Research in Developmental Disabilities 112 (2021) 103881

Contents lists available at ScienceDirect

Research in Developmental Disabilities

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

The co-occurrence of Attention-Deficit/Hyperactivity Disorder

and mathematical difficulties: An investigation of the role of basic
numerical skills
Elena von Wirth a, b, *, Katharina Kujath c, Lea Ostrowski b, Ellen Settegast b,
Sarah Rosarius b, Manfred Döpfner a, b, Christin Schwenk d, Jörg-Tobias Kuhn d
a
University of Cologne, Faculty of Medicine and University Hospital Cologne, Department of Child and Adolescent Psychiatry, Psychosomatics and
Psychotherapy, Cologne, Germany
b
University of Cologne, Faculty of Medicine and University Hospital Cologne, School for Child and Adolescent Cognitive Behavior Therapy (AKiP),
Cologne, Germany
c
University of Muenster, Institute of Psychology, Muenster, Germany
d
TU Dortmund University, Faculty of Rehabilitation Sciences, Dortmund, Germany

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

No. of reviews completed is 2 Background: Attention-deficit/hyperactivity disorder (ADHD) and dyscalculia, also called math­
ematics disorder, frequently co-occur, yet the etiology of this comorbidity is poorly understood.
Keywords: Aims: This study investigated whether impairments in the understanding of numbers and mag­
ADHD nitudes (basic numerical skills) are a unique risk factor for mathematical difficulties (MD) or a
Mathematical difficulties
shared risk factor that could help to explain the association between ADHD and MD.
Comorbidity
Methods and procedures: Basic numerical skills were assessed with eight subtests in children (age
Basic numerical skills
Numerical processing 6–10 years, N = 86) with clinically significant ADHD symptoms and/or MD and typically
Risk factor developing children (control group). This double dissociation design allowed to test for main and
interaction effects of ADHD and MD using both classical and Bayesian analysis of variance
(ANOVA).
Outcomes and results: Children with MD were impaired in transcoding, complex number and
magnitude comparison, and arithmetic fact retrieval. They were not impaired in tasks assessing
core markers of numeracy, which might be explained by the sample including children with
mathematical difficulties instead of a diagnosed dyscalculia. ADHD was not associated with
deficits in any of the tasks. The evidence for an additive combination of cognitive profiles was
weak.
Conclusions and implications: Impairments in basic numerical skills are uniquely associated with
MD and do not represent a shared risk factor for ADHD symptoms and MD.

What this paper adds?

A better understanding of the comorbidity between ADHD and dyscalculia, also called mathematics disorders, is needed to

* Corresponding author at: University of Cologne, Faculty of Medicine and University Hospital Cologne, School for Child and Adolescent Cognitive
Behavior Therapy (AKiP), Pohligstr. 9, 50969, Cologne, Germany.
E-mail address: [email protected] (E. von Wirth).

https://doi.org/10.1016/j.ridd.2021.103881
Received 7 May 2020; Received in revised form 26 August 2020; Accepted 23 January 2021
Available online 16 February 2021
0891-4222/© 2021 Published by Elsevier Ltd.
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

facilitate clinical decision-making. Previous work demonstrated that children with dyscalculia have difficulties on basic numerical
tasks such as counting, comparing numbers or magnitudes, writing numbers to dictation, estimating the position of a number on a
number line, and arithmetic fact retrieval. Only few studies investigated these skills in children with ADHD, and their results are
mixed. Some studies suggest that children with ADHD perform poor on basic numerical tasks; others found no impairments in this
group. The present study investigated basic numerical skills in four groups of children with clinically significant ADHD symptoms
(ADHD), mathematical difficulties (MD), clinically significant ADHD symptoms and mathematical difficulties (ADHD + MD), and
typically developing children. A novel technique (Bayesian ANOVAs and hypothesis testing) was employed as a supplement to null
hypothesis significance testing (NHST) ANOVA. Results revealed that children with MD were impaired in complex number processing.
Children with clinically significant ADHD symptoms showed no impairments on any of the basic numerical tasks, supporting the
hypothesis that deficits in numerical processing are uniquely associated with MD. This finding has clinical implications regarding the
assessment of MD in children with ADHD symptoms.

1. Introduction

Accurate assessment and diagnosis of dyscalculia, also called mathematics disorder, can be impeded by symptoms of attention-
deficit/hyperactivity disorder (ADHD). Since deficits in basic numerical skills have been found to be characteristic of children with
dyscalculia (e.g., Geary, Hoard, & Bailey, 2012; Geary, Hoard, Nugent, & Byrd-Craven, 2008; Landerl, Bevan, & Butterworth, 2004;
Moura et al., 2013), tasks assessing these skills might help to differentiate between dyscalculia and poor math performance caused by
symptoms of inattention, impulsivity or hyperactivity. However, it is to date not clear, whether impairments in the understanding of
numbers and magnitudes are indeed uniquely associated with dyscalculia. Only few studies have investigated basic numerical pro­
cessing in children with ADHD and their results are inconsistent (e.g., Colomer, Re, Miranda, & Lucangeli, 2013; Ganor-Stern &
Steinhorn, 2018). The aim of the present study was, therefore, to investigate whether deficits in basic numerical skills are uniquely
associated with mathematical difficulties (MD) or a shared risk factor for ADHD and MD in order to facilitate clinical decision-making
and to improve our understanding of the etiology of the comorbidity of the two disorders.

1.1. Hypotheses for the comorbidity between ADHD and specific learning disorders

Children with ADHD show high comorbidity rates with dyslexia and dyscalculia, an observation, which is not only mirrored in
prevalence studies (Gross-Tsur, Manor, & Shalev, 1996), but has found its way into the 5th edition of the Diagnostic and Statistical
Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2013). The comorbidity of ADHD and reading and/or spelling
disorder is more consistently reported and seems somewhat higher than for mathematical disorder (e.g. Schuchardt, Fischbach,
Balke-Melcher, & Mähler, 2015; Schulte-Körne & Haberstroh, 2018). The majority of studies concentrate on dyslexia instead of
dyscalculia, and estimates of comorbidity rates vary with the framework of primary and conditional (i.e. comorbid) diagnosis: DuPaul,
Gormley, and Laracy (2013) reported a comorbid learning disorder in 31–45 percent of ADHD students and vice versa. Mayes and
Calhoun (2006) started with ADHD and found that every fourth of respective children (26 %) also fulfilled criteria of developmental
dyscalculia. Departing from the learning disorder, Willcutt et al. (2013) reported an ADHD pattern (all subtypes) in 28 % of the
children with a math disability, 32 % of the children with a reading disability and even 37 % of the children with a combined reading
and math disability. An even higher rate of ADHD in children with a mathematical learning disability was found by Desoete (2008).
The phenocopy hypothesis represented one of the first attempts to explain this co-occurrence beyond chance (Pennington, Groisser,
& Welsh, 1993). It proposed that reading problems associated with dyslexia could cause a phenotypic manifestation of ADHD: Poor
readers would show symptoms of ADHD (i.e. inattention, hyperactivity, impulsivity) without sharing the particular cognitive deficits
associated with ADHD (e.g., deficits in working memory and executive function). Despite early support for the phenocopy hypothesis,
more recent studies provide strong evidence that comorbidities between neurodevelopmental disorders are not artificial (e.g.,
Friedman, Chhabildas, Budhiraja, Willcutt, & Pennington, 2003; Landerl & Moll, 2010; Peterson et al., 2017). The multiple cognitive
deficit model (Pennington, 2006) postulates that each developmental disorder has its own profile of risk factors (both etiologic and
cognitive), with some risk factors being unique to the disorder and others being shared by another disorder. Studies on the comorbidity
between ADHD and dyslexia corroborate this model (e.g., Peterson et al., 2017; Willcutt et al., 2010). McGrath et al. (2011), for
example, showed that weakness in phonological awareness was a unique predictor of reading disability and weakness in response
inhibition was a unique predictor of ADHD. Processing speed was a shared predictor of both reading disability and ADHD and
accounted for their comorbidity.
To date, only few studies have investigated which cognitive risk factors could account for the comorbidity between ADHD and
dyscalculia. The multiple pathways model of ADHD suggested by Sonuga-Barke (2005) proposes that two uncorrelated processes,
namely executive/inhibitory deficits and delay aversion, each contribute to the prediction of ADHD. Findings showing a stable
relationship between central executive functions, including working memory, and mathematical performance (Friso-van den Bos, Van
der Ven, Kroesbergen, & Van Luit, 2013; Peng, Namkung, Barnes, & Sun, 2016) suggest that weaknesses in executive functions might
be a shared risk factor that has an effect on both ADHD and mathematical difficulties. Basic numerical processing is closely related to
mathematical skills (De Smedt, Noël, Gilmore, & Ansari, 2013) and deficits in numerical processing are often assumed to be a unique
risk factor for dyscalculia that is not shared by other developmental disorders. However, only few studies have investigated whether
children with ADHD show numerical processing deficits to provide empirical support for this assumption.

