You are on page 1of 17

Journal of Experimental Child Psychology 117 (2014) 12–28

Contents lists available at ScienceDirect

Journal of Experimental Child


Psychology
journal homepage: www.elsevier.com/locate/jecp

What basic number processing measures in


kindergarten explain unique variability in
first-grade arithmetic proficiency?
Dimona Bartelet a,⇑, Anniek Vaessen b, Leo Blomert b,1, Daniel Ansari c
a
Top Institute for Evidence Based Education Research, Maastricht University, 6200 MD Maastricht, The Netherlands
b
Faculty of Psychology and Neuroscience, Maastricht University, 6200 MD Maastricht, The Netherlands
c
Department of Psychology, University of Western Ontario, London, Ontario N6A 5C2, Canada

a r t i c l e i n f o a b s t r a c t

Article history: Relations between children’s mathematics achievement and their


Received 13 February 2013 basic number processing skills have been reported in both cross-
Revised 19 August 2013 sectional and longitudinal studies. Yet, some key questions are cur-
Available online 12 October 2013
rently unresolved, including which kindergarten skills uniquely
predict children’s arithmetic fluency during the first year of formal
Keywords:
schooling and the degree to which predictors are contingent on
Non-symbolic number processing skills
Symbolic number processing skills
children’s level of arithmetic proficiency. The current study
Arithmetic proficiency assessed kindergarteners’ non-symbolic and symbolic number pro-
Elementary school cessing efficiency. In addition, the contribution of children’s under-
Task-specific effects lying magnitude representations to differences in arithmetic
Unique predictors achievement was assessed. Subsequently, in January of Grade 1,
their arithmetic proficiency was assessed. Hierarchical regression
analysis revealed that children’s efficiency to compare digits,
count, and estimate numerosities uniquely predicted arithmetic
differences above and beyond the non-numerical factors included.
Moreover, quantile regression analysis indicated that symbolic
number processing efficiency was consistently a significant predic-
tor of arithmetic achievement scores regardless of children’s level
of arithmetic proficiency, whereas their non-symbolic number pro-
cessing efficiency was not. Finally, none of the task-specific effects
indexing children’s representational precision was significantly
associated with arithmetic fluency. The implications of the results
are 2-fold. First, the findings indicate that children’s efficiency to
process symbols is important for the development of their arith-
metic fluency in Grade 1 above and beyond the influence of

⇑ Corresponding author.
E-mail address: dimona.bartelet@maastrichtuniversity.nl (D. Bartelet).
1
In remembrance of coauthor Leo Blomert, who passed away after illness just before the final revision of this article.

0022-0965/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jecp.2013.08.010
D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28 13

non-numerical factors. Second, the impact of children’s non-sym-


bolic number processing skills does not depend on their arithmetic
achievement level given that they are selected from a nonclinical
population.
Ó 2013 Elsevier Inc. All rights reserved.

Introduction

A key aim of primary education is teaching children arithmetic skills. Yet, some children in primary
school experience severe difficulties in automatizing arithmetic facts (Swanson & Jerman, 2006).
According to developmental theory, children’s mathematics capacity emerges from a cumulative pro-
cess (Entwisle, Alexander, & Olson, 2005). Therefore, to understand the origins of arithmetic perfor-
mance differences, it is necessary to acquire insights into which basic number processing skills in
kindergarten uniquely predict individual differences in arithmetic skills at the onset of formal
education.
Empirical studies have increasingly addressed the issue to which extent basic number processing
skills in kindergarten explain early mathematics achievement, but most were restricted to one time
point in development (e.g., Holloway & Ansari, 2009; Inglis, Attridge, Batchelor, & Gilmore, 2011;
Muldoon, Towse, Simms, Perra, & Menzies, 2013). Although Duncan and colleagues (2007) reported
that knowledge of mathematics concepts in kindergarten was a more important predictor of later
achievement in elementary and middle school than reading and attention skills, only a few studies
have examined longitudinal data (Aunio & Niemivirta, 2010; Desoete, Ceulemans, de Weerdt, & Piet-
ers, 2010; Kolkman, Kroesbergen, & Leseman, 2013; Libertus, Feigenson, & Halberda, 2013; Mazzocco,
Feigenson, & Halberda, 2011b; Stock, Desoete, & Roeyers, 2009b, 2010). Furthermore, many studies
have not assessed non-symbolic and symbolic number processing skills simultaneously to isolate un-
ique predictors of arithmetic performance. However, this is essential to elucidate whether children’s
(perhaps innate) ability to process non-symbolic magnitudes or their capacity to process culturally in-
vented numerical symbols is a more important predictor of early arithmetic achievement. Therefore,
in the current study, we examined one non-symbolic and three different symbolic number processing
skills in kindergarten to determine which skills predict unique variance in first-grade arithmetic
performance.
Another theoretical issue is whether individual differences in kindergarteners’ magnitude repre-
sentations, children’s efficiency to process number information, or both account for significant indi-
vidual variation in arithmetic performance in Grade 1. To answer this question, studies need to
incorporate different outcome measures. Task-specific effect measures, such as distance and ratio ef-
fects, are postulated to tap the representations underlying children’s performance on the number pro-
cessing tasks (Maloney, Risko, Preston, Ansari, & Fugelsang, 2010). Other cognitive processes driving
children’s attainment are captured by efficiency measures such as mean accuracy and mean reaction
time (RT). The former quantifies children’s ability to process magnitude information, and the latter
quantifies their speed of processing magnitude information. All of the measures have been related
to mathematics achievement, but no measure was consistently found to be a significant predictor
of mathematics performance (see review by De Smedt, Noël, Gilmore, & Ansari, 2013). Hence, it re-
mains unclear which measure is most sensitive to the prediction of early arithmetic achievement dif-
ferences. In the current study, a third efficiency measure was introduced, namely a combined
accuracy–speed outcome (acc–RT). It has long been established that individuals’ response behavior
might be characterized by a decrease in response speed to ensure a correct answer or vice versa (Scho-
uten & Bekker, 1967), but very few studies have included an outcome measure that takes a potential
speed–accuracy trade-off into account (Sasanguie, De Smedt, Defever, & Reynvoet, 2012; Sasanguie,
Van den Bussche, & Reynvoet, 2012). Thus, all of these measures generated by basic number process-
ing tasks capture different performance features and, therefore, cannot be used interchangeably (Price,
Palmer, Battista, & Ansari, 2012). Moreover, the inclusion of these measures allowed us to analyze how
much of the variance in arithmetic achievement could be attributed to individual differences in
14 D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28

underlying magnitude representations and how much could be attributed to processing efficiency of
non-symbolic and symbolic magnitudes.
Finally, researchers have hypothesized that the relationship between basic number processing
skills and arithmetic performance is not only age dependent but also nonlinear (Bonny & Lourenco,
2013; Inglis et al., 2011). Yet, whether the explanatory power of children’s number processing effi-
ciency skills is contingent on their arithmetic proficiency has rarely been studied in nonclinical sam-
ples. Recently, Bonny and Lourenco (2013) tested preschoolers’ non-symbolic comparison skills and
mathematics abilities concurrently. Splitting their nonclinical sample into two groups based on chil-
dren’s mathematics performance, they found non-symbolic comparison skills to significantly explain
the low mathematics achievers’ test scores but not the higher mathematics achievers’ test scores. This
finding demonstrates the value of investigating whether the importance of basic number processing
predictors varies for children with different levels of arithmetic proficiency.

Basic number processing predictors and mathematics ability differences

In the current study, four basic number processing measures were included, namely a non-sym-
bolic and symbolic magnitude comparison paradigm, a counting paradigm, and an estimation para-
digm. They were selected because they have been postulated to underlie early mathematics
performance and have been used regularly in previous studies to measure basic number processing
(e.g., Mazzocco, Feigenson, & Halberda, 2011a; Muldoon et al., 2013; Sasanguie, Göbel, Moll, Smets,
& Reynvoet, 2013). Moreover, each task allowed for the indexing of task-specific effects. In the follow-
ing, each of the tasks and the corresponding task-specific effects are discussed.

