ArticlePDF Available

Psychometric assessment of individual differences in second language reading anxiety for identifying struggling students in classrooms


Abstract and Figures

Assessing learners’ individual differences helps identify students who need teacher support in classrooms. Previous studies have examined second language (L2) achievement based on reading anxiety because reading is an input-based activity essential for successful L2 learning. This study applied a latent rank model to identify L2 learners who are likely to be struggling or successful in classrooms according to their L2 reading anxiety symptoms. Moreover, a psychometric function was developed to determine the cutoff anxiety scores that discriminate against their substantial differences. The model was applied to responses from the Foreign Language Reading Anxiety Scale (FLRAS) provided by 335 Japanese learners of English. The results showed that the FLRAS classified students into three ranked groups with ordinal information regarding L2 reading anxiety. Rank 1 exhibited good conditions in L2 reading anxiety. Rank 2 reported high anxiety toward unfamiliar grammar during L2 reading. Rank 3 had even higher anxiety levels, especially for vocabulary and grammatical knowledge deficits and reading difficulty. The cutoff anxiety scores estimated by the model detected students who failed their L2 class with 79% accuracy. Theoretical, methodological, and pedagogical issues in language anxiety were discussed in terms of diagnosis and different approaches to teaching L2 reading.
Content may be subject to copyright.
Frontiers in Psychology 01
Psychometric assessment of
individual dierences in second
language reading anxiety for
identifying struggling students in
* and ShuichiTakaki
1 Department of English Studies, Kobe City University of Foreign Studies, Kobe, Japan, 2 Faculty of
Human Development and Culture, Fukushima University, Fukushima, Japan
Assessing learners’ individual dierences helps identify students who need
teacher support in classrooms. Previous studies have examined second
language (L2) achievement based on reading anxiety because reading is an
input-based activity essential for successful L2 learning. This study applied a
latent rank model to identify L2 learners who are likely to be struggling or
successful in classrooms according to their L2 reading anxiety symptoms.
Moreover, a psychometric function was developed to determine the cuto
anxiety scores that discriminate against their substantial dierences. The
model was applied to responses from the Foreign Language Reading
Anxiety Scale (FLRAS) provided by 335 Japanese learners of English. The
results showed that the FLRAS classified students into three ranked groups
with ordinal information regarding L2 reading anxiety. Rank 1 exhibited
good conditions in L2 reading anxiety. Rank 2 reported high anxiety toward
unfamiliar grammar during L2 reading. Rank 3 had even higher anxiety levels,
especially for vocabulary and grammatical knowledge deficits and reading
diculty. The cuto anxiety scores estimated by the model detected students
who failed their L2 class with 79% accuracy. Theoretical, methodological, and
pedagogical issues in language anxiety were discussed in terms of diagnosis
and dierent approaches to teaching L2 reading.
L2 reading, L2 achievement, individual dierences, anxiety, pedagogical screening,
a latent rank model
Second language (L2) anxiety is operationalized as a predictor of the L2 achievement
(Teimouri etal., 2019; Zhang, 2019). For example, reading is an input-based activity essential
for successful L2 learning but high anxiety toward reading impedes input and intake
processing (Horwitz, 2001). L2 reading anxiety is considered inuential in the Japanese
learners’ achievement in English classrooms (Matsuda and Gobel, 2004) because a task type
TYPE Original Research
PUBLISHED 18 August 2022
DOI 10.3389/fpsyg.2022.938719
Kaiqi Shao,
Hangzhou Dianzi University,
Abdullah Alamer,
Imam Mohammad Ibn Saud Islamic
University (IMSIU), SaudiArabia
Jianling Zhan,
Guangdong University of Foreign Studies,
Akira Hamada
This article was submitted to
Language Sciences,
a section of the journal
Frontiers in Psychology
RECEIVED 08 May 2022
ACCEPTED 12 July 2022
PUBLISHED August 202218
Hamada A and Takaki S (2022)
Psychometric assessment of individual
dierences in second language reading
anxiety for identifying struggling students in
Front. Psychol. 13:938719.
doi: 10.3389/fpsyg.2022.938719
© 2022 Hamada and Takaki. This is an
open-access article distributed under the
terms of the Creative Commons Attribution
License (CC BY). The use, distribution or
reproduction in other forums is permitted,
provided the original author(s) and the
copyright owner(s) are credited and that
the original publication in this journal is
cited, in accordance with accepted
academic practice. No use, distribution or
reproduction is permitted which does not
comply with these terms.
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 02
required for them is mediating a text (e.g., translating and
summarizing English documents in Japanese). Mediation activities
are in high need in monolingual classrooms and workplaces
(Lambert, 2010). Considering that the individual dierences in L2
reading anxiety are associated with learning behaviors in a
classroom and subsequent L2 achievement (e.g., Sellers, 2000;
Alderson et al., 2016; Hamada and Takaki, 2021a, 2021b), it is
important to diagnose strengths and weaknesses, identify specic
diculties, and place students into dierent learning environments.
Ganschow and Sparks (2001) highlighted the importance of
pedagogical screening, namely, identifying individuals who are
likely to bestruggling in L2 classrooms in order to place them in
an appropriate learning environment. For example, the Foreign
Language Reading Anxiety Scale (FLRAS) developed by Saito
etal. (1999) can examine individual dierences in anxiety toward
L2 reading and identify specic factors evoking L2 reading anxiety
(Zhao etal., 2013). Students may further beclassied into several
groups by predetermined cuto points (e.g., low, average, and high
anxiety groups). While this sort of categorization is practical to
determine what groups need a special intervention, some studies
showed insignicant associations between L2 achievement and
the groups divided by anxiety scores (Phillips, 1992; Marcos-
Llinás and Garau, 2009; Wu, 2011). is suggests that the arbitrary
cuto points will cause the misclassication of students.
is study applied a latent rank model to categorize students
into ranked groups according to L2 reading anxiety symptoms.
e latent rank model is a statistical method that categorizes
students into ranked groups (Shojima, 2007). e ranked groups
will provide information about what kind of L2 reading anxiety
characteristics they have and whether they are struggling learners
in L2 classrooms or not. Here, the traditional methods of group
categorization are reviewed in terms of L2 anxiety scores and
predictive relations to L2 achievement. We then explain the
framework and advantages of applying the latent rank model in
pedagogical screening. Based on the results of this study, the
applicability of the latent rank model and theoretical and
pedagogical implications are discussed.
Literature review
L2 reading anxiety and achievement
e denition of L2 anxiety is “the worry and negative
emotional reaction aroused when learning or using a second
language” (MacIntyre, 1999, p.24). L2 anxiety has been examined
using Foreign Language Classroom Anxiety Scale (FLCAS) of
Horwitz et al. (1986) based on the idea that anxiety involves a trait,
state, and situation-specic construct (MacIntyre and Gardner,
1991; see also Dörnyei and Ryan, 2015). More recently, language-
skill-specic anxieties have been examined in terms of their
separability: listening, reading, speaking, and writing (Cheng etal.,
1999; Saito etal., 1999; Elkhafai, 2005; Pae, 2013; Cheng, 2017). In
L2 reading, Saito etal. (1999) argued that L2 reading anxiety occurs
consistently when performing L2 reading. ey developed the
FLRAS to reect the gradation of L2 reading anxiety as a continuous
variable and showed that it can beseparated from the general L2
anxiety measured by the FLCAS. Each statement of the FLRAS
involves two descriptions about a specic situation in L2 reading
(e.g., “Whenever Iencounter unfamiliar grammar when reading a
foreign language”) and a subsequent symptom (e.g., “I get upset”).
is psychometric instrument has been adopted to describe
individual dierences in L2 reading anxiety and investigate the
reciprocal relationships between L2 reading anxiety and
achievement (e.g., Zhao etal., 2013; Jee, 2016; Sparks etal., 2018a,b;
Hamada and Takaki, 2021a) similar to other studies that used the
FLCAS (e.g., Horwitz etal., 1986; Phillips, 1992; Ganschow and
Sparks, 1996; Hewitt and Stephenson, 2012; Shao etal., 2013).
Comprehensive narrative reviews (MacIntyre and Gardner,
1991; Horwitz, 2001; MacIntyre, 2017) and systematic research
syntheses (Teimouri et al., 2019; Zhang, 2019) support the
negative relationships between L2 anxiety and achievement
including the domain of L2 reading. According to MacIntyre
(2017) and MacIntyre and Gardner (1991), the advent of situation-
specic approaches to L2 anxiety made a signicant contribution
to investigating its negative impact on L2 achievement. ey
indicated initial studies on L2 anxiety produced conicting
ndings due to a lack of theoretical (i.e., distinction of state-, trait-,
and situation-specic constructs of anxiety) and methodological
(i.e., decits in measurement tools for each anxiety type)
sophistications. Horwitz (2001) concluded the negative
relationships between L2 anxiety and L2 achievement. Recently,
the precise association between L2 reading anxiety and
achievement was calculated by two meta-analyses; Teimouri etal.
(2019) and Zhang (2019) showed small-to-medium negative
correlations of 0.38 (k = 8, 95% CI [0.47, 0.29]) and of 0.23
(k = 7, 95% CI [0.34, 0.11]), respectively.
Although the FLRAS has been validated with respect to the
negative relations between L2 reading anxiety and outcome
measures, causal inferences based solely on such negative
associations have also been criticized. Sparks and his colleagues
claimed that the FLRAS merely reects learners’ self-assessments
of their language learning skills when considering several
confounding variables aecting both L2 reading anxiety and L2
achievement. For example, FLRAS scores were found to
benegatively correlated with rst language literacy and literacy-
related measures prior to beginning L2 learning (Sparks etal.,
2018a). Sparks etal. (2018b) further suggested a mediation model
of L2 reading anxiety to raise awareness of spurious correlations
with outcome measures. In fact, a mediation analysis by Hamada
and Takaki (2021b) indicated that the proportion of variance
explained by L2 reading anxiety for achievement signicantly
decreased when L2 reading prociency played a mediating role.
