You are on page 1of 19

Journal of Experimental Child Psychology 164 (2017) 68–86

Contents lists available at ScienceDirect

Journal of Experimental Child


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

The unique and shared contributions of arithmetic


operation understanding and numerical
magnitude representation to children’s
mathematics achievement
Terry Tin-Yau Wong
The Education University of Hong Kong, Tai Po, New Territories, Hong Kong

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

Article history: The current study examined the unique and shared contributions
Received 21 November 2016 of arithmetic operation understanding and numerical magnitude
Revised 18 May 2017 representation to children’s mathematics achievement. A sample
of 124 fourth graders was tested on their arithmetic operation
understanding (as reflected by their understanding of arithmetic
Keywords:
principles and the knowledge about the application of arithmetic
Mathematics
Arithmetic principles
operations) and their precision of rational number magnitude rep-
Commutativity resentation. They were also tested on their mathematics achieve-
Relation to operand ment and arithmetic computation performance as well as the
Arithmetic operations potential confounding factors. The findings suggested that both
Numerical magnitude arithmetic operation understanding and numerical magnitude rep-
resentation uniquely predicted children’s mathematics achieve-
ment. The findings highlight the significance of arithmetic
operation understanding in mathematics learning.
Ó 2017 Elsevier Inc. All rights reserved.

Introduction

Mathematics is a compulsory subject for elementary school students. Mastery of mathematical


skills is related to better educational attainment, financial status, and psychological well-being
(Parsons & Bynner, 2005; Ritchie & Bates, 2013). An increasing number of studies have been conducted

E-mail address: tinyau@eduhk.hk

http://dx.doi.org/10.1016/j.jecp.2017.07.007
0022-0965/Ó 2017 Elsevier Inc. All rights reserved.
T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86 69

to examine how children learn these important skills (Geary, 2010). Solving mathematical problems,
even the simplest ones such as 3 + 2 = __, requires the understanding of both numerical magnitude
(i.e., how much do 3 and 2 represent?) and arithmetic operations (i.e., what is the meaning of +?).
Inadequate understanding in either aspect would make mathematics learning difficult. For instance,
whereas an inadequate understanding of numerical magnitude may hinder children from doing arith-
metic computation (Wong, Ho, & Tang, 2014), children will probably get stuck on mathematical prob-
lem solving if they do not know which arithmetic operations are needed for calculating the area of a
triangle in geometry, and for finding out the averages in statistics, even though they may have good
understanding about numerical magnitude. Although it seems obvious that understanding of both
numerical magnitude and arithmetic operations is essential to mathematics learning, the unique con-
tributions of the two constructs remain largely unexplored. The current study, therefore, was con-
ducted to fill this important gap.

Numerical magnitude representation

Numerical magnitude representation is probably the most extensively researched topic within the
field of mathematical cognition. The major underlying assumption behind the huge body of research is
that the representation of numerical magnitude serves as the basis on which our mathematics skills
can be developed (e.g., Dehaene, 2001; Feigenson, Dehaene, & Spelke, 2004; Geary, 2013; Siegler,
2016). In the integrated theory of numerical development, Siegler and his colleagues (Siegler, 2016;
Siegler, Thompson, & Schneider, 2011) have proposed four major developments in numerical magni-
tude knowledge and how these developments are related to our mathematics skills. The first develop-
ment involves the refinement of nonsymbolic numerical magnitude representation. Humans, like
many other animals, are equipped with an innate ability to represent nonsymbolic numerosity. Such
ability allows us to discriminate between numerosities that differ by a certain ratio. For example, 6-
month-old infants can discriminate between arrays of 8 and 16 dots but not between arrays of 8 and
12 dots (Xu & Spelke, 2000). This ability, also known as the number acuity, improves with age.
Whereas kindergarteners are able to reliably discriminate arrays of 9 versus 12 dots, the ratio is sharp-
ened to approximately 9:10 during adulthood (Halberda & Feigenson, 2008; Piazza et al., 2010).
Because this approximate sense of number does not allow us to work with precise quantities, we need
to develop a symbolic number system and associate the symbolic numbers with this nonsymbolic
numerical representation (Dehaene, 2005). This is the second major development of numerical mag-
nitude knowledge. Children acquire the meaning of the number symbols through making such asso-
ciations. After they have acquired the meanings of the small numbers, they bootstrap their
understanding about these small numbers to the larger numbers. For example, by analogizing 1500
to 10,000 as 15 to 100, children can better understand the magnitude of large numbers based on their
representation of magnitudes of small numbers (Thompson & Opfer, 2010). In the final step of numer-
ical magnitude development, children learn about rational numbers. They need to realize that
although many of the rules that apply to whole numbers (e.g., each whole number has a unique pre-
decessor and successor and is represented by a unique symbol) do not apply to rational numbers, all
rational numbers can be represented as magnitudes on the mental number line. With the interference
of whole number knowledge (also known as the whole number bias; Ni & Zhou, 2005), mastering the
magnitude of rational numbers seems to be particularly difficult. Even for tasks as simple as fraction
comparison, the performance of undergraduates is far from perfect (DeWolf, Grounds, Bassok, &
Holyoak, 2014). It seems that not every adult reaches this final stage of numerical development.