Magnitude comparison skills


In a non-symbolic magnitude comparison task, children are required to process and compare
imprecise non-symbolic quantities (e.g., dots) (Mundy & Gilmore, 2009). This skill is thought to be
supported by an inborn approximate number system (ANS) that provides a noisy, ratio-dependent
representation of magnitudes (Dehaene, 2011). Specifically, the signal distributions of numerosities
increasingly overlap with decreasing distance between numerosities (Feigenson, Dehaene, & Spelke,
2004). The preciseness of this representation is believed to be revealed by the task-specific numerical
distance effect (NDE) and numerical ratio effect (NRE) (Moyer & Landauer, 1967; Price et al., 2012).
The NDE refers to the observation that humans’ response accuracy and speed are superior for num-
erosities that are further apart (e.g., four and nine) as opposed to the ones that are close together (e.g.,
five and seven). The NRE, also referred to as the numerical size or magnitude effect, pertains to the
observation that if distance is kept constant, it takes longer to compare high ratio magnitude pairs
(e.g., nine and eight) than to compare lower ratio pairs (e.g., four and five) despite equal numerical
distance. Because the NDE and NRE were found to be highly correlated, they have frequently been
used interchangeably (Price et al., 2012). The NDE is already observed in infants (Izard, Sann, Spelke,
& Streri, 2009) and animals (Cantlon, Platt, & Brannon, 2009). For example, newborn infants can dif-
ferentiate two large numerosities if their ratio is approximately 1:3 (Izard et al., 2009). With increas-
ing age, humans can successfully compare smaller ratios, suggesting that their representation of
numerosities becomes more precise (Halberda & Feigenson, 2008).
To compare exact symbolic quantities, researchers hypothesize that humans need to connect the
culturally transmitted discrete symbols to their corresponding ANS representation (Dehaene, 2011).
This proposition is based on the observation that the symbolic magnitude comparison task, a variant
of a magnitude comparison task in which two symbolic stimuli (e.g., 4 and 6) are displayed, generates
the NDE and NRE as well (Moyer & Bayer, 1976; Mundy & Gilmore, 2009). A number of previous stud-
ies have reported a smaller symbolic NDE, but not a smaller non-symbolic NDE, to be significantly re-
lated to mathematics performance (Holloway & Ansari, 2009; Sasanguie, De Smedt, et al., 2012). This
led researchers to hypothesize that differences in mathematics proficiency might be determined by
people’s preciseness of mapping abstract symbols to the ANS representations rather than by ANS pre-
cision (Mundy & Gilmore, 2009; Sasanguie Göbel et al., 2013). However, in addressing this issue in old-
er children, specifically 8- to 10-year-olds, Lonnemann, Linkersdörfer, Hasselhorn, and Lindberg
D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28 15

(2011) found both their non-symbolic and symbolic NDE to be concurrently associated with mathe-
matics achievement, which is not in line with this proposition.
Recently, two studies investigated whether children’s non-symbolic and symbolic magnitude rep-
resentation explains unique variance in mathematics achievement assessed at a later point in time. In
addition, they addressed the predictive power of preschoolers’ efficiency to compare non-symbolic
and symbolic magnitudes as well. Sasanguie, Van den Bussche, and colleagues (2012) administered
non-symbolic and symbolic magnitude comparison tasks, number line estimation tasks, and number
priming tasks to 72 children attending either Grade 1, 2, or 3. For both magnitude comparison tasks,
they computed an efficiency measure, specifically adjusted RT scores that take into account a potential
speed–accuracy trade-off, and the NDE slope. All number processing measures, in addition to chil-
dren’s grade level and spelling achievement, were included in the regression model, but only chil-
dren’s symbolic comparison speed, non-symbolic estimation accuracy, spelling ability, and grade
level explained unique variance in mathematics achievement. Neither the non-symbolic NDE measure
nor the symbolic NDE measure was found to be a significant predictor of mathematics achievement
1 year later. Recently, Sasanguie Göbel (2013) related a similar regression model, not including the
number priming tasks, to children’s timed arithmetic achievement. Again, children’s preciseness of
their non-symbolic and symbolic magnitude representation, indexed by the Weber fraction and
NDE, respectively, was not a unique predictor. Only children’s efficiency to compare Arabic digits, in-
dexed by mean RT, and their spelling achievement uniquely predicted arithmetic performance 1 year
later. Thus, both studies suggest that children’s efficiency to compare symbolic magnitudes, and not
non-symbolic magnitudes, is the best predictor of elementary mathematics achievement. Yet, other
studies addressing the role of children’s number processing efficiency measures have produced con-
tradictory results.
Desoete and colleagues (2010) administered three magnitude comparison tasks with different
notations (dot, Arabic, and verbal number word) to kindergarteners, but only accuracy on the non-
symbolic comparison measure uniquely explained arithmetic fact retrieval 1 year later. In addition,
Libertus and colleagues (2013) found preschool children’s non-symbolic processing efficiency to be
predictive of early mathematics achievement after partialling out differences in expressive vocabulary
and initial mathematics achievement. Finally, Mazzocco and colleagues (2011b) observed preschool
children’s non-symbolic comparison efficiency to explain first-grade differences in mathematics
achievement, but not in subtests assessing children’s expressive vocabulary, perceptual organization,
or numerical lexical retrieval. This suggests that non-symbolic comparison efficiency is a predictor
specific to mathematics achievement. Note, however, that Libertus and colleagues (2013) and Maz-
zocco and colleagues (2011b) did not examine children’s symbolic representations simultaneously,
thereby not allowing for the estimation of the unique variance in mathematics achievement explained
by non-symbolic magnitude comparison abilities.

Estimation ability
Numerical estimation tasks assess children’s ability to approximate a quantity (Mazzocco et al.,
2011a). Similar to the symbolic magnitude comparison task, children need to have an approximate
understanding of the non-symbolic magnitude expressed by the symbolic magnitude in order to suc-
cessfully complete estimation tasks. It is suggested that being more skillful in approximating magni-
tudes limits the range of possible answers from which children need to choose, which in turn might
decrease the likelihood of making mistakes and increase arithmetic proficiency (Booth & Siegler,
2008).
Estimation tasks asking participants to repeatedly estimate non-symbolic magnitudes enable
researchers to record the precise difference between an estimate and target numerosity multiple
times (Ansari, Donlan, & Karmiloff-Smith, 2007). This allows them to directly quantify an individual’s
mapping precision between each discrete number word and its ANS representation rather than infer-
ring it indirectly through metrics such as NDE (Ansari et al., 2007). It is hypothesized that mapping
precision decreases in proportion to the size of the target numerosity (Whalen, Gallistel, & Gelman,
1999). To investigate this, previous studies used the coefficient of variation (COV), a measure of dis-
persion that takes the magnitude of numerosities into account. The COV is defined as the ratio of var-
iability in estimates, namely standard deviation to mean (SD/M) estimate scores (Ansari et al., 2007).
16 D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28

Therefore, the COV should remain constant across numerosities if an individual’s mapping variability
increases proportional to the target quantity.
Mazzocco and colleagues (2011a) examined the retrospective relation between the mapping pre-
cision, quantified by the COV, of children in Grade 9 and the arithmetic achievement scores in Grades
3 to 6. They administered an estimation paradigm requiring children to repeatedly estimate arrays of
dots. Results revealed that higher COV scores, which are believed to reflect poorer mapping precision
between the ANS and number words, correlates with weaker arithmetic proficiency. To the best of our
knowledge, no previous study predictively related the estimation measure used in the present study,
to in early primary education.

