 2018, 679). Peltonen (ibid.) examined fluency from 
two different perspectives: quantitatively by analysing rank-order correlations and regression 
models, and qualitatively by selecting a number of participants and analysing their L1 and L2 
productions more closely. The results of the rank-order correlation analyses showed statistically 
significant positive correlations between the majority of the fluency measures, the strength of 
the correlations varying from medium to large effects (Peltonen 2018, 687). The strongest 
correlations were found in the mean length of end-clause SPs, mean length of SPs, mid-clause 
SPs/min and articulation rate, which are part of the temporal fluency measurements (Peltonen 
2018, 684). Similarly, the regression analyses revealed that the L2 fluency measures could be 
predicted from the equivalent L1 measures at least to some extent (Peltonen 2018, 689). In 
regard to the role of L2 proficiency, the results seemed to be in line with the previous studies, 
as they indicated that the L1-L2 correlations were stronger for the more proficient group 
(Peltonen 2018, 688). 
The same question of L1 and L2 interrelations has also been studied in writing contexts. 
It has been found that writers resort to their L1-based strategies for a variety of different 
functions, such as planning and organizing the text, solving linguistic and stylistic problems, as 
well as monitoring and controlling the writing process by changing languages (Manchón 2013, 
109; Manchón, Roca de Larios and Murphy 2007, 239242). Moreover, research has provided 
evidence that the use of L1 strategies is related to L2 proficiency (Manchón 2013, 109; Wang 
 16 
 
2003). If the L2 proficiency is low, L1 has mainly a compensatory function, which is reflected, 
for instance, as a higher number of borrowings and translations (Manchón 2013, 110; Wang 
2003; Navés, Miralpeix and Celaya 2005; Sasaki 2000). As the L2 proficiency develops, 
fluency increases, and learners start to use L1 as a means for more higher-level writing 
processes and overall aspects of language generation, such as solving rhetorical and discourse 
problems (Manchón, Roca de Larios and Murphy 2007, 242; Wang 2003; Manchón 2013, 110). 
Although there are many studies which examine the use of L1 in L2 writing, especially 
concerning writing skills and strategies as the previous paragraph demonstrates, the number of 
studies focusing solely on the fluency aspect in writing has been scarce. Those writing fluency 
studies, which have used and compared both L1 and L2 data, have mostly focused on L2 fluency 
in relation to different proficiency levels (see e.g. Chenoweth and Hayes 2001) or studied 
fluency from a developmental perspective (see e.g. Lindgren, Spelman Miller and Sullivan 
2008; Yang and Sun 2015; Kobayashi and Rinnert 2013). Chenoweth and Hayes (2001) 
explored the relationship between writing fluency and linguistic experience by studying native 
speakers of English (n=13) who were learning French or German as a third-semester or as a 
fifth- or sixth-semester course. In the analysis, they used the texts composed by the students, as 
well as think-aloud methods and videotapes of the writing sessions (Chenoweth and Hayes 
2001, 86-87). The results revealed that the L2 experience had an impact on the fluency of the 
texts, which they measured in their study by words written per minute (Chenoweth and Hayes 
2001, 89). When investigating and comparing L2 and L1 fluency separately, the findings 
showed that although the more proficient learners were more fluent in their L2, the average 
number of words written in minute being 10.91 compared to 10.61 produced by the less 
proficient group, they were also surprisingly less fluent in their L1, the average being 13.831 
words while the other group produced 20.1 words (Chenoweth and Hayes 2001, 89-90). The 
authors suggested that the fifth-year students of their study happened to have lower verbal skills 
in general, which led to a reduced fluency in both, L1 and L2 (ibid.). In addition, the authors 
found out that differences in fluency were related to the length of P-Bursts, the burst sizes being 
longer with the more proficient group, and in general longer in L1 compared to L2 with all the 
learners (Chenoweth and Hayes 2001, 91-92).  
Lindgren, Spelman Miller and Sullivan (2008) examined the development of fluency of 
Swedish high-school students (n=17) longitudinally over the course of two consecutive years 
by comparing texts the students wrote in their L1 Swedish and L2 English. The fluency of the 
texts was evaluated by using measurements of fluency, burst and fluency during burst, and 
they analysed the data with the help of ANOVA and linear regression models (Lindgren, 
 17 
 
Spelman Miller and Sullivan 2008, 138-141). The results indicated that the linguistic 
experience had a positive effect on all the three fluency measures, as well as on the quality of 
the text, measured by grade, and the number of conceptual revisions, meaning that learners 
focused more on the conceptual aspects of writing once their proficiency developed 
(Lindgren, Spelman Miller and Sullivan 2008, 146-147). In regard to differences between L1 
and L2 fluency measures, the results revealed that students were more fluent in their L1 
writing compared to L2 (ibid.). However, interestingly, the findings also indicated that they 
improved more in fluency measures in L1 over a two-year period (ibid.). The authors 
suggested that one explanation for the results was that the students developed their writer 
status more in L1 during this time, while fluency problems continued to be present in L2 
writing (ibid.). The measures between L1 and L2 also revealed that the learners employed 
different writing and revision processes in these two languages, where they, for instance, 
revised more in L2 and paused less in L1 (ibid.).  
Yang and Sun (2015) focused on the development of L2 complexity, accuracy and 
fluency by studying undergraduate multilingual learners’ (n=5) L1 (Chinese), L2 (English) 
and L3 (French) writing longitudinally during one academic year. In their study, fluency was 
measured by the mean number of words per t-unit, in which they also took into account the 
syntactic differences of languages by defining the concept differently in Chinese (Yang and 
Sun 2015, 301). The results showed that the developmental patterns in CAF measures were 
not linear and they were characterized with constants changes, increases and backsliding 
(Yang and Sun 2015, 306). In addition, when the authors examined the relationship between 
complexity, accuracy and fluency within one language, as well as across all the languages, the 
findings implied there to be a relation between them to a certain degree, although the only 
statistically significant correlation was found between L1 and L3 accuracy (Yang and Sun 
2015, 304-305). Thus, the study further demonstrates that multilingual learners’ different 
languages interact with each other (Yang and Sun 2015, 307). 
As the results of these studies demonstrate, learners exhibit some differences in writing 
processes and writing fluency between the L1 and L2. However, as these studies have had a 
slightly different emphasis and have mainly focused on the possible differences in fluency 
between L1 and L2 data, the actual L1-L2 connections of writing fluency on the individual 
level have still remained an under-researched issue. The relationship between L1 and L2 
writing proficiency, on the contrary, has been a subject of many studies, and many of them 
have in fact been able to attest a relation between the two constructs (see e.g. Sasaki and 
Hirose 1996; Schoonen et al. 2003; Schoonen et al. 2011). Consequently, as fluency is 
 18 
 
generally regarded as one of the possible indicators of L2 proficiency (see section 2.1.), the 
results of these writing proficiency studies could indicate that also L1 fluency somewhat 
correlates with L2 fluency. 
This section has discussed the possible effects of L1 fluency on L2 fluency and introduced 
some of the earlier studies which have been conducted on the subject. Nevertheless, it has to be 
noted that since many of these studies have examined oral fluency instead of a written one, the 
results might not be entirely comparable to the current research. After all, as it was already 
mentioned in section 2.2., individuals’ oral and written fluency might differ from one another 
due to the divergent cognitive processes required to produce the modes. Despite these issues, it 
was important to have some reference points, and therefore, the findings presented in this 
section will also work as a starting point for the current study.  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 19 
 
4 Research methodology 
In this section, the research methodology will be presented in more detail. First, the research 
questions and the hypotheses will be explained. Second, the data collection methods will be 
described in more detail, and finally, the methods of analysing the data. 
4.1 The Aim of the Research 
The data were collected as a part of the larger research project, which explores how formulaic 
sequences are used and how they can be recognized in three morphologically different 
languages (Swedish, English, Finnish). It has collected written online data from upper-
secondary school students in the form of essays. The linguistic background of the participants 
also varies, as students with diverse L1s (Finnish, Swedish) have composed essays in different 
L2s (English, Swedish and/or Finnish depending their L1). 
The present research used the already collected data for the purpose of examining and 
comparing the essays in terms of their fluency. The objective was to examine the connections 
between first and second language fluency by analysing the features of the writing process. The 
L1-L2 connections were explored first quantitatively by using correlation and regression 
models. Subsequently, a more profound qualitative approach was used in which the essays were 
examined individually with the help of the categorization of writer profiles. The following 
research questions were addressed:  
1) To what degree do measures of L1 fluency correlate with equivalent measures of L2 
fluency?  
2) To what extent can L2 measures of fluency be predicted from the equivalent L1 
measures of fluency? 
3) To what extent do writer profiles in L1 and L2 correspond to one another? 
Based on the findings of the previous studies, it was hypothesized that the present research 
would also find a relation between L1 and L2 fluency measures; the correlation potentially 
being stronger for the more proficient learners. Additionally, it was expected that the 
participants could somewhat be categorized to different writer profiles, and that these profiles 
would also correspond to one another in the individual’s L1 and L2. In other words, it was 
presumed that writers who belong to a certain group in their L1 writing would also possess 
similar characteristics in their L2 writing, which would further reinforce the relation between 
L1 and L2 fluency. By using a mixed-methods approach, where quantitative research questions 
 20 
 
were complemented with a more profound qualitative analysis on personal writing 
characteristics, the aim was to study L2 fluency from a new perspective, and thus contribute to 
the currently under-researched field of L1-L2 writing fluency connections.  
4.2 Data Collection and Processing 
The online data consisted of essays written by Finnish upper secondary school students (N=16, 
3 males, 13 females; aged between 16 and 17 years) in their L1 (Finnish) and L2 (English and 
Swedish). Given the scope of the present study, only the essays written in Finnish and English 
were analysed. However, for future research, it could be useful to examine not only the relation 
between an individual’s L1 and L2, but also the role of the possible L3s or L4s in the process. 
In the data collection phase, the students were asked to write the essays by means of the 
keystroke recording tool Sriptlog (Strömqvist and Karlsson 2002), which is a programme 
designed to record every keystroke, mouse and cursor movement during the writing session. 
First, the students were given 30 minutes to plan and compose an argumentative essay in 
Swedish, and then another 30 minutes to write one in English. Lastly, they had 30 minutes to 
complete a similar argumentative essay in their L1 (Finnish). The length of the essays was 
instructed to be around 150-250 words and they had to be composed without the help of 
dictionaries or other resources. In order to determine the language proficiency level of 
participants, the L2 essays were later evaluated by two reviewers who gave the grade according 
to the CEFR-scale. Two of the essays were graded as A2 (referred in this study as P1 and P2), 
six as B1 (P3-P8), seven as B2 (P9-P15) and one as C1 level (P16). The instructions given to 
the students regarding L2 English essay were as follows: 
 
“Youth for Understanding organization is inviting texts from the young from different parts of 
the world to create a website depicting life around the globe. The topic of the text is 'Today's 
people only care about themselves and not about the whole community '- Please write a text in 
English presenting and arguing for your views and opinions. Write a title yourself.” 
 
