You are on page 1of 14

Contemporary Educational Psychology 65 (2021) 101949

Contents lists available at ScienceDirect

Contemporary Educational Psychology


journal homepage: www.elsevier.com/locate/cedpsych

The relationship between math anxiety and math achievement in young


children is mediated through working memory, not by number sense, and it
is not direct☆
Monika Szczygieł
Institute of Psychology, Pedagogical University of Krakow, ul. Podchorążych 2, 30-084 Krakow, Poland

A R T I C L E I N F O A B S T R A C T

Keywords: Math anxiety is considered a predictor of math achievement, although the cognitive mechanism whereby math
Math anxiety anxiety impairs math achievement is unclear. The paper presents the results of cross-sectional (N = 241) and
Working memory longitudinal (N = 369) studies conducted among early school-aged children on the cognitive mechanism
Number sense
whereby math anxiety impairs math achievement. The following hypotheses were tested: (1) math anxiety
Math achievement
directly affects math achievement; (2) in accordance with processing efficiency and attentional cognitive theories,
math anxiety indirectly affects math achievement through working memory; (3) in accordance with the cognitive
deficit model, math anxiety indirectly affects math achievement through number sense. The results mostly
confirm the mediating role of working memory and undermine the mediating role of number sense and the direct
path in the relationship between math anxiety and math achievement. Because previous studies undertaken in
adults show the direct path from math anxiety to math achievement and the role of symbolic number processing
in explaining the relationship between the two, the methodological and developmental aspects of the obtained
results are discussed in the paper.

1. Introduction solving of mathematical problems in a wide variety of everyday life and


academic situations” (Richardson & Suinn, 1972, p. 551). Math anxiety
1.1. The relationship between math anxiety and math achievement is negatively related to math achievement (Zhang, Zhao, & Kong, 2019).
However, the debate regarding the nature of the relationship between
From both an individual and a society-wide perspective, having math anxiety and math achievement has been ongoing for some time.
adequate mathematical skills is fundamental in contemporary societies. Suárez-Pellicioni, Núñez-Peña, and Colomé (2016) presented three
Poor mathematical skills are associated not only with a low individual explanations for the relationship between math anxiety and math
social position but also with significant costs for the general public in the achievement. First, in accordance with the cognitive deficit model (Fer­
areas of social policy and the economy (Ansari, De Smedt, & Grabner, guson, Maloney, Fugelsang, & Risko, 2015; Maloney, Risko, Ansari, &
2012; Butterworth, Varma, & Laurillard, 2011; OECD, 2010; Parsons & Fugelsang, 2010; Maloney, Ansari, & Fugelsang, 2011), deficits in basic
Bynner, 2006). The quality of early math education is especially numerical processing skills lead to poor math achievement, which in
important for the development of further math skills in children. The turn leads to negative affect. Second, according to the attention control
acquisition of advanced mathematical knowledge and skills is made theory (Eysenck, Derakshan, Santos, & Calvo, 2007) and its predecessor,
possible when students master the basics of mathematics. Learning the processing efficiency theory (Eysenck & Calvo, 1992), the cognitive
mathematics requires regularity, which is why low achievement at an resources of math-anxious individuals may be consumed by worrisome
early stage of education may determine future low math performance. thoughts, ruminations, and distractors when solving mathematical
One of the correlates of students’ math achievement is math anxiety tasks; these reduce the level of math performance and may consequently
(Namkung, Peng, & Lin, 2019), which is defined as “feelings of tension increase math anxiety. The final explanation concerns negative experi­
and anxiety that interfere with the manipulation of numbers and the ences whilst learning mathematics that result in the avoidance of math;


I would like to thank all those involved in the project: researchers, school principals, teachers, parents, and pupils. This work was supported by the National
Science Centre (Poland; grant number 2015/19/N/HS6/00791).
E-mail address: monika.szczygiel@up.krakow.pl.

https://doi.org/10.1016/j.cedpsych.2021.101949

Available online 10 February 2021


0361-476X/© 2021 Elsevier Inc. All rights reserved.
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

math avoidance reduces math performance, which in turn results in the effect. Working memory is defined as a cognitive system that is
development of negative emotions (Hembree, 1990; Namkung et al., responsible for the storing and processing of information for a short
2019). The data confirming the validity of the model described by period of time (Baddeley & Hitch, 1974). It is considered both a pre­
Suárez-Pellicioni et al. (2016) were provided for adults by Skagerlund, dictor of math achievement (in young children: Jordan, Glutting, &
Östergren, Västfjäll, and Träff (2019). They observed an indirect path Ramineni, 2010; in older children: Sowinski et al., 2015; in adolescents:
through number sense, an indirect path through working memory, and a De Smedt, Taylor, Archibald, & Ansari, 2010; and in adults: Skagerlund
direct path from math anxiety to math achievement. et al., 2019) and a mediator in the relationship between math anxiety
and math achievement (in young and older children: Justicia-Galiano
1.2. The universality of the three-path model of math anxiety effect on et al., 2017; Soltanlou et al., 2019; and in adults: Skagerlund et al.,
math achievement 2019). Therefore, it can be assumed that negative thoughts associated
with mathematics can consume the cognitive resources of young chil­
Although the three-path model was confirmed in adults, some doubts dren, thus leading to a decrease in the level of mathematical perfor­
with its universality across the lifespan have been indicated (Lutten­ mance (Ashcraft & Kirk, 2001). This explanation is all the more credible
berger, Wimmer, & Paechter, 2018). The main problem concerns the because, at the initial stage of mathematical education, children have to
paucity of data on the mechanism that underlies the relationship be­ master many new mathematical facts, and this is associated with sig­
tween math anxiety and math achievement in young children. The nificant cognitive effort. However, in all likelihood, only one study
verification of this model in a group of young children is justified among children between the third and fifth grades (Justicia-Galiano
because it has been observed that early school-aged children feel some et al., 2017) has confirmed so far the mediating role of working memory
degree of math anxiety (Cargnelutti, Tomasetto, & Passolunghi, 2017; in the relationship between math anxiety and math achievement.
Ramirez, Chang, Maloney, Levine, & Beilock, 2016; Ramirez, Gunder­ Therefore, testing the indirect path from math anxiety through working
son, Levine, & Beilock, 2013) and this anxiety is negatively related to memory to math achievement among early school-aged children is
their math achievement (Ramirez et al., 2013; Vukovic, Kieffer, Bailey, justified.
& Harari, 2013; Wu, Barth, Amin, Melcarne, & Menon, 2012). It is likely The last path concerns the role of number sense in the relationship
that similar mechanisms underlie the relationship between math anxiety between math anxiety and math achievement. Number sense, defined as
and math achievement in both young and older children, but one may a fundamental ability to automatically and efficiently process numerical
wonder whether all three paths presented by Suárez-Pellicioni et al. magnitude information (Dehaene, 2011), is considered a predictor of
(2016) adequately describe the situation of young children. The hy­ math achievement (Chen & Li, 2014; Libertus, Feigenson, & Halberda,
pothesis that math anxiety directly affects math achievement is espe­ 2011) but not necessarily of math anxiety. Some researchers indicate
cially debatable. Although the other two explanations regarding the that a less precise representation of numerical magnitude is related to
mediating role of working memory and number sense are more likely, math anxiety (Douglas & LeFevre, 2017; Lindskog, Winman, & Poom,
there is limited evidence for their validity in children (Justicia-Galiano, 2017; Maloney et al., 2010, 2011; Skagerlund et al., 2019), but this
Martín-Puga, Linares, & Pelegrina, 2017). Arguments for and against the relationship was not observed by others (Braham & Libertus, 2018;
universality of the three-path model from childhood to adulthood are Dietrich, Huber, Moeller, & Klein, 2015). There is also a debate as to
presented below. whether math anxiety potentially determines nonverbal number skills or
The first explanation concerns the direct path between math anxiety vice versa. Lindskog et al. (2017) checked both possibilities and showed
and math achievement. Skagerlund et al. (2019) observed a direct path that the approximate number system predicts math anxiety, which in
from math anxiety to math achievement in adults, and they hypothe­ turn leads to low math performance. Furthermore, math outcomes
sized that it could be a result of avoiding math through many years of predict math anxiety, which in turn reduces approximate number sys­
education. They underlined that the avoidance effect hypothesis might tem acuity. Recently, Skagerlund et al. (2019) indicated that math
be strengthened if the direct path from math anxiety to math achieve­ anxiety hampers number processing, which in turn results in low math
ment in children was weak or the effect did not occur. It is likely that performance, but this is not the case in all types of mathematical tasks.
avoidance behaviors may adequately explain the mechanism of math Although this relationship can be two-sided (Herts, Beilock, & Levine,
anxiety effect on math achievement in people with many mathematical 2019), Skagerlund et al. (2019) showed that math anxiety indirectly
experiences (Skagerlund et al., 2019). Although children starting school affects math achievement through poor basic number-processing abili­
can differ in their math learning experiences and their levels of math ties. Researchers (Braham & Libertus, 2018; Maloney et al., 2011) sug­
anxiety (Ganley & McGraw, 2016; Sorvo et al., 2017), it is more credible gest that math anxiety diminishes math performance through less
that avoiding mathematics is not a common problem at this stage of precise representations of numerical magnitude, which means that
education (Ginsburg, 2006). Nevertheless, one should not dismiss the proper approximate number system acuity may serve as a protective
explanation that one year of schooling is enough for children to start factor for individuals with high math anxiety when they do math tasks.
avoiding mathematics, which in turn may explain the relationship be­ This hypothesis is probable. Therefore, it is worth ascertaining whether
tween their math anxiety and math achievement. Since the correlation good number sense in early school age reduces the harmful effects of
between math anxiety and math achievement may occur after a few math anxiety on math results.
months of schooling (Dowker, Cheriton, Horton, & Mark, 2019), and it
has not yet been determined how much time must pass before the 1.3. Current study
avoidance of mathematics contributes to poor math achievement and
deepens math anxiety, it is an open question whether the direct path The main objective of the current paper is to verify the universality of
from math anxiety to math achievement occurs in young children. If the the model that assumes the mediation effect of working memory be­
direct path does not occur or is weak in children, it will contribute to the tween math anxiety and math achievement, the mediation effect of the
validity of the hypothesis that direct path in adults results from math approximate number system on the relationship between math anxiety
avoidance (Skagerlund et al., 2019). If there is a moderate or strong and math achievement, and the direct path from math anxiety to math
direct path in children, it undermines the explanation of the effect achievement (Skagerlund et al., 2019; Suárez-Pellicioni et al., 2016).
through avoidance behavior. Therefore, there is a need to revise the While some studies among adults (Skagerlund et al., 2019), school-aged
hypothesis about the factors that explain the relationship between math children (Soltanlou et al., 2019) and young children (Justicia-Galiano
anxiety and math achievement in adults. et al., 2017) confirm the first hypothesis, the mediating role of number
The second explanation concerning the relationship between math sense is definitely more debatable because of the paucity of evidence in
anxiety and math achievement refers to the working memory mediation adults (Douglas & LeFevre, 2017; Skagerlund et al., 2019) and the lack