Counting skills
Dot enumeration tasks assess children’s mastery of the cardinality principle, which refers to their
understanding that the last count word used in a sequence corresponds to the exact number of objects
constituting the set. Through counting, children learn that each numerosity corresponds to one exact
symbolic quantity (Cordes & Gelman, 2005). With respect to arithmetic, counting skills allow children
to assign an exact meaning to disjoint numerosity sets and combine them step by step to derive a dis-
crete answer (Butterworth, 2005; Le Corre & Carey, 2007). In line with this, children widely rely on
counting skills, such as finger counting, when they first learn to solve arithmetic problems
(Butterworth, 2005).
Previous studies argue for a relationship between children’s enumeration efficiency and their
mathematics achievement. Muldoon and colleagues (2013) repeatedly analyzed preschool children’s
enumeration ability and mathematics achievement. They found that at the beginning and end of
the school year, children who successfully enumerated 20 dots scored significantly better on the con-
currently administered mathematics achievement test than children who failed to correctly count 20
dots. Reigosa-Crespo and colleagues (2012) extended these findings by observing that children’s dot
enumeration efficiency is correlated with arithmetic fluency in second- to ninth-grade children after
controlling for Arabic comparison efficiency. In a study by Aunio and Niemivirta (2010), a composite
measure of counting proficiency, which included enumeration items, was administered to an exten-
sive kindergarten sample and found to be predictive of arithmetic ability in Grade 1. Note that this
study did not address the unique explanatory power of children’s enumeration efficiency. Stock and
colleagues (2009b) found kindergarteners’ counting accuracy to be significantly correlated with arith-
metic fluency differences in Grade 1. However, when they took children’s intelligence, non-symbolic
comparison, and logical ability skills into account, counting was not a significant predictor. This result
was confirmed by Stock and colleagues (2010), who administered the same test battery, except for the
intelligence measure, to a different sample.
In addition to efficiency measures, some studies have computed linear regression slopes that sum-
marize the positive relationship between cardinality (numerical size) and counting time. The steep-
ness of the slopes generally differs for small and large numerosity ranges. It is postulated that this
reflects a distinction between two underlying cognitive mechanisms, namely subitizing and counting
(Schleifer & Landerl, 2011). Subitizing refers to the fast and accurate recognition of small numerosities
(up to three or four objects) (Kaufmann, Lord, Reese, & Volkmann, 1949), whereas counting indexes
the strength of the mapping between larger number words and their semantic representation (Schle-
ifer & Landerl, 2011).
A recent study found that first-grade children’s subitizing slope contributed significantly to calcu-
lation proficiency measured concurrently (Penner-Wilger et al., 2007). This is in line with studies
reporting children with mathematics learning disabilities to have steeper subitizing slopes than con-
trols (Landerl, Bevan, & Butterworth, 2004; Schleifer & Landerl, 2011). To the best of our knowledge,
no predictive studies have been conducted in young children, nor have researchers addressed the role
of the counting slope in explaining arithmetic differences.

Non-numerical control variables

Previous studies have also associated mathematics achievement with factors other than basic num-
ber processing measures, including school, gender, intelligence, and general processing speed (Geary,
D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28 17

2011; Penner & Paret, 2008; Stock, Desoete, & Roeyers, 2009a). A general processing speed measure
specifies people’s ability and speed to serially encode and process low-level (perceptual) information
without needing to rely on more advanced cognitive capacities (e.g., accessing letter meaning) or
strategies (e.g., counting vs. retrieval) (Passolunghi & Lanfranchi, 2012).
In a longitudinal study, Geary (2011) assessed children’s general processing speed, intelligence, and
working memory, in addition to their magnitude comparison, conceptual counting, and number line
estimation skills, in Grades 1 to 5. They found that processing speed, intelligence, and the visuospatial
sketchpad (a working memory component) uniquely predicted achievement growth in mathematics
above and beyond the basic number processing measures. The results of a study by Stock and col-
leagues (2009a) point toward an association between school and arithmetic fluency. They assessed
the explanatory power of kindergarteners’ conceptual counting skills for arithmetic fluency in Grade
1, observing an intraclass correlation of .40. Several non-numerical factors, such as intelligence and
general processing speed, were included in the current study to investigate which basic number pro-
cessing skills affect arithmetic achievement above and beyond the non-numerical control variables.

The current study

The preceding review of the extant literature suggests that children’s ability to represent and pro-
cess non-symbolic and symbolic magnitudes is related to mathematics achievement differences. Using
four different basic number processing paradigms, we assessed kindergarten children’s processing
efficiency and representation of non-symbolic and symbolic magnitudes and their relationship to
first-grade arithmetic proficiency after controlling for several non-numerical factors.
The previous discussion points to three theoretical issues that were addressed in this study. First,
we investigated which number processing efficiency measures assessed in kindergarten uniquely pre-
dict first-grade arithmetic fluency after partialling out the influence of non-numerical factors. We
were primarily interested in whether non-symbolic number processing efficiency or symbolic number
processing efficiency best explains individual differences in calculation skills. Second, we examined
whether the relationship between children’s number processing efficiency and arithmetic achieve-
ment depends on the level of children’s calculation performance. Third, we sought to clarify whether
children’s underlying magnitude representation or their number processing efficiency skills predom-
inantly contribute to differences in early arithmetic achievement.

Method

Participants

The initial sample consisted of 248 kindergarten children (134 boys and 114 girls). This sample was
derived from an extensive test standardization study. For this purpose, kindergarten to sixth-grade
children from six primary schools spread across The Netherlands were recruited. The children com-
pleted the cognitive measures during the second half of the school year. Subsequently, in January of
Grade 1, the arithmetic fluency test was administered to 209 children (113 boys and 96 girls); the
other 39 children from the initial sample were not tested due to illness, grade retention, switching
of schools, or parents’ refusal to give consent.

Procedure

After the contacted schools agreed to participate, parents were informed by letter. They could deny
consent by returning an enclosed form. Each child was tested individually in a quiet room at school by
trained project workers during one session lasting approximately 30 min. Only the arithmetic achieve-
ment test was centrally administered in January of Grade 1. Both the nonverbal IQ test (Raven’s Col-
oured Progressive Matrices [CPM]; Raven, Court, & Raven, 1995) and the arithmetic achievement test
(Tempo Test Automatiseren [TTA]; de Vos, 2010) were paper-and-pencil tests. The remaining mea-
sures were computerized using a specialized response box with four buttons to precisely record
18 D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28

children’s accuracy (% correct) and RT (ms/item) scores (Blomert, Vaessen, & Ansari, 2013). The reli-
abilities of the computerized outcome measures ranged from .83 to .94 except for the accuracy score
of the baseline response task, which had a reliability of .66 (Blomert et al., 2013).

Tasks

Arithmetic achievement test


Arithmetic fact retrieval was operationalized using the standardized TTA (de Vos, 2010). Its objec-
tive is to quantify the degree of automatization of basic arithmetic facts attained by children from
Grade 1 to Grade 6. The entire test consists of four separate worksheets containing addition, subtrac-
tion, multiplication, and division problems of increasing difficulty. Because the multiplication and
division subtests are not standardized for Grade 1, only the addition and subtraction worksheets were
administered to the current sample. Then, a composite score was computed by summing the number
of correct responses. For each worksheet, children were instructed to calculate as many of the 50 oper-
ations as possible within 2 min.

Dot comparison
In the dot comparison task, two arrays of dots, containing numerosities from 1 to 9 (Level 1) or
from 10 to 39 (Level 2), were presented on a computer screen simultaneously and randomly. The side
of the larger numerosity was counterbalanced to ensure that for each ratio the larger array was
equally often presented on the right- and left-hand sides. Children were told that they would see
two planes with dots. Subsequently, they were instructed to push the button corresponding to the
numerosity that they thought was larger. It was explicitly stated that they were not supposed to
count. The presentation time of the stimuli was not restricted. Hence, a new trial was initiated only
after children pushed a button. The displays differed in numerosity based on one of four numerical
ratios (.25–.33, .50, .66, and .75). Each ratio was presented 16 times, resulting in 64 test stimuli and
3 practice trials. After half of the trials, a break was included to prevent fatigue, and participants deter-
mined when to continue. Lastly, to ensure that total area and total perimeter were not predictive of
numerosity, we included area-matched and perimeter-matched trials. Specifically, for half of the stim-
uli the total area of the dot arrays are the same, which causes the total perimeter to be greater on the
side with more dots. For the other half of the stimuli, the total perimeter of the dots arrays are the
same, which causes the total area to be greater on the side with less dots. These area-matched and
perimeter-matched trials were randomly presented in each presentation condition in an effort to pre-
vent participants from developing a strategy that relied on the relative size of the dot arrays.

Arabic comparison
The symbolic comparison task was composed of 4 practice and 32 test items that randomly dis-
played two one-digit Arabic numbers that differed based on one of four numerical ratios (.25–.33,
.50, .66, and .75). Each ratio was presented eight times. Also in this task, the side of the larger number
was counterbalanced for each ratio. It was explained to children that they would see two numbers and
needed to determine which number was larger by pushing the corresponding button. As in the dot
comparison task, a new item was presented only after children had responded.