For the L1 Finnish essay the students were instructed as follows:   
“Ruotsin kielen opetuksesta suomenkielisissä kouluissa käydään paljon keskustelua. Miksi 
ruotsin kieltä kannattaa opiskella? Ota kantaa ja pohdi, mitä etua ja hyötyä ruotsin osaamisesta 
on opinnoissa ja työelämässä. Perustele näkemyksesi. Kirjoita kantaaottava teksti Suomen 
Lukiolaisten liiton kampanjaan monipuolisesta kielitaidosta.” 
 
 21 
 
“Mandatory Swedish in Finnish schools has been a hot topic of conversation. Why do you think 
students should study Swedish? Write an argumentative text on the subject for The Union of 
Upper Secondary School Students in Finland, which campaigns for a versatile knowledge of 
languages. You should argue your opinions clearly and discuss what advantages learning 
Swedish might give to further studies and working life.” (my translation) 
 
The fluency of the writing process was measured by employing a set of variables which 
have been used by previous studies and, therefore, have shown to characterize well the 
underlying writing process. These included measures which can automatically be extracted 
from ScriptLog logfiles, namely the total number of typed characters (including spaces, letters 
and punctuation marks), total writing time, the total number of pauses, total pause time, average 
pause length and mean transition time. In addition, the length of the initial planning phase (from 
the beginning to the start of the second sentence), the length of the revision phase (time reserved 
for revising text after completing the first draft) and the number of revisions were calculated 
based on the LIN files. The original categorization of writer profiles by Van Waes and Schellens 
(2003) and the categories used by Mutta (forthcoming) were also mainly based on these 
variables. Furthermore, the measures of fluency (linear), fluency (product), burst and fluency 
during burst were applied as they have often been used in the writing fluency studies (e.g. 
Palviainen, Kalaja and Mäntylä 2012; Schoonen et al. 2003; Lindgren, Spelman Miller and 
Sullivan 2008). The measures were operationalized as follows: 
 Fluency (linear) = number of characters produced per minute 
 Fluency (product) = number of characters produced per minute in the final text 
 Burst = total number of typed characters / (total number of revisions + total number of 
pauses) 
 Fluency during Burst = total number of typed characters / (total writing time – total 
pausing time) 
Bursts were further examined by calculating separately P-Bursts, which in this study denotes 
the mean number of characters the writer produces between pauses, and R-Bursts, which refers 
to the mean number of revisions which writers perform during one burst. Despite the fact that 
these two notions have not been widely used in previous studies, they were incorporated into 
the current study in the hope of examining the concept of Burst more thoroughly. In regard to 
pauses, a threshold level of 2000 milliseconds was used to determine which interruptions could 
be considered as pauses during the writing process, and which could be excluded as so-called 
“technical pauses”, resulting from writers correcting typological errors. Furthermore, it enabled 
the comparison of results with the previous studies which have used the same time criterion 
 22 
 
(e.g. Mutta, forthcoming). Revisions included all the backspaces and deletions which writer 
performed during the writing process in order to edit the text. It has to be noted that some studies 
have included also insertions of characters to the notion of revision (e.g. Palviainen, Kalaja and 
Mäntylä 2012; Lindgren, Spelman Miller and Sullivan 2008). However, as it was difficult to 
determine when the writer was editing the text instead of producing it, insertions of the 
characters were left out from the current research. For future research, it might be useful to 
combine computer-based analysis with a retrospective think-aloud method, in which writers 
could verbalize their thoughts after the composing process to explain in which instances they 
edited the text and where they produced new pieces of text. This way, the individual writing 
processes would become more explicit also for research purposes. 
The burst refers to the average number of characters typed between pauses and/or 
revisions while fluency during burst to the speed of typing between pauses or revisions, which 
can both be seen as indicators of the general smoothness of the composition process. Similarly, 
the variable of mean transition time (TT) is related to the speed of writing, since it measures 
how fast the writer proceeds between letters in a word (Kowal 2014, 236). Mean transition time 
is related to the automaticity in writing, the assumption being that a writer with higher 
automaticity is able to retrieve language items more quickly and, as a result, produce text faster 
compared to a writer whose cognitive processed have yet to automatize (ibid.). 
4.3 Methods of Analysis 
Once all the values were collected from the participants’ personal ScriptLog files, the statistical 
analyses were performed by means of SPSS. First, the Shapiro-Wilk test was conducted in order 
to investigate whether the data were approximately normally distributed. The assumption of 
normality is a prerequisite for many statistical procedures, including parametric tests (Razali 
and Yap 2011, 21). The Shapiro-Wilk test was chosen to assess the normality, as it has been 
generally used with small sample sizes and it has been showed to be one of the best indicators 
of non-normality among the other tests (Razali and Yap 2011, 25; Shapiro, Wilk and Chen 
1968, 1371). Since some of the variables were non-normally distributed in the data, a 
nonparametric test was selected for the purpose of examining the correlations. Nonparametric 
tests are preferred in cases where the sample size is rather small and there is a possibility of 
marked departures from the normal distribution (Dörnyei 2007, 227; Şen 2017, 69). Therefore, 
to answer the research question number 1, the Spearman rank-order correlation coefficients 
were conducted to test the correlation of the fluency measures between L1 and L2. It was 
 23 
 
decided that not all the possible variables were tested, but those which were regarded as good 
indicators of overall written fluency. These variables included measures related to pausing 
behaviour (total pause time, the number and mean length of pauses, pause time in relation to 
the total typing time), speed of writing (fluency product, fluency linear), as well as the concept 
of burst (burst, P-burst, R-burst, fluency during burst) and revisions. Besides calculating the 
correlations for the group as a whole, they were also computed separately for groups with 
different proficiency levels in order to see whether the degree of correlation between L1-L2 
fluency measurements would be dependent on the L2 proficiency. Since the sample size was 
rather small and there were not enough participants for all the six CEFR levels, it was decided 
that levels A2 and B1 formed one group (n=8), while B2 and C2 formed another (n=8). This 
enabled comparing and testing two different proficiency groups. The effect sizes were 
interpreted by using Plonsky and Oswald’s (2014, 889) levels, where rs close to .25 is 
considered small, .40 medium, and .60 large, since they have been frequently applied in L2 
research. To examine the second research question, simple linear regression analyses were 
performed to those variables where correlations were found. By using regression analyses, one 
is able to look into which factors independently contribute to explaining the variance in the 
studied variable (Larson-Hall 2016, 224-225), and thus it suited well for the present research as 
well, as the purpose was to explain and predict the phenomenon profoundly. The proficiency 
level was also taken into account in the analysis by carrying out hierarchical multiple 
regressions, where the L1 fluency measure was added first, and the proficiency level next to 
see whether the explained variance in L2 measures would increase by adding a proficiency 
group. The assumptions of regression analysis, including linear relationship, normal 
distribution of errors, homoscedasticity, checking for outliers and multicollinearity (for 
hierarchical regression), were checked and met for these variables, so there was no need for 
transforming the data. 
After the quantitative description of the data, the writing processes of the participants 
were examined individually with the help of Van Waes and Schellens’ (2003) and Mutta’s 
(forthcoming) categorization of writer profiles. Although the variables used were mostly the 
same, there were some changes in interpreting the results. For instance, in the original Van 
Waes and Schellens’ categorization, the second-phase writers were characterised to have more 
revisions in their second phase compared to the first one. However, as none of the participants 
actually revised the texts more during the second phase, it was decided that in this group would 
belong those who revised clearly more than average during the second phase. Moreover, the 
present research did not examine the type, level nor the location of pauses. Instead, the total 
 24 
 
number of pauses and the mean length of them were calculated and they were then considered 
in relation to the total typing time. In the analysis, individuals’ values were compared to the 
mean values of the group in order to see which writer profile they would belong to. A qualitative 
description of the data also enabled a closer examination of different proficiency levels. This 
was necessary since due to the small number of participants, the quantitative analysis alone 
would not have been enough to demonstrate the possible differences between the groups. In the 
qualitative analysis of the proficiency groups, the participants of the same CEFR levels were 
grouped together and examined individually, and their values were then compared to the mean 
values of the group. By this, the present research aimed to get more insight into the possible 
differences of the proficiency groups. The results of the analysis will be presented next. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 25 
 
5 Fluency in Finnish Students’ Writing 
This section will start by presenting the descriptive statistics for all the variables in L1 and L2 
for the whole sample. It will then move on to reporting the results of the quantitative analysis, 
namely, the correlations between the fluency measures and regressions analyses. These are 
further complemented with a group level qualitative analysis. Lastly, the participants’ essays 
are examined individually and the fluency-related features of their writing processes in L1 and 
L2 are explored more profoundly. 
The descriptive statistics for the whole group in L1 and L2 are presented in Table 1.  
 
Table 1. Means (and Standard Deviations) of the Fluency Measures  
Measure L1 
(Finnish) 
M (SD) 
L2  
(English) 
M (SD) 
Total pause time (s) 327.75  
(14.15) 
530.51 
(162.98) 
The number of pauses 35.19 
(12.58) 
66.63 
(23.53) 
The mean length of pauses (s) 9.83 
(3.95) 
8.56 
(3.71) 
Pause time / typing time (%) 44.50 
(12.00) 
52.25 
(14.74) 
The number of revisions 329.69 
(222.59) 
396.88 
(332.35) 
Fluency (product) 99.64 
(27.37) 
65.27 
(27.25) 
Fluency (linear) 128.81 
(29.56) 
89.39 
(32.60) 
Burst 5.24 
(2.24) 
4.26 
(2.43) 
P-Burst 46.65 
(16.63) 
24.36 
(11.59) 
R-Burst 9.67 
(5.73) 
6.00 
(4.26) 
Fluency during Burst 236.54 
(45.78) 
188.38 
(40.82) 
The length of the initial planning phase (s)  27.94 
(22.92) 
95.81 
(147.53) 
The length of the revision phase (s) 82.00 
(119.27) 
114.81 
(129.63) 
Mean transition time (s) 0.17 
(0.04) 
0.19 
(0.06) 
 
 26 
 
As can be seen from the table, the participants seemed to be more fluent in their L1 overall, 
which is line with the previous findings (see e.g. see Silva, 1993 for a review of studies). The 
only exceptions were the mean length of pauses, which was longer in L1, and R-Burst, which 
indicated that the participants revised the L1 text more during one burst compared to L2 text. 
Although the pauses were longer in L1, the mean number of pauses was significantly lower 
compared to L2 texts and P-Bursts were longer, thus indicating that even though the participants 
might have allocated more time on thinking what to write next, once they started to write the 
process was smooth. Similarly, R-Burst sizes might also be reflected by the smoother writing 
process; when writing in L1, writers do not have to pay extra attention to the language 
structures, and therefore, they might easily make extra keystrokes by mistake, especially when 
typing fast with a keyboard. This hypothesis is also supported by the values of the mean 
transition time, which in L1 was shorter, meaning that the participants typed slightly faster in 
their L1. Overall, the mean transition time was short for almost all of the participants in both 
languages, indicating that they were all experienced users of a computer keyboard. The only 
exception in the group was P2, whose value in L1 was 0.3 and in L2 0.38. Hence, it seems that 
P2 could have been slightly less experienced with using a keyboard compared to others. 
However, overall in this study, the mean transition time did not provide any information on the 
possible individual differences in fluency, since most of the participants seemed to be within 
the range of normal individual variation in this variable. 
5.1 The Correlations between L1 and L2 Fluency Measures 
The first research question examined to which degree the L1 fluency measures correlate with 
the equivalent L2 fluency measures. The results of the Spearman’s rank order correlations are 
presented in Table 2. 
 