Several longitudinal studies also demonstrated that the earlier L2
achievement predicted the later development of anxiety (Alamer
and Lee, 2021; Sparks and Alamer, 2022).
Despite the limitations to the ndings of the negative
correlation, L2 reading anxiety has been used to examine L2
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 03
achievement (e.g., Wu, 2011; Zhao etal., 2013; Xiao and Wong,
2014; Jee, 2016). However, the continuous scores of the FLRAS are
not always informative when identifying students who will
bestruggling in L2 classrooms due to a lack of information about
cuto points. In such pedagogical screening, a psychometric
function has to beapplied to the psychometrics to determine the
cuto points that can discriminate the substantial dierences of
learners’ individual dierences (Hasselblad and Hedges, 1995;
Finch and French, 2018). is idea is incorporated into testing
research as the diagnostic classication models related to the item
response theory and diagnostic assessments (Liu and Jiang, 2018,
2020; Ravand and Baghaei, 2020). A review of Ravand and
Baghaei (2020) suggested that the diagnostic classication models
can compute a psychometric function to classify respondents
according to multiple categorical attributes with mastery and
non-mastery statuses. Liu and Jiang (2018, 2020) and Shojima
(2007, 2008) further developed a graded classication method to
discriminate respondents’ latent trait levels.
Establishing cuto points and psychometric functions could
also solve the standard error of measurement with psychometrics
problem. Psychological instruments cannot assess the underlying
construct without any measurement errors. erefore, great care
should betaken when identifying individual dierences in L2
reading anxiety among learners using one-point increments.1
Instead, it is pedagogically signicant to classify learners into
several groups that have substantially dierent levels of L2 reading
anxiety. Converting a continuous variable into categorical groups
can inform us if dierent groups show dierent L2 reading anxiety
symptoms. Such classications could determine teaching
approaches appropriate for particular groups in a classroom (e.g.,
Ganschow and Sparks, 1991, 2001; Oxford and Ehrman, 1992;
Swanson, 2017; Finch and French, 2018; Crowther etal., 2021).
Establishing cuto points and the latent
rank model
As the Standards for Educational and Psychological Testing
(American Educational Research Association, 2014) stated, cuto
points must beset on the basis of a clearly dened rationale,
including any description of how they are determined. When
cuto points do not function as intended, some students might
be misclassied into a group that does not represent their
symptoms toward L2 reading anxiety. According to Hasselblad
and Hedges (1995), determining cuto points from continuous
scales is known as a discriminant problem, in which cuto points
can be established if the distance between two groups is the
1 The standard error of measurement estimates how repeated measures
of individuals on the same instrument tend to bedistributed around their
true score. The formula is SD*sqrt(1  Cronbach’s α). Since Cronbach’s α
of the FLRAS is generally high (M = 0.87), when the SD of the FLRAS score
is 10, the standard error of measurement will be3.61 (Teimouri etal., 2019).
largest. is distance is represented by standardized mean
dierences (i.e., eect sizes) like Cohens d and Hedges g. eir
meta-analysis also suggested the importance of reporting the exact
accuracy of screening tests to reduce misclassication.
However, previous studies have never applied these
screening test features to classify students into categorical
groups. In case of the FLCAS (Horwitz et al., 1986), Ganschow
and Sparks (1996), and Marcos-Llinás and Garau (2009)
adopted the method of overall means and standard deviations
(SDs) in classications. Students who scored one or more SDs
above the overall means were identied as a high-anxiety
group, those between ±1 SDs from the mean were identied as
an average-anxiety group, and those with one or more SDs
below the mean were identied as a low-anxiety group. A
similar way to convert anxiety scores is using 25, 50, and 75%
quantiles (Phillips, 1992; Hewitt and Stephenson, 2012).
Another method used by Shao etal. (2013) determined the
denite thresholds like “[s]cores above 132 signify high
anxiety; scores between 99 and 132 denote a middle level of
anxiety, and scores below 99 imply little or no anxiety”
(p. 920).2 As Ravand and Baghaei (2020) suggested, their
generalizability to other populations cannot be ensured
because responses to each questionnaire item depend on both
item and respondent traits. Nevertheless, the same classication
approach has been adopted in L2 reading anxiety research.
Among previous studies included in the meta-analysis by
Teimouri etal. (2019), overall means and SDs (Wu, 2011),
quantiles (Sellers, 2000), and denite cuto points (Zhao etal.,
2013; Xiao and Wong, 2014; Jee, 2016) were employed.
Although L2 anxiety research postulated that students with
higher anxiety are more likely to have lower L2 achievement (e.g.,
Horwitz, 2001), sometimes null or contradicted results were
obtained when using the cuto points set by each study. For
example, Sellers (2000) and Wu (2011) showed insignicant
dierences in L2 reading achievement between low, average, and
high anxiety groups. e denite cuto points were only used to
interpret the qualitative dierences among student groups (Zhao
etal., 2013; Xiao and Wong, 2014; Jee, 2016). By integrating the
interview data with the FLRAS scores, Zhao etal. (2013) noted
that the items whose average scores were above 3.00 should
represent signicant sources of L2 reading anxiety. However, these
previous studies did not validate whether the cuto points
function as intended by examining the relationships to L2
achievement. ese methodological decits must beresolved to
advance theoretical and practical discussions on the relationships
between L2 reading anxiety and achievement.
Regarding statistical classication methods, cluster analysis
has frequently been used in L2 research on individual dierences
2 Despite a lack of any specific explanation, these cuto points seem to
bedetermined based on the Likert-scale; for example, the score of 99
indicates that learners are likely to answer “(3) neither agree nor disagree
to 33 items.
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 04
(Crowther etal., 2021). is technique can identify a number of
groups that are dierent from each other in terms of whether
those within a group have similar target characteristics. However,
since comparisons across clusters are based on descriptive (e.g.,
means) and inferential (e.g., analysis of variance) statistics, the
cuto points that dierentiate each group will bedicult to
reproduce (Pastor etal., 2007). erefore, recent studies have
employed a latent trait approach, such as latent class/prole
analysis, to label learners’ individual dierences (e.g., Swanson,
2017). In the present study, extended model of the latent prole
analysis—the latent rank model—is applied to the FLRAS for
screening practicality. Similar to the diagnostic classication
models (Liu and Jiang, 2018, 2020), the latent rank model can
estimate the number of latent ranks of psychometrics (see
Shojima, 2007, 2008, for mathematical details). Similar to latent
class/prole analysis, the latent rank model allows for applying
the FLRAS’ possible cuto points to dierent populations
because it incorporates the item response theory to estimate the
latent trait of ranked groups. More importantly, latent rank
analysis diers from the other methods in that it can identify
groups with ordinal information without having to perform post
hoc comparisons (Shojima, 2009).
In this study, we investigated the number of latent ranks
included in the FLRAS that may underlie the diagnostic
classication of struggling learners in L2 classrooms. Previous
studies using conventional classication methods provide limited
perspectives on the characteristics of learners’ individual
dierences in L2 reading anxiety. e present study attempts to
qualitatively categorize the diagnostic information regarding L2
reading anxiety. To that end, the study sought to answer the three
research questions below.
1. Are there any cuto points in the FLRAS for the
pedagogical screening of L2 reading anxiety?
2. What kind of L2 reading anxiety characteristics can
bediagnosed for each rank estimated by the FLRAS?
3. Can the latent ranks of the FLRAS identify struggling
learners in L2 classrooms?
Materials and methods
Participants for the FLRAS latent rank model examination
included 335 Japanese learners of English as a foreign language
(EFL) from eight classrooms of three universities located in
urban, suburban, and rural areas (female = 134, male = 201). eir
ages ranged from 18 to 22 years (average = 18.98), and they were
taught English as a compulsory school subject from grades 7 to
12. ey majored in diverse academic elds, such as the
humanities, art, law, social sciences, English, education,
engineering, mathematics, chemistry, and business. All
participants enrolled in 2–4 English courses for general purposes
as required for graduation. Response data from this sample were
used to construct a latent rank model that determines the FLRAS’
possible thresholds.
Responses from another sample were collected as a validation
dataset that examined whether dierences in ranked groups
estimated by the latent rank model predicted success levels in L2
(i.e., EFL) classrooms. Data were included from 158 Japanese EFL
learners (female = 22, male = 136) from four classrooms of a
university located in an urban city. eir ages ranged from 18 to
19 years (average = 18.32), and they had been taught English as a
compulsory school subject from grades 7 to 12. eir major was
engineering. At the university, they enrolled in an English course
for general purposes during the survey.
The foreign language reading anxiety scale
A Japanese-translated version of the FLRAS (Hamada and
Takaki, 2021a) was used to measure Japanese EFL students’
reading anxiety (see Table1) because the assessment by this scale
was more comprehensive than any of the other brief measurements
(Cheng, 2017). e word English in each statement was used
instead of the original words French, Russian, and Japanese in the
FLRAS (Saito et al., 1999, pp. 205–207). is psychometric
instrument consisted of 20 self-report items with a ve-point
Likert scale: (1) strongly disagree, (2) disagree, (3) neither agree
nor disagree, (4) agree, and (5) strongly agree. e sequence of the
questionnaire statements was rearranged using a random-
number method.
Based on the factor structure of the FLRAS (Matsuda and
Gobel, 2004; Hamada and Takaki, 2021a; see also Saito etal.,
1999), each item was labeled as reading diculty (Items 1–9), self-
ecacy in reading (Items 12–18), and language distance (Items
10–11 and 19–20). As Saito etal. (1999) suggested, these specic
statements could bequalitatively interpreted as dierent situation-
specic anxieties that might interfere with L2 learning. Specically,
low anxious students are more likely to befull of self-ecacy in
L2 reading and subsequently reach high L2 achievement (Mills
etal., 2007). e language distance indicates specic anxieties
toward unfamiliar writing systems and cultural material (Saito
etal., 1999).
