See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/310161672

Learning difficulties, academic well-being and educational pathways among

adolescent students

Thesis · November 2016

DOI: 10.13140/RG.2.2.30090.47049

CITATIONS READS

3 3,745

1 author:

Johan Korhonen
Åbo Akademi University
37 PUBLICATIONS   225 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Early numeracy Intervention study for low performing children in South African schools View project

(Eye) Tracking individual differences in numeracy development. Interplay between skills, motivation and well-being. (iSeeNumbers) View project

All content following this page was uploaded by Johan Korhonen on 14 November 2016.

The user has requested enhancement of the downloaded file.

 Johan Korhonen | Learning difficulties, academic well-being and educational pathways among adolescent students | 2016
Johan Korhonen

Learning difficulties, academic well-being

and educational pathways among
adolescent students

9 789521 234637

ISBN 978-952-12-3463-7
  

Learning difficulties,
academic well-being and
educational pathways among
adolescent students

Johan Korhonen

Special Education
Faculty of Education and Welfare Studies
Åbo Akademi University
Vasa, Finland, 2016
 
Custos
Docent Karin Linnanmäki, Åbo Akademi University

Supervisors
Docent Karin Linnanmäki, Åbo Akademi University
Professor Pirjo Aunio, University of Helsinki

Pre-examiners
Professor Markku Niemivirta, University of Helsinki
Professor Ulf Träff, Linköping University, Sweden

Opponent
Professor Rauno Parrila, University of Alberta, Canada

ISBN 978-952-12-3463-7 (Print)

ISBN 978-952-12-3464-4 (PDF)
Painosalama Oy – Turku, Finland 2016

 
Abstract

The purpose of this thesis was to investigate the interplay between learning
difficulties in mathematics and reading, academic well-being and educational
pathways. Particular emphasis was placed on mathematical learning difficulties, that
was the main interest of research in all four studies. In Study I, the relations between
mathematics performance and different language skills (word-, reading
comprehension, and spelling) were investigated in ninth-grade students (N = 810).
Study II focused on the relative importance of mathematics and reading achievement
on students’ (N = 1152) educational aspirations together with interest and well-being
measures. Study III investigated change and stability of mathematical learning
difficulties and academic well-being across the transition to upper secondary
education in adolescent students (N = 980). Finally, educational dropout was
predicted with learning difficulties and academic well-being in Study IV.
Structural equation modelling techniques were used in all the studies and resulted
in the following findings. Students’ reading skills were strongly related to their
mathematics performance while spelling was not (Study I). Latent profile analysis
identified a group of students with combined difficulties in mathematics and reading
but no “single difficulty” groups (Study IV). Mathematics was more important
compared to reading for boys’ educational aspirations, while for girls reading was more
important (Study II). School burnout predicted educational aspirations indirectly
through interest while academic self-concept had a positive direct effect on students’
educational aspirations (Study II). Students with mathematical learning difficulties
and low-achieving students did not differ from their peers in academic self-concept
development from grade nine to upper secondary education although they
experienced lower levels of academic self-concept in grade nine (Study III). However,
students with mathematical learning difficulties in general upper secondary education
displayed opposite change patterns in school burnout compared to typically achieving
students. Both learning difficulties and academic well-being predicted educational
dropout (Study IV). In general, learning difficulties and academic well-being were
related (Study II, III, & IV) but a group of students with average performance in
reading and mathematics, but negative academic well-being was identified (Study IV).
To conclude, learning difficulties and academic well-being are related and both
shape students’ educational pathways. Moreover, mathematical learning difficulties
co-occur frequently with difficulties in reading in this age group. These findings
emphasize the importance to identify individual differences not only in students’
academic achievement, but also concerning their well-being, and support both areas
in school.

Keywords: dropout, educational pathways, learning difficulties, well-being

3
 
Abstrakt

Syftet med denna avhandling var att undersöka sambandet mellan inlärningssvårigheter
i matematik och läsning, akademiskt välbefinnande och utbildningsvägar.
Inlärningssvårigheter i matematik var i fokus i alla fyra delstudier. I studie I undersöktes
samband mellan matematikprestationer och olika färdigheter inom läsning (ord-,
läsförståelse och stavning) i ett sampel av elever i årskurs nio (N = 810). Studie II
fokuserade på den relativa betydelsen av matematik- och läsfärdigheter för studerandes
(N = 1152)) utbildningsmålsättningar tillsammans med mått på deras intresse och
välbefinnande. I studie III undersöktes förändring och stabilitet av inlärningssvårigheter
i matematik och akademiskt välbefinnande hos studerande (N = 980) under övergången
till andra stadiets utbildning. Slutligen predicerades skolavbrott i andra stadiet med hjälp
av inlärningssvårigheter och akademiskt välbefinnande i studie IV.
Strukturekvationsmodeller användes i alla delstudier och resulterade i följande fynd.
Studerandes läsfärdigheter hade ett starkt samband med deras matematikfärdigheter
medan deras stavningsfärdigheter inte var relaterade till deras matematikfärdigheter
(Studie I). Latenta profilanalyser identifierade en grupp studerande med svårigheter i både
läsning och matematik medan ”en-svårighetsgrupper” inte identifierades (Studie IV).
Matematikfärdigheter var viktigare för pojkarnas utbildningsmålsättningar medan
läsfärdigheter var viktigare för flickornas utbildningsmålsättningar (Studie II). Skoltrötthet
hade en negativ indirekt effekt på studerandes utbildningsmålsättningar via intresse medan
akademisk självuppfattning hade en positiv direkt effekt (Studie II). Studerande med
inlärningssvårigheter i matematik och studerande med svaga prestationer i matematik
skiljde sig inte från sina klasskamrater med avseende på utvecklingen av deras akademisk
självuppfattning från årskurs nio till andra stadiet även om de uppvisade lägre akademisk
självuppfattning i årskurs nio (Studie III). Däremot uppvisade studerande med
inlärningssvårigheter i matematik i gymnasiet ett motsatt mönster beträffande
utvecklingen av skoltrötthet jämfört med normalpresterande studerande. Både
inlärningssvårigheter och akademiskt välbefinnande predicerade skolavbrott (Studie IV).
Överlag fanns det ett negativt samband mellan inlärningssvårigheter och akademiskt
välbefinnande (Studie II, III, & IV), men också en grupp studerande med
medelprestationer och negativt akademiskt välbefinnande kunde identifieras (Studie IV).
Sammanfattningsvis kan man konstatera att det finns ett samband mellan
inlärningssvårigheter och akademiskt välbefinnande och att båda formar studerandes
utbildningsvägar. Vidare kan man konstatera att inlärningssvårigheter i matematik
sammanfaller ofta med inlärningssvårigheter i läsning i denna åldersgrupp. Dessa fynd
understryker vikten av att inte endast identifiera individuella skillnader i skolprestationer
utan också i akademiskt välbefinnande och stöda båda områden i skolan.

Sökord: skolavbrott, utbildningsvägar, inlärningssvårigheter, välbefinnande

4
 
Acknowledgements

This long adventure, finally culminating in this thesis, would not have been
possible to finish without the support of many people. Contrary to some beliefs,
research is not something you do alone but in cooperation with others. You need
a lot of support and guidance, from your colleagues but also from your family
and friends. I have been lucky in both departments.
I wish to thank my supervisor Docent Karin Linnanmäki for igniting my
interest in mathematical learning difficulties and encouraging me to start this
adventure. During our time as colleagues, both in different research projects and
at the department, she has always guided me in the right direction but at the same
time also given me the freedom to shape my own path. I am also very grateful
that she introduced me to my second supervisor Professor Pirjo Aunio. I would
not be the researcher I am today without Pirjo. From the very beginning she has
challenged me to work harder, aspire higher, and of course, do better research. I
am extremely lucky that I have had two great supervisors that complement each
other and who both also have become good friends to me.
I would sincerely like to thank my pre-examiners, Professor Markku
Niemivirta and Professor Ulf Träff for their constructive feedback on my thesis.
I also feel honored to have Professor Rauno Parrila as my opponent.
During this time, I have also had the pleasure to share the ups and downs of
doing research with the other members of Pirjos research group; Riikka
Mononen, Heidi Hellstrand, Ulrika Ekstam, Eija Väisänen, Henrik Husberg and
Anna Tapola. Thank you for taking the time to read and comment all manuscript
drafts during the years and for all the fun times during seminars and conferences.
I also want to express my gratitude for having Anna as a coauthor in one of the
studies. Your critical thinking and knowledge in motivational theories made all
the difference. My thanks also go to my colleagues at Åbo Akademi University
(Kristina Ström, Christel Sundqvist, Kajsa Lindström, Anki Risberg, and Tanja
Östman) for all the peer-support and encouraging words during the years. I
would also like to thank my fellow PhD student in the project, Camilla Svens-
Liavåg for all the nice discussions about research and other topics. Pekka Räsänen

5
 
(Niilo Mäki Institute), thank you for all the inspiring discussions concerning
mathematical learning difficulties and for good company during conferences.
I want to express my gratitude to Svenska Kulturfonden, Harry Schaumans
stiftelse and Högskolestiftelsen for supporting our project financially.
My well-being during this project has been greatly enhanced by my dear
friends. Franken, thank you for being such a great friend since preschool. I also
want to express my joy of having the support and company of Ale and Hencki
throughout my later educational career, first as classmates in general upper
secondary, and then as roommates during our University studies.
Finally, I want to express my gratitude to my family. I grew up in a big and
loving family where my parents Solveig and Hannu always put us children first.
Thank you for all the love and support. My dear brothers, Mikael, Patrik and
Tomi have always been entertaining company and good friends. I also feel very
lucky to have a loving person beside me who has supported me during this
adventure. Thank you Maria for being there. Our three children have all come to
us during this project and have made all things so much more fun. Thank you
Alma, Ella and Isak for being you.

Vasa, October, 2016

Johan Korhonen

6
 
Table of Contents
Abstract .................................................................................................................3 

Abstrakt .................................................................................................................4 

Acknowledgements ...............................................................................................5 

List of Original Publications ..............................................................................10 

1.  Introduction ...................................................................................................11 

1.1.  Learning Difficulties ........................................................................................ 11 
1.1.1.  Learning difficulties in mathematics ................................................ 11 
1.1.2.   Learning difficulties in reading.......................................................... 13 
1.1.3.   Learning difficulties in both mathematics and reading ................. 14 
1.2.  Academic Well-Being ...................................................................................... 17 
1.2.1.   Academic self-concept ........................................................................ 18 
1.2.2.   School burnout .................................................................................... 18 
1.2.3.   Perceived learning difficulties ............................................................ 19 
1.3.  Educational Pathways ...................................................................................... 19 
1.3.1.   Educational aspirations ...................................................................... 20 
1.3.1.1. Expectancy-value theory ...................................................... 21
1.3.1.2. Internal/external frame of reference model ....................... 23
1.3.1.3. Gendered pathways to educational aspirations................. 23
1.3.2.   Transition to upper secondary education ........................................ 24 
1.3.2.1. Academic self-concept and the ‘big fish in a little pond’
effect ......................................................................................... 25
1.3.3.   School dropout .................................................................................... 26 
1.4.  The Present Study ............................................................................................ 27 
1.4.1.   Context: The Finnish education system ........................................... 30 

2.  Methods ..........................................................................................................32 
2.1.  Participants and Procedure ............................................................................ 32 

7
 
 2.2.  Measures ........................................................................................................... 33 
2.2.1.   Mathematics skills .............................................................................. 33 
2.2.2.   Reading skills ....................................................................................... 34 
2.2.3.   Academic self-concept ....................................................................... 34 
2.2.4.   Perceived learning difficulties ........................................................... 35 
2.2.5.   School burnout .................................................................................... 35 
2.2.6.   Individual interest............................................................................... 35 
2.2.7.   Educational aspirations ...................................................................... 36 
2.2.8.   Upper secondary education track ..................................................... 36 
2.2.9.   School dropout .................................................................................... 36 
2.2.10. Parental educational background ..................................................... 36 
2.3.  Data Analysis ................................................................................................... 37 
2.3.1.   Missing data analysis .......................................................................... 37 
2.3.2.   Confirmatory factor analysis and latent factor path analysis ....... 37 
2.3.3.   Multiple group confirmatory factor analysis .................................. 38 
2.3.4.   Longitudinal confirmatory factor analysis ...................................... 39 
2.3.5.   Latent change score analysis.............................................................. 39 
2.3.6.   Latent profile analysis......................................................................... 40 
2.3.7.   Analyses of variance and covariance ................................................ 40 
2.3.8.   Chi-square tests and adjusted residuals ........................................... 41 
2.3.9.   Configural frequency analysis ........................................................... 41 

3.  Overview of the Original Studies.................................................................. 42 

3.1.  Study I ............................................................................................................... 42 
3.1.1.   Aims...................................................................................................... 42 
3.1.2.   Participants and procedure ............................................................... 42 
3.1.3.   Measures .............................................................................................. 42 
3.1.4   Data analysis ........................................................................................ 42 
3.1.5.   Results .................................................................................................. 43 
3.1.6.   Discussion ............................................................................................ 43 
3.2.  Study II.............................................................................................................. 44 
3.2.1.   Aims...................................................................................................... 44 
3.2.2.   Participants and procedure ............................................................... 44 
3.2.3.   Measures .............................................................................................. 44 

8
 
 3.2.4.   Data analysis ........................................................................................ 45 
3.2.5.   Results ................................................................................................... 45 
3.2.6.   Discussion ............................................................................................ 46 
3.3.  Study III ............................................................................................................. 47 
3.3.1.   Aims ...................................................................................................... 47 
3.3.2.   Participants and procedure ................................................................ 47 
3.3.3.   Measures ............................................................................................... 48 
3.3.4.   Data analysis ........................................................................................ 48 
3.3.5.   Results ................................................................................................... 49 
3.3.6.   Discussion ............................................................................................ 50 
3.4.  Study IV ............................................................................................................. 51 
3.4.1.   Aims ...................................................................................................... 51 
3.4.2.   Participants and procedure ................................................................ 52 
3.4.3.   Measures ............................................................................................... 52 
3.4.4.   Data analysis ........................................................................................ 53 
3.4.5.   Results ................................................................................................... 53 
3.4.6.   Discussion ............................................................................................ 54 

4.  General Discussion ........................................................................................55 

4.1.  Learning Difficulties in Mathematics and Reading ..................................... 56 
4.2.  Learning Difficulties and Academic Well-Being ......................................... 57 
4.3.  Changes in Academic Well-Being Related to Mathematical Learning
Difficulties Across the Transition to Upper Secondary Education ........... 58 
4.4.  Learning Difficulties and Academic Well-Being Predicting Educational
Aspirations and Potential for Dropping Out ............................................... 60 
4.5.  Strengths and Limitations ............................................................................... 62 
4.7.  Pedagogical Implications ................................................................................ 65 

References ............................................................................................................67 

Original Publications ..........................................................................................79 

 

9
 
List of Original Publications

Study I: Korhonen, J., Linnanmäki, K., & Aunio, P. (2012). Language and
mathematical performance: A comparison of lower secondary
school students with different level of mathematical skills.
Scandinavian Journal of Educational Research, 56(3), 333-344.

Study II: Korhonen, J., Tapola, A., Linnanmäki, K., & Aunio, P. (2016).
Gendered pathways to educational aspirations: The role of
achievement, academic self-concept, interest, and school burnout.
Learning and Instruction, 46, 21-33.

Study III: Korhonen, J., Linnanmäki, K., & Aunio, P. (submitted). Change and
stability in mathematical learning difficulties and academic well-
being in adolescent students.

Study IV: Korhonen, J., Linnanmäki, K., & Aunio, P. (2014). Learning
difficulties, academic well-being and educational dropout: A
person-centered approach. Learning and Individual Differences, 31,
1-10.

Author contribution
Johan Korhonen is the first author for all four manuscripts that are included in
this doctoral thesis. Korhonen is responsible for all the analyses used in this
dissertation study and has written all manuscripts under the supervision of
Docent Karin Linnanmäki and Professor Pirjo Aunio.

10
 
1. Introduction

The effectiveness of a given educational system can be determined based on its

capability to equip its students with the relevant academic competencies as well
as its support of their general well-being and motivation to learn. In addition, the
quality—and especially equality—of an educational system can be measured in
terms of its effectiveness in promoting the progress of each individual student
irrespective of factors that might serve as disadvantages along the student’s
educational journey (e.g., the student’s social or ethnic background or learning
difficulties [LDs]).
This work investigates the impact and interrelationships of LDs and academic
well-being on students’ educational pathways, as assessed at the end of their
compulsory education. This work advances the current knowledge on the
relationship between reading difficulties (RDs) and mathematical learning
difficulties (MLDs), their impact on educational aspirations and likelihood of
dropout along with academic well-being and any changes and/or stability of
MLDs and academic well-being in adolescent students across the transition to
upper secondary school.

1.1. Learning Difficulties

In this study the focus is on learning difficulties in mathematics and reading.
These difficulties are first described separately and then together to provide an
overview of what we know of the co-occurrence of LDs in these two content
areas.