Understanding of arithmetic operations

Compared with numerical magnitude representation, we have relatively little understanding about
how children understand arithmetic operations. The term arithmetic operation understanding refers to
the sense about the four arithmetic operations, such as when they should be used and the implicit reg-
ularities behind them, instead of the ability to perform arithmetic computations. In other words, it
refers to whether individuals know they should use multiplication to find out the total costs required
70 T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86

for buying a dozen cupcakes that cost $5 each, and whether they know that the answer of 12  5
should be equal to 5  12, but not whether they can work out the answer.
The existing literature on understanding of arithmetic operations focuses on how people under-
stand various arithmetic principles, which are the rules and regularities of arithmetic operations
(Prather & Alibali, 2009). Examples of arithmetic principles include commutativity (e.g., the order of
the operands is irrelevant in some operations, e.g., 4 + 7 = 7 + 4), relation to operands (e.g., addition
or multiplication of a natural number results in an increase in numerical value, whereas subtraction
or division results in the opposite, e.g., 78 + 91 > 91), and direction of effect (the change of value in
the operands affects the value of the result; e.g., for addition and multiplication with natural numbers,
increasing either of the operands increases the value of the sum/product). Previous studies suggested
that children’s understanding of arithmetic principles emerged during early childhood. Children as
young as 5 years already know that order of the addends is irrelevant in addition problems (Canobi,
Reeve, & Pattison, 2002; Farrington-Flint, Canobi, Wood, & Faulkner, 2010; Wilkins, Baroody, &
Tiilikainen, 2001). However, complete mastery of more advanced arithmetic principles, such as rela-
tion to operands and direction of effect, is not achieved even during adulthood (Dixon, Deets, &
Bangert, 2001; Siegler & Lortie-Forgues, 2015).
In the current study, the focus was on the two within-operation principles—commutativity and
relation to operand—because these two principles seem to be more directly related to children’s math-
ematics achievement. First, understanding of the commutativity principles may result in better arith-
metic fact retrieval. When children understand commutativity, they may store the arithmetic facts
(e.g., the multiplication table) in a more efficient manner. Previous studies found that the effect of
arithmetic problem-solving training (e.g., solving __ = 6  9) is generalized to the nontrained items
that are the commutativity pairs of the trained items (e.g., 6  9 vs. 9  6; Rickard, Healy, & Bourne,
1994), suggesting that commutativity pairs are stored in the same ‘‘mental units.” This may result
in more efficient retrieval of the arithmetic facts because (a) there will be less interference from the
nontarget arithmetic facts and (b) the arithmetic facts are activated twice as frequently due to their
commutativity pairs. Tentative support of the above argument can be found in some previous findings
suggesting that children who have better conceptual understanding of the arithmetic principles tend
to use retrieval more often (Canobi, Reeve, & Pattison, 1998; Canobi, Reeve, & Pattison, 2003) and have
better performance in arithmetic fact retrieval (Andersson, 2008). Second, better understanding of the
relation to the operand principle may help us in rejecting impossible answers. When we solve an addi-
tion problem (with both addends being positive numbers) and we find that the answer is smaller than
either of the addends, we know that something has gone wrong and we may attempt to solve the
problem again. This may result in higher problem-solving accuracy because we have a chance to cor-
rect wrong answers. However, these facilitation effects (arithmetic fact retrieval and rejection of arith-
metic errors) can occur only when the problem solvers understand the relevant arithmetic principles
(i.e., commutativity and relations to operand, respectively). This highlights the significance of under-
standing of arithmetic principles in children’s mathematics learning.
Apart from the understanding of arithmetic principles, the understanding of arithmetic operation
should also include the knowledge about when and where to use these arithmetic operations. Know-
ing that you need to use subtraction to calculate the change while shopping and knowing that you
need to use multiplication to calculate the area of a rectangle are examples of such knowledge. This
knowledge may play an important role in arithmetic word problem solving. When arithmetic word
problems are introduced in schools, students are taught to solve these problems (e.g., Sam has $8.
He has $2 more than John does. How much does John have?) through writing down the corresponding
equations (e.g., 8–2 = __).1 During this process, identifying the right arithmetic operations (i.e., whether
one should use + or ) plays a critical role. Without identifying the correct arithmetic operation(s), the
equations are bound to be wrong, which will lead to wrong answers even if the computations are done
accurately. This is supported by the finding that when problem solvers err in arithmetic word problems,

1
It should be noted that the formation of equation is not a necessary step for arithmetic word problem solving. Some children
may solve arithmetic word problems informally (e.g., finding the total cost of 12 lemons that cost $5 each by counting 10, 20, 30,
40, 50, and 60; Carraher, Carraher, & Schliemann, 1985). In that case, the role of arithmetic operation understanding in arithmetic
word problem solving is less obvious.
T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86 71

most of the errors are due to the use of wrong arithmetic operations (e.g., writing 8 + 2 instead of 8 2 in
the above example) (Lewis & Mayer, 1987; Muth, 1984).