L2 reading proficiency test
e standardized English reading prociency test (TOEIC
Bridge®; Educational Testing Service, 2007) was used to measure
participants’ L2 reading prociency. It had a multiple-choice format
and consisted of 50 items. Responses were marked dichotomously
(score range = 0–50). e test scores were used to examine the
association between L2 reading anxiety and prociency. As dened
in language testing (Bachman and Palmer, 2010), the reading
prociency test evaluated a static trait of learners’ reading skills
while the L2 achievement reected mastery of the just-completed
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 05
courses in which students were enrolled (Ross, 1998; see also
Teimouri etal., 2019; Zhang, 2019).
L2 course achievement assessment
e course grade from the other sample was used to indicate L2
achievement (see also Zhang, 2019). Since there were no participants
with learning disabilities, this study dened struggling students as
those who might drop out from a classroom even if they continued
to learn to read. As noted, participants took the achievement test in
partial fulllment of their English course for general purposes. e
test consisted of integrated reading-to-write task performance (40%),
independent listening skills (40%), and spoken interaction (20%) that
were introduced and practiced in the L2 classrooms to evaluate the
degree to which participants achieved learning goals (Bachman and
Palmer, 2010). e rating categories of the university were excellent
(90–100), very good (80–89), good (70–79), fair (60–69), and failing
(0–59). e course grade was used as a dependent variable to explore
whether the psychometric function could predict the participants
success (i.e., excellent to good) and fair-failing in the classroom.3
3 Based on Sparks etal. (2008), this study recognized students whose
grade was fair as being potentially struggling in L2 classrooms because
they would have failed the class if they missed a few more points on the
achievement test.
e survey was conducted during the authors’ regular L2
classes. Participants were notied of the study’s purpose and how
their personal data would be used. ey provided written
informed consent.
First, the L2 reading prociency test was implemented in
35 min. Next, the participants received detailed information on
how to answer the FLRAS and completed 20 self-report items
at their own pace. ey were also asked not to answer the
questions based on the specic class in which the questionnaire
was administered (see Matsuda and Gobel, 2004; Hamada
and Takaki, 2021a). ere was no set time limit but the
administration time was approximately 15 min. Apart from the
survey, the end-of-quarter test for the L2 achievement
assessment of the other sample was conducted approximately
2 months aer the FLRAS had been implemented to examine
whether the preceding L2 reading anxiety aected the degree
of success in the L2 classroom.
Data analysis
Questionnaires with missing values (0.89%) were excluded
resulting in the nal sample of 335 participants. e reverse code
TABLE1 Means with 95% CIs and SDs for each Foreign Language Reading Anxiety Scale (FLRAS) statement.
No. Statements M95% CI SD
Factor 1: Reading diculty (Cronbachs α = 0.82, 95% CI [0.78, 0.87])
1. I get upset when Iamnot sure whether Iunderstand what Iamreading in English. 3.60 [3.50, 3.70] 0.95
2. When reading English, Ioen understand the words but still cannot quite understand what the author saying. 3.28 [3.17, 3.39] 1.01
3. When Iamreading English, Iget so confused Icannot remember what Iamreading. 3.20 [3.08, 3.31] 1.05
4. I feel intimidated whenever Isee a whole page of English in front of me. 3.27 [3.14, 3.40] 1.18
5. I amnervous when Iamreading a passage in English when Iamnot familiar with the topic. 2.87 [2.76, 2.99] 1.08
6. I get upset whenever Iencounter unknown grammar when reading English. 3.56 [3.45, 3.67] 0.99
7. When reading English, Iget nervous and confused when Ido not understand every word. 3.44 [3.34, 3.55] 0.95
8. It bothers me to encounter words Icannot pronounce while reading English. 2.61 [2.49, 2.73] 1.11
9. I usually end up translating word by word when I’m reading English. 2.94 [2.83, 3.06] 1.05
Factor 2: Self-ecacy in reading (Cronbach’s α = 0.77 [0.73, 0.81])
12. I enjoy reading English. 2.73 [2.62, 2.84] 1.03
13. I feel condent when Iamreading in English. 2.45 [2.33, 2.56] 1.06
14. Once youget used to it, reading English is not so dicult. 3.26 [3.15, 3.37] 1.00
15. e hardest part of learning English is learning to read. 2.76 [2.66, 2.86] 0.92
16. I would behappy just to learn to speak English rather than having to learn to read as well. 3.35 [3.24, 3.45] 1.00
17. I do not mind reading to myself, but Ifeel very uncomfortable when Ihave to read English aloud. 2.81 [2.69, 2.93] 1.13
18. I amsatised with the level of reading ability in English that Ihave achieved so far. 1.88 [1.78, 1.98] 0.89
Factor 3: Language distance (Cronbachs α = 0.72 [0.68, 0.76])
10. By the time youget past the funny letters and symbols in English, it is hard to remember what youare reading about. 2.81 [2.70, 2.93] 1.06
11. I amworried about all the new symbols youhave to learn in order to read English. 2.75 [2.62, 2.84] 1.09
19. English culture and ideas seem very foreign to me. 2.17 [2.06, 2.27] 0.98
20. You have to know so much about English history and culture in order to read English. 3.13 [3.03, 3.24] 1.00
n = 335.
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 06
scale items (Items 12, 13, 14, and 18) were corrected aer reporting
the descriptive statistics (see Table1) so that a high value manifests
the same type of response on the other items. An item-total
correlation analysis showed no negatively correlated items with
the total anxiety scores (range = 0.00–0.65). All the materials and
data used in this study are available at the IRIS Digital Repository.
To answer the rst research question, a self-organized
mapping neural network was adapted in a latent rank analysis
using Exametrika version 5.5 (Shojima, 2019). Following Shojima
(2008), two criteria were considered to determine the number of
latent ranks of the FLRAS. First, the estimated ranks were aligned
ordinally and the principal components increased monotonically
because the observed data contained ordinal graded responses.
Under this condition, the latent rank model that t the observed
data best was selected based on the Akaike information criterion
(AIC) and Bayesian information criterion (BIC). en, the
probabilities of which ranked group the participants belonged to
were calculated (i.e., rank membership prole; Shojima, 2007).
e thresholds of L2 reading anxiety scores between the adjacent
two ranks were identied when certain anxiety scores signicantly
changed the rank membership prole. For example, an anxiety
score of 60 indicated if a participant belonged in Rank 1 or 2 with
a 60 and 40% probability, respectively, and a score of 61 indicated
if a participant belonged to Rank 1 or 2 with a 40 and 60%
probability, respectively, the cuto point for discriminating
between Rank 1 and 2 was determined as the anxiety score of 61.
In relation to the second research question, an implicational
analysis was conducted to describe the L2 reading anxiety
characteristics of each ranked group. e implicational analysis and
subsequent scaling are methods to display individual and group
variations of data to reveal both underlying systematicity in the data
and a theoretical explanatory model (Andersen, 1978). In this study,
the group average scores for each item were further rounded to the
nearest rst decimal point to examine which FLARS items
participants responded to positively and negatively. Namely, the
scores of 1.00–1.49, 1.50–2.49, 2.50–3.49, 3.50–4.49, and 4.50–5.00
were converted to 1, 2, 3, 4, and 5, indicating the participants strongly
disagreed, disagreed, neither disagreed nor agreed, agreed, and strongly
disagreed with particular statements. Using this approximated data,
an implicational scaling was created, in which the questionnaire
items were listed in descending order from the least to most anxious
situations in L2 reading as perceived by participants.
Finally, the third research question was investigated by
binominal logistic regression to predict the probabilities of
participants’ success in L2 classrooms based on their L2 reading
anxiety. L2 achievement was an indicator of success in the classroom,
binarily converted into “Success” (> = 70: Grades Excellent, Very
Good, and Good) and “Fair-Failing” (< 70: Grades Fair and Failing).
To evaluate the detective power for pedagogical screening, 70% of
the observed data was randomly split into a training set for building
a detective model. e remaining data were used as a test set for
evaluating this model. In addition, this study compared two
mediation models to evaluate the direct eect of L2 reading anxiety
even when L2 reading prociency was a mediating variable. If the
L2 reading anxiety merely reected the learners’ self-perception of
L2 reading diculties, its direct eect on L2 achievement would
disappear (i.e., a complete mediation model). In contrast, it could
bepossible that the direct eect of L2 reading anxiety remained
signicant while L2 reading prociency played a mediating role.
ese analyses were conducted using R-4.1.3 (R Core Team, 2021).
The FLRAS cuto points
Table1 displays the descriptive statistics of the FLRAS. e
measurement reliability was adequate (Cronbachs α = 0.83,
95% CI [0.81, 0.86]). e descriptive statistics for total FLRAS
scores were as follows: M = 61.71, 95% CI [60.63, 62.79],
SD = 10.02, Min = 28, Max = 91, and SE = 0.55. erefore, the
standard error of measurement for the FLRAS was 4.12. e
descriptive statistics of the L2 reading prociency test were as
follows: M = 31.61, 95% CI [30.55, 32.67], SD = 9.89, Min = 4,
Max = 49, and SE = 0.54. Internal consistency of the test was
adequately high (Cronbachs α = 0.91, 95% CI [0.89, 0.93]).
According to the 95% CIs of the means, no oor or ceiling
eects were found.
