Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Psychological Research (2021) 85:1662–1672

https://doi.org/10.1007/s00426-020-01356-7

ORIGINAL ARTICLE

Non‑symbolic representation is modulated by math anxiety


and cognitive inhibition while symbolic representation not
Dongxiao Guan1,2 · Jiru Ai1 · Yaru Gao1 · Hongxia Li1 · Bijuan Huang1 · Jiwei Si1

Received: 10 October 2019 / Accepted: 8 May 2020 / Published online: 29 May 2020
© Springer-Verlag GmbH Germany, part of Springer Nature 2020

Abstract
The present study aimed to investigate whether there are basic numerical processing deficits in high math anxiety (HMA)
individuals and examine the effects of cognitive inhibition on the performance of high and low math anxiety (LMA) indi-
viduals. 35 undergraduate students were recruited to perform a numerical comparison task, a numerical Stroop task, a dot
comparison task, and a dot Stroop task. Results showed the following: (1) Compared with LMA group, HMA group reacted
more slowly and exhibited more fixation counts in non-symbolic representation task. (2) Inhibition condition significantly
increased HMA individuals’ fixation durations, whereas the fixation durations of LMA individuals were similar between
inhibition and non-inhibition condition. The results demonstrated that MA might temporarily occupy individuals’ working
memory resources. (3) In non-symbolic representation task, the effect of numerical ratio on fixation counts was larger for
HMA group than that for LMA group under inhibition condition, indicating MA might arise from a basic level deficit in
numerosity processing.

Introduction that 59% of students reported worries that “math classes are
difficult for them” and that 30% of students reported feel-
Mathematics anxiety (MA), as a negative emotional reaction ing helpless when solving math problems (OECD, 2013).
characterized by anxiety, nervousness, fear and avoidance of Researchers have reached broad consensus that there is a
stress when individuals deal with mathematical problems slight to moderate connection between MA and poor math
(Young, Wu, & Menon, 2012; Si, Xu, Feng, Xu, & Zhou, performance (e.g., Ashcraft & Krause, 2007; Devine, Faw-
2014), has become a global phenomenon. Data from the Pro- cett, Szücs, & Dowker, 2012; Zakaria, Zain, Ahmad, &
gram for International Student Assessment (PISA) showed Erlina, 2012). Furthermore, MA also has a significant nega-
tive impact on individuals’ long-term career achievement.
In order to minimize such negative effects of MA, a better
* Jiwei Si
sijiwei1974@126.com understanding of the underlying mechanism of MA’s effects
is required.
Dongxiao Guan
guandongxiao@foxmail.com Currently, several theories have been formulated to
explain the effect of MA. The most dominant theory was
Jiru Ai
1715380552@qq.com proposed by Ashcraft and colleagues (Ashcraft & Faust,
1994; Ashcraft & Kirk, 2001). Considering that the rumi-
Yaru Gao
gyr1997@163.com nation caused by MA would consume individuals’ working
memory (WM) resources that are used to perform math-
Hongxia Li
dongfangxia125@163.com ematical tasks, they claimed that MA has complex effects.
First, there is no obvious effect of MA in simple mathemat-
Bijuan Huang
18766173466@163.com ics tasks. Second, the effect of MA will become more sig-
nificant as the complexity of math-related task situations or
1
School of Psychology, Shandong Normal University, No. 1, the difficulty of mathematical problem escalates. Compared
University Road, Science Park, Changqing District, Jinan, with low-mathematics anxiety (LMA) individuals, high-
Shandong, China
mathematics anxiety (HMA) individuals would perform
2
School of Business Administration, Northeastern University, worse on complex arithmetic problems involving high WM
Shenyang, China