Dot enumeration
The counting task consisted of 6 practice and 45 test items. On a black screen, children saw ran-
domly one to nine white dots of varying sizes in a nonlinear order. Moreover, as in the dot comparison
task, half of the trials were area matched and half of the trials were perimeter matched to control for
the influence of perimeter and cumulative surface area. One third of the stimuli displayed quantities of
three or less. The remaining items presented quantities ranging from four to nine. Each numerosity
was presented five times, although no display contained the same number of dots as its immediate
predecessor. Children were asked to count the dots and give their answers aloud. Children needed
to push the green button of the response box simultaneously to ensure RT registration. The presenta-
tion time of the stimuli was not fixed. Hence, a new item was displayed only after children had
responded.
D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28 19

Estimation
In the estimation task, collections of 1, 2, 3, 4, 7, 11, or 16 white dots were presented for 750 ms on
a black computer screen in random order. Children were told that they would see very briefly a num-
ber of dots and needed to determine immediately how many dots were displayed. If they were hesi-
tant to do so, the test administrator encouraged children to estimate or guess and reminded them that
they were not expected to count or know the exact number (Ansari et al., 2007). The spatial arrange-
ment of the dot clusters was random to avoid processing biases due to recognition of familiar patterns
(Mandler & Shebo, 1982). Furthermore, as in the dot comparison and dot enumeration tasks, half of
the trials had equivalent area and the other half had equivalent perimeter to ensure that total area
and total perimeter were not predictive of numerosity. Each of 6 practice and 67 test items was fol-
lowed by a mask after 750 ms. Three short breaks were included at fixed time points to avoid fatigue.

Baseline response task


The baseline response task (baseline RT) consisted of 20 trials in which children saw a row of four
empty squares. Children were told that an animation figure would appear in one of these squares.
When they saw the animation figure, they needed to push the button corresponding to its location
as fast as possible.

Nonverbal IQ
The CPM is a normed, untimed, visuospatial reasoning test for children in the age range from 5 to
11 years (Raven et al., 1995). Children saw a colored pattern and were asked to select the missing
piece out of six choices. Van Bon (1986) reported reliabilities of .80 or higher for the Dutch version.

Results

Children’s mean age in kindergarten and mean performance on the different measures are reported
in Table 1. Except for the estimation task, trials on the computerized measures with a response faster
than 200 ms were excluded. In addition to mean accuracy (% correct), mean RT (ms) and mean acc–RT
(number of items correct/s) scores were calculated. The latter was derived by dividing the total num-
ber of correct items by the total response time in seconds. In the estimation task, extreme outliers,
defined as responses of 100 or higher, were excluded before calculating accuracy and COV (SD/M)
scores. Finally, arithmetic achievement was operationalized as the composite score of the addition
and subtraction fact retrieval subtests of the TTA.

Table 1
Age (in kindergarten) and task performance of the follow-up kindergarten sample.

Accuracy Reaction time (ms) Acc–RT (number


(% correct) of items correct/s)
Mean SD Mean SD Mean SD Mean SD
Age (months) 71.8 4.4
Raven IQ (ss 0–10) 6.5 2.2
Arithmetic achievement (rs 0–55)a 17.8 7.3
Number-specific cognitive tasks
Dot comparison 88.7 10.2 1893 519 0.50 0.14
Arabic comparison 88.9 12.7 2161 774 0.46 0.17
Dot enumeration 89.8 10.2 4391 1167 0.22 0.06
Estimation 47.7 5.5
Baseline RT 95.7 7.3 1105 243 0.90 0.17

Note: ss, standard score; rs, raw score.


a
Administered in Grade 1.
20 D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28

Arithmetic achievement and number processing efficiency

At first, a Pearson correlation analysis was conducted to analyze the relationship between the effi-
ciency measures and arithmetic fact retrieval (see Table 2). To investigate whether the results are con-
sistent across different efficiency measures, children’s accuracy, RT, and acc–RT scores were included
if recorded. To correct for multiple pairwise comparisons, Holm’s correction method was used to com-
pute an adjusted significance level (Aickin & Gensler, 1996). Many of the measures were significantly
associated with each other (p 6 .001) and significantly correlated with arithmetic achievement
(p 6 .001). However, children’s dot comparison speed (r = –.14, p > .001), their Arabic comparison abil-
ity (r = .15, p > .001), and their counting ability (r = .21, p > .001) were not significantly associated with
children’s arithmetic performance.
In addition, two hierarchical multiple regression analyses were conducted to investigate the signif-
icant simple correlations between children’s number efficiency measures and arithmetic fluency in
more detail. In both regressions, the control variables time elapsed between kindergarten assessment
and follow-up testing (DL difference T1 and T2), gender, baseline RT, and school were entered in Step
1. Because school is a categorical variable, we operationalized it using five school dummies. In addi-
tion, in Step 2, the efficiency measures revealing significant simple correlations were included. The
process of entering all control variables in a first block, followed by the addition of all remaining
explanatory variables of interest in a second block, allows us to distinguish the variance in arithmetic
achievement attributable to the basic number processing measures from the variance captured by the
control variables (Pedhazur, 1997).
In the first regression analysis, efficiency on the timed number processing tasks was indexed by
acc–RT score. Because the estimation task was untimed, children’s efficiency on this measure was in-
dexed by accuracy scores. The results are reported in Table 3. A total of 45% of the variance was ex-
plained by the complete model, F(14, 194) = 11.49, p < .001. The number processing measures
accounted for an additional 25% of the variance, F(4, 194) = 21.91, p < .01. To further investigate the
importance of the second predictor block, the effect size was estimated using Cohen’s f2. Cohen ad-
vanced the following guidelines for effect size interpretation: small = .02, medium = .15, and
large = .35 (Cohen, Cohen, West, & Aiken, 2003). Based on these references, we found a large effect size
of .46. In addition, the regression results revealed that except for dot comparison (p > .05), the number
processing variables included in the model (p < .01) as well as the control variables gender (p < .05)
and School Dummy 4 (p < .01) were significant individual predictors.
In the second regression analysis, children’s efficiency on the timed number processing tasks was
indexed by accuracy and RT scores separately. Again, only the measures significantly correlated with
arithmetic fluency were included. The results are reported in Table 4. The complete model explained
42% of the variance in arithmetic fact retrieval, F(14, 194) = 10.23, p < .01, and the number processing

Table 2
Pearson correlations between arithmetic achievement test and basic number processing skills (after applying Holm’s correction
method).

Measure 1 2 3 4 5 6 7 8 9 10 11
1. Arithmetic achievement .24* .14 .22* .15 .38* .45* .21 .22* .37* .38*
2. Dot comparison (acc) .09 .32* .10 .07 .15 .25* .20 .23* .35*
3. Dot comparison (RT) .85* .06 .39* .38* .05 .22* .28* .02
4. Dot comparison (acc–RT) .08 .41* .45* .15 .27* .36* .15
5. Arabic comparison (acc) .20 .45* .13 .06 .04 .16
6. Arabic comparison (RT) .85* .08 .22 .28* .17
7. Arabic comparison (acc–RT) .16 .19 .31* .26*
8. Counting (acc) .21 .18 .25*
9. Counting (RT) .84* .23*
10. Counting (acc–RT) .34*
11. Estimation (acc)
*
p < .05 (after Holm’s correction).
D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28 21

Table 3
Results of hierarchical multiple regression analysis with dependent arithmetic achievement test and adding basic number
processing skills in Step 2.

B SE B b Significance (p)
Step 1
Constant 8.17 10.08 .42
Age in months 0.04 0.10 .02 .69
DL difference T1 and T2 0.22 0.44 .03 .62
Gender 1.71 0.82 .12 .04*
Nonverbal IQ 0.05 0.09 .04 .57
Baseline RT 0.00 0.00 .02 .69
School Dummy 2 1.84 2.76 .04 .51
School Dummy 3 0.84 1.34 .05 .53
School Dummy 4 6.36 1.28 .35 .00**
School Dummy 5 1.15 1.32 .06 .39
School Dummy 6 1.20 1.37 .06 .38
Step 2
Dot comparison (acc–RT) 1.80 3.38 .04 .59
Arabic comparison (acc–RT) 12.11 2.83 .28 .00**
Counting (acc–RT) 23.06 8.21 .17 .01*
Estimation (acc) 0.40 0.08 .30 .00**

Note: R2 = .20 for Step 1; DR2 = .25 for Step 2.