1.1.1. Learning difficulties in mathematics

Learning difficulties in mathematics can take various forms (e.g., difficulties in
basic arithmetic and word problem solving) and have different underlying causes
(Toll, Kroesbergen, & Van Luit, 2016; Träff, 2013). Furthermore, the literature
uses different terminology—such as mathematical disability (MD),
developmental dyscalculia (DD), mathematical learning difficulties (MLD) and
low achievement (LA) in mathematics—to describe the mathematics-related LDs

11
 
found in an average student population (for a more detailed discussion, see
Mazzocco, 2007). Mathematical disability, developmental dyscalculia and
mathematical learning difficulties (which are terms that are often used
interchangeably) in children have a prevalence of 3% to 6% in the school-aged
population, which is comparable to the rate of dyslexia (Shalev, 2007; Rubinstein
& Henrik, 2009). This deficit can be tracked to the neural level, where it is
manifested in weaker brain activation in the intraparietal sulcus and the middle
and inferior frontal gyrus of both hemispheres (Kaufmann, Wood, Rubinstein,
& Henrik, 2011; Price et al., 2007; Rotzer et al., 2009). Thus, these conditions have
a biological cause.
According the latest knowledge, a defining feature of MLD is difficulty
processing numerical information (De Smedt & Gilmore, 2011; Skagerlund &
Träff, 2014). More specifically, two possible and potentially related causes—
deficits in the approximate number system (ANS) (Mazzocco et al., 2011; Olsson,
Östergren, & Träff, 2016) and with symbolic numerical magnitude processing
(De Smedt & Gilmore, 2011; Skagerlund & Träff, 2014)—have been reported to
characterize MLD. ANS refers to an innate preverbal ability to represent and
manipulate quantities and is believed to form the basis for the acquisition of the
symbolic number system used to learn formal arithmetic (Dehaene, 2011; Piazza,
2010). Symbolic numerical magnitude processing is usually assessed with a
symbolic numerical magnitude comparison task (De Smedt, Noël, Gilmore, &
Ansari 2013), where one is instructed to indicate the numerically larger of two
presented magnitudes in a symbolic format (i.e., Arabic digits).
A recent meta-analysis incorporating 284 effect sizes from 17,201 participants
showed that symbolic numerical magnitude processing is more strongly related
to mathematical achievement than nonsymbolic comparisons (Schneider et al.,
2016). Moreover, symbolic magnitude processing has been found to be the
strongest predictor of arithmetic skills at the beginning of primary school (Lyons,
Price, Vaessen, Blomert, & Ansari, 2014), to mediate the relationship between
informal and formal mathematics skills (Merkley & Ansari, 2016) and to predict
mathematical skills more strongly than phonological awareness predicts reading
skills (Vanbinst, Ansari, Ghesquiere, & De Smedt, 2016). Symbolic numerical
magnitude processing has also been shown to be related to developmental

12
 
dyscalculia (Meijas, Grégoire, & Noël, 2012) and mathematics achievement in
adults (Linsen, Torbeyns, Verschaffel, Reynvoet, & De Smedt, 2016).
On a behavioural level, this manifests in difficulties with basic arithmetic
tasks. Specifically, children with MLD rely on counting-based strategies when
solving arithmetic problems and do not easily shift to more effective strategies
like fact retrieval from their long-term memory. To identify students with MLD
in both research settings and in practice, some researchers have advocated for
the use of the persistence criterion (Fletcher et al., 2005; Stock, Desoete, &
Roeyers, 2010; Mazzocco & Räsänen, 2013). For example, to determine the
persistence of MLD in a student, he or she has to score under a given cut-off score
on a mathematics test for 2 or more consecutive years. Research findings have
emphasised the validity of this approach, as there are students who only meet the
criteria for MLD at one time point instead of two or more; this is also in line with
the response to intervention practices for defining MLDs (e.g., Fuchs et al., 2007).
However, not all children struggling with mathematics meet the criteria for
having an MLD. In research settings, it is common practice to identify students
with mathematics-related LDs by using a cut-off criteria of up to 35%. Low-
achievement (LA) in mathematics has a wider definition than MLD and includes
all students who struggle in mathematics for various reasons, including
sociocultural and educational factors (Mazzocco, 2007). In addition, working
memory (WM) (Friso-van den Bos, Van der Ven, Kroesbergen, & Van Luit,
2013), language skills (Lefevre et al., 2010), motivation (Murayama, Pekrun,
Lichtenfeld, & vom Hofe, 2012), self-concept (Seaton et al., 2014) and
socioeconomic background (Sirin, 2005) have all been shown to influence
students’ mathematics skills. Therefore, it is important to distinguish between
students with MLD and LA students to accurately advance the knowledge base
on why students struggle with mathematics and to determine the best indicators
of these difficulties (Mazzocco & Räsänen, 2013.

1.1.2. Learning difficulties in reading

Similar to mathematics, reading difficulties (RDs) can take various forms (e.g.,
fluency and/or comprehension) and have multiple underlying causes. It has
become conventional to distinguish children who have specific RDs (e.g.,

13
 
developmental dyslexia) from children who have RDs in the context of more
general learning problems (Snowling, 2005). Specifically, an RD or
developmental dyslexia is a mild hereditary neurological disorder that manifests
as a persistent difficulty in learning to read in children with otherwise normal
intellectual functioning and educational opportunities (Ramus, 2004). More
specifically, students with RD tend to have pervasive deficiencies in word
identification, fluency, phonological (letter sound) decoding, and spelling
(Vellutino & Fletcher, 2007). The prevalence of specific reading difficulties
depends on the specific cut-off point taken as indicative of RD, but studies have
shown that 3% to 10 % of the population has RD (Snowling, 2000).
On a neural level, when comparing people with RD and normal readers,
studies have shown that there is difference between brain structure and function,
particularly in the left hemisphere temporal regions (Shaywitz et al., 2002).Well-
established markers for RD at the cognitive level are phonological processing
deficits (Vellutino, Fletcher, Snowling, & Scanlon, 2004) and deficits in rapid
automatized naming (RAN) (Kirby, Georgiou, Martinussen, & Parrila, 2010;
Landerl et al., 2013). Furthermore, the associations among phonological
processing, RAN and reading skills are moderated by the orthographic
complexity of the language under study (e.g., Finnish versus English) (Georgiou,
Torppa, Manolitsis, Lyytinen, & Parrila, 2013). As stated earlier, not all students
who are struggling in reading show these domain-specific cognitive deficits and
should not be categorized as having an RD. This broader group, which
encompasses all students with difficulties in reading regardless of the underlying
cause, is usually defined as low-achieving students in reading. In this group,
problems in reading may arise from factors like WM (Daneman & Carpenter,
1980), motivation (Guthrie et al., 2007), self-concept (Retelsdorf, Köller, &
Möller, 2014) and socioeconomic background (Sirin, 2005).

1.1.3. Learning difficulties in both mathematics and reading

Many students have problems in attaining average achievement in both reading
and mathematics (Dirks, Spyer, van Lieshoult, & de Sonneville, 2008; Light &
DeFries, 1995). This rate of co-occurrence is substantially greater than would be
predicted from the individual prevalence rates of MLDs and RDs (Dirks et al.,

14
 
2008). Students with comorbid difficulties display more severe and global
functional difficulties than students with only MLDs (Andersson & Lyxell, 2007;
Fuchs & Fuchs, 2002; Mazzocco & Myers, 2003). These students also seem to
have broader problems with language learning, especially with spelling and
reading comprehension, than children with only one of these difficulties (Dirks
et al., 2008). For example, both Ostad (1998) and Shafrir and Siegel (1994) have
reported an overlap between spelling problems and MLDs.
Based on the literature, there is no shared understanding for the origins of the
comorbidity of RD and MLDs. One possible cause of this co-occurrence on the
cognitive level is working memory capacity (Hanich, Jordan, Kaplan, & Dick,
2001; Siegel & Ryan, 1989). Research findings suggest that students with low-
achievement in mathematics have a general working memory processing deficit;
thus, this could partially explain the shared variance in mathematical skills and
language skills (Kyttälä, 2008; Pickering & Gathercole, 2004). However, Landerl,
Fussenegger, Moll and Willburger (2009) found two different cognitive profiles
for children with dyscalculia and dyslexia, namely a phonological deficit in
dyslexia and a number-processing deficit in the case of dyscalculia. Their
conclusion concerning children with comorbid difficulties is that they have
additive cognitive deficits that indicate no evidence for shared cognitive risk
factors. There are other findings that suggest that the presence of both RDs and
MLDs seems to be caused by a fundamental deficit in phonological processing,
which affects, for instance, arithmetic fact retrieval (Geary, 1993; Geary & Hoard,
2001; Rourke & Conway, 1997).
However, dysfunction in phonological processing does not seem to be a
plausible explanation, as approximately half of the children with MLDs do not
have RDs (Badin, 1983; Ostad, 1998), nor is it clear why these students also show
fact retrieval deficits while their phonetic abilities are intact (Jordan, Hanich, &
Kaplan, 2003). Augustyniak, Murphy and Phillips (2005) have concluded that
mathematical ability seems to be based on a more complex set of abilities than
reading ability. Furthermore, findings show that some areas in mathematics are
more dependent on reading abilities than others (Jordan et al., 2003; Vukovic &
Lesaux, 2013a).

15
 
 The measures used to operationalize reading skills also vary among studies.
Measures such as word decoding, rapid naming, spelling and reading
comprehension are often used to operationalize reading skills, and in many
studies, composite scores consisting of two or more of these measures are used
to examine the relationship between mathematics and reading (Dirks et al., 2008;
Swanson & Beebe-Frankenberger, 2004; Wise et al., 2008). Only a few studies
have examined the unique contribution of different language skills to
mathematics performance (Vukovic & Lesaux, 2013b). Dirks and colleagues
(2008) have demonstrated that using different measures of reading skills results
in different children being identified as having RDs and, thus, different children
being treated as having both RDs and MLDs. So far, it seems that there is no
consensus of the underlying causes for RDs and MLDs and why these difficulties
sometimes overlap each other.
The terminology and definitions for RD and MLD vary across studies,
although a common factor visible in school practice is that students struggling
with reading and mathematics learning can most often be found in the lowest
performing group on standardized mathematics and reading tests. A traditional
and widely used approach to define subtypes of learning difficulties is to divide
students into groups according to RD only, MLD only and combined difficulties
(MLDRD) (Anderson & Lyxell, 2007; Andersson, 2010; Jordan et al., 2003;
Vukovic, Lesaux, & Siegel, 2010). Interestingly, the MLD-only and RD-only
groups have been found to perform at the same level on word problem tasks and
significantly better than the MLDRD group (Jordan et al., 2003).
Consistent with these results, Vukovic and colleagues (2010) have
demonstrated that both children with dyslexia and children with specific reading
comprehension difficulties performed equally on word problems and
significantly poorer than normal readers. However, the children with reading
comprehension difficulties did not differ from normal readers in arithmetic fact
fluency and operations, whereas the dyslexia group performed on a lower level.
They concluded that children with decoding problems (dyslexia) had a higher
risk of also experiencing MLD. Concerning MLD, there is evidence that both the
MLD-only and MLDRD groups showed fundamental deficits in factual

16
 
knowledge and that this is a primary characteristic for MLDs (Andersson, 2010;
Jordan et al., 2003).
A recent meta-analysis by Swanson, Jerman and Zheng (2009) investigated
differences in cognitive features in students with MLDs, typically achieving
students, RD students and MLDRD students. The clearest differences were found
between the typically achieving students and those with MLD, namely in working
memory and literacy skills, whereas the MLD and RD groups differed in variation
in working memory and mathematical problem-solving skills. The only thing
that differentiated the MLD students from MLDRD students was IQ and long-
term memory, with the MLDRD group displaying lower levels for both.
These results lend support to the hypothesis that grouping that aims to create
qualitatively (type) different subtypes only creates quantitatively (level) different
subtypes of learning difficulties. To summarize, we can conclude that in the
younger children (aged 7-13), the MLD-only and MLDRD groups performed
worse than the RD-only and typically achieving students on the more technical
aspects of mathematics (factual knowledge, place value, etc.) and that the
children with dyslexia performed poorer compared to the children with reading
comprehension and typically achieving children in the more technical aspects of
reading (e.g., word reading) (Andersson, 2010; Jordan, et al., 2003; Vukovic et
al., 2010). However, the MLDRD group seemed to perform on a lower level on
almost all measures compared to the other groups, and the RD-only and MLD-
only groups performed at the same level on mathematical word problem tasks. It
is therefore tempting to look at both mathematical and reading skills as
continuums where the order of the traditional subtypes of learning difficulties is
similar; i.e., the MLDRD group is at the lower end and the typically achieving
children are at the higher end of the continuum. The MLD-only and RD-only
groups fall between and their order are defined by which skill is currently
assessed.

1.2. Academic Well-Being

School is a central factor in the lives of adolescents (Eccles & Roeser, 2009), and
therefore it is important to define well-being in relation to the school context.

17
 
How students perceive themselves as learners in school and how they experience
their schooling have been shown to influence their well-being (Goetz,
Cronjaeger, Frenzel, Ludtke, & Hall, 2010; Tuominen-Soini, Salmela-Aro, &
Niemivirta, 2008, 2012). Furthermore, students’ academic well-being is viewed
as an important indicator of the educational process (van Petegem, Aelterman,
Rossel, & Creemers, 2007) and has received growing attention in recent years
(Holopainen, Lappalainen, Junttila, & Savolainen, 2012; Rueger, Malecki, &
Demaray, 2010; Tuominen-Soini et al., 2008, 2012). However, there is no
consensus around the definition or measurements of academic well-being, so
positive and negative indicators of well-being, such as self-esteem, school value
and stress, are commonly used as indicators of adolescent academic well-being
(Pollard & Lee, 2003). In the present study, academic well-being is
operationalized with academic self-concept, school burnout, and perceived
learning difficulties.

1.2.1. Academic self-concept

Self-concept is theorised as a hierarchical multidimensional construct
(Shavelson, Hubner, & Stanton, 1976; Marsh & Shavelson, 1985). The most
relevant dimension of self-concept in the educational settings is arguably
academic self-concept. Academic self-concept can be defined as mental
representations of one’s competencies in academic domains (Marsh & Craven,
1997). These ability beliefs are influenced by prior achievement and social
comparison processes (Marsh, 1986). Academic self-concept is positively related
to achievement (Marsh, Hau, & Kong, 2002; Valentine, DuBois, & Cooper, 2004),
approaches to learning (Burnett, Pillay, & Dart, 2003), social interaction
(Gurney, 1986) and general happiness (Harter, 1990). According to the OECD,
self-concept and related constructs are “closely tied to students’ economic
success and long-term health and wellbeing” (2003, p. 9).

1.2.2. School burnout

School burnout can be defined as exhaustion due to school demands, a cynical
and detached attitude toward one’s school and feelings of inadequacy as a
student; it can develop as a result of a high imbalance between perceived

18
 
demands and personal resources (Salmela-Aro, Kiuru, Leskinen, & Nurmi, 2009;
Salmela-Aro & Upadyaya, 2014a). These subdimensions have been shown to be
associated with adolescent students’ motivational strivings, whereby a tendency
to minimize schoolwork and effort (i.e., work avoidance orientation) increases
the likelihood of cynicism and sense of inadequacy. It is noteworthy that even
high-achieving and motivated students, especially girls, have been found to be
vulnerable to exhaustion (Tuominen-Soini et al., 2012; Tuominen-Soini &
Salmela-Aro, 2014). Furthermore, previous studies have found that girls
experience higher levels of school burnout than boys (Salmela-Aro, Kiuru, &
Nurmi, 2008; Salmela-Aro & Tynkkynen, 2012). School burnout is also related
to low academic achievement (Kiuru, Aunola, Nurmi, Leskinen, & Savolainen,
2008) and depression (Bakker et al., 2000; Salmela-Aro, Savolainen, &
Holopainen, 2009).

1.2.3. Perceived learning difficulties

Perceived learning difficulties have not received much attention in the literature.
Theoretically, perceived learning difficulties are different from academic self-
concept and more related to self-efficacy (Bong & Skaalvik, 2002). While self-
concept refers to an individual’s’ general perceptions of the self in a given
domain, self-efficacy refers to the expectations and convictions of what the
individual can accomplish in a given situation. Perceived learning difficulties
refer to difficulties with learning in a particular situation or setting (e.g., difficulty
taking notes when the teacher explains specific content in class). In the few
studies incorporating perceived learning difficulties, it has been shown to be
related to academic well-being, academic achievement and secondary education
placement (DeGarmo & Martinez, 2006; Hakkarainen, Savolainen, &
Holopainen, 2012).

1.3. Educational Pathways

To better understand the relationship between learning difficulties and academic
well-being, it is useful to take into the consideration the context that this
relationship is formed in. Adolescent students are generally in the middle of their

19
 
educational journey, making it a natural point to start investigating the
relationship between learning difficulties and academic well-being in
conjunction with their educational aspirations, transition to upper secondary
school and possibly dropping out of school. The various processes leading to
these important educational outcomes have been described and explained using
several prominent theories incorporating achievement and motivational beliefs
(Eccles et al., 1983; Marsh, 1986, 1987). In the following sections, the adolescent
educational pathway and the processes influencing it are described, and the
variables under study (i.e., learning difficulties and academic well-being) are
situated within the theories used to explain these processes.

1.3.1. Educational aspirations

Educational aspirations in adolescence have been used to explain educational
disparities in general and individuals’ occupational choices and attainment later
in life (Domina et al., 2011; Rojewski, 2005). There is no commonly shared
definition for the term ‘educational aspirations’, but it is often used to refer to a
person’s goals and plans within an academic setting (Trebbels, 2015).
Researchers have defined and measured educational aspirations via the number
of career plans per adolescent (Mendez & Crawford, 2002), the prestige of their
educational aspirations (Leung, Conoley, & Scheel, 1994; Viljaranta, Nurmi,
Aunola, & Salmela-Aro, 2009) and the level of self-set educational goals
(Vasalampi, Salmela-Aro, & Nurmi, 2009; Watt et al., 2012).
Educational aspirations can also be defined as being either idealistic or
realistic. While idealistic aspirations refer to the attainment level that a student
hopes to achieve, realistic aspirations reflect the actual perceived likelihood of
success and more pragmatic expectations of completing a certain level of
education (Rojewski, 2005). However, in several studies focusing on educational
choices, the operationalisation of educational aspirations has covered both
idealistic and realistic alternatives, with no clear distinction being made between
the two terms Chow, Eccles, & Salmela-Aro, 2012; Durik, Vida, & Eccles, 2006;
Guo, Marsh, Parker, Morin, & Yeung, 2015). In some studies, girls have reported
higher educational aspirations (Mahaffy & Ward, 2002; Mau & Bikos, 2000),
whereas in other studies, boys’ aspirations have been higher (Inoue, 1999;

20
 
Mendez & Crawford, 2002). There have also been studies that found no gender
differences in educational aspirations (e.g., Ireson & Hallam, 2009; Watt et al.,
2012). Besides the possible gender differences in the level of educational
aspiration, it also seems that the processes or paths that lead to these aspirations
may differ by gender (Domene, Shapka, & Keating, 2006; Watt et al., 2012).

1.3.1.1. Expectancy-value theory

Students’ abilities and academic success (both in reading and mathematics)
shape their educational aspirations (Durik et al., 2006; Shapka, Domene, &
Keating, 2006). However, academic achievement is not, by itself, a sufficient
factor to explain students’ educational aspirations. For example, irrespective of
girls’ high mathematics and sciences grades, they are underrepresented in
advance STEM courses and occupational fields. Further, it has been documented
that along with the possible direct effects, the influence of academic achievement
may be indirect and mediated by different motivational constructs (Parker, Nagy,
Trautwein, & Lüdtke, 2014).
According to the expectancy-value model (Eccles et al., 1983; Wigfield &
Eccles, 1992), students’ expectancies of success at school and subjective task
values are key predictors of academic decision making that are related but
distinct. Expectancies of educational success are defined as an individual’s
competence perceptions about how well they will perform on future academic
tasks and subjective ability beliefs (Eccles et al., 1983).
In many of the studies, students’ competence perceptions have been
conceptualized through academic self-concept. Subjective task value is usually
operationalized in terms of intrinsic value (i.e., interest) and importance value
(i.e., utility and attainment value combined) (Jacobs et al., 2002; Watt et al.,
2012). Individual interest refers to the tendency to reengage and enjoy a
particular content domain (Frentzel, Goetz, Pekrun, & Watt, 2010; Renninger &
Hidi, 2011). Individual interest towards an academic domain is likely to develop
when engagement with it is experienced as inherently rewarding (Renninger &
Su, 2012). Once interest deepens, it becomes intertwined with a student’s
personal values and may even form a part of their identity (Renninger, 2009:
Renninger & Hidi, 2011). Both academic self-concept and individual interest

21
 
have been shown to positively predict educational aspirations in a number of
studies (Guo, Marsh, Morin, Parker, & Kaur, 2015; Guo, Marsh, Parker, Morin,
& Yeung, 2015; Watt et al., 2012).
Within the expectancy-value framework, some studies have also examined
how self-perceived drawbacks predict educational choices (Perez, Comley, &
Kaplan., 2014). The concept of cost has been utilized to describe the perceived
negative consequences of engaging in learning, such as required effort or
psychological well-being (for a more detailed discussion, see Eccles et al., 1983).
In addition, these studies have identified that cost measures both predict and
correlate negatively with student interests (Flake, Barron, Hulleman, McCoach,
& Welsh, 2015; Gaspard et al., 2014) and educational aspirations (Battle &
Wigfield, 2003; Luttrell et al., 2010; Perez et al., 2014). In fact, the
conceptualisations of cost and school burnout share some common features. For
example, the exhaustion component of school burnout is related to both
dimensions of cost, namely, the amount of effort and emotional and
psychological demands, whereas inadequacy and cynicism are more similar to
the latter dimension (Flake et al., 2015; Gaspard et al., 2014; Salmela-Aro, Kiuru
et al., 2009).
However, there are also some important differences between school burnout
and cost. First, school burnout is a more general negative emotion toward school
(Salmela-Aro, Savolainen, & Holopainen, 2009), whereas cost is more domain or
situation specific (Wigfield & Eccles, 2000). Second, school burnout develops
over a longer period of time, whereas cost is more situational and anticipatory.
In fact, arguably, school burnout can be regarded as a consequence of
experiencing prolonged cost (Shaufeli & Bakker, 2004; Tuominen-Soini et al.,
2008). To summarize, despite the differences between cost and burnout, both
have been theorized and shown to hinder student motivation and engagement
by using up their psychological resources and decreasing their positive affect
(Barron & Hulleman, 2015; Salmela-Aro & Upadyaya, 2014a; Tuominen-Soini &
Salmela-Aro, 2014). As a consequence, students’ capability and desire to set and
achieve ambitious educational goals may also be diminished (Nurmi & Salmela-
Aro, 2002).