13
Vol:.(1234567890)
Psychological Research (2021) 85:1662–1672 1663

load (e.g., Ashcraft & Kirk, 2001; Sun, Si, & Xu, 2012). participants to perform both non-symbolic dot comparison
However, the above theory cannot explain the differences and symbolic number comparison.
of the processing of numerical magnitudes between HMA Scholars from diverse domains, such as cognitive psy-
individuals and LMA individuals (e.g., Maloney, Risko, chology, have agreed that emotion and cognition are linked
Ansari, & Fugelsang, 2010). Generally speaking, people in a complex way (Wang, Lukowski, Hart, Lyons, & Petrill,
do not rely heavily on WM resources when representing 2015). In line with this view, Eysenck, Derakshan, San-
numerical magnitudes. tos and Calvo (2007) suggested that anxiety can damage
The representation of numerical magnitude is usually the inhibitory function of attentional control, which is an
assessed by different variants of the numerical compari- extremely important WM ability. To be specific, inhibitory
son task, namely, symbolic number comparison task and function can resist intense internal tendencies or external
non-symbolic numerosity comparison. In numerical cog- temptations by controlling individuals’ attention, behaviors,
nition research, it is frequently posited that the processing thoughts and/ or emotions. And the function can help indi-
of symbolic numbers is the same as the processing of the viduals to do something more appropriate or necessary at the
non-symbolic numerosities. This is because the above two same time (Diamond, 2013). In other words, by inhibiting
number formats are represented by one shared magnitude dominant reactions, individuals can reduce or even elimi-
system—the Approximate Number System, or shortly ANS nate the interference of unnecessary information in the cur-
(Marinova, Sasanguie, & Reynvoet, 2020). Some studies rent task, and cope with changing circumstances flexibly.
found that individuals with MA might have deficits in the Therefore, attention resources of anxious individuals may be
ANS. In other words, HMA individuals and LMA individu- distributed preferentially to threatening stimuli irrelevant to
als might be not only different in solving difficult math prob- the current task, which results in delay in the processing of
lems. For example, Maloney, Ansari, & Fugelsang (2011) the cognitive task (Eysenck et al., 2007).
pointed out that the influence of numerical distance (i.e., Inhibitory function is related not only to anxiety, but also
the absolute distance between two compared numbers) on to ANS. Fuhs and Mcneil (2013) found that children’s inhib-
performance was larger for HMA group than that for LMA itory function was significantly correlated with the ability of
group in symbolic numerical comparison task, indicating ANS. The evidence from neuroscience further revealed that
the magnitude representation of HMA individuals is less there was an overlap in the brain activation region between
precise than that of their LMA peers. Utilizing the method ANS and inhibitory function (Cantlon, Platt, & Brannon,
of event-related brain potentials (ERPs), Núñez-Peña and 2009). Using the Stroop paradigm, Szucs, Devine, Soltesz,
Suárez-Pellicioni (2014) further verified the result reported Nobes and Gabriel (2013) conducted an empirical research
by Maloney et al. (2011). However, both researches used to examine the difference of the size of the inhibitory effect
only a symbolic number comparison task to access ANS between groups in numerical comparison. They found that
acuity. there was a stronger inhibitory effect in children with devel-
In addition to the numerical distance effect (NDE), the opmental dyscalculia than in children with typical develop-
numerical ratio effect (NRE) is also a common phenom- ment. It is worth noting that the Stroop paradigm is a typical
enon in both non-symbolic magnitude representation and task type to test cognitive inhibition, which is the foundation
symbolic magnitude representation (Moyer & Landauer, of inhibition function (Bai, Jia, & Wang, 2016). Thus trials
1967). NRE implies that participants’ reaction time (RT) in Stroop task usually contain task-irrelevant visual cues that
will be longer and error rates (ERs) will be higher when the need inhibiting from conscious aspect.
numerical ratio between the two stimuli (i.e., the relative
distance between the two numbers) is closer to 1 (Lyons,
Nuerk, & Ansari, 2015). For example, people’s performance The present study
is typically worse when comparing 8 and 9 (ratio = 0.89)
than that when comparing 3 and 9 (ratio = 0.33). Accord- Altogether, MA and inhibition function may be able to
ing to ANS theory, the smaller the ratio of two numbers jointly affect individuals’ performance in numerical repre-
is, the greater the overlap of the ANS representations will sentation tasks, and there appears to be no study examin-
be (Dehaene, Dehaene-Lambertz, & Cohen, 1998; Dietrich, ing this issue. The present study investigated exactly this by
Stefan, Korbinian, & Elise, 2015). Therefore, NRE is just an combining number/numerosity comparison task and Stroop
index of the precision of ANS, as NDE is. Only a few studies paradigm. First, we set the numerical comparison task and
investigated whether the sizes of NRE are different for HMA the dot comparison task as a non-inhibition condition (i.e.,
group and LMA group (e.g., Colomé, 2018). To understand baseline condition). Under the condition, each pair of dot
the influence of MA on numerical representation more in- arrays consisted of physically identical dots (i.e., the surface
depth, the present study and previous researches are unique area of single dot was the same) and each pair of numbers
in two aspects. First, NRE was examined. Second, we asked used the same font size. Second, the numerical Stroop task

13
1664 Psychological Research (2021) 85:1662–1672

and the dot Stroop task were set to an inhibition condition. this study. They were selected from a total of 254 under-
Unlike the non-inhibition condition, we manipulated visual graduate students from Shandong Normal University. All
characteristics of two compared stimuli under the inhibition subjects gave written informed consent. MA was measured
condition. Specifically, dot arrays included physically differ- by using the Chinese version of the Revised Mathematical
ent dots (i.e., the surface area of single dot was not constant) Anxiety Rating Scale (R-MARS; Si & Liu, 2014), and then
and numbers were written in the different font size. In order we ranked the participants’ scores from high to low. HMA
to compare a pair of numbers/ dot sets under the inhibition group constituted roughly the top 15% of the overall distri-
condition, participants needed to inhibit the inference effect bution (38 students), and LMA group constituted roughly the
of visual cues on numerical magnitude processing. Third, bottom 15% of the overall distribution (38 students). After
each task contained four numerical ratios (0.5, 0.6, 0.7 and that, we invited the subjects to participate in the following
0.8) between the two stimuli to be compared. Each subject experiments. Finally, 35 subjects took part in the experi-
was required to complete all tasks. In this way, we could ments, and they were divided into two groups: 19 students
directly test whether MA and cognitive inhibition modulate with HMA (74% female, Mage = 19.37, SD = 1.16), and 16
the numerical representation in young adults. students with LMA (88% female, Mage = 18.38, SD = 0.72).
Considering that eye-tracking technology can provide All participants completed all the tasks and were paid for
sensitive and objective information about attentional pro- corresponding compensation. This study was approved by
cesses (Duchowski, 2007; Price, Wilkey, & Yeo, 2017), the regional ethics committee.
many researchers have used eye-tracking to explore cogni-
tive mechanisms underlying the numerical representation
and the arithmetic processing (e.g., Merkley & Ansari, 2010; Materials
Hunt, Clark-Carter, & Sheffield, 2015; Price et al., 2017).
Besides, Merkley and Ansari (2010) suggested that cognitive Revised Mathematics Anxiety Rating Scale (R‑MARS)
processes represented by eye movement measures might be
different from the processes reflected by RT measure. There- The Chinese version of the R-MARS (Si & Liu, 2014) was
fore, this study recorded RTs and indexes of eye movement utilized to evaluate adults’ math anxiety. It consists of 21
simultaneously. items with a 5-point Likert scale, which indicates how anx-
Evidence indicates that there is a close relationship ious participants feel in different math-related situations.
between arithmetic skills and numerical magnitude repre- The sum of all item scores equals the total score of the
sentations. For example, Goffin and Ansari (2016) found R-MARS. In the present study, the instrument has shown
that the reaction time NDE correlated significantly with strong internal consistency (Cronbach’s α = 0.94).
the arithmetic skill indexed by fluency of simple arithme-
tic. To exclude potential effects of differences in arithmetic The French Kit
skills, the study controlled the score of the French Kit as a
covariate, which reflects the level of individuals’ arithmetic The current study used two sub-tests of the French Kit of
skill. We hypothesized that (a) the numerical ratio and MA Cognitive Factors (French, Ekstrom, & Price, 1963) to index
would interact with participants’ performance on numeri- participants’ arithmetic fluency. The addition sub-test com-
cal magnitude representations, and HMA individuals would prises three-term addition problems; the subtraction/multi-
display a larger NRE compared with LMA individuals; (b) plication sub-test includes double-digit subtraction problems
inhibition condition would have a negative impact on HMA and double-digit by single-digit multiplication problems.
individuals’ performance, but it would not influence LMA Each sub-test contains two pages (each page contains 60
individuals; (c) under inhibition conditions, the numerical problems). The administration time for each page is 2 min.
representation of HMA individuals would be more inaccu- Participants were told to solve the problems as quickly and
rate, and they would show a stronger NRE relative to their accurately as possible. The total scores of the test are calcu-
LMA peers. lated by adding up items solved correctly across four pages.