2
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

The MAQC was used for the whole group, whilst the mAMAS-E was used
only for the part of the group because it was only published and adapted
into Polish after the study had begun. Therefore, it was decided to use it
additionally among the children that had not yet been studied.
The mAMAS-E was developed on the basis of the mAMAS (Modified
Abbreviated Math Anxiety Scale; Carey, Hill, Devine, & Szűcs, 2017),
whose psychometric properties had been confirmed in a series of tests
(Szczygieł, 2019). The mAMAS-E, which is similar to the mAMAS, is a
nine-item scale and contains two subscales: testing (mAMAS-E testing)
and learning (mAMAS-E learning). However, when compared to the
mAMAS, the mAMAS-E differs in the response scale and the test pro­
cedure that was used for the adaptation of the scale to the capabilities of
Fig. 1. Hypothesized conceptual model of pathways between math anxiety, early school-aged children. A three-point response scale was used
working memory, number sense, and math achievement in early school- instead of a five-point Likert scale, and children’s math anxiety was
aged children. assessed in individual semi-structured interviews instead of in a group
procedure. To eliminate differences between the children in the level of
of evidence in children. Based on previous studies, it may be assumed their language skills, the instructions and statements were read aloud by
that math anxiety burdens working memory resources, thus impairing the researcher, and the children’s answers were marked on a response
math performance in children. In addition, it is hypothesized that math card by the interviewer. Before the testing session, the children were
anxiety undermines more basic number-processing abilities, which in given an example of how to answer further questions. The children were
turn reduces math achievement in children. A direct path from math asked: “Do you feel math anxiety in such situations? Yes, a little, or no?”,
anxiety to math achievement has been observed in adults, and it is likely and their responses were rated in the following way: no – 0 points, a
that this results from years of avoiding math activity. If so, then math little – 1 point, yes – 2 points. The children were given explanations of
anxiety’s direct effect on math achievement in young children should what anxiety means (“Anxiety means feelings of worry, nervousness, or
not be present or should be weak. However, as it is not known how long fear”). Therefore, the minimum and maximum scores in the mAMAS-E
it takes for children to develop a math-avoidance strategy, it is therefore ranged between 0 and 18 for the total score, 0–10 for the learning
justified to test this path in elementary school children. score, and 0–8 for the testing score. The higher the number of points, the
To verify these three hypotheses, multiple mediator models (see higher the level of math anxiety. The reliability of the mAMAS-E (N =
Fig. 1) were tested in two studies: a cross-sectional study among first- to 121) was checked with Cronbach’s α = 0.72 (total score), α = 0.67
third-grade children (study 1) and a longitudinal study among first- to (testing score), α = 0.58 (learning score). The relatively low reliability of
second-grade children (study 2). The two studies differed in methodo­ the learning subscale is due to the small number of questions, the short
logical details and were carried out to ensure that the results did not answer scale, and the very low variance in the children’s math anxiety
depend on the type of tasks or the research plan and sample used. assessments (most children reported a low level of math anxiety).
The MAQC (Szczygieł, 2020a) measures general math anxiety (MA-
2. Study 1 G) and was developed on the basis of two scales: the Mathematics
Anxiety Scale in Young Children (Harari, Vukovic, & Bailey, 2013) and
2.1. Method the Scale for Early Mathematics Anxiety (Wu et al., 2012). Both scales
were used in the preliminary study, but the results revealed many
2.1.1. Participants problems with measuring children’s emotions: the questions were not
The selection of schools for the study was random and included all appropriate to the language abilities of young children, the Polish
public elementary schools in Krakow (Poland). However, a given educational curriculum differs from the US curriculum, and the children
school’s decision to participate in the project depended upon its man­ had difficulties assessing their emotions on a four- or five-point response
agement. The cross-sectional research was conducted among early scale. Therefore, several changes were needed to reliably and accurately
school-aged children in nine public elementary schools in this city. Two measure math anxiety in young children in the Polish educational
hundred and forty-one students whose parents allowed them to partic­ context. Based on both scales, a 12-item unidimensional scale was
ipate in the project took part in the research: 46 children (23 girls) from created. Evidence on the convergent and divergent validity of the MAQC
the first grade (mean age: 7 years and 3 months, SD = 0.62, range = can be found in Szczygieł (2020a). The MAQC score mainly strongly
6.1–8.3 years), 101 children (60 girls) from the second grade (mean age: correlated with math anxiety measured with other math anxiety scales,
8 years and 2 months, SD = 0.55, range = 7.1–9.3 years), and 94 chil­ while it mainly moderately correlated with general and test anxiety.
dren (51 girls) from the third grade (mean age: 9 years and 4 months, SD Moreover, children with a high level of math self-esteem had a lower
= 0.69, range = 8.0–11.2 years). None of the students participating in MAQC score than children with medium math self-esteem, whilst chil­
the study had been diagnosed with a math learning disability. The so­ dren with medium and high self-esteem in Polish did not differ from
cioeconomic status of the children’s families was established based on each other in their math anxiety level. Similar to the mAMAS-E, the
the level of education, income, and occupation of both parents. To following features were employed: a simple instruction, a three-point
obtain one measure of the socioeconomic status, principal component answer scale (no – 0 points, a little – 1 point, yes – 2 points), and an
analysis was conducted and raw scores were converted to sten scores. individual semi-structured interview. Children were asked, e.g., “When
The families of most children had a moderate or high socioeconomic you are in a math class and the teacher says that you will be discussing a
status. The study was carried out at schools from the top, middle, and new topic, are you nervous? Yes, a little or no?” Minimum and maximum
bottom of the ranking of elementary schools in the city. The ranking was scores in the MAQC fall within the range of 0–24. The higher the score
based on the final test results (Dolna, 2016). on the MAQC, the higher the level of math anxiety. The internal reli­
ability of the scale in the present study is α = 0.72, N = 241.
2.1.2. Measures Working Memory Because prior studies had indicated that various
Math Anxiety Two scales were used to measure math anxiety among components of working memory played distinct roles in math perfor­
early school-aged children: the Modified Abbreviated Math Anxiety mance (Soltanlou, Pixner, & Nuerk, 2015; Soltanlou et al., 2019), two
Scale for Elementary Children (mAMAS-E; Szczygieł, 2019) and the backward span tasks were used to assess verbal and visuospatial working
Math Anxiety Questionnaire for Children (MAQC; Szczygieł, 2020a). memory efficiency: the WISC-R Backward Digit Span task (the Polish

3
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

language edition of the Wechsler Intelligence Scale for Children – allowed for the completion of the task was about 4 min. The indicator of
Revised; Matczak, Piotrowska, & Ciarkowska, 1991) and the Corsi approximate number system acuity used in the study was the average
Block-Tapping task (Kessels, van Zandvoort, Postma, Kappelle, & de correctness in the whole task. In the present study, Cronbach’s α index
Haan, 2000). for the task was 0.87, N = 232.
Verbal working memory (VWM) was measured using the WISC-R Digit Math Achievement Because no standardized mathematical tests
Span Backward task (DSB) according to the procedure described in the were available for use in Poland, the mathematical achievement of the
test manual. This task required the children to repeat a list of digits that children was measured with tasks prepared on the basis of the core
were spoken by the researcher. The list began with two elements; when curriculum for elementary schools and mathematical education mate­
the children successfully repeated them, more elements were added rials recommended by the Ministry of Education in Poland (MATH). No
until a maximum of eight was reached. Each span level included two dimensions of the constructed tasks were assumed. Children from the
lists. If the children incorrectly repeated two items consecutively, the first, second, and third grades received a varied set of mathematical
test session ended. The total result was calculated by adding up the tasks that tested different math skills to be mastered by children during
number of items correctly recalled for each list length. The minimum the school year. The mathematical tasks tested the following areas: first
and maximum scores are between 0 and 14 points. The split-half reli­ grade – knowledge of numbers, counting, addition and subtraction,
ability of the task was calculated with the Spearman-Brown formula and discovering rules, knowledge of money, knowledge of geometric figures,
is below the cut-off (r = 0.63, N = 239). reading a tape measure; second grade – addition and subtraction,
Visuospatial working memory (VSWM) was measured using the multiplication and division, reading a tape measure, spatial orientation,
manual version of the Corsi Block-Tapping task (CBT). The children discovering rules, clock reading; third grade – addition and subtraction,
were asked to recall the order of a sequence of blocks touched by the multiplication and division, reading a tape measure, discovering rules,
researcher, from last to first. The test session was preceded by a training clock reading, knowledge of dates and money. The children performed
session. The task consisted of a series from two to nine items. There were mathematical tasks on their own, but to eliminate the impact of reading
two sequences to perform in each series. On the successful repetition of skills on mathematical performance the researcher read the instructions
the series, the number of items increased until the children performed aloud. This was possible because each child had an individual meeting
two consecutively incorrect attempts at repeating the series. After that, with the researcher. The time allowed for the completion of the task
the testing session stopped. Each successful attempt was rewarded with depended on a child’s individual needs and was usually 20–40 min. The
one point, while an incorrect one scored no point. The minimum and maximum score that could be obtained in the tasks for each grade year
maximum score in the CBT is from 0 to 18 points. The split-half reli­ was 36. The psychometric properties of the mathematical tests were
ability of the CBT was calculated with the Spearman-Brown formula and established. All correlations between the mean score and math items
is borderline acceptable (r = 0.67, N = 233). Both tasks that measure were from weak to strong, positive, and significant (p < .05–0.001): first
working memory are progressive, which is why a relatively poor reli­ grade – r = 0.39–0.69; second grade – r = 0.29–0.71; third grade – r =
ability can be observed. 0.32–0.74. The internal consistency of the tests was also checked. The
Number Sense The non-symbolic comparison task (NSC) is a well- reliability of the mathematical achievement tasks was satisfactory: first
known and valid task that measures approximate number system acu­ grade α = 0.80, second grade α = 0.83, third grade α = 0.89. Finally, the
ity (Dietrich et al., 2015; Lafay, St-Pierre, & Macoir, 2019). The task was level of task difficulty in each grade year was checked. The average level
constructed as follows: Sets of magnitudes (white dots) were placed on a of difficulty for the mathematical task was similar in each class (0.77,
black background in each trial. Stimuli were generated in MATLAB in 0.71, 0.71 for the first, second, and third grades, respectively), and the
accordance with the guidelines of Gebuis and Reynvoet (2011). The size tasks may be referred to as “easy” or “moderately difficult” (Janowicz,
of the set was determined on the basis of the Weber fraction and took the 2017).
following rounded sizes: 0.125, 0.25, 0.33, 0.6, 1. The boards consisted
of 8, 10, 12, 13, 14, 18, 20, 22, 26, or 32 dots. Both the size of the set and 2.1.3. Procedure
the ratio of the number of sets were balanced in the task (the same The cross-sectional study was carried out in public schools after prior
number of sets of a given number, and the same number of sets for each approval had been obtained from the head of each school and the par­
ratio displayed on the screen). The trials differed in terms of the ents of the children. Ethical permission was obtained from the Scientific
coherence of the spatial cues with each number. For a coherent trial, Research Ethics Committee of the Institute of Psychology, Jagiellonian
larger and more dots were displayed on one board, whilst smaller and University of Krakow. The described study is part of a larger research
fewer dots were displayed on the other board. In this way, the selection project in which the individual and environmental predictors of math
of a larger set is easier as larger dots mean a larger magnitude set. An achievement of early school-aged children were tested (see Szczygieł,
incoherent trial was more difficult. An incoherent trial occurred when a 2020b, 2020c). The study was conducted in 2017 and consisted of two
set of larger and fewer dots was compared with a set of more numerous individual meetings with children in April/ May (first test session) and
and smaller dots. Both in the training and in the test session, the boards May/June (second test session). The data were collected by four re­
presented both the coherent and the incoherent conditions equally, in searchers who had previously been trained in the research procedure.
accordance with the recommendations of Nuerk, Weger, and Willmes Before starting the study, the children were asked to agree to participate
(2004). In total, 30 trials in the coherent condition and 30 in the inco­ in it; they were informed that they could ask questions and refuse to
herent condition were constructed, with each trial being displayed perform a task or participate in the meetings at any time. During the first
twice. The order of trials was random. The task was presented on a test session, the children performed tasks that measured their verbal and
computer screen with DMDX software (Forster & Forster, 2003). The visuospatial working memory resources and approximate number sys­
children were asked to select the more numerous set of dots from two tem acuity; also, they were asked about their level of math anxiety.
sets. The elements of the set were different-sized dots. The children During the second test session, the children were asked to perform
selected the larger number of dots using the buttons marked on the mathematical tasks. The average interval between the two sessions was
keyboard (e.g. ‘A’ on the left and ‘L’ on the right). If the set on the left 47 days. The first and second meeting with the children lasted for an
side of the screen was more numerous, the children were expected to average of 25 min (15–45 min) and 35 min (10–45 min), respectively.
press the ‘A’ key; if there was a more numerous set on the right side of All tasks instructions were read aloud by the researcher.
the screen, the children were expected to press the ‘L’ key. The stimulus
display time was 7 s, after which the board disappeared and the next one 2.1.4. Analysis
appeared in its place. The task consisted of 4 training trials and 120 trials The children solved various mathematical tasks, but the level of
in the test session, with a break in the middle of the task. The time difficulty of the test was similar in each grade. A one-way analysis of