Figure1 shows changes in the principal components from 2-
to 5-rank models. is indicated the principal components
increased monotonically only in the 2- and 3-rank models. In
contrast, the results suggested no substantial dierences in L2
reading anxiety between Ranks 2 and 3in the 4-rank model and
between Ranks 2, 3, and 4in the 5-rank model. e observed data
t the 3-rank model (AIC = 18,215; BIC = 18,680) better than the
2-rank model (AIC = 18,536; BIC = 18,845). erefore, the
subsequent analyses were conducted using the 3-rank model of
the FLRAS.
Table2 displays the descriptive statistics of L2 reading anxiety
for each rank and thresholds between the adjacent two ranks. A
Kruskal–Wallis test4 showed signicant dierences in the L2
reading anxiety scores between the adjacent two ranks,
χ2(2) = 257.86, p < 0.001, with large eect sizes (Ranks 1–2:
p < 0.001, d = 2.00, 95% CI [1.68, 2.32]; Ranks 2–3: p < 0.001,
d = 1.84, 95% CI [1.51, 2.17]). is suggests that the L2 reading
anxiety scores considerably increased as per ranking. e
thresholds were the anxiety scores where the probabilities of the
participants belonging to each ranked group diered between the
adjacent two ranks. As shown in Figure 2, participants with
anxiety scores below 57 were highly likely to belong to Rank 1.
Participants with anxiety scores between 58 and 67 were grouped
into Rank 2. Participants with anxiety scores above 68 were in
Rank 3, showing the highest L2 reading anxiety.
4 Since there were some cases where dependent variables did not satisfy
the normality assumption, this study used the non-parametric test to
compare the outcomes.
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 07
Diagnostic characteristics of L2 reading
A Kruskal–Wallis test showed a signicant main eect of
L2 reading anxiety on L2 reading prociency, χ2(2) = 30.98,
p < 0.001 (s ee Table2). e participants in Rank 1, who showed
the least L2 reading anxiety, had better L2 reading prociency
than those in Rank 2 (p < 0.001, d = 0.65, 95% CI [0.38, 0.91])
and in Rank 3 (p < 0.001, d = 0.67, 95% CI [0.40, 0.94]). In
contrast, there was no signicant dierence between Ranks 2
and 3in L2 reading prociency (p = 0.842, d = 0.03, 95% CI
[0.30, 0.25]).
Table3 shows changes in average response scales for each
item from Ranks 1 to 3. Item discriminability5 also indicates
how big dierences among the three ranks were found. As
overall results indicated that the anxieties manifested by each
statement were likely to increase from Ranks 1 to 3, the FLRAS
could discriminate the individual dierences in L2 reading
anxiety. Specically, anxiety toward reading diculty (Items
1–9) was a strong discriminator of the learners (range = 0.45–
0.76). Although self-ecacy in reading also discriminated the
5 In the latent rank model, the values of item discriminability can
be considered in a similar way to factor loadings (Shojima, 2007,
2008,2009). This study used the conventional.30 and over (Finch and
French, 2018) when interpreting the discriminative power of each
questionnaire item.
characteristics of the three ranks (range = 0.31–0.49), Items 16
(0.12) and 18 (0.23) showed less discriminative power.
Language distance was also able to identify dierences between
the three ranks by Items 10 (0.58) and 11 (0.65), but not by
Items 19 (0.25) and 20 (0.15).
Table4 shows an implicational scaling that describes the
dierent participant characteristics by the ranked group.
Overall, anxiety toward language distance was not a stronger
cause of L2 reading anxiety than the other two factors. While
the factor of self-ecacy in reading also showed similar results,
Item 13 was related to relatively high anxiety on the scale.
Statements regarding reading diculty were located at the
relative bottom of the implicational scaling. is suggested that
anxiety toward reading diculty was the major source of L2
Changes in the principal component values for the 2- to 5-rank models.
TABLE2 Dierences in L2 reading anxiety, its subscales, and L2 reading proficiency between three latent ranks.
Rank 1 (n = 132) Rank 2 (n = 101) Rank 3 (n = 102)
Measures M95% CI SD M95% CI SD M95% CI SD
Overall L2 reading anxiety 52.50 [51.46, 53.54] 6.05 62.84 [62.11, 63.58] 3.72 72.50 [71.24, 73.76] 6.43
Reading diculty 2.66 [2.57, 2.75] 0.53 3.32 [3.25, 3.39] 0.36 3.88 [3.78, 3.97] 0.48
Self-ecacy in reading 2.82 [2.75, 2.90] 0.43 2.99 [2.92, 3.07] 0.38 2.66 [2.56, 2.76] 0.51
Language distance 2.08 [1.99, 2.17] 0.52 2.78 [2.67, 2.89] 0.54 3.23 [3.11, 3.35] 0.61
L2 reading prociency 35.45 [33.96, 36.93] 8.58 29.31 [27.22, 31.40] 10.59 29.57 [27.81, 31.32] 8.93
e thresholds between Ranks 1 and 2 and Ranks 2 and 3 were 57/58 and 67/68, respectively.
Probability density curves of the rank membership profiles. Two
vertical lines indicate the thresholds between the adjacent ranks.
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 08
reading anxiety. More specically, participants in Rank 1
responded, “disagree” and “neither disagree nor agree” for almost
all statements. Participants in Rank 2 also neither disagreed nor
agreed to the statements but showed high anxiety toward
unfamiliar grammatical features during L2 reading (Item 6).
Participants in Rank 3 were likely to negatively respond to
statements regarding reading diculty and condence in L2
reading (Item 13). e orthographic dierences between
Japanese and English were also a source of their high L2 reading
anxiety (Item 10).
TABLE3 Average response scales for each item among the three ranks and item characteristics.
Rank 1 (n = 132) Rank 2 (n = 101) Rank 3 (n = 102) Item
Item number and labels MSD MSD MSD discriminability
1: Reading diculty 3.14 1.05 3.48 0.63 4.33 0.57 0.52
2: Reading diculty 2.85 1.01 3.33 0.71 3.79 1.03 0.48
3: Reading diculty 2.58 0.99 3.27 0.66 3.92 0.94 0.58
4: Reading diculty 2.52 1.07 3.32 0.86 4.20 0.90 0.76
5: Reading diculty 2.17 0.83 3.26 0.77 3.40 1.15 0.59
6: Reading diculty 3.05 1.06 3.50 0.70 4.27 0.69 0.51
7: Reading diculty 2.94 1.02 3.47 0.64 4.08 0.68 0.47
8: Reading diculty 2.05 0.93 2.95 0.80 3.00 1.29 0.45
9: Reading diculty 2.44 0.99 2.85 0.80 3.69 0.92 0.54
10: Language distance 2.20 0.81 2.74 0.77 3.69 0.98 0.58
11: Language distance 2.07 0.89 2.94 0.72 3.43 1.13 0.65
12: Self-ecacy in reading 2.34 0.99 2.80 0.63 3.18 1.20 0.36
13: Self-ecacy in reading 3.10 1.07 3.44 0.75 4.25 0.95 0.49
14: Self-ecacy in reading 2.40 0.88 2.72 0.72 3.20 1.19 0.41
15: Self-ecacy in reading 2.36 0.80 2.95 0.70 3.10 1.07 0.31
16: Self-ecacy in reading 3.31 1.03 3.32 0.79 3.42 1.14 0.12
17: Self-ecacy in reading 2.11 0.96 3.20 0.71 3.32 1.20 0.48
18: Self-ecacy in reading 4.19 0.80 3.55 0.91 4.59 0.67 0.23
19: Language distance 1.61 0.70 2.74 0.81 2.32 1.05 0.25
20: Language distance 3.08 1.13 3.03 0.71 3.31 1.04 0.15
High values of Items 12, 13, 14, and 18 (reverse coded) indicate high anxiety. Generally, the discriminability among ranks became low when the items did not show monotonic increase.
TABLE4 Implicational analysis summary results.
Approximated response scale
Item number and labels Rank 1 Rank 2 Rank 3
11: Language distance low 2average 3average 3
12: Self-ecacy in reading low 2average 3average 3
14: Self-ecacy in reading low 2average 3average 3
15: Self-ecacy in reading low 2average 3average 3
17: Self-ecacy in reading low 2average 3average 3
5: Reading diculty low 2average 3average 3
8: Reading diculty low 2average 3average 3
10: Language distance low 2average 3 high 4
3: Reading diculty low 2average 3 high 4
4: Reading diculty low 2average 3 high 4
9: Reading diculty low 2average 3 high 4
1: Reading diculty average 3average 3 high 4
2: Reading diculty average 3average 3 high 4
7: Reading diculty average 3average 3 high 4
13: Self-ecacy in reading average 3average 3 high 4
6: Reading diculty average 3 high 4 high 4
Items 16, 18, 19, and 20 were removed from the implicational scaling due to extremely low item discriminability.
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 09
Pedagogical screening
e practicality of the FLRAS thresholds was investigated
using the other sampled population. Table5 shows the descriptive
statistics of their L2 reading anxiety scores and L2 achievement
assessment for the three ranked groups. Kruskal–Wallis tests
showed signicant main eects of the ranked groups on both L2
reading anxiety, χ2(2) = 107.34, p < 0.001, and L2 achievement,
χ2(2) = 34.78, p < 0.001. Multiple comparisons with Holms
adjustment demonstrated that the participants in Rank 3 reached
considerably less L2 achievement than Rank 1 (p < 0.001, d = 1.51,
95% CI [0.98, 2.04]) and Rank 2 (p < 0.001, d = 1.66, 95% CI [1.17,
2.15]). ere was no outstanding dierence between Rank 1 and
Rank 2 (p = 0.650, d = 0.09, 95% CI [0.45, 0.27]), although their
L2 reading anxiety scores diered substantially (p < 0.001,
d = 2.86, 95% CI [2.36, 3.36]). e correlation between their L2
reading prociency and achievement was r = 0.37 (95% CI [0.27,
0.46]), suggesting both tests measured dierent constructs of L2
performance as intended (Ross, 1998).