Existing empirical evidence

Although both the representation of numerical magnitude and the understanding of arithmetic
operations are logically linked to children’s mathematics achievement, the efforts devoted to investi-
gate the two constructs differ greatly. The relation between the magnitude construct and children’s
mathematics achievement is well established. Children’s ability to discriminate numerosities, for
example, has been found to be related to both concurrent (Halberda, Mazzocco, & Feigenson, 2008)
and future (Libertus, Feigenson, & Halberda, 2013) mathematics achievement. Such a relation is fur-
ther confirmed by recent meta-analyses (Chen & Li, 2014; Schneider et al., 2017). Similarly, the ability
to map symbolic numbers onto nonsymbolic numerosity representation has also been found to be pre-
dictive of children’s mathematical skills (Booth & Siegler, 2006). In fact, such mapping mediates the
relation between nonsymbolic numerical representation and our mathematical skills (Libertus,
Odic, Feigenson, & Halberda, 2016; Wong, Ho, & Tang, 2016). Finally, and most important, adolescents’
magnitude knowledge about rational numbers is found to be strongly related to their mathematics
achievement (DeWolf, Bassok, & Holyoak, 2015; Torbeyns, Schneider, Xin, & Siegler, 2015). The med-
ian correlation between the fraction number line and mathematics achievement was as high as .65
(Torbeyns et al., 2015). Besides the correlational evidence, there is also experimental evidence sup-
porting the magnitude–mathematics relation. For instance, training children on nonsymbolic numer-
ical comparison and nonsymbolic arithmetic tasks results in improvement in their symbolic
arithmetic skills (Hyde, Khanum, & Spelke, 2014; Park & Brannon, 2013). Games that are designed
to improve the association between symbolic numbers and nonsymbolic magnitude representation
also help children to improve their mathematical skills (Kucian et al., 2011; Maertens, De Smedt,
Sasanguie, Elen, & Reynvoet, 2016; Ramani, Siegler, & Hitti, 2012). Finally, the effect of interventions
attempting to improve adolescents’ fraction magnitude knowledge has been found to be generalizable
to mathematics achievement outcomes (Fuchs et al., 2014). From all of the evidence above, there is
little doubt that numerical magnitude knowledge plays an important role in the acquisition of math-
ematical skills.
It should be noted that although the current study followed the practice of many other studies by
using both number comparison tasks and number line tasks to tap the understanding of numerical
magnitude (Fazio, Bailey, Thompson, & Siegler, 2014; Torbeyns et al., 2015), some recent evidence sug-
gests that the number line task may be tapping proportional reasoning instead of numerical magni-
tude representation (Barth & Paladino, 2011) and that the effects of training on these two tasks are
not generalized to each other (Maertens et al., 2016). To examine whether the two types of tasks
reflect the same underlying construct (i.e., the representation of numerical magnitude), the factor
structure of these tasks was explored in the current study.
There have been few empirical studies done to investigate the relation between arithmetic opera-
tion understanding and children’s mathematics learning. These studies mainly focused on children’s
understanding of arithmetic principles. Whereas some studies have observed significant relations
between arithmetic principle understanding and different mathematical outcomes such as more flex-
ible strategy use in mathematical problem solving (Canobi et al., 1998; Canobi et al., 2003) and higher
problem-solving accuracy (Andersson, 2008; Dowker, 2014), other studies have failed to observe a sig-
nificant relation between the two (Prather, 2012; Siegler & Lortie-Forgues, 2015; Watchorn et al.,
2014). The evidence on the link between applications of arithmetic operations and mathematics
achievement is even scarcer. In conducting a comprehensive literature search in PsycINFO, only a sin-
gle measure of this construct (i.e., the Necessary Arithmetic Operation Test; Ekstrom, French, &
Harman, 1976) was found. In this test, children are given some arithmetic word problems and are
asked to select the arithmetic operations required to solve the problems without actually solving
them. Although the content of the test is quite mathematics specific, its strong correlations with other
reasoning tests (e.g., Raven’s standard progressive matrices, Sternberg verbal analogies; Snow, 1980)
led researchers to categorize it as a general measure of intelligence (Colom, Rebollo, Palacios, Juan-
Espinosa, & Kyllonen, 2004; Snow, 1980). Its relation with mathematics achievement, however, has
72 T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86

received little attention. To summarize, despite the logical relation between arithmetic operation
understanding and mathematics achievement, there is currently little evidence supporting such a
relation.

Cultural differences in mathematics achievement

The cultural differences in mathematics achievement between Asian and Western countries are
well documented. For example, in the recent Trends in International Mathematics and Science Study
(TIMSS), Asian countries occupied the top five positions in terms of mathematics achievement in the
fourth grade, and their performance was significantly better than all of the other participating coun-
tries (Mullis, Martin, Foy, & Hooper, 2016). Whereas the more regular number-naming systems in
many Asian countries probably contribute to an earlier mastery of the number system (Miller,
Smith, Zhu, & Zhang, 1995), these differences seem to be amplified by the cultural difference in both
the valuation of mathematics (Geary, 1996; Miller, Kelly, & Zhou, 2005) and the instructions that stu-
dents receive in mathematics classrooms (Leung, 2005; Stigler, Gallimore, & Hiebert, 2000). For exam-
ple, compared with their Western counterparts, Asian students spend less time on practicing routine
procedures and spend more time on analyzing problems (Stigler et al., 2000). The mathematics class-
rooms in Asia are also suggested to be more coherent and fully developed (Leung, 2005). These cul-
tural differences may account for the outstanding mathematics achievement in Asian countries.
Despite these cultural differences in mathematics learning, most of the studies on children’s mathe-
matics learning are conducted in Western contexts. The corresponding picture in the Asian context
remains largely unexplored (see Siegler & Mu, 2008; Torbeyns et al., 2015; Wong et al., 2016, for some
exceptions). Therefore, the current study, which was conducted in Hong Kong, provided a unique
opportunity to study children’s mathematics learning in an understudied context.

The current study

Given the imbalance in the effort made to investigate the two constructs of mathematical knowl-
edge, the aims of the current study were twofold. First, with the limited and inconsistent findings on
the relation between arithmetic operation understanding and mathematics achievement despite the
logical relation between the two, the current study aimed to clarify the above relation. A sample of
fourth graders was tested on both their arithmetic operation understanding (as reflected by their
understanding of arithmetic principles and their knowledge about the application of arithmetic oper-
ations) and their numerical magnitude representation. To be developmentally sensitive, the mea-
sures of arithmetic operation understanding mainly involved natural numbers (except for one
item in the relation to operand task that involved the number 1/2), whereas rational numbers were
used to test numerical magnitude representation. This was mainly because although numerical mag-
nitude representation of natural numbers is already very well developed by the fourth grade (Booth
& Siegler, 2006), rational number knowledge is emerging. The assessment of children’s rational num-
ber magnitude representation, therefore, may provide a more sensitive measure of children’s numer-
ical magnitude knowledge at this age. On the other hand, it is not possible to test arithmetic
operation understanding with rational numbers among fourth graders because the multiplication
and division of rational numbers has not been introduced to the participants in schools. Participants
were also assessed with two mathematics achievement tasks. Their general cognitive capacities, as
well as their reading skills, were also included as controls. Children’s understanding of arithmetic
operations was expected to correlate with their concurrent mathematics achievement even after
controlling for the effects of their general cognitive skills as well as their rational number magnitude
representation. After the first goal had been reached, the current study went one step further to
examine the unique and shared concurrent contributions of the magnitude construct and the oper-
ation construct to children’s mathematics achievement. Both constructs were expected to correlate
with children’s mathematics achievement, and their contributions were expected to be roughly sim-
ilar. Clarifying the unique contributions of the two constructs will help researchers and educators to
more appropriately allocate their attention to these constructs. For example, if the operation con-
struct were found to be as important as the magnitude construct to children’s mathematics learning,
T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86 73

it would make sense to devote more effort to investigating the operation construct. Teachers might
also need to pay more attention to children’s understanding of arithmetic operations in mathematics
classrooms.