2
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

1.2. Numerical processing in children with dyscalculia

The 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2013)
defines dyscalculia as a specific learning disorder affecting calculation, arithmetic fact retrieval, and basic numerical processing.
Studies have shown that basic numerical skills, such as comparing magnitudes or numbers, writing numbers, ordering numbers by size,
or estimating the position of a number on a number line, developmentally precede and predict children’s calculation and arithmetic
skills (Hawes, Nosworthy, Archibald, & Ansari, 2019; Von Aster & Shalev, 2007). In addition, children with dyscalculia are impaired in
a variety of tasks assessing number and magnitude understanding (e.g., dot enumeration, Reigosa-Crespo et al., 2012; number line
estimation, Geary et al., 2008; writing numbers to dictation [transcoding], Moura et al., 2013; symbolic magnitude comparison,
Landerl et al., 2004; arithmetic fact retrieval, Geary et al., 2012).

1.3. Numerical processing in children with ADHD

ADHD is characterized by a persistent and age-inappropriate pattern of inattention and / or hyperactive / impulsive behavior
(DSM-5; American Psychiatric Association, 2013). Studies have shown that ADHD is associated with poor academic achievement,
including lower grades and lower test scores in mathematics (Loe & Feldman, 2007; Tosto, Momi, Asherson, & Malki, 2015). This
seems to be caused by executive function deficits such as weaknesses in working memory (Daley & Birchwood, 2010; Friedman,
Rapport, Orban, Eckrich, & Calub, 2018) and by difficulties selecting efficient strategies to accomplish arithmetic tasks (Sella, Re,
Lucangeli, Cornoldi, & Lemaire, 2019). It is not clear, however, whether children with ADHD show impairments in basic numerical
skills that might contribute to their mathematical difficulties. Only a very limited number of studies have investigated basic numerical
skills in samples of children with ADHD and their results are inconsistent.
A few studies observed that children with ADHD performed poor on basic numerical tasks. For example, Zentall, Smith, Lee, and
Wieczorek (1994) found that boys with ADHD were slower and less accurate in number recognition and transcoding compared with
boys without ADHD. Colomer et al. (2013) reported that substantial proportions of children with ADHD displayed severe impairments
(-2 standard deviations [SD] below age mean [M]) in standardized tasks of counting (36 %), transcoding (18 %), and arithmetic fact
retrieval (11.5 %). Only four percent of the sample was severely impaired (− 2 SD below age mean) in numerical knowledge, which was
conceptualized as a composite score of different subtasks including number comparison (symbolic magnitude comparison) and
number ordering (arranging Arabic numbers from the greatest to the least, and vice versa). In contrast, Kaufmann and Nuerk (2008)
found that children with ADHD performed poorer than a non-ADHD control group on a number comparison task. The two groups did
not differ on other measures of number and magnitude processing assessed in this study (number line, counting, transcoding, dot
enumeration).
Other studies suggest that ADHD is not associated with difficulties in basic numerical processing. For example, Ganor-Stern and
Steinhorn (2018) showed that adults with ADHD are impaired in calculation processes that rely on working memory, but not on
estimation skills that are based on sense of magnitude. González-Castro, Cueli, Areces, Rodríguez, and Sideridis (2016) showed that
children with ADHD aged 6–9 years performed better on tasks assessing counting, quantity comparison, and number facts than
children with mathematics disorder did. The study design did not include an unimpaired control group.
Kuhn, Ise, Raddatz, Schwenk, and Dobel (2016) compared the performance of children with ADHD symptoms and poor mathe­
matics achievement (developmental dyscalculia, DD) in a double dissociation design including four groups (ADHD-only, DD-only, DD
+ ADHD, typically developing [TD]). Children with mathematics difficulties (DD-only, DD + ADHD) displayed deficits in all tasks
assessing basic numerical skills. Children with ADHD symptoms (ADHD-only, DD + ADHD) displayed selective difficulties in dot
enumeration but their performance on magnitude comparison and transcoding tasks did not differ from the non-ADHD group
(DD-only, TD). Children with DD + ADHD showed an additive combination of the deficits associated with DD and ADHD on most tasks,
which might be interpreted as evidence against a unique etiology of the comorbid group (Landerl, Göbel, & Moll, 2013). However, the
interpretation of the results was hindered by the fact that the test battery used for group assignment included tasks assessing basic
numerical skill. In addition, many of the studies summarized above are limited by their small sample size (e.g. total sample N < 35 in
studies by Colomer et al., 2013; Ganor-Stern & Steinhorn, 2018; Kaufmann & Nuerk, 2008).

1.4. The present study

The present study aims to investigate basic numerical skills in children with ADHD symptoms and/or mathematics difficulties (MD)
and typically developing children (control group, CG) in order to determine whether impairments in basic numerical skills are a unique
or shared risk factor for these disorders. We used a double dissociation design with four groups of children and hypothesized that
deficits in basic numerical skills are uniquely associated with MD. We expected children with MD, but not children with ADHD
symptoms, to show impairments in tasks assessing basic numerical processing. In addition, we expected to find additive impairments in
the comorbid group (ADHD + MD). Data was analyzed using both a classical analysis of variance (ANOVA) to test for group differences
and a Bayesian ANOVA to test for null effects.

3
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

2. Method

2.1. Sample and procedure

Eighty-six children (43 girls, 43 boys) aged 6–10 years were included in the study. They met the following inclusion criteria: a)
intelligence ≥ 70, b) grade level 2–4, c) German as first language or language of instruction since first grade, and d) absence of a mental
disorder other than ADHD or oppositional defiant disorder (ODD) (assessed with the FBB-SCREEN [see 2.3.4]). ODD was not defined as
an exclusion criterion since it shows high comorbidity with ADHD (Nock, Kazdin, Hiripi, & Kessler, 2007). None of the participants had
uncorrected visual or hearing impairments. Children were recruited at two study sites in Germany via newspaper advertisements,
flyers in schools and medical practices, and face-to-face contacts at the two sites.
All participants were offered a 20 € gift voucher and free participation in a computer-based training of number skills (CODY
training; Kuhn & Holling, 2014) as incentives. The study site Muenster (Institute of Psychology, University of Muenster, Germany)
recruited children for the MD group and the control group. The MD group was recruited via a newspaper article that included in­
formation on MD, the purpose of the study, and the incentives. The CG was recruited from local elementary schools for participation in
a longitudinal study on predictors of mathematical achievement (Kuhn, Schwenk, Souvignier, & Holling, 2019). Children who had at
least average mathematical abilities were called and asked to participate in the present study. The study site Cologne (University
Hospital Cologne, Germany) recruited children for the ADHD and the ADHD + MD group. Children were recruited from a university
outpatient clinic and via recruitment flyers in medical practices. The flyers included the purpose of the study, inclusion criteria for the
ADHD and MD groups, and information about the incentives.
The study protocol was approved by the appropriate ethics committee and all procedures were carried out in accordance with the
Declaration of Helsinki. After obtaining parental consent and an assent of the child, two testing sessions were scheduled on two
different days. Tasks assessing intelligence (CFT, see 2.3.1), mathematical ability (see 2.3.5), reading ability (see 2.3.6), and arithmetic
fact retrieval (see 2.3.7) were administered in the first session. Children who were recruited at study site Cologne were tested indi­
vidually. Children in the CG and the MD group were tested in groups of 2–6 children by staff of the study site Muenster. All children
administered the six basic numerical skills subtests (subtests 1–6, see 2.3.7) individually on a computer during the second session.
While children performed the basic numerical skills subtests, we asked parents or caregivers to participate in a semi-structured
interview assessing the diagnostic criteria of ADHD. If the interview could not be conducted (e.g., due to parents not accompa­
nying the child, or parents needing to take care of younger siblings), we asked parents to participate in a telephone interview. If this
was refused, information in ADHD was extracted from behavioral ratings.
Children who received medication to treat ADHD (psychostimulants, n = 18) performed all tests on medication-free days to ensure
that the dependent variables were tested under the same conditions in all participants. Seven of the children who received stimulant
treatment (39 % of 18) performed poor on the mathematical ability test (DEMAT, T score < 40). Since stimulant treatment has been
found to improve academic test performance (Kortekaas-Rijlaarsdam, Luman, Sonuga-Barke, & Oosterlaan, 2019), these seven chil­
dren performed the DEMAT again under medication. When tested under medication, four children showed average or above-average
performance on the DEMAT (T scores ≥ 40) and three children showed below-average performance (T score < 40). To increase the
validity of the group assignment, DEMAT performance under medication was used for group allocation.