A logistic regression model established by the training dataset
showed that L2 reading anxiety explained the variances of success
probabilities in the L2 classrooms (β = 0.15, SE = 0.04, z = 4.16,
p < 0.001). e psychometric function, predicting the outcome of
an observation given a predictor variable (L2 reading anxiety), is
an S-shaped curve. As plotted in Figure3, the FLRAS thresholds
indicated that the probability of success in L2 classrooms that
dierentiated between Ranks 1 and 2 was 88%. Such probability
between Ranks 2 and 3 was 63%. e accuracy rate for detecting
the struggling students in the L2 classrooms was 79% in the
test dataset.
Finally, Figure4 shows the standardized path coecients from
L2 reading anxiety to prociency (β = 0.52, 95% CI [0.70, 0.33],
p < 0.001), from prociency to achievement (β = 0.21, 95% CI
[0.03, 0.44], p = 0.097), and from anxiety to achievement
(β = 0.31, 95% CI [0.61, 0.02], p = 0.037). ese results indicate
a partial mediation model, in which L2 reading anxiety aected the
degree of L2 achievement partially because of the mediating role of
L2 reading prociency. Importantly, Figure5 indicates that the
direct eect of L2 achievement on L2 reading anxiety was also
signicant (β = 0.26, 95% CI [0.49, 0.04], p = 0.022). is mo del
t the observed data (AIC = 1,748, BIC = 1,779) better than the
former model (AIC = 3,021, BIC = 3,055). Taken together, although
the mediating eects of L2 reading prociency can never beignored,
the direct eect of L2 reading anxiety might beconsidered for the
factor aecting pedagogical screening. However, it is highly possible
that the degree of L2 achievement determined the magnitude of L2
reading anxiety.
is study applied a latent rank model to the FLRAS for
pedagogical screening of the students who would bestruggling in
L2 classrooms. Reading is an essential cognitive activity for L2
learning (e.g., Grabe, 2009) but demanding for learners who feel
highly anxious toward reading in an L2 (Saito etal., 1999; Sellers,
2000; Matsuda and Gobel, 2004; Zhao etal., 2013; Jee, 2016;
Hamada and Takaki, 2021a,b). Because high L2 reading anxiety
can beassociated with reading attitude in a classroom (Yamashita,
2007), wepredicted that particular groups of learners who showed
certain symptoms of L2 reading anxiety led to dierent levels of
L2 achievement. e latent rank model provided evidence that the
FLRAS can diagnose L2 reading anxiety of struggling students in
L2 classrooms. e three discrete groups showed dierent
TABLE5 Means with 95% CI and SD for L2 reading anxiety and L2 achievement.
L2 reading anxiety L2 achievement
Groups n M 95% CI SD M95% CI SD
Rank 1 48 53.23 [52.19, 54.27] 3.58 82.79 [79.52, 86.07] 11.28
Rank 2 82 62.33 [61.69, 62.97] 2.93 83.76 [81.36, 86.15] 10.91
Rank 3 28 70.57 [69.56, 71.58] 2.60 64.43 [59.17, 69.69] 13.56
e L2 achievement test reliability was adequate [Cronbach’s α = 0.74, 95% CI (0.68, 0.80)].
A probability curve with a 95% CI of the success in the L2
classrooms modeled by the logistic regression. A jitter-plot
represents the actual points of each observation (The ratio of
success to fair-failing: Rank 1 = 41/7, Rank 2 = 70/12, and Rank
3 = 10/18). Dashed lines indicate the thresholds of the L2 reading
anxiety scores that discriminate between the probabilities of the
success in the L2 classrooms and Ranks 1–3.
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 10
symptoms of L2 reading anxiety and L2 achievement. Moreover,
the psychometric function applicable to the FLRAS could predict
the probability of success in L2 classrooms with 79% accuracy. In
line with these ndings, the theoretical and methodological issues
for psychometric assessment of individual dierences in L2
reading anxiety will bediscussed.
e rst research question addressed FLRAS cuto points that
can discriminate dierences in L2 reading anxiety among groups of
L2 learners. e results showed that it could dierentiate the
characteristics among only three groups. Dierences among ranked
groups were not clear for classifying participants into four or more
groups (see Figure1). e FLRAS’ standard error of measurement
also indicated that the true score of L2 reading anxiety per
participant varied from 4.12 to 4.12. ese ndings suggested the
FLRAS was not reliable enough to discriminate L2 learners on its
20–100 continuous scale. Although previous studies have used the
raw scores (e.g., Saito etal., 1999; Matsuda and Gobel, 2004; Wu,
2011; Zhao etal., 2013; Xiao and Wong, 2014; Jee, 2016), it should
be noted that individual anxiety scores do not always reect
substantial dierences in individual L2 reading anxiety.
Specically, the latent rank analysis showed the score range
of the FLRAS can bemapped into a three-point discrete scale. By
grouping participants with the latent rank information, their L2
reading prociency was found to signicantly dier between the
low-anxiety group (Rank 1) and the other two groups (Ranks 2
and 3). Consistent with the present result, dierences between
average- and high-anxiety groups were sometimes unclear in
previous studies (Phillips, 1992; Ganschow and Sparks, 1996;
Hewitt and Stephenson, 2012). However, these studies commonly
provided evidence that the low-anxiety group was always the
most procient in L2 prociency tests. Although there were
dierences in the questionnaires used, the present result was
consistent with Ganschow and Sparks (1996) showing that the
low-anxiety group was the most procient in L2 reading. Given
the relatively weak correlations between L2 anxiety and
prociency (Teimouri etal., 2019; Zhang, 2019), it is reasonable
that group dierences in L2 prociency were not large.
e second research question explored what kind of
characteristics can be diagnosed for each ranked group by the
FLRAS. e results of the implicational analysis found qualitative
dierences between the three ranked groups (see Table5). More
specically, reading diculty was the strongest factor that
dierentiated the ranked groups, followed by self-ecacy in reading,
and language distance. is result was fully consistent with previous
A mediation model of the eects of L2 reading anxiety on L2 achievement. Values in brackets are 95% CIs.
A mediation model of the eects of L2 achievement on L2 reading anxiety. Values in brackets are 95% CIs.
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 11
studies that showed the relative proportions of variances explained
by these three factors. Matsuda and Gobel (2004) and their
replication study (Hamada and Takaki, 2021a) revealed that reading
diculty explained the largest variance of the FLRAS responses,
followed by self-ecacy in reading and language distance. e result
also supported the evidence that cognitive processes and linguistic
knowledge are major components of L2 reading (Grabe, 2009),
resulting in a source of perceived reading diculty.
More specically, participants categorized into Rank 1
exhibited good conditions in L2 reading anxiety. ey responded
with less impacts for dierences in orthographic features and
writing system on their L2 reading anxiety (Items 10 and 11).
Reading was also a part of their enjoyment (Item 12) and not
dicult to learn in L2 classrooms (Items 14, 15, and 17) even
though their condence in L2 reading was slightly high (Item 13)
compared to the other specic situations of self-ecacy. Reading
diculty caused by cognitive processing involved in L2 reading did
not make them uneasy (Items 3, 4, 5, 8, and 9). Instead, anxieties
toward linguistic knowledge such as unfamiliar words (Item 7) or
grammar (Item 6) were higher among participants in Rank 1.
Participants in Rank 2 showed similar trends, only responding
negatively to unfamiliar grammar during L2 reading. However, their
anxieties toward several aspects substantially increased compared to
participants in Rank 1. First, the level of L2 reading anxieties related
to language distance (Items 10 and 11) and self-ecacy (Items 12,
14, 15, and 17) increased from low to average. Likewise, perceived
reading diculty of participants in Rank 2 was generally higher than
that of Rank 1. e L2 reading anxiety of Rank 3 spiked even further,
particularly regarding several reading diculties. eir anxiety levels
were on average only toward unfamiliar topics of a passage (Item 5)
and word decoding (Item 8) in L2 reading. Compared to participants
in Ranks 1 and 2, they did not feel condent during L2 reading. e
orthographic dierences between Japanese and English were also a
source of their high L2 reading anxiety (Item 10). In contrast, their
self-ecacy in L2 reading did not dier from Rank 2 students. ese
results suggest that while highly anxious students perceived their L2
reading ability as low due to insucient cognitive processing, they
might feel that L2 reading is not fun, but not painful either.
ese qualitative dierences among the ranked groups highlight
the importance of considering the relative inuences of situation-
specic reading anxiety when interpreting the FLRAS responses.
Previous studies provided diagnostic information by comparing
dierent cultural groups of learners (Saito et al., 1999) and
qualitative analyses of interview protocols (Zhao etal., 2013). Other
studies used denite cuto points based on the Likert-scale (Xiao
and Wong, 2014; Jee, 2016). e present ndings added a more ne-
grained view that the FLRAS can diagnose individual dierences in
L2 reading anxiety. Such diagnostic information is useful to identify
the strengths and weakness of L2 readers (Alderson etal., 2016) and
examine relationships with L2 learning problems that lead to L2
achievement (Ganschow and Sparks, 1991, 2001).