22
 
1.3.1.2. Internal/external frame of reference model
The internal/external (I/E) frame of reference model (Marsh, 1986; Marsh et al.,
2015) postulates that students’ self-perceptions are dependent not only on social
(external) comparisons, but also on dimensional (internal) comparisons, in
which achievement in one school subject is evaluated in reference to another
school subject (Möller et al., 2009). Traditionally, this model has been used to
explain the relationship among mathematics, verbal achievement and self-
concept. In the external comparison process, high mathematics achievement is
supposed to lead to higher mathematics self-concept and high verbal
achievement is supposed to lead to higher verbal self-concept. However,
according to the internal comparison process, higher mathematics achievement
will result in lower verbal self-concept after the effect of verbal achievement is
taken into account. Consequently, while students achieving well in mathematics
also often perform well in reading, they consider themselves ‘as being good at’
mainly one or the other of the domains. There are robust findings across age
groups, gender and countries concerning the presence of this mechanism
(Möller, Pohlmann, Köller, & Marsh, 2009).
The I/E frame of reference model has been extended to the expectancy-value
model (Eccles, 2009). According to Eccles (2009), the external and internal
comparisons of one’s abilities and values are a natural part of adolescents’
identity development process, whereby they start to identify themselves more
strongly with some academic domains than with others. This identification
process, in turn, is likely to be influenced by gender stereotypes: out of two
academic domains that an adolescent achieves well in, he or she will be more
prone to emphasize and prefer the one that corresponds to gender-typical
expectations and norms. As a consequence, adolescents’ academic choice
behaviour and expectations concerning their educational and occupational goals
may also be influenced (e.g., Nagy, Trautwein, Baumert, Köller, & Garrett, 2006;
Parker et al., 2012).

1.3.1.3. Gendered pathways to educational aspirations

In this study, one focus have been on the relative importance of reading and
mathematics achievement (and well-being measures) for students’ educational

23
 
aspirations, which can be investigated in the context of the expectancy-value
model and the I/E model. Studies focusing solely on one academic domain have
found mixed results concerning gender differences in the relationship (i.e.,
moderator effects) among academic self-concept, individual interest and
educational aspirations (e.g., Durik et al., 2006; Watt et al., 2012).
Interestingly, in studies contrasting two or more academic domains, gender
has been found to moderate the effects of these variables on educational
aspirations and choices. For example, Nagy and colleagues (2006) demonstrated
that during secondary school, boys’ self-concept and interest in mathematics
affected their decision making concerning advanced biology course choices,
whereas for girls, the choices in mathematics and biology were independent of
the contrasted domain’s motivational beliefs.
In line with this, a few studies utilising a person-centred methodological
approach have investigated how different interest profiles measured by interest
in different domains predict educational aspirations (Chow & Salmela-Aro,
2011; Chow et al., 2012; Viljaranta et al., 2009; Watt, 2005). In these studies, boys
were observed to be more likely to have a profile of high interest or task values in
mathematics and science compared to girls, who were more likely to display a
profile with high reading interest (Chow et al., 2012; Viljaranta et al., 2009).
While in Chow and colleagues’ study (2012) the interest profiles fully
mediated the relationship between gender and educational aspirations in the
physical and information technology–related sciences, Viljaranta and colleagues
(2009) found that their task-value profiles predicted educational aspirations for
girls but not for boys. They hypothesized that boys´ educational aspirations
might originate from an objective criterion (e.g., academic achievement),
whereas girls seemed to base their expectations on broader considerations,
including their values and motivation.

1.3.2. Transition to upper secondary education

Educational transitions have been found to be related to a decrease in students’
motivation and well-being (e.g., Eccles & Midgley, 1989; Otiz, Grouzet, &
Pelletier, 2005). So far, most studies have focused on middle school transitions
(i.e., junior high school to senior high school) in early adolescence and have been

24
 
conducted in the United States. In Finland, the transition from compulsory
education to upper secondary education (high school to college) is an important
step in adolescent students’ educational pathways (Nyyssölä, 2004). More
specifically, the general upper secondary schools prepare students for university
studies (i.e., the academic track), whereas the vocational upper secondary schools
(i.e., the vocational track) are more practically oriented, preparing students for
working life. Admittance to the academic track usually requires higher academic
skills compared to the vocational track, and it is generally recommended that
adolescents move to a school form that is in accordance with their academic skills
(Eccles, 1993; Hakkarainen, Holopainen, Savolainen, 2012; Savolainen et al.,
2008). It is expected that these students will show higher motivation and well-
being when they are in a more congruent educational setting (Vasalampi,
Salmela-Aro, & Nurmi, 2010) compared to a compulsory education setting that
is the same for all students. However, there is a lack of studies that have
specifically looked at changes in well-being among students with learning
difficultiess. Studies that have examined the development of school burnout have
not specifically looked at students with MLDs; rather, the focus has been on
differences in educational tracks and gender (Salmela-Aro, Kiuru, & Nurmi,
2008; Salmela-Aro, Savolainen, & Holopainen, 2009; Salmela-Aro & Tynkkynen,
2012; Salmela-Aro & Upadyaya, 2014b; Tuominen-Soini & Salmela-Aro, 2014).
However, in a study by Salmela-Aro and colleagues (2008), low-achieving
students (operationalized by a low grade point average) were identified as having
higher levels of school burnout compared to their peers. In addition, girls in both
tracks and all students in the academic track reported an increase in school
burnout (Salmela-Aro et al., 2008; Salmela-Aro & Tynkkynen, 2012; Salmela-Aro
& Upadyaya, 2014b).

1.3.2.1. Academic self-concept and the ‘big fish in a little pond’ effect
In this study, academic self-concept is also considered to reflect students’
academic well-being. In general, academic self-concept development has been
found to be related to academic achievement in a reciprocal manner (e.g., Marsh,
Hau, & Kong, 2002). However, the educational setting that students are situated
in also has an impact on their academic self-concept. The ‘big fish in a little pond’

25
 
effect (BFLPE) (Marsh, 1987; Marsh & Craven, 2002; Nagengast & Marsh, 2011)
can be used to describe the effects of one’s frame of reference in educational
settings. The BFLPE is based upon a social comparison process (for a detailed
account of the theoretical background of the BFLPE, see Marsh et al., 2008),
where individual student achievement has a positive effect on academic self-
concept while school-average achievement has a negative effect on academic self-
concept. Students compare themselves with their peers, and, therefore,
individual academic self-concept is shaped not only by a student’s own
achievement, but also by the achievement of other students in the same class or
school. Thus, students have lower academic self-concept when they are in high-
ability classes or schools compared to students of the same ability that are
educated in mixed or low-ability classes or schools. Interestingly, these kind of
effects have also been observed when comparing the academic and vocational
track in a sample of Finnish students concerning their education-related goal
appraisals (Vasalampi et al., 2010).
Students with MLDs have lower academic self-concept in general compared
to their peers (Bear, Minke, & Manning, 2002; Zeleke, 2004). The literature on
learning difficulties and the BFLPE is scarce. In the only study conducted with
this focus, Marsh, Tracey and Craven (2006) demonstrated that students with
learning difficulties in inclusive settings had lower academic self-concept
compared to students with learning difficulties in segregated settings and offered
the BFLPE as a potential explanation for this result.

1.3.3. School dropout

Dropping out of school can have far-reaching negative consequences on an
individual level. Students who drop out are more likely to be unemployed (Sum,
Khatiwada, McLaughlin, & Palma, 2009), have a lower income level (Levin,
Belfield, Muenning, & Rouse, 2007; Rumberger & Lamb, 2003) and experience
lower levels of general well-being (Bynner & Parsons, 2002; Lamb, 2011). In
addition to these individual costs, there are also social costs associated with
increased welfare needs and reduced taxation revenue (Owens, 2004).
As graduation from secondary school is currently viewed as the minimum
level of educational attainment for the successful inclusion of young people into

26
 
society, it is important to identify at-risk groups already present in
comprehensive education. Converging evidence suggests that the single most
important dropout predictor is low academic achievement (Battin-Pearson et al.,
2000; Janosz, LeBlanc, Boulerice, & Tremblay, 1997, 2000), which is often related
to a lower socioeconomic background of a student’s family (Duncan & Brooks-
Gunn, 2001; Fall & Roberts, 2012). However, learning difficulties in mathematics
and reading have not been analysed separately in these studies on dropout rates.
A large body of research has also determined that students classified as having
low academic well-being face a higher risk of dropping out of school
(Archambault, Janosz, Morizot, & Pagani, 2009; Fall & Roberts, 2012; Fortin,
Marcotte, Potvin, Royer, & Joly, 2006; Janosz et al., 1997, 2000).

1.4. The Present Study

Although considerable efforts have been made to understand the dynamics
between learning difficulties in reading and mathematics and their relationship
to other educational outcomes and well-being measures, there are still gaps in
our shared understanding of these processes. Different conceptualizations of the
constructs under study, the inclusion of different variables in research designs
and different theoretical starting points have made it challenging to integrate
findings from different studies. This dissertation study makes an attempt to
provide a more holistic picture of adolescent students’ educational pathways with
focus on learning difficulties in reading and mathematics and academic well-
being.
In this study, MLDs are defined as problems in the domain of mathematics
that are attributed to either environmental or individual factors. This is a broader
definition of MLD than, for example, mathematical disability (Geary, 1994) or
developmental dyscalculia (Butterworth, 2005). Similarly, RDs are defined as
problems with reading that are attributed to either environmental or individual
factors. However, an attempt is made in this study to separate low-achieving
students and students with MLDs using the persistence criterion (Fletcher et al.,
2005) and to investigate possible differences in the development of academic
well-being measures in adolescent students in the low-achieving and MLD

27
 
groups. In the present work, academic well-being is operationalized through
academic self-concept, perceived learning difficulties and school burnout.
Academic well-being is highly relevant for the educational process (Tuominen-
Soini et al., 2008, 2012) and is closely related to academic achievement
(Hakkarainen et al., 2012; Valentine et al., 2004).
Even though there is extensive evidence on the link between learning
difficulties and educational outcomes (e.g., school dropout and educational
aspirations), there are research gaps that the present study aims to fill. Firstly,
previous studies have more or less treated students with learning difficulties as a
homogenous group, which means that there is a lack of knowledge of whether
some types of learning difficulties impose a higher risk of dropping out of school
or predict an individual’s educational aspirations differently.
Secondly, there are varying research results regarding the relationship
between reading difficulties and mathematical learning difficulties. Even though
much research has been conducted within the framework of the RD, MLD, and
MLDRD typologies, results from these same studies support the hypothesis that
the differences between the groups are more quantitative than qualitative.
Moreover, these studies have not utilized person-centred methods in creating
classifications and have not focused on adolescent students. The assessment
methods used to measure reading skills also vary between studies. In many
studies, composite scores consisting of two or more scales of reading skills (e.g.,
word reading and reading comprehension) have been used to examine the
relationship between mathematics and reading (see, for example, Dirks et al.,
2008; Swanson & Beebe-Frankenberger, 2004; Wise et al., 2008). There seems to
be a gap in the research literature concerning studies that examine the unique
contribution of different reading skills to mathematics performance.
Based on the literature on educational aspirations, which has mostly been
situated either within the expectancy-value model or I/E frame of reference
model, the variables included in our study play pivotal roles in the processes that
shape these aspirations. Furthermore, the literature clearly demonstrates that
these processes are somewhat different for boys and girls and that incorporating
achievement and motivational beliefs measures from the two domains
(mathematics and reading) are warranted. Hence, we have incorporated

28
 
individual interest in mathematics and reading into our empirical model as
mediators of the effects of mathematics and reading achievement according to
the expectancy-value model. Finally, there is some evidence that students’
academic well-being should also be included when trying to understand
educational aspirations (e.g., Perez et al., 2014).
This dissertation is based on four original publications, which are referred to
in the text by Roman numerals (studies I to IV). The overall aim of the present
dissertation study was to investigate the interplay among LDs in mathematics
and reading, academic well-being and educational pathways (Figure 1) in
adolescent students. The general aim was pursued based on the following
research questions:

1) How are mathematical learning difficulties (MLD) and reading

difficulties (RD) related in adolescent students (studies I and IV)?

2) How are learning difficulties and academic well-being related (studies II,
III and IV) in adolescent students?

3) How are changes in academic well-being related to mathematical

learning difficulties (MLD) during the transition from comprehensive
school to upper secondary school (study III)?

4) How can learning difficulties and academic well-being explain

educational aspirations and school dropout in adolescent students
(studies II and IV)?

29
 
Figure 1. Theoretical model of adolescent students’ educational pathways where LDs and
academic well-being predict educational outcomes; ASC = academic self-concept, SBO
= school burnout, PLD = perceived learning difficulties, ED ASP = educational
aspirations, educational track = general or vocational upper secondary education.

1.4.1. Context: The Finnish education system

Finland is officially bilingual, with Finnish and Swedish being the national
languages. The Swedish-speaking minority constitutes approximately 5% (n =
291,219) of the population (Official Statistics of Finland, 2012a). The education
systems are identical for both language groups but, unusual for a minority, the
Swedish-speaking population has lower unemployment rates (Saarela & Finnäs,
2003) and higher general well-being (Nyqvist, Finnäs, Jakobsson, & Koskinen,
2008).
In Finland, comprehensive school is a 9-year compulsory general schooling
for all children aged 7 to 16 years. It is comprised of primary school (Grades 1 to
6) and lower secondary school (Grades 7 to 9). At the end of comprehensive
school (Grade 9), students choose between upper secondary general education
(i.e., the academic track) and vocational education (i.e., the vocational track). A
majority of students (50%) who finish compulsory education in Finland choose

30
 
upper secondary education, whereas 41% of the students choose vocational
education. Upper secondary education prepares students for university studies,
whereas vocational education is more practically oriented (for a detailed
description, see Rinne & Järvinen, 2011). This choice is one of the most
important decisions that they will make regarding the ultimate level of education
that they will attain (Nyyssölä, 2004).
When students apply for a secondary education programme, there is a risk
that they will not be admitted to an educational program or that they drop out at
an early stage if their academic performance is low. Studies have shown that LDs
increase the risk of not being admitted to an educational program and that these
students tend to choose the vocational education track (Järvinen & Vanttaja,
2006; Savolainen, 2001; Vanttaja & Järvinen, 2004). Students with LDs that get
admitted are at a high risk of dropping out and leaving school without finishing
their studies, particularly post-compulsory schooling (Jahnukainen, 2001).
The dropout rate at Finnish comprehensive schools is minimal (0.04%;
Official Statistics of Finland, 2012b), but in the secondary education
programmes, the school dropout rate is far larger (5.5%; Official Statistics of
Finland, 2012b). One reason for the high completion of comprehensive
education is the extensive special educational support system in Finnish schools
(Simola, Rinne, & Kivirauma, 1999). In Finland, on average, children from
homes with low socioeconomic status have a higher probability of being excluded
from education and beginning their working life between the ages of 16 to 18
than the rest of the population (Järvinen & Vanttaja, 2001). The relationship
between parental background and achievement in Finland is also well
documented, although this effect is weaker than in many other countries (OECD,
2010).

31
 
2. Methods

2.1. Participants and Procedure

In all of the four original studies used in this dissertation, the data were drawn
from the Vem Väljer Vad (VVV, ‘Who Chooses What’) study, which is a project
led by Karin Linnanmäki (Åbo Akademi University). The project has been
mainly funded by the Swedish Cultural Foundation in Finland and
Högskolestiftelsen i Österbotten. VVV is an ongoing follow-up study whose
overall aim is to investigate adolescent students’ educational pathways from
compulsory to tertiary education with a focus on learning difficulties and
academic well-being. The project started in 2007, and I have been involved in the
planning and implementation of the data collection from the beginning of this
study. Data have been collected from 14 different Swedish-speaking secondary
schools located all over the geographic area inhabited by the Swedish-speaking
minority group in Finland.
The data used in the present work were from two measurement occasions.
The first data collection was executed when the students were in Grade 9 in 2008,
and the second data collection was conducted 2 years later when the same
students were in secondary education. Information on the measures, number of
participants and the time points included in the original studies are summarized
in Table 1.

32
 
Table 1
Measures, number of participants and time points included in the studies
Measure N Study Study Study Study
I II III IV
Time point 1
Mathematics 980 X X X X
Reading comprehension 1028 X X X
Word comprehension 1000 X X X
Spelling 1019 X X
Academic self-concept 866 X X X
School burnout 921 X X X
Perceived learning difficulties 921 X
Interest in math 870 X
Interest in reading 1034 X
Educational aspirations 859 X
Parental educational background 734 X

Time point 2
Mathematics 712 X
Academic self-concept 650 X
School burnout 693 X
Upper secondary education track 1066 X
School dropout 1152 X

2.2. Measures
For this dissertation study, data were collected with assessment tools to measure
students’ skills in reading and mathematics. Academic well-being was measured
with questionnaires, and students were also asked to report gender, parental
educational level and educational aspirations. Teachers reported students’
interest in mathematics and reading. Information concerning dropout and
educational track was gathered from schools.

2.2.1. Mathematics skills

The mathematics skills of students were assessed with the standardized KTLT
test (Räsänen & Leino, 2005). We used the Swedish version of the scale, which
has been standardized for the Finnish Swedish population. This scale assesses
students’ performance in basic arithmetic (addition, subtraction, multiplication

33
 
and division), word problem solving, algebra, geometry and unit conversion
skills. It is also used for screening purposes for identifying at-risk students. This
scale is intended for Grades 7 to 9 (13 to 16 years of age), and it consists of four
parallel versions (A, B, C and D). The KTLT test contains 40 items, with a correct
answer worth 1 point and an incorrect answer worth no points. Thus, the results
range from 0 to 40. Students have 40 minutes to complete the scale, which is in
the form of a paper-and-pencil test.