2.2. Group allocation

Children were allocated into one of four groups depending on their mathematical ability and the severity of ADHD symptoms.
Children in the ADHD group (n = 33) had to meet the following criteria: a) average or above-average mathematical skills (T score ≥ 40)
and b) a clinical diagnosis of ADHD (confirmed by the semi-structured interview DCL-ADHS, see 2.3.2) (n = 28) or, in the absence of
interview data, a scale score ≥1.5 on the Inattention scale (IN) and/or the Hyperactivity-Impulsivity scale (HY-IMP) of the parent
rating scale FBB-ADHS (see 2.3.3) (n = 5). The cut-off of 1.5 for the ADHD group was derived from data showing a mean of M = 1.53
(SD = 0.59) for FBB-ADHS scores in a large sample of children with a clinical diagnosis of ADHD (see 2.3.3). Children were assigned to
the mathematics difficulties (MD) group (n = 16) if they a) showed below-average mathematical skills (T score < 40) and b) did not
display symptoms of ADHD (scale score <1.0 on both scales of the FBB-ADHS). The definition of MD used in this study did not require a
substantial IQ-math achievement discrepancy. Children were assigned to the ADHD + MD group (n = 19) if they a) showed below-
average mathematical skills (T score < 40) and b) had a clinical diagnosis of ADHD (confirmed by DCL-ADHS, n = 10) or, in the
absence of interview data, a scale score ≥1.5 on the IN scale and/or the HY-IMP scale of the FBB-ADHS (n = 9). Children in the control
group (n = 18) fulfilled the following criteria: a) average or above-average mathematical skills (T score ≥ 40) and b) score <1.0 on both
scales of the FBB-ADHS (n = 7) or, in the absence of a completed FBB-ADHS, a value of 0 (“Not at all”) on all three ADHD screening
items of the FBB-SCREEN (n = 11).

2.3. Measures

2.3.1. Intelligence (CFT)

Depending on children’s age, either the short version of the Grundintelligenztest Skala 1 Revision (CFT 1-R, Engl.: Culture Fair In­
telligence Test CFT 1-R; Weiß & Osterland, 2013) or the short version of the Grundintelligenztest Skala 2-Revision (CFT 20-R, Engl.:
Culture Fair Intelligence Test CFT 20-R; Weiß, 2006) were administered. The manuals report high parallel form reliabilities (CFT 1-R: r
= .85–.91; CFT 20-R: r = .80–.82) and several analyses that support the tests’ internal, external and factorial validity.

4
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

2.3.2. Clinical interview for ADHD (DCL-ADHS)

The Diagnose-Checkliste für Aufmerksamkeitsdefizit − /Hyperaktivitätsstörungen (DCL-ADHS, Engl.: Diagnostic Checklist for ADHD)
assesses the criteria for ADHD according to the DSM-5 and the 10th edition of the International Classification of Diseases (ICD-10;
World Health Organization, 1994). Psychologists and experienced psychology students completed the checklist based on a
semi-structured interview with one or both parents (Interviewleitfaden für Externale Störungen [ILF-EXTERNAL], Engl.: Interview
guideline for externalizing disorders). DCL-ADHS and ILF-EXTERNAL are both part of the German DSM-5 and ICD-10-based Diagnostic
System for the Assessment of Mental Disorders in Children and Adolescents (DISYPS-III, Döpfner & Görtz-Dorten, 2017). The manual
reports high internal consistencies for the DCL-ADHS (Cronbach’s α = .89–.95) and satisfactory external validity (correlations with
ADHD rating scales: r = .45–.75).

2.3.3. ADHD parent rating scale (FBB-ADHS)

The rating scale Fremdbeurteilungsbogen ADHS (FBB-ADHS, Engl.: Parent rating scale ADHD) is also part of DISYPS-III (Döpfner &
Görtz-Dorten, 2017). The scale consists of 20 items that describe specific ADHD symptom behavior (e.g., “Is easily distracted”, “Blurts
out answers”). Parents indicate their agreement with each item on a four-point Likert scale ranging from 0 (not at all) to 3 (very much).
There is a 9-item Inattention scale (IN) and an 11-item Hyperactivity-Impulsivity scale (HY-IMP). Scale scores are computed by
averaging the responses to items on each scale, with higher scores indicating worse problem behavior. Research has shown that the
FBB-ADHS is a reliable (internally consistent, Cronbach’s α = .73–.90) instrument whose two-component structure has been confirmed
by factor analyses (Erhart, Döpfner, Ravens-Sieberer, & The Bella Study Group, 2008). The manual (Döpfner & Görtz-Dorten, 2017)
reports a mean scale score of M = 1.53 (SD = 0.59) of parent ratings (FBB-ADHS total scale) in a large clinical sample of children with a
clinical diagnosis of ADHD (n = 1,294) and a mean scale of score of M = 0.54 (SD = 0.52) in a community sample (n = 713). Another
recent publication demonstrated that the FBB-ADHS differentiates reliably between healthy children and children with a clinical
diagnosis of ADHD (AUC = .88; Hamadache, Günther, Hoberg, & Zaplana Labarga, 2020).

2.3.4. ADHD screening (FBB-SREEN)

The parent-rated screening instrument FBB-SCREEN (Fremdbeurteilungsbogen-Screening [Engl.: Proxy rating scale Screening];
Döpfner & Görtz-Dorten, 2017) comprises 35 items that assess a wide range of symptom behavior in children and adolescents. Three
items assess ADHD symptom behaviors (inattention, hyperactivity, impulsivity). Items were selected based on the diagnostic criteria of
the DSM-5 and ICD-10. In accordance with the FBB-ADHS, each screening item is rated on a four-point Likert scale ranging from 0 (not
at all) to 3 (very much). The manual reports satisfactory internal consistency for the ADHD subscale (Cronbach’s α = .78).

2.3.5. Mathematical ability (DEMAT)

Depending on their grade level, children completed either the subtests assessing arithmetic skills of the Deutscher Mathematiktest für
zweite Klassen (DEMAT 2+, Engl.: German mathematics test for second grades; Krajewski, Liehm, & Schneider, 2004), the Arithmetic
scale of the Deutscher Mathematiktest für dritte Klassen (DEMAT 3+, Engl.: German mathematics test for third grades; Roick, Gölitz, &
Hasselhorn, 2004) or the Arithmetic scale of the Deutscher Mathematiktest für vierte Klassen (DEMAT 4, Engl.: German mathematics test
for fourth grades; Gölitz, Roick, & Hasselhorn, 2006). The DEMAT tests are standardized, curriculum-based measures. All subtests are
timed (1.5–4 min). Raw scores were transformed to standardized T scores (M = 50, SD = 10). If the arithmetic subtests of the DEMAT
2+ were administered, a mean T score calculated by averaging the T scores of the five subtests. If DEMAT 3+ or DEMAT 4 was
administered the T score of the Arithmetic scale was used for group allocation.
The arithmetic subtests of DEMAT 2+ include Addition and Subtraction (eight two-digit addition and subtraction tasks, e.g. “? + 15
= 34”; time limit: 4 min), Duplication (three tasks that ask children to double two-digit numbers, e.g. “What is the double of 43”; time
limit: 1.5 min), Division (three two-digit division tasks, e.g. “24:4 =? ; time limit: 2.5 min), Divide in Half (three tasks that ask children
to divide a number in half, e.g. “What is half of 26?”; time limit: 1.5 min), and Money (four calculation tasks with money, e.g., “45 Cent
+? = 1 Euro”; time limit: 2.5 min). The Arithmetic scale of DEMAT 3+ includes the subtests Number line (three tasks that ask children
to determine the three-digit number that lies exactly between two given numbers; time limit: 3 min), Addition (four three-digit
addition tasks, e.g. “106 + 54? =???”; time limit: 3 min), Subtraction (four three-digit subtraction tasks, e.g. “763 + 356 =???”;
time limit: 3 min), and Multiplication (four two to three-digit multiplication tasks, e.g. “24 × 6 =?” ; time limit: 3 min). The Arithmetic
scale of DEMAT 4 includes the subtests Number line (three tasks that ask children to choose out of five options the multidigit number, e.
g. “9000”, that is marked on a number line; time limit: 1.5 min), Addition (four multidigit addition tasks, e.g. “1066 + 5574 =?” ; time
limit: 3 min), Subtraction (four multidigit subtraction tasks, e.g. “–7634–3565 =?” ; time limit: 3 min), Multiplication (four multidigit
multiplication tasks, e.g. “20240 × 4 =?” ; time limit: 3.5 min), and Division (four multidigit division tasks, e.g. “1007 : 4 =?” ; time
limit: 3.5 min).
The manuals report research showing that the subtests used in the present study are internally consistent (DEMAT 2+: Cronbach’s α
= .63–.87; DEMAT 3+: Cronbach’s α = .76–.79; DEMAT 4: Cronbach’s α = .77–81) and that correlations between DEMAT scores and
math grades are high (r = .61–.70), thus supporting the external validity of the DEMAT tests.

2.3.6. Reading ability (SLS 2-9)

The reading test Lese-Screening für die Klassenstufen 2-9 (SLS 2-9, Engl.: Salzburg reading screening test for grades 2-9; Wimmer &
Mayringer, 2014) is a standardized measure of reading ability. Children silently read sentences with simple semantic and syntactic
structure (e.g., “Trees can speak.”) and mark each sentence as right or wrong by circling a checkmark or a cross. There is time limit of 3
min. Raw scores are transformed in a standardized reading quotient (M = 100, SD = 15). The manual reports high parallel form

5
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

reliability (r = .87–.95) and satisfactory external validity.