*
p < .05.
**
p < .01.

Table 4
Results of hierarchical multiple regression analysis with dependent arithmetic achievement test and adding basic number
processing skills (accuracy and reaction time separately) in Step 2.

B SE B b Significance (p)
Step 1
Constant 1.19 10.75 .91
Age in months 0.07 0.10 .04 .52
DL difference T1 and T2 0.28 0.45 .04 .54
Gender 2.33 0.83 .16 .01**
Nonverbal IQ 0.07 0.09 .05 .44
Baseline RT 0.00 0.00 .00 1.00
School Dummy 2 1.50 2.78 .03 .59
School Dummy 3 0.52 1.36 .03 .70
School Dummy 4 6.64 1.29 .37 .00**
School Dummy 5 1.09 1.34 .06 .42
School Dummy 6 1.33 1.40 .07 .34
Step 2
Dot comparison (acc) 0.08 0.04 .12 .05
Arabic comparison (RT) 0.00 0.00 .23 .00**
Counting (RT) 0.00 0.00 .05 .37
Estimation (acc) 0.44 0.08 .33 .00**

Note: R2 = .20 for Step 1; DR2 = .22 for Step 2.


**
p < .01.

measures accounted for an additional 22% of the variance, F(4, 194) = 18.40, p < .01. Moreover, Cohens
f2 was found to be large at .38. Children’s ability to compare non-symbolic magnitudes and their
counting speed did not contribute additional individual explanatory power to the model (p > .05),
whereas their estimation ability and Arabic comparison speed did (p < .01).
22 D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28

A contingent relationship?

To address whether the relationship between children’s number processing efficiency and arithme-
tic fact retrieval is contingent on their arithmetic proficiency, we first checked for heteroskedasticity.
The significant Breusch–Pagan test, P(v2(14) = 36.25) = .001, points toward a nonlinear relationship.
Subsequently, we ran a quantile regression analysis in Stata. This statistical method models the rela-
tionship between an outcome variable and its predictor variables at different points of the conditional
distribution of the outcome measure (Cameron & Trivedi, 2009). We estimated the rate of change in
the 16th, 50th, and 84th percentiles of arithmetic achievement brought about by a 1-unit change in
the predictor variables. The two extreme percentile ranks correspond to a performance score of 1
SD above or below the mean under a normal curve. They were chosen because we were primarily
interested in the explanatory regression models for children who exhibited arithmetic performance
substantially higher or lower than the mean.
Children’s number processing efficiency measures, indexed by acc–RT scores, and the non-numer-
ical control factors were inserted as predictors into the quantile regression analyses. The results of the
quantile regression analysis are presented in Table 5. The basic number processing tasks Arabic com-
parison (p < .05) and estimation (p < .01) were significant individual predictors across all quantile lev-
els, whereas the counting measure (p < .05) explained unique variance only at the highest quantile
level of arithmetic performance. Regarding the control variables, only School Dummy 4 (p < .05)
was a unique predictor, although this variable no longer significantly explained variance at the highest
quantile level. Additional analysis in Stata revealed that the variable School Dummy 4 had a signifi-
cantly higher impact at the lower quantile of arithmetic performance. However, the regression coef-
ficients of the basic number processing measures, including counting, did not differ significantly
across the quantile levels. A visual comparison of the quantile and linear regression coefficients
showed that the latter sometimes underestimates or overestimates the rate of change explained by
variables, depending on the quantile of arithmetic achievement analyzed. Finally, the unique

Table 5
Results of quantile regression analysis with dependent arithmetic achievement test for the 16th, 50th, and 84th percentiles.

Quantile Regression
16th percentile 50th percentile 84th percentile
B SE B Significance B SE B Significance B SE B Significance
(p) (p) (p)
Constant 13.54 12.40 .28 6.77 12.09 .58 21.34 16.01 .18
Age in months 0.02 0.11 .88 0.00 0.14 .98 0.03 0.16 .83
DL difference T1 and T2 0.70 0.69 .31 0.07 0.52 .89 0.20 0.56 .73
Gender 1.99 1.09 .07 1.50 1.09 .17 1.67 1.25 .19
Nonverbal IQ 0.07 0.12 .59 0.01 0.13 .95 0.05 0.15 .73
Baseline RT 0.00 0.00 .64 0.00 0.00 .43 0.00 0.00 .59
School Dummy 2 0.36 3.09 .91 1.51 3.52 .67 4.50 3.85 .24
School Dummy 3 1.23 1.22 .32 1.02 1.59 .52 2.05 2.28 .37
School Dummy 4 10.07 1.65 .00** 4.79 2.35 .04* 3.21 1.79 .08
School Dummy 5 0.07 1.79 .97 1.22 1.64 .46 2.24 2.75 .42
School Dummy 6 0.31 1.43 .83 0.26 1.87 .89 1.76 2.02 .38
Dot comparison 1.09 3.62 .76 0.62 4.86 .90 0.61 6.77 .93
(acc–RT)
Arabic comparison 7.78 3.90 .05* 11.78 4.36 .01** 16.89 5.97 .00**
(acc–RT)
Counting (acc–RT) 10.02 10.66 .35 25.86 14.23 .07 34.99 14.17 .01*
Estimation (acc) 0.42 0.12 .00** 0.32 0.11 .01** 0.51 0.17 .00**

Note: Pseudo R2 = .37 for 16th percentile model; pseudo R2 = .23 for 50th percentile model; pseudo R2 = .29 for 84th percentile
model.
*
p < .05.
**
p < .01.
D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28 23

explanatory power of gender observed in the linear regression model was absent in all three quantile
regression models.

Arithmetic achievement and magnitude representations

We examined whether the task-specific measures of numerical magnitude processing were signif-
icantly related to individual differences in children’s arithmetic achievement. First, we needed to ver-
ify that the postulated task-specific effects were observed in the current sample. Because mean
accuracy scores were high on the timed number processing measures, which considerably reduced
the discernible variability, RT scores were used to compute task-specific effects. Given that no RT
scores were available for the estimation task, the COV was calculated using accuracy scores.

Symbolic and non-symbolic NRE


First, a 2 (Task)  4 (Ratio) repeated-measures analysis of variance (ANOVA) was executed. Because
the assumption of sphericity was violated, the Greenhouse–Geisser adjustments are reported for ratio
and task  ratio. A significant ratio effect (NRE), F(2.64, 548.77) = 104.57, p < .001, and a significant task
effect, F(1, 208) = 23.91, p < .001, were found. To further examine the relationship between the non-
symbolic and symbolic comparison NRE, numerical ratio slopes were calculated. The size of the
numerical ratio slope (M = 1831, SD = 1566) was larger for the non-symbolic measure than for the
symbolic measure (M = 903, SD = 1585). A dependent t test showed that the non-symbolic slope
was significantly steeper than the symbolic one, t(208) = 6.08, p < .001.

Counting-specific effects
Behavioral performance on an enumeration task is frequently characterized by a subitizing and
counting effect. Because the quantity of objects falling within the subitizing range increases across
development, accuracy and RT scores were explored to decide on the appropriate cutoff point for
the current sample (Fischer, Gebhardt, & Hartnegg, 2008). A sharp and continuous increase in RT
and a drop in accuracy scores below 99% after numerosity 3 were observed. This indicated that the
enumeration of 1 to 3 objects constitutes the subitizing range, whereas the items with more dots re-
flect the counting range. A repeated-measures ANOVA with number as the within-participant variable
confirmed the number effect for accuracy, F(3.53, 734.64) = 97.13, p < .001, and RT scores,
F(4.25, 819.95) = 591.07, p < .001. Note that the Greenhouse–Geisser adjustment was used because
the data violated the sphericity assumption. For the purpose of analyzing the relationship between
arithmetic achievement and both the subitizing and counting effects, we computed a subitizing slope
and a counting slope.