2.2.2. Reading skills

The LS reading ability test (Johansson, 2005) is a standardized test for Grades 7
to 9 (13 to 16 years of age) and Grade 1 in upper secondary school (16 to 17 years
of age). The test identifies students with RDs. The subtests we used were word
comprehension, reading comprehension and a spelling test. The word
comprehension and reading comprehension scales had multiple-choice tasks,
and in the spelling test, the words that needed to be written were presented
verbally for the students.
Word comprehension (40 items). These items measured students’ word
comprehension skills in Swedish. There were 40 sentences that were missing the
last word. Five words were listed after each sentence, and the students had to
choose the right word that fit the sentence.
Reading comprehension (5 texts). These items measured students’ reading
comprehension skills in Swedish. There were five short texts that the students
read. After that, they chose the correct title from four suggestions for each text
and then chose the correct statements from six statements about the content of
the text.
Spelling (50 items). These items measured students’ spelling skills in Swedish.
The test administrator read a sentence and then repeated words that the students
must write down correctly.

2.2.3. Academic self-concept

Academic self-concept was measured using items from the Perceived
Competence Scale for Children (Harter, 1982). The original scale consists of 28
items and is a paper-and-pencil test. We used a modified version that had two

34
 
additional items, one of which measured academic self-concept. For each item,
students were given two statements. First, they were asked to decide if they are
more like the young people described on the left side of the statement or more
like those depicted on the right side. Secondly, they were asked to mark whether
each statement was completely true or just partly true for them. The responses
were then scored on a 4-point scale ranging from least competent (1) to most
competent (4). Because the original scale was created in English and available in
Finnish and Norwegian, we utilised a multilingual team to ensure the validity of
the instrument by translating the questionnaire from Norwegian to Swedish and
from Finnish to Swedish and comparing the similarity of both translations.

2.2.4. Perceived learning difficulties

Perceived learning difficulties were assessed using nine items measuring how
much difficulty the students experienced in their schoolwork (e.g., take notes in
class, understand instructions on the whiteboard, etc.). Each item was assessed
using a 7-point Likert scale ranging from 1 (not at all) to 7 (very much). The
items originally came from the School Health Promotion study (Rimpelä, 2003)
and were translated to Swedish for this purpose.

2.2.5. School burnout

School burnout was assessed using the School Burnout Inventory (SBI), which is
a nine item scale developed by Salmela-Aro and Näätänen (2005). The inventory
consisted of three subscales: exhaustion at school (EXH), cynicism towards the
meaning of school (CYN) and sense of inadequacy as a student (INAD). Each
subscale comprised three items which were assessed using a 6-point Likert scale
ranging from 1 (completely disagree) to 6 (completely agree). A multilingual
team translated the scale from Finnish to Swedish.

2.2.6. Individual interest

We used teacher ratings to measure students’ interest in mathematics and
Swedish (i.e., their mother tongue). The students’ mathematics and Swedish
teachers were asked to evaluate the students’ motivation and activity in class and
rate the students’ interest in these subjects on a 5-point Likert scale. They

35
 
received a name list of their students by class (e.g., class 9b) and were requested
to fill in the ratings for every student.

2.2.7. Educational aspirations

We measured the students’ idealistic and realistic educational aspirations
according to two statements on a 3-point Likert scale (vocational upper secondary
education, polytechnic education and university education): ‘Highest academic
degree I want to achieve’ and ‘Highest academic degree I will probably achieve’.

2.2.8. Upper secondary education track

Information on students’ educational track (vocational or academic) was
collected from schools. First, information from all 14 comprehensive schools
where the students had applied was collected. Second, the upper secondary
schools were contacted to get confirmation that the students were studying there.
If this was not the case, the students themselves were contacted to get the
information.

2.2.9. School dropout

School dropout in this study refers to students not enrolled in education 2 years
after Grade 9. Of the original sample of 1,152 students, 55.4% (n = 639) were
studying at general upper secondary schools, 36.9% (n = 426) were studying at
vocational upper secondary school and 3% (n = 35) of the students were not on
any educational path at the time of the second data collection. Of these latter
students, 37.1% had never started secondary school, 51.4% dropped out from a
vocational upper secondary school and only 11.4% dropped out from a general
upper secondary school. Furthermore, there was a heterogeneous group of
students (n= 52) that included those who were studying abroad, combining
general and vocational upper secondary studies or studying at folk high schools
that do not give qualifications for specific professions.

2.2.10. Parental educational background

Students were asked to report their mother’s and father’s highest educational
degrees on a 6-point scale (1 = no education, 2 = short course-based vocational

36
 
education, 3 = vocational education, 4 = vocational education in combination
with the matriculation exam, 5 = polytechnic and 6 = university degree).

2.3. Data Analysis

A common feature in all studies was the use of structural equation models to
analyse the data. Structural equation models refer to general statistical
procedures for multiequation systems that include continuous or categorical
latent variables, multiple indicators of constructs, errors of measurement, errors
in equations, and observed variables (Bollen, 1989). In addition to structural
equation models, some more traditional analyses (e.g., ANOVA) were also used
in this dissertation study. The data analyses were conducted with SPSS and the
Mplus statistical program (Muthén & Muthén, 1998-2011).

2.3.1. Missing data analysis

Missing data were imputed using the expectation-maximisation (EM) algorithm
(Dempster, Laird, & Rubin, 1977) implemented in SPSS in study II, III, and IV.
Concerning study III, only the time point two KTLT scores were imputed.
Additionally, the full information maximum likelihood estimation procedure
implemented in Mplus that uses all available data in the analyses was
implemented in the latent change score models in study III, and in the structural
equation path model in study I. These approaches are superior to deletion
methods (e.g., listwise deletion) as they produce unbiased parameter estimates
when data are missing at random (MAR), while listwise deletion requires that
data are missing completely at random (MCAR) (Enders, 2010).

2.3.2. Confirmatory factor analysis and latent factor path analysis

In confirmatory factor analysis (CFA) an a priori factor model is fitted to the
data. The factor model (measurement model) consists of a set of observed
variables that serve as factor indicators and the latent factors that are not
observed directly. This allows the researcher to test if the a priori factor structure
that is based on theoretical knowledge fit the data, and to control for
measurement error in the latent factor(s) (Brown, 2006). The Mplus statistical

37
 
program was used to estimate the parameters of the model using maximum
likelihood estimation. Goodness of fit was evaluated using a combination of fit
indices. We used the Comparative Fit Index (CFI), the Tucker–Lewis Index (TLI)
and the Root Mean Square Error of Approximation (RMSEA) as model-fit
indicators. The CFI and TLI vary along a 0-to-1 continuum, and values greater
than 0.90 and 0.95 typically reflect acceptable and excellent fit to the data,
respectively. RMSEA values of less than 0.05 and 0.08 reflect a close fit and a
reasonable fit to the data, respectively (Marsh, Hau, &Wen, 2004).
Latent factor path analysis (SEM) combine CFA and multivariate regression
analysis. The structural relations between the latent constructs) and possible
covariates are of interest in SEM. Usually, the researcher starts with CFA to
investigate the plausibility of the measurement model, and after that moves to
investigate the structural part of the model with SEM. Model fit indices are used
in a similar way as in CFA to examine how well the model fits the data. CFA and
SEM were used in study I, while multiple group CFA and multiple group SEM
were used in study II. In study III we used longitudinal CFA and latent change
score analysis and in study IV, CFA and latent profile analysis.

2.3.3. Multiple group confirmatory factor analysis

Multiple group CFA is an extension of the CFA model. The goal with this analysis
is to determine if the measures used in the study are invariant across groups (i.e.,
measure the same constructs in the same way) to be able to reliably compare groups
on the latent constructs under study. To test for measurement invariance, a series
of nested models are specified, where the endpoints are the least restrictive model
with no invariance constraints, and the most restrictive model that constrains all
parameters to be the same across all groups (Bollen, 1989). If the goal is to compare
groups in structural relations between the constructs, it is sufficient if the factor
loadings are invariant across groups (metric invariance). Invariant factor loadings
and intercepts are required if the goal is to compare groups in latent means (scalar
invariance). One can use chi-square difference testing to measure the statistical
significance between nested models (Bentler, 1990). Another viable option is to
compare fit indices between models (Chen, 2007). Since chi-square difference
testing tends to be biased toward significance when the sample size is large (see

38
 
Marsh, Hau, Balla, & Grayson, 1998), we decided to compare the models in terms
of fit indices (CFI and RMSEA). According to Chen (2007), support for the more
parsimonious model requires a change in CFI (ΔCFI) of less than .01 or a change
in RMSEA (ΔRMSEA) of less than .015.

2.3.4. Longitudinal confirmatory factor analysis

Longitudinal CFA is another extension of the CFA model bearing resemblance
to multiple group CFA. Instead of establishing measurement invariance across
groups, the goal of longitudinal CFA is to ensure the researcher that the
constructs under study are invariant over time. With this approach, the
researcher can confidently rule out the possibility that changes in the
measurement model, or measurement error would account for the temporal
changes in the constructs under study. The invariance testing follows similar
steps as in multiple group CFA. A series of nested models are compared from the
least restrictive to the most restrictive model. If the goal is to study changes over
time, scalar invariance is required (equal factor loadings and intercepts). As in
multiple group CFA we decided to compare the models in terms of fit indices
(CFI and RMSEA). According to Chen (2007), support for the more
parsimonious model requires a change in CFI (ΔCFI) of less than .01 or a change
in RMSEA (ΔRMSEA) of less than .015.

2.3.5. Latent change score analysis

Latent change score (LCS) analysis is a special case of longitudinal structural
equation modelling (McArdle & Nesselroade, 2014). The purpose of LCS analysis
is to investigate differences in latent means over time. When using observed
change scores in analyses it is not possible to separate “true” change from
measurement error (e.g., Cronbach & Furby, 1970). LCS are part of the model
rather than part of the data, as the LCS can be modelled as a latent variable in the
model. There are several different ways to specify the model to obtain a LCS
factor (McArdle & Nesselroade, 2014; Steyer, Partchev, & Shanahan, 2000), but
a common feature in all of these approaches are that they require longitudinal
measurement invariance of the constructs under study. In the approach used in
study III, both the initial level (LEVEL) and change (CHANGE) were modelled

39
 
as latent variables using the same rationale as in latent growth curve modelling.
In the model, factor loadings are invariant over time, all intercepts are fixed at
zero, and residuals of the corresponding items over time points are correlated.
The coefficients of the LEVEL are fixed at one and the path from the CHANGE
factor to time point one latent factor is fixed at zero, and the path to time point
two latent factor is fixed at one as in conventional latent growth curve modelling.
However, in this approach the residual variances of the time point one and two
latent factors are fixed at zero so that the LEVEL = time point one factor score,
and CHANGE = time point two factor score – time point one factor score.

2.3.6. Latent profile analysis

To classify students into homogenous groups with similar patterns of reading and
mathematics performance, and academic well-being we used latent profile analysis
(LPA) in study IV. LPA is a probabilistic or model-based variant of traditional
cluster analysis (Vermunt & Magidson, 2002), which goal is to identify the smallest
number of latent classes (groups) that adequately describe the associations among
observed continuous variables. In the analyses one class is added each step until
the model optimally fits the data. For choosing the best fitting model, Bayesian
Information Criteria (BIC) and Vuong-Lo-Mendell-Rubin (VLMR) likelihood
ratio test were used as the statistical criteria. A decrease in BIC when an additional
class is added indicates an improvement in model fit. Regarding VLMR, a resulting
p value less than .05 indicates that the estimated model is preferable over the
reduced model (Lo, Mendell, & Rubin, 2001). Furthermore, the usefulness and
interpretableness of the latent classes were also considered.

2.3.7. Analyses of variance and covariance

Analysis of variance (ANOVA) is used to investigate group differences in a
dependent variable. Three one-way ANOVAs were used in study I to investigate
differences between mathematics performance groups in three language
measures. Analysis of covariance (ANCOVA) is used to investigate group
differences in a dependent variable while controlling for a continuous variable
(or several). Three ANCOVAs were used in study I to investigate differences
between mathematics performance groups in one language measure at a time

40
 
while controlling for the two other language measures. In study IV, two-way
ANCOVAs were performed to investigate differences between the clusters
(identified by means of LPA) in the performance and well-being measures, while
controlling for the effects of gender and parental educational level.

2.3.8. Chi-square tests and adjusted residuals

Independent samples chi-square test (χ2) is used to determine if there is an
association between two categorical variables. The χ2 –test was used in study III
to investigate the association between MLD groups and educational track and
gender. While the association between the latent profiles and dropout was
investigated in study IV. In a crosstabulation, observed frequencies are compared
to expected frequencies, and the χ2-test determines if the difference is statistically
significant. To determine exactly which cells’ observed frequencies differ from
the expected frequencies, we investigated the adjusted residual in each cell. If the
residual is over the critical value of 1.96 or under -1.96 in a z-distribution the
observed frequencies differ significantly from the expected frequencies.

2.3.9. Configural frequency analysis

To examine the stability and change of mathematical learning difficulties over time
(study III), configural frequency analysis was applied to the data (von Eye, Spiel, &
Wood, 1996). With this approach it is possible to identify patterns of change (or
stability) that are more frequent (type) or less frequent (antitype) than would be
expected based on some chance model. The first order configural frequency
analysis, which assumes that variables under study may show main effects but no
interaction effects, was selected as the baseline model for expected frequencies. The
observed frequencies from the cross classification of the time point 1 and time
point 2 math grouping were compared to the expected frequencies to identify types
and antitypes of change and stability in the three groups.

Vivamus imperdiet, nibh ornare varius eleifend, mauris ligula ultrices ante, sed dapibus leo neque
id augue. In aliquam, massa in posuere ullamcorper, ante dui suscipit lectus, in venenatis tortor
neque id odio.

41
 
3. Overview of the Original Studies

3.1. Study I

3.1.1. Aims
To The aim of this study was to investigate the connection between different
language skills (i.e., word comprehension, reading comprehension and spelling)
and mathematical performance from an LD point of view.

3.1.2. Participants and procedure

The participants in this study were 810 Swedish-speaking Grade 9 students (418
girls and 392 boys) from 14 different lower secondary schools in Finland.
Researchers and trained research assistants conducted the measurements with
groups of students in their own schools during ordinary lessons. For the purpose
of this study, the students were divided into eight performance groups based on
their scores on the mathematical achievement test.

3.1.3. Measures

Mathematical skills
The mathematical skills of the students were assessed with the standardized
KTLT paper-and-pencil test (Räsänen & Leino, 2005). The KTLT test contains
40 items, with a correct answer being worth 1 point and an incorrect answer
worth no points. The students had 40 minutes to complete the scale.

Language skills
Students’ language skills were assessed with the LS reading ability test
(Johansson, 2005). The subtests we used were word comprehension, reading
comprehension and a spelling test. The word comprehension and reading
comprehension scales had multiple-choice tasks, and for the spelling test, the
words that need to be written were presented verbally for the students.

3.1.4 Data analysis

Analysis of variance techniques (ANOVA and ANCOVA) were used to examine
group differences in language skills between the mathematical performance

42
 
groups. Furthermore, structural equation modelling was utilized to investigate
the unique predictive value of reading and spelling for mathematical skills.

3.1.5. Results
The one-way ANOVA tests revealed significant differences and large effect sizes
in all measures between the groups. The lowest performing students in
mathematics also had the lowest scores in all of the language measures. These
analyses were followed up by 3 one-way ANCOVAs, where one of the language
skills was set as the dependent variable and the two others were added in the
analysis as covariates. The groups still differed in word and reading
comprehension after the adjustments by the covariates, but the group differences
in spelling disappeared when word and reading comprehension was controlled
for. Next, we specified a structural equation model (SEM) where students’
mathematical performance was regressed on a reading factor and a spelling
factor. The model fit the data well and revealed that reading indeed predicted
mathematical performance but that spelling did not, thus supporting the results
from the ANCOVAs.

3.1.6. Discussion
In this study, the relationship between language and mathematical performance
was investigated. The group comparisons showed that students with low
mathematical performance also had low scores in word and reading
comprehension as well as in spelling. This is in line with previous findings that
many students have problems in both mathematics and reading. However, the
groups did not differ in spelling when word and reading comprehension was
controlled for. These results suggest that spelling and mathematics do not seem
to have a strong relationship in this age group. This was further supported by the
results from the SEM analyses, where the path from spelling to mathematics was
non-significant. This is not in line with previous studies that found a relationship
between spelling and mathematics. The explanation for these contradicting
results could be that most previous studies (1) investigated younger children and
(2) did not control for other language skills.

43
 
 This study has some limitations. It was cross-sectional, whereby the
directionality of the effects in our model would need to be replicated in a
longitudinal design. It would also be of interest to use measures that are more
focused on certain domains in mathematics; in that case, it would be possible to
pinpoint where language skills overlap with mathematical skills more
specifically. Future studies should apply more detailed measures of both skills in
a longitudinal setting.

3.2. Study II

3.2.1. Aims
The aim of this study was to examine lower secondary school students’ pathways
to educational aspirations. The study used a multigroup SEM to investigate the
predictions of academic self-concept, school burnout, achievement and interest
in mathematics and reading in relation to the educational aspirations of boys and
girls.

3.2.2. Participants and procedure

The participants in this study were 1,152 Swedish-speaking Finnish students (576
girls and 576 boys). All students were measured at the end of Grade 9 (mean age
= 15.8 years, SD = 4.9).

3.2.3. Measures

Mathematics achievement
The mathematical skills of the students were assessed with the standardized
KTLT paper-and-pencil test (Räsänen & Leino, 2005). The KTLT test contains
40 items, with a correct answer being worth 1 point and an incorrect answer
worth no points. The students had 40 minutes to complete the scale.

Reading achievement
Students’ reading skills were assessed with the LS reading ability test (Johansson,
2005). The subtests we used were word comprehension and reading
comprehension that both consisted of multiple-choice tasks.

44
 
Academic self-concept
Academic self-concept was measured with 8 items from the Perceived
Competence Scale for Children (Harter, 1982).

School Burnout
School burnout was assessed using the SBI inventory, which is 9-item scale
developed by Salmela-Aro and Näätänen (2005). The inventory consists of three
subscales: exhaustion, cynicism and inadequacy.

Interest in mathematics and Swedish

To measure students’ interest in mathematics and Swedish (i.e., mother tongue),
we used teacher ratings.

Educational aspirations
We measured students’ idealistic and realistic educational aspirations according
to two statements on a 3-point Likert scale (vocational upper secondary
education, polytechnic education and university education): ‘Highest academic
degree I want to achieve’ and ‘Highest academic degree I will probably achieve’.

3.2.4. Data analysis

We conducted multiple-group confirmatory factor analyses (CFA) to establish
measurement invariance for boys and girls and multiple group SEM path
modelling to answer our research questions.

3.2.5. Results
Overall, academic self-concept was the strongest predictor of educational
aspirations for both genders. Mathematics achievement was more important than
reading achievement for boys’ educational aspirations, while for girls, reading
achievement was more important. The effect from reading achievement to
aspirations was indirect, mediated by interest in reading for both genders.
Mathematics achievement predicted aspirations directly for boys, but for girls, the
effect was mediated by an interest in mathematics. However, interest in
mathematics did not predict boys’ educational aspirations. School burnout had a
negative indirect effect on aspirations through interest for both genders.
Surprisingly, school burnout also had a direct positive effect on aspirations for girls.