2.3.7. Basic numerical skills

Basic numerical skills were assessed with eight different subtests. Six subtests (1-6) were derived from the computer-administered
task battery CODY-Mathetest für die 2.-4. Klasse (CODY math assessment for second to fourth grades; Kuhn, Schwenk, Raddatz, Dobel, &
Holling, 2017). Each CODY task started with practice items with feedback, followed by test items. Instructions were presented au­
ditorily using headphones. The manual reports good test-retest reliabilities (r = .72–.76) and satisfactory external validity. Subtests 7
and 8 were taken from the standardized paper-and-pencil arithmetic test Diagnostisches Inventar zu Rechenfertigkeiten im Grundschulalter
(DIRG, Engl.: Diagnostic inventory of arithmetic skills at elementary school age; Grube, Weberschock, Blum, & Hasselhorn, 2010). The
test items are preceded by two practice items. The manual reports high test-retest reliabilities (ADD FACTS: r = .85; SUB FACTS: r =
.86) and satisfactory external validity.
1) Dot enumeration. One to nine black dots were presented in the middle of the screen. Participants were supposed to enumerate
them as quickly as possible, and then to press the corresponding number key on the keyboard. The test included 18 items (time limit: 2
min). The median of correct answers was used as test score.
2) Number comparison. Two single-digit numbers were presented left and right on screen. Participants were asked to decide as
quickly as possible which number was larger by pressing one of two response keys. Numerical distance between stimuli was sys­
tematically manipulated to be small (1–3) or large (4–6). The test included 24 items (time limit:1.5 min). The median of correct
answers was used as test score.
3) Mixed comparison. The setup was identical to number comparison, except that a single-digit number and a cloud of one to nine
dots were presented on screen.
4) Transcoding. Participants heard numbers and were asked to type them using the keyboard. Each number could be heard a
maximum of two times. The test consisted of eight class level-specific items (time limit: 2 min). The number of correct items was used
as test score.
5) Number sets. Participants saw a target number on screen. Below the target, a number set consisting of two or three cells was
shown. Each cell showed a structured set of objects or a number. Participants had to decide as quickly as possible by pressing one of two
response keys whether the total magnitude in the number set matched the target number. Two target numbers (5 and 9) were used for
1.5 min each. This test was a speed test, and the number of correct answers minus the number of false alarms was used as test score.
6) Number line estimation. Participants saw a number between 1 and 99 on screen. Below, a number line with endpoints 0 and 100
was shown. Participants were supposed to click onto the number line at the position that corresponded to the shown target number.
This test consisted of 23 items (time limit: 3.5 min). Mean deviation between target number and answer was taken as test score.
7) Arithmetic fact retrieval – Addition (ADD FACTS). Single-digit addition problems were presented and participants were asked
to solve as many as possible within a time limit of 2 min.
8) Arithmetic fact retrieval – Subtraction (SUB FACTS). The setup was identical to ADD FACTS, except that subtraction problems
were used.

2.4. Statistical analyses

First, we checked ANOVA assumptions, using the Levene test for assessing heteroscedasticity and the Shapiro-Wilks test to check
whether residuals were normally distributed. Next, group differences in basic numerical skills were analyzed by 2 × 2 analyses of
variance (ANOVA) with two between-subjects variables (ADHD: yes/no, MD: yes/no), and the basic numerical skill tasks as the
dependent variables. A significant main effect of ADHD for a given basic numerical task would indicate that the performance of
children with ADHD (ADHD, ADHD + MD) differs significantly from the performance of children without ADHD (CG, MD). A sig­
nificant main effect of MD would indicate that the performance of children with MD (MD, ADHD + MD) differs significantly from the
performance of children without MD (CG, ADHD). The absence of an interaction would suggest that the effects of ADHD and MD are
additive and statistically independent (i.e. the performance of the ADHD + MD group results additively from those of the ADHD and
MD groups). The 2 × 2 ANOVAs were performed using the R software (R Core Team, 2019), and the Bonferroni–Holm correction was
used to correct for type I errors (Holm, 1979).
Because we theoretically expected null results concerning several parameters in the ANOVA model (ADHD and interaction effects),
we utilized Bayesian ANOVAs and hypothesis testing as a supplement to null hypothesis significance (NHST) ANOVA in order to
investigate whether evidence for null effects could be found. The Bayesian statistical framework offers several advantages in com­
parison to classical NHST-based ANOVAs. For example, hypothesis testing in a Bayesian framework allows to directly quantify evi­
dence for two competing hypotheses or models, i.e. the null and alternative hypothesis. In contrast, p-values in a NHST framework
simply provide the probability for finding a more extreme test statistic than the observed one, given that the null hypothesis is true.
Hence, no evidence for the alternative hypothesis, but only against the null hypothesis is provided. In Bayesian hypothesis testing, null
and alternative models can be directly compared (e.g., using Bayes factors; Wagenmakers et al., 2018), and a decision can be based on
the strength of evidence found. Bayes factors indicate the degree to which a null or an alternative model are more plausible when
taking the observed data into account (change from prior to posterior odds). A Bayes factor of 3:1 or larger is regarded as moderate
evidence for the alternative hypothesis, whereas a Bayes factor of 1:3 or less is regarded as moderate evidence for the null hypothesis
(Lee & Wagenmakers, 2013). Bayes factor values between 1:3 and 3:1 are regarded as anecdotal, providing no clear evidence for either
null or alternative hypothesis. Hence, Bayesian hypothesis testing can provide evidence for the null hypothesis, which is not possible in
NHST settings (Rouder, Morey, Speckman, & Province, 2012), but it can also show that evidence is too inconclusive to make a decision

6
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

between two hypotheses.

We used the R software (R Core Team, 2019), version 3.6.0 as an environment to run the BayesFactor package (Morey et al., 2018)
to conduct Bayesian ANOVAs and hypothesis testing. We used standard priors for Bayesian ANOVA factorial designs as described in
Rouder et al. (2012) for all analyses (r scale value = 0.5). For each Bayesian ANOVA, 1,000,000 Monte Carlo simulations were
generated. To conduct a sensitivity analysis, we repeated all Bayesian analyses using the brms package (Bürkner, 2017), which samples
based on a Hamiltonian Monte Carlo algorithm. Informative priors distributions chosen were N(5,10) for main effect MD, N(0,10) for
main effect ADHD, N(0,10) for the interaction effect, N(40,20) for the intercept and Cauchy(0,10) for the error variance. Because all
dependent variables were based on the T scale (M = 50, SD = 10), these prior distributions were regarded as informative and in line
with theoretical expectations. We also checked chain convergence for all parameters visually and statistically in order to follow
guidelines for reporting Bayesian analyses (Depaoli & Van de Schoot, 2017).

3. Results

3.1. Sample characteristics

Table 1 displays group means and standard deviations for children’s age, mathematical ability, ADHD symptom score, intelligence,
and reading ability. It also shows group means (M) and standard deviations (SD) of the subtests assessing basic numerical skills.
Analyses of variance (ANOVA) with the between-subjects variable group (CG/MD/ADHD/ADHD + MD) revealed no significant group
differences regarding children’s age (p = .62), intelligence (p = .43), or reading skills (p = .29). Eleven children attended grade 2 (CG: n
= 2, MD: n = 3, ADHD: n = 2, ADHD + MD: n = 4), 36 children attended grade 3 (CG: n = 6, MD: n = 4, ADHD: n = 19, ADHD + MD: n =
7), and 39 children attended grade 4 (CG: n = 10, MD: n = 9, ADHD: n = 12, ADHD + MD: n = 8).

3.2. Results of 2 × 2 NHST ANOVA

We found that the homoscedasticity assumption was met for all measures (all Holm-Bonferroni adjusted p > .05), but the
assumption of normally-distributed residuals was violated in one measure (number comparison). However, because ANOVA is robust
with respect to this finding (Schmider, Ziegler, Danay, Beyer, & Bühner, 2010), we decided to use the regular ANOVA approach.
Table 2 provides the results of the 2 × 2 NHST ANOVA. Children with MD performed significantly worse than children without MD
on the Transcoding task (F(1,80) = 13.73, p < .01, Cohen’s d = − 0.81), on Number sets (F(1,79) = 16.44, p < .01, Cohen’s d = − 0.80),
and on the SUB FACTS task (F(1,80) = 18.40, p < .01, Cohen’s d = − 0.86). Children with MD also performed worse than children
without MD did on the ADD FACTS task (Cohen’s d = − 0.46), but this effect failed to reach the level of significance after the
Bonferroni-Holm correction. There were no significant main effects of MD on the dependent variables Dot enumeration (Cohen’s d =
0.15), Number comparison (Cohen’s d = − 0.22), Mixed comparison (Cohen’s d = − 0.10), or Number line estimation (Cohen’s d =
0.05).
There were no significant differences between the ADHD and the non-ADHD group on the dependent variables Dot enumeration

Table 1
Descriptive Statistics.
CG (n = 18) MD (n = 16) ADHD (n = 33) ADHD + MD (n = 19)

Measure M (SD) M (SD) M (SD) M (SD)