 
 
 
 
 
 
 
 27 
 
Table 2. Correlations between L1 and L2 Fluency Measures for the Whole Sample and for the 
Two Groups Separately 
 
Measure Groups 
combined 
Group A2-B1 Group B2-C2 
Total pause time .226 .476 -.024 
The number of pauses .418 .443 .708* 
The mean length of pauses .708* .905* .515 
Pause time / typing time .354 .635 .187 
The number of revisions .191 .381 -.071 
Fluency (product) .159 .310 -.095 
Fluency (linear) .229 .238 .167 
Burst .071 .190 .084 
P-Burst .491 .476 .714* 
R-Burst .037 .381 -.168 
Fluency during Burst .068 .214 -.214 
 Note * p <.05 
 
For the whole group, the results indicated only a weak and statistically not significant 
correlation for the majority of the measures. The only exception was the mean length of pauses, 
where the correlation was found to be strong and statistically significant (rs = .708, p = .002). 
The variables of P-Burst (rs = .491, p = .053) and the number of pauses (rs = .418, p = .107) 
also showed moderate correlations, although they did not quite reach the required statistical 
significance level. However, it has to be noted that as P-Burst was quite close to the required 
level even with a small number of participants, a larger number of data might already be enough 
the show a positive significant correlation between the L1 and L2 in this variable. Overall, based 
on the correlation analysis, it seems that there could potentially be a relation between the 
individual’s writing processes in L1 and L2 in regard to pausing behaviour, since measures 
which covered pauses showed the strongest correlations. The more complex measures which 
also entailed revisions in some form, on the other hand, did not seem to correlate as strongly in 
 28 
 
L1 and L2 writing processes. The possible explanations for this will be discussed in more detail 
in the following section. 
The results of the group-level analysis revealed surprisingly somewhat stronger 
correlations for the less proficient group, although the only statistically significant positive 
correlation was found in the mean length of pauses (rs = .905, p = .002). In the second group, 
on the other hand, the measures of number of pauses (rs = .708, p = .050) and P-Burst (rs = 
.714, p = .047) were strongly connected. Based on these results, it is difficult to say overall 
whether the L1 and L2 measures correlate stronger for one group over the other, since only a 
few of the measures showed statistically significant correlations.  
A closer qualitative analysis of the different proficiency groups seemed to corroborate 
these findings. For instance, in the A2 level (P1+P2), the writers’ total pausing time was above 
the group average, and the size of their P-Bursts was somewhat shorter compared to other 
participants. However, when these same writers were composing texts in their L1, they did not 
share the same characteristics. Although they still spent slightly above the average time pausing 
while writing, their P-Burst sizes varied. P1’s burst size was one the highest of the whole group 
in L1, while P2 still produced very few characters between the pauses compared to others. 
Similarly, there was individual variation in other proficiency levels. For instance, one of the B2 
level participants, P9, had very similar writing processes in L1 and L2. He paused quite many 
times for a short period of time in both languages, which also made the P-Bursts and R-Bursts 
in both languages relatively short. On the contrary, P10 from B2 level had a very dissimilar 
composing processes, as he revised the L2 text the least compared to others, making thus the 
R-Burst size small, whereas in L1, his R-Bursts were the longest of the whole group. Therefore, 
based on these findings, it seems that there might be some differences between the groups with 
different proficiency levels. However, the findings do not provide any conclusive answer as to 
whether the connections between L1 and L2 fluency measures actually increase when learners 
become more proficient in L2. The fluency characteristics of the individual writers in L1 and 
L2 are discussed more in section 4.3. Before that, the connections between the L1 and L2 
fluency measures are further explored with simple and hierarchical regression analysis. 
5.2 Predicting L2 Fluency 
The second research question examined how much of the variability in L2 measures could be 
predicted on the basis of the equivalent L1 fluency measures. To address this question, simple 
linear regressions were conducted on those measures, where statistically significant correlations 
 29 
 
were found in the previous section. These measures included the mean length of pauses, the 
number of pauses and P-Burst. After this, the possibility of proficiency level increasing the 
explained variance was examined with hierarchical multiple regression analysis. The results of 
the simple regression analysis are presented in Table 3.  
 
Table 3. Results of the Simple Regression Models  
Measure B Std. 
error 
Beta F(1,14) p Adj. 
R2 
The mean length of pauses .591 .195 .630 9.223 .009 .354 
The number of pauses .675 .466 .361 2.094 .170 .068 
P-Burst .430 .147 .617 8.604 .011 .336 
 
As can be seen from the table, the models predicting L2 measures from L1 measures were 
significant for the mean length of pauses, where the L1 measure accounted for 35% of the 
variance in L2 measure, and for P-Burst, where the L1 measure explained 34% of the 
variability. Hierarchical regression analyses revealed the proficiency level not to be a 
statistically significant predictor in the models, despite the fact the explained variance increased 
in two of the variables. For P-Bursts, including the proficiency level in the model increased the 
explained variance from 34% to 47% (Adj. R2 = .471, F (2,13) = 7.676, p = 0.052). Similarly, 
for the number of pauses, the explained variance increased from 7% to 21% (Adj. R2 = .206, F 
(2,13) = 2.954, p = .087). However, the increase was not statistically significant, as was not the 
original model where only the effect of L1 measures was tested. In regard to the mean length 
of pauses, adding the proficiency level as a predictor did not increase the explained variance. 
Overall, the results of the regression analyses showed that L2 fluency measures could be 
predicted from L1 measures to some extent, and therefore they support the findings of the 
correlational analyses. However, the role of the proficiency level in L1-L2 connections remains 
unclear, since it did not turn out to be a statistically significant predictor in the models in this 
case. 
 30 
 
5.3 Qualitative Analysis: Individual Writer Profiles in L1 and L2 
The third research question focused on exploring the possible individual differences in L1 and 
L2 fluency. For this purpose, the essays were analysed individually for their fluency 
characteristics with the help of the categorization of different writer profiles. 
 
P1 (L2 proficiency level A2): 
P1 is a great example of an initial planner in his L2. His initial planning time exceeded 
significantly the average time in L2 writing. Moreover, although P1 paused relatively few times, 
he spent around 75% of the total writing time pausing, and the average length of pauses was 
markedly longer compared to others. Based on the revision strategies, P1 could also be 
classified in the same group, as he revised the text very little, especially during the second phase 
of writing. Similarly, in his L1 essay, P1’s pauses were relatively long and the number of overall 
revisions small, although in L1 he did not devote as much time to initial planning as in his L2 
essay. It is also noteworthy, that P1’s total typing time was shorter than the average in both, L1 
and L2 essays. Overall, the writing processes in L1 and L2 seemed to be very similar in regard 
to pausing and revision behaviour. 
 
P2 (L2 proficiency level A2): 
P2 differed slightly in his L1 and L2 writing processes. In his L2 essay, P2 had a long initial 
planning phase, he paused very often, and the total pause time was above average, although the 
mean length of pauses stayed below average. Furthermore, P1 made very few revisions overall. 
Thus, he could be regarded mainly as an initial planner in L2. In L1 writing, however, P2 had 
mainly characteristics of a non-stop writer. He revised the text very little and paused relatively 
few times, although the total pause time and mean length of pauses showed somewhat average 
values. Moreover, he did not allocate much time to initial planning and completed the task fairly 
quickly, although it has to be noted that he also typed the least during the assignment, which 
might have contributed on its part to the short completion time. P2’s pausing and typing 
behaviour might have also been influenced by slightly longer mean transition time compared 
to others, indicating that he might not be as experienced with using a keyboard.  
 
P3 (L2 proficiency level B1):  
P3 could be categorized as a non-stop writer in both languages. In L2, he completed the writing 
task more quickly than others and spent only little time on initial planning. In addition, he 
 31 
 
paused few times and the mean length of pauses and the total pause time were shortest compared 
to all the other participants. Similarly, in his L1 writing, P3’s initial planning phase, mean length 
of pauses and the total time spent pausing all clearly laid lower than the average. As for his 
revision behaviour, P3 seemed to have similar strategies in L1 and L2, since he revised both 
texts very little, especially during the second phase of writing. 
 
P4 (L2 proficiency level B1): 
In L2, P4 revised the text more than average focusing on the first phase. In his L1 writing, P4 
revised less, but all the revisions were made in a similar manner during the first phase. The 
number of pauses and total pause length were quite close to the average values in both 
languages, although the mean length of pauses was slightly above average. The length of the 
initial phase did not surpass the average level in L2 writing, while in L1 it did. Overall, P4 could 
be categorized as a fragmentary first-phase writer in L2 and as an initial planner in L1. 
 
P5 (L2 proficiency level B1): 
P5 could be regarded as a fragmentary first-phase writer in his L2. Although the initial planning 
time was relatively long for the profile in question, P5 revised the text a lot during the first 
phase, and he paused many times for a relatively short period, which made the process highly 
fragmented. In his L1 writing, although P5 did not revise the text as much as in L2, the revisions 
were also centred on the first phase and the pauses were relatively short. However, as the 
number of pauses, total pause time and total typing time were all below average, P5 had mainly 
the characteristics of a non-stop writer in his L1. 
 
P6 (L2 proficiency level B1):  
In L2, P6 revised the text significantly more compared to other participants. A closer 
examination of the writing process revealed that P6 had accidentally used a copy-paste –
shortcut in the middle of writing, which pasted the earlier essay written in Swedish inside the 
text. He had to remove the Swedish text before continuing the essay, which naturally left a trace 
to the programme. Therefore, the values related to P6’s revision behaviour might be slightly 
erroneous. Despite this incident, P6 revised the text a lot also otherwise. He also paused quite 
often for a short period of time, for which he could be considered mainly as a fragmentary first-
phase writer. In L1, P6’s values did not differ significantly from average values, with the 
exception of mean length of pauses, which was shorter and initial planning phase, which was 
longer than average. Consequently, P6 was predominantly an initial planner in L1. 
 32 
 
P7 (L2 proficiency level B1): 
In L2, P7 paused more than any other participant did, although the length of pauses remained 
short. In L1, on the contrary, the number of pauses was slightly below average, and the length 
of them was not considerably shorter compared to others. Despite the differing pausing 
behaviour, P7 had similar revision strategies. Overall, the number of revisions remained quite 
close to the mean values, and they were mainly conducted during the first phase. Additionally, 
P7’s allocation of time to the initial planning was below average in both languages. However, 
due to the differing pausing behaviour, P7 did not share the same writer qualities in his L1 and 
L2. In L2, he was more of a fragmentary first-phase writer, whereas in L1, he was closer to an 
average writer. 
 
P8 (L2 proficiency level B1): 
In L2, P8 conducted very few revisions overall, and they occurred mostly during the first phase. 
Other variables seemed to be rather close to the average values, apart from the mean length of 
pause, which exceeded the average length, making also the total pause slightly above average. 
However, due to the small number of revisions and the fact that many of the other variables 
stayed slightly below the mean values, P8 could be mainly regarded as a non-stop writer in his 
L2. On the contrary, in L1 writing, P8 revised significantly more than average, although the 
revisions were also made in a similar manner than in L2, i.e. during the first phase of writing. 
In other variables, P8 seemed to have values somewhat close to the average values of the whole 
group. In general, P8 did not seem to belong clearly to any of the profiles in his L1, but based 
on revision behaviour, he could be mostly regarded as a fragmentary first-phase writer. 
 