45
 
3.2.6. Discussion
The aim of this study was to examine gendered pathways to educational
aspirations. We found that student achievement in mathematics and reading and
academic well-being were related to educational aspirations alongside
motivational beliefs and that these relationships somewhat varied as a function
of gender.
Overall, mathematics achievement was more important than reading
achievement for boys’ educational aspirations, while girls showed the opposite
pattern. This finding partly reflects the rationale behind the I/E frame of
reference model (Marsh, 1986) and gender-typical comparisons. It seems that
boys identify themselves more strongly with the mathematics domain than with
the reading domain and, consequently, place more emphasis on mathematics
achievement, while the opposite pattern may be more likely for girls.
However, there was still small significant indirect effects through interest in
the gender-atypical domain for both genders. This finding suggests that within
the domains that are less important for a particular student, in order for the
student’s achievement to have an effect on his/her educational aspirations, the
student needs to be interested in the particular subject. It is also worth noting
that mathematics achievement had a direct effect on educational aspirations for
boys but not for girls. This is in line with previous studies that have found that
boys tend to rely more on their own achievement in particular subjects in their
decision-making processes, while girls’ choices are more directed by their
interest towards the subject.
Academic self-concept had a positive direct effect on educational aspirations
and was the strongest predictor among the variables for both boys and girls. This
could reflect the nature of the transition to upper secondary education in the
Finnish school system. In general, it is more difficult to get admitted to the
academic track compared to the vocational track, and therefore students’ ability
beliefs are important. Academic self-concept is also a domain-general construct
whereas the interest and achievement variables are domain-specific and hence
share some of the variance in the prediction of educational aspirations. More
importantly, having domain-specific self-concept in the model could have
revealed additional gendered pathways to educational aspirations.

46
 
 In line with recent studies on the consequences of students’ socio-emotional
strain, school burnout negatively predicted interest in both reading and
mathematics and, through them, educational aspirations. However, we also
found a positive direct effect of school burnout on aspirations for girls, which
could indicate that ambitious and success-oriented girls invest more effort in
school and may experience higher levels of school burnout while at the same time
holding high educational goals and plans.

3.3. Study III

3.3.1. Aims
The aim of this study was to investigate the change and stability of MLDs and
academic well-being in adolescent students. Students were divided into three
groups based on their mathematics performance measured in Grade 9 and 2
years later when they were in an upper secondary school. Furthermore, students
with persistent MLDs were compared to low-achieving and typically-achieving
students on measures of academic self-concept and school burnout over two time
points.

3.3.2. Participants and procedure

The participants in this study were drawn from 1,152 Swedish-speaking
Finnish students (576 girls and 576 boys). All students were measured at the
end of Grade 9 (mean age = 15.8 years, SD = 4.9). Of the original 1,152 students,
980 students completed the mathematics achievement test at time point 1. In
this study, these students were included in the analyses. The follow-up took
place during the spring of 2010 when the students were attending either a
vocational or general upper secondary school (mean age = 17.9 years, SD = 4.9
months). The first data collection (2008) was performed by the researchers and
trained research assistants who conducted the measurements with groups of
students in their own schools during ordinary lessons. The second data
collection (2010) was performed by trained research assistants who conducted
the measurements with groups of students in their respective schools during or
after the school day.

47
 
3.3.3. Measures

Mathematical skills
The mathematical skills of the students were assessed using the standardized
KTLT paper-and-pencil test (Räsänen & Leino, 2005). The KTLT test contains
40 items, with a correct answer worth 1 point and an incorrect answer worth no
points. The students had 40 minutes to complete the scale.

Academic self-concept
Academic self-concept was measured with 8 items from the Perceived
Competence Scale for Children (Harter, 1982).

School Burnout
School burnout was assessed by the SBI, which is 9-item scale developed by
Salmela-Aro and Näätänen (2005). The inventory consists of three subscales:
EXH, CYN and INAD.

Upper secondary educational track

Information on the students’ educational track (vocational or academic) was
collected from the schools. First, information from all 14 comprehensive schools
where the students had applied was collected. Second, the upper secondary
schools were contacted to get confirmation of whether the students were
studying there. If this was not the case, the students themselves were contacted
to get the information.

3.3.4. Data analysis

Configural frequency analysis was used to investigate the change and stability of
typically-achieving students, low-achieving students and students with MLDs.
We used an independent sample chi-square test (χ2) to determine if there were
an association between group membership and educational track and between
group membership and gender. Longitudinal CFAs were conducted to establish
measurement invariance of the academic self-concept and school burnout
constructs of the two time points. Latent change score modelling was used to
investigate changes in the academic well-being measures from Grade 9 to upper
secondary school and if typically-achieving students, Low-achieving students
and students with MLDs differed in this aspect. Missingness of data was also

48
 
analysed, and the expectation-maximization algorithm was used to impute
missing values of the mathematics achievement test at time point 2. Furthermore,
both maximum likelihood estimation procedures (which assume that data are
missing at random) and pattern-mixture modelling (which can be used when
data are not missing at random) were used in the latent change score models to
ensure unbiased results in relation to missing data.

3.3.5. Results
We used the 25th percentile as the cut-off value for typically-achieving, low-
achieving was defined as scores between the 10th and 25th percentile, and MLD
was defined as scoring under the 10th percentile. All three groups showed
stability (91%, 70% and 87%) from Grade 9 to secondary school, meaning that
these configurations occurred more often than expected from chance (p < .001).
Four untypical patterns (antitypes of change) were also identified, meaning that
these change configurations occurred less often than expected from chance (p <
.001). These were all configurations that incorporated students who had been
identified as typically-achieving at one time point and low-achieving or MLD at
the other time point. Group membership was associated with each students’
secondary education track [χ2(2, N = 820) = 134.49, p < .001], showing that
students in the typically-achieving group (z = 11.2, p < .001) were more likely to
study at general upper secondary schools, whereas students in the low-achieving
(z = -6.9, p < .001) and MLD (z = -8.3, p < .001) groups were more likely to study
at vocational upper secondary schools. The gender distribution was not
significantly different across groups [χ2(2, N = 885) = .44, p = .80], indicating that
girls and boys showed similar prevalence rates of low-achieving and MLD.
Both academic self-concept and school burnout were invariant over the two
time points. Concerning academic self-concept, both the low-achieving and
MLD groups had lower scores in Grade 9 compared to the typically-achieving
group. The magnitude of the regression paths from the low-achieving and MLD
groups were similar, indicating no differences in academic self-concept in Grade
9 between them. The students in the three groups showed similar changes in
academic self-concept from Grade 9 to upper secondary school. The academic-
track students had higher academic self-concept in Grade 9, but their change was

49
 
negative, while the vocational-track students had lower academic self-concept in
Grade 9, but they increased in upper secondary school. The interaction between
the educational track and low-achieving group was marginally significant (p =
.06) on the intercept in academic self-concept. Low-achieving students that
continued to the academic track did not differ in academic self-concept
compared to the low-achieving students continuing to the vocational track as
much as students within the typically-achieving and MLD groups.
Vocational track students in general experienced higher levels of school
burnout in Grade 9, but there was also a significant educational track by MLD
group interaction (p < .05) on level of school burnout. The MLD students
continuing to the academic track clearly showed a different pattern. They
experienced the highest levels of school burnout, while within the other groups,
the academic-track students had lower school burnout compared to the
vocational-track students. Change in school burnout was predicted by
educational track, educational track and low-achieving group and educational
track and MLD group. The typically-achieving students in the academic track
showed the highest increase in school burnout compared to all other groups,
while the typically-achieving students in vocational education experienced a
decrease in school burnout. The interaction pattern was similar in the low-
achieving students, although the academic-track students within this group had
more or less stable school burnout across both time points. The MLD group’s
interaction with the educational track was different; the vocational-track
students had a smaller decrease in school burnout compared to the other groups,
whereas the academic-track students showed the biggest decrease in SB
compared to all other groups.

3.3.6. Discussion
The aim of this study was to investigate change and stability in MLDs and
academic well-being across the transition to upper secondary school. The
configural frequency analyses showed that the grouping of students into
typically-achieving, low-achieving and MLD groups was stable during this 2-
year period. However, consistent with the literature, not all students classified
as MLD remained in this group. This supports the use of the persistence

50
 
criterion when identifying students with MLD. The changes in academic self-
concept were mainly explained by the educational track of the students—the
academic-track students experienced a decrease in academic self-concept, while
the vocational-track students experienced an increase in academic self-concept.
This result mirrors findings concerning the social comparison effect in the
literature (e.g., the BFPLE). The low-achieving and MLD groups had lower
academic self-concept in Grade 9 compared to the typically-achieving group,
but similar changes in academic self-concept. The only difference found
between the low-achieving students and students with MLD was that the low-
achieving students in the academic track did not differ as much in academic self-
concept compared to the vocational-track low-achieving students as was the
case within the other groups. Educational track continued to be an important
predictor also when investigating school burnout. However, here the interaction
effects played a more dominant role, and especially students with MLD in the
academic track exhibited differing patterns in both the level (Grade 9) and
change of school burnout. It seems that these students had to put a lot of effort
into their studies in Grade 9 to be able to continue to the academic track. This
is reflected by their very high levels of school burnout compared to all other
groups. Interestingly, they also experience the sharpest decline in school
burnout in upper secondary school, which could indicate that they mainly chose
other non-mathematics subjects in the academic track, thus lessening the impact
of their MLDs on their school burnout. It is worth noting that the typically-
achieving students in the academic track experienced an increase in school
burnout, while the low-achieving students in the academic track show a more
stable trend in their school burnout.

3.4. Study IV

3.4.1. Aims
The aim of this study was to examine lower secondary students’ performance
(reading and mathematics) and academic well-being (academic self-concept,
perceived learning difficulties and school burnout) profiles and the relationship
between these profiles and dropping out of school.

51
 
3.4.2. Participants and procedure
The participants in this study were 1,152 Swedish-speaking Finnish students (576
girls and 576 boys). All students were measured at the end of Grade 9 (mean age
= 15.8 years, SD = 4.9). Research assistants tracked down students by phone and
social media who did not participate in the second data collection. This way we
could determine educational status (i.e., whether they were in school or had
dropped out) for all but two students.

3.4.3. Measures

Mathematical skills
The mathematical skills of the students were assessed with the standardized
KTLT paper-and-pencil test (Räsänen & Leino, 2005). The KTLT test contains
40 items, with a correct answer worth 1 point and an incorrect answer worth no
points. The students had 40 minutes to complete the scale.

Reading skills
Students’ reading skills were assessed with the LS reading ability test
(Johansson, 2005). The subtests we used were word comprehension and
reading comprehension, which both consist of multiple-choice tasks, and a
spelling task.

Academic self-concept
Academic self-concept was measured with 8 items from the Perceived
Competence Scale for Children (Harter, 1982).

Perceived learning difficulties

Perceived learning difficulties were assessed by 9 items measuring how much
difficulty a student experienced while competing their schoolwork (e.g., taking
notes in class, understanding instructions on the whiteboard, etc.). Each item
was assessed using a 7-point Likert scale ranging from 1 (not at all) to 7 (very
much).

School burnout
School burnout was assessed by the SBI, which is 9-item scale developed by
Salmela-Aro and Näätänen (2005). The inventory consists of three subscales:
exhaustion, cynicism and inadequacy.

52
 
3.4.4. Data analysis
To classify students into homogenous groups with similar patterns of reading
and mathematics performance and academic well-being, we used latent profile
analysis (LPA). LPA is a probabilistic or model-based variant of the traditional
cluster analysis (Vermunt & Magidson, 2002), the goal of which is to identify the
smallest number of latent classes (groups) that adequately describe the
associations among observed continuous variables. We used an independent
sample chi-square test (χ2) to determine if there were an association between
group membership and dropout. Missing data were handled with imputation of
missing values with the expectation-maximization algorithm, and the proposed
factor structure of the measures was tested with CFAs.

3.4.5. Results
The CFAs confirmed the factor structure of the performance and academic well-
being measures. In order to group students into different academic performance
and well-being profiles, a series of LPAs was conducted. The analyses revealed
that a four-group solution best described the data. The four groups were labelled
according to the mean score of their profiles in the academic performance and
well-being measures as high-performing, average-performing, low-performing
and negative academic well-being.
ANCOVAs were conducted to investigate group differences in the latent
factor scores of all the measures while controlling for the effects of parental
educational background and gender. Six students had missing information on
gender and were therefore not included in the analyses. The students in the high-
performing group (n = 391, 34%) performed significantly better on all
performance measures compared to the other groups, reported lower levels of
school burnout and perceived learning difficulties and displayed the highest level
of academic self-concept. The average-performing students (n = 473, 41%)
represented students with typical performance and average levels of academic
well-being. They tended to perform better and display higher academic well-
being compared to the low-performing group and the negative academic well-
being group.

53
 
 The low-performing students (n = 204, 18%) exhibited severe learning
difficlties in reading, spelling and mathematics (mean scores < -1.27 SD) and
reported lower levels of academic well-being compared to the high-performing
and average-performing students. The negative academic well-being group (n =
78, 7%) represented students that, despite their average performance levels,
reported the lowest levels of academic well-being. They had lower academic self-
concept and reported higher levels of perceived learning difficulties and school
burnout, even when compared to the low-performing group. The χ2-test revealed
that students from the low-performing group and negative academic well-being
group exhibited a higher risk of dropping out of school compared to the other
groups. We also investigated if students who had dropped out from school had
lower parental educational levels, but this was not the case.

3.4.6. Discussion
We found support for four distinct academic performance and well-being
profiles that showed similar performance levels in mathematics, reading and
spelling within groups. The low-performing group can be viewed as
corresponding to the MLDRD group, as these students performed at a lower level
in mathematics and all reading measures compared to the other groups. We also
identified one group of students who, despite their rather average performance,
experienced the lowest levels of academic well-being. This group exhibited an
increased risk for dropping out of school, which is in line with previous studies.
A common feature for these students is that they exhibit maladaptive patterns of
functioning at school, and consequently, they are in danger of dropping out of
school. Furthermore, one group of average-performing students and one group
of high-performing students were identified.
Academic performance in reading, spelling and mathematics was strongly
related to academic well-being except for in the group with average performance
and low well-being. These findings support previous studies that point to the
need to also consider academic well-being in school settings, as there are clearly
students who do not have difficulties performing academically but still do not
feel comfortable in school.

54
 
4. General Discussion

This dissertation study investigated the relationships among learning difficulties,

academic well-being and educational pathways. Overall, the results of this
dissertation demonstrated that learning difficulties in mathematics and reading
were closely related and had a strong influence on educational aspirations and
school dropout (Figure 2). Learning difficulties in mathematics and reading were
also associated with lower academic well-being. Furthermore, academic well-
being was found to be related to school dropout and educational aspirations, and
it developed differently in general upper secondary schools and vocational upper
secondary schools.

Figure 2. Empirical model of adolescent students’ educational pathways based on the

results of this dissertation study, where learning difficulties and academic well-being
predict educational outcomes; ASC = academic self-concept, SBO = school burnout, PLD
= perceived learning difficulties, ED ASP = educational aspirations, educational track =
general or vocational upper secondary education. Note: The red arrows represent
relationships found only for girls, and the blue arrows represent relationships found only
for boys.

55
 
4.1. Learning Difficulties in Mathematics and Reading
The first aim of this dissertation was to investigate the relationship between
learning difficulties in mathematics and reading. In study I, the focus was on
examining how different subskills in reading are related to mathematical
performance. Previous work has mainly focused on how reading skills in general
(e.g., Jordan et al., 2003) or dyslexia in particular (Landerl et al., 2009; Simmons
& Singleton, 2008) are related to mathematical performance. In line with
previous research, a strong relationship between mathematical performance and
word and reading comprehension was found, but only a weak or non-existent
relationship between spelling and mathematical performance.
Previous studies (Ostad, 1998) have found mathematical performance and
spelling to be related, but these studies have mainly focused on younger children
and did not control for other reading skills. The results from study IV support
and complement these findings. In study I, the relationship was studied using
variable-centred methods (i.e., SEM path-model), whereas in study IV, a person-
centred method was adopted to study learning difficulties in mathematics and
reading (latent profile analysis). The analyses identified one group of students
with learning difficulties in both reading and mathematics.
The empirical data did not support the traditional grouping used in learning
difficulties research (RD-only, MLD-only and MLDRD groups). However, both
MLDs and RDs may incorporate different subtypes (e.g., Skagerlund & Träff,
2014), and not all students who show low achievement in mathematics/reading
at one time point have persistent and severe learning difficulties (e.g., Fletcher et
al., 2005). In this study, only performance measures were included in the latent
profile analyses, but to separate, for example, students with RD from low-
achieving students, cognitive indicators such as phonological processing
(Vellutino et al., 2004) or naming speed (Kirby et al., 2010) would have needed
to be included. In previous studies low-achieving students have been found to
differ from typical-achieving students in general domain skills (e.g., working
memory), while students with MLDs and RDs differ from low- and typical-
achieving students in domain-specific measures (e.g., number sense, naming

56
 
speed). It is likely that the learning difficulties group identified in study IV also
included low-achieving students.
These results indicate that if learning difficulties in respective domains are
defined with a low score on a performance measure, a lot of students will be
categorized as having MLDRD in this age group. This result could reflect the
viewpoint that some shared-domain general skills like working memory (e.g.,
Kyttälä, 2008) can explain this co-occurrence. Moreover, as the results from
study I highlight, mathematical skills in general require good language skills in
terms of word and reading comprehension (LeFevre et al., 2010; Vukovic &
Lesaux, 2013b), which also contributes to the strong relationship between
mathematics and reading. Particularly in this age group, mathematics often
involves word problem solving that requires good language skills and good
working memory capacity.

4.2. Learning Difficulties and Academic Well-Being

Overall, students’ mathematics performance (studies II, III and IV) and reading
performance (studies II and IV) were related to lower academic well-being. More
specifically, students with learning difficulties in mathematics or/and reading
had lower academic self-concept and higher school burnout and perceived
learning difficulties in Grade 9 compared to their typically performing peers. In
light of the literature and the conceptualization of academic well-being in
general, as well as the indicators used in this study in particular, these results are
expected. Academic self-concept refers to students’ beliefs in their ability in
academic domains and incorporates a social comparison dimension (Marsh,
1986). Therefore, it is hardly surprising that students who perform at a lower
level compared to their peers do not feel as competent in school settings. If we
had measured domain-specific self-concepts in reading and mathematics, the
results might have changed. For example, students with only MLD would
probably exhibit low mathematics self-concept but might experience similar
levels of reading self-concept as typically-achieving students. Based on the I/E
frame of reference model (Marsh, 1986), it might even be that they would show

57
 
relatively high reading self-concept compared to their actual skill level as a result
of the internal comparison process.
Concerning school burnout, the higher levels experienced by students with
learning difficulties probably reflects that they have to work harder in school,
thus depleting their psychological resources (Tuominen-Soini & Salmela-Aro,
2014). Another, interpretation of the relationship between school burnout and
learning difficulties found in the literature on motivation and well-being is that
these students give up easily on academic tasks and do not invest effort in their
school work (Tuominen-Soini et al., 2012). Similarly, perceived learning
difficulties were also related to learning difficulties. This is understandable, as
students with learning difficulties most likely struggle to follow the general
instructions given in the classroom and are in need of additional support. It is
also worth noting that we also found a group of students in study IV that showed
average performance in reading and mathematics but exhibited lower levels of
academic well-being compared to the learning difficulties group. This finding
reflects that low academic well-being at a given point in time is not always related
to learning difficulties and, consequently, learning difficulties are not always the
cause for low academic well-being. However, it is plausible to assume that
negative academic well-being can, over time, result in lower achievement in these
students.