Age in years 9.3 (0.7) 9.0 (1.0) 9.0 (0.8) 9.0 (1.2)
Math (DEMATa) 47.3 (6.0) 32.7 (5.4) 45.6 (5.5) 34.7 (3.5)
ADHD symptom score (FBB-ADHDb)
Total scale 0.5 (0.2) 0.4 (0.2) 1.3 (0.6) 1.3 (0.4)
Inattention 0.8 (0.3) 0.7 (0.3) 1.6 (0.6) 1.8 (0.4)
Hyperactivity-Impulsivity 0.2 (0.4) 0.2 (0.2) 1.1 (0.7) 0.9 (0.6)
Intelligence (CRFc) 99.0 (11.6) 92.9 (9.7) 99.3 (11.9) 96.5 (12.7)
Reading (SLS 2-9d) 92.5 (10.6) 93.6 (12.4) 91.0 (14.4) 97.9 (12.3)
Basic numerical skills
Dot enumerationa 46.8 (11.3) 44.7 (10.0) 42.0 (9.1) 46.0 (15.9)
Number comparisona 54.9 (7.9) 46.7 (16.0) 48.5 (11.8) 49.3 (10.1)
Mixed comparisona 49.1 (9.4) 48.3 (16.2) 51.4 (9.8) 50.4 (15.0)
Transcodinga 45.0 (10.9) 33.6 (8.4) 46.9 (10.4) 41.6 (9.7)
Number setsa 50.9 (7.7) 39.3 (11.3) 43.5 (10.9) 36.2 (10.3)
Number line estimationa 46.1 (10.9) 43.2 (9.8) 41.3 (9.1) 43.7 (9.0)
ADD FACTSa 41.4 (11.5) 34.6 (8.1) 36.1 (11.2) 32.2 (8.2)
SUB FACTSa 42.5 (11.8) 28.7 (6.4) 36.6 (11.8) 31.1 (6.4)

Note. CG = Control group; MD = Mathematics difficulties; ADHD = Attention-deficit/hyperactivity disorder; ADD FACTS = Arithmetic fact retrieval –
Addition; SUB FACTS = Arithmetic fact retrieval – Subtraction.
a
T-values (M = 50, SD = 10).
b
Scale scores (range 0–3).
c
Intelligence quotient (M = 100, SD = 15).
d
Reading quotient (M = 100, SD = 15).

7
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

Table 2
Results of 2 × 2 NHST ANOVA.
Measure ADHD MD ADHD*MD

Dot enumeration F(1,80) = 0.45, p= .504 F(1,80) = 0.13, p = .723 F(1,80) = 1.37, p= .245
Number comparison F(1,80) = 0.50, p= .480 F(1,80) = 1.97, p = .165 F(1,80) = 2.82, p= .097
Mixed comparison F(1,80) = 0.63, p= .432 F(1,80) = 0.11, p = .743 F(1,80) = 0.00, p= .983
Transcoding F(1,80) = 4.82, p= .031 a F(1,80) = 13.73, p < .001 F(1,80) = 1.80, p= .184 MD < non-MD
Number sets F(1,79) = 4.96, p= .029a F(1,79) = 16.44, p < .001 F(1,79) = 0.87, p= .353 MD < non-MD
Number line estimation F(1,80) = 1.00, p= .321 F(1,80) = 0.02, p = .904 F(1,80) = 1.51, p= .223
ADD FACTS F(1,82) = 2.89, p= .093 F(1,82) = 5.55, p = .021a F(1,82) = 0.41, p= .525
SUB FACTS F(1,80) = 0.59, p= .444 F(1,80) = 18.40, p < .001 F(1,80) = 3.41, p= .069 MD < non-MD

Note. ADD FACTS = Arithmetic fact retrieval – Addition; SUB FACTS = Arithmetic fact retrieval – Subtraction.
a
Not significant after the Bonferroni-Holm correction.

(Cohen’s d = − 0.20), Number comparison (Cohen’s d = − 0.18), Mixed comparison (Cohen’s d = 0.19), Transcoding (Cohen’s d =
0.51), Number sets (Cohen’s d = − 0.39), Number line estimation (Cohen’s d = − 0.26), ADD FACTS (Cohen’s d = − 0.34), or SUB
FACTS (Cohen’s d = − 0.12) after the Bonferroni-Holm correction. No interaction effect (ADHD x MD) was found for any of the
measures. Fig. 1 shows bars charts of group means with error bars.

3.3. Results of Bayesian ANOVA

Visual inspection of convergence plots and statistical measures suggested convergence of Markov Chain Monte Carlo chains (all
Rhat < 1.10; Gelman & Rubin, 1992). Results from both Bayesian analyses (using packages BayesFactor and brms) were comparable,
with brms results comprising tendentially smaller Bayes factors. We report only results using the BayesFactors package here.
The Bayesian factorial ANOVAs (see Table 3) revealed that there was strong evidence for a main effect of MD on SUB FACTS,
Transcoding and Number sets, indicating impairments in these tasks in children with MD. In contrast, there was evidence that MD did
not affect performance in the Dot enumeration, Number comparison, or Mixed comparison tasks. No main effect of ADHD on numerical
tasks emerged in the Bayesian analysis, there was rather clear evidence in several tasks for a null effect (SUB FACTS, Dot enumeration,
Number comparison, Mixed comparison). The results did not change when we compared the full main effects model (MD + ADHD)
with the single-effects models (MD or ADHD, respectively, see columns 4 and 5 in Table 3). Concerning a possible interaction effect of
MD and ADHD, no clear evidence could be found: Although most Bayes factors were smaller than 1, favoring the null hypothesis
against an interaction effect, they were larger than 1:3, indicating only weak evidence. We could rule out an interaction effect only in
the case of mixed comparisons.
Finally, we investigated which ANOVA model described the data best. In the case of SUB FACTS, Transcoding, and Number sets, a
MD-only model was the best description of the data. In the case of Dot enumeration, clearly a null model (containing a single intercept
only) described the data best. In the other cases, the null model represented the data best, because there was only weak evidence for a
better-fitting model.
To summarize, Bayesian ANOVAs indicated that children with MD were impaired in more complex number processing tasks and
retrieval of subtraction facts, whereas no impairment was found in basic core markers of numerical skills and addition facts. Further,
no effects of ADHD on any task were found, either because we found direct evidence for the lack of an effect, or because evidence was
inconclusive. Finally, we found no support for an interaction between MD and ADHD in our data.

4. Discussion

The current study investigated basic numerical skills in four groups of children (ADHD, MD, ADHD + MD, typically developing) to
determine whether impairments in basic numerical skills are a unique risk factor for MD or a shared risk factor that could help to
explain the association between MD and ADHD. As we predicted based on past research, children with MD showed poor numerical
processing. More specifically, both classical (NHST) and Bayesian ANOVA showed that children with MD performed worse on tasks
assessing complex number processing (writing number to dictation [Transcoding]: Cohen’s d = − 0.81, comparing numbers and
magnitudes [Number sets]: Cohen’s d = − 0.80) and arithmetic fact retrieval (subtraction fact retrieval [SUB FACTS]: Cohen’s d = −
0.86) compared to children without MD. The observed impairments are substantial (Cohen’s d between − 0.80 and − 0.86) and can
possibly explain the mathematical difficulties experienced by children with MD.
In contrast to previous studies (e.g., Geary et al., 2008; Reigosa-Crespo et al., 2012), children with MD were not impaired on tasks
assessing core markers of numeracy (dot enumeration, simple magnitude comparison tasks, number line estimation; Cohen’s d be­
tween − 0.22 and 0.15). Bayesian ANOVA clearly showed that the lack of a significant MD vs. non-MD group difference for these core
markers was indeed based on a null effect. A possible explanation is that impairments in core markers are often found in severe forms of
mathematics disorder (Mazzocco, Feigenson, & Halberda, 2011). Results in our sample may have been affected by our decision to
include children with mathematical difficulties based on their performance on arithmetic tasks instead of studying children with a
clinically confirmed diagnosis of dyscalculia. The MD subsample, as a group, showed considerable deficits in their mathematical skills
(MD group: T score M = 32.7; ADHD + MD group: T score M = 34.7; see Table 1). However, there was heterogeneity in our MD sample
and, hence, not all children display severe mathematical deficits.

8
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

Fig. 1. Group comparisons for basic numerical tasks.

Note. Results are presented as mean scores (T scores: M = 50, SD = 10). ADD FACTS = Arithmetic fact retrieval – Addition; SUB FACTS = Arithmetic
fact retrieval – Subtraction; CG = Control group; MD = Mathematics difficulties; ADHD = Attention deficit-/hyperactivity disorder.