Symbolic magnitude representation task


Methods
Participants were asked to select the number with larger
Participants numerical magnitude from a pair of two-digit numbers pre-
sented simultaneously (horizontally). The two numbers were
Using an estimated large effect size (f = 0.40, Cohen, 1988), displayed in black font on a white background. The selection
for 80% power at α = 0.05, the present study would need of number pairs in this study (total of 20 pairs) referred to
N = 34 participants, hence 35 participants were recruited in the number of dot comparison tasks selected in Teng (2015).

13
Psychological Research (2021) 85:1662–1672 1665

The number ranged from 5 to 28. There were four numeri- Procedure
cal ratios between the two numbers to be compared, which
were approximately 0.5, 0.6, 0.7 and 0.8. For the numerical The study was composed of two sessions. The first session
comparison task, numbers were presented in the same font was a group test, in which participants were asked to com-
size in all 40 trials (see Fig. 1a). For the numerical Stroop plete the Chinese version of the R-MARS and two sub-tests
task, the number with the larger numerical magnitude and of the French Kit of Cognitive Factors. Participants were
larger font size had to be compared with the number with tested in their classroom during the first session. The sec-
the smaller numerical magnitude and smaller font size in ond session (eye-tracking session) was the individual test;
20 trials. While the number with larger numerical magni- participants were tested in a quiet room that was dimly lit.
tude but smaller font size was compared to the number with Eye-tracking session contained four blocks (i.e., numerical
smaller numerical magnitude but larger font size in the other comparison task, numerical Stroop task, dot comparison
20 trials (see Fig. 1b). As for the above two tasks, the larger task, and dot Stroop task), which were counterbalanced
number was in the left side of the screen in half the trials, among the participants to avoid an order effect. Each block
and the larger number was in the right side of the screen for comprised six practice trials and forty experimental trials.
the other half. The order of trials was balanced by Latin square design.
Each trial started with a fixation point (1000 ms). A pair
Non‑symbolic magnitude representation task of number or two sets of dots were then appeared for up to
2000 ms until the subjects pressed one of the response but-
Participants had to select the set with more dots from two dot tons, and was followed by a blank screen for 300 ms. We
sets presented simultaneously (horizontally), which included scheduled a short break between two blocks. An example of
black dots against a white background. The quantity of dots the numerical comparison task was shown in Fig. 2.
in this study was identical with symbolic magnitude repre-
sentation. For the dot comparison task, the two sets to be Eye‑movement data acquisition
compared consisted of physically identical dots (see Fig. 1c).
For the dot Stroop task, the set of dots with more and physi- Participants’ eye movements were recorded with an Eye-
cally larger dots (i.e., the total surface area of the dots was link 1000 eye-tracking system (SR Research, Mississauga,
larger), had to be compared with the set of dots with fewer Ontario, Canada) at a sampling frequency of 1000 Hz. Chin
and physically smaller dots (i.e., the total surface area of and forehead rests were used in order to minimize head-
the dots was smaller) in 20 trials. While the set of dots with movement artifacts. All stimuli were presented on a 19′′
more but physically smaller dots (i.e., the total surface area monitor and resolution of 800 × 600 pixels by using E-prime
of the dots was smaller) was compared to an array with fewer 2.0 software (Psychology Software Tools Inc., Sharpsburg,
but physically larger dots (i.e., the total surface area of the PA, USA). This study defined two rectangular interest areas
dots was larger; see Fig. 1d) in the other 20 trials. As for the that were slightly larger than each stimulus presentation area
above two tasks, the set with more dots was in the left side (left and right). The size of the interest area was consistent
of the screen in half the trials, and the set with more dots was between non-symbolic and symbolic conditions. The two
in the right side of the screen for the other half. interest areas were for all subsequent analyses of eye move-
ment data (fixation count and fixation duration). It should
be noted that fixation count (FC) meant the total number of
the fixation on two interest areas, and fixation duration (FD)
meant the total time that individuals spent in fixating on two
interest areas.

Data analysis

We coded the ratio 0.5, 0.6 as small ratio, and the ratio 0.7,
0.8 as large ratio. In order to investigate the effect of numeri-
cal ratio, cognitive inhibition and MA on task performance,
2 (MA Group: HMA vs. LMA) × 2 (Numerical Ratio: small
vs. large) × 2 (Cognitive Inhibition Condition: inhibition
vs. non-inhibition) repeated-measures Analysis of Variance
Fig. 1  Four magnitude comparison tasks: a numerical compari-
(ANOVAs) was performed for mean RT, mean FC and mean
son task, b numerical Stroop task, c dot comparison task, and d dot FD, with MA Group as between-subject variable, Numerical
Stroop task Ratio and Cognitive Inhibition Condition as within-subject

13
1666 Psychological Research (2021) 85:1662–1672

Fig. 2  Single experimental


procedure (e.g., numerical
comparison task)

variables. The above analysis was conducted in the symbolic were solved slower (M = 910.42, SD = 16.20) than these
magnitude representation and the non-symbolic quantity with small ratio (M = 881.39, SD = 14.92). Moreover, neither
magnitude representation, respectively. Only trials with cor- the main effects of MA Group, Cognitive Inhibition Con-
rect response were included in our analysis. Trials on which ditions, the two-way (MA Group × Numerical Ratio, MA
RT was shorter than 100 ms or exceeded 3 standard devia- Group × Cognitive Inhibition Condition, nor Cognitive Inhi-
tions above each participant’s mean were excluded (1.5%). bition Condition × Numerical Ratio) nor three-way interac-
In addition, the present study reported behavioral results of tions (MA Group × Numerical Ratio × Cognitive Inhibition
individuals with different MA levels while controlling for Condition) reached significance, ps> 0.05.
the total score of The French Kit.
Fixation count

Results The main effect of Numerical Ratio, MA group, Cognitive


inhibition condition, and the two-way or three-way interac-
Symbolic magnitude representations tions were not significant, ps> 0.05.