Method

Participants

The current sample consisted of 124 Chinese-speaking fourth graders with a mean age of 10 years 1
month (SD = 4 months). In this sample, 66 were boys. All participants were recruited from the partic-
ipant pool of a longitudinal study on children’s mathematics development, and the original sample
came from 17 different kindergartens all over Hong Kong. The sample in the current study did not dif-
fer from the participants who refused to participate in their initial performance.2 None of them had
been reported to have any intellectual disabilities/delayed intellectual development. In Hong Kong,
the four operations are taught during the first 2 years of primary school, so children in the final sample
should have received adequate instructions on the four operations. They should also have adequate expe-
rience in using all four operations.

Measures

Arithmetic operation concepts


Children’s understanding of arithmetic operations was assessed using three tasks designed by the
author based on the existing literature. All three tasks were conducted in a paper-and-pencil format,
with the items being read to the participants to ensure that they were not hindered by their reading
skills.
Two tasks were conducted to assess the participants’ understanding of arithmetic principles. In the
commutativity judgment task, participants were given four statements (e.g., + = + , one
for each operation), and they needed to judge whether the statements were correct by circling the cor-
responding sign (tick or cross). Shapes, instead of numbers, were used to prevent the participants from
performing computations, which could become a serious confounding factor in the examination of the
relations with mathematics achievement. A similar approach had been adopted by others (Canobi
et al., 2002; Patel & Canobi, 2010). To answer these items correctly, participants needed to know that
the commutativity principle applies to some arithmetic operations (i.e., addition and multiplication)
but not the others (i.e., subtraction and division). To account for the chance factor, participants got
one mark only if they got both addition and subtraction items correct, and they got another mark only
if they got both multiplication and division items correct, resulting in a total score of 2. The resulting
reliability was excellent, with Cronbach’s a being .92.
In the second task, which assessed the participants’ understanding of the relation to the operand
principle, participants were given statements such as ‘‘If I want to increase the value of a number, I
can transform it by (+ / /  / ) 10,” and they were asked to circle all of the operations that made
the given statements correct (e.g., + and  in the above case). They were told beforehand that all of the
numbers in the statements were positive. There were six items in this part. There was one correct
answer for the first two items and two correct answers for the remaining four items (see Table 1b
in Results for details). To include items with a wider range of difficulty, the fraction 1/2 was involved
in the last item: ‘‘If I want to increase the value of a number, I can transform it by (+ / /  / ) 1/2.”
Each correctly identified operation in the second part yielded one mark. To improve reliability, one of
the items that involved two answers was deleted, resulting in a total score of 8 for this task. The result-
ing reliability was satisfactory, with Cronbach’s a being .71.

2
We compared the performance of the dropped-out sample and the final sample on the measures that we had conducted in
Time 1 using independent-samples t tests. In all four measures that are relevant to the current study (i.e., working memory,
number comparison, number line, and arithmetic computation), none of the comparisons turned out to be significant (all ts < 1.9,
ps > .05).
74 T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86

In the arithmetic operation application task (modified from the Necessary Arithmetic Operation
Test; Ekstrom et al., 1976), participants were given some arithmetic word problems (e.g., ‘‘In a primary
school there are 104 boys and 96 girls. If every 8 students line up in a line, how many lines are for-
med?”). Unlike the usual arithmetic word problem solving, in which the problem solvers are required
to complete the whole problem-solving process and find out the answers, participants were asked
only to identify the operation(s) needed for solving the problems. They were told that there could
be more than one answer for each item. To increase the variability in item difficulty, the word prob-
lems in this task vary in different dimensions such as the type of problems (combine, compare, or
change), the position of the unknown (start or result unknown), and the number of operations
involved (one, two, or three). There were six items in total. Each correctly identified operation yielded
one mark, resulting in a total score of 13 for this task.

Rational number magnitude representation


Four tasks were conducted to assess the participants’ rational number magnitude representation,
with the first three tasks being modified from Fazio et al. (2014). In the fraction number line task, par-
ticipants were shown a number line from 0 to 1 and a fraction. Their task was to locate the fraction in
the right place on the number line by moving the mouse. The decimal number line task mimicked the
fraction number line task except that the participants needed to locate decimal numbers, instead of
fractions, on the number line. The decimals involved were equivalent to the corresponding fractions
in the fraction number line task except that they were rounded to different place value positions
(one third of the items were rounded to the tenth digit, hundredth digit, and thousandth digit, respec-
tively). For both number line tasks, participants had a chance to familiarize themselves with the num-
ber line on the instruction page. There were 24 experimental items for both tasks, which were evenly
distributed on the number line. Performance in these two tasks was measured using the percentage
absolute error (PAE), which is a commonly used indicator for assessing the participants’ performance
in number line estimation tasks (e.g., Fazio et al., 2014; Lyons, Price, Vaessen, Blomert, & Ansari, 2014).
In the fraction comparison task, two fractions were shown to the participants, who needed to select
the fraction that was numerically larger. The pairs of fractions to be compared belonged in four ratio
bins (i.e., 1.15–1.28, 1.28–1.43, 1.48–1.65, and 2.46–2.71), and there were 10 pairs of fractions in each
ratio bin. To ensure that the participants attended to both the numerators and the denominators, the
pairs of fractions were designed in a way that the larger fraction consisted of either (a) a larger numer-
ator and an equal denominator, (b) an equal numerator and a smaller denominator, (c) a larger numer-
ator and a larger denominator, (d) a larger numerator and a smaller denominator, or (e) a smaller
numerator and a smaller denominator.
In the last rational number magnitude representation task, the decimal comparison task, partici-
pants were shown two decimals. They needed to decide on the relative magnitude of the two decimals
by circling the right sign (> / < / =). There were 10 items with varying difficulties in this task. Some of
the items had incongruent digit values (e.g., 0.02 vs. 0.1), some of them had incongruent lengths (e.g.,
0.203 vs. 0.54), and some of them had zeros at the end (e.g., 0.400 vs. 0.4). For both comparison tasks,
accuracies were used in the analyses.
All rational numbers that appeared in the rational number magnitude tasks ranged from 0 to 1
because it was the range of numbers being emphasized in the local fourth-grade curriculum.