Our results further suggest that ADHD is not associated with deficits in basic numerical skills across a range of different tasks. There
were no significant ADHD vs. non-ADHD group differences in the NHST ANOVA (Cohen’s d between − 0.34 and 0.51). Similarly, no
effects of ADHD symptoms were found in the Bayesian ANOVA, either because there was direct evidence for the lack of an effect, or
because evidence was inconclusive. This finding contrasts with previous evidence of selective deficits in some numerical processing
tasks (e.g. Colomer et al., 2013; Kaufmann & Nuerk, 2008), and supports recent studies that found no impairments in numerical
processing in ADHD samples (e.g., Ganor-Stern & Steinhorn, 2018). Therefore, the present finding provides empirical evidence for the
assumption that deficits in numerical processing are uniquely associated with MD. Within the framework of the multiple cognitive
deficit model of developmental disorders (Pennington, 2006), our results can be interpreted as supporting the hypothesis that deficits

9
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

Table 3
Model comparisons (Bayes factors) of group main effects and interactions based on Bayesian ANOVA.
Model comparisons
Measure MD vs. Null ADHD vs. Null ME mode ME mode Full model vs. Best model
model model vs. MD vs. ADHD ME model

Dot enumeration 0.28 0.33 0.32 0.27 0.53 Null model

Number comparison 0.36 0.31 0.33 0.39 0.93 (Null model)
Mixed comparison 0.25 0.32 0.33 0.26 0.29 (Null model)
Transcoding 64.56 2.19 1.47 43.33 0.57 MD
Number sets 49.07 0.87 2.39 135.36 0.43 MD
Number line estimation 0.24 0.42 0.42 0.23 0.59 (Null model)
ADD FACTS 1.47 0.65 0.93 2.11 0.36 (Null model)
SUB FACTS 114.34 0.26 0.35 152.65 1.13 MD

Note. Bayes factors of model comparisons, comparing unrestricted (mentioned first) and restricted models (mentioned second), respectively. MD vs.
Null model: Main effect MD vs. empty null model (intercept only); ADHD vs. Null model: Main effect ADHD vs. empty null model (intercept only); ME
model vs. MD: both main effects model (MD + ADHD) vs. main effect MD model; ME model vs: ADHD: both main effects model (MD + ADHD) vs.
main effect ADHD model; Full model vs. ME model: both main effects and interaction model (MD + ADHD + MD*ADHD) vs. both main effects model
(MD + ADHD). Bayes factors suggesting evidence for unrestricted model are in boldface. Bayes factors suggesting evidence for the restricted model
are in italics. Model designations in brackets indicate that the null model was assumed because only weak evidence for other models was found. ADD
FACTS = Arithmetic fact retrieval – Addition; SUB FACTS = Arithmetic fact retrieval – Subtraction.

in basic numerical skills are a unique risk factor for dyscalculia that is not shared by ADHD.
Finally, we investigated whether the cognitive profiles of the comorbid group result additively from the combination of the two
disorders. In line with prior results (e.g., Kuhn et al., 2016; Landerl, Fussenegger, Moll, & Willburger, 2009), our data showed that
interaction effects were absent in the NHST paradigm, suggesting additivity. Bayesian ANOVA revealed, however, that the strength of
evidence for assuming additivity of cognitive profiles was relatively weak as most Bayes factors were between 1:3 and 1 (range of
anecdotal evidence for H0). Hence, our results do not provide unequivocal support for the additivity assumption, but rather suggest
that more evidence is needed to clearly answer whether ADHD and MD have independent underlying cognitive causes.
There are several limitations worth mentioning. First, although an effort was made to include only children with a clinical ADHD
diagnosis based on a clinical interview, some children (n = 14) were included in the ADHD subsample based on clinically significant
ADHD symptom ratings. An independent samples t-test revealed that parents of children with a clinical diagnosis of ADHD (n = 38)
reported significantly higher symptom ratings on the FBB-ADHS than parents of children allocated to the ADHD group on the basis of
clinically significant ADHD symptom ratings (n = 14, t = 2.16, p = .04). However, further independent-samples t-tests demonstrated
that the subsample of children with available interview data (n = 38) performed significantly better on the subtests ADD FACTS, SUB
FACTS, Transcoding, and Number sets (all p < .05) than the subsample of children without interview data (n = 14). The two subgroups
did not differ significantly on the remaining basic numerical subtests. It is, therefore, unlikely that the lack of significant differences in
basic numerical skills between the ADHD and the non-ADHD group found in the present study can be attributed to the inclusion of
children without interview data in the ADHD sample. However, future studies should use multiple sources of information on ADHD
including clinical interviews and behavioral ratings by participants and significant others.
Second, some results of our study could be related to insufficient statistical power. An a priori power analysis for frequentist
ANOVAs showed that our design approximated a satisfactory power of .80 for medium and large effect sizes (Cohen‘s f > .30), whereas
power for smaller effect sizes was substantially reduced (e.g., power of .63 for Cohen’s f = .25). Importantly, Bayesian ANOVAs showed
that several statistically insignificant effects were indeed in line with evidence for the null hypothesis (e.g., four ADHD main effects),
while other insignificant results remained inconclusive, with Bayes factors between 3 and 1/3 (e.g., several interaction effects).
However, in most cases, evidence pointed towards a lack of ADHD effects or a lack of interaction between ADHD and MD effects.
Third, previous studies were criticized for using a test battery for group assignment that included tasks assessing basic numerical
skills (e.g., Kuhn et al., 2016). Group differences in basic numerical skills were therefore to be expected. In the present study, we
decided to use children’s performance on the DEMAT arithmetic subtests for group allocation, since these subtests focus on arithmetic
skills. Still, there is some overlap between the DEMAT subtests and the CODY tasks representing basic numerical skills. This is in part
due to the difficulty of constructing a task that assesses basic numerical skills without relying too heavily on other functions. The CODY
subtest Number sets, for example, contains an addition exercise, as children are required to sum a set of objects and a digit number.
Similarly, the Dot enumeration task used in the present study requires children to press the corresponding key on the keyboard, and
therefore tests number knowledge. The latter observation does not seem to influence the conclusions drawn from the present results, as
dot enumeration and number knowledge are both core numerical skills.
Another limitation worth mentioning is the fact that the tasks assessing intelligence, mathematical ability, reading ability, and
arithmetic fact retrieval were conducted either individually or in groups, depending on the study site. Group testing may negatively
influence test performance in children with ADHD or MD due to higher distractibility or test anxiety. Future studies should, therefore,
ensure that the same procedures for data collection are used in all participants.
Finally, the lack of clinical diagnosis in the MD groups, the heterogeneity of the sample in terms of maturational level and
mathematical ability, and the size of the sample limit the generalizability of the results to a specific subsample of children in
elementary school with mathematical difficulties and/or clinically significant levels of ADHD symptoms.

10
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

Despite these limitations, the present study provides clear support for the hypothesis that impairments in basic numerical skills are
uniquely associated with MD, and do not represent a shared risk factor for ADHD and MD. This finding has several implications. First, it
disconfirms previous reports of selective impairments in basic numerical skills in children with ADHD (e.g., Colomer et al., 2013;
Kaufmann & Nuerk, 2008; Zentall et al., 1994) and extends the results of a recent study demonstrating no difficulties in numerical tasks
in adults with ADHD (e.g., Ganor-Stern & Steinhorn, 2018) to children with clinically relevant ADHD symptoms. From a theoretical
perspective, our findings corroborate and refine the multiple cognitive deficit model of developmental disorders (Pennington, 2006) by
showing that deficits in numerical processing represent a unique risk factor for MD that is not shared by ADHD. Second, some tentative
clinical implications can be drawn from our findings. Since ADHD symptoms per se were not associated with impaired basic numerical
skills, difficulties in numerical tasks seem to be indicative of MD, even in children with clinically significant ADHD symptoms. In line
with the diagnostic criteria of the DSM-5 (American Psychiatric Association, 2013), which describe problems with number sense and
memorization of arithmetic facts as symptoms of dyscalculia, we suggest that an assessment of basic numerical skills should be an
integral part of the diagnostic evaluation of MD.
Clearly, more research with larger samples is needed to determine the diagnostic accuracy of tasks assessing complex number
processing. Future studies should investigate basic numerical skills in larger samples of children with clinically confirmed diagnoses of
ADHD and/or dyscalculia to be able to examine the diagnostic accuracy of tasks assessing basic numerical skills and to investigate the
influence of executive function deficits on numerical task performance.

CRediT authorship contribution statement

Elena von Wirth: Conceptualization, Methodology, Formal analysis, Project administration, Writing - original draft, Funding
acquisition. Katharina Kujath: Investigation, Writing - review & editing. Lea Ostrowski: Investigation, Writing - review & editing.
Ellen Settegast: Investigation. Sarah Rosarius: Investigation, Writing - review & editing. Manfred Döpfner: Conceptualization,
Methodology, Writing - review & editing. Christin Schwenk: Investigation, Writing - review & editing. Jörg-Tobias Kuhn:
Conceptualization, Methodology, Software, Formal analysis, Project administration, Writing - review & editing, Funding acquisition.

Declaration of Competing Interest

JTK is author of the CODY tasks used in the present study and receives royalties from the publisher. EvW, KK, LO, ES, SR, MD, and
CS are not aware of any biases that might be perceived as affecting the objectivity of this manuscript and declare that they have no
competing interests.

Acknowledgements

This work was supported by the Koeln Fortune Program/Faculty of Medicine, University of Cologne, Germany [grant KF150/
2015]; and the German Federal Ministry of Education and Research [grant no. 01GJ 1302].