P9 (L2 proficiency level B2): 
P9 seemed to have similar pausing behaviour in both, L1 and L2, composing processes. He 
paused many times, while the average length of pauses remained short. Moreover, he did not 
allocate much time to initial planning in either of the languages. He revised both texts quite the 
average amount, and after the completion of the first draft, he hardly altered the texts at all. 
Thus, P9 had the characteristics of a fragmentary first-phase writer in both languages. 
 
P10 (L2 proficiency level B2): 
P10 revised the least the L2 essay compared to others. In addition, all the values related to 
pausing behaviour (total pausing time, total typing time, the total number of pauses, mean pause 
length) were slightly below the average, with the exception of initial planning time, which 
 33 
 
exceeded the average length. Despite the long initial planning phase, P10 could be classified 
predominantly as a non-stop writer in his L2. On the contrary, in L1, P10 was one of the 
participants who revised the most on their texts during the writing process. In addition, all the 
pausing variables exceeded the average values and the initial planning phase was the longest of 
the group. As a result, P10 was clearly an initial planner in his L1. 
 
P11 (L2 proficiency level B2): 
P11 was quite close to the average values when it comes to the total pausing time and the 
number of revisions. He paused fewer times than average, although the mean length of pauses 
was above the average. He did not spend much time on initial planning. P11 was, therefore, 
quite hard to categorize, as he could be regarded as somewhat average writer with second-phase 
writers’ pausing behaviour. Interestingly, in L1, P11 has similar values compared to his L2. He 
paused less than average, while the mean length of pauses was long. The total pausing time and 
the number of revisions were similarly close to the average values, and he did not allocate much 
time to initial planning in L1 either. Hence, it could be argued that he shared the same writing 
profile in both languages, albeit it was difficult to determine which profile. 
 
P12 (L2 proficiency level B2): 
P12 was a fine example of a non-stop writer in both languages. He paused few times, revised 
the texts less than average, and began to write the texts without devoting much time to 
preliminary planning. Moreover, the mean length of the pauses and the total pause time stayed 
clearly below the average. 
 
P13 (L2 proficiency level B2): 
P13’s composing process was characterised by a large number of revisions made during the 
second phase of writing in both, L1 and L2 essays. When it comes to pausing behaviour, the L1 
and L2 writing differed slightly. In L2, the number of pauses stayed below average, whereas 
the mean length of them was above the average values. In contrast, P13 paused more than 
average in L1, while the mean length of pauses was below average. The initial planning time 
also varied between the two languages. Therefore, it cannot be inferred that the writing profiles 
are exactly the same ones in L1 and L2, as there are some differences in regard to pausing 
behaviour. Nevertheless, they are not entirely dissimilar either, since the revision behaviour 
was identical. Solely based on the revision behaviour, he could be classified as a second-phase 
writer. 
 34 
 
 
P14 (L2 proficiency level B2): 
P14 revised more than average in L2 and many of the revisions were made during the revision 
phase. As for his pausing behaviour, he did not spend much time on initial planning and he 
paused more than average, while the mean length of pauses stayed close to the average level. 
Despite this, he could be regarded mainly as a second-phase writer in L2. In L1, P14 similarly 
paused more than average, and the mean length of pauses was below average. He also revised 
the text more than average overall, although unlike in L2, he did not alter the text after finishing 
the first draft. Hence, P14 had the characteristics of a fragmentary first-phase writer in his L1. 
 
P15 (L2 proficiency level B2): 
P15 clearly shared similar writing processes in his L1 and L2 writing. After the completion of 
the drafts, he spent a long time revising the texts and made a significantly higher number of 
revisions during the second phase compared to others. When analysing the type of revisions, it 
turned out that many of them were made at a level above the word. In regard to pausing 
behaviour, P15 spent roughly the same amount of time pausing (~60%) in both of the languages. 
Although he clearly did not allocate much time on initial planning in either of the languages, 
the mean length of pauses surpassed the average level in both, L1 and L2. Despite the relatively 
short initial planning phase, P15 is a good example of a second-phase writer in both languages. 
 
P16 (L2 proficiency level C1): 
In L2 writing, P16 had mainly the characteristics of a non-stop writer. Although he revised the 
text more than average, he did not pause very often, and total time spent pausing clearly laid 
below average. Moreover, P16 devoted little time for initial planning and he composed the 
essay in a shorter period of time. In his L1 essay, on the contrary, P16’s pauses were long, 
making also the total pausing time longer, and he spent quite some time on initial planning. 
However, similarly to his L2 essay, he revised the L1 text quite a lot, although many of the 
revisions were realized after completing the first draft. Therefore, P16 could be categorized as 
a second-phase writer in his L1 writing. 
 
Comparison of different writer profiles 
Overall, the results of the qualitative analysis showed that seven out of 16 participants shared 
the same writer profile in their L1 and L2 composing processes. In addition, there were cases, 
such as P14, who seemed to possess similar writer characteristics in some areas, but were still 
 35 
 
classified into two separate profiles due to the categorization. From those who shared the same 
profile in L1 and L2, one participant had the characteristics of an initial planner, two were 
clearly non-stop writers, while two were second-phase writers. In addition, there was one person 
who was classified as a fragmentary first-phase writer and one who had mostly values of an 
average writer. Thus, it seems that all the writer profiles were somewhat evenly presented in 
the data, also within those who shared the same profile in their L1 and L2. In regard to different 
proficiency levels, the groups differed slightly in their writer profiles. While there were some 
initial planners among the G1 participants (2/8), none of the G2 members were classified in this 
profile. There were also more fragmentary first-phase writers in G1 (4/8) compared to G2 (1/8). 
The G2 members, on the contrary, had many second-phase writers (4/8), whereas none of the 
G1 members had similar characteristics. One possible explanation for the results could be that 
the more proficient group members were also more motivated in general to write the essay, and 
thus they devoted some time after the completion of the draft to reading the text and making 
sure that language was correct. Therefore, it could be that they also made more revisions to the 
text during the second-phase compared to G1 members. Moreover, the fact that initial planners 
were found only among the less proficient group members is interesting. It could be that this 
profile is somehow linked to the proficiency level of the writer, indicating that less proficient 
writers need more time at the beginning of the writing process to think and plan, as their level 
in L2 is not strong enough for them to start writing immediately first things that come to their 
mind. However, as there does not seem to exist any previous results confirming these 
hypotheses, they should be further investigated in longitudinal studies focusing on the 
development of individual writers. Interestingly, non-stop writers were found in both of the 
groups, two participants in G1 and three in G2, although the characteristics of non-stop writers 
are linked to a fluent writing process in many ways, and thus one might expect to find it to be 
a more prominent profile within the more proficient writers. However, based on these results, 
it seems there is no single profile in particular which could be linked to features of fluency. 
These results are also in line with Mutta’s (forthcoming) findings, which showed that there was 
not a link between different writer profiles and the text quality. 
Interestingly, when examining participants in a group level, it seems that the G2 members 
shared the same profile more often (5/8) compared to the less proficient group G1 (2/8). 
Although these results are slightly contradictory to the outcome of the correlation analyses, they 
could indicate that the relationship between L1 and L2 fluency shifts with developing 
proficiency, as some of the previous studies have suggested (e.g. Huensch and Tracy-Ventura 
2017). The implications of the results will be discussed in more detail next. 
 36 
 
6 Discussion 
The goal of the present study was to investigate the extent to which writers’ L1 fluency 
characteristics contribute to fluency in L2 productions. The rank-order correlations 
demonstrated positive correlations between the majority of L1 and L2 fluency measures ranging 
from weak to strong effects both within the whole group and for the two groups separately, yet 
only a subset of these correlations was statistically significant. The strongest correlation for the 
whole group was found in the mean length of pauses. Moreover, the measures of the number of 
pauses and P-Burst also correlated significantly depending on the group. The regression 
analyses on these measures revealed that L2 measures could be predicted from the L1 measures 
in the mean length of pauses and P-Burst, although the proficiency level had a minor role 
compared to the findings of the correlational analysis. A further individual examination of the 
essays also corroborated the results of the quantitative part of the study, as they indicated that 
participants’ writer profiles are connected to some extent, reinforcing thus the connections 
between the L1 and L2 fluency. These results support the findings of the earlier studies, which 
have found a relation between L1 and L2 oral fluency measures, especially in temporal 
measures of fluency (Peltonen 2018; De Jong et al. 2015). Similarly, in the current study, the 
strongest correlations between the L1 and L2 measures were found in variables related to 
writers’ pausing behaviour. Moreover, it is noteworthy, that these correlations were statistically 
significant even with a small sample size, indicating that there exists a strong connection in 
learners’ L1 and L2 pausing behaviour, irrespective of the mode of how language is produced 
(oral or written). The variables related to revisions, on the other hand, did not seem to correlate 
much in the quantitative part of the study, although the closer examination of the essays showed 
there to be some similarities between individual writers’ revision behaviour in some cases. 
Overall, the results of the present study seem to be in line also with those few writing fluency 
studies (e.g. Yang and Sun 2015), which have attested a link to some degree between the CAF 
measures of multilingual writers’ different languages or the relation between L1 and L2 writing 
proficiency (e.g. Sasaki and Hirose 1996; Schoonen et al. 2003; Schoonen et al. 2011). 
Moreover, as Chenoweth and Hayes (2001) demonstrated in their study that the length of P-
Bursts was related to writing fluency, the fact that the variable correlated between L1 and L2 
in this study supports the hypothesis that the L1 fluency behaviour somewhat contributes to the 
L2 fluency. 
In addition to exploring the L1-L2 writing fluency within the whole sample group, the 
role of the proficiency level in L1-L2 connections was also examined. The results revealed there 
 37 
 
to be some differences between the different proficiency groups in how L1 and L2 fluency are 
connected. Nevertheless, they did not provide a clear answer as to whether the connections were 
stronger for one group over the other, since the findings of the quantitative and qualitative parts 
were slightly contradictory to one another. These results are not entirely in line with previous 
studies (see e.g. Peltonen 2018; Huensch and Tracy-Ventura 2017), which have generally 
shown a stronger relationship for the more proficient participant group. There might be several 
reasons for the differing results. Firstly, due to the small sample group, all the proficiency levels 
from A2-C2 did not have enough participants, which meant that some of the proficiency levels 
had to be combined for the correlational analyses. This might have affected the outcome, since 
combining levels could have evened out some of the possible differences between the two 
extremes. Secondly, the proficiency level of the writers was evaluated based solely on one 
essay, which might not be a valid indicator of the true competence of the participants. Naturally, 
it is more accurate than for instance school level or the age of the learner, which many of the 
earlier studies have used, but for future research, it should be combined with a more profound 
and extensive assessment of the proficiency. Moreover, the group as a whole seemed to be 
rather homogenous, for the majority of the participants belonged to either B1 or B2 level. 
Therefore, the differences in L2 competence and fluency might not be as prominent within this 
sample as within a group including a larger number of A2 and C2 level participants. It could 
also be that that the more proficient writers happened to have lower verbal skills in general, 
thus also affecting the results on its part, as Chenoweth and Hayes’ (2001) study demonstrated. 
Taking all the above-mentioned issues into account, it is possible that the quantitative analysis 
of the fluency aspects might not be entirely valid when it comes to the role of the proficiency 
level. Therefore, the results of the qualitative analysis of the writing processes might illustrate 
the differences between the groups more accurately. More research on the subject is, however, 
needed. 
In regard to the variables used in the study, it seemed that many of the reliable indicators 
of fluency, such as Fluency (product, linear), Burst, R-Burst and Fluency during Burst, 
correlated surprisingly little between L1 and L2 in the present research. The notions of Burst 
and R-Burst included the number of revisions as part of the formula, which might have impacted 
the results, since revisions in general did not seem to correlate in the quantitative analysis. 
However, it does not explain why Fluency (product, linear) and Fluency during Burst produced 
such low effects, while P-Burst correlated strongly. After all, these variables have been found 
to be valid indicators of fluency and thus have been used in many previous studies (see e.g. 
Palviainen, Kalaja and Mäntylä 2012; Lindgren, Spelman Miller and Sullivan 2008). There is, 
 38 
 