4
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

Table 1
Descriptive Statistics and Pearson Correlations between Variables in Study 1.
M (SD) 1 2 3 4 5 6

1 ANS 0.81 (0.09)


2 VSWM 6.05 (1.63) 0.12
3 VWM 4.08 (1.32) 0.10 0.26***
4 MA-G 4.27 (3.63) − 0.10 − 0.17** − 0.15*
5 MA-T 1.80 (1.93) 0.01 − 0.16 − 0.22* 0.53***
6 MA-L 0.69 (1.28) − 0.24* − 0.18* − 0.24* 0.53*** 0.48***
7 MATH 26.28 (6.21) 0.28*** 0.27*** 0.29*** − 0.36*** − 0.37*** − 0.33***

Note: * p < .05, ** p < .01, *** p < .001; N = 223 (correlations with MA-T and MA-L was calculated based on N = 114). Correlations for math achievements were
calculated using z-scores for each grade level. ANS – approximate number system (the Non-Symbolic Comparison task), VSWM – visuospatial working memory (the
Corsie Block Tapping task), VWM – verbal working memory (WISC-R Digit Span Backward task), MA-G – general math anxiety (MAQC), MA-T – testing math anxiety
(mAMAS-E testing), MA-L – learning math anxiety (mAMAS-E learning), MATH – math achievement.

Table 2
Direct and Indirect Effect of Math Anxiety on Mathematical Performance in Study 1.
Standardized coefficient Model fit

Model Predictor Mediator Output Indirect Total N χ2 CFI RMSEA [90%] SRMR

1 MA WM MATH − 0.19 − 0.53*** 114 χ2(5) = 7.51; p = .19 0.97 0.07 [0.00− 0.16] 0.04
ANS − 0.02
2 MA-G WM MATH − 0.13* − 0.36*** 223 χ2(2) = 0.43; p = .81 1.00 0.00 [0.00− 0.08] 0.01
ANS − 0.02

Note: * p < .05, ** p < .01, *** p < .001. MA – math anxiety is a latent variable that consists of testing (mAMAS-E testing) and learning math anxiety (mAMAS-E
learning), MA-G – general math anxiety (MAQC), WM – working memory is a latent variable that consists of visuospatial (the Corsie Block Tapping task) and verbal
(WISC-R Digit Span Backward task) working memory, ANS – approximate number system (the Non-Symbolic Comparison task), MATH – math achievement.

variance was performed to test whether grade level affects the math
result. The grade effect was not observed, F(2,237) = 1.76; p = .17; this
enabled the inclusion, in the analysis, of one indicator of math
achievement in early school-aged children. The descriptive statistics and
correlation analysis were prepared in PS IMAGO PRO 5.1, and the
mediation analysis was tested in the lavaan and mediation packages for
the R environment (Rosseel, 2012; Tingley, Yamamoto, Hirose, Keele, &
Imai, 2014). The effect size for Pearson’s correlation was provided based
on Evans’ proposal (1996): r < 0.20 very weak, 0.20–0.39 weak,
0.40–0.59 moderate, 0.60–0.79 strong, >0.80 very strong correlation.
The model fit indices cut-off values were established in accordance with
Hu and Bentler (1999): CFI – Comparative Fit Index (>0.90), SRMR –
Standardized Root Mean Square Residual (<0.08), RMSEA – Root Mean
Square Error of Approximation (<0.06). The Maximum Likelihood
Fig. 2. Tested multi-mediation model for the relationship between math anx­
estimator was used in all mediation analyses. Listwise deletion was
iety and math performance in the first to third grade children (study 1). MA –
applied in correlation analyses and mediation models. Listwise deletion
math anxiety is a latent variable that consists of testing (mAMAS-E testing) and
was applied because math anxiety was measured with the MAQC in the
learning math anxiety (mAMAS-E learning), WM – working memory is a latent
whole group and additionally with the mAMAS-E in around 1/2 group. variable that consists of visuospatial (the Corsie Block Tapping task) and verbal
Therefore, there were mostly no single gaps in the data, but there was no (WISC-R Digit Span Backward task) working memory, ANS – approximate
data for around 50% of the children in the mAMAS-E. Since data were number system (the Non-Symbolic Comparison task), MATH – math achieve­
missing completely at random (Rubin, 1976): Little’s MCAR test: χ 2(23) = ment. * p < .05, ** p < .01, *** p < .001.
24.83, p = .34, listwise deletion is suitable (Soley & Bori, 2013). To
maintain the consistency of data analyses, correlations are presented In the second step of the analysis, the multi-mediation effects of
with the MAQC (N = 223) and with the mAMAS-E (N = 114) using working memory and the approximate number system in the relation­
listwise deletion separately for each dataset. ship between math anxiety and math achievement were tested (see
Table 2). Math anxiety was measured with two tasks: the mAMAS-E and
3. Results the MAQC. Since the first questionnaire was only used amongst half of
the students in the study, it was decided to build two separate models.
The descriptive statistics and Pearson’s correlation coefficient for all Both models tested working memory and the approximate number
variables are presented in Table 1. Math achievement correlates posi­ system as mediators between math anxiety and math achievement. The
tively and weakly with the approximate number system and with verbal first model assumed that the mAMAS-E was a latent variable consisting
and visuospatial working memory; it correlates negatively and weakly of two observable factors: testing and learning. The second model
with all measures of math anxiety. Math anxiety correlates negatively included the MAQC as an observable variable. In both models, working
and weakly with the approximate number system only when learning memory (WM) was a latent variable consisting of two observable vari­
math anxiety is considered. Almost all indicators of math anxiety ables: verbal (VWM) and visuospatial (VSWM) working memory. The
negatively and weakly correlate with verbal and visuospatial working approximate number system (ANS) and math achievement (MATH)
memory. were treated as observable variables.