References

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Washington, DC: American Psychiatric Association.
Bürkner, P.-C. (2017). Brms: An R package for Bayesian multilevel models using Stan. Journal of Statistical Software, 80(i01), 1–28. https://doi.org/10.18637/jss.v080.
i01.
Colomer, C., Re, A. M., Miranda, A., & Lucangeli, D. (2013). Numerical and calculation abilities in children with ADHD. Learning Disabilities: A Contemporary Journal,
11(2), 1–15. Retrieved from http://www.ldw-ldcj.org/.
Daley, D., & Birchwood, J. (2010). ADHD and academic performance: Why does ADHD impact on academic performance and what can be done to support ADHD
children in the classroom. Child: Care, Health and Development, 36(4), 455–464. https://doi.org/10.1111/j.1365-2214.2009.01046.x.
De Smedt, B., Noël, M. P., Gilmore, C., & Ansari, D. (2013). How do symbolic and non-symbolic numerical magnitude processing skills relate to individual differences
in children’s mathematical skills? A review of evidence from brain and behavior. Trends in Neuroscience and Education, 2, 48–55. https://doi.org/10.1016/j.
tine.2013.06.001.
Depaoli, S., & Van de Schoot, R. (2017). Improving transparency and replication in Bayesian statistics: The WAMBS-checklist. Psychological Methods, 22(2), 240–261.
https://doi.org/10.1037/met0000065.
Desoete, A. (2008). Co-morbidity in mathematical learning disabilities: Rule or exception? The Open Rehabilitation Journal, 1, 15–26. https://doi.org/10.2174/
1874943700801010015.
Döpfner, M., & Görtz-Dorten, A. (2017). Diagnostik-System für psychische Störungen nach ICD-10 und DSM-5 für Kinder- und Jugendliche III (DISYPS-III) [Diagnostic system
for mental disorders in child and adolescents according to ICD-10 and DSM-5 DISYPS-3]. Göttingen, Germany: Hogrefe.
DuPaul, G. J., Gormley, M. J., & Laracy, S. D. (2013). Comorbidity of LD and ADHD: Implications of DSM-5 for assessment and treatment. Journal of Learning
Disabilities, 46, 43–51. https://doi.org/10.1177/0022219412464351.
Erhart, M., Döpfner, M., Ravens-Sieberer, U., & The Bella Study Group. (2008). Psychometric properties of two ADHD questionnaires: Comparing the Conners’ scale
and the FBB-HKS in the general population of German children and adolescents – Results of the BELLA study. European Child and Adolescent Psychiatry, 17(Suppl.
1), 106–115. https://doi.org/10.1007/s00787-008-1012-1.
Friedman, M. C., Chhabildas, N., Budhiraja, N., Willcutt, E. G., & Pennington, B. F. (2003). Etiology of the comorbidity between RD and ADHD: Exploration of the non-
random mating hypothesis. American Journal of Medical Genetics Part B (Neuropsychiatric Genetics), 120B, 109–115. https://doi.org/10.1002/ajmg.b.20029.
Friedman, L. M., Rapport, M. D., Orban, S. A., Eckrich, S. J., & Calub, C. A. (2018). Applied problem solving in children with ADHD: The mediating roles of working
memory and mathematical calculation. Journal of Abnormal Child Psychology, 46(3), 491–504. https://doi.org/10.1007/s10802-017-0312-7.
Friso-van den Bos, I., Van der Ven, S. H. G., Kroesbergen, E. H., & Van Luit, J. E. H. (2013). Working memory and mathematics in primary school children: A meta-
analysis. Educational Research Review, 10, 29–44. https://doi.org/10.1016/j.edurev.2013.05.003, 2013.
Ganor-Stern, D., & Steinhorn, O. (2018). ADHD and math – The differential effect on calculation and estimation. Acta Psychologica, 188, 55–64. https://doi.org/
10.1016/j.actpsy.2018.05.013.

11
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

Geary, D. C., Hoard, M. K., & Bailey, D. H. (2012). Fact retrieval deficits in low achieving children and children with mathematical learning disability. Journal of
Learning Disabilities, 45(4), 291–307. https://doi.org/10.1177/0022219410392046.
Geary, D. C., Hoard, M. K., Nugent, L., & Byrd-Craven, J. (2008). Development of number line representations in children with mathematical learning disability.
Developmental Neuropsychology, 33(3), 277–299. https://doi.org/10.1080/87565640801982361.
Gelman, A., & Rubin, D. B. (1992). Inference from iterative simulation using multiple sequences. Statistical Science, 7(4), 457–472. https://doi.org/10.1214/ss/
1177011136.
Gölitz, D., Roick, T., & Hasselhorn, M. (2006). DEMAT 4: Deutscher Mathematiktest für vierte Klassen [German mathematics test for fourth grades]. Göttingen, Germany:
Hogrefe.
González-Castro, P., Cueli, M., Areces, D., Rodríguez, C., & Sideridis, G. (2016). Improvement of word problem solving and basic mathematics competencies in
students with attention deficit/hyperactivity disorder and mathematical learning difficulties. Learning Disabilities Research & Practice, 31(3), 142–155. https://doi.
org/10.1111/ldrp.12106.
Gross-Tsur, V., Manor, O., & Shalev, R. S. (1996). Developmental dyscalculia: Prevalence and demographic features. Develomental Medicine and Child Neurolology, 38
(1), 25–33. https://doi.org/10.1111/j.1469-8749.1996.tb15029.x.
Grube, D., Weberschock, U., Blum, M., & Hasselhorn, M. (2010). Diagnostisches Inventar zu Rechenfertigkeiten im Grundschulalter: DIRG [Diagnostic inventory of arithmetic
skills at elementary school age]. Göttingen, Germany: Hogrefe.
Hamadache, S., Günther, T., Hoberg, K., & Zaplana Labarga, S. (2020). Is the QbMini a valid instrument for ADHD assessment? Journal of Attention Disorders. https://
doi.org/10.1177/1087054720903361.
Hawes, Z., Nosworthy, N., Archibald, L., & Ansari, D. (2019). Kindergarten children’s symbolic number comparison skills relates to 1st grade mathematics
achievement: Evidence from a two-minute paper-and-pencil test. Learning and Instruction, 59, 21–33. https://doi.org/10.1016/j.learninstruc.2018.09.004.
Holm, S. (1979). A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics, 6, 65–70.
Kaufmann, L., & Nuerk, H. (2008). Basic number processing deficits in ADHD: A broad examination of elementary and complex number processing skills in 9- to 12-
year-old children with ADHD-C. Developmental Science, 11(5), 692–699. https://doi.org/10.1111/j.1467-7687.2008.00718.x.
Kortekaas-Rijlaarsdam, A. E., Luman, M., Sonuga-Barke, E., & Oosterlaan, J. (2019). Does methylphenidate improve academic performance? A systematic review and
meta-analysis. European Child & Adolescent Psychiatry, 28, 155–164.
Krajewski, K., Liehm, S., & Schneider, W. (2004). DEMAT 2+: Deutscher Mathematiktest für zweite Klassen [German mathematics test for second grades] Göttingen.
Germany: Hogrefe.
Kuhn, J.-T., & Holling, H. (2014). Number sense or working memory? The effect of two computer-based trainings on mathematical skills in elementary school.
Advances in Cognitive Psychology, 10, 59–67.
Kuhn, J.-K., Ise, E., Raddatz, J., Schwenk, C., & Dobel, C. (2016). Basic numerical processing, calculation, and working memory in children with dyscalculia and/or
ADHD symptoms. Zeitschrift für Kinder- und Jugendpsychiatrie und Psychotherapie, 44(5), 365–375.
Kuhn, J.-T., Schwenk, C., Raddatz, J., Dobel, C., & Holling, H. (2017). CODY-M 2-4: CODY-Mathetest für die 2.-4. Klasse [CODY math assessment for second to fourth
grades]. Düsseldorf, Germany: Kaasa health.
Kuhn, J.-T., Schwenk, C., Souvignier, E., & Holling, H. (2019). Arithmetische Kompetenz und Rechenschwäche am Ende der Grundschulzeit: Die Rolle
statusdiagnostischer und lernverlaufsbezogener Prädiktoren [Arithmetic skills and mathematical learning difficulties at the end of elementary school: The role of
summative and formative predictors]. Empirische Sonderpaedagogik, 11, 95–117.
Landerl, K., & Moll, K. (2010). Comorbidity of learning disorders: Prevalence and familial transmission. Journal of Child Psychology and Psychiatry, 51(3), 287–294.
https://doi.org/10.1111/j.1469-7610.2009.02164.x.
Landerl, K., Bevan, A., & Butterworth, B. (2004). Developmental dyscalculia and basic numerical capacities: A study of 8-9-year old students. Cognition, 93(2), 99–125.
https://doi.org/10.1016/j.cognition.2003.11.004.
Landerl, K., Fussenegger, B., Moll, K., & Willburger, E. (2009). Dyslexia and dyscalculia: Two learning disorders with different cognitive profiles. Journal of
Experimental Child Psychology, 103(3), 309–324. https://doi.org/10.1016/j.jecp.2009.03.006.
Landerl, K., Göbel, S., & Moll, K. (2013). Core deficit and individual manifestations of developmental dyscalculia (DD): The role of comorbidity. Trends in Neuroscience
and Education, 2(2), 38–42. https://doi.org/10.1016/j.tine.2013.06.002.
Lee, M. D., & Wagenmakers, E.-J. (2013). Bayesian cognitive modeling: A practical course. Cambridge: Cambridge University Press.
Loe, I. M., & Feldman, H. M. (2007). Academic and educational outcomes of children with ADHD. Journal of Pediatric Psychology, 32(6), 643–654. https://doi.org/
10.1093/jpepsy/jsl054.
Mayes, S. D., & Calhoun, S. L. (2006). Frequency of reading, math, and writing disabilities in children with clinical disorders. Learning and Individual Differences, 16,
145–157. https://doi.org/10.1016/j.lindif.2005.07.004.
Mazzocco, M. M., Feigenson, L., & Halberda, J. (2011). Impaired acuity of the approximate number system underlies mathematical learning disability (dyscalculia).
Child Development, 82(4), 1224–1237. https://doi.org/10.1111/j.1467-8624.2011.01608.x.
McGrath, L. M., Pennington, B. F., Shanahan, M. A., Santerre-Lemmon, L. E., Barnard, H. D., Willcutt, E. G., … Olson, R. K. (2011). A multiple deficit model of reading
disability and attention-deficit/hyperactivity disorder: Searching for shared cognitive deficits. Journal of Child Psychology and Psychiatry, 52(5), 547–557. https://
doi.org/10.1111/j.1469-7610.2010.02346.x.
Morey, R. D., Rouder, J. N., Jamil, T., Urbanek, S., Forner, K., & Ly, A. (2018). Package ‘BayesFactor’. Retrieved from https://cran.r-project.org/web/packages/
BayesFactor/BayesFactor.pdf.
Moura, R., Wood, G., Pinheiro-Chagas, P., Lonnemann, J., Krinzinger, H., Willmes, K., … Haase, V. G. (2013). Transcoding abilities in typical and atypical
mathematics achievers: The role of working memory and procedural and lexical competencies. Journal of Experimental Child Psychology, 116(3), 707–727. https://
doi.org/10.1016/j.jecp.2013.07.008.
Nock, M. K., Kazdin, A. E., Hiripi, E., & Kessler, R. C. (2007). Lifetime prevalence, correlates, and persistence of oppositional defiant disorder: Results from the
national comorbidity survey replication. Journal of Child Psychology and Psychiatry, 48(7), 703–713. https://doi.org/10.1111/j.1469-7610.2007.01733.x.
Peng, P., Namkung, J., Barnes, M., & Sun, C. (2016). A meta-analysis of mathematics and working memory: Moderating effects of working memory domain, type of
mathematics skill, and sample characteristics. Journal of Educational Psychology, 108(4), 455–473. https://doi.org/10.1037/edu0000079.
Pennington, B. F. (2006). From single to multiple deficit models of developmental disorders. Cognition, 101(2), 385–413. https://doi.org/10.1016/j.
cognition.2006.04.008.
Pennington, B. F., Groisser, D., & Welsh, M. C. (1993). Contrasting cognitive deficits in attention deficit hyperactivity disorder versus reading disability. Developmental
Psychology, 29(3), 511–523. https://doi.org/10.1037/0012-1649.29.3.511.
Peterson, R. L., Boada, R., McGrath, L. M., Willcutt, E. G., Olson, R. K., & Pennington, B. F. (2017). Cognitive prediction of reading, math, and attention: Shared and
unique influences. Journal of Learning Disabilities, 50(4), 408–421. https://doi.org/10.1177/0022219415618500.
R Core Team. (2019). R: A language and environment for statistical computing. Retrieved from. Vienna, Austria: R Foundation for Statistical Computing https://www.R-
project.org/.
Reigosa-Crespo, V., Valdés-Sosa, M., Butterworth, B., Estévez, N., Rodríguez, M., Santos, E., … Lage, A. (2012). Basic numerical capacities and prevalence of
developmental dyscalculia: The Havana survey. Developmental Psychology, 48(1), 123–135. https://doi.org/10.1037/a0025356.
Roick, T., Gölitz, D., & Hasselhorn, M. (2004). Deutscher Mathematiktest für dritte Klassen (DEMAT 3+) [German mathematics test for third grades]. Göttingen, Germany:
Hogrefe.
Rouder, J. N., Morey, R. D., Speckman, P. L., & Province, J. M. (2012). Default Bayes factors for ANOVA designs. Journal of Mathematical Psychology, 56(5), 356–374.
https://doi.org/10.1016/j.jmp.2012.08.001.
Schmider, E., Ziegler, M., Danay, E., Beyer, L., & Bühner, M. (2010). Is it really robust? Reinvestigating the robustness of ANOVA against violations of the normal
distribution assumption. Methodology, 6, 147–151. https://doi.org/10.1027/1614-2241/a000016.