4.3. Changes in Academic Well-Being Related to Mathematical

Learning Difficulties Across the Transition to Upper
Secondary Education
Study III investigated changes in academic well-being measures from Grade 9 to
upper secondary school in relation to MLDs. In Grade 9, the students who
continued to the academic track had higher levels of academic self-concept
compared to students who continued to the vocational track. This was expected,
as academic self-concept is strongly related to achievement level (e.g., Marsh et
al., 2002). Furthermore, as expected, the low-achieving and MLD students
showed lower levels of academic self-concept compared to the typically-
achieving students (Bear et al., 2002; Zeleke, 2004). On the whole sample level,

58
 
there was an increase in academic self-concept in the vocational track and a
decrease in academic self-concept in the academic track compared to Grade 9.
This result fits well with the BFLPE (Marsh & Craven, 2002), as students in the
academic track face tougher competition in terms of social comparisons
compared to compulsory education, while the opposite pattern is true for
vocational-track students. Although this effect has also been found when
comparing students with learning difficulties in inclusive versus segregated
settings (Marsh et al., 2006), we did not find interaction effects between
educational track and the typically-, low-achieving and MLD groups on change
in academic self-concept. This probably reflects that in both tracks, the low-
achieving and MLD students are still performing at the lower end, resulting in
similar social comparison processes.
Students from all three groups who continued to the vocational track reported
a decrease in school burnout compared to Grade 9. However, differences were
found between all groups among students in the academic track. Typically-
achieving students in the academic track exhibited an increase in school burnout
compared to Grade 9, which is line with previous studies (Salmela-Aro et al.,
2008; Salmela-Aro & Tynkkynen, 2012; Salmela-Aro & Upadyaya, 2014b).
Interestingly low-achieving students showed a stable trend, and MLD students
showed a decreasing trend in school burnout compared to Grade 9. These
students have probably chosen to focus on other subjects than mathematics, thus
reducing the negative impact of their difficulties in mathematics on their
academic well-being. Therefore, it seems plausible that in Grade 9 these students
invested effort into mathematics because they had been aiming for the academic
track, thus experiencing more school burnout, but once they were in the
academic track, they directed their motivation and effort to areas other than
mathematics. This pattern was even more pronounced in the MLD students in
the academic track who had the highest level of school burnout in Grade 9 but
exhibited the steepest decrease in school burnout during the educational
transition. This finding is similar to a study by Salmela-Aro and Upadyaya
(2014b), who found that those students with a high initial and decreasing change
in school burnout held lower educational aspirations than groups with either
low-increasing or low-stable school burnout profiles. The results in study III

59
 
reflect similar processes in the MLD group (in the academic track), where these
students had decided not to pursue educational pathways that included
mathematics, thus reducing the negative impact of their MLD on their academic
well-being. As stated previously, including domain-specific measures of self-
concept would provide a more detailed picture of how students’ beliefs in their
ability develop in relation to their learning difficulties in mathematics. Similarly,
school burnout is also considered a domain-general measure, and therefore, for
example, perceived cost in the domain of mathematics could be of interest here.

4.4. Learning Difficulties and Academic Well-Being Predicting

Educational Aspirations and Potential for Dropping Out
The fourth aim of this study was to investigate the contribution of learning
difficulties in mathematics and reading and academic well-being on outcomes
such as educational aspirations (study II) and school dropout (study IV).
Concerning educational aspirations, student achievement in mathematics and
reading and academic well-being were related to educational aspirations
alongside motivational beliefs and that these relationships somewhat varied as a
function of gender. These findings indicate that overall, students who are low-
achieving or who have learning difficulties in mathematics and reading hold
lower educational aspirations compared to their typically-achieving peers.
Moreover, in line with previous research, boys identify themselves more strongly
with the mathematics domain than with the reading domain and, consequently,
place more emphasis on mathematics achievement, while the opposite pattern
may be more likely for girls (Eccles, 2009; Nagy et al., 2006; Meece, Glienke, &
Burg, 2006). Therefore, it is plausible to assume that MLDs are more detrimental
for boys’ future educational pathways, while RDs are more detrimental for girls.
Another interesting finding is that boys tend to rely more on their own subject
achievement (i.e., mathematics performance) in their decision-making processes
(Nagy et al., 2006; Viljaranta et al., 2009), while girls’ choices are more directed
by their interest towards the subject (Chow et al., 2012; Watt et al., 2012).
The results also show that learning difficulties are not by themselves sufficient
to explain educational aspirations. Academic well-being in terms of academic

60
 
self-concept and school burnout uniquely predicted educational aspirations in
the model. Students that hold higher ability beliefs (academic self-concept) set
higher educational goals, while students that experience school burnout loose
interest in their school work, which leads to lower educational aspirations. This
pattern was similar for boys and girls, but one interesting difference concerning
school burnout emerged. Higher levels of school burnout among girls seems to
directly influence their educational aspirations positively. This could indicate
that ambitious and success-oriented girls invest more effort in school and may
experience higher levels of school burnout, while at the same time holding high
educational goals and plans (Tuominen-Soini et al., 2008; Tuominen-Soini &
Salmela-Aro, 2014).
Study IV focused on explaining dropout from secondary schools by measures
of performance and well-being. In this study, a person-centred approach was
used to group students and did not result in separate MLD or RD groups, but
only one group of students with combined difficulties in mathematics and
reading. Consequently, it is not possible to determine if MLD or RD predicts
dropout more strongly based on this study. Consistent with previous findings
(Battin-Pearson, et al., 2000; Janosz, et al., 2000), learning difficulties were a good
predictor of dropout from secondary school. Students that formed the lowest-
performing group (combined MLD and RD) had the highest risk of dropping out
of school, whereas students from the average-performance and high-
performance group exhibited the lowest risk of dropping out school. However,
the results also demonstrated that there is a group of students that, while having
average performance levels in reading and mathematics, experience low
academic well-being and consequently have a higher risk of dropping out of
school.
Taken together, these findings show that learning difficulties and academic
well-being are related to both educational aspirations and school dropout.
Moreover, gender was found to moderate the effects of reading and mathematics
achievement on educational aspirations while not being related to dropout. As
gender differences emerged when contrasting the mathematics and reading
domains concerning aspirations, results might have differed concerning dropout
if MLD and RD profiles had been found. The results also show the importance

61
 
to link studies on learning difficulties and well-being to existing theories to get a
more holistic picture of the processes influencing educational outcomes and to
better understand the results obtained. More specifically, the expectancy-value
framework helped us build the model for students’ educational aspirations and
include the interest variables as mediators between the achievement measures
and educational aspirations, whereas the I/E frame of reference model was useful
for interpreting the gendered effects from reading and mathematics achievement
to aspirations, indicating that internal comparison processes can lead to students
opting out from educational pathways that focus on the domain that they feel
less competent in. An example of this would be girls choosing not to pursue
STEM-related educational paths (Nagy et al., 2006; Watt et al., 2012) even though
they perform at similar level in mathematics compared to boys (Else-Quest,
Hyde, & Linn, 2010).

4.5. Strengths and Limitations

This dissertation study has some strengths but also several limitations that need
to be taken into account when interpreting the results. In all the studies latent
constructs were used to take into account measurement error and structural
equation modelling techniques were utilized to obtain model fit indices.
However, concerning the achievement measures, instead of individual items,
item parcels or single indicators of the measures were used. This was done to
reduce the complexity of the models but the trade-off of this approach was that
not all measurement error was parcelled out in the measurement models
concerning reading, mathematics and spelling. Another overall feature of this
dissertation study was that different theoretical starting points were used in the
original studies. This can be seen as a strength as the overall results are linked
and advance previous research in different ways. However, this approach does
not come without limitations as the operationalisations of key-constructs (e.g.,
academic self-concept versus mathematical self-concept) do not always match
that of previous research in a given field. This was due to the fact the empirical
work this thesis is based on, did not use one specific theoretical model as a
starting point, rather a more data-driven approach was adopted.

62
 
 As study I and other studies (e.g., Vukovic et al., 2010) indicate, the
relationship between mathematics and reading varies as a function of how these
constructs are operationalized. Some subskills in reading are more strongly
related to mathematics compared to others, and some areas in mathematics have
higher demands on language than others (Jordan et al., 2003; Vukovic & Lesaux,
2013). Furthermore, both MLD and RD incorporate different subtypes of
difficulties that show deficits in different areas. It is of particular importance to
distinguish between low-achievement and learning difficulties to be able to
advance our current knowledge in MLD and RD research. This can be partly
achieved by including multiple measurement occasions (the persistence
criterion; Fletcher et al., 2005) and incorporating both groups alongside
typically-achieving students in research designs (the thrichotomous approach;
Mazzocco & Räsänen, 2013), as was done in study III.
This dissertation study only included performance measures, but research in
both MLD and RD has identified underlying cognitive skills that are indicative
of these difficulties. More specifically, number processing incorporating
nonsymbolic (ANS) and symbolic magnitude comparison skills are indicative of
MLD (De Smedt & Gilmore, 2011; Mazzocoo et al., 2011; Skagerlund & Träff,
2014). Similarly, deficits in phonologic processing and naming speed are core
features of RD (Vellutino & Fletcher, 2005). Future studies on MLD could—in
addition to the persistence criterion and the thrichotomous approach—also
include one of these cognitive measures when operationalizing MLD.
To overcome the limitations associated with the use of cut-off scores in RD
and MLD research, which vary considerably across studies (from 2.5% to 35%),
person-centred methods offer an alternative way of classifying students.
Advances in statistical research methods offer researchers the possibility of using
model-based groupings of students (e.g., LPA; Vermunt & Magidson, 2002),
thereby omitting the need to use arbitrary cut-off scores in research designs. This
was done in study IV where latent profile analysis was used to classify students.
However, study III used cut-off scores to identify students with low-
performance, and mathematical learning difficulties respectively. This resulted
in another classification of students compared to study IV making it more
difficult to draw coherent conclusions of the findings across studies III and IV.

63
 
This decision was driven by the fact that we had not measured all constructs over
time (Table 1) and the focus in study III was especially on mathematical learning
difficulties.
Concerning educational pathways and educational aspirations in particular,
study II demonstrates the need to incorporate both the reading and mathematics
domains in the model when trying to explain these outcomes and the processes
leading to them. A limitation in our study was that we used a domain general
construct in operationalizing academic self-concept. This probably kept us from
finding more nuanced predictions from the mathematics and reading domains
concerning students’ educational aspirations. Our results also highlight the need
of incorporating possible gender effects in the model.
We have investigated changes in academic self-concept and school burnout
from Grade 9 to upper secondary school among typically-achieving students,
low-achieving students and students with MLD. A limitation in this study was
the high attrition rate at time point 2. Better cooperation with the upper
secondary schools might have decreased the number of students who did not
participate in the second data collection. Some schools allowed the research
assistant to conduct the measurements during the ordinary school day, while in
other schools the data collection had to be performed after the school day.
Another limitation concerning this study was that we only measured students’
academic self-concept and school burnout once before the transition to upper
secondary school. It would be interesting to include multiple measurements
during compulsory education to be able to make stronger claims about the
changes in academic well-being in the educational and vocational tracks,
respectively.
Finally, from a statistical methods view, multiple-group SEM, where both the
measurement model and the structural paths between the constructs under study
are estimated separately for typically-achieving students and students with
learning difficulties, would offer the best and most rigorous approach to
investigating similarities and differences between these groups (Wilson &
Rupley, 2013). A challenge with this type of analysis is the sample size
requirements for the learning difficulties group that, depending on the
complexity of the model, would be challenging to achieve.

64
 
4.7. Pedagogical Implications
This dissertation study also has some pedagogical implications. Overall, the
results concerning the relationship between learning difficulties in reading and
mathematics indicate that a large number of students with MLDs also struggle in
reading. Therefore, interventions that not only support mathematical skills but
also help increase reading comprehension are needed in this age group. A
promising method that has been found effective for students with comorbid
difficulties in reading and mathematics is Schema Broadening Instruction (Fuchs
et al., 2009), which builds on the work of Jitendra and Hoff (1996) and Jitendra,
DiPipi and Perron-Jones (2002) concerning math instruction for word problem
solving. This instructional method incorporates teaching students to understand
the underlying mathematical structure of the problem type, to recognise the basic
problem type and to solve the problem type. Additionally, teaching for transfer
is explicitly incorporated.
Furthermore, in this study, lower academic self-concept was found to be
related to students with learning difficulties. Meta-analytic findings suggest that
interventions that target both academic skills and academic self-concept are the
most effective (O’Mara, Marsh, Craven, & Debus, 2006). Furthermore, students
with learning difficulties seem to profit from interventions that explicitly target
self-concept as opposed to those that merely focus on academic skills training
(Hattie, 1992; O’Mara et al., 2006). Therefore, incorporating elements (e.g.,
praise and feedback) that aim to enhance mathematical self-concept are
warranted. Prior research (Kamins & Dweck, 1999; O’Mara et al., 2006) has
identified praise and feedback as effective components in enhancing self-
concept. Moreover, results from study II and III highlight the importance of
academic self-concept in the decision-making processes concerning educational
pathways, further emphasising the importance of enhancing students’
competence perceptions in academic domains.
A group of students with average performance but low levels of academic
well-being and an increased risk for dropout from upper secondary school was
identified in study IV. This finding emphasises the need to consider academic
well-being more systematically in school settings. In Finnish schools, students

65
 
are systematically screened for possible learning difficulties and given additional
support mainly in the core subjects, such as their mother tongue and
mathematics. Unfortunately, students’ well-being is not followed up in the same
way. Practical means to intervene and to provide all adolescents with adequate
coping strategies should be developed. Concrete advice on how to prevent and
handle feelings of exhaustion and negative affect might save these students from
more severe problems later on, as study-related demands are likely to increase
over the course of their education.
Taken together, this dissertation study indicates that it is possible to identify
students with learning difficulties and low well-being in upper secondary
schools, and that these factors are related to their chances to successfully pursue
their personal educational pathway.
 

66
 
References

Andersson, U. (2010). Skill development in Battle, A., & Wigfield, A. (2003). College
different components of arithmetic and women’s value orientations toward family,
basic cognitive functions: Findings from a 3- career, and graduate school. Journal of
year longitudinal study of children with Vocational Behavior, 62, 56–75.
different types of learning difficulties. Bear, G. G., Minke, K. M., & Manning, M. A.
Journal of Educational Psychology, 102(1), (2002). Self-concept of students with
115-134. learning disabilities: A meta-analysis.
Andersson, U., & Lyxell, B. (2007). Working School Psychology Review, 31(3), 405-427.
Memory Deficits in Children with Bentler, P. M. (1990). Comparative fit indices in
Mathematical Difficulties: A General or structural models. Psychological Bulletin,
Specific Deficit? Journal of Experimental 107(2), 238–246.
Child Psychology, 96, 197-228.
Bollen, K. A. (1989). Structural equations with
Archambault, I., Janosz, M., Morizot, J., & latent variables. New York, NY: John Wiley
Pagani, L. (2009). Adolescent behavioral, & Sons.
affective, and cognitive engagement in
Bong, M., & Skaalvik, E. M. (2002). Academic
school: Relationship to dropout. Journal of
self-concept and self-efficacy: How different
School Health, 79(9), 408-415.
are they really? Educational Psychology
Augustyniak, K., Murphy, J., & Phillips, D. K. Review, 15(1), 1-40.
(2005). Psychological perspectives in
Brown, T. A. (2006). Confirmatory factor
assessing mathematics learning needs.
analysis for the applied research. London:
Journal of Instructional Psychology, 32(4),
The Guilford Press.
277-286.
Burnett, P. C., Pillay, H., & Dart, B. C. (2003).
Badin, N. A. (1983). Dyscalculia and Nonverbal
The influences of conceptions of learning
Disorders of Learning. In H.R. Myklebust
and learner self-concept on high school
(Ed.), Progress in learning disabilities (Vol.5,
students´ approaches to learning. School
pp. 235-264). New York: Stratton.
Psychology International, 24, 54-66.
Bakker, A. B., Schaufeli, W. B., Demerouti, E.,
Butterworth, B. (2005). Developmental
Janssen, P. P. M., Van der Hulst, R., &
dyscalculia. In J.I.D. Campbell (Ed.), The
Brouwer, J. (2000). Using equity theory to
Handbook of Mathematical Cognition (pp.
examine the difference between burnout
455-467). Hove: Psychology Press.
and depression. Anxiety, Stress and Coping,
13(3), 247-268. Bynner, J., & Parsons, S. (2002). Social
exclusion and the transition from school to
Barron, K. E., & Hulleman, C. S. (2015).
work: The case of young people not in
Expectancy-value-cost model of
education, employment or training
motivation. In J. D. Wright (Ed.),
(NEET). Journal of Vocational Behavior,
International Encyclopedia of the Social and
60(2), 289-309.
Behavioral Sciences (pp. 503–509). Oxford:
Elsevier Ltd. Chen, F. F. (2007). Sensitivity of goodness of
fit indexes to lack of measurement
Battin-Pearson, S., Newcomb, M.D., Abbott,
invariance. Structural Equation Modeling,
R.D., Hill, K.G., Catalano, R.F., & Hawkins,
14, 464–504.
J.D. (2000). Predictors of early high school
dropout: A test of five theories. Journal of
Educational Psychology, 92(3), 568-582.