5
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

second study were different from those in study 1. The study involved
children who had received their parents’ consent to participate in the
project (N = 369, 181 girls). The children attended 28 classes across
twelve elementary schools. The full measurement was made four times:
at the beginning (N = 348), in the middle (N = 337), and at the end of
the first grade (N = 317), and at the end of the second grade (N = 263).
The number of the children who participated in subsequent measure­
ments decreased because some of them had changed school, some were
absent, and some withdrew from participation in the study. The mean
age of the children during the first measurement was 7 years 2 months
(SD = 0.42; range = 6.0–8.3), during the second measurement it was 7
years 5 months (SD = 0.43; range 6.3–8.6), and at the end of the first
grade it was 7 years 8 months (SD = 0.41; range = 6.6–8.7). The children
were one year older when the last measurement was conducted. Only
those children who did not have a diagnosis of mathematics learning
Fig. 3. Tested multi-mediation model for the relationship between math anx­
iety and math performance in the first to third grade children (study 1). MA – disability participated in the study. The children were from families with
general math anxiety (MAQC), WM – working memory is a latent variable that a medium or a high socioeconomic status. Similar to the first study, the
consists of visuospatial (the Corsie Block Tapping task) and verbal (WISC-R indicator for the children’s families’ socioeconomic status was ascer­
Digit Span Backward task) working memory, ANS – approximate number sys­ tained on the basis of parental education, income, and occupation. The
tem (the Non-Symbolic Comparison task), MATH – math achievement. * p < principal component analysis was conducted and raw scores were con­
.05, ** p < .01, *** p < .001. verted to sten scores. The schools used in the study were at the top,
middle, or bottom of the ranking of elementary schools in the city based
Both tested models (see Fig. 2, Fig. 3, and Table 2) obtained the on the final test results (Dolna, 2016).
required χ 2 values and the model fit indices were very good (see Table 2,
Hu & Bentler, 1999). In the first model, the following were observed: a 4.2. Measures
nonsignificant path from math anxiety through working memory to
math achievement (β = − 0.19, p = .09), no mediation effect of the 4.2.1. Math anxiety
approximate number system (β = − 0.02, p = .33), and a negative and As with the cross-sectional study, the mAMAS-E (Szczygieł, 2019)
weak direct path from math anxiety to math achievement (β = − 0.32, p was used to assess the level of math anxiety of the children across the
= .04). A direct path and two indirect paths explained 42% of the four measurement points. Internal consistency for the total score
variance in math achievement (R2 = .42). In the second model, the checked with Cronbach’s α: at the beginning (N = 347), in the middle (N
negative and weak mediation effect of working memory was observed = 333), at the end of first (N = 309), and second grade (N = 262) was
(β = − 0.13, p = .03), with no mediation of approximate number sense (β 0.78, 0.71, 0.73, and 0.76, respectively. Cronbach’s α established for the
= − 0.02, p = .22). In this model, a significant, negative, and weak direct learning scale was 0.69, 0.54, 0.59, 0.57; for the testing scale it was 0.72,
path from math anxiety to math achievement was observed (β = − 0.22, 0.69, 0.69, 0.73. One of the reasons for the low reliability of the learning
p < .01). Math anxiety, working memory, and the approximate number subscale is the very low variance in the results among young children
system explained 32% of the variance in math achievement (R2 = .32). (most children reported weak math anxiety).
Therefore, the results of the first study are inconclusive and indicate (1)
a negative and weak direct path from math anxiety to math achieve­ 4.2.2. Working memory
ment, (2) negative and weak mediating roles of working memory or the Verbal and visuospatial working memory was measured in the same
lack of mediation, and (3) no mediation effect of number sense. way as in the cross-sectional study. As with the first study, two trials
were used for each list length. The split-half reliability of the Backward
4. Study 2 Digits Span task calculated with the Spearman-Brown formula is r =
0.61 (N = 345) and the split-half reliability of the Corsi Block Tapping
The first study had some methodological limitations and its results task is r = 0.69 (N = 346). Task reliability is similar to the first study
are inconsistent in some points. First, the p values of the results of the because there were only two attempts and the task had an increasing
working memory mediation effect and direct path varied from <0.01 to level of difficulty.
0.09, which increases the type I error and makes the results inconclu­
sive. Second, the first model was built on a relatively small number of 4.2.3. Number sense
observations. Third, in both models, the dependent variable was calcu­ In the longitudinal study, the Numeracy Screener was used (Nos­
lated as one indicator to include the results of the children from first to worthy, Bugden, Archibald, Evans, & Ansari, 2013). This is a pencil and
third grade in one model. It is likely, however, that there are differences paper test that measures symbolic and non-symbolic numerical magni­
between the children from particular grades in their levels of math tude processing. Detailed information on the construction of the test (e.
anxiety, working memory efficiency, and number sense, although there g. order, ratio, density of stimuli) can be found in the authors’ work
were no differences in the average level of mathematical tasks. There­ (Nosworthy et al., 2013). During the magnitude comparison task, the
fore, the second study was conducted without these limitations and an children compared pairs of magnitudes (ranged 1–9) in symbolic (56
analogous model was tested in the longitudinal plan. pairs of digits) or non-symbolic (56 dot arrays pairs) forms and chose
which one in each pair was larger. The children in both conditions were
to make a comparison within one minute. Both test sessions were pre­
4.1. Method ceded by training (three attempts with the researcher and nine practice
items performed by the children themselves). Half of the children
4.1.1. Participants received the symbolic comparison task first and then the non-symbolic
As with the first study, schools were randomly selected for the study version; the other half received the tasks in reverse order. In both con­
from amongst all the schools in the city (Krakow, Poland), but it was up ditions, the task score that could be obtained ranged from 0 to 56 points.
to the individual schools’ management and parents to give consent to Internal consistency was not ascertained because of the time-limited test
children’s participation in the project. The children participating in the and the zero variance in some items.

6
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

Table 3
Descriptive Statistics and Pearson Correlations between Variables in Study 2.
M (SD) 1 2 3 4 5 6 7 8 9 10 11

1 ANS T1 31.14 (7.16)


2 VSWM T1 5.24 (1.70) 0.13*
3 VWM T1 3.41 (1.16) 0.12* 0.35***
4 MA-T T2 2.05 (2.11) − 0.10 − 0.02 − 0.17**
5 MA-L T2 0.78 (1.33) − 0.05 − 0.13* − 0.09 0.49***
6 MA-T T3 1.93 (2.06) − 0.02 − 0.05 − 0.14* 0.32*** 0.32***
7 MA-L T3 0.84 (1.44) 0.01 − 0.08 − 0.04 0.55*** 0.55*** 0.53***
8 MA-T T4 2.50 (2.22) − 0.12 − 0.17* − 0.17* 0.37*** 0.29*** 0.50*** 0.50***
9 MA-L T4 1.26 (1.70) − 10 − 22*** − 0.15* 0.26*** 0.37*** 0.38*** 0.33*** 0.50***
10 MATH-T2 48.21 (6.77) 0.15* 0.41*** 0.42*** − 0.08 − 0.15* − 0.20** − 0.14* − 0.27*** − 0.25***
11 MATH-T3 47.12 (7.68) 0.17* 0.38*** 0.38*** − 0.05 − 0.14* − 0.15* − 0.07 − 0.26*** − 0.27*** 0.70***
12 MATH-T4 42.50 (9.92) 0.10 0.29*** 0.40*** − 0.04 − 0.20** − 0.17* − 0.09 − 0.24*** − 0.26*** 0.63*** 0.66***

Note: *p < .05, ** p < .01, *** p < .001; N = 239–348. ANS – approximate number system (the Non-Symbolic Comparison task), VSWM – visuospatial working memory
(the Corsie Block Tapping task), VWM – verbal working memory (WISC-R Digit Span Backward task), MA-T – testing math anxiety (mAMAS-E testing), MA-L – learning
math anxiety (mAMAS-E learning), MATH – math achievement. T1 – the beginning of the first grade, T2 – the middle of the first grade, T3 – the end of the first grade, T4
– the end of the second grade.

4.2.4. Math achievement questions, refuse to perform a task, or withdraw from the meetings at
As with the first study, the mathematical tasks in each measurement any time. For all tasks, the researcher read the instructions aloud to
were prepared in accordance with the mathematical core curriculum for compensate for differences in the level of the children’s reading skills.
elementary schools and with materials recommended by the Ministry of The average length of the meeting was dependent on the research stage
Education in Poland. The tests in the longitudinal study examined the and individual needs of each child. The first meeting lasted about 30–40
practical application of mathematical knowledge in the following areas: min, during which verbal and visuospatial working memory resources
the middle of the first grade (MATH-T2) – knowledge of numbers, and approximate number system acuity were measured. Subsequent
counting, discovering rules, knowledge of geometric figures, spatial meetings, during which the level of math anxiety and math achievement
orientation; the end of the first grade (MATH-T3) – knowledge of were measured, took approximately 20–30 min per child.
numbers, counting, addition and subtraction, discovering rules,
knowledge of money, knowledge of geometric figures, reading a tape 4.4. Analysis
measure; the end of the second grade (MATH-T4) – addition and sub­
traction, multiplication and division, reading a tape measure, spatial Similar to the first study, the descriptive statistics and correlation
orientation, discovering rules, clock reading. The procedure for testing analysis were calculated in PS IMAGO PRO 5.1; the mediation models
math achievement in the longitudinal study was the same as in the first were tested in the lavaan and mediation packages in the R environment
study. Each time, the children were given mathematical tasks in written (Rosseel, 2012; Tingley et al., 2014). Since “number sense” refers to non-
form, although task instructions were read aloud by the researcher. The symbolic numerical magnitude processing, the non-symbolic part of the
tasks were solved by the children themselves. There was no time limit to Numeracy Screener was included in the analysis and is presented in the
complete the tasks, but most children finished the test in 20–40 min. The results section. However, because some researchers understand basic
point score that could be obtained in each measurement ranged from numerical skills as number processing, additional analysis was con­
0 to 62. The psychometric properties of the tests were established. The ducted (including the symbolic part of the task), the results of which are
internal reliability tested with Cronbach’s α is 0.85 (MATH-T2), 0.86 presented in the Appendix. Similar to the first study, the effect size of
(MATH-T3), and 0.91 (MATH-T4). Correlations between mathematical Pearson’s correlation was assessed in accordance with Evans (1996),
tasks and the total score at each measurement point were significant (p and model fit indices were rated according to Hu and Bentler (1999).
< .05–0.001), positive, and from weak to strong (T2 r = 0.15–0.51; T3 r Moreover, as with the first study, the Maximum Likelihood estimator
= 0.13–0.61; T4 r = 0.18–0.61). The average level of difficulty of the was used in all mediation analyses. Pairwise deletion of missing data
tasks in the three measurements was 0.78, 0.76, 0.69, respectively. was used in the correlation analyses and listwise deletion of missing data
Therefore, the test was easy in the first grade and moderately difficult in was applied in mediation models. Mediation models were tested three
the second grade (Janowicz, 2017). times. However, some children dropped out of the study. Therefore,
there were children who did not participate in all measurements. Similar
4.3. Procedure to the first study, missing data also were random (Little’s MCAR test:
χ 2(119) = 141.39, p = .08) and listwise deletion was justified (Soley & Bori,
The data from the longitudinal study was collected in four meetings 2013). The correlation matrix presents results from four-time points. So
with the children. The meetings were conducted at the beginning as not to substantially limit the number of observations, it was decided
(September/October 2017), in the middle (February/March 2018), and to show the results of pairwise deletion.
at the end (May/June 2018) of the first grade, and at the end of the
second grade (May/June 2019). The presented data are part of a larger 4.5. Results
research project concerning individual and environmental predictors of
math achievement in early school-aged children (see Szczygieł, 2020d, First of all, the descriptive statistics and Pearson’s correlation co­
Szczygieł & Pieronkiewicz, accepted). The study was carried out in the efficients between measured variables at four time points were calcu­
schools after prior approval had been given by the head of each school lated (see Table 3). Math achievement was positively but very weakly
and the parents of the children. Ethical permission was obtained from related to the non-symbolic comparison task (except the last measure­
the Scientific Research Ethics Committee of the Institute of Psychology, ment). Both verbal and visuospatial working memory were positively
Jagiellonian University of Krakow. Similar to the cross-sectional study, and weakly/moderately related to math achievement in the first and
the children were tested individually by a researcher trained in the second grades. Generally, negative and very weak or weak correlations
research procedure. Before each meeting, the children were asked to were observed between math achievement and math anxiety, but the
agree to participate in it and were informed that they could ask existence of this relationship and its strength depended on the math

7
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

Fig. 4. Tested multi-mediation model for the relationship between math anx­ Fig. 6. Tested multi-mediation model for the relationship between math anx­
iety and math performance in the middle of the first grade (study 2). MA (T2) – iety and math performance at the end of the second grade (study 2). MA (T4) –
math anxiety is a latent variable that consists of testing (mAMAS-E testing) and math anxiety is a latent variable that consists of testing (mAMAS-E testing) and
learning math anxiety (mAMAS-E learning), WM (T1) – working memory is a learning math anxiety (mAMAS-E learning), WM (T1) – working memory is a
latent variable that consists of visuospatial (the Corsie Block Tapping task) and latent variable that consists of visuospatial (the Corsie Block Tapping task) and
verbal (WISC-R Digit Span Backward task) working memory, ANS (T1) – verbal (WISC-R Digit Span Backward task) working memory, ANS (T1) –
approximate number system (the Non-Symbolic Comparison task), MATH (T2) approximate number system (the Non-Symbolic Comparison task), MATH (T4)
– math achievement. T1 – the beginning of the first class, T2 – the middle of the – math achievement. T1 – the beginning of the first class, T4 – the end of the
first class. * p < .05, ** p < .01, *** p < .001. four class. * p < .05, ** p < .01, *** p < .001.