thus, a need for further research on these variables, and they should be studied with a larger 
number of data in order to get valid results. However, in addition to these, it is important to 
include also the simple variables in the analysis, since they might show more clearly the 
possible similarities and differences between individuals.  
The categorization of Van Waes and Schellens (2003) was indispensable in helping to 
classify the participants’ writing characteristics. Nevertheless, in some cases, it was difficult to 
categorize the participants if they possessed characteristics of various different profiles. In those 
cases, the participants had to be classified based on their prevailing features, despite the fact 
they might not share all the described characteristics of the profile in question. One possible 
explanation for the difficulty of classification in this study might be related to the fact that the 
participants were so-called digital natives, who have used technical devices from their early 
childhood onwards. Their typing skills, and thus also writer profiles, might differ from Van 
Waes and Schellens’ categorization, which was, after all, created almost 20 years ago. 
Therefore, in future studies, it might be useful to explore and compare also other possible writer 
profile categorizations (see e.g. Torrance, Thomas and Robinson 2000), and decide which might 
work best for the data in question. The more in-depth qualitative analysis provided more insight 
into the individual writing processes, especially regarding the pausing and revision behaviour, 
which in this study seemed to correlate rather differently. As can be seen from the results of the 
qualitative examination, although in some cases learners revised the texts in a similar manner 
(e.g. P1, P3, P7), there were also individuals whose revisions strategies were the total opposite 
of one another in L1 and L2 (e.g. P4, P8, P10). There thus seems to be a bit more variation 
between individuals’ L1 and L2 in revisions compared to pauses. A possible explanation for 
the results could be that revision behaviour is simply not a feature that learners share in their 
L1 and L2 in writing, at least not as strongly compared to pausing behaviour. However, this 
hypothesis needs to be further studied on L2 learners of various ages and from diverse language 
backgrounds before one can make any definite conclusions. The differing correlational results 
might also be related to different writing and revision processes in L1 and L2, as Lindgren, 
Spelman Miller and Sullivan (2008) found out in their study. In other words, it could be that 
writers’ revision strategies and processes differ from one another when writing L1 compared to 
L2, while pausing strategies remain similar between the languages. However, this also needs to 
be studied further. Moreover, it has to be pointed out that the notion of revision was defined 
somewhat differently in the present study compared to some of the earlier studies, which could 
have affected the outcome as well. A different definition of the term, or even different 
categorization of writer profiles, might yield differing results. Overall, the findings of the 
 39 
 
qualitative part of the study suggested that writer profiles were connected to another in 
individuals’ L1 and L2, corroborating thus also Mutta’s (2007) results. 
Since the present research was one of the first ones to examine written data to find out 
L1-L2 connections, there were not any previous findings which could have worked as a 
reference point for the results. Therefore, the results had to be mostly compared on spoken 
fluency studies which have used a similar research design, although, as mentioned in section 
2.2, the constructs of speaking and writing require different cognitive processes, and thus might 
not be comparable to one another. However, since the findings of the present study somewhat 
corroborated the results obtained from similar oral fluency studies, it seems that these two 
modes are not entirely unconnected. In future studies, it might thus be useful to study both, 
writing and speaking fluency features, from the same participants in order to compare the two 
modes on an individual level and find their possible differences. Suggestions for other future 
studies are discussed next, along with the final conclusions. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 40 
 
7 Conclusion 
The present study examined the relationship between L1 and L2 fluency in writing among 
Finnish learners of English by using a mixed-methods approach, where an in-depth qualitative 
analysis of the essays was employed together with quantitative statistical methodologies. Based 
on the findings, it can be concluded that L1 and L2 writing fluency are connected to some 
degree, although the relationship seems to be rather complex and depend on various different 
factors, such as the proficiency level of the learner. Therefore, the findings also highlight the 
importance of employing diverse research methodologies, both quantitative and qualitative, in 
exploring the connections, for they provide more insights into this complex relationship. In 
future investigations, the computer-based analysis should be complemented with learners’ own 
retrospective thoughts, since they provide more detailed and accurate information on the writing 
processes. Although keystroke recording tools, such as ScriptLog, are valuable in giving 
quantitative information on writers’ behaviour, they do not uncover what writers are actually 
thinking while pausing or revising their texts. 
Besides the lack of complementary think-aloud procedures, the present research has also 
other limitations which should be acknowledged. Firstly, there is naturally the question of 
diverse motivational and affective factors, which can always affect the outcome when studying 
humans. In this case, the participants wrote altogether three essays consecutively during a 90-
minute time period. Naturally, the lack of concentration and motivation, and even the fact that 
the writing task was a so-called low stakes assignment, might have affected the production 
process, especially during the last essay. However, the order of the essays was well-thought, 
since the L1 essay was left out to be the last one, thus eliminating the risk that the participants 
might translate or transfer the ideas directly from L1 to L2. Due to the small scope of the present 
research, the essays written in Swedish were left out from the analysis. The connections 
between the possible L3 fluency to L2s and L1s should, nevertheless, be investigated more 
profoundly in the future. Secondly, one of the limitations of the present study was the small 
number of data, consisting mainly of females, and the fact it was conducted as a single task. 
Future research could investigate the issue more thoroughly by using a larger number of 
participants with diverse language backgrounds and examine the L1-L2 connections 
longitudinally, so that the effect of the proficiency level in the process becomes clearer. 
Moreover, instead of using only argumentative texts, learners could write different types of 
essays, e.g. descriptive or narrative, so that the task itself does not become an obstacle to 
anyone.  
 41 
 
Despite the possible limitations, the results of the study provided valuable insight into the 
complex relationship between L1 and L2 writing fluency, which has long been an 
underresearched subject compared to the spoken fluency studies. The findings have also 
significant implications especially for pedagogical contexts. In case L1 writing fluency does 
play a role in explaining aspects of L2 writing fluency, like the findings of the present study 
suggest, teachers could take this into account when assessing the fluency of the texts or when 
facilitating learners’ writing processes to become more fluent. The learners themselves are also 
able to utilize this information in order to become aware of their individual writing processes. 
Moreover, and perhaps of more importance, the results of this study can work as a starting point 
for more extensive investigations of the L1-L2 connections, especially for writing contexts 
involving typologically different language pairs. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 42 
 
References 
Chenoweth, N. Ann, John R. Hayes. 2001. “Fluency in Writing: Generating Text in L1 and 
L2.” Written Communication 18, no. 1: 8098. Accessed 19 April 2019. 
http://dx.doi.org/10.1177/0741088301018001004. 
Council of Europe. 2001. “Common European Framework of Reference for Languages: 
Learning, teaching, assessment.” Accessed 4.7.2019. 
https://www.coe.int/en/web/portfolio/the-common-european-framework-of-reference-
for-languages-learning-teaching-assessment-cefr-. 
De Jong, Nivja H, Rachel Groenhout, Rob Schoonen, and Jan H Hulstijn. 2015. “Second 
Language Fluency: Speaking Style or Proficiency? Correcting Measures of Second 
Language Fluency for First Language Behaviour.” Applied Psycholinguistics 36, no. 2: 
223243. Accessed 19 April 2019. http://dx.doi.org/10.1017/S0142716413000210. 
Derwing, Tracey M., Murray J. Munro, Ronald I. Thomson, and Marian J. Rossiter. 2009. 
“The Relationship between L1 Fluency and L2 Fluency Development.” Studies in 
Second Language Acquisition 31, no. 4: 533557. Accessed 19 April 2019. 
http://dx.doi.org/10.1017/S0272263109990015. 
Dörnyei, Zoltán. 2007. Research Methods in Applied Linguistics. Oxford: Oxford University 
Press. 
Ellis, Rod, and Gary Barkhuizen. 2005. Analysing Learner Language. Oxford: Oxford 
University Press. 
Fillmore, Charles J., Daniel Kempler, and William S-Y. Wang. 1979. Individual Differences 
in Language Ability and Language Behaviour. New York: Academic Press. 
Hall, Chris. 1990. “Managing the Complexity of Revising across Languages.” TESOL 
Quarterly 24, no. 1: 4360. Accessed 2 July 2019. http://dx.doi.org/10.2307/3586851. 
Hang, Du. 2013. “The Development of Chinese Fluency During Study Abroad in China.” 
Modern Language Journal 97, no. 1: 131143. Accessed 19 April 2019. 
http://dx.doi.org/10.1111/j.1540-4781.2013.01434.x.  
Hayes, J. R., and L. S. Flower. 1980. "Identifying the organization of writing processes." In 
Cognitive Processes in Writing, edited by L. W. Gregg and E. R. Steinberg, 3-30, 
Hillsdale, NJ: Lawrence Erlbaum.  
Huensch, Amanda, and Nicole Tracy-Ventura. 2017. “Understanding Second Language 
Fluency Behaviour: The Effects of Individual Differences in First Language Fluency, 
Cross-linguistic Differences, and Proficiency over Time.” Applied Psycholinguistics 38, 
no. 4: 755785. Accessed 19 April 2019. 
http://dx.doi.org/10.1017/S0142716416000424. 
Kaufer, Davis S., John R. Hayes, and Linda Flower. 1986. "Composing Written Sentences." 
Research in the Teaching of English 20, no. 2: 121149. Accessed 5 July 2019. 
https://www.jstor.org/stable/40171073. 
Kim, Youyjin, and Nicole Tracy-Ventura 2013. “The Role of Task Repetition in L2 
Performance Development: What Needs to Be Repeated During Task-based 
Interaction?” System 41, no. 3: 829840. Accessed 26 September 2019. 
http://dx.doi.org/10.1016/j.system.2013.08.005. 
Kobayashi, Hiroe, and Carol Rinnert. 2013. “L1/L2/L3 Writing Development: Longitudinal 
Case Study of a Japanese Multicompetent Writer.” Journal of Second Language Writing 
22, no. 1: 433. Accessed 14 July 2019. http://dx.doi.org/10.1016/j.jslw.2012.11.001.  
Kowal, Iwona. 2014. “Fluency in Second Language Writing: a Developmental Perspective.” 
Studia Linguistica Universitatis Iagellonicae Cracoviensis 131: 229246. Accessed 8 
May 2019. http://dx.doi.org/10.4467/20834624SL.14.013.2321. 
 43 
 