Estimation effect
The variability in children’s verbal responses on an estimation task was captured by the COV, which
was computed as SD/M per numerosity. Because it has been suggested that the processing of small and
large numerosities might be supported by different cognitive processes (Schleifer & Landerl, 2011),
two separate repeated measures were conducted, including numerosities falling either within the sub-
itizing range (1, 2, and 3) or within the counting range (5, 7, 11, and 16). The COV was constant for the
subitizing range, F(1, 207) = 0.01, p = .916, but not for the counting range, F(1, 207) = 15.56, p < .001. A
Bonferroni post hoc comparison including the counting range numerosities revealed that only mean
COVs for 11 and 16 dots were constant. Unexpectedly, both COVs were smaller than the COV of num-
erosity 7. If this decrease in variability was caused by a deficient understanding of the meaning of two-
digit numbers inducing children to repeatedly respond with the same number word when viewing
displays containing two-digit numerosities, then the SD for numerosities 11 and 16 should be smaller
than that for numerosity 7. However, the SD for numerosities, 7, 11, and 16 were 2.59, 3.7, and 6.4,
respectively.

Task-specific effects and arithmetic achievement


To examine the relationship between task-specific effects and arithmetic achievement, a Pearson’s
correlations analysis was conducted (see Table 6). However, after Holm’s correction, none of the
24 D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28

Table 6
Pearson correlations between arithmetic achievement test and task-specific effects.

Measure 1 2 3 4 5 6 7
1. Arithmetic achievement .10 .13 .16 .03 .10 .01
2. Non-symbolic NRE .02 .03 .10 .05 .05
3. Symbolic NRE .01 .09 .02 .02
4. Subitizing effect (slope) .04 .04 .05
5. Counting effect (slope) .13 .02
6. Estimation COV subitizing range .64*
7. Estimation COV counting range
*
p < .05.

measures was significantly associated with arithmetic fluency (p > .05). Hence, the relationship be-
tween the task-specific effects and arithmetic performance was not investigated further.

Discussion

The first aim of this study was to clarify which preschool number processing efficiency measures
uniquely predict first-grade arithmetic achievement while controlling for non-numerical factors. Sec-
ond, we wanted to create a more refined picture of this relationship. Therefore, we investigated
whether the relationship between children’s number processing efficiency and arithmetic perfor-
mance varied depending on children’s arithmetic proficiency. Moreover, we examined whether the
preciseness of children’s underlying magnitude representations or their number processing efficiency
accounts for most variation in arithmetic proficiency.
Kindergarteners’ efficiency to process non-symbolic and symbolic magnitudes was significantly
correlated with calculation fluency at the onset of formal education. However, after controlling for
non-numerical factors, the tasks assessing children’s efficiency to process symbolic magnitudes and
the control variables gender and school were the only unique predictors. Quantile regression analysis
revealed that children’s Arabic comparison and estimation efficiency, but not their dot comparison
efficiency, were significant individual predictors across all three quantile regression models. Children’s
counting efficiency explained unique variance only in the 84th percentile model. Yet, the regression
coefficients of the counting measure did not differ significantly across the quantile regression models.
The non-numerical factor school no longer contributed individual explanatory power in the highest
percentile of arithmetic achievement. Additional analysis showed that the regression coefficient of
the variable school in the 84th quantile model was significantly different from the coefficient found
in the 16th and 50th quantile models. Finally, no significant simple correlations were observed be-
tween task-specific effect measures, which are thought to index individual differences in magnitude
representations, and arithmetic performance.

Contribution of specific cognitive skills

In accordance with previous studies, we found kindergarteners’ efficiency to compare symbolic


magnitudes, but not non-symbolic magnitudes, to be uniquely predictive of first-grade arithmetic flu-
ency above and beyond several control factors (De Smedt & Gilmore, 2011; Kolkman et al., 2013;
Sasanguie Göbel et al., 2013; Sasanguie, Van den Bussche, et al., 2012; Stock et al., 2010). Children’s
procedural counting efficiency and their ability to verbally estimate numerosities have been studied
less extensively as individual predictors of arithmetic fluency. The current finding that children’s
counting and estimation efficiency explained a unique part of the variance in arithmetic fluency as
well further supports the hypothesis that children’s efficiency to process symbolic magnitudes is an
important kindergarten predictor of early arithmetic performance. However, Stock and colleagues
(2010) did not find a significant relationship between children’s counting skills and arithmetic
achievement. Moreover, some researchers observed non-symbolic comparison skills to be related to
arithmetic skills (Desoete et al., 2010; Libertus et al., 2013; Mazzocco et al., 2011b). Some of the
D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28 25

reasons that could account for these divergent findings include the use of accuracy scores as an index
of number processing efficiency, different operationalization of the tasks and the arithmetic achieve-
ment test, and differences in the age and size of the samples (Price et al., 2012). Furthermore, those
studies that found an association between non-symbolic comparison and arithmetic did not use sym-
bolic comparison tasks concurrently and therefore could not, unlike the current study, estimate the
amount of unique variance explained by non-symbolic magnitude comparison performance.
These contradictory findings suggest that the role of children’s basic number processing skills in
the development of mathematics achievement is not straightforward. Researchers have emphasized
that advanced mathematics abilities build on lower ones (Bonny & Lourenco, 2013; Entwisle et al.,
2005). It might be that the initial learning of arithmetic requires the support of not only a broad net-
work of basic number processing skills but also more basic ones, whereas automatizing one’s arithme-
tic abilities relies more on number skills closer in content and developmental time. In line with such
an argument is the hypothesis that children’s efficiency to compare non-symbolic magnitudes might
be a significant individual predictor of arithmetic fluency only for lower achieving children. But the
quantile regression results in the current study do not disclose a differential effect of the non-symbolic
number processing measure. Children’s non-symbolic number processing efficiency did not explain
unique variance in arithmetic achievement, whereas their symbolic number processing efficiency
was a consistent predictor. This indicates that early arithmetic development is primarily explained
by basic number processing skills that require an understanding of symbolic notations. In addition,
the results show that in a nonclinical sample, the relationship between number processing efficiency
and arithmetic performance does not vary for different points in the conditional distribution of arith-
metic scores. Both findings are contrary to Bonny and Lourenco (2013), who observed a significant cor-
relation between children’s non-symbolic magnitude processing efficiency for children with lower
mathematics scores but not for children with higher scores. However, the researchers measured chil-
dren’s non-symbolic number processing efficiency and mathematics achievement concurrently and
did not include multiple number skills simultaneously in their explanatory model.
The current study also shed light on the individual contribution of non-numerical factors to calcu-
lation fluency. Children attending School 4 performed significantly worse than the reference group de-
spite the fact that they used the same mathematics curriculum. However, other characteristics could
possibly account for the observed difference in performance. For example, the reference school was
located in a more rural area, the number of enrolled students was substantially lower, and it had a
Catholic denomination as opposed to a public one. Together with the previous study by Stock and col-
leagues (2009a) in which a substantial intraclass correlation was reported, the current finding implies
that school characteristics should be included as control variables in future research. Furthermore, in
line with past studies, gender (Geary, Saults, Liu, & Hoard, 2000; Penner & Paret, 2008) was found to be
uniquely predictive of average arithmetic differences, with girls scoring lower than boys. Finally, nei-
ther nonverbal IQ nor general response speed uniquely predicted arithmetic achievement differences,
indicating that these variables are of little additional informative value for the prediction of arithmetic
fluency at the beginning of formal schooling. Although this is in line with a number of studies (Jordan,
Kaplan, Nabors Oláh, & Locuniak, 2006; Krajewski & Schneider, 2009; Passolunghi, Vercelloni, & Sch-
adee, 2007), it contradicts others (Geary, 2011; LeFevre et al., 2006; Penner & Paret, 2008). A possible
explanation for these conflicting findings could be the difference in operationalization of IQ and gen-
eral response speed. Note that Passolunghi and Lanfranchi (2012) found verbal IQ to be related to
mathematics achievement, whereas performance IQ was not. Furthermore, the variance in arithmetic
fluency explained by children’s general response speed might have been captured by the timed num-
ber processing measures. In support of this argument, we found the general response speed measure
to be strongly correlated with the timed number processing tasks.