Response time Fixation duration

Descriptive statistics outcomes were shown in Table 1. RT There was a MA Group × Cognitive Inhibition Condition
results revealed a significant main effect of Numerical Ratio, interaction (see Fig. 3a), F(1, 32) = 6.67, p < 0.05, η2p = 0.17.
F(1, 32) = 4.22, p < 0.05, η2p = 0.12. Trials with large ratio When each group was analyzed separately, only the

Table 1  Means and standard deviations of response time, fixation duration, fixation counts for LMA and HMA groups under different cognitive
inhibition conditions and numerical ratios in symbolic and non-symbolic magnitude representation tasks
Cognitive inhibi- Numerical ratio Math Symbolic magnitude representation task Non-symbolic magnitude representation task
tion condition anxiety
level RT FD FC RT FD FC

Non-inhibition Large ratio HMA 874.71 (96.26) 680.68 (166.44) 3.17 (0.70) 922.85 (119.31) 713.97 (180.29) 3.61 (0.58)
condition LMA 844.94 (89.65) 725.69 (91.61) 2.77 (0.69) 840.28 (135.00) 716.85 (100.02) 3.18 (0.60)
Small ratio HMA 839.12 (100.03) 662.57 (140.99) 3.05 (0.63) 808.37 (70.65) 631.67 (125.03) 3.27 (0.49)
LMA 808.90 (80.14) 701.86 (81.61) 2.59 (0.65) 745.19 (84.80) 653.50 (66.55) 2.73 (0.49)
Inhibition condi- Large ratio HMA 984.15 (96.86) 837.24 (96.82) 3.37 (0.85) 1027.95 (138.12) 809.66 (156.73) 3.95 (0.62)
tion LMA 935.91 (121.57) 763.75 (174.54) 3.00 (0.69) 922.00 (132.88) 743.51 (118.45) 3.47 (0.76)
Small ratio HMA 958.67 (74.21) 800.97 (98.19) 3.40 (0.66) 866.40 (83.55) 685.38 (107.86) 3.35 (0.56)
LMA 917.22 (124.30) 743.50 (157.74) 2.89 (0.67) 790.23 (95.77) 651.16 (89.92) 3.10 (0.75)

RT, FD and FC are the means and standard deviations (in black bracket) of response time, fixation duration and fixation counts, respectively
HMA high math anxiety, LMA low math anxiety

13
Psychological Research (2021) 85:1662–1672 1667

an interaction between MA and cognitive inhibition. As we


have expected, inhibition condition significantly increased
HMA individuals’ fixation durations, whereas LMA individ-
uals spent similar time in fixating interest areas under both
inhibition and non-inhibition conditions. Therefore, the cog-
nitive inhibition condition appeared to negatively influence
HMA participants’ performance on numerical magnitude
representation, and it did not impact LMA participants’ per-
formance. Unexpectedly, the MA Group × Numerical Ratio
interaction did not reach significance, no matter in RT, FC,
or FD analyses. These results did not replicate the findings
obtained by Maloney et al. (2011) and Núñez-Peña and
Suárez-Pellicioni (2014), which showed that HMA group
displayed a larger distance effect in number comparison
tasks compared with LMA group. Instead, there appeared to
be no deficit in processing numerical information for HMA
individuals. Finally, we failed to find the three-way interac-
tion (MA Group × Numerical Ratio × Cognitive Inhibition
Condition) in all of RT, FC, or FD analyses. Our results
indicated that cognitive inhibition might not relate to the
processing of numerical magnitudes.

Non‑symbolic magnitude representations

Response time

Descriptive statistics outcomes are shown in Table 1. RT


results revealed a significant main effect of Numerical Ratio,
F(1, 32) = 4.82, p < 0.05, η2p = 0.13. Participants took sig-
nificantly longer time to respond to trials with large ratio
(M = 928.16, SD = 20.42) relative to trials with small ratio
(M = 802.67, SD = 13.49). Notably, the main effect of MA
Group also reached significance, F(1, 32) = 5.23, p < 0.05,
η2p = 0.14. HMA individuals reacted slower (M = 906.35,
Fig. 3  Fixation duration (FD; mean and standard error) as a function
SD = 23.30) than their LMA counterparts (M = 824.48,
of cognitive inhibition condition in high math anxiety (HMA) and
low math anxiety (LMA) groups in symbolic/ non-symbolic mag- SD = 25.55). However, there was no main effect of Cogni-
nitude representation: a symbolic magnitude representation, b non- tive Inhibition Condition, no two-way (MA Group × Numeri-
symbolic magnitude representation cal Ratio, MA Group × Cognitive Inhibition Condition, nor
Cognitive Inhibition Condition × Numerical Ratio) or three-
participants in HMA group showed significant differences way interactions (MA Group × Numerical Ratio × Cognitive
between the two cognitive conditions, F(1, 32) = 21.63, Inhibition Condition), ps > 0.05.
p < 0.001, η2p = 0.40, whose fixation durations were longer
under inhibition condition than under non-inhibition condi- Fixation count
tion. As for LMA group, the difference between the two cog-
nitive conditions did not reach significance, F(1, 32) = 0.39, The main effect of Numerical Ratio was significant, F(1,
p = 0.54, η2p = 0.01. There was no main effect of Numeri- 32) = 5.94, p < 0.05, η2p = 0.16. The fixation counts were
cal Ratio, MA Group, Cognitive Inhibition Condition, nei- different between HMA and LMA group, F(1, 32) = 4.41,
ther MA Group × Numerical Ratio, Cognitive Inhibition p < 0.05, η2p = 0.12. Paired comparison showed the fixation
Condition × Numerical Ratio, nor MA Group × Numeri- counts of HMA group (M = 3.55, SD = 0.48) were more
cal Ratio × Cognitive Inhibition Condition interactions, than that of their LMA peers (M = 3.12, SD = 0.61). More
ps> 0.05. importantly, the triple interaction MA Group × Numerical
To sum up, we observed the NRE in the representation of Ratio × Cognitive Inhibition Condition interaction reached
symbolic magnitude. More importantly, FD results showed significance [F(1, 32) = 4.57, p < 0.05, η2p = 0.13]. Separate