Outcome measures
Arithmetic computation. An arithmetic task was employed to assess the arithmetic skills of the partic-
ipants. The task was constructed based on the local third- to fifth-grade mathematics curriculum.
There were 35 items in the task. Of these 35 items, 26 were single-step problems and 9 involved mul-
tiple steps. In addition, 13 of the items involved integers, whereas the others involved fractions and
decimal numbers. Items were presented to the participants in both visual form (in a question card)
and verbal form (the experimenters read out the problem). Participants were given rough work sheets
for calculation. Each correctly answered item yielded one mark, resulting in a total score of 35.

General mathematics achievement. The Learning and Achievement Measurement Kit 3.0 Fourth Grade
Mathematics (LAMK 3.0; Hong Kong Education Bureau, 2015) was conducted to measure the
T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86 75

participants’ mathematics achievement. The LAMK 3.0 was constructed by the local Education Bureau
based on the local curriculum for the purpose of assessing students’ achievement. The current version
(Fourth Grade Mathematics) consisted of 39 items covering all major topics in the mathematics cur-
riculum, including arithmetic, shape and space, measure, and simple statistics. Because some of the
items contained sub-items, the total score of the task was 50. Participants were given 45 min to com-
plete the task. The raw score was used in the following analysis.

Control measures
Nonverbal intelligence. The short form (Sets A, B, and C) of the Raven’s standard progressive matrices
(Raven, 1956) was conducted to assess the participants’ nonverbal intelligence. In each item, partici-
pants were shown a large visual pattern with a missing part and needed to select the right piece of
pattern, out of six to eight options, that could fill the missing part. There were 36 items in the short
form. The raw score was converted into a scaled score based on the local norm.

Working memory. The working memory capacity of the participants was assessed using the backward
digit span task. The experimenters presented a series of digits to the participants orally at a pace of one
digit per second. Participants needed to recall the digits in a backward manner after listening. A prac-
tice trial was given to the participants before the experimental trials. There were 10 experimental tri-
als in five difficulty levels. Starting with items with two digits, items in each successive difficulty level
had one more digit to be recalled compared with the items in the previous level. Each correctly
recalled item yielded one mark. The task was discontinued when the participants scored zero in a
level.

Reading. The word reading subtest in the Hong Kong Test of Specific Learning Difficulties in Reading
and Writing for Primary School Students–Second Edition (Ho et al., 2007) was conducted to measure
the reading skills of the participants. Participants were asked to read aloud 150 Chinese two-character
words, and each correctly read word scored one mark. The task was discontinued when the partici-
pants failed to read 15 consecutive words. The raw score was converted into a scaled score based
on the local norm.

Procedures

Participants were recruited from the participant pool of a longitudinal study on children’s mathe-
matics development (it should be noted that the previous longitudinal study did not include any of the
key measures included in the current study). Parental consents were obtained before the participants
were assessed by trained experimenters in participants’ homes. The testing lasted for approximately
2 h. Souvenirs and coupons were given to the participants as a token of appreciation for their effort.
In addition to the key measures used in the current study, all of which were conducted when the
participants were in the fourth grade, the author included the participants’ domain-general cognitive
skills and reading skills in the current analyses as controls. This information was obtained in the pre-
vious longitudinal study. These measures were conducted when the participants were in kindergarten
(backward digit span), first grade (Raven’s standard progressive matrices), and second grade (word
reading).

Results

Before the analysis, the univariate outliers (data points that are 3 standard deviations or above/be-
low the means of the corresponding variables) were first winsorized (replaced by values 3 standard
deviations above/below the mean). Because the participants came from different schools, the effect
of school was examined. The effect of school on the outcome variables was not significant (Wald
Zs = 0.70 and 1.10 for general mathematics achievement and arithmetic computation, respectively,
ps > .25) and was not considered further.
76 T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86

Participants’ performance in arithmetic operation tasks

The choices of the participants in the arithmetic operation tasks are listed in Tables 1a–1c. It is
noted that although most participants were correct in approving the commutativity statements for
addition and multiplication, fewer than 70% of the participants correctly rejected the commutativity
statements for subtraction and division. This suggests that children tend to wrongly overgeneralize
the commutativity principle to subtraction and division. In the relation to operand task, most of the
participants chose the correct answers but were less likely to realize that, besides addition, multipli-
cation would also increase the original value. The same applied to the item concerning subtraction and
division. These findings suggest that children tend to undergeneralize the relation to operand principle
to multiplication and division, which makes sense given that they had less experience on these oper-
ations. Furthermore, very few participants realized that dividing, but not multiplying, a number with
1/2 increases the numerical value, which is understandable given that multiplication with fractions
has not been introduced by fourth grade. Finally, in the application of arithmetic operation task, most
of the correct answers were correctly identified. Yet, the cumulative proportion of participants choos-
ing each of the four operations was much lower than the expected value in the final two items
(accounting for up to 217% and 241% only; with three correct answers, the cumulative percentage
should be 300%), suggesting that children tend to miss required arithmetic operations in complex
word problems.

Descriptive statistics and correlations

The means, standard deviations, reliabilities, and correlations among variables are presented in
Table 2. Except for backward digit span, all of the tasks were reliable (Cronbach as > .70). The three
tasks designed for assessing the participants’ understanding of arithmetic operations were significantly
correlated with each other (rs  .42, ps < .001, median r = .51), and so were the four tasks assessing the
participants’ rational number magnitude representation (|r|s  .27, ps < .01, median r = .41). All of the
tasks assessing the magnitude and operation constructs significantly and strongly correlated with the
two outcomes (|r|s  .40, ps < .001, median r = .52). These correlations suggested that both the magni-
tude and operation constructs were strongly related to children’s mathematics outcomes.