Larson-Hall, Jenifer. 2016. A Guide to Doing Statistics in Second Language Research Using 
SPSS and R. New York: Routledge, Taylor & Francis Group. EBSCOhost. 
Lennon, Paul. 1990. “Investigating Fluency in EFL: A Quantitative Approach.” Language 
Learning 40, no. 3: 387417. Accessed 12 April 2019. http://dx.doi.org/10.1111/j.1467-
1770.1990.tb00669.x.  
Lindgren, Eva, Kristyan Spelman Miller, and Kirk P.H. Sullivan. 2008. “Development of 
Fluency and Revision in L1 and L2 Writing in Swedish High School Years Eight and 
Nine.” ITL International Journal of Applied Linguistics 156: 133151. Accessed 2 July 
2019. http://dx.doi.org/10.2143/ITL.156.0.2034428. 
Maclay, Howard, and Charles E. Osgood. 1959. “Hesitation Phenomena in Spontaneous 
English Speech.” Word 15, no. 1: 1944.  Accessed 19 April 2019. 
http://dx.doi.org/10.1080/00437956.1959.11659682. 
Manchón, Rosa M, Julio Roca de Larios, and Liz Murphy. 2007. “A Review of Writing 
Strategies: Focus on Conceptualizations and Impact of the First Language.” In 
Language learner strategies: thirty years of research and practice, edited by Andrew D. 
Cohen and Ernesto Macaro, 229250. Oxford: Oxford University Press. 
Manchón, Rosa M. 2013. “Writing”. In Psycholinguistics of Bilingualism, edited by François 
Grosjean and Ping Li, 100115. Chicester: John Wiley & Sons. 
Mutta, Maarit. 2007. “La Conscience Métapragmatique et L’attitude Métacognitive 
Épistémique des Scripteurs Universitaires: la Révision de Texte en Temps Réel. 
[Metapragmatic Consciousness and Epistemic Stance of University Writers: The Case 
of Online Revision]. Dissertation, University of Turku. 
. 2019 in press. “L2 Fluency and Writer Profiles.” In Fluency in L2 Learning and Use, 
edited by Pekka Lintunen, Maarit Mutta and Pauliina Peltonen. Bristol: Multilingual 
Matters.  
Navés, Teresa, Immaculada Miralpeix, and M. Luz Celaya. 2005. “Who Transfers More… 
and What? Crosslinguistic Influence in Relation to School Grade and Language 
Dominance in EFL.” International Journal of Multilingualism 2, no. 2: 113134. 
Accessed 12 July 2019. http://dx.doi.org/10.1080/14790710508668380. 
Olkkonen, Sanna. 2017. “Second and Foreign Language Fluency from a Cognitive 
Perspective: Inefficiency and Control of Attention in Lexical Access.” Dissertation, 
University of Jyväskylä. 
Ong, Justina, and Lawrence Jun Zhang. 2010. “Effects of Task Complexity on the Fluency 
and Lexical Complexity in EFL Students’ Argumentative Writing.” Journal of Second 
Language Writing 19, no. 4: 218233. Accessed 14 July 2019. 
http://dx.doi.org/10.1016/j.jslw.2010.10.003. 
Palviainen, Åsa, Paula Kalaja, and Katja Mäntylä. 2012. “Development of L2 Writing: 
Fluency and Proficiency.” AFinLA-E: Soveltavan Kielitieteen Tutkimuksia 4: 4759. 
Accessed 19 April 2019. https://journal.fi/afinla/article/view/7037. 
Peltonen, Pauliina. 2018. “Exploring Connections Between First and Second Language 
Fluency: A Mixed Methods Approach.” The Modern Language Journal 102, no. 4: 
676692. Accessed 19 April 2019. http://dx.doi.org/10.1111/modl.12516. 
Plonsky, Luke, and Frederick L. Oswald. 2014. “How Big Is “Big”? Interpreting Effect Sizes 
in L2 Research.” Language Learning 64, no. 4: 878912. Accessed 15 September 2019. 
http://dx.doi.org/10.111/lang.12079. 
Razali, Nornadiah Mohd, and Bee Wah Yap. 2011. “Power Comparisons of Shapiro-Wilk, 
Kolmogorov-Smirnov, Lilliefors and Anderson-Darling Tests.” Journal of Statistical 
Modeling and Analytics 2, no.1: 21-33. Accessed 8 November 2019. 
https://bit.ly/2NTJJeZ. 
 44 
 
Roca de Larios, Julio, Rosa M. Manchón, and Liz Murphy. 2006. “Generating Text in Native 
and Foreign Language Writing: A Temporal Analysis of Problem-Solving Formulation 
Processes.” Modern Language Journal 90, no. 1: 100114. Accessed 2 July 2019. 
http://dx.doi.org/10.1111/j.1540-4781.2006.00387.x. 
Sasaki, Miyuki, and Keiko Hirose. 1996. “Explanatory Variables for EFL Students’ 
Expository Writing.” Language Learning 46, no. 1: 137168. Accessed 12 July 2019. 
http://dx.doi.org/10.1111/j.1467-1770.1996.tb00643.x.  
Sasaki, Miyuki. 2000. “Toward an Empirical Model of EFL Writing Processes: An 
Exploratory Study.” Journal of Second Language Writing 9, no. 3: 259291. Accessed 
12 July 2019. http://dx.doi.org/10.1016/S1060-3743(00)00028-X. 
Schmidt, Richard. 1992. “Psychological Mechanisms Underlying Second Language Fluency.” 
Studies in Second Language Acquisition 14, no. 4: 357385. Accessed 12 April 2019. 
http://nflrc.hawaii.edu/PDFs/SCHMIDT%20Psychological%20mechanisms%20underly
ing%20second%20language%20fluency.pdf. 
Schoonen, Rob, Amos Van Gelderen, Kees De Glopper, Jan Hulstijn, Annegien Simis, 
Patrick Snellings, and Marie Stevenson. 2003. “First Language and Second Language 
Writing: The Role of Linguistic Knowledge, Speed of Processing, and Metacognitive 
Knowledge.” Language Learning 53, no. 1: 165202. Accessed 2 July 2019. 
http://dx.doi.org/10.1111/1467-9922.00213. 
Schoonen, Rob, Amos Van Gelderen, Reinoud D. Stoel, Jan Hulstijn, and Kees de Glopper. 
2011. “Modelling the Development of L1 and EFL Writing Proficiency of Secondary 
School Students.” Language Learning 61, no. 1: 3179. Accessed 12 July 2019. 
http://dx.doi.org/10.1111/j.1467-9922.2010.00590.x. 
Schwartz, Mimi. 1983. “Revision Profiles: Patterns and Implications.” College English 45, 
no. 6: 549558. Accessed 8 July 2019. http://dx.doi.org/10.2307/377139. 
Segalowitz, Norman. 2010. Cognitive Bases of Second Language Fluency. New York & 
London: Routledge. 
. 2016. “Second Language Fluency and Its Underlying Cognitive and Social 
Determinants.”  IRAL 54, no. 2: 7995. Accessed 8 May 2019. 
http://dx.doi.org/10.1515/iral-2016-9991. 
Şen, Zekâi. 2017. Innovative Trend Methodologies in Science and Engineering. Springer 
International Publishing. 
Shapiro, S.S., M.B. Wilk, and H.J. Cohen. 1968. “A Comparative Study of Various Tests for 
Normality.” Journal of the American Statistical Association 63, no. 324: 1343-1372. 
Accessed 8 November 2019. http://dx.doi.org/10.2307/2285889. 
Silva, Tony. 1993. “Toward an Understanding of the Distinct Nature of L2 Writing: the ESL 
Research and Its Implications.” TESOL Quarterly 27, no. 4: 657677. Accessed 2 July 
2019. http://dx.doi/10.2307/3587400. 
Skehan, Peter. 2009. “Modelling Second Language Performance: Integrating Complexity, 
Accuracy, Fluency, and Lexis.” Applied Linguistics 30, no. 4: 510532. Accessed 12 
April 2019. http://dx.doi.org/10.1093/applin/amp047. 
Spelman Miller, Kristyan. 2000. “Academic Writers On-line: Investigating Pausing in the 
Production of Text.” Language Teaching Research 4, no. 2: 123148. Accessed 2 July 
2019. http://dx.doi.org/10.1191/136216800675510135. 
Spelman Miller, Kristiyan, Eva Lindgren, and Kirk P.H. Sullivan. 2008. “The Psychologic 
Dimension in Second Language Writing: Opportunities for Research and Pedagogy 
Using Computer Keystroke Logging.” Tesol Quarterly 42, no. 3: 433454. Accessed 14 
July 2019. http://dx.doi.org/10.1002/j.1545-7249.2008.tb00140.x. 
 45 
 
Stevenson, M.M.C. 2005. “Reading and Writing in a Foreign Language: A Comparison of 
Conceptual and Linguistic Processes in Dutch and English.” Dissertation, Universiteit 
van Amsterdam. 
Strömqvist, Sven, and Henrik Karlsson. 2002. “Scriptlog for Windows – User’s Manual. 
Technical Report.” University of Lund, Department of Linguistics, and University 
College of Stavanger, Centre for Reading Research. 
Thai, Chau, and Frank Boers. 2016. “Repeating a Monologue Under Increasing Time 
Pressure: Effects on Fluency, Complexity, and Accuracy.” TESOL Quarterly 50, no. 2: 
369393. Accessed 19 April 2019. http://dx.doi.org/10.1002/tesq.232. 
Torrance, Mark, Gkyn V. Thomas, and Elizabeth J. Robinson. 2000. “Individual Differences 
in Undergraduate Essay-Writing Strategies: A Longitudinal Study.” Higher Education 
39, no. 2: 181200. Accessed 24 September 2019. 
http://dx.doi.org/10.1023/A:1003990432398. 
Torrance, Mark, and Guido Nottbusch. 2012. “Written Production of Single Words and 
Simple Sentences.” In Past, Present, and Future Contributions of Cognitive Writing 
Research to Cognitive Psychology, edited by Virginia Wise Berninger, 403422, New 
York: Taylor & Francis Group. 
Van Waes, Look, and Peter Jan Schellens. 2003. “Writing Profiles: the Effect of the Writing 
Mode on Pausing and Revision Patterns of Experienced Writers.” Journal of 
Pragmatics 35, no. 6: 829853. Accessed 26 September 2019. 
http://dx.doi.org/10.1016/S0378-2166(02)00121-2. 
Van Waes, Look, and Mariëlle Leijten. 2015. “Fluency in Writing: A Multidimensional 
Perspective on Writing Fluency Applied to L1 and L2.” Computers and Composition 
38: 7995. Accessed 14 July 2019. http://dx.doi.org/10.1016/j.compcom.2015.09.012. 
Wang, Lurong. 2003. “Switching to First Language among Writers with Differing Second-
language Proficiency.” Journal of Second Language Writing 12, no. 4: 347375. 
Accessed 26 September 2019. http://dx.doi.org/10.1016/j.jslw.2003.08.003. 
Wolfe-Quintero, Kate, Shunji Inagaki, and Hae-Young Kim. 1998. Second Language 
Development in Writing: Measures of Fluency, Accuracy and Complexity. Honolulu: 
Second Language Teaching & Curriculum Center. 
Yang, Wenxing and Ying Sun. 2015. “Dynamic Development of Complexity, Accuracy and 
Fluency in Multilingual Learners’ L1, L2 and L3 Writing.” Theory and Practice in 
Language Studies 5, no. 2: 298308. Accessed 26 September 2019. 
http://dx.doi.org/10.17507/tpls.0502.09. 
 