13
1668 Psychological Research (2021) 85:1662–1672

ANOVAs were conducted for the inhibition condition difference) with reference to Núñez-Peña and Suárez-
and the non-inhibition condition, with MA Group as the Pellicioni (2014). ANOVAs were performed for the NRE,
between-subjects factor and Numerical Ratio as the within- with Cognitive Inhibition Condition as within-subject
subject factor (see Fig. 4). The main effect of Numerical factors and MA Group as the between-subjects factor.
Ratio [F(1, 32) = 5.13, p < 0.05, η2p = 0.14] and the Numeri- The MA Group × Cognitive Inhibition Condition interac-
cal Ratio × MA Group interaction [F(1, 32) = 5.74, p < 0.05, tion was significant, F(1, 32) = 4.57, p < 0.05, ηp2 = 0.13,
η2p = 0.15] were significant only for inhibition condition, revealing that NRE in HMA group was larger than that in
showing that both HMA group and LMA group demon- LMA group for the inhibition condition, F(1, 32) = 5.74,
strated NRE [F(1, 32) = 89.43, p < 0.001, η2p = 0.74; F(1, p < 0.05, ηp2 = 0.15. Nonetheless, the difference of NRE
32) = 27.73, p < 0.001, η2p = 0.46, respectively]. between HMA group and LMA group was not signifi-
On this basis, in order to further determine whether cant under the non-inhibition condition, F(1, 32) = 0.62,
the NRE of HMA group was stronger than LMA group, p = 0.44, ηp2 = 0.02.
we constructed the index of NRE (large ratio–small ratio

Fixation duration

There was a MA Group × Cognitive Inhibition Condi-


tion interactionsee (Fig. 3b), F(1, 32) = 6.52, p < 0.05,
ηp2 = 0.17. Simple effect analysis results showed that the
fixation durations of HMA group was longer under the
inhibition condition than that under the non-inhibition
condition, F(1, 32) = 13.10, p = 0.001, η p2 = 0.29. How-
ever, for LMA group, the difference of fixation durations
between the inhibition condition and the non-inhibition
condition was not significant, F(1, 32) = 0.06, p = 0.80,
ηp2 = 0.00. No main effect of Numerical Ratio, MA Group,
Cognitive Inhibition Condition, and no other two-way or
three-way interactions were found, ps > 0.05.
To sum up, we also observed NRE in the non-symbolic
magnitude representation. Furthermore, HMA individuals’
fixation durations significantly increased under inhibition
condition than under non-inhibition condition, and the
inhibition condition had no effect on LMA group. These
results were in line with our hypothesis. It should be noted
that we found the differences caused by MA in non-sym-
bolic magnitude representations. Specifically, the RTs of
HMA group were longer than that of LMA group and the
FCs of HMA group were also more than that of LMA
group. Although we did not confirm the data obtained
by Maloney et al. (2011) and Núñez-Peña and Suárez-
Pellicioni (2014), our results opposed the view that MA
has complex effects: there is no obvious effect of MA in
simple mathematics tasks (Ashcraft & Faust, 1994; Ash-
craft & Kirk, 2001). Instead, our results of non-symbolic
representation tasks supported that HMA individual might
have a basic deficit in numerical manipulating (Maloney
et al., 2010). Above all, we observed an expected three-
way interaction MA Group × Numerical Ratio × Cogni-
tive Inhibition Condition in FC analysis. Further analysis
Fig. 4  Fixation count (FC; mean and standard error) as a function of described that under inhibition conditions, HMA individu-
numerical ratio in high math anxiety (HMA) and low math anxiety
als showed a stronger NRE relative to their LMA peers.
(LMA) groups in non-inhibition/ inhibition condition in non-sym-
bolic magnitude representation: a non-inhibition condition, b inhibi- Our results indicated that cognitive and MA could modu-
tion condition late non-symbolic representations.