Magnitude representations and arithmetic fluency

Some researchers have postulated that differences in mathematics achievement can be attributed
to the preciseness of children’s ANS and the analogue representation of symbolic magnitudes (Dehae-
ne, 2011; Mundy & Gilmore, 2009). However, the absence of significant associations between arithme-
tic fluency and multiple task-specific effects in the current study speak against this view.
26 D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28

Contrary to our results, Bugden and Ansari (2011), who studied first- and second-grade children,
found the symbolic NRE to uniquely explain arithmetic fluency measured concurrently. To the best
of our knowledge, previous studies with younger children have examined only the predictive relation
between NDE, a metric strongly correlated with the NRE, and calculation fluency. For example, De
Smedt, Verschaffel, and Ghesquière (2009) reported symbolic NDE in Grade 1 to uniquely predict
mathematics achievement in Grade 2. Using the COV as a metric indexing the mapping precision be-
tween the ANS and symbolic number words, Mazzocco and colleagues (2011a) found children’s map-
ping precision in Grade 9 to be retrospectively correlated with their arithmetic achievement scores in
Grades 3 to 6. Yet, our study differs from these previous studies in certain aspects, including the sam-
ple size and the age of the children.
In addition, the subitizing and counting effect was not significantly correlated with early arithmetic
fluency in the current study. This finding is not in line with a previous study by Penner-Wilger and
colleagues (2007) in which the subitizing slope was linearly related to calculation skills in nonclinical
first-grade children. However, their measures were administered concurrently and analyzed a differ-
ent age group. In addition, they computed the subitizing slope slightly different, first calculating med-
ian response times for the subitizing range values (1–3) for each child and subsequently drawing a
regression line through these medians. In our study, we computed a slope through the observed reac-
tion times for all subitizing range items (1–3) for each child.
The current findings lead to the conclusion that children’s efficiency to process symbolic magni-
tudes and not the precision or acuity of their analog representation of non-symbolic and symbolic
magnitudes is a central, predictive preschool competence for arithmetic fluency development. Never-
theless, the absence of significant findings for the task-specific effects might ensue from a low reliabil-
ity of the task-specific effect measures. Regarding the NDE, researchers recently found this measure to
have low, albeit significant, split-half reliabilities and, therefore, proposed that one should interpret
results based on an NDE measure carefully (Maloney et al., 2010).
Furthermore, consistent with the notion of a strong difference between symbolic and non-symbolic
magnitude processing, it was recently proposed that the symbolic magnitude representation system
might not be mapped onto the ANS but rather is a distinct exact system (Lyons, Ansari, & Beilock,
2012). The findings of a recent study by Sasanguie, Defever, Maertens, and Reynvoet (2013) support
this view. Their analysis revealed that kindergarteners’ non-symbolic comparison accuracy was not
significantly correlated with their symbolic comparison ability 6 months later. The current finding
that children’s symbolic NDE and estimation COV were both not significantly associated with arithme-
tic performance suggests that children’s analog representation of numerical symbols in kindergarten
is not important to early arithmetic development. However, we cannot exclude the possibility that
individual differences in arithmetic performance are driven by a deficient, qualitatively different exact
symbolic representational system.

Conclusions

The current study suggests that children’s efficiency to process discrete symbols is important to the
development of adequate calculation fluency above and beyond non-numerical factors. Children’s
non-symbolic magnitude comparison efficiency, on the other hand, did not uniquely explain arithme-
tic performance regardless of children’s level of arithmetic proficiency. This suggests that non-sym-
bolic magnitude comparison efficiency is not a critical predictor of early arithmetic proficiency.
Finally, children’s magnitude representations indexed by task-specific effects were not significantly
correlated with arithmetic achievement. To conclude, children’s efficiency to process symbolic magni-
tudes, rather than underlying magnitude representations, should be considered as potential candi-
dates to be addressed in successful teaching or remediation programs (Gersten, Jordan, & Flojo,
2005) aimed at young children.

References

Aickin, M., & Gensler, H. (1996). Adjusting for multiple testing when reporting research results: The Bonferroni vs. Holm
methods. American Journal of Public Health, 86, 726–728.
D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28 27

Ansari, D., Donlan, C., & Karmiloff-Smith, A. (2007). Typical and atypical development of visual estimation abilities. Cortex, 43,
758–768.
Aunio, P., & Niemivirta, M. (2010). Predicting children’s mathematical performance in Grade One by early numeracy. Learning
and Individual Differences, 20, 427–435.
Blomert, L., Vaessen, A., & Ansari, D. (2013). Differentiaal diagnostiek van dyscalculie: Cognitieve analyse van rekenen [Dyscalculia
differential diagnosis: Cognitive analysis of math]. Amsterdam: Boom Test Publishers.
Bonny, J. W., & Lourenco, S. F. (2013). The approximate number system and its relation to early math achievement: Evidence
from the preschool years. Journal of Experimental Child Psychology, 114, 375–388.
Booth, J. L., & Siegler, R. S. (2008). Numerical magnitude representations influence arithmetic learning. Child Development, 79,
1016–1031.
Bugden, S., & Ansari, D. (2011). Individual differences in children’s mathematical competence are related to the intentional but
not automatic processing of Arabic numerals. Cognition, 118, 32–44.
Butterworth, B. (2005). The development of arithmetical abilities. Journal of Child Psychology and Psychiatry, 46, 3–18.
Cameron, A. C., & Trivedi, P. K. (2009). Microeconometrics using Stata (rev. ed.). College Station, TX: Stata Press.
Cantlon, J. F., Platt, M., & Brannon, E. M. (2009). Beyond the number domain. Trends in Cognitive Sciences, 13, 83–91.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd
ed.). New York: Routledge.
Cordes, S., & Gelman, R. (2005). The young numerical mind: When does it count? In J. I. D. Campbell (Ed.), Handbook of
mathematical cognition (pp. 127–142). New York: Psychology Press.
De Smedt, B., & Gilmore, C. K. (2011). Defective number module or impaired access? Numerical magnitude processing in first
graders with mathematical difficulties. Journal of Experimental Child Psychology, 108, 278–292.
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 & Education, 2, 48-55. http://dx.doi.org/10.1016/j.tine.2013.06.001.
De Smedt, B., Verschaffel, L., & Ghesquière, P. (2009). The predictive value of numerical magnitude comparison for individual
differences in mathematics achievement. Journal of Experimental Psychology, 103, 469–479.
de Vos, T. (2010). Tempo test automatiseren. Amsterdam: Boom Test Publishers.
Dehaene, S. (2011). The number sense: How the mind creates mathematics (2nd ed.). New York: Oxford University Press.
Desoete, A., Ceulemans, A., de Weerdt, F., & Pieters, S. (2010). Can we predict mathematical learning disabilities from symbolic
and non-symbolic comparison tasks in kindergarten? Findings from a longitudinal study. British Journal of Educational
Psychology, 82, 64–81.
Duncan, G. J., Dowsett, C. J., Claessens, A., Magnuson, K., Huston, A. C., Klebanov, P., et al (2007). School readiness and later
achievement. Developmental Psychology, 43, 1428–1446.
Entwisle, D. R., Alexander, K. L., & Olson, L. S. (2005). First grade and educational attainment by age 22: A new story. American
Journal of Sociology, 110, 1458–1502.
Feigenson, L., Dehaene, S., & Spelke, E. (2004). Core systems of number. Trends in Cognitive Sciences, 8, 307–314.
Fischer, B., Gebhardt, C., & Hartnegg, K. (2008). Subitizing and visual counting in children with problems acquiring basic
arithmetic skills. Optometry and Vision Development, 39, 24–29. Retrieved from http://www.covd.org/Home/OVDJournal/
tabid/104/Default.aspx.
Geary, D. C. (2011). Cognitive predictors of achievement growth in mathematics: A 5-year longitudinal study. Developmental
Psychology, 47, 1539–1552.
Geary, D. C., Saults, S. J., Liu, F., & Hoard, M. K. (2000). Sex differences in spatial cognition, computational fluency, and
arithmetical reasoning. Journal of Experimental Child Psychology, 77, 337–353.
Gersten, R., Jordan, N. C., & Flojo, J. R. (2005). Early identification and interventions for students with mathematics difficulties.
Journal of Learning Disabilities, 38, 293–304.
Halberda, J., & Feigenson, L. (2008). Developmental change in the acuity of the ‘‘number sense’’: The approximate number
system in 3-, 4-, 5-, and 6-year-olds and adults. Developmental Psychology, 44, 1457–1465.
Holloway, I. D., & Ansari, D. (2009). Mapping numerical magnitudes onto symbols: The numerical distance effect and individual
differences in children’s mathematics achievement. Journal of Experimental Child Psychology, 103, 17–29.
Inglis, M., Attridge, N., Batchelor, S., & Gilmore, C. (2011). Non-verbal number acuity correlates with symbolic mathematics
achievement: But only in children. Psychonomic Bulletin & Review, 18, 1222–1229.
Izard, V., Sann, C., Spelke, E. S., & Streri, A. (2009). Newborn infants perceive abstract numbers. Proceedings of the National
Academy of Sciences of the United States of America, 106, 10382–10385.
Jordan, N. C., Kaplan, D., Nabors Oláh, L., & Locuniak, M. N. (2006). Number sense growth in kindergarten: A longitudinal
investigation of children at risk for mathematics difficulties. Child Development, 77, 153–175.
Kaufmann, E. L., Lord, M. W., Reese, T. W., & Volkmann, J. (1949). The discrimination of visual number. American Journal of
Psychology, 62, 498–525.
Kolkman, M. E., Kroesbergen, E. H., & Leseman, P. P. M. (2013). Early numerical development and the role of non-symbolic and
symbolic skills. Learning and Instruction, 25, 95–103.
Krajewski, K., & Schneider, W. (2009). Early development of quantity to number–word linkage as a precursor of mathematical
school achievement and mathematical difficulties: Findings from a four-year longitudinal study. Learning and Instruction, 19,
513–526.
Landerl, K., Bevan, A., & Butterworth, B. (2004). Developmental dyscalculia and basic numerical capacities: A study of 8–9-year-
old students. Cognition, 93, 99–125.
Le Corre, M., & Carey, S. (2007). One, two, three, four, nothing more: An investigation of the conceptual sources of the verbal
counting principles. Cognition, 105, 395–438.
LeFevre, J.-A., Smith-Chant, B. L., Fast, L., Skwarchuk, S.-L., Sargla, E., Arnup, J. S., et al (2006). What counts as knowing? The
development of conceptual and procedural knowledge of counting from kindergarten through Grade 2. Journal of
Experimental Child Psychology, 93, 285–303.
28 D. Bartelet et al. / Journal of Experimental Child Psychology 117 (2014) 12–28