due to the assumption that the cognitive skills of children do not rapidly
change over two years. Math anxiety and math achievement were tested
three times to check whether there were changes in the results
depending on the grade year. Therefore, three models were tested to
check the mediating role of working memory and the approximate
number system in the relationship between math anxiety and math
achievement.
All models (see Figs. 4–6 and Table 4) are characterized by a good χ 2
value and a very good model fit (see details in Table 4, Hu & Bentler,
1999). The results indicate that working memory negatively, weakly,
and fully mediated the relationship between math anxiety and math
achievement in the middle of the first grade (β = − 0.20, p < .05) and at
the end of the second grade (β = − 0.27, p < .01) but not at the end of the
first grade (β = − 0.10, p = .06). The approximate number system did not
mediate the relationship between math anxiety and math achievement
at any time (the middle of the first grade β = − 0.01, p = .83, the end of
the first grade β = − 0.01, p = .96, the end of the second grade β = − 0.01,
Fig. 5. Tested multi-mediation model for the relationship between math anx­
p = .43) and a direct path from math anxiety to math achievement was
iety and math performance at the end of the first grade (study 2). MA (T3) – not found at any time (the middle of the first grade β = 0.01, p = .99, the
math anxiety is a mean score of testing (mAMAS-E testing) and learning math end of the first grade β = − 0.02, p = .76, the end of the second grade β =
anxiety (mAMAS-E learning), WM (T1) – working memory is a latent variable − 0.13, p = .20). The variance in math achievement was explained by
that consists of visuospatial (the Corsie Block Tapping task) and verbal (WISC-R direct and indirect paths in over 40% (R2 = 0.48 in the middle of the
Digit Span Backward task) working memory, ANS (T1) – approximate number first grade, R2 = 0.43 at the end of the first grade, R2 = 0.42 at the end
system (the Non-Symbolic Comparison task), MATH (T3) – math achievement. of the second grade).
T1 – the beginning of the first class, T3 – the end of the first class. * p < .05, ** p
< .01, *** p < .001.
5. Discussion

anxiety type and the measurement point. Math anxiety negatively and
In previous studies in adults, adolescents, and children, a relation­
very weakly correlated with verbal and visuospatial working memory,
ship between math anxiety and math achievement had been observed,
but the occurrence of this relationship depended on measurement time.
but only a limited number of studies had tested the mechanism under­
Math anxiety generally did not correlate with the approximate number
lying this relationship. The main objective of the present studies was to
system.
determine the cognitive mechanism whereby math anxiety impairs math
Next, the multi-mediator models were tested. In the first and third
achievement in early school-aged children. In both the cross-sectional
models, math anxiety (MA) was defined by two observable variables:
and the longitudinal study, the direct path from math anxiety to math
testing and learning. In the second model, the mAMAS-E summary index
achievement, the mediation effect of working memory, and the
(MA-E) was taken into account due to negative variances and the need to
approximate number system were tested. In five out of eight tested
reduce the complexity of the model. In all models, working memory
models presented in the paper and in the Appendix, the mediating role of
(WM) was a latent variable and included two variables: verbal (VWM)
working memory between math anxiety and math achievement was
and visuospatial (VSWM) working memory. The approximate number
confirmed. The role of non-verbal number sense was not demonstrated
system (ANS) and math achievement (MATH) were included in the
in any model. The occurrence of the direct path was observed in two out
models as observable variables. Working memory and non-symbolic
of eight models (in the first study but not in the second study).
number sense were tested once (at the beginning of the first grade)
Therefore, the main observation that working memory is a mediator

8
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

Table 4
Direct and Indirect Effect of Math Anxiety on Math Achievements in Study 2.
Standardized coefficient Model fit

Model Predictor Mediator Output Indirect Total N χ2 CFI RMSEA [90%] SRMR

1 MA (T2) WM (T1) MATH-T2 − 0.20* − 0.19* 325 χ2(5) = 11.44; p = .04 0.97 0.06 [0.01− 0.11] 0.03
ANS (T1) − 0.01
2 MA-E (T3) WM (T1) MATH-T3 − 0.10 − 0.12* 298 χ2(2) = 0.56; p = .76 1.00 0.00 [0.00− 0.08] 0.01
ANS (T1) − 0.01
3 MA (T4) WM (T1) MATH-T4 − 0.27** − 0.39*** 255 χ2(5) = 2.28; p = .81 1.00 0.00 [0.00− 0.05] 0.02
ANS (T1) − 0.01

Note: * p < .05, ** p < .01, *** p < .001.


MA – math anxiety is a latent variable that consists of testing (mAMAS-E testing) and learning math anxiety (mAMAS-E learning), MA-E – mean score of testing
(mAMAS-E testing) and learning (mAMAS-E learning) math anxiety, WM – working memory is a latent variable that consists of visuospatial (the Corsie Block Tapping
task) and verbal (WISC-R Digit Span Backward task) working memory, ANS – approximate number system (the Non-Symbolic Comparison task), MATH – math
achievement. T1 – the beginning of the first grade, T2 – the middle of the first grade, T3 – the end of the first grade, T4 – the end of the second grade.

in the relationship between math anxiety and math achievement in Soltanlou et al., 2019). The results of the current studies indicate that, in
young children is in accordance with the models of Suárez-Pellicioni early childhood, appropriate interventions are already needed to pre­
et al. (2016) and Skagerlund et al. (2019). However, the universality of vent the emergence and development of math anxiety and to reduce the
both models is challenged by the fact that there is mostly no direct path high level of math anxiety among some young students. Although there
from math anxiety to math achievement and there is no indirect path is some controversy regarding the long-term efficacy of working mem­
from math anxiety to math achievement through number sense. The ory training (Melby-Lervåg & Hulme, 2013), the results partially explain
results indicate that the three-path model suits adults well but not young why short-term interventions that consist in relieving working memory
children. resources may reduce math anxiety and increase task performance
First, the lack of a direct path in the tested children in the second (Park, Ramirez, & Beilock, 2014; Ramirez & Beilock, 2011).
study strengthens the hypothesis presented by Skagerlund et al. (2019) Third, as opposed to Skagerlund et al. (2019), number sense turned
that the direct effect in adults may stem from avoiding mathematics in out not to mediate the relationship between math anxiety and math
education. The tested children had had math classes for several months achievement in young children. Regardless of whether the models with
up to a maximum of several years, and their levels of math anxiety were non-symbolic (study 1 and study 2) or symbolic (study 2, Appendix)
relatively low. Math education in the first, second, and third grades is number sense were tested, the results were the same. Therefore, the
often combined with fun, which demonstrates that math can be a obtained results challenge assumptions about the role of deficits in nu­
pleasant activity that children engage in rather than avoid. Indeed, merical magnitude information processing as a factor that contributes to
previous works indicated that children like mathematics and math is a the negative effect of math anxiety on math performance. The results
natural part of young children’s activity (Ginsburg, 2006; Dowker et al., may be explained by methodological differences between young chil­
2019). However, it is important to test an analogous model of young dren and adults and their different educational situations. Math anxiety
children, older children, and adolescents in the future. It may turn out is associated with the use of mathematical symbols, while Skagerlund
that it is wrong to think that the negative effects of avoiding mathe­ et al. (2019) used symbolic comparison tasks to measure number sense
matics can only be observed after many years of education. It is possible efficiency. The correlation between the math anxiety and the symbolic
that this strategy could develop within just a few years. Such hypothesis comparison task score may therefore result from the fact that people
is strengthened by the results from the first study that reveals a weak and have high levels of math anxiety because they are not good at symbol
negative direct path from math anxiety to math achievement. The dif­ operations. Indeed, this relationship may stem from the fact that
ferences between both studies may be explained by the methodological learners avoid math lessons for many years of mathematics education.
differences, especially the grade year of the tested children. The analyses This trend cannot yet be observed in young children. Developmental
in the first study were conducted in children older than those tested in factors are important when considering differences between the situa­
the second study: the former involved children from the end of the first, tion of children and that of adults because the acuity of number sense
second, and third grade, whereas the latter tested children in the middle (Nys et al., 2013; Park & Brannon, 2013; Park et al., 2014) and the level
and at the end of the first grade and at the end of the second grade. of math anxiety (Hembree, 1990; Wang, Oh, Malanchini, & Borriello,
Second, the results are in accordance with studies showing the 2020) may increase during math education. Additionally, previous
importance of cognitive resources as a mediator between math anxiety studies revealed that symbolic and non-symbolic magnitudes are coded
and math achievement in younger and older children (Justicia-Galiano differently and are not represented in an abstract and modality-
et al., 2017; Soltanlou et al., 2019) and in adults (Skagerlund et al., independent manner. Therefore, symbolic and non-symbolic compari­
2019). The obtained results mostly indicate that the negative effect of son tasks should not be used interchangeably (Bulthe, De Smedt, & Op
math anxiety on working memory and consequently on the performance de Beeck, 2014; Dietrich et al., 2015; Lyons, Nuerk, & Ansari, 2015). The
of mathematical tasks is universal across the lifespan. Even a low last problem is the ecological validity of the math anxiety scale used by
average level of math anxiety manifested in a group of young children Skagerlund et al. (2019), which covers many situations unsuited to
weakens their cognitive resources. According to the processing efficiency adults, e.g. being given a surprise math test in a class or sitting in a math
theory and the attentional cognitive theory, the cognitive resources of class. Further research is required to determine what underlies the dif­
math-anxious individuals are burdened by distractors, worrisome ferences in results that show the existence (or not) of the role of number
thoughts, and ruminations while solving math problems; this diminishes sense. Nevertheless, the obtained results do not allow us to conclude that
the level of math performance and consequently strengthens math good approximate number system acuity is a factor that protects against
anxiety (Ashcraft & Kirk, 2001; Eysenck & Calvo, 1992; Eysenck et al., the harmful effects of math anxiety on math achievement among early
2007; Ramirez et al., 2013). Working memory is one of the most school-aged children.
examined mediators of anxiety/math anxiety and math performance, Therefore, the universal validity of the cognitive deficit model (Malo­
and it undoubtedly plays an important role in explaining the mechanism ney et al., 2011; Ferguson et al., 2015) has to be challenged. Admittedly,
of math anxiety effect (Justicia-Galiano et al., 2017; Ng & Lee, 2015; supporters of this model assume that math anxiety results from a basic

9
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

Table 5
Pearson Correlation between Symbolic Number Processing and Other Variables in Study 2.
1 2 3 4 5 6 7 8 9 10 11 12
ANS T1 VSWM T1 VWM T1 MA-T T2 MA-L T2 MA-T T3 MA-L T3 MA-T T4 MA-L T4 MATH-T2 MATH-T3 MATH-T4

NS-S T1 0.80*** 0.17** 0.22*** − 0.10 − 0.08 − 10 0.00 − 0.14* − 0.14* 0.24*** 0.25*** 0.22***

Note: *p < .05, ** p < .01, *** p < .001; N = 239–369. NS-S – symbolic number sense (the Symbolic Comparison task), ANS – approximate number system (the Non-
Symbolic Comparison task), VSWM – visuospatial working memory (the Corsie Block Tapping task), VWM – verbal working memory (WISC-R Digit Span Backward
task), MA-T – testing math anxiety (mAMAS-E testing), MA-L – learning math anxiety (mAMAS-E learning), MATH – math achievement. T1 – the beginning of the first
grade, T2 – the middle of the first grade, T3 – the end of the first grade, T4 – the end of the second grade.
NS-S: M = 30.89, SD = 7.16.