13
Psychological Research (2021) 85:1662–1672 1669

Discussion Archibald, & Ansari, 2019). Recently, Huang, Zhao, Li,


Yang and Si (2019) chose four groups with different levels
The current study examined the effect of MA and cog- of MA and arithmetic skill to complete a symbolic repre-
nitive inhibition on the symbolic magnitude representa- sentation task. RT results showed that individuals with
tion and the non-symbolic magnitude representation. We high arithmetic skill responded faster than individuals with
hoped that eye-movement measures could provide unique low arithmetic skill, and there was no difference between
evidences about the numerical magnitude representation HMA group and LMA group. Furthermore, there was a
of individuals with different MA levels, and beyond the MA × arithmetic skill interaction in electrophysiological
findings described by reaction time. results, suggesting that basic arithmetic skill could modu-
In symbolic magnitude representation as well as non- late the MA-performance link. Arithmetic skill (as meas-
symbolic magnitude representation, this study found a reli- ured by the French kit) was entered as a covariate in our
able NRE reflected by RT (symbolic/non- symbolic task) analysis, whereas previous studies, such as Maloney et al.
and FC (non-symbolic task) measurements. However, we (2011) and Núñez-Peña and Suárez-Pellicioni (2014), did
did not replicate the findings reported by previous studies not take any measure of arithmetic skill or control that as
(e.g., Maloney et al., 2011; Núñez-Peña and Suárez-Pelli- a covariate. Thus, there could be potential effects of arith-
cioni, 2014): there was an interaction between MA and the metic skill on the relationship between MA and numerical
numerical distance in the symbolic number comparison magnitude representation in their studies.
task. For the absence of this interaction in our symbolic Although the MA group × numerical ratio interaction was
representation task, the reason could be one of several. not observed in the non- symbolic magnitude representation
First, our task included trials involving two-digit num- task, we found that HMA individuals reacted more slowly
ber pairs, and previous studies used a one-digit number and used more fixation counts in non-symbolic tasks than
comparison task (e.g., Maloney et al., 2011; Núñez-Peña their LMA counterparts. These results were consistent with
and Suárez-Pellicioni, 2014). Compared with single-digit the findings reported by Maloney et al. (2010). They found
numbers, the comparison of two-digit numbers seems to that LMA individuals performed significantly better than
require additional processes, such as Huber, Nuerk, Reips HMA individuals in the counting range, which indicated
and Soltanlou (2017) suggested that the tens and units that MA could arise from a basic level deficit in numerical
of two- digits numbers are operated separately when two processing. This view challenged the dominant account of
numbers are compared. Therefore, we speculated these mathematics anxiety by Ashcraft and colleagues. Given that
additional cognitive processes might modulate or even working memory resources could be consumed by rumina-
obscure the difference of the NRE between HMA group tions induced by anxiety, Ashcraft and Faust (1994) assumed
and LMA group (e.g., Huber, Moeller, Nuerk, & Willmes, that MA would affect individuals’ performance only for
2013; Huber et al., 2017). Further investigations should complex mathematical problems that were depended on
control the influence of these other processes when using abundant cognitive resources. In other words, their view
multi-digit number comparison tasks. Second, besides has two main ideas: (a) the impact of MA on performance
MA, other individual differences, such as gender and has complexity effect, and (b) MA undermines individu-
time spent in playing computer games, could also influ- als’ performance on mathematical problem solving through
ence individuals’ performance when they compared a pair briefly consuming WM resources (Maloney et al., 2010).
of two-digit numbers (Huber et al., 2017). Specifically, Our RT and FC results in the non-symbolic task contra-
the response time of female subjects was slower and the dicted the first idea. We confirmed that MA could also affect
time spent playing computer games was positively corre- individuals’ performance in simple mathematical task (e.g.,
lated with reaction speed. As for our study, the proportion non-symbolic magnitude representations), suggesting that
of female subjects in HMA group and LMA group was HMA individuals might have a defect in representing non-
not completely equal, and we did not ask participants to symbolic magnitude.
report how long they spent in playing computer games. Expectedly, FD results showed that there was just an
These factors might cause the failure to repeat the findings interaction between MA and cognitive inhibition conditions
observed by previous studies (e.g., Maloney et al., 2011; in both the symbolic magnitude representation task and the
Núñez-Peña and Suárez-Pellicioni, 2014), and it should be non-symbolic magnitude representation task. Specifically,
taken into account the influence of these factors in future HMA individuals were more sensitive to the setting of inhi-
research. bition conditions, because their fixation times increased
Finally, some studies verified the correlation between significantly under inhibition conditions. Considering that
arithmetic skills and numerical magnitude representa- the fixation duration depends on the amount of cognitive
tions (e.g., Goffin & Ansari, 2016; Hawes, Nosworthy, processing, also known as cognitive load (Rayner, 1998), we
suggested that HMA individuals might pay more cognitive

13
1670 Psychological Research (2021) 85:1662–1672

resources to inhibition conditions than to non-inhibition a greater NRE effect, which also reflected the nature that the
conditions. Hence, these results supported the second idea ANS of HMA individuals was more imprecise. The results
proposed by Ashcraft and Faust (1994): MA undermines further supported that the poor performance of HMA indi-
individuals’ mathematical relevant performance by briefly viduals in the math-related task might be due to their basic
consuming WM resources. Similarly, Ashkenazi (2018) used and low-level defects in quantitative processing.
the numerical Stroop paradigm and found that HMA indi- Most numerical cognitive models assumed that both sym-
viduals showed higher processing of information unrelated bolic and non-symbolic quantities are processed by ANS,
to numerical after mathematical priming. This result may be and the processing is according to Weber’s law. However,
due to the fact that there is a vigilance-avoidance pattern in we did not obtain the similar three-way interaction (MA
the attention bias of anxious individuals. Pizzie and Kraemer group × numerical ratio × cognitive inhibition) in symbolic
(2017) verified this view from the perspective of cognitive representation task. Krajcsi, Lengyel, and Kojouharova
neuroscience. They observed that HMA participants first (2018) reported that the ANS model can better predict the
showed stronger amygdala reactivity while responding to performance of non-symbolic dot comparisons, while its
transient mathematical stimuli, and then they exhibited a dis- predictions for symbolic India-Arabic numbers are less pre-
engagement bias away from math stimuli in behavior. Their cise. This is to say that non-symbolic dot comparison tasks
findings meant that anxious individuals would prefer to may be more suitable for measuring individuals’ ANS acu-
focus on threatening information during the early automatic ity. Besides, some researchers also suggested that NRE in
stage of processing, while in later period (more strategic pro- symbolic tasks might be related to the numerical comparison
cessing stages), they tend to shift their attention away from process, rather than ANS acuity (e.g., Dietrich et al., 2015;
threats (Pizzie & Kraemer, 2017; Ashkenazi, 2018). With Carrie, Danielle, & Christine, 2016). Therefore, the absence
regard to our study, the existence of inhibition condition of the three-way interaction in the symbolic task might be
might make the task more difficult, which caused that more caused by the fact that HMA individuals had no defects in
cognitive resources were needed to complete the task and processing numerical magnitudes. Taken together, these
anxiety could impair individuals’ attention control, which results revealed that the ANS acuity of HMA individuals
made HMA individuals fail to allocate attention resources might be less precise than that of LMA individuals, and
to current task effectively. Hence, HMA participants spent there appeared to be no difference between HMA group and
more fixation times in comparing two stimuli. LMA group in representing numerical magnitudes.
Just as expected, we found a three-way interaction MA
group × numerical ratio × cognitive inhibition in the non- Educational implications
symbolic representation task, which reflected by FC meas-
urements. To be clear, HMA group displayed larger NRE in The present study showed that HMA individuals might have
the dot Stroop task than LMA group. Processing efficiency defects in representing the numerosity magnitude. Specifi-
theory (Eysenck & Calvo, 1992) claimed that anxiety can cally, their ANS seem to be less precise. Maloney et al.
occupy part of limited WM resources, which will reduce the (2011) pointed out the difficulty that HMA have in com-
cognitive resources allocated to the current task. Eysenck pleting complex problems may fundamentally result from
and Calvo (1992) suggested that there may be a compensa- imprecise representations of numerical magnitude. Braham
tion effect when the current task is relatively simple (i.e., and Libertus (2018) found there is a positive relationship
requires less working memory resources). Therefore, the per- between the precision of ANS and applied problem solv-
formance effectiveness, reflected by behavior measurement ing, which is only present in HMA students. Therefore, the
indicators, might be not affected by anxiety. When the task precision of ANS could be served as a ‘protective factor’
is relatively difficult (i.e., need more cognitive resources), for HMA individuals (Braham & Libertus, 2018). The good
which will lead to the competition of cognitive resources ANS acuity is likely to improve HMA students’ high-level
between anxiety and the task requirement, and damage the math skills, which are used to solve difficult mathematics
anxiety individuals’ performance effectiveness. According problems. When HMA individuals have less difficulty in
to this view, we thought that cognitive resources of HMA solving complex math problems, the levels of anxiety would
individuals were sufficient under non-inhibition condition; decrease. Besides, research from children with develop-
they could consciously compensate for quantitative process- mental dyscalculia examined that a short-term numerosity
ing. Consequently, there seemed to be no difference in NRE training could boost these children’s arithmetic performance
between HMA group and LMA group. Instead, under inhibi- (Cheng et al. 2019). As for HMA students, we suggest that
tion condition, HMA individuals’ cognitive resources were their parents and teachers can conduct the training of non-
used to deal with the interference information in the task and symbolic quantity representation, which could promote
the ruminations induced by anxiety, thus the compensation these students’ ANS acuity and improve their performance
effect did not exist. Hence, HMA group inevitably showed on more difficult problems.