12
E. von Wirth et al. Research in Developmental Disabilities 112 (2021) 103881

Schuchardt, K., Fischbach, A., Balke-Melcher, C., & Mähler, C. (2015). Die Komorbidität von Lernschwierigkeiten mit ADHS-Symptomen im Grundschulalter.
Zeitschrift für Kinder-und Jugendpsychiatrie und Psychotherapie, 43, 185–193. https://doi.org/10.1024/1422-4917/a000352.
Schulte-Körne, G., & Haberstroh, S. (2018). Diagnostik und Behandlung der Rechenstörung. AWMF-Register-Nr. 028/046. Retrieved from: S3 Richtlinie www.awmf.org/
en/clinical-practice-guidelines/detail/ll/028-046.html.
Sella, F., Re, A. M., Lucangeli, D., Cornoldi, C., & Lemaire, P. (2019). Strategy selection in ADHD characteristics children: A study in arithmetic. Journal of Attention
Disorders, 23(1), 87–98. https://doi.org/10.1177/1087054712438766.
Sonuga-Barke, E. J. S. (2005). Causal models of attention-deficit/hyperactivity disorder: From common simple deficits to multiple developmental pathways. Biological
Psychiatry, 567, 1231–1238, 10.1016/j.biopsych.2004.09.008.
Tosto, M. G., Momi, S. K., Asherson, P., & Malki, K. (2015). A systematic review of attention deficit hyperactivity disorder (ADHD) and mathematical ability: Current
findings and future implications. BMC Medicine, 13, 204. https://doi.org/10.1186/s12916-015-0414-4.
Von Aster, M. G., & Shalev, R. S. (2007). Number development and developmental dyscalculia. Developmental Medicine & Child Neurology, 49(11), 868–873. https://
doi.org/10.1111/j.1469-8749.2007.00868.x.
Wagenmakers, E.-J., Marsman, M., Jamil, T., Ly, A., Verhagen, J., Love, J., … Morey, R. (2018). Bayesian inference for psychology. Part I: Theoretical advantages and
practical ramifications. Psychonomic Bulletin & Review, 25, 35–57. https://doi.org/10.3758/s13423-017-1343-3.
Weiß, R. H. (2006). CFT 20-R – Grundintelligenztest Skala 2 mit Wortschatztest (WS) und Zahlenfolgentest (ZF)-Revision [Culture Fair Intelligence Test, Scale 20-R].
Göttingen, Germany: Hogrefe.
Weiß, R. H., & Osterland, J. (2013). CFT 1-R – Grundintelligenztest Skala 1 – Revision [Culture Fair Intelligence Test, Scale 1]. Göttingen, Germany: Hogrefe.
Willcutt, E. G., Betjemann, R. S., McGrath, L. M., Chhabildas, N. A., Olson, R. K., DeFries, J. C., … Pennington, B. F. (2010). Etiology and neuropsychology of
comorbidity between RD and ADHD: The case for multiple-deficit models. Cortex, 46(10), 1345–1361. https://doi.org/10.1016/j.cortex.2010.06.009.
Willcutt, E. G., Petrill, S. A., Wu, S., Boada, R., DeFries, J. C., Olson, R. K., … Pennington, B. F. (2013). Comorbidity between reading disability and math disability:
Concurrent psychopathology, functional impairment, and neuropsychological functioning. Journal of Learning Disabilities, 46, 500–516. https://doi.org/10.1177/
0022219413477476.
Wimmer, H., & Mayringer, H. (2014). SLS 2–9. Salzburger Lese-Screening für die Schulstufen 2–9 [Salzburg reading screening test for grades 2-9]. Göttingen, Germany:
Hogrefe.
World Health Organization. (1994). The ICD-10 classification of mental and behavioural disorders: Diagnostic criteria for research. Geneva: WHO.
Zentall, S. S., Smith, Y. N., Lee, Y. B., & Wieczorek, C. (1994). Mathematical outcomes of attention-deficit hyperactivity disorder. Journal of Learning Disabilities, 27(8),
510–519. https://doi.org/10.1177/002221949402700806.

13