Finally, the third question was related to the practical, but
ignored use of the FLRAS and other psychometrics in L2 anxiety
research. e results showed the psychometric function of the FLRAS
could accurately identify students who were likely to besuccessful or
struggling in L2 classrooms. In other words, L2 reading anxiety
played a signicant role in the odds of being successful L2 learners or
not (Alderson etal., 2016). No doubt, variations related to high and
low perceptions of L2 reading anxiety helped guess who would
bestruggling in L2 classrooms and those considered good L2 readers,
respectively. In fact, the probabilities of success in L2 classrooms
varied considerably according to the three ranked groups. As shown
in Figure 3, the S-shaped curve for Rank 1 was a gradual slope
compared to Ranks 2 and 3. is suggested that a student labeled as
a prospectively successful L2 learner (Rank 1) was likely to achieve
particular learning goals in L2 classrooms. e aforementioned
results supported this nding because Rank 1 students were likely to
manifest the lowest anxiety toward reading diculty and language
distance. ey were also full of self-ecacy despite relatively low
condence in L2 reading. ese arguments were consistent with
several studies that showed individual dierences in L2 reading
anxiety as the psychological factors dening strengths of successful
L2 readers (Saito etal., 1999; Mills et al., 2007; Zhao etal., 2013;
Xiao and Wong, 2014; Alderson etal., 2016; Jee, 2016).
Figure3 also shows many Rank 2 students were successful in
their classrooms. Because they did not show high L2 reading
anxiety with respect to reading diculty, self-ecacy in reading,
and language distance, the means of their L2 achievement test did
not dier from those of students in Rank 1. However, the actual
data points indicated the growth of the number of students who
received a fair or failing grade around the threshold between Ranks
2 and 3. In line with this result, the probability of success in L2
classrooms dropped to 63% as the students’ L2 reading anxiety
score approached to 67. Although the implicational analysis did
not produce any characteristics of the L2 reading anxiety of Rank
2, it should beinterpreted with caution when they showed relatively
strong overall L2 reading anxiety. Particularly, students who
manifested strong anxiety toward unfamiliar grammar and much
less condence in L2 reading could be labeled as potentially
unsuccessful in L2 classrooms (see also Zhao etal., 2013).
As noted, students in Rank 2 were not found to beprospectively
unsuccessful in L2 classrooms, although their L2 reading
prociency was not as good as that of the Rank 3 students. is
result is explainable from the viewpoint of the dierent natures of
L2 reading prociency and achievement tests. While prociency
tests involve contents unrelated to the language courses, the
contents of achievement tests must berelated to course learning in
which learners were engaged (Ross, 1998; Bachman and Palmer,
2010). Given that less anxious learners were likely to bemore active
in L2 classroom learning (e.g., Horwitz et al., 1986; Saito etal.,
1999; Horwitz, 2001; Yamashita, 2007; Zhao et al., 2013), it is
possible that the Rank 2 students could achieve course learning
goals because of relatively low L2 reading anxiety. e weak
correlation between L2 reading prociency and achievement also
supports the interpretation that anxiety, self-ecacy, and
condence in L2 reading aected the degree of class engagement
and enjoyment more than L2 reading prociency (Matsuda and
Gobel, 2004; Mills etal., 2007). Consistent with Alderson etal.
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 12
(2016), the student group with low anxiety, but low prociency can
beregarded as being in a developmental stage of L2 reading.
As expected, great care should betaken with Rank 3 students.
e results indicated the probability of success in L2 classrooms
decreased precipitously when their L2 reading anxiety scores crossed
the second threshold of the FLRAS (> = 68). e means of their L2
achievement test were also much lower than Rank 1 and 2 students.
Because the majority of students who received a fair or failing grade
were classied into Rank 3, the latent rank model has the potential
to identify the students being struggling in L2 classrooms. Consistent
with Ganschow and Sparks (1991), students who were labeled as
potentially unsuccessful in L2 learning were inferior in L2 reading
skills. Unlike the students of Rank 2, it is possible that the double
bindings caused by low prociency and high anxiety in L2 reading
hurt them, leading to the lowest L2 achievement among the groups.
Moreover, the results were consistent with Alamer and Lee (2021)
and Sparks and Alamer (2022) that lower L2 achievement increased
the magnitude of L2 anxiety. Although the relationships between L2
anxiety and prociency will determine student achievement in L2
classroom learning (Horwitz, 2001; Dörnyei and Ryan, 2015;
MacIntyre, 2017), it is also important to consider that the promising
solution to reducing L2 reading anxiety is to develop L2 reading skills.
e present ndings emphasize the importance of understanding
learners’ aective proles to classify them into suitable learning
environments. Proling data regarding specic anxieties in response
to L2 reading will determine what kind of instruction is necessary
for each group. For example, the perceived diculties in L2 reading
dierentiated the ranked groups (see Table 4), and the priority
should beto improve the skill and knowledge necessary for reading
comprehension. is perspective is consistent with the mediation
analysis results, in which the participants perceived higher anxiety
as a result of lower L2 reading prociency. Aer improving the level
of L2 reading prociency, teachers may beable to help the students
develop their self-ecacy to reduce L2 reading anxiety further.
Given the associational nature of language anxiety and prociency
(Teimouri etal., 2019), the language anxiety scales can befunctioned
as basic diagnostic testing.
Most L2 learners perceive L2 anxiety in classrooms, to which
teachers do not attribute adequate importance (Tran etal., 2013).
Given the clear importance of assessing individual dierences in L2
learning, the present study applied the latent rank model to identify
struggling students in L2 classrooms. e results showed the FLRAS
was not sensitive enough to discriminate L2 reading anxiety on its
continuous scale. Instead, the FLRAS could categorize students into
three ranked groups according to substantial dierences in L2
reading anxiety symptoms. e psychometric function provided by
the estimated cuto points also helped determine success
probabilities in L2 classrooms. ese ndings signicantly contribute
to improving the learning experiences in L2 classrooms as well as the
assessment quality of individual dierences in L2 learning.
Toward future research, several factors other than L2
anxiety must beincorporated to identify struggling students
in L2 learning. For example, Ganschow and Sparks (1991)
showed the predictive power of learners’ L2 learning history,
developmental history, academic learning history, and tests
and classroom learning characteristics in identifying students
with L2 learning disabilities. The present study conducted
brief screening in educational settings; therefore, the
integration of potential cognitive and affective factors
determining L2 achievement will advance theoretical and
methodological discussions in research on individual
differences in L2 learning.
Data availability statement
e datasets presented in this study can befound in online
repositories. e names of the repository/repositories and
accession number(s) can befound at: https://www.iris-database.
Ethics statement
e studies involving human participants were reviewed and
approved by Nihon University. e patients/participants provided
their written informed consent to participate in this study.
Author contributions
All authors listed have made a substantial, direct, and
intellectual contribution to the work and approved it for publication.
is study was supported by Grants-in-Aid for Scientic
Research (B) no. 20H01287 and for Young Scientists (B) no.
18K12443 from the Japan Society for the Promotion of Science.
e authors wish to acknowledge the editor and the reviewers
for their valuable comments to improve an earlier version of
this manuscript.
Conflict of interest
e authors declare that the research was conducted in the
absence of any commercial or nancial relationships that could
beconstrued as a potential conict of interest.
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 13
Publisher’s note
All claims expressed in this article are solely those of the
authors and do not necessarily represent those of their aliated
organizations, or those of the publisher, the editors and the
reviewers. Any product that may be evaluated in this article, or
claim that may be made by its manufacturer, is not guaranteed or
endorsed by the publisher.
Alamer, A., and Lee, J. (2021). Language achievement predicts anxiety and not the
other way around: a cross-lagged panel analysis approach. Lang. Teach. Res. doi:
10.1177/13621688211033694 [Epub ahead of print].
Alderson, J. C., Huhta, A., and Nieminen, L. (2016). Characteristics of weak and strong
readers in a foreign language. Mod. Lang. J. 100, 853–879. doi: 10.1111/modl.12367
American Educational Research Association (2014). Standards for Educational
and Psychological Testing. Washington, D.C.: e Author.
Andersen, R. W. (1978). An implicational model for second language research.
Lang. Learn. 28, 221–282. doi: 10.1111/j.1467-1770.1978.tb00134.x
Bachman, L. F., and Palmer, A. S. (2010). Language Assessment in Practice:
Developing Language Assessments and Justifying eir Use in the Real World. Oxford,
UK: Oxford University Press.
Cheng, Y.-S. (2017). Development and preliminary validation of four brief
measures of L2 language-skill-specic anxiety. System 68, 15–25. doi: 10.1016/j.
Cheng, Y.-S., Horwitz, E. K., and Schallert, D. L. (1999). Language anxiety:
Dierentiating writing and speaking components. Lang. Learn. 49, 417–446. doi:
Crowther, D., Kim, S., Lee, J., Lim, J., and Loewen, S. (2021). Methodological
synthesis of cluster analysis in second language research. Lang. Learn. 71, 99–130.
doi: 10.1111/lang.12428
Dörnyei, Z., and Ryan, S. (2015). e Psychology of the Language Learner Revisited.
New York, NY: Routledge.
Educational Testing Service (2007). TOEIC Bridge® Ocial Guide & Question
Collection. Tokyo, Japan: Institute for International Business Communication
Elkhafai, H. (2005). Listening comprehension and anxiety in the Arabic language
classroom. Mod. Lang. J. 89, 206–220. doi: 10.1111/j.1540-4781.2005.00275.x
Finch, W. H., and French, B. F. (2018). Educational and Psychological Measurement.
New York, NY: Routledge
Ganschow, L., and Sparks, R. (1991). A screening instrument for the identication
of foreign language learning problems. Foreign Lang. Ann. 24, 383–398. doi:
Ganschow, L., and Sparks, R. (1996). Anxiety about foreign language learning
among high school women. Mod. Lang. J. 80, 199–212. doi: 10.1111/j.1540-4781.1996.
Ganschow, L., and Sparks, R. (2001). Learning diculties and foreign language
learning: a review of research and instruction. Lang. Teach. 34, 79–98. doi: 10.1017/
Grabe, W. (2009). Reading in a Second Language: Moving From eory to Practice.
New York, NY: Cambridge University Press.