13
Psychological Research (2021) 85:1662–1672 1671

Acknowledgements The authors are grateful to Ms. Tingting Wang, Devine, A., Fawcett, K., Szücs, D., & Dowker, A. (2012). Gender dif-
Xiao Ding, and Yalin Yang, Mr Weixing Yang, and Jiajia Zhang for ferences in mathematics anxiety and the relation to mathematics
their distinguished contributions to collecting the experimental data. performance while controlling for test anxiety. Behavioral and
This work was supported by the National Natural Science Foundation Brain Functions, 8, 33. https​://doi.org/10.1186/1744-9081-8-33.
of China (31371048; 31971010) and Ministry of Education Humanities Diamond, A. (2013). Executive functions. Annual Review of Psychol-
and Social Sciences Research Foundation (18YJA190014). ogy, 64(1), 135–168.
Dietrich, J. F., Stefan, H., Korbinian, M., & Elise, K. (2015). The
influence of math anxiety on symbolic and non-symbolic mag-
Compliance with ethical standards nitude processing. Frontiers in Psychology, 6, 1621. https​://doi.
org/10.3389/fpsyg​.2015.01621​.
Conflict of interest The author(s) declared no potential conflict of in- Duchowski, A. (2007). Eye tracking methodology: Theory and prac-
terest with respect to the research, authorship, and/or publication of tice. London: Springer.
this article. Eysenck, M. W., & Calvo, M. G. (1992). Anxiety and performance:
The processing efficiency theory. Cognition and Emotion, 6(6),
Ethical statement This study was funded by 31371048, 31971010, and 409–434.
18YJA190014. The manuscript does not contain clinical studies or Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. (2007).
patient data. Studies were approved by the local ethics committee and Anxiety and cognitive performance: Attentional control theory.
all experimental manipulations were in accordance with the approved Emotion, 7(2), 336–353.
guidelines. French, J. W., Ekstrom, R. B., & Price, L. A. (1963). Kit of refer-
ence tests for cognitive factors. Princeton: Educational Testing
Informed consent All human participants had signed an informed Service.
consent form. Fuhs, M. W., & McNeil, N. M. (2013). ANS acuity and mathematics
ability in preschoolers from low-income homes: Contributions
of inhibitory control. Developmental Science, 16(1), 136–148.
Goffin, C., & Ansari, D. (2016). Beyond magnitude: Judging ordinal-
ity of symbolic number is unrelated to magnitude comparison
References and independently relates to individual differences in arithmetic.
Cognition, 150, 68–76.
Hawes, Z., Nosworthy, N., Archibald, L., & Ansari, D. (2019). Kin-
Ashcraft, M. H., & Faust, M. W. (1994). Mathematics anxiety and
dergarten children’s symbolic number comparison skills predict
mental arithmetic performance: An exploratory investigation.
1st grade mathematics achievement: Evidence from a two-minute
Cognition and Emotion, 8(2), 97–125.
paper-and-pencil test. Learning and Instruction, 59, 21–33.
Ashcraft, M. H., & Kirk, E. P. (2001). The relationships among work-
Huang, B. J., Zhao, X. M., Li, H. X., Yang, W. X., & Si, J. W. (2019).
ing memory, math anxiety, and performance. Journal of Experi-
Arithmetic skill may refine the performance of individuals with
mental Psychology: General, 130(2), 224–237.
high math anxiety, especially in the calculation task: An ERP
Ashcraft, M. H., & Krause, J. A. (2007). Working memory, math per-
study. Scientific Reports, 9, 13283. https​://doi.org/10.1038/s4159​
formance, and math anxiety. Psychonomic Bulletin and Review,
8-019-49627​-7.
14(2), 243–248.
Huber, S., Moeller, K., Nuerk, H. C., & Willmes, K. (2013). A com-
Ashkenazi, S. (2018). Intentional and automatic processing of numeri-
putational modeling approach on three-digit number processing.
cal information in mathematical anxiety: Testing the influence of
Topics in Cognitive Science, 5(2), 317–334.
emotional priming. Cognition and Emotion, 32(8), 1–8.
Huber, S., Nuerk, H. C., Reips, U. D., & Soltanlou, M. (2017). Indi-
Bai, X., Jia, L., & Wang, J. (2016). Emotional priming effects on dif-
vidual differences influence two-digit number processing, but not
ficult Stroop task for trait anxiety. Journal of Psychological Sci-
their analog magnitude processing: A large-scale online study.
ence, 39(1), 8–12.
Psychological Research, 83(4), 1444–1464.
Braham, E. J., & Libertus, M. E. (2018). When approximate number
Hunt, T. E., Clark-Carter, D., & Sheffield, D. (2015). Exploring the
acuity predicts math performance: The moderating role of math
relationship between mathematics anxiety and performance: An
anxiety. PLoS One, 13(5), e0195696. https​://doi.org/10.1371/
eye-tracking approach. Applied Cognitive Psychology, 29(2),
journ​al.pone.01956​96.
226–231.
Cantlon, J. F., Platt, M. L., & Brannon, E. M. (2009). Beyond the num-
Krajcsi, A., Lengyel, G., & Kojouharova, P. (2018). Symbolic number
ber domain. Trends in Cognitive Sciences, 13(2), 83–91.
comparison is not processed by the analog number system: Dif-
Carrie, G., Danielle, H., & Christine, S. (2016). How math anxiety
ferent symbolic and non-symbolic numerical distance and size
relates to number–space associations. Frontiers in Psychology, 7,
effects. Frontiers in Psychology, 9, 124. https​://doi.org/10.3389/
1401. https​://doi.org/10.3389/fpsyg​.2016.01401​.
fpsyg​.2018.00124​.
Cheng, D. Z., Xiao, Q., Cui, J. X., Chen, C. S., Zeng, J. Y., Chen, Q.,
Lyons, I. M., Nuerk, H. C., & Ansari, D. (2015). Rethinking the impli-
& Zhou, X. L. (2019). Short-term numerosity training promotes
cations of numerical ratio effects for understanding the devel-
symbolic arithmetic in children with developmental dyscalculia:
opment of representational precision and numerical processing
The mediating role of visual form perception. Developmental Sci-
across formats. Journal of Experimental Psychology: General,
ence. https​://doi.org/10.1111/desc.12910​.
144(5), 1021–1035.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences.
Maloney, E. A., Ansari, D., & Fugelsang, J. A. (2011). The effect of
Hillsides: Lawrence Erlbaum Associates.
mathematics anxiety on the processing of numerical magnitude.
Colomé, À. (2018). Representation of numerical magnitude in math-
The Quarterly Journal of Experimental Psychology, 64(1), 10–16.
anxious individuals. Quarterly Journal of Experimental Psychol-
Maloney, E. A., Risko, E. F., Ansari, D., & Fugelsang, J. (2010). Math-
ogy, 72(3), 424–435.
ematics anxiety affects counting but not subitizing during visual
Dehaene, S., Dehaene-Lambertz, G., & Cohen, L. (1998). Abstract
enumeration. Cognition, 114(2), 293–297.
representations of numbers in the animal and human brain. Trends
Marinova, M., Sasanguie, D., & Reynvoet, B. (2020). Numerals do
in Neuroscience, 21(8), 355–361.
not need numerosities: Robust evidence for distinct numerical