Table 1a
Choice of the participants in commutativity judgment task.

Item no. Item Approving% Disapproving%


1 + = + 97.6 2.4

2 = 38.7 61.3

3 92.7 7.3
 = 
4  =  32.3 67.7

Table 1b
Choice of the participants in the relation to operand task.

Item Item Chose Chose Chose Chose


no. +  
1 If I want to find the triple of a number, I can (+ / /  / ) 3 14.5 1.6 93.5 4.0
2 If I want to find the half of a number, I can (+ / /  / ) 2 0.8 5.6 3.2 96.8
3a If I want to keep the value of a number unchanged, I can (+ / /  / ) 1 8.1 7.3 72.6 58.9
4 If I want to increase the value of a number, I can (+ / /  / ) 10 94.4 0 87.1 1.6
5 If I want to decrease the value of a number, I can (+ / /  / ) 4 2.4 95.2 0 83.1
6 If I want to increase the value of a number, I can (+ / /  / ) 12 88.7 2.4 65.3 8.1

a
Items deleted to improve the reliability of the task.
T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86 77

Table 1c
Choice of the participants in the application of arithmetic operation task.

Item Item % choosing Correlation with LAMK


no.
+ X  + X 
1 Lily has 91 stickers, 18 fewer than May. How many 73.4 27.4 0 3.2 .21 .20 NA .03
stickers does May have?
2a Longman has $125. After receiving $150 from his 88.7 81.5 2.4 12.1 .41 .34 .18 .13
mother, he has just the right amount to buy a toy car
that costs $399 together with his younger brother.
How much does Longman’s younger brother have?
3 The fruit shop has 6 boxes of oranges in the morning. 8.1 80.6 83.9 13.7 .02 .41 .31 .24
There are 60 oranges in each box. When the shop is
closed at night, there are 128 oranges left. How many
oranges were sold today?
4 There are 104 boys and 96 girls in the fourth grade. If 83.1 0.8 12.1 87.1 .53 .07 .15 .22
they are to line up in 8, how many lines will be
formed?
5 A mother brings her sixth-grade daughter to a buffet. 53.2 93.5 7.3 62.9 .39 .20 .02 .52
The charge is $140 for an adult and children are
charged half the original price. The mother pays $500.
How much change will she get back?
6 Basketball costs $166 each. Now it is sold at half the 77.4 8.9 78.2 76.6 .15 .07 .48 .52
original price. Each bag of badminton costs $28. How
much does it cost if I buy one basketball and 3 bags of
badminton?

Note: It should be noted that although some of the problems could be solved by alternative strategies, resulting in the choice of
arithmetic operations other than the ‘‘correct answers”, the choice of arithmetic operations other than the ‘‘correct answers”
were all negatively correlated with children’s mathematics achievement. It is therefore unlikely that children who chose
arithmetic operations other than the ‘‘correct answers” were truly aware of these alternative strategies.
a
Participants got 2 marks for choosing either subtraction only or both addition and subtraction.

Factor structure of rational number magnitude and operation constructs

Before examining the unique contributions of the two constructs to children’s mathematics
achievement, the factor structure of the measures from both constructs was examined using confirma-
tory factor analysis with the package ‘‘lavaan” in R (Rosseel, 2012). Three factor models have been pro-
posed. In the first model, all seven measures formed a single factor of mathematical knowledge. In the
second model, the three operation measures formed a factor of operation and the four rational number
magnitude measures formed a factor of rational number magnitude. In the third model, the rational
number magnitude factor was further divided into two factors according to the task demand (number
line vs. comparison). The fits of the models were evaluated based on the following criteria: insignifi-
cant chi-square (v2) test, comparative fit index (CFI) > .95, root mean square error of approximation
(RMSEA) < .006, and standardized root mean square residual (RMSR) < .08 (Hu & Bentler, 1999). Based
on these criteria, all models except Model 1 fit the data well. Therefore, the Akaike information crite-
rion (AIC) and Bayesian information criterion (BIC) values of the models were examined. The model
with the lowest AIC and BIC values was considered the best-fit model (i.e., Model 2 in the current
case). The confirmatory factor analysis (CFA), therefore, suggested that a two-factor model, with a
rational number magnitude factor capturing all of the rational number magnitude tasks and an oper-
ation factor tapping all of the tasks involving the understanding of arithmetic operations, fit the data
best (see Table 3 for details). The two factors strongly correlated with each other (r = .80, p < .001). The
factor scores were saved for the following analyses.

Predicting mathematical outcomes with rational number magnitude and operation factors

To examine the predictive power of the operation constructs on the two mathematical outcomes,
as well as the unique contributions of the rational number magnitude and operation constructs, four
78
Table 2

T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86


Descriptive statistics and correlations.

Measure Mean (SD) a Correlations


Comm. RtO Ap. Op. FNL DNL FC DC LAMK Arith. Raven BDS
Commutativity 1.23 (.94) .92 –
Relation to operand 6.48 (1.07) .71 .42*** –
Application of arithmetic operation 10.23 (2.39) .76 .51*** .64*** –
Fraction number line 12.77 (8.63) .91 .46*** .44*** .43*** –
Decimal number line 17.37 (11.13) .89 .39*** .28** .36*** .45*** –
Fraction comparison 30.57 (4.72) .75 .28** .18* .36*** .33*** .27** –
Decimal comparison 7.85 (2.38) .80 .44*** .43*** .56*** .50*** .52*** .36*** –
LAMK 40.14 (8.02) .91 .48*** .53*** .66*** .51*** .45*** .44*** .61*** –
Arithmetic 19.29 (5.45) .86 .56*** .52*** .59*** .47*** .40*** .40*** .56*** .73*** –
Raven’s 111.65 (12.70) .85 .34*** .36*** .46*** .36*** .36*** .26** .40*** .47*** .50*** –
Backward digit span 5.35 (1.31) .58 .09 .15+ .24** .06 .16+ .05 .16+ .31** .23* .24** –
Reading 11.81 (3.24) .97 .25** .27** .47*** .28** .25** .19* .39*** .48*** .44*** .27** .14

Note. Comm, commutativity; RtO, relation to operand; ApOp, application of arithmetic operation; FNL, fraction number line; DNL, decimal number line; FC, fraction comparison; DC,
decimal comparison, LAMK, Learning and Achievement Measurement Kit 3.0; Arith, arithmetic; Raven, Raven’s standard progressive matrices; BDS, backward digit span.
+
p < .10.
*
p < .05.
**
p < .01.
***
p < .001.
T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86 79

Table 3
Fit indices of the CFA models.