Hamada, A., and Takaki, S. (2021a). Approximate replication of Matsuda and
Gobel (2004) for psychometric validation of the foreign language reading anxiety
scale. Lang. Teach. 54, 535–551. doi: 10.1017/S0261444819000296
Hamada, A., and Takaki, S. (2021b). Eects of multidimensional foreign language
reading anxiety on achievement in Japanese EFL classrooms. System 101:102613.
doi: 10.1016/j.system.2021.102613
Hasselblad, V., and Hedges, L. V. (1995). Meta-analysis of screening and diagnostic
tests. Psychol. Bull. 117, 167–178. doi: 10.1037/0033-2909.117.1.167
Hewitt, E., and Stephenson, J. (2012). Foreign language anxiety and oral exam
performance: a replication of Phillips’s MLJ study. Mod. Lang. J. 96, 170–189. doi:
Horwitz, E. K., Horwitz, M. B., and Cope, J. (1986). Foreign language classroom
anxiety. Mod. Lang. J. 70, 125–132. doi: 10.1111/j.1540-4781.1986.tb05256.x
Horwitz, E. K. (2001). Language anxiety and achievement. Annu. Rev. Appl.
Linguist. 21, 112–126. doi: 10.1017/S0267190501000071
Jee, M. J. (2016). Exploring Korean heritage language learners’ anxiety: ‘we are not
afraid of Korean!’. J. Multiling. Multicult. Dev. 37, 56–74. doi: 10.1080/
Lambert, C. (2010). A task-based needs analysis: putting principles into practice.
Lang. Teach. Res. 14, 99–112. doi: 10.1177/1362168809346520
Liu, R., and Jiang, Z. (2018). Diagnostic classication models for ordinal item
responses. Front. Psychol. 9:2512. doi: 10.3389/fpsyg.2018.02512
Liu, R., and Jiang, Z. (2020). A general diagnostic classication model for rating
scales. Behav. Res. Methods 52, 422–439. doi: 10.3758/s13428-019-01239-9
MacIntyre, P. D. (1999). “Language anxiety: A review of the research for language
teachers,” in Aect in Foreign Language and Second Language Learning. ed. D. J.
Young (New York, NY: McGraw-Hill), 24–45.
MacIntyre, P. D. (2017). “An overview of language anxiety research and trends in
its development,” in New Insights Into Language Anxiety: eory, Research and
Educational Implications. eds. C. Gkonou, M. Daubney and J. M. Dewaele (Bristol,
UK: Multilingual Matters).
MacIntyre, P. D., and Gardner, R. C. (1991). Methods and results in the study of
anxiety and language learning: a review of the literature. Lang. Learn. 41, 85–117.
doi: 10.1111/j.1467-1770.1991.tb00677.x
Marcos-Llinás, M., and Garau, M. J. (2009). Eects of language anxiety on three
prociency-level courses of Spanish as a foreign language. Foreign Lang. Ann. 42,
94–111. doi: 10.1111/j.1944-9720.2009.01010.x
Matsuda, S., and Gobel, P. (2004). Anxiety and predictors of performance in the
foreign language classroom. System 32, 21–36. doi: 10.1016/j.system.2003.08.002
Mills, N., Pajares, F., and Herron, C. (2007). Self-ecacy of college intermediate
French students: relation to achievement and motivation. Lang. Learn. 57, 417–442.
doi: 10.1111/j.1467-9922.2007.00421.x
Oxford, R., and Ehrman, M. (1992). Second language research on individual
dierences. Annu. Rev. Appl. Linguist. 13, 188–205. doi: 10.1017/S0267190500002464
Pae, T.-I. (2013). Skill-based L2 anxieties revisited: eir intra-relations and inter-
relations with general foreign language anxiety. Appl. Lingust. 34, 232–252. doi:
Pastor, D. A., Barron, K. E., Miller, B. J., and Davis, S. L. (2007). A latent prole
analysis of college students’ achievement goal orientation. Contemp. Educ. Psychol.
32, 8–47. doi: 10.1016/j.cedpsych.2006.10.003
Phillips, E. M. (1992). e eects of language anxiety on students’ oral test performance
and attitudes. Mod. Lang. J. 76, 14–26. doi: 10.1111/j.1540-4781.1992.tb02573.x
R Core Team (2021). R: A Language and Environment for Statistical
Computing. R Foundation for Statistical Computing. Available at: https://www.R-
Ravand, H., and Baghaei, P. (2020). Diagnostic classication models: recent
developments, practical issues, and prospects. Int. J. Test. 20, 24–56. doi:
Ross, S. (1998). Self-assessment in second language testing: a meta-analysis and
analysis of experiential factors. Lang. Test. 15, 1–20. doi: 10.1177/02655322
Saito, Y., Horwitz, E. K., and Garza, T. J. (1999). Foreign language reading anxiety.
Mod. Lang. J. 83, 202–218. doi: 10.1111/0026-7902.00016
Sellers, V. D. (2000). Anxiety and reading comprehension in Spanish as a foreign
language. Foreign Lang. Ann. 33, 512–520. doi: 10.1111/j.1944-9720.2000.tb01995.x
Shao, K., Yu, W., and Ji, Z. (2013). An exploration of Chinese EFL students'
emotional intelligence and foreign language anxiety. Mod. Lang. J. 97, 917–929. doi:
Shojima, K. (2007). Neural test theory. DNC Research Note, 07-02. Available at:
Shojima, K. (2008). Neural test theory: A latent rank theory for analyzing test
data. DNC Research Note, 08-01. Available at:
Shojima, K. (2009). “Neural test theory,” in New Trends in Psychometrics. eds. K.
Shigemasu, A. Okada, T. Imaizumi and T. Hoshino (Tokyo, Japan: Universal
Academy Press), 417–426.
Shojima, K. (2019). Exametrika (Version 5.5). [Computer soware]. Available at:
Sparks, R. L., and Alamer, A. (2022). Long-term impacts of L1 language skills on
L2 anxiety: e mediating role of language aptitude and L2 achievement. Lang.
Teach. Res. doi: 10.1177/13621688221104392 [Epub ahead of print].
Hamada and Takaki 10.3389/fpsyg.2022.938719
Frontiers in Psychology 14
Sparks, R. L., Humbach, N., and Javorsky, J. (2008). Individual and longitudinal
dierences among high and low-achieving, LD, and ADHD L2 learners. Learn.
Individ. Dier. 18, 29–43. doi: 10.1016/j.lindif.2007.07.003
Sparks, R. J., Luebbers, J., Castañeda, M., and Patton, J. (2018a). High school
Spanish students and foreign language reading anxiety: Déjà vu all over again all
over again. Mod. Lang. J. 102, 533–556. doi: 10.1111/modl.12504
Sparks, R. L., Patton, J., and Luebbers, J. (2018b). L2 anxiety and the foreign
language reading anxiety scale: listening to the evidence. Foreign Lang. Ann. 51,
738–762. doi: 10.1111/an.12361
Swanson, H. L. (2017). A latent transition analysis of English learners with
reading disabilities: do measures of cognition add to predictions of late
emerging risk status? Top. Lang. Disord. 37, 114–135. doi: 10.1097/TLD.000
Teimouri, Y., Goetze, J., and Plonsky, L. (2019). Second language anxiety and
achievement. A meta-analysis. Stud. Second. Lang. Acquis. 41, 363–387. doi:
Tran, T. T. T., Baldauf, R. B. Jr., and Moni, K. (2013). Foreign language anxiety:
understanding its status and insiders’ awareness and attitudes. TESOL Q. 47,
216–243. doi: 10.1002/tesq.85
Wu, H. J. (2011). Anxiety and reading comprehension performance in English as
a foreign language. Asian EFL J. 13, 273–307.
Xiao, Y., and Wong, K. F. (2014). Exploring heritage language anxiety: a study of
Chinese heritage language learners. Mod. Lang. J. 98, 589–611. doi: 10.1111/
Yamashita, J. (2007). e relationship of reading attitudes between L1 and L2: an
investigation of adult EFL learners in Japan. TESOL Q. 41, 81–105. doi:
Zhang, X. (2019). Foreign language anxiety and foreign language performance: a
meta-analysis. Mod. Lang. J. 103, 763–781. doi: 10.1111/modl.12590
Zhao, A., Guo, Y., and Dynia, J. (2013). Foreign language reading anxiety: Chinese
as a foreign language in the UnitedStates. Mod. Lang. J. 97, 764–778. doi: 10.1111/j.
... In a recent study, Hamada and Takaki (2022) applied a latent rank model to identify Japanese learners of English who were struggling or successful in L2 classrooms according to their L2 reading anxiety symptoms on the FLRAS. Their findings showed that the FLRAS classified students into three ranked groups: low L2 reading anxiety, high anxiety about unfamiliar grammar knowledge in L2 reading, and even higher anxiety about L2 vocabulary and L2 grammar knowledge deficits. ...
... In addition, findings support speculation that L2 aptitude on the MLAT, a measure of the ability to think about language, is itself a cognitive resource. Speculation about cognitive (language-based) resources is supported by Hamada and Takaki's (2022) findings, which showed that students who report higher anxiety (on the FLRAS) have weaker grammar and vocabulary knowledge in the L2 as well as lower L2 reading skills. The findings from the present investigation also lend additional support to Alamer and Lee's (2021) and Botes et al.'s (2020b) study which showed that language achievement precedes language anxiety in the L2 classroom. ...
... Sparks (1995) has contended that L2 anxiety scales like the FLRAS (and FLCAS) are likely to be contaminated by language achievement because the items ask students, directly and explicitly, about their language skills. The results of this study, which found that the relationship between L1 achievement and L2 anxiety is both direct and indirect as mediated by language variables [L2 aptitude on the MLAT, L2 achievement, L1 metalinguistic (literacy) knowledge], suggest that language educators and researchers should consider developing instruments that measure language anxiety uncontaminated by language ability and/or use language anxiety surveys as a proxy for students' language skills, i.e., higher anxiety and weaker language skills, and vice versa (see Hamada & Takaki, 2022). As noted by Hamada and Takaki (2021), the factorial structure of FLRAS differs from context to context and there is no strong agreement on the measurement model of the scale. ...