 
 
 
 
 
 
 46 
 
Finnish summary 
Tämän tutkielman tavoitteena oli kartoittaa äidinkielen sujuvuuden vaikutusta vieraan kielen 
sujuvuuteen kirjoitetussa kielessä. Sujuvuuden vaikutusta tutkittiin vertailemalla 
suomenkielisten englanninoppijoiden (n=16) äidinkielellään (L1) kirjoittamia esseitä heidän 
toisella kielellä (L2), eli tässä tapauksessa englanniksi, kirjoittamiin esseisiin. Aineisto kerättiin 
ScriptLog-tietokoneohjelman avulla, ja sitä analysoitiin yhdistämällä laadullista ja määrällistä 
tutkimusmenetelmää. Kokonaisuutena tutkielman tavoitteena oli siis syventää käsitystä vieraan 
kieleen sujuvuudesta nimenomaan kirjoitetussa kielessä sekä tutkia siihen vaikuttavia tekijöitä. 
Sujuvuutta on tutkittu paljon toisen kielen oppimisen ja omaksumisen tutkimuksen (SLA) 
tieteenalalla. Monet aikaisemmista tutkimuksista ovat keskittyneet mm. L2 sujuvuuden 
kehittymisen tutkimiseen (kts. Kowal 2014) tai tutkineet L2 sujuvuuden yhteyttä eri 
taitotasoihin (kts. Palviainen, Kalaja ja Mäntylä 2012). Näissä tutkimuksissa sujuvuutta on 
kuitenkin pidetty yhtenä vieraan kielen oppimisen piirteenä, joka liittyy vieraan kielen 
kielitaitoon eikä niinkään yksilön muihin piirteisiin. Viime vuosina äidinkielen vaikutuksista 
vieraan kielen sujuvuuteen ollaan kiinnostuttu enemmän, kun on havaittu, että yksilöiden 
sujuvuudessa esiintyy vaihtelevuutta myös heidän äidinkielellään. Tämä voisi tarkoittaa, että 
sujuvuus olisikin ennemmin yksilön ominaisuus kuin vieraan kielen oppimisen piirre (kts. 
Derwing ym. 2009). 
Äidinkielen ja vieraan kielen välisiä yhteyksiä on tutkittu aiemmin jonkin verran 
puhutussa kielessä (kts. esim. Peltonen 2018; Derwing ym. 2009; De Jong ym. 2015; Huensch 
& Tracy-Ventura 2017), mutta ei niinkään kirjoitetussa kielessä. Tutkimukset ovat osoittaneet, 
että äidinkielen ja vieraan kielen sujuvuuden välillä on yhteys, joskin se riippuu monista 
osatekijöistä, kuten oppijan vieraan kielen taitotasosta tai äidinkielen ja vieraan kielen 
samankaltaisuudesta.  
Tutkielman alussa esitellään vieraan kielen sujuvuuden eri määritelmiä ja mittareita, joita 
on käytetty enimmäkseen puhutun aineiston tutkimuksissa. Vieraan kielen sujuvuutta on 
vuosien saatossa määritelty monella eri tapaa, mutta yleisesti sillä on viitattu vaivattomaan ja 
sulavaan tapaan tuottaa kieltä. Lisäksi sitä pidetään yhtenä yleisen kielitaidon mittareista. 
Toisistaan osittain poikkeavat sujuvuuden määritelmät ovat kuitenkin aiheuttaneet haasteita sen 
tutkimiselle. Segalowitz (2010, 4849) onkin jaotellut sujuvuuden eri ulottuvuudet kolmeen eri 
kategoriaan helpottaakseen sen tutkimista. Hänen mukaansa sujuvuutta voidaan tarkastella 
ensinnäkin kognition toiminnan kautta (engl. cognitive fluency), joka viittaa kognitiivisten 
toimintojen sujuvuuteen, eli kuinka sujuvasti esimerkiksi prosessoimme vieraskielistä puhetta 
 47 
 
tai tekstiä. Sujuvuutta voidaan tutkia myös kuulijan tai lukijan näkökulmasta (engl. perceived 
fluency), joka perustuu kuulijan/lukijan näkemykseen kuinka sujuvaa kieltä puhuja tuottaa. 
Näiden lisäksi sujuvuutta voidaan myös mitata erilaisilla mittareilla, joihin kuuluvat 
esimerkiksi tauot, epäröinnit ja korjaukset (engl. utterance fluency). Skehan (2003, 8) on 
vuorostaan jaotellut nämä mittarit eri kategorioihin: taukosujuvuuden (engl. breakdown 
fluency), vauhtisujuvuuden (engl. speech rate), korjaussujuvuuden (repair fluency) ja 
automatisoitumisen (automatization) mittareihin. Vaikka näitä mittareita onkin käytetty 
enimmäkseen puhutun kielen sujuvuuden tutkimuksessa, niitä voidaan soveltaa myös 
kirjoitetun kielen tutkimukseen. Tästä syystä myös tämä tutkielma pohjaa sujuvuuden 
määritelmän Segalowitzin ja Skehanin malleihin ja keskittyy tutkimaan sujuvuutta eri 
mittareilla arvioituna. Tutkielmassa sujuvuudella siis viitataan sekä taukosujuvuuteen, eli 
esimerkiksi taukojen määrään ja pituuteen, sekä korjaussujuvuuteen, eli siihen, kuinka paljon 
kirjoittavat korjaavat tekstiään prosessin aikana.  
Kirjoitetun kielen sujuvuuden mittarit ovat yleensä jaettu valmiin tuotetun tekstin 
sujuvuuden mittareihin (product-oriented measures) ja kirjoittamisprosessin sujuvuuden 
mittareihin (process-oriented measures) (Palviainen, Kalaja ja Mäntylä 2012, 48). Valmiin 
tekstin sujuvuus perustuu lukijan arvioon siitä, minkälaista sujuva teksti on. Koska lukija ei näe 
itse kirjoittamisprosessia, hänen on perustettava käsityksensä tekstin kielellisiin piirteisiin, 
kuten tekstin rakenteeseen, koherenssiin ja koheesioon. Kirjoittamisprosessin sujuvuutta taas 
on mitattu erilaisilla nopeuden, tekstin pituuden ja ajan mittareilla (Palviainen, Kalaja ja 
Mäntylä 2012, 50). Näiden mittareiden takana on ajatus siitä, että teksti on sujuvaa silloin, kun 
kirjoittaja pystyy vaivattomasti siirtymään kirjoittamisen eri kognitiivisten prosessien, eli 
suunnittelun, kääntämisen ja arvioimisen, välillä (Palviainen, Kalaja ja Mäntylä 2012, 4950). 
Vieraan kielen oppijoilla nämä prosessit eivät välttämättä ole kehittyneet aivan äidinkielen 
tasolle, mikä johtaa mm. erilaiseen taukokäyttäytymiseen ja/tai suurempaan muokkausten 
määrään tekstissä (Schoonen ym. 2003, 171; Hall 1990; Spelman Miller 2000). Taukojen ja 
muokkausten tutkimisen lisäksi kirjoitusprosessin sujuvuutta ollaan tutkittu niin sanottujen 
”kirjoitusjaksojen” (engl. Burst) avulla. Kirjoitusjaksoilla viitataan sanojen määrään, jonka 
kirjoittaja tuottaa kahden tauon tai muokkauksen välillä (Chenoweth ja Hayes 2001, 88). 
Sanojen sijasta tutkijat käyttävät nykyään merkkien määrää, sillä merkki on helpompi 
määritellä ja analysoida eri kielten välillä erityisesti tietokoneella kerätyissä aineistoissa 
Palviainen, Kalaja ja Mäntylä 2012, 51).  
Kirjoitusprosesseihin liittyvät vahvasti myös erilaiset kirjoitusprofiilit, jotka perustuvat 
taukojen ja muokkausten määrään ja laatuun. Van Waesin ja Schellensin (2003) typologiaa on 
 48 
 
käytetty aiemmissa tietokonepohjaisissa kirjoitusprosessien tutkimuksissa (kts. esim. Mutta, 
tuleva). Van Waes ja Schellens (2003) käyttivät aineiston keräämiseen tavallisen kynä ja paperi 
-metodin lisäksi Keytrap-tietokoneohjelmaa, joiden pohjalta he loivat kirjoittajaprofiilit. Van 
Waesin ja Schellensin (2003) luokittelua voi näin ollen käyttää myös tämän tutkimuksen 
metodina, kun tarkoitus oli analysoida kirjoitusprosesseja myös laadullisesti. Heidän 
luokittelussa kirjoittajat jaetaan viiteen eri profiiliin perustuen taukojen ja muokkausten 
määrään, kestoon, tyyppiin ja paikkaan, sekä tekstin pituuteen ja sen kirjoittamiseen 
kuluneeseen aikaan (Van Waes ja Schellens 2003, 836837). Näihin profiileihin kuuluvat 
alkusuunnittelijat (engl. initial planners), fragmentaariset ensimmäisen vaiheen kirjoittajat 
(engl. fragmentary first-phase writers), toisen vaiheen kirjoittajat (engl. second-phase writers), 
non-stop kirjoittajat (engl. non-stop writers) sekä keskivertokirjoittajat (engl. average writers) 
(Van Waes ja Schellens 2003, 844845). 
Teoreettisen viitekehyksen lopuksi tässä tutkielmassa kartoitettiin aikaisempien 
tutkimusten tuloksia, mitkä ovat keskittyneet tarkastelemaan äidinkielen ja vieraan kielen 
sujuvuuden välisiä yhteyksiä. Koska kirjoitettua kieltä ei ole vielä tutkittu nimenomaan 
sujuvuuden näkökulmasta vertaillen äidinkieltä ja vierasta kieltä, tutkielmassa esitellyt 
tutkimustulokset pohjaavat lähinnä puhuttuun kieleen. Tutkimustuloksista on käynyt ilmi, että 
L1 ja L2 sujuvuuden välillä on yhteys (Derwing yms. 2009; De Jong yms. 2015; Huensch ja 
Tracy-Ventura 2017; Peltonen 2018), joskin se saattaa riippua monista osatekijöistä, kuten L1 
ja L2 välisistä yhtäläisyyksistä ja eroavaisuuksista (Derwing yms. 2009). Peltonen (2018) ja 
Huensch ja Tracy-Ventura (2017) onnistuivat myös osoittamaan, että oppijan vieraan kielen 
taitotaso saattaisi vaikuttaa siihen, kuinka vahvasti L1 sujuvuus vaikuttaa L2 sujuvuuteen. 
Äidinkielen ja vieraan kielen yhteyksiä on tutkittu myös kirjoitetussa kielessä, mutta 
tutkimukset ovat aiemmin keskittyneet lähinnä kirjoitustaidon vertailuun L1 ja L2 välillä. 
Näissä tutkimuksissa on huomattu, että äidinkielen kirjoitustaidolla on vaikutus myös vieraan 
kielen kirjoitustaitoon (Sasaki ja Hirose 1996; Schoonen yms. 2003; Schoonen yms. 2011). 
Lisäksi on todistettu, että oppijat turvautuvat äidinkielessään käyttämiin kirjoitusstrategioihin 
myös vieraalla kielellä kirjoittaessaan (Manchón 2013, 109; Manchón, Roca de Larios ja 
Murphy 2007, 239242). Näiden tutkimustulosten pohjalta voitiin olettaa, että myös 
sujuvuuden näkökulmasta löydettäisiin yhteys L1 ja L2 välillä tässä tutkielmassa. Osassa 
kirjoitussujuvuuden tutkimuksista on myös käytetty L1 aineistoa L2 aineiston lisäksi, joskin 
näissä tutkimuksissa on joko keskitytty kirjottamissujuvuuden kehittymiseen (kts. esim. 
Lindgren, Spelman Miller ja Sullivan 2008; Yang ja Sun 2015; Kobayashi ja Rinnert 2013) tai 
L2 sujuvuuden ja eri taitotasojen välisiin yhteyksiin (esim. Chenoweth ja Hayes 2001). Näin 
 49 
 