Model df v2 AIC BIC CFI RMSEA SRMR


One factor 14 28.399 4187.290 4246.516 .929 .091 .047
Two factors (magnitude vs. operation) 13 16.343 4175.216 4237.262 .984 .046 .040
Three factors (number line vs. comparison vs. operation) 11 15.742 4178.482 4246.169 .977 .059 .041

Note. 2, chi-square; AIC, Akaike information criterion; BIC, Bayesian information criterion; CFI, comparative fit index; RMSEA,
root mean square error of approximation; RMSR, root mean square residual.

sets of hierarchical regressions were conducted. In the first two sets of regression, general mathemat-
ics achievement was the dependent variable. The controlled variables (i.e., IQ, working memory, and
reading) were put into the first block. In Model 1A, the rational number magnitude factor was put into
the second block and the operation factor was put into the third block. This allowed the author to
examine the unique contribution of the operation factor to children’s general mathematics achieve-
ment while controlling for the potential confounding factors (e.g., intelligence, working memory, read-
ing, rational number magnitude representation). In Model 1B, the order of entry was reversed, with
the operation factor being put into the second block and the rational number magnitude factor being
put into the third block. In this way, the author could compare the unique and shared variance of the
two constructs on general mathematics achievement. The other two sets of regression models (Models
2A and 2B) were identical to the previous two except that the dependent variable was changed to
arithmetic computation.
In Model 1A, a significant amount of the variance in general mathematics achievement was
accounted for by the controlled variables. Adding the rational number magnitude factor significantly
increased the amount of explained variance (p < .001). After accounting for the controlled variables
and the rational number magnitude factor, the operation factor was still a significant predictor of gen-
eral mathematics achievement (p = .041). When the order of entry was reversed in Model 1B, the oper-
ation factor accounted for a significant amount of the variance in general mathematics achievement
(p < .001), and the rational number magnitude factor further increased the amount of explained vari-
ance (p = .006) (see Table 4 for details). In the final model, both the rational number magnitude factor
(b = .362, p = .006) and the operation factor (b = .273, p = .041) were significant predictors of general
mathematics achievement. The findings suggested that whereas a large amount of the contribution
of the two factors to general mathematics achievement was shared (19.5%), both the rational number
magnitude factor and operation factor accounted for significant amounts of unique variance in chil-
dren’s general mathematics achievement.
Next, the above regression analyses were repeated, this time with arithmetic computation being
the dependent variable. In Model 2A, after taking the controlled variables into account, the rational
number magnitude factor explained a significant amount of the variance in children’s arithmetic com-
putation (p < .001). Further inclusion of the operation factor increased the amount of explained vari-
ance significantly (p = .029). In Model 2B, the order of entry of the two constructs was reversed. The
operation factor now accounted for a significant amount of the variance in arithmetic computation
(p < .001), and the rational number magnitude factor accounted for a marginally significant amount
of the additional outcome variance (p = .091) (see Table 4 for details). In the final model, children’s
arithmetic skills were significantly predicted by the operation factor (b = .320, p = .029) and marginally
significantly predicted by the rational number magnitude factor (b = .243, p = .092). Again, the findings
suggested that the two constructs shared a large amount of their predictive power over the arithmetic
computation (15.3%). Whereas the operation factor accounted for a significant amount of unique vari-
ance in children’s arithmetic computation performance (1.9%), the rational number magnitude factor
accounted for a marginally significant amount of unique variance (1.1%).
To further examine the predictive power of each of the key measures, two multiple linear regres-
sions were conducted, with all of the variables being included in the models as predictors. Mathemat-
ics achievement and arithmetic computation were included as the dependent variables in the two
models. In terms of predicting mathematics achievement, both rational number comparison measures
turned out to be significant predictors (fraction comparison: b = .166, p = .011; decimal comparison:
80 T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86

Table 4
Regression analysis predicting general mathematics achievement and arithmetic.

Model Step Predictor Mathematics achievement Arithmetic


R2 DR2 DR2 B b R2 DR2 DR2 B b
95% CI 95% CI
1 Raven’s .385 .040 .063 .358 .067 .155*
Backward digit .870 .143* .246 .069
span
Reading .402 .163* .236 .140*
A 2 Magnitude .605 .220*** 1.478 .362** .522 .164*** .674 .243+
3 Operation .619 .014* [.0001, 4.065 .273* .541 .019* [.0008, 3.249 .320*
.0669] .0626]
*** * ***
B 2 Operation .594 .209 4.065 .273 .530 .172 3.249 .320*
3 Magnitude .619 .025** [.0006, 1.478 .362** .541 .011+ [.0001, .674 .243+
.0768] .0465]
C 2 Fraction number .630 .090 .097 .576 .022 .035
line
Decimal .038 .053 .003 .006
number line
*
Fraction .282 .166 .172 .149*
comparison
Decimal .576 .171* .328 .143+
comparison
Commutativity .604 .071 1.383 .238**
Relation to .994 .132+ .910 .178*
operand
Application of .608 .181+ .075 .033
arithmetic
operations

Note. CI, confidence interval.