Table 6
Direct and Indirect Effect of Math Anxiety on Math Achievement in Study 2.
Standardized coefficient Model fit

Model Predictor Mediator Output Indirect Total N χ2 CFI RMSEA [90%] SRMR

1 MA (T2) WM (T1) MATH-T2 − 0.20* − 0.18* 333 χ2(5) = 9.12; p = .10 0.98 0.05 [0.00− 0.10] 0.02
NS-S (T1) − 0.01
2 MA-E (T3) WM (T1) MATH- T3 − 0.10 − 0.12* 306 χ2(2) = 1.32; p = .52 1.00 0.00 [0.00− 0.10] 0.01
NS-S (T1) − 0.01
3 MA (T4) WM (T1) MATH-T4 − 0.25** − 0.36*** 262 χ2(5) = 2.74; p = .74 1.00 0.00 [0.00− 0.06] 0.02
NS-S (T1) − 0.01

Note: * p < .05, ** p < .01, *** p < .001. MA – math anxiety is a latent variable that consists of testing (mAMAS-E testing) and learning math anxiety (mAMAS-E
learning), MA-E – mean score of testing (mAMAS-E testing) and learning (mAMAS-E learning) math anxiety, WM – working memory is a latent variable that consists of
visuospatial (the Corsie Block Tapping task) and verbal (WISC-R Digit Span Backward task) working memory, NS-S – symbolic number system (the Symbolic Com­
parison task), MATH – math achievement. T1 – the beginning of the first grade, T2 – the middle of the first grade, T3 – the end of the first grade, T4 – the end of the
second grade.

Fig. 7. Tested multi-mediation model for the relationship between math anx­
iety and math performance in the middle of the first grade. MA (T2) – math
anxiety is a latent variable that consists of testing (mAMAS-E testing) and
Fig. 8. Tested multi-mediation model for the relationship between math anx­
learning math anxiety (mAMAS-E learning), WM (T1) – working memory is a
iety and math performance at the end of the first grade. MA (T3) – math anxiety
latent variable that consists of visuospatial (the Corsie Block Tapping task) and
is a latent variable that consists of testing (mAMAS-E testing) and learning math
verbal (WISC-R Digit Span Backward task) working memory, NS-S (T1) –
anxiety (mAMAS-E learning), WM (T1) – working memory is a latent variable
symbolic number sense (the Symbolic Comparison task), MATH (T2) – math
that consists of visuospatial (the Corsie Block Tapping task) and verbal (WISC-R
achievement. T1 – the beginning of the first class, T2 – the middle of the first
Digit Span Backward task) working memory, NS-S (T1) – symbolic number
class. * p < .05, ** p < .01, *** p < .001.
sense (the Symbolic Comparison task), MATH (T3) – math achievement. T1 –
the beginning of the first class, T3 – the end of the first class. * p < .05, ** p <
low-level deficit in numerical processing and it mediates the develop­ .01, *** p < .001.
ment of a higher level of math skills, while in current studies and the
research conducted by Skagerlund et al. (2019) the assumption made to discussions about the relationship between math anxiety and number
was that math anxiety affects basic number skills. Nevertheless, whether sense as previous studies have confirmed its occurrence (Maloney et al.,
math anxiety impairs number sense or number sense affects math anx­ 2010, 2011; Lindskog et al., 2017; Douglas & LeFevre, 2017; Núñez-
iety, the correlation between both variables should be observable. Peña & Suárez-Pellicioni, 2014; Skagerlund et al., 2019) or absence
However, the obtained results do not provide strong evidence for this (Braham & Libertus, 2018; Dietrich et al., 2015; Hart et al., 2016; Wang
thesis. The analyses, which included various types of math anxiety et al., 2014). Although the obtained results challenge the cognitive
(general, learning, and testing), show that correlations between math deficits model, it may be assumed that people who have weak number
anxiety and the approximate number system are insignificant, or very sense do not cope with mathematics and start avoiding it. At the same
weak and negative. Therefore, the results represent another contribution time, it deepens their math anxiety, which – rather than sharpen

10
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

that the low variance of math anxiety in young children presents a


developmental pattern of math anxiety intensity. Generally, most young
children are not math anxious or feel weak math anxiety (Dowker et al.,
2019; Ginsburg, 2006; Petronzi, Staples, Sheffield, Hunt, & Fitton-
Wilde, 2019). In the working memory tests, the participants had only
two attempts to perform each of the sequences in the tasks. Future
studies should be broadened to include more attempts to perform the
sequences in the tasks that measure working memory. Finally, many
effects were borderline significant or at the level of p < .05. This is a
challenge for the results and indicates a need for further research on this
topic in children.
Finally, educational implications and further research directions
should be discussed. From the practical point of view, the most impor­
tant result is the one that shows the weak and negative mediation of
Fig. 9. Tested multi-mediation model for the relationship between math anx­ working memory in the relationship between math anxiety and math
iety and math performance at the end of the second grade. MA (T4) – math achievement. Math anxiety burdens working memory resources in
anxiety is a latent variable that consists of testing (mAMAS-E testing) and young children, which in turn negatively affects their mathematical
learning math anxiety (mAMAS-E learning), WM (T1) – working memory is a performance. It indicates the great responsibility of parents and teachers
latent variable that consists of visuospatial (the Corsie Block Tapping task) and to limit the math anxiety level already in early school-aged children.
verbal (WISC-R Digit Span Backward task) working memory, NS-S (T1) – Among the reasons of math anxiety reported by young children there are
symbolic number sense (the Symbolic Comparison task), MATH (T4) – math concerns about failure during math problem solving, negative reactions
achievement. T1 – the beginning of the first class, T4 – the end of the four class.
of parents, teachers, and classmates, bad grades, and math problem
* p < .05, ** p < .01, *** p < .001.
solving under time pressure (Petronzi, Staples, Sheffield, Hunt, 2019;
Szczygieł & Pieronkiewicz, accepted). Worries about the consequences
approximate number system acuity – contributes in turn to their low of failure as well as limited time for solving math problems decrease
math achievement. This process continues over the years and the indi­ children’s working memory resources, which results in failure and
vidual factors determine each other. Such mechanisms would explain extended response time. The circle of fear and mistakes is closed.
why almost no relationship is observed between math anxiety and the Therefore, although it sounds like a truism, it is important to create a
approximate number system in the case of children, although it appears friendly and safe atmosphere where every students is approached in a
to be the case in older people. This hypothesis is reinforced by the fact personalized manner. Learners should have enough time to solve math
that, in the present studies, the correlation between the approximate problems (Jameson, 2019) and have a chance to correct their answers
number system and math achievement in young children was signifi­ (Núñez-Peña, Bono, & Suárez-Pellicioni, 2015). The optimization of
cant. This means that math anxiety and number sense independently math class conditions may improve children’s cognitive functioning.
correlate with math outcomes from the beginning of education. There­ The children could in turn use working memory resources more
fore, it is possible that math anxiety and approximate number system efficiently.
acuity develop in separate ways and, in adulthood, they interact and It is also worth considering whether the obtained results and rec­
consequently affect some math learning outcomes (Braham & Libertus, ommendations may be useful for students with special educational
2018). Nevertheless, the level of math anxiety does not predict number needs, e.g. with dyscalculia and attention deficit hyperactivity disorder
sense acuity in childhood. What is more, the obtained results suggest (ADHD). Taking into account deficits in number sense in the former
that number sense may be important in regression models when it is the (Wilson & Dehaene, 2007) and deficits in working memory in both
only predictor but not when general cognitive skills are entered into the groups (Alloway & Passolunghi, 2011; Menon, 2016), it is likely that the
regression model (Fuhs & McNeil, 2013; Gilmore et al., 2013; Szűcs, strategy of avoiding math starts earlier in children with severe mathe­
Devine, Soltesz, Nobes, & Gabriel, 2014). Therefore, cross-sectional and matical difficulties than in children without cognitive and mathematical
developmental research is needed to determine the dynamics in the deficits. Moreover, children with math difficulties may have a higher
change of relationships between math anxiety, number sense, general level of math anxiety than children without such problems (Devine, Hill,
cognitive skills, and math achievement, including complex mediation Carey, & Szűcs, 2018), and math anxiety may additionally impair the
models. It is noteworthy that the direct path and two indirect paths former’s working memory resources more. It is also interesting to test
explained from 32% to 48% of the variance of the results in mathe­ the mechanism of math anxiety effect on math performance in learners
matical achievement. This means that math anxiety, working memory, with ADHD and precise number sense as well as those with overlapping
and number sense are important predictors of mathematical achieve­ problems: ADHD and dyscalculia. Due to their greater difficulty in
ment in early school-aged children. learning math and their higher level of anxiety, one can expect that the
Subsequently, the methodological issues should be discussed. mechanism of math anxiety effect on math performance in children with
Although a large group of children was tested in the two studies, some dyscalculia and/or ADHD may be more similar to the mechanism
limitations with the generalizability of the results on elementary school described in adults than in typical young children.
children should be noted. Both studies were conducted in children from
high/medium socioeconomic status families living in a big city. Consent 6. Conclusion
for the children’s participation in the study was obtained during a
teacher-parent conference in which not all parents participated. It can be The results generally show that working memory mediates the
assumed, therefore, that an additional limitation of the study is that the relationship between math anxiety and math achievement in early
tested children have parents who are involved in their education. school-aged children, whereas the approximate number system path
Another limitation is the relatively low reliability of some of the scales. does not allow an explanation as to how math anxiety impacts mathe­
The low reliability of the math anxiety scales can mostly be explained by matical achievement. In accordance with processing efficiency and
the low variance of results that arises because the response scale con­ attentional cognitive theories, math anxiety generates distracting
sisted of only three categories and therefore produced homogeneous thoughts that reduce working memory resources and in turn impede
results. The problem with math anxiety measurement in young children mathematical operations. Efficient working memory is a factor that
is very common (Primi et al., 2020). However, it should be indicated partly limits the negative effect of math anxiety on the level of

11
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

mathematical tasks in children. Precise number sense does not mediate the three-path model are necessary among children and adolescents to
the relationship between math anxiety and math achievement, so the elucidate the mechanism of action of math anxiety on math achievement
validity of the cognitive deficit model in young children has been chal­ across the lifespan.
lenged. Finally, a direct path from math anxiety to math achievement
was mostly not found, which suggests that the direct effect of math Data availability statement
anxiety on math achievement observed in adults may stem from the fact
that they avoided math at school. Further studies on the universality of The data are available: https://doi.org/10.17605/OSF.IO/SXVQG.