67
 
Chow, A., & Salmela-Aro, K. (2011). Task values Dirks, E., Spyer, G., van Lieshoult, E. C. D. M., &
across subject domains: A gender de Sonneville, L. (2008). Prevalence of
comparison using a person-centered combined reading and arithmetic
approach. International Journal of disabilities. Journal of Learning Disabilities,
Behavioral Development, 35, 202–209. 41(4), 460–473.
Chow, A., Eccles, J., & Salmela-Aro, K. (2012). Domene, J. F., Shapka, J. D., & Keating, D. P.
Task value profiles across subjects and (2006). Educational and career-related help-
aspirations to physical and IT-related seeking in high school: An exploration of
sciences in the United States and Finland. students’ choices. Canadian Journal of
Developmental Psychology, 48(6), 1612– Counselling, 40, 145–159. Retrieved from
1628. http://cjcrcc.ucalgary.ca/cjc/index.php/rcc/
Cronbach, L. J., & Furby, L. (1970). How we article/view/297/678
should measure "change": Or should we? Domina, T., Conley, A. M., & Farkas, G. (2011).
Psychological Bulletin, 74(1), 68-80. The link between educational expectations
Daneman, M., & Carpenter, P. A. (1980). and effort in the college-for-all era. Sociology
Individual differences in working memory of Education, 84(2), 93-112.
and reading. Journal of Verbal Learning and Duncan, G. J., & Brooks-Gunn, J. (2001).
Verbal Behavior, 19(4), 450-466. Poverty, welfare reform, and children’s
De Smedt, B., & Gilmore, C. K. (2011). Defective achievement. In B. Biddle (Ed.), Social class,
number module or impaired access? poverty and education: Policy and practice
Numerical magnitude processing in first (pp. 49–76). New York, NY: Routledge-
graders with mathematical difficulties. Falmer.
Journal of Experimental Child Psychology, Durik, A. M., Vida, M., & Eccles, J. S. (2006).
108(2), 278-292. Task values and ability beliefs as predictors
De Smedt, B., Noël, M-P., Gilmore, C., & Ansari, of high school literacy choices: A
D. (2013). How do symbolic and non- developmental analysis. Journal of
symbolic numerical magnitude processing Educational Psychology, 98(2), 382–393.
skills relate to individual differences in Eccles, J. (2009). Who am I and what am I going
children's mathematical skills? A review of to do with my life? Personal and collective
evidence from brain and behavior. Trends in identities as motivators of action.
Neuroscience and Education, 2(2), 48-55. Educational Psychologist, 44(2), 78–89.
DeGarmo, D. S., & Martinez, C. R. (2006). A Eccles, J. S., & Midgley, C. (1989). Stage-
culturally informed model of academic well- environment fit: Developmentally
being for Latino youth: The importance of appropriate classrooms for young
discriminatory experiences and social adolescents. In C. Ames & R. Ames (Eds.),
support. Family Relations, 55(3), 267–278. Research on motivation in education (pp.
Dehaene, S. (2011). The number sense: How the 139–186). San Diego: Academic Press.
mind creates mathematics, revised and Eccles, J. S., Adler, T. F., Futterman, R., Goff, S.
updated edition. New York: Oxford B., Kaczala, C. M., Meece, J. L., & Midgley,
University Press. C. (1983). Expectancies, values and
Dempster, A. P., Laird, N. M., & Rubin, D. B. academic behaviors. In J. T. Spence (Ed.),
(1977). Maximum likelihood from Achievement and achievement motives (pp.
incomplete data via the EM algorithm. 74–146). San Francisco, CA: W. H.
Journal of the Royal Statistical Society, Series Freeman.
B, 39(1), 1-38. Eccles, J. S., Midgley, C., Wigfield, A., Buchanan,
C. M., Reuman, D., Flanagan, C., & Mac, I.

68
 
 D. (1993). Development during adolescence: Friso-van den Bos, I., van der Ven, S. H. G.,
The impact of stage-environment fit on Kroesbergen, E. H., & van Luit, J. E. (2013).
young adolescents’ experiences in school Working memory and mathematics in
and in families. American Psychologist, 48, primary school children: A meta-analysis.
90–101. Educational Research Review, 10, 29-44.
Eccles, J. S., Roeser, R. W. (2009). Schools, Fuchs, L. S., Fuchs, D., Compton, D. L., Bryant,
academic motivation, and stage- J. D., Hamlett, C. L., & Seethaler, P. M.
environment fit. In R. M. Lerner, & L. (2007). Mathematics screening and progress
Steinberg (Eds.), Individual bases of monitoring at first grade: Implications for
adolescent development (3rd ed.). Handbook responsiveness to intervention. Exceptional
of adolescent psychology, Vol. 1. (pp. 404- Children, 73, 311-330.
434). Hoboken, NJ: Wiley. Fuchs, L.S., & Fuchs, D. (2002). Mathematical
Else-Quest, N. M., Hyde, J. S., & Linn, M. C. problem-solving profiles of students with
(2010). Cross-national patterns of gender mathematical disabilities with and without
differences in mathematics: A meta- comorbid reading disabilities. Journal of
analysis. Psychological Bulletin, 136(1), 103– Learning Disabilities, 35, 563 – 573.
127. Fuchs, L.S., Powell, S.R., Seethaler, P.M., Cirino,
Enders, C. K. (2010). Applied missing data P.T., Fletcher, J.M., Fuchs, D., … & Zumeta,
analysis. London: The Guilford Press. R.O. (2009). Remediating number
Fall, A-M., & Roberts, G. (2012). High school combination and word problem deficits
dropouts: Interactions between social among students with mathematics
context, self-perceptions, school difficulties: A randomized control trial.
engagement, and student dropout. Journal Journal of Educational Psychology, 101(3),
of Adolescence, 35, 787-798. 561 – 576.

Flake, J. K., Barron, K. E., Hulleman, C., Gaspard, H., Dicke, A-L., Flunger, B., Schreier,
McCoach, B. D., & Welsh, M. E. (2015). B., Häfner, I., Trautwein, U., & Nagengast,
Measuring cost: The forgotten component B. (2014). More value through greater
of expectancy-value theory. Contemporary differentiation: Gender differences in value
Educational Psychology 41, 232–244. beliefs about math. Journal of Educational
Psychology. doi:http://psycnet.apa.org/doi/
Fletcher, J. M., Denton, C., & Francis, D. J.
10.1037/edu0000003
(2005). Validity of alternative approaches
for the identification of learning disabilities: Geary, D. (1994). Children’s mathematical
Operationalizing unexpected development. Washington, DC: American
underachievement. Journal of Learning Psychological Association.
Disabilities, 38, 545–552. Geary, D. C. (1993). Mathematical Disabilities:
Fortin, L., Marcotte, D., Potvin, P., Royer, É., & Cognitive, Neuropsychological, and Genetic
Joly, J. (2006). Typology of students at risk Components. Psychological Bulletin, 114(2),
of dropping out of school: Description by 345-362.
personal, family and school factors. Geary, D., & Hoard, M. (2001). Numerical and
European Journal of Psychology of arithmetical weaknesses in learning-
Education, 21(4), 363-383. disabled children: Relation to dyscalculia
Frenzel, A. C., Goetz, T., Pekrun, R., & Watt, H. and dyslexia. Aphasiology, 15(7), 635-647.
M. G. (2010). Development of mathematics Georgiou, G. K., Torppa, M., Manolitsis, G.,
interest in adolescence: Influences of gender, Lyytinen, H., & Parrila, R. (2012).
family, and school context. Journal of Longitudinal predictors of reading and
Research on Adolescence, 20, 507–537. spelling across languages varying in

69
 
 orthographic consistency. Reading and Harter, S. (1990). Processes underlying
Writing, 25(2), 321-346. adolescent self-concept formation. In R.
Goetz, T., Cronjaeger, H., Frenzel, A. C., Ludtke, Montemayor, G. Adams, & T. Gullotta
O., & Hall, N. C. (2010). Academic self- (Eds.), From childhood to adolescence: A
concept and emotion relations: Domain transitional period? (pp. 205-239).
specificity and age effects. Contemporary Thousand Oaks, CA: Sage.
Educational Psychology, 35(1), 44-58. Hattie, J. A. (1992). Self-concept. Hillsdale, NJ:
Guo, J., Marsh, H., Morin, A. J. S., Parker, P. D., Lawrence Erlbaum Associates, Inc.
& Kaur, G. (2015). Directionality of the Holopainen, L., Lappalainen, K., Junttila, N., &
associations of high school expectancy- Savolainen, H. (2012). The role of social
value, aspirations, and attainment: A competence in the psychological well-being
longitudinal study. American Educational of adolescents in secondary education.
Research Journal, 52(2), 371–402. Scandinavian Journal of Educational
Guo, J., Marsh, H., Parker, P. D., Morin, A. J. S., Research, 56(2), 199-212.
& Yeung, A. S. (2015). Expectancy-value in Inoue, Y. (1999). The educational and
mathematics, gender and socioeconomic occupational attainment process. New York,
background as predictors of achievement NY: University Press of America.
and aspirations: A multi-cohort study. Ireson, J., & Hallam, S. (2009). Academic self-
Learning and Individual Differences, 37, concepts in adolescence: Relations with
161–168. achievement and ability grouping in schools.
Gurney, P. (1986). Self-esteem in the classroom: Learning and Instruction, 19, 201–213.
Theoretical perspectives and assessment Jacobs, J. E., Lanza, S., Osgood, D. W., Eccles, J.
issues. School Psychology International, 7, S., & Wigfield, A. (2002). Changes in
199-209. children’s self-competence and values:
Guthrie, J. T., Hoa, A. L. W., Wigfield, A., Tonks, Gender and domain differences across
S. M., Humenick, N. M., & Littles, E. (2007). grades one through twelve. Child
Reading motivation and reading Development, 73, 509–527.
comprehension growth in the later Jahnukainen, M. (2001). Two models for
elementary years. Contemporary preventing students with special needs from
Educational Psychology, 32(3), 282-313. dropping out of education in Finland.
Hakkarainen, A., Holopainen, L., & Savolainen, European Journal of Special Needs
H. (2012). Mathematical and reading Education, 16(3), 245-258.
difficulties as predictors of school Janosz, M., LeBlanc, M., Boulerice, B., &
achievement and transition to secondary Tremblay, R.E. (1997). Disentangling the
education. Scandinavian Journal of Weight of School dropout predictors: A test
Educational Research, on two longitudinal samples. Journal of
doi:10.1080/00313831.2012.696207 Youth and Adolescence, 26(6), 733-762.
Hanich, L. B., Jordan, N. C., Kaplan, D., & Dick, Janosz, M., LeBlanc, M., Boulerice, B., &
J. (2001). Performance across different areas Tremblay, R.E. (2000). Predicting different
of mathematical cognition in children with types of school dropouts: A typological
learning difficulties. Journal of Educational approach with two longitudinal samples.
Psychology, 93, 615-626. Journal of Educational Psychology, 92(1),
Harter, S. (1982). The perceived competence 171-190.
scale for children. Child Development, 53, Järvinen, T., & Vanttaja, M. (2006).
87-97. Koulupudokkaiden työurat [The Careers of

70
 
 Students Dropping Out of School]. reading: From prediction to instruction.
Yhteiskuntapolitiikka, 71(1), 14–22. Reading Research Quarterly, 45(3), 341-362.
Järvinen, T., Vanttaja, M. (2001). Young people, Kiuru, N., Aunola, K., Nurmi, J-E., Leskinen, E.,
education and work: Trends and changes in & Savolainen, H. (2008). Peer group
Finland in the 1990s. Journal of Youth influence and selection in adolescents'
Studies, 4(2), 195-207. school burnout: A longitudinal study.
Jitendra, A. K., & Hoff, K. (1996). The effects of Merrill-Palmer Quarterly, 54, 23-55.
schema-based instruction on the word- Kyttälä, M. (2008). Visuospatial working
problem-solving performance of students memeory in adolescents with poor
with learning disabilities. Journal of performance in mathematics: variation
Learning Disabilities, 29, 421–431. depending on reading skills. Educational
Jitendra, A. K., DiPipi, C. M., & Perron-Jones, N. Psychology, 28, 273-289.
(2002). An exploratory study of schema- Lamb, S. (2011). Pathways to school completion:
based word-problem-solving instruction for An international comparison. In S. Lamb, E.
middle school students with learning Markussen, R. Teese., N. Sandberg, & J.
disabilities: An emphasis on conceptual and Polesel (Eds.), School dropout and
procedural understanding. Journal of Special completion: International comparative
Education, 36, 23–38. studies in theory and policy (pp. 21-73).
Johansson, M-G. (2005). LS. Handledning. London: Springer.
Klassdiagnoser i läsning och skrivning för Landerl, K., Fussenegger, B., Moll, K., &
högstadiet och gymnasiet. [LS. Manual. Willburger, E. (2009). Dyslexia and
Diagnostics in reading and writing for dyscalculia: Two learning disorders with
middle school and upper secondary different cognitive profiles. Journal of
students]. M.-G. Johansson & Experimental Child Psychology, 103, 309-
Psykologiförlaget AB. Stockholm, Sweden: 324.
Psykologiförlaget. Landerl, K., Ramus, F., Moll, K., Lyytinen, H.,
Jordan, N. C., Hanich, L. B., & Kaplan, D. (2003). Leppänen, P. H. T., Lohvansuu, K.,
A longitudinal study of mathematical O´Donovan, M… Schulle-Körne, G. (2013).
competencies in children with specific Predictors of developmental dyslexia in
mathematics difficulties versus children European orthographies with varying
with comorbid mathematics and reading complexity. Journal of Child Psychology and
difficulties. Child Development, 74(3), 843– Psychiatry, 54(6), 686-694.
850. LeFevre, J., Fast, L., Skwarchuk, S., Smith-Chant,
Kamins, M., & Dweck, C. S. (1999). Person vs. B. L., Bisanz, J., Kamawar, D., et al. (2010).
process praise and criticism: Implications Pathways to mathematics: Longitudinal
for contingent self-worth and coping. predictors of performance. Child
Developmental Psychology, 35, 835–847. Development, 81, 1753-1767.
Kaufmann, L., Wood, G., Rubinstein, O., & Leung, S. A., Conoley, C. W., & Scheel, M. J.
Henrik, A. (2011). Meta-Analyses of (1994). The career and educational
Developmental fMRI studies investigating aspirations of gifted high school students: A
typical and atypical trajectories of number retrospective study. Journal of Counseling
processing and calculation. Developmental and Development, 72, 298–303.
Neuropsychology, 36(6), 763-787. Levin, H., Belfield, C., Muenning, P., & Rouse,
Kirby, J. R., Georgiou, G. K., Martinussen, R., & C. (2007). The costs and benefits of an
Parrila, R. (2010). Naming speed and excellent education for all of America´s

71
 
 children. New York, NY: Teachers College Marsh, H. W., & Craven, R. G. (2002). The
Press. pivotal role of frames of reference in
Light, G. J., & De Fries, J. C. (1995). Comorbidity academic self-concept: The “big-fish-little-
of reading and mathematical disabilities: pond” effect. In F. Pajares & T. Urdan (Eds.),
Genetic and environmental etiologies. Adolescence and education: Vol. 2. Academic
Journal of Learning Disabilities, 28(2), 96– motivation of adolescents. Greenwich, CT:
106. Information Age Publishing.
Linsen, S., Torbeyns, J., Verschaffel, L., Marsh, H. W., Hau, K. T., & Kong, C-K. (2002).
Reynvoet, B., & De Smedt, B. (2016). The Multilevel causal ordering of academic self-
association between symbolic and concept and achievement: Influence of
nonsymbolic numerical magnitude language of instruction (English compared
processing and mental versus algorithmic with Chinese) for Hong Kong students.
subtraction in adults. Acta Psychologia, 165, American Educational Research Journal,
34-42. 39(3), 727-763.
Lo, Y., Mendell, N. R., & Rubin, D. B. (2001). Marsh, H. W., Hau, K. T., Balla, J. R., & Grayson,
Testing the number of components in a D. (1998). Is more ever too much? The
normal mixture. Biometrika, 88, 767–778. number of indicators per factor in
confirmatory factor analysis. Multivariate
Luttrell, V. R., Callen, B. W., Allen, C. S., Wood,
Behavioral Research, 33, 181–220.
M. D., Deeds, D. G., & Richard, D. C. S.
(2010). The mathematics value inventory for Marsh, H. W., Seaton, M., Trautwein, U.,
general education students: Development Lüdtke, O., Hau, K.T., O’Mara, A.J., &
and initial validation. Educational and Craven, R.G. (2008). The big-fish-little-
Psychological Measurement, 70, 142–160. pond effect stands up to critical scrutiny:
Implications for theory, methodology, and
Lyons, I. M., Price, G. R., Vaessen, A., Blomert,
future research. Educational Psychology
L., Ansari, D. (2014). Numerical predictors
Review, 3, 319-350.
of arithmetic success in grades 1–6.
Developmental Science, 17(5), 714-726. Marsh, H. W., Tracey, D., & Craven, R. (2006).
Multidimensional self-concept structure for
Mahaffy, K., & Ward, S. (2002). The gendering
preadolescents with mild intellectual
of adolescents’ childbearing and educational
disabilities: A hybrid multigroup –MIMC
plans: Reciprocal effects and the influence of
approach to factorial invariance and latent
social context. Sex Roles, 46, 403–417.
mean differences. Educational and
Marsh, H. W. (1986). Verbal and math self- Psychological Measurement, 66(5), 795-818.
concepts: An internal/external frame of
Marsh, H., Hau, K. T., & Wen, Z. (2004). In
reference model. American Educational
search for golden rules: Comment on the
Research Journal, 23, 129–149.
hypothesis testing approaches to setting
Marsh, H. W. (1987). The big-fish-little-pond cutoff values for fit indexes and dangers in
effect on academic self-concept. Journal of overgeneralising Hu & Bentler’s (1999)
Educational Psychology, 79, 280-295. findings. Structural Equation Modeling, 11,
Marsh, H. W., & Craven, R. G. (1997). Academic 320–341.
self-concept: Beyond the dustbowl. In G. D. Marsh, H.W., Abdulkabbar, A. S., Parker, P. D.,
Phye (Ed.), Handbook of classroom Morin, A. J. S., Abdelfattah, F., Nagengast,
assessment: Learning, achievement, and B., Möller, J., & Abu-Hilal, M. M. (2015).
adjustment, Educational psychology series The internal/external frame of reference
(pp. 131-198). San Diego, CA: Academic model of self-concept and achievement
Press. relations: Age-cohort and cross-cultural

72
 
 differences. American Educational Research adolescent boys and girls. Journal of
Journal, 52(1), 168-202. Secondary Gifted Education, 13, 96–107.
Mau, W. C., & Bikos, L. H. (2000). Educational Merkel, R., & Ansari, D. (2016). Why numerical
and vocational aspirations of minority and symbols count in the development of
female students: A longitudinal study. mathematical skills: evidence from brain
Journal of Counseling & Development, 78, and behavior. Current Opinion in Behavioral
186–194. Sciences, 10, 14-20.
Mazzocco, M. M. M. (2007). Defining and Möller, J., Pohlman, B., Köller, O., & Marsh, H.
differentiating mathematical learning W. (2009). A meta-analytic path analysis of
disabilities and difficulties. In D. B. Berch & the internal/external frame of reference
M. M. M. Mazzocco (Eds.), Why is math so model of academic achievement and
hard for some children? The nature and academic self-concept. Review of
origins of mathematical learning difficulties Educational Research, 79(3), 1129-1167.
and disabilities (pp. 49–60). Baltimore, MD: Murayama, K., Pekrun, R., Lichtenfeld, S., &
Brookes. vom Hofe, R. (2012). Predicting long-term
Mazzocco, M. M. M., & Myers, G. F. (2003). growth in students’ mathematics
Complexities in Identifying and Defining achievement: The unique contributions of
Mathematics Learning Disability in the motivation and cognitive strategies. Child
Primary School-Age Years. Annals of Development, 84(4), 1475-1490.
Dyslexia, 53, 218-253. Muthén, L. K., & Muthén, B. O. (1998–2011).
Mazzocco, M. M. M., & Räsänen, P. (2013). Mplus user’s guide (6th ed.). Los Angeles,
Contributions of longitudinal studies to CA: Muthén & Muthén.
evolving defnitions and knowledge of Nagengast, B., & Marsh, H. W. (2011). The
developmental dyscalculia. Trends in negative effect of school-average ability on
Neuroscience and Education, 2, 65-73. science self-concept in the UK, the UK
Mazzocco, M. M. M., Feigenson, L., & Halberda, countries and the world: The big-fish-little-
J. (2011). Impaired acuity of the pond effect for PISA 2006. Educational
approximate number system underlies Psychology, 31(5), 629-656.
mathematical learning disability Nagy, G., Trautwein, U., Baumert, J., Köller, O.,
(dyscalculia). Child Development, 82, 1224– & Garrett, J. (2006). Gender and course
1237. selection in upper secondary education:
McArdle, J. J., & Nesselroade, J. R. (2014). Effects of academic self-concept and
Longitudinal data analysis using structural intrinsic value. Educational Research and
equation models. Washington, DC, USA: Evaluation: An International Journal on
American Psychological Association. Theory and Practice, 12(4), 323–345.
Meece, J. L., Glienke, B. B., & Burg, S. (2006). Nurmi, J.-E., & Salmela-Aro, K. (2002). Goal
Gender and motivation. Journal of School construction, reconstruction and depressive
Psychology, 44, 351–373. symptoms in a life-span context: The
Mejias, S., Grégoire, J., & Noël, M-P. (2012). transition from school to work. Journal of
Numerical estimation in adults with and Personality, 70, 385–420.
without developmental dyscalculia. Nyqvist, F., Finnäs, F., Jakobsson, G, &
Learning and Individual Differences, 22(1), Koskinen, S. (2008). The effect of social
164-170. capital on health: The case of two language
Mendez, L. M. R., & Crawford, K. M. (2002). groups in Finland. Health & Place, 14(2),
Gender-role stereotyping and career 347-360.
aspirations: A comparison of gifted early