Libertus, M. E., Feigenson, L., & Halberda, J. (2013). Is approximate number precision a stable predictor of math ability? Learning
and Individual Differences, 25, 126–133.
Lonnemann, J., Linkersdörfer, J., Hasselhorn, M., & Lindberg, S. (2011). Symbolic and non-symbolic distance effects in children
and their connection with arithmetic skills. Journal of Neurolinguistics, 24, 583–591.
Lyons, I. M., Ansari, D., & Beilock, S. L. (2012). Symbolic estrangement: Evidence against a strong association between numerical
symbols and the quantities they represent. Journal of Experimental Psychology: General, 141, 635–641.
Maloney, E. A., Risko, E. F., Preston, F., Ansari, D., & Fugelsang, J. (2010). Challenging the reliability and validity of cognitive
measures: The case of the numerical distance effect. Acta Psychologica, 134, 154–161.
Mandler, G., & Shebo, B. J. (1982). Subitizing: An analysis of its component processes. Journal of Experimental Psychology: General,
11, 1–22.
Mazzocco, M. M. M., Feigenson, L., & Halberda, J. (2011a). Impaired acuity of the approximate number system underlies
mathematical learning disability (dyscalculia). Child Development, 82, 1224–1237.
Mazzocco, M. M. M., Feigenson, L., & Halberda, J. (2011b). Preschoolers’ precision of the approximate number system predicts
later school mathematics performance. PLoS One, 6, 1–8.
Moyer, R. S., & Bayer, R. H. (1976). Mental comparison and the symbolic distance effect. Cognitive Psychology, 8, 228–246.
Moyer, R. S., & Landauer, T. K. (1967). Time required for judgements of numerical inequality. Nature, 215, 1519–1520.
Muldoon, K., Towse, J., Simms, V., Perra, O., & Menzies, V. (2013). A longitudinal analysis of estimation, counting skills, and
mathematical ability across the first school year. Developmental Psychology, 49, 250–257.
Mundy, E., & Gilmore, C. K. (2009). Children’s mapping between symbolic and nonsymbolic representations of number. Journal
of Experimental Child Psychology, 103, 490–502.
Passolunghi, M. C., & Lanfranchi, S. (2012). Domain-specific and domain-general precursors of mathematical achievement: A
longitudinal study from kindergarten to first grade. British Journal of Educational Psychology, 82, 42–63.
Passolunghi, M. C., Vercelloni, B., & Schadee, H. (2007). The precursors of mathematics learning: Working memory, phonological
ability, and numerical competence. Cognitive Development, 22, 165–184.
Pedhazur, E. J. (1997). Multiple regression in behavioral research: Explanation and prediction (3rd ed.). Orlando, FL: Harcourt Brace.
Penner, A. M., & Paret, M. (2008). Gender differences in mathematics achievement: Exploring the early grades and the extremes.
Social Science Research, 37, 239–253.
Penner-Wilger, M., Fast, L., LeFevre, J., Smith-Chant, B. L., Skwarchuk, S., Kamawar, D., et al (2007). The foundations of numeracy:
Subitizing, finger agnosia, and fine-motor ability. In D. S. McNamara & J. G. Trafton (Eds.), Proceedings of the 29th annual
cognitive science society (pp. 1385–1390). Austin, TX: Cognitive Science Society.
Price, G. R., Palmer, D., Battista, C., & Ansari, D. (2012). Nonsymbolic numerical magnitude comparison: Reliability and validity
of different task variants and outcome measures, and their relationship to arithmetic achievement in adults. Acta
Psychologica, 140, 50–57.
Raven, J., Court, J. H., & Raven, J. C. (1995). Coloured progressive matrices. Oxford, UK: Oxford Psychologists Press.
Reigosa-Crespo, V., Valdés-Sosa, M., Butterworth, B., Estévez, N., Rodríguez, M., Santos, E., et al (2012). Basic numerical
capacities and prevalence of developmental dyscalculia: The Havana survey. Developmental Psychology, 48, 123–135.
Sasanguie, D., De Smedt, B., Defever, E., & Reynvoet, B. (2012). Association between basic numerical abilities and mathematics
achievement. British Journal of Developmental Psychology, 30, 344–357.
Sasanguie, D., Defever, E., Maertens, B., & Reynvoet, B. (2013). The approximate number system is not predictive for symbolic
number processing in kindergartners. Quarterly Journal of Experimental Psychology.
Sasanguie, D., Göbel, S. M., Moll, K., Smets, K., & Reynvoet, B. (2013). Approximate number sense, symbolic number processing,
or number–space mappings: What underlies mathematics achievement? Journal of Experimental Child Psychology, 114,
418–431.
Sasanguie, D., Van den Bussche, E., & Reynvoet, B. (2012). Predictors for mathematics achievement? Evidence from a
longitudinal study. Mind, Brain, and Education, 6, 119–128.
Schleifer, P., & Landerl, K. (2011). Subitizing and counting in typical and atypical development. Developmental Science, 14,
280–291.
Schouten, J. F., & Bekker, J. A. M. (1967). Reaction time and accuracy. Acta Psychologica, 27, 143–153.
Stock, P., Desoete, A., & Roeyers, H. (2009a). Mastery of the counting principles in toddlers: A crucial step in the development of
budding arithmetic abilities? Learning and Individual Differences, 19, 419–422.
Stock, P., Desoete, A., & Roeyers, H. (2009b). Predicting arithmetic abilities: The role of preparatory arithmetic markers and
intelligence. Journal of Psychoeducational Assessment, 27, 237–251.
Stock, P., Desoete, A., & Roeyers, H. (2010). Detecting children with arithmetic disabilities from kindergarten: Evidence from a 3-
year longitudinal study on the role of preparatory arithmetic abilities. Journal of Learning Disabilities, 43, 250–268.
Swanson, H. L., & Jerman, O. (2006). Math disabilities: A selective meta-analysis of the literature. Review of Educational Research,
76, 249–274.
Van Bon, W. H. J. (1986). Raven’s Coloured Progressive Matrices: Manual of Dutch norms. Lisse, The Netherlands: Swets &
Zeitlinger.
Whalen, J., Gallistel, C. R., & Gelman, R. (1999). Nonverbal counting in humans: The psychophysics of number representation.
Psychological Science, 10, 130–137.

You might also like