13
1672 Psychological Research (2021) 85:1662–1672

representations for symbolic and non-symbolic numbers. Psy- Si, J. W., Xu, Y. L., Feng, H. M., Xu, X. H., & Zhou, C. (2014). Dif-
chological Research. https​://doi.org/10.1007/s0042​6-019-01286​ ferences of arithmetic strategy use in adults with different math
-z. (on line). anxieties: An ERP study. Acta Psychologica Sinica, 46(12),
Merkley, R., & Ansari, D. (2010). Using eye tracking to study numeri- 1835–1849.
cal cognition: The case of the ratio effect. Experimental Brain Sun, Y., Si, J. W., & Xu, Y. L. (2012). The effect of math anxiety on
Research, 206(4), 455–460. college students’/children’s strategy use in computational esti-
Moyer, R. S., & Landauer, T. K. (1967). Time required for judgments mation: A comparative study. Psychological Development and
of numerical inequality. Nature, 215(5190), 1519–1520. Education, 28(3), 263–270.
Núñez-Peña, M. I., & Suárez-Pellicioni, M. (2014). Less precise rep- Szucs, D., Devine, A., Soltesz, F., Nobes, A., & Gabriel, F. (2013).
resentation of numerical magnitude in high math-anxious indi- Developmental dyscalculia is related to visuo-spatial memory and
viduals: An ERP study of the size and distance effects. Biological inhibition impairment. Cortex, 49(10), 2674–2688.
Psychology, 103, 176–183. Teng, J. (2015). The relationship between children’ non-symbolic
OECD. (2013). PISA 2012 results: Ready to learn: Students’ engage- numerical representation and mathematics achievement (Unpub-
ment, drive and self-beliefs (Vol. III). Paris: OECD Publishing. lished master’s thesis). Shanghai: East China Normal University.
Pizzie, R. G., & Kraemer, D. J. M. (2017). Avoiding math on a rapid Wang, Z., Lukowski, S. L., Hart, S. A., Lyons, I. M., & Petrill, S. A.
timescale: Emotional responsivity and anxious attention in math (2015). Is math anxiety always bad for math learning? The role of
anxiety. Brain and Cognition, 118, 100–107. math motivation. Psychological Science, 26(12), 1863.
Price, G. R., Wilkey, E. D., & Yeo, D. J. (2017). Eye-movement pat- Young, C. B., Wu, S. S., & Menon, V. (2012). The neurodevelopmental
terns during nonsymbolic and symbolic numerical magnitude basis of math anxiety. Psychological Science, 23(5), 492–501.
comparison and their relation to math calculation skills. Acta Zakaria, E., Zain, N. M., Ahmad, N. A., & Erlina, A. (2012). Mathe-
Psychologica, 176, 47–57. matics anxiety and achievement among secondary school students.
Rayner, K. (1998). Eye movements in reading and information pro- American Journal of Applied Science, 9(6), 890–893.
cessing: 20 years of research. Psychological Bulletin, 124(3),
372–422. Publisher’s Note Springer Nature remains neutral with regard to
Si, J. W., & Liu, W. Z. (2014). The Chinese version of revised mathe- jurisdictional claims in published maps and institutional affiliations.
matics anxiety rating scale (R-MARS). In J. L. Shen & Y. H. Chen
(Eds.), Handbook of research instruments of educational psychol-
ogy in China (pp. 92–97). Beijing: Higher Education Press.

13
Reproduced with permission of copyright owner. Further reproduction
prohibited without permission.

You might also like