73
 
Nyyssölä, K. (2004). Siirtymävaiheiden Parker . P., Nagy, G., Trautwein, U., Lüdtke, O.
koulutusvalinnat—koulutuspoliittinen (2014). Predicting career aspirations and
tarkastelu [Educational choices in university majors from academic ability and
transition—perspective of educational self-concept: a longitudinal application of
policy]. Kasvatus, 35, 222–229. the internal–external frame of reference
O’Mara, A. J., Marsh, H. W., Craven, R. G., & model. In I. Schoon, & J. S. Eccles (Eds.),
Debus, R. L. (2006). Do self-concept Gender Differences in Aspirations and
interventions make a difference? A Attainment: A Life Course Perspective (pp.
synergistic blend of construct validation and 224–246). Cambridge: University Press.
meta-analysis. Educational Psychologist, Parker, P. D., Schoon, I., Tsai, Y.-M., Nagy, G.,
41(3), 181–206. Trautwein, U., & Eccles, J. S. (2012).
OECD (2003). Student engagement at school: A Achievement, agency, gender, and
sense of belonging and participation. Paris: socioeconomic background as predictors of
OECD. postschool choices: A multicontext study.
Developmental Psychology, 48(6), 1629 –
OECD. (2010). PISA 2009 results: What students
1642.
know and can do: Student performance in
reading, mathematics and science. Paris: Perez, T., Cromley, J. G., & Kaplan, A. (2014).
Author. The role of identity development, values and
costs in college STEM retention. Journal of
Official Statistics of Finland (2012a). Population
Educational Psychology, 106(1), 315–329.
structure [e-publication].
ISSN=1797-
Piazza, M. (2010). Neurocognitive start-up tools
5395. Helsinki: Statistics Finland [referred:
for symbolic number representations.
13.10.2012].
Access method: Trends in Cognitive Science, 14(12), 542-551.
http://www.stat.fi/til/vaerak/index_en.html.
Pickering, S. J., & Gathercole, S. E. (2004).
Official Statistics of Finland (2012b). Distinctive working memory profiles in
Discontinuation of education [e- children with special educational needs.
publication].
ISSN=1798-9302. Helsinki: Educational Psychology, 24, 393-408.
Statistics Finland [referred: 13.10.2012]. Pollard, E.L., & Lee, P.D. (2003). Child well-
Access method: http://www.stat.fi/til/ being: A systematic review of the literature.
kkesk/index_en.html. Social Indicator Research, 61(1), 59–78.
Olsson, L., Östergren, R., & Träff, U. (2016). Price, G. R., Holloway, I., Räsänen, P.,
Developmental dyscalculia: A deficit in the Vesterinen, M., & Ansari, D. (2007).
approximate number system or an access Impaired parietal magnitude processing in
deficit? Cognitive Development, 39, 154-167. developmental dyscalculia. Current Biology,
Ostad, S. (1998). Comorbidity between 17, R1042–R1043.
mathematics and spelling difficulties. Ramus, F. (2004). Neurobiology of dyslexia: A
Logopedics Phoniatrics Vocology (Log Phon reinterpretation of the data. Trends in
Vocol), 23(4), 145-154. Neuroscience, 27(12), 720-726.
Otis, N., Grouzet, M. E., & Pelletier, L. C. (2005). Räsänen, P., & Leino, L. (2005). KTLT.
Latent motivational change in an academic Laskutaidon testi. Opas yksilö- tai
setting: A 3-year longitudinal study. Journal ryhmämuotoista arviointia varten [KTLT.
of Educational Psychology, 97, 170–183. Arithmetic skills test. Manual for individual
Owens, J. (2004). A review of the social and non- or group assessment]. Jyväskylä, Finland:
market returns to education. Wales, UK: NMI.
Education and Learning Network. Renninger, K. A. & Su, S. (2012). Interest and its
development. In R. Ryan (Ed.), The Oxford

74
 
 handbook of human motivation (pp. 167– early adolescence: Comparisons across
187). New York: Oxford University Press. gender. Journal of Youth and Adolescence,
Renninger, K. A., & Hidi, S. (2011). Revisiting 39(1), 47–61.
the conceptualization, measurement and Rumberger, R. W., & Lamb, S. P. (2003). The
generation of interest. Educational early employment and further education
Psychologist, 46(3), 168–184. experiences of high school dropouts: A
Retelsdorf, J., Köller, O., & Möller, J. (2014). comparative study of the United States and
Reading achievement and reading self- Australia. Economics of Education Review,
concept – Testing the reciprocal effects 22, 353-366.
model. Learning and Instruction, 29, 21-30. Saarela, J., & Finnäs, F. (2003). Unemployment
Rimpelä, M. (2003). Kouluterveyskysely and native language: The Finnish case. The
[Questionnaire for school health] [referred: Journal of Socio-Economics, 32(1), 59-80.
6.11.2012]. Access method: Salmela-Aro, K., & Näätänen, P. (2005). Method
http://info.stakes.fi/kouluterveyskysely/EN/ of assessing adolescents´ school burnout [in
index.htm Finnish]. Helsinki, Finland: Edita.
Rinne, R., & Järvinen, T. (2011). Dropout and Salmela-Aro, K., & Tynkkynen, L. (2012).
completion in upper secondary education in Gendered pathways in school burnout
Finland. In S. Lamb, E. Markussen, R. among adolescents. Journal of Adolescence,
Teese., N. Sandberg, & J. Polesel (Eds.), 35(4), 929–939.
School dropout and completion: Salmela-Aro, K., & Upadyaya, K. (2014a). The
International comparative studies in theory demands–resources model in school
and policy (pp. 215-232). London: Springer. context. British Journal of Educational
Rojewski, J. W. (2005). Occupational Psychology, 84,137–151.
aspirations: Constructs, meanings, and Salmela-Aro, K., & Upadyaya, K. (2014b).
applications. In S. D. Brown, & R. W. Lent Developmental trajectories of school
(Eds.), Career development and counselling: burnout: Evidence from two longitudinal
Putting theory and research to work (pp. 131- studies. Learning and Individual Differences,
154). Hoboken, NJ: Wiley. 36, 60–68.
Rotzer, S., Loenneker, T., Kucian, K., Martin, E., Salmela-Aro, K., Kiuru, N., & Nurmi, J-E.
Klaver, P., & von Aster, M. Dysfunctional (2008). The role of educational track in
neural network of spatial working memory adolescents’ school burnout: A longitudinal
contributes to developmental dyscalculia. study. British Journal of Educational
Neuropsychologia, 47(13), 2859-2865. Psychology, 78(4), 663–689.
Rourke, B. P., & Conway, J. A. (1997). Salmela-Aro, K., Kiuru, N., Leskinen, E., &
Disabilities of arithmetic and mathematical Nurmi, J-E. (2009). School burnout
reasoning: Perspectives from neurology inventory (SBI): Reliability and validity.
neuropsychology. Journal of Learning European Journal of Psychological
Disabilities, 30(1), 34-46. Assessment, 25, 48–57
Rubinsten, O., & Henik, A. (2009). Salmela-Aro, K., Savolainen, H., & Holopainen,
Developmental dyscalculia: heterogeneity L. (2009). Depressive symptoms and school
might not mean different mechanisms. burnout during adolescence: Evidence from
Trends in Cognitive Science 13, 92–99. two cross-lagged longitudinal studies.
Rueger, S.Y., Malecki, C.K., & Demaray, M.K. Journal of Youth and Adolescence, 38(10),
(2010). Relationship between multiple 1316-1327.
sources of peceived social support and
psychological and academic adjustment in

75
 
Savolainen, H. (2001). Explaining mechanisms of dyslexia. Biological Psychiatry, 52(2), 101-
educational career choice. A follow-up study 110.
of the educational choices of a group of youths Siegel, L. S., & Ryan, E. B. (1989). The
that finished compulsory education in 1990. development of working memory in
University of Joensuu. Publications in normally achieving and subtypes of learning
education, 69. disabled children. Child Development, 60,
Savolainen, H., Ahonen, T., Aro, M., Tolvanen, 973-980.
A., & Holopainen, L. (2008). Reading Simmons, F. R., & Singleton, C. (2008). Do weak
comprehension, word reading and spelling phonological representations impact on
as predictors of school achievement and arithmetic development? A review of
choice of secondary education. Learning and research into arithmetic and dyslexia.
Instruction, 18, 201–210. Dyslexia, 14, 77−94.
Schaufeli, W., & Bakker, A. (2004). Job Simola, H., Rinne, R., & Kivirauma, J. (1999).
demands, job resources and their Finland. National changes in education and
relationship with burnout and engagement; education governance, In S. Lindblad & T.
a multi-sample study. Journal of Popkewitz (Eds.), Education governance and
Organizational Behavior, 25, 293–299. social integration and exclusion: National
Schneider, M., Beeres, K., Coban, L., Merz, S., cases of educational systems and recent
Schmidt, S., Stricker, J., & De Smedt, B. reforms (pp. 42-64). Uppsala: Uppsala
(2016). Associations of non-symbolic and Univeristy.
symbolic numerical magnitude processing Sirin, S. R. (2005). Socioeconomic status and
with mathematical competence: a meta- academic achievement: A meta-analytic
analysis. Developmental Science. review of research. Review of Educational
doi: 10.1111/desc.12372 Research, 75(3), 417-453.
Shafrir, U. & Siegel, L. S. (1994). Subtypes of Skagerlund, K., & Träff, U. (2014). Number
Learning Disabilities in Andolescents and processing and heterogeneity of
Adults. Journal of Learning Disabilities, developmental dyscalculia: Subtypes with
27(2), 123-134 different cognitive profiles and deficits.
Shalev, R. S. (2007). Prevalence of Journal of Learning Disabilities. doi:
developmental dyscalculia. In D. B. Berch & 10.1177/0022219414522707
M. M. M. Mazzocco (Eds.), Why is math so Snowling, M. J. (2005). The science of dyslexia:
hard for some children? The nature and A review of contemporary approaches. In
origins of mathematical learning difficulties M. Turner & J. Rack (Eds.), The Study of
and disabilities (pp. 49–60). Baltimore, MD: Dyslexia (pp. 77-90). New York: Springer.
Brookes.
Steyer, R., Partchev, I., & Shanahan, M. J. (2000).
Shapka, J. D., Domene, J. F., & Keating, D. P. Modeling true intraindividual change in
(2006). Trajectories of career aspirations structural equation models: The case of
through adolescence and young adulthood: poverty and children's psychosocial
Early math achievement as a critical filter. adjustment. In T. D. Little, K. U. Schnabel,
Educational Research and Evaluation, 12(4), & J. Baumert (Eds.), Modeling longitudinal
347–358. and multilevel data: Practical issues, applied
Shaywitz, B. A., Shaywitz, S. E., Pugh, K. R., approaches, and specific examples (pp. 109-
Menci, W. E., Fulbright, R. K., Skudlarski, 126). Mahwah, NJ: Erlbaum.
P., Constable, R. T… Gore, J. C. (2002). Stock, P., Desoete, A., & Roeyers, H. (2010).
Disruption of posterior brain systems for Detecting children with arithmetic
reading in children with developmental disabilities from kindergarten: Evidence

76
 
 from a 3-year longitudinal study on the role orientations and academic well-being across
of preparatory arithmetic abilities. Journal of the transition to upper secondary education.
Learning Disabilities, 43(3), 250-268. Learning and Individual Differences, 22, 290-
Sum, A., Khatiwada, I., McLaughlin, J., & Palma, 305.
S. (2009). The consequences of dropping out Valentine, J., DuBois, D., & Cooper, H. (2004).
of high school. Boston, MA: Center for Labor The relation between self-beliefs and
Market Studies. academic achievement: A meta-analytic
Swanson, H. L., & Beebe-Frankenberger, M. review. Educational Psychologist, 39(2), 111-
(2004). The Relationship Between Working 133.
Memory and Mathematical Problem Van Petegem, K., Aelterman, A., Rossel, Y., &
Solving in Children at Risk and Not at Risk Creemers, B. (2007). Students’ perceptions
for Serious Math Difficulties. Journal of as moderator for students’ wellbeing. Social
Educational Psychology, 96(3), 471-491. Indicators Research, 83(3), 447–463.
Swanson, H. L., Jerman, O., & Zheng, X. (2009). Vanbinst, K., Ansari, D., Ghesquiere, P., De
Math disabilities and reading disabilities: Smedt, B. (2016). Symbolic numerical
Can they be separated? Journal of magnitude processing is as important to
Psychoeducational Assessment, 27(3), 175- arithmetic as phonological awareness is to
196. reading. PLoS ONE, 11(3), e0151045.
Toll, S. W. M., Kroesbergen, E. H., & Van Luit, Vanttaja, M., & Järvinen, T. (2004).
J. E. H. (2016). Visual working memory and Koulutuksesta ja työstä karsiutuneet: Vailla
number sense: Testing the double deficit ammattitutkintoa ja työpaikkaa vuonna
hypothesis in mathematics. British Journal 1985 olleiden nuorten myöhemmät
of Educational Psychology. doi: elämänvaiheet [Excluded from Education
10.1111/bjep.12116 and Work: Young People with Neither a
Träff, U. (2013). The contribution of general Degree Nor a Job in 1985 and Their Later
cognitive abilities and number abilities to Stages of Life]. Yhteiskuntapolitiikka, 69(5),
different aspects of mathematics in children. 472–480.
Journal of Experimental Child Psychology, Vasalampi, K., Salmela-Aro, K., & Nurmi, J-E.
116(2), 139-156. (2009). Adolescents’ self-concordance,
Trebbels, M. (2015). The transition at the end of school engagement, and burnout predict
compulsory full-time education: Educational their educational trajectories. European
and future career aspirations of native and Psychologist, 14(4), 332–341.
migrant students. Wiesbaden: Springer. Vasalampi, K., Salmela-Aro, K., & Nurmi, J-E.
Tuominen-Soini, H., & Salmela-Aro, K. (2014). (2010). Education-related goal appraisals
Schoolwork engagement and burnout and self-esteem during the transition to
among Finnish high school students and secondary education: A longitudinal study.
young adults: Profiles, progressions, and International Journal of Behavioral
educational outcomes. Developmental Development, 34(6), 481-490.
Psychology, 50(3), 649–662. Vellutino, F. R., & Fletcher, J. M. (2005).
Tuominen-Soini, H., Salmela-Aro, K., & Developmental dyslexia. In M. J. Snowling &
Niemivirta, M. (2008). Achievement goal C. Hulme (Eds.), The science of reading: A
orientations and subjective well-being: A handbook (pp. 362-378). Malden: Blackwell
person-centred analysis. Learning and Publishing.
Instruction, 18, 251-266. Vellutino, F. R., Fletcher, J. M., Snowling, M. J.,
Tuominen-Soini, H., Salmela-Aro, K., & & Scanlon, D. M. (2004). Specific reading
Niemivirta, M. (2012). Achievement goal disability (dyslexia): what have we learned in

77
 
 the past four decades? The Journal of Child Gendered motivational processes affecting
Psychology and Psychiatry, 45(1), 2-40. high school mathematics participation,
Vermunt, J. K., & Magidson, J. (2002). Latent educational aspirations, and career plans: A
class cluster analysis. In J. A. Hagenaars, & comparison of samples from Australia,
A. L. McCutcheon (Eds.), Applied latent Canada, and the United States.
class analysis (pp. 89–106). Cambridge, UK: Developmental Psychology, 48(6), 1594–
Cambridge University Press. 1611.

Viljaranta, J., Nurmi, J.-E., Aunola, K., & Wigfield, A., & Eccles, J. S. (1992). The
Salmela-Aro, K. (2009). The role of task development of achievement task values: A
values in adolescents’ educational tracks: A theoretical analysis. Developmental Review,
person-oriented approach. Journal of 12, 1–46.
Research on Adolescence, 19, 786–798. Wigfield, A., & Eccles, J. S. (2000). Expectancy–
von Eye, A., Spiel, C., & Wood, P. K. (1996). value theory of achievement motivation.
Configural frequency analysis in Contemporary Educational Psychology,
psychological research. Applied Psychology, 25(1), 68–81.
45(4), 301-352. Wilson, V. L. & Rupley W. H. (2013) Design for
Vukovic, R. K., & Lesaux, N. K. (2013a). The learning disabilities experimental and quasi-
relationship between linguistic skills and experimental research. In H. L. Swanson, K.
arithmetic knowledge. Learning and R. Harris, & S. Graham (Eds.), Handbook of
Individual Differences, 23, 87-91. learning disabilities 2nd ed. (pp. 593–606).
New York, NY: The Guilford Press.
Vukovic, R. K., & Lesaux, N. K. (2013b). The
language of mathematics: Investigating the Wise, J. C., Kyeong Pae, H., Wolfe, C. B., Sevcik,
ways language counts for children´s R. A., Morris, R. D., Lovett, M., & Wolf, M.
mathematical development. Journal of (2008). Phonological Awareness and Rapid
Experimental Child Psychology, 115, 227- Naming Skills of Children with Reading
244. Disabilities and Children with Reading
Disabilities Who Are At Risk for
Vukovic, R. K., Lesaux, N. K., & Siegel, L. S.
Mathematics Difficulties. Learning
(2010). The mathematics skills of children
Disabilities Research & Practice, 23(3), 125-
with reading difficulties. Learning and
136.
Individual Differences, 20, 639-643.
Zeleke, S. (2004). Self-concepts of students with
Watt, H. M. (2005). Exploring adolescent
learning disabilities and their normally
motivations for pursuing maths-related
achieving peers: A review. European
careers. Australian Journal of Educational
Journal of Special Needs Education, 19(2),
and Developmental Psychology, 5, 107–116.
145–170.
Watt, H. M., Shapka, J. D., Morris, Z. A., Durik,
A. M., Keating, D. P., & Eccles, J. S. (2012).

78
 
 Johan Korhonen | Learning difficulties, academic well-being and educational pathways among adolescent students | 2016
Johan Korhonen

Learning difficulties, academic well-being

and educational pathways among
adolescent students

9 789521 234637

ISBN 978-952-12-3463-7

View publication stats