Full-text available
Although studies have consistently found negative correlations between second language (L2) anxiety and L2 achievement, the anxiety concept is criticized for the failure to consider confounding variables, particularly first language (L1) achievement, in the relationship between language anxiety and language achievement. Also, past studies tend to rely on cross-sectional data to conclude causality. To assess long-term causal relationships between L1 and L2 achievement, the study used structural equation modeling to examine pathways by which L1 achievement is associated with L2 reading anxiety after two years via several mediators: L1 working memory, L1 print exposure, and L1 metalinguistic knowledge; L2 aptitude (MLAT); and L2 reading, writing, and listening comprehension. Students (n = 293) in a U.S. secondary school were administered these measures and followed through first-year Spanish. Findings showed the effect of L1 achievement on later L2 reading anxiety was direct and indirect through L2 aptitude, L2 achievement, and L1 metalinguistic knowledge. Effects of L2 achievement and L1 metalinguistic knowledge on L2 reading anxiety were direct. An out-of-sample analysis for model prediction power supported the external validity of the study results. We recommend that teachers and learners deal with L1/L2 skills, rather than anxiety per se, in order to reduce sense of L2 reading anxiety.
Full-text available
Although researchers have found strong relationships between first language (L1) skills and second language (L2) anxiety, less is known about how L1 skills in primary school impact L2 anxiety years after L1 has been acquired. Previous research lacked evidence for the mediating role of L2 aptitude and L2 achievement in the relationship. This study applies structural equation modelling to examine pathways by which L1 skills are associated with L2 anxiety via the mediators. Sixty-five students were administered L1 skill measures from 1st–5th grades, then followed over 10 years when completing L2 courses in high school. We examined the model across six different ages and found the effect of L1 skills on L2 anxiety was indirect through L2 aptitude and L2 achievement, indicating the impact is better understood through the mediators. The effect of L2 achievement on L2 anxiety is direct, and large in magnitude. Results remained stable across different ages and invariant across gender.
Full-text available
A fine-grained understanding of individual differences in reading in English as a foreign language (EFL) is critical because a significant amount of EFL reading is essential for successful EFL learning. This study examined the effects of multidimensional EFL reading anxiety on learners' achievement in EFL classrooms. In a preliminary study, an item pool comprising approximately 150 statements was compiled and tested with multidimensional item response theory for the assessment of cognitive, metacognitive, and classroom aspects of perceived EFL reading anxiety. The main study provided evidence of the replicability of the multidimensionality of EFL reading anxiety and its differential effects on EFL reading proficiency and course achievement. The results showed that cognitive anxiety influenced EFL reading proficiency negatively, but not course achievement. Classroom anxiety inhibited EFL course achievement more strongly than did reading proficiency. Metacognitive anxiety was a significant predictor of both EFL reading proficiency and course achievement. Mediation analysis revealed that although the anxiety scales partly reflected learners' levels of EFL reading proficiency, the direct effects of EFL reading anxiety on course achievement were still significant. Theoretical, methodological, and pedagogical implications are discussed with respect to the multidimensional model of EFL reading anxiety.
Full-text available
While language anxiety is known to be a strong predictor of the second language (L2) achievement for decades, there has been a relative lack of empirical attention to establishing which one-anxiety or achievement-influences the other. The present study, based on a cross-lagged panel analysis, examined the causal relationships between language anxiety and L2 achievement, each of which was measured at three times points across 17 weeks. The results supported the directional relationship from L2 achievement at Time 2 to language anxiety at Time 3 while the opposite directional relationship from language anxiety at Time 2 to L2 achievement at Time 3 was not supported. Thus, our results suggest that language achievement precedes anxiety and not the other way round. The moderation analysis further sheds light on the importance of the earlier L2 achievement in the development of anxiety later on, by understanding learners' motivational profiles. That is, learners with high autonomous motivation and high achievement at Time 1 exhibited a further decrease in anxiety at Time 3. On the other hand, learners with high autonomous motivation but low achievement showed a higher level of anxiety at Time 3. Overall, the present study makes a noteworthy contribution to one of the most debatable issues in the field and concludes with some practical implications for L2 educators.
Full-text available
This study proposes and evaluates a general diagnostic classification model (DCM) for rating scales. We applied the proposed model to a dataset to compare its performance with traditional DCMs for polytomous items. We also conducted a simulation study based on the applied study condition in order to evaluate the parameter recovery of the proposed model. The findings suggest that the proposed model shows promise for (1) accommodating much smaller sample sizes by reducing a large number of parameters for estimation; (2) obtaining item category response probabilities and individual scores very similar to those from a traditional saturated model; and (3) providing general item information that is not available in traditional DCMs for polytomous items.
Full-text available
This paper reports an approximate replication of Matsuda & Gobel (2004) for the psychometric validation of the Foreign Language Reading Anxiety Scale (FLRAS). Their study examined the structural aspects of the FLRAS developed by Saito, Horwitz & Garza (1999). The results showed that the FLRAS measured three different subcomponents of foreign language reading anxiety; none of the factors predicted foreign language performance in content-based and four-skill classes. The present study aimed to reconfirm the psychometric validity of the FLRAS because it has been widely employed to make foreign language reading anxiety researchable. Our study retained the same methodology, with the exception of the measurements of classroom performance and reading proficiency. Matsuda & Gobel's conclusions were reproduced by showing a weak relationship between classroom performance and foreign language reading anxiety measured by the three-factor model of the FLRAS. However, this study newly demonstrated a strong association of reading-anxiety subcomponents with learners' reading proficiency. The administration, scoring, and interpretation methods of the FLRAS were reconsidered based on the replicated results.
Full-text available
More than three decades after their introduction, diagnostic classification models (DCM) do not seem to have been implemented in educational systems for the purposes they were devised. Most DCM research is either methodological for model development and refinement or retrofitting to existing nondiagnostic tests and, in the latter case, basically for model demonstration or constructs identification. DCMs have rarely been used to develop diagnostic assessments right from the start with the purpose of identifying individuals’ strengths and weaknesses (referred to as true applications in this study). In this article, we give an introduction to DCMs and their latest developments along with guidelines on how to proceed to employ DCMs to develop a diagnostic test or retrofit to a nondiagnostic assessment. Finally, we enumerate the reasons why we believe DCMs have not become fully operational in educational systems and suggest some advice to make their advent smooth and quick.
Full-text available
Second language (L2) anxiety has been the object of constant empirical and theoretical attention for several decades. As a matter of both theoretical and practical interest, much of the research in this domain has examined the relationship between anxiety and L2 achievement. The present study meta-analyzes this body of research. Following a comprehensive search, a sample of 97 reports were identified, contributing a total of 105 independent samples (N 5 19,933) from 23 countries. In the aggregate, the 216 effect sizes (i.e., correlations) reported in the primary studies yielded a mean of r 5 2.36 for the relationship between L2 anxiety and language achievement. Moderator analyses revealed effects sizes to vary across different types of language achievement measures, educational levels, target languages, and anxiety types. Overall, this study provides firm evidence for both the negative role of L2 anxiety in L2 learning and the moderating effects of a number of (non)linguistic variables. We discuss the findings in the context of theoretical and practical concerns, and we provide direction for future research. Anxiety is arguably the most researched affective variable in the field of second language acquisition (SLA). For more than half a century, SLA researchers have been conducting studies to explore second language (L2) learners' anxiety in relation to a host of linguistic and nonlinguistic variables (for reviews,
Full-text available
The purpose of this study is to develop and evaluate two diagnostic classification models (DCMs) for scoring ordinal item data. We first applied the proposed models to an operational dataset and compared their performance to an epitome of current polytomous DCMs in which the ordered data structure is ignored. Findings suggest that the much more parsimonious models that we proposed performed similarly to the current polytomous DCMs and offered useful item-level information in addition to option-level information. We then performed a small simulation study using the applied study condition and demonstrated that the proposed models can provide unbiased parameter estimates and correctly classify individuals. In practice, the proposed models can accommodate much smaller sample sizes than current polytomous DCMs and thus prove useful in many small-scale testing scenarios.
We present a review of second language researchers’ use of cluster analysis, an advanced statistical method still uncommon but increasingly used to identify groups or patterns in a dataset and to examine group differences. After describing key methodological considerations in conducting cluster analysis, we present a methodological synthesis of 65 studies published between 1989 and 2018 that employed cluster analysis. We specifically review the use of cluster analysis for themes of usage and reporting practices. Our findings indicate that hierarchical cluster analysis and K‐means cluster analysis were the most commonly used cluster methods, but the widespread use of these two methods tended not to be accompanied by sound reporting practices, particularly when justifying cluster solutions. In our analysis, we highlight concerns related to reporting and evaluation. For future use and to inform methodological practices in second language research, we briefly report on a sample study of cluster analysis that uses open data.
This meta‐analysis investigated the relationship between foreign language (FL) anxiety and FL performance. Fifty‐five independent samples with more than 10,000 participants were surveyed. It was found that the overall correlation between FL anxiety and FL performance was −.34 (p < .01). FL listening anxiety had the strongest correlation with FL listening performance. Both FL reading anxiety and test anxiety had a weaker correlation with FL performance as compared to other types of anxiety. The anxiety–performance correlation remained stable across groups with different FL proficiency levels, suggesting that the role of FL anxiety should not be ignored regardless of the FL learners’ proficiency level. As compared to language family, lexical similarity was found to have a more decisive modulating effect on the anxiety–performance correlation. However, language family and lexical similarity may interact to affect the anxiety–performance correlation. Finally, a metaregression analysis showed that age could affect the correlation between FL anxiety and performance.