Appendix

Since many researchers test the cognitive deficit model by using symbolic or non-symbolic magnitude processing tasks, it was additionally checked in
the longitudinal study whether the obtained results differed when a symbolic comparison task (NS-S) was used instead of a non-symbolic one (ANS).
The symbolic comparison task was used in accordance with the proposal of Nosworthy et al. (2013).
First, Pearson’s correlation with other variables was checked, and this was followed by the testing of the mediation models (see Tables 5–6,
Fig. 7–9).
The results show that symbolic number sense was positively and weakly related to verbal and visuospatial working memory and math achieve­
ment. The relationship between symbolic number sense and math anxiety was insignificant or significant but very weak and negative, and it depended
on the type of math anxiety and the time point of measurement.
In the second step, the mediation effect of working memory and symbolic number sense was tested in an analogous way to models with a non-
symbolic comparison task.
All models (see Figs. 6–8, and Table 6) are characterized by a good χ 2 value and a very good model fit (Hu & Bentler, 2009). The results show that
working memory negatively, very weakly or weakly, and fully mediates the relationship between math anxiety and math achievement in the middle of
the first grade (β = − 20, p < .01) and at the end of the second grade (β = − 0.25, p < .05). There was no mediation effect of working memory at the end
of first grade (β = − 0.01, p = .82). Symbolic number sense did not mediate the relationship between math anxiety and math achievement any time (the
middle of the first grade β = 0.01, p = .93, the end of the first grade β = − 0.01, p = .91, the end of the second grade β = − 0.01, p = .92) and the direct
effect was not observed any time (the middle of the first grade β = 0.02, p = .83, the end of the first grade β = − 0.01, p = .82, the end of the second
grade β = − 0.11, p = .25). The variance in math achievement was explained by direct and indirect paths in over 40% (R2 = 0.49 in the middle of the
first grade, R2 = 0.43 at the end of the first grade, R2 = 0.41 at the end of the second grade). The results are almost identical to those obtained in
models with non-symbolic comparison tasks. Summing up, math anxiety mostly decreases math achievement via working memory but not directly or
by number sense.

References Developmental Science, 13(3), 508–520. https://doi.org/10.1111/


j.14677687.2009.00897.x.
Dietrich, J. F., Huber, S., Moeller, K., & Klein, E. (2015). The influence of math anxiety
Alloway, T. P., & Passolunghi, M. C. (2011). The relations between working memory and
on symbolic and non-symbolic magnitude processing. Frontiers in Psychology, 6,
arithmetical abilities: A comparison between Italian and British children. Learning
1621.
and Individual Differences, 21, 133–137.
Dolna, J. (2016). Test szóstoklasisty 2016: ranking szkół podstawowych. Nasze Miasto
Ansari, D., De Smedt, B., & Grabner, R. H. (2012). Introduction to the special section on
Kraków. Downloaded from: http://krakow.naszemiasto.
“Numerical and Mathematical Processing”. Mind, Brain and Education, 6(3),
pl/artykul/test-szostoklasisty-2016-ranking-szkol-podstawowych,3763864,artgal,t,
117–118.
id,tm.html [08.05.2020].
Ashcraft, M. H., & Kirk, E. P. (2001). The relationships among working memory, math
Douglas, H. P., & LeFevre, J. (2017). Exploring the influence of basic cognitive skills on
anxiety, and performance. Journal of Experimental Psychology, 130, 224–237.
the relation between math performance and math anxiety. Journal of Numerical
Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H. Bower (Ed.), The
Cognition, 3, 642–666. https://doi.org/10.5964/jnc.v3i3.113.
psychology of learning and motivation: Advances in research and theory (Vol. 8, pp.
Dowker, A., Cheriton, O., Horton, R., & Mark, W. (2019). Relationships between attitudes
47–89). New York: Academic Press.
and performance in young children’s mathematics. Educational Studies in
Braham, E. J., & Libertus, M. E. (2018). When approximate number acuity predicts math
Mathematics, 100, 211–230.
performance: The moderating role of math anxiety. PLoS ONE, 13(5), Article
Evans, J. D. (1996). Straightforward statistics for the behavioral sciences. Pacific Grove:
e0195696.
Brooks/Cole Publishing.
Bulthe, J., De Smedt, B., & Op de Beeck, H. (2014). Format-dependent representations of
Eysenck, M. W., & Calvo, M. G. (1992). Anxiety and performance. The processing
symbolic and non-symbolic numbers in the human cortex as revealed by multi-voxel
efficiency theory. Cognition and Emotion, 6, 409–434.
pattern analyses. NeuroImage, 87, 311–322. https://doi.org/10.1016/j.
Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. (2007). Anxiety and cognitive
neuroimage.2013.10.049.
performance: Attentional control theory. Emotion, 7(2), 336–353.
Butterworth, B., Varma, S., & Laurillard, D. (2011). Dyscalculia: From brain to education.
Ferguson, A. M., Maloney, E. A., Fugelsang, J. A., & Risko, E. F. (2015). On the relation
Science, 332, 1049–1053.
between math and spatial ability: The case of math anxiety. Learning and Individual
Carey, E., Hill, F., Devine, A., & Szűcs, D. (2017). The modified Abbreviated Math
Differences, 39. https://doi.org/10.1016/j.lindif.2015.02.007.
Anxiety Scale: A valid and reliable instrument for use with children. Frontiers in
Forster, K. I., & Forster, J. C. (2003). DMDX: A Windows display program with
Psychology, 8(11). https://doi.org/10.17863/CAM.8377.
millisecond accuracy. Behavior Research Methods, Instruments, and Computers, 35,
Cargnelutti, E., Tomasetto, C., & Passolunghi, M. C. (2017). How is anxiety related to
116–124.
math performance in young students? A longitudinal study of grade 2 to grade 3
Fuhs, M. W., & McNeil, N. M. (2013). ANS acuity and mathematics ability in preschoolers
children. Cognition and Emotion, 31(4), 755–764. https://doi.org/10.1080/
from low-income homes: Contributions of inhibitory control. Developmental Science,
02699931.2016.1147421.
16, 136–148.
Chen, Q., & Li, J. (2014). Association between individual differences in non-symbolic
Ganley, C. M., & McGraw, A. L. (2016). The development and validation of a Revised
number acuity and math performance: A meta-analysis. Acta Psychologica, 148,
Version of the Math Anxiety Scale for Young Children. Frontiers in Psychology, 7
163–172. https://doi.org/10.1016/j.actpsy.2014.01.016.
(1181). https://doi.org/10.3389/fpsyg.2016.01181.
Dehaene, S. (2011). The number sense. How the mind creates mathematics. New York:
Gebuis, T., & Reynvoet, B. (2011). Generating nonsymbolic number stimuli. Behavior
Oxford University Press.
Research Methods, 43(4), 981–986.
Devine, A., Hill, F., Carey, E., & Szűcs, D. (2018). Cognitive and emotional math
Gilmore, C., Attridge, N., Clayton, S., Cragg, L., Johnson, S., Marlow, N., … Inglis, M.
problems largely dissociate: Prevalence of developmental dyscalculia and
(2013). Individual differences in inhibitory control, not non-verbal number acuity,
mathematics anxiety. Journal of Educational Psychology, 110(3), 431–444. https://
correlate with mathematics achievement. PLoS One, 8, Article e67374.
doi.org/10.1037/edu0000222.
Ginsburg, H. (2006). Mathematical play and playful mathematics: A guide for early
De Smedt, B., Taylor, J., Archibald, L., & Ansari, D. (2010). How is phonological
education. In D. G. Singer, R. M. Golinkoff, & K. Hirsh-Pasek (Eds.), Play = learning:
processing related to individual differences in children’s arithmetic skills?
How play motivates and enhances children’s cognitive and social-emotional growth (pp.
145–165). New York: Oxford University Press.