ollen L1 ja L2 väliset yhteydet ovat jääneet vähemmälle huomiolle näissä tutkimuksissa, lukuun 
ottamatta Yangin ja Sunin (2015) tutkimustuloksia, jotka osoittivat, että kirjoittajan sujuvuus, 
tarkkuus ja kompleksisuus vaikuttavat jossain määrin toisiinsa kirjoittajan eri kielten välillä.  
Tutkimuksessa käytetty aineisto on kerätty osana Turun yliopiston tutkimushanketta, 
jonka tarkoituksena on tarkastella sanastokeskittymiä morfologisesti erilaisissa kielissä (suomi, 
ruotsi, englanti). Tätä tutkimusta varten aineistosta valittiin suomenkielisten lukiolaisten vain 
suomeksi ja englanniksi kirjoittamat esseet analysointia varten. Aineistonkeruutilanteessa 
oppilailla oli aikaa yhteensä 90 minuuttia kirjoittaa argumentoiva essee ensiksi ruotsiksi, sitten 
englanniksi, ja lopuksi suomeksi. Kirjoitusprosessit tallennettiin ScriptLog-ohjelman avulla 
Lopuksi esseet arvioitiin Eurooppalaisen viitekehyksen taitotasoihin perustuen (2 arvioijaa). 
Taitotasoa käytettiin tässä tutkielmassa koehenkilöiden kielitaidon mittarina. 
Tutkimuskysymykset olivat seuraavat: 
1) Miten paljon L1 sujuvuuden mittarit korreloivat vastaavien L2 mittareiden kanssa? 
2) Missä määrin L2 sujuvuuden mittarit voidaan ennustaa vastaavista L1 mittareista? 
3) Miten paljon kirjoitusprofiilit vastaavat toisiaan L1 ja L2 tuotoksissa? 
Hypoteesina oletettiin, että myös tässä tutkielmassa L1 ja L2 sujuvuuden väliltä löytyisi 
yhteyksiä, kuten aiemmat tutkimukset ovat osoittaneet. Samoin odotettiin, että L1 ja L2 
sujuvuuden yhteys olisi vahvempi kielitaidoltaan taitavimmilla kirjoittajilla.  
Sujuvuuden mittaamiseen käytettiin samoja mittareita, joita aiemmat tutkimukset ovat 
käyttäneet. Näihin kuuluivat kirjoitettujen merkkien määrä, kokonaiskirjoitusaika, taukojen 
määrä ja kesto, suunnitteluun käytetty aika (aika alusta toisen lauseen alkuun saakka), 
muokkaamiseen käytetty aika (ensimmäisen version kirjoituksen jälkeinen aika), sekä 
muokkausten määrä. Tämän lisäksi käytettiin kompleksisempia sujuvuuden mittareita, kuten 
”kirjoitusjakso” (engl. Burst), ”sujuvuus kirjoitusjakson aikana” (engl. Fluency during Burst), 
sekä laskettiin, kuinka monta merkkiä keskimäärin koehenkilöt tuottivat minuutissa (engl. 
Fluency linear) ja kuinka monta merkkiä he tuottivat keskimäärin minuutissa lopulliseen 
tekstiin nähden (engl. Fluency product). Kirjoitusjaksoista tarkasteltiin vielä erikseen, kuinka 
paljon koehenkilöt keskimäärin muokkasivat tekstiä yhden jakson aikana (engl. R-Burst) ja 
kuinka paljon he keskimäärin tuottivat merkkejä kahden tauon välissä (engl. P-Burst). Aineistoa 
tutkittiin määrällisesti käyttämällä Spearmanin järjestyskorrelaatiokerrointa ja 
regressioanalyysia, jonka jälkeen tuotoksia tutkittiin laadullisesti kirjoitusprofiilien avulla, ja 
pyrittiin näin tunnistamaan yksilöiden sujuvuuden piirteitä. Koska aineisto oli lukumäärältään 
pieni, eri taitotasot yhdistettiin toisiinsa (A2+B1 ja B2+C1), jotta voitiin vertailla myös 
taitotason vaikutusta korrelaatiokertoimiin. 
 50 
 
Spearmanin korrelaatiokertoimen tulokset osoittivat, että L1 ja L2 sujuvuuden mittarit 
korreloivat toistensa kanssa jossain määrin, joskin vain osa tuloksista oli tilastollisesti 
merkittäviä. Myös korrelaatioiden vahvuus vaihteli taitotasoryhmien välillä. Ainoa 
tilastollisesti merkittävä mittari koko ryhmälle oli taukojen keskimääräisen pituus, joka korreloi 
vahvasti (rs = 708, p = .002) L1 ja L2 välillä. Ryhmien välinen vertailu osoitti, että 
korrelaatiokertoimet olivat suurimmaksi osaksi yllättäen hieman vahvemmat taitotasoltaan 
huonommalla ryhmällä, joskin vain yksi näistä mittareista, taukojen keskimääräinen pituus, oli 
tilastollisesti merkittävä (rs = .905, p = .002). Taitotasoltaan vahvemmalla ryhmällä taasen 
taukojen määrä (rs = .708, p = .050) sekä P-Burst (rs = .714, p = .047) korreloivat tilastollisesti 
merkittävästi. Tulosten perusteella on vaikea sanoa, onko L1 ja L2 sujuvuuden välinen yhteys 
vahvempi jommallakummalla ryhmällä, sillä vain osa mittareista korreloi tilastollisesti 
merkittävästi. Regressioanalyysin tulokset osoittivat, että L2 sujuvuuden mittarit voitiin jossain 
määrin ennustaa vastaavien L1 mittareista kahdessa kolmesta mittarista, joissa tilastollisesti 
merkittävää korrelaatiota löytyi. Keskimääräisen taukojen pituuden kohdalla L1 mittari selitti 
35% ja P-Burstin kohdalla 34% variaatiosta vastaavassa L2 mittarissa. Hierarkkisen 
regressionanalyysin tulokset paljastivat, ettei taitotaso ollut tilastollisesti merkittävä ennustaja 
malleissa. 
Laadullisen tutkimuksen tulokset osoittivat, että kaiken kaikkiaan seitsemällä 
koehenkilöllä oli sama kirjoitusprofiili L1 ja L2 tuotoksissa. Tämän lisäksi oli muutamia 
henkilöitä, joilla selkeästi oli samantyyppisiä piirteitä molemmissa kirjoitusprosesseissaan, 
mutta heidän kokonaisprofiilinsa olivat silti erilaiset. Eri profiilit esiintyivät myös tasaisesti 
koko aineistossa. Eroja löytyi kuitenkin eri tasoryhmien väliltä. Siinä missä taitotasoltaan 
heikommassa ryhmässä oli enemmän alkusuunnittelijoita, taitotasoltaan vahvemman ryhmän 
jäsenet olivat useammin toisen vaiheen kirjoittajia. Yhtenä selityksenä eroille voisi olla, että 
vahvemman ryhmän jäsenet olivat motivoituneempia muokkaamaan tekstiä vielä ensimmäisen 
luonnoksen kirjoittamisen jälkeenkin. Heikomman ryhmän jäsenet taas saattavat tarvita 
enemmän aikaa kirjoitusprosessin aluksi suunnitellakseen tekstiä, sillä heidän kielitaitonsa ei 
ole vielä riittävä siihen, että he voisivat alkaa suoraan kirjoittamaan vieraskielistä tekstiä. Non-
stop kirjoittajia löytyi toisaalta molemmista ryhmistä, eli näyttäisi siltä, ettei ainakaan 
sujuvuuden piirteitä voida yhdistää yhteen tiettyyn kirjoitusprofiiliin. Mitä tulee eri 
tasoryhmiin, tulokset osoittivat myös, että kielitaidoltaan vahvemmilla kirjoittajilla oli 
useammin sama kirjoittajaprofiili (5/8) L1 ja L2 välillä verrattuna heikompiin kirjoittajiin (2/8). 
Nämä tulokset ovat hieman ristiriidassa määrällisen tutkimuksen tulosten kanssa, mutta 
toisaalta ne tukevat mm. Peltosen (2018) ja Huenschin ja Tracy-Venturan (2017) tuloksia. 
 51 
 
Kaiken kaikkiaan tämän tutkielman tulokset osoittivat, että L1 ja L2 sujuvuuden välillä 
on jonkinlainen yhteys kirjoitetussa kielessä. Määrällisen tutkimuksen perusteella parhaiten 
korreloivat mittarit, jotka liittyivät kirjoittajien taukokäyttäytymiseen. Myös Peltonen (2018) ja 
De Jong yms. (2015) löysivät samanlaisia tuloksia suullisen aineiston tutkimuksissaan, mikä 
saattaisi tarkoittaa, että kirjoittaminen ja puhuminen eivät olisikaan loppujen lopuksi täysin 
erillisiä toisistaan. Muokkausten määrässä ja tavassa ei ollut yhtä paljon nähtävissä 
samankaltaisuutta L1 ja L2 välillä, vaikka esseiden laadullinen tarkasteleminen osoittikin, että 
monet kirjoittajista myös muokkasivat tekstejään samoilla tavoin. Yhtenä syynä tuloksille voi 
olla, että tekstin muokkaamisen suhteen L1 ja L2 ei ole yhtä suurta yhteyttä verrattuna taukojen 
määrän ja pituuteen. Tosin myös muokkausten määritelmällä voi olla vaikutusta tämän 
tutkielman tuloksiin. Yllättäen myös monet tekstien yleistä sujuvuutta mittaavat mittarit, kuten 
kirjoitusjaksot, korreloivat heikosti L1 ja L2 välillä. Tutkimuksia aiheesta tarvitaan siis lisää, ja 
erityisesti suuremmalla aineistonmäärällä, sekä toisenlaisilla muokkausten määritelmillä. L1 ja 
L2 välisiä yhteyksiä kirjoitetussa kielessä voidaan tutkia myös käyttämällä ääneen ajattelu  
-metodia tekstintallennusohjelmien tukena, jolloin päästään syvemmälle yksilöiden 
kirjoitusprosesseihin. Määrällisen ja laadullisen tutkimuksen yhdistäminen on elintärkeää näin 
monisyisen ilmiön tutkimiseen, kuten tämänkin tutkielman tulokset osoittivat.