12
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

Harari, R. R., Vukovic, R. K., & Bailey, S. P. (2013). Mathematics anxiety in young OECD. (2010). The high cost of low educational performance: The long-run economic impact
children: An exploratory study. The Journal of Experimental Education, 81(4), of improving educational outcomes. Paris: OECD.
538–555. Park, J., & Brannon, E. M. (2013). Training the approximate number system improves
Hart, S. A., Logan, J. A. R., Thompson, L., Kovas, Y., McLoughlin, G., & Petrill, S. A. math proficiency. Psychological Science, 24(10), 2013–2019. https://doi.org/
(2016). A latent profile analysis of math achievement, numerosity, and math anxiety 10.1177/0956797613482944.
in twins. Journal of Educational Psychology, 108(2), 181–193. https://doi.org/ Park, D., Ramirez, G., & Beilock, S. L. (2014). The role of expressive writing in math
10.1037/edu0000045. anxiety. Journal of Experimental Psychology, 20(2), 103–111.
Hembree, R. (1990). The nature, effects, and relief of mathematics anxiety. Journal for Parsons, S., & Bynner, J. (2006). Does numeracy matter more? London: NRDC.
Research in Mathematics Education, 21(1), 33–46. https://doi.org/10.2307/749455. Petronzi, D., Staples, P., Sheffield, D., Hunt, T., & Fitton-Wilde, S. (2019). Further
Herts, J. B., Beilock, S. L., & Levine, S. C. (2019). The role of parents’ and teachers’ math development of the children’s mathematics anxiety scale UK (CMAS-UK) for ages
anxiety in children’s math learning and attitudes. In I. C. Mammarella, S. Caviola, & 4–7 years. Educational Studies in Mathematics, 100, 231–249.
A. Dokwer (Eds.), Mathematics anxiety. What is known and what is still to be understood Petronzi, D., Staples, P., Sheffield, D., & Hunt, T. (2019). Acquisition, development and
(pp. 190–210). New York, NY: Routledge. maintenance of math anxiety in young children. In I. C. Mammarella, S. Caviola, &
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure A. Dokwer (Eds.), Mathematics anxiety. What is known and what is still to be understood
analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: (pp. 77–102). New York, NY: Routledge.
A Multidisciplinary Journal, 6(1), 1–55. Primi, C., Donati, M. A., Izzo, V. A., Guardabassi, V., O’Connor, P. A., Tomasetto, C., &
Jameson, M. M. (2019). Time, time, time: Perception of the causes of mathematics Morsanyi, K. (2020). The Early Elementary School Abbreviated Math Anxiety Scale
anxiety in highly maths anxious female adult learners. Adult Education Quarterly, (the EES-AMAS): A new adapted version of the AMAS to measure math anxiety in
1–17. https://doi.org/10.1177/0741713619896324. young children. Frontiers in Psychology, 11, 1014. https://doi.org/10.3389/
Janowicz, J. (2017). Jak statystyka może pomóc w odczytaniu wyników sprawdzianu. fpsyg.2020.01014.
Warszawa: Nowa Era. Pobrane z: http://www.sprawdzian6.nowaera. pl/pdf/Jak_s Ramirez, G., & Beilock, S. L. (2011). Writing about testing worries boosts exam
tatystyka_moze_pomoc.pdf (10.05.2020). performance in the classroom. Science, 331, 211–213. https://doi.org/10.1126/
Jordan, N. C., Glutting, J., & Ramineni, C. (2010). The importance of number sense to science.1199427.
mathematics achievement in first and third grades. Learning and Individual Ramirez, G., Chang, H., Maloney, E. A., Levine, S. C., & Beilock, S. L. (2016). On the
Differences, 20, 82–88. relationship between math anxiety and math achievement in early elementary
Justicia-Galiano, M. J., Martín-Puga, M. E., Linares, R., & Pelegrina, S. (2017). Math school: The role of problem solving strategies. Journal of Experimental Child
anxiety and math performance in children: The mediating roles of working memory Psychology, 141, 83–100. https://doi.org/10.1016/j.jecp.2015.07.014.
and math self-concept. The British Journal of Educational Psychology, 87(4), 573–589. Ramirez, G., Gunderson, E. A., Levine, S. C., & Beilock, S. L. (2013). Math anxiety,
https://doi.org/10.1111/bjep.12165. working memory, and math achievement in early elementary school. Journal of
Kessels, R. P., van Zandvoort, M. J., Postma, A., Kappelle, L. J., & de Haan, E. H. (2000). Cognition and Development, 14(2), 187–202.
The Corsie Block-Tapping Task: Standarization and normative data. Applied Richardson, F. C., & Suinn, R. M. (1972). The mathematics anxiety rating scale:
Neuropsychology, 7(4), 252–258. Psychometric data. Journal of Counseling Psychology, 19(6), 551–554.
Lafay, A., St-Pierre, M.-C., & Macoir, J. (2019). Impairment of non-symbolic number Rosseel, Y. (2012). lavaan: An R package for structural equation modeling. Journal of
processing in children with mathematical learning disability. Journal of Numerical Statistical Software, 48(2), 1–36.
Cognition, 5(1), 86–104. https://doi.org/10.5964/jnc.v5i1.177. Rubin, D. B. (1976). Inference and missing data. Biometrika., 63, 581–592.
Melby-Lervåg, M., & Hulme, C. (2013). Is working memory training effective? A meta- Skagerlund, K., Östergren, R., Västfjäll, D., & Träff, U. (2019). How does mathematics
analytic review. Developmental Psychology, 49(2), 270–291. https://doi.org/ anxiety impair mathematical abilities? Investigating the link between math anxiety,
10.1037/a0028228. working memory, and number processing. PLoS ONE, 14(1), Article e0211283.
Libertus, M. E., Feigenson, L., & Halberda, J. (2011). Preschool acuity of the approximate Soltanlou, M., Pixner, S., & Nuerk, H.-C. (2015). Contribution of working memory in
number system correlates with school math ability. Developmental Science, 14(6), multiplication fact network in children may shift from verbal to visuo-spatial: A
1292–1300. https://doi.org/10.1111/j.1467-7687.2011.01080.x. longitudinal investigation. Frontiers in Psychology, 6, 1062.
Lindskog, M., Winman, A., & Poom, L. (2017). Individual differences in nonverbal Soltanlou, M., Artemenko, C., Dresler, T., Fallgatter, A. J., Ehlis, A.-C., & Nuerk, H.-C.
number skills predict math anxiety. Cognition, 159, 156–162. https://doi.org/ (2019). Math anxiety in combination with low visuospatial memory impairs math
10.1016/j.cognition.2016.11.014. learning in children. Frontiers in Psychology, 10(89). https://doi.org/10.3389/
Luttenberger, S., Wimmer, S., & Paechter, P. (2018). Spotlight on math anxiety. fpsyg.2019.00089.
Psychology Research and Behavior Management, 11, 311–322. https://doi.org/ Sorvo, R., Koponen, T., Viholainen, H., Aro, T., Räikkönen, E., Peura, P., … Aro, M.
10.2147/PRBM.S141421. (2017). Math anxiety and its relationship with basic arithmetic skills among primary
Lyons, I. M., Nuerk, H.-C., & Ansari, D. (2015). Rethinking the implications of numerical school children. British Journal of Educational Psychology, 87(3), 309–327. https://
ratio effects for understanding the development of representational precision and doi.org/10.1111/bjep.12151.
numerical processing across formats. Journal of Experimental Psychology, 144(5), Sowinski, C., LeFevre, J. A., Skwarchuk, S. L., Kamawar, D., Bisanz, J., & Smith-Chant, B.
1021–1035. https://doi.org/10.1037/xge0000094. (2015). Refining the quantitative pathway of the pathways to mathematics model.
Maloney, E. A., Ansari, D., & Fugelsang, J. A. (2011). The effect of mathematics anxiety Journal of Experimental Child Psychology, 131, 73–93. https://doi.org/10.1016/j.
on the processing of numerical magnitude. The Quarterly Journal of Experimental jecp.2014.11.004.
Psychology, 64(1), 10–16. Suárez-Pellicioni, M., Núñez-Peña, M. I., & Colomé, A. (2016). Math anxiety: A review of
Maloney, E. A., Risko, E. F., Ansari, D., & Fugelsang, J. (2010). Mathematics anxiety its cognitive consequences, psychophysiological correlates, and brain bases.
affects counting but not subitizing during visual enumeration. Cognition, 114(2), Cognitive, Affective, and Behavioral Neuroscience, 16(1), 3–22. https://doi.org/
293–297. https://doi.org/10.1016/j.cognition.2009.09.013. 10.3758/s13415-015-0370-7.
Matczak, A., Piotrowska, A., & Ciarkowska, W. (1991). Skala inteligencji D. Wechslera dla Szczygieł, M. (2019). How to measure math anxiety in young children? Psychometric
dzieci – wersja zmodyfikowana. Warszawa: Pracownia Testów Psychologicznych PTP. properties of the modified Abbreviated Math Anxiety Scale for Elementary Children
Menon, V. (2016). Working memory in children’s math learning and its disruption in (mAMAS-E). Polish Psychological Bulletin, 4(50), 303–315.
dyscalculia. Current Opinion in Behavioral Sciences, 10, 125–132. Szczygieł, M. (2020a). More evidence that math anxiety is specific to math in young
Namkung, J., Peng, P., & Lin, X. (2019). The relation between mathematics anxiety and children. International Electronic Journal of Elementary Education, 12(5), 429–438.
mathematics performance among school-aged students: A meta-analysis. Review of https://doi.org/10.26822/iejee.2020562133.
Educational Research, 89(3), 459–496. https://doi.org/10.3102/ Szczygieł, M. (2020b). When does math anxiety in parents and teachers predict math
0034654319843494. anxiety and math achievement in elementary school children? The role of gender
Ng, E., & Lee, K. (2015). Effects of trait test anxiety and state anxiety on children’s and grade year. Social Psychology of Education, 23, 1023–1054. https://doi.org/
working memory task performance. Learning and Individual Differences, 40, 141–148. 10.1007/s11218-020-09570-2.
https://doi.org/10.1016/j.lindif.2015.04.007. Szczygieł, M. (2020c). Sukcesy i porażki w uczeniu się matematyki. Uwarunkowania
Nosworthy, N., Bugden, S., Archibald, L., Evans, B., & Ansari, D. (2013). A two-minute osiągnięć matematycznych uczniów edukacji wczesnoszkolnej. [Successes and failures in
paper-and-pencil test of symbolic and nonsymbolic numerical magnitude processing learning mathematics. Determinants of mathematical achievements in early childhood
explains variability in primary school children’s arithmetic competence. Plos One, 8 education.]. Krakow: Libron.
(7), Article e67918. Szczygieł, M. (2020d). Gender, general anxiety, math anxiety and math achievement in
Nuerk, H. C., Weger, U., & Willmes, K. (2004). On the perceptual generality of the unit- early school-age children. Issues in Educational Research, 30(3), 1126–1140.
decade compatibility effect. Experimental Psychology, 51(1), 72–79. Szczygieł, M., & Pieronkiewicz, B. (accepted). Exploring the nature of math anxiety in
Núñez-Peña, M. I., & Suárez-Pellicioni, M. (2014). Less precise representation of young children: Intensity, prevalence, reasons. Mathematical Thinking and Learning.
numerical magnitude in high math-anxious individuals: An ERP study of the size and Szűcs, D., Devine, A., Soltesz, F., Nobes, A., & Gabriel, F. (2014). Cognitive components
distance effects. Biological Psychology, 103, 176–183. https://doi.org/10.1016/j. of a mathematical processing network in 9-year-old children. Developmental Science,
biopsycho.2014.09.004. 17(4), 506–524.
Núñez-Peña, M. I., Bono, R., & Suárez-Pellicioni, M. (2015). Feedback on students’ Tingley, D., Yamamoto, T., Hirose, K., Keele, L., & Imai, K. (2014). mediation: R package
performance: A possible way of reducing the negative effect of math anxiety in for causal mediation analysis. Journal of Statistical Software, 59(5), 1–38. http
higher education. International Journal of Educational Research, 70, 80–87. https:// ://www.jstatsoft.org/v59/i05/.
doi.org/10.1016/j.ijer.2015.02.005. Vukovic, R. K., Kieffer, M. J., Bailey, S. P., & Harari, R. R. (2013). Mathematics anxiety in
Nys, J., Ventura, P., Fernandes, T., Querido, L., Leybaert, J., & Content, A. (2013). Does young children: Concurrent and longitudinal associations with mathematical
math education modify the approximate number system? A comparison of schooled performance. Contemporary Educational Psychology, 38, 1–10.
and unschooled adults. Trends in Neuroscience and Education, 2(1), 1–10. https://doi. Wang, Z., Hart, S. A., Kovas, Y., Lukowski, S., Soden, B., … Petrill, S. A. (2014). Who is
org/10.1016/j.tine.2013.01.001. afraid of math? Two sources of genetic variance for mathematical anxiety. Journal of

13
M. Szczygieł Contemporary Educational Psychology 65 (2021) 101949

Child Psychology and Psychiatry, 55(9), 1056–1064. https://doi.org/10.1111/ Wu, S. S., Barth, M., Amin, H., Melcarne, V., & Menon, V. (2012). Math anxiety in second
jcpp.12224. and third graders and its relation to mathematics achievement. Frontiers in
Wang, Z., Oh, W., Malanchini, M., & Borriello, G. A. (2020). The developmental Psychology, 3(162), 1–11.
trajectories of mathematics anxiety: Cognitive, personality, and environmental Zhang, J., Zhao, N., & Kong, Q. P. (2019). The relationship between math anxiety and
correlates. Contemporary Educational Psychology, 61. https://doi.org/10.1016/j. math performance: A meta-analytic investigation. Frontiers in Psychology, 10, 1613.
cedpsych.2020.101876. https://doi.org/10.3389/fpsyg.2019.01613.
Wilson, A. J., & Dehaene, S. (2007). Number sense and developmental dyscalculia. In
D. Coch, G. Dawson, & K. W. Fischer (Eds.), Human behavior, learning, and the
developing brain: Atypical development (pp. 212–238). The Guilford Press.

14

You might also like