+
p < .10.
*
p < .05.
**
p < .01.
***
p < .001.

b = .171, p = .034), and the understanding of the relation to operand principle (b = .132, p = .094) and
the application of arithmetic operation (b = .181, p = .053) turned out to be marginally significant pre-
dictors. (Both of the operation measures were marginally significant because the two measures shared
their contributions to general mathematics achievement. When either of the measures was removed,
the other one became highly significant; e.g., application of arithmetic operation without relation to
operand: b = .254, p = .003; relation to operand without application of arithmetic operation:
b = .203, p = .004.) Children’s mathematics achievement was also predicted by their working memory
capacity (b = .150, p = .014) and their reading skills (b = .169, p = .012). The pattern shifted a little bit
when arithmetic computation was the dependent variable. Understanding of arithmetic principles
now played a larger role in predicting the outcome (commutativity understanding: b = .238,
p = .002; relation to operand understanding: b = .178, p = .036). Whereas fraction comparison contin-
ued to be a significant predictor (b = .149, p = .032), the predictive power of decimal comparison
became marginally significant (b = .143, p = .097). Children’s arithmetic computation performance
was also predicted by their nonverbal intelligence (b = .159, p = .032) and their reading skills
(b = .171, p = .017).

Discussion

The current study was conducted to examine (a) the relation between children’s understanding of
arithmetic operations and their mathematics achievement and (b) the unique and shared contribu-
tions of the numerical magnitude representation and arithmetic operation understanding to children’s
mathematics achievement. A sample of fourth graders was assessed on their rational number
T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86 81

magnitude representation and understanding of arithmetic operations as well as their arithmetic com-
putation skills and general mathematics achievement. The results suggested that whereas the contri-
butions of the magnitude and operation aspects of mathematics to the mathematical outcomes were
largely shared, both aspects accounted for unique variance in the mathematical outcomes. Theoretical
and practical implications are discussed below.

Significance of arithmetic operation understanding

The understanding of arithmetic operations is logically related to our mathematics achievement in


various ways. For example, by understanding the commutativity principle, arithmetic facts may be
stored in a more efficient manner (Rickard et al., 1994). The inadequate knowledge about when the
arithmetic operations are used is also suggested to be one of the causes of children’s errors in arith-
metic word problem solving (Lewis & Mayer, 1987). Despite the theoretical relation between arith-
metic operation understanding and mathematics achievement, direct evidence is rare and
inconsistent. The understanding of arithmetic principles, for instance, has been found to be related
to mathematics achievement in some previous studies (Andersson, 2008; Canobi, 2005; Canobi
et al., 2003; Dowker, 2014) but not others (Prather, 2012; Siegler & Lortie-Forgues, 2015; Watchorn
et al., 2014). The current study was conducted to reexamine the issue. To ensure that the observed
relation was not driven by the shared variance with general abilities and numerical magnitude repre-
sentation, the effect of these variables was controlled. The findings suggested that children’s under-
standing of arithmetic operations was significantly related to both mathematical outcomes, echoing
previous studies that had yielded positive findings (Andersson, 2008; Canobi, 2005; Canobi et al.,
2003; Dowker, 2014).
Furthermore, the current findings elaborated on the previous findings regarding the relation
between arithmetic principle understanding and mathematics achievement in two major ways. First,
whereas previous studies that examined children’s understanding of commutativity usually found
that this understanding was related to their arithmetic problem-solving skills (Andersson, 2008;
Canobi, 2005; Dowker, 2014), most of these studies focused on the correct application of the principle
instead of the overgeneralization of this principle to other operations (i.e., subtraction and division).
The current findings suggest that the overgeneralization of this principle is also related to children’s
arithmetic computation. Children who overgeneralize the commutativity principle may commit errors
in arithmetic computation such as treating 32–7 as 37–2. Students, therefore, should be taught to be
aware of the limit of the application of the commutativity principle. On the other hand, the current
study observed a significant relation between the understanding of the relation to operand principle
and children’s performance in arithmetic computation. This contrasts with the findings of Siegler and
Lortie-Forgues (2015) and Prather (2012). These differences in findings may be explained by various
factors such as the cultural differences (the other two studies were conducted in a Western setting),
the number notations involved, and the mathematical outcomes being considered. For example, while
the first study found that the students’ understanding of the relation to operand principle concerning
fractions (a more advanced understanding that involved the understanding of both the principle and
the properties of fractions) was largely unrelated to their fraction arithmetic, the second study found
that the effects of relation to operand intervention did not generalize to students’ number sentence
construction skills. The current findings suggest that when we focus on integers (in contrast to
Siegler & Lortie-Forgues, 2015, who focused on fractions), children’s understanding of the relation
to operand principle is actually related to their arithmetic computation skills (in contrast to
Prather, 2012, who focused on number sentence construction skills). The current findings, therefore,
further elaborate on the specific relations between the understanding of different arithmetic princi-
ples and different components of mathematics achievement.

Unique and shared contributions of arithmetic operation understanding and numerical magnitude
representation

After establishing the relation between arithmetic operation understanding and mathematics
achievement, the other major goal of the current study was to examine the unique and shared contri-
82 T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86

butions of arithmetic operation understanding and numerical magnitude representation to children’s


mathematics achievement. The results suggested that both the understanding of arithmetic operation
and the representation of rational number magnitude independently correlated with the two mathe-
matical outcomes, meaning that both of them play unique roles in children’s mathematics learning.
Furthermore, a large portion of their contributions to the mathematical outcomes was shared, sug-
gesting that the two constructs are related. Sensitivity to numerical magnitude, for example, may help
children to acquire some of the arithmetic principles (e.g., relation to operand). The interaction
between the magnitude and operation constructs should be further investigated in future studies.

Representation of rational number magnitudes

Besides showing that the understanding of both rational number magnitude and arithmetic oper-
ations uniquely contributed to children’s mathematics achievement with similar magnitudes, the cur-
rent data allowed the author to examine the factor structure behind the commonly used rational
number magnitude measures. Whereas number comparison and number line tasks were commonly
perceived as reflecting the same underlying construct—numerical magnitude understanding (Fazio
et al., 2014)—playing linear board games and training with the fraction number line improved both
number line and number comparison performance (Hamdan & Gunderson, 2017; Siegler & Ramani,
2009); other studies suggested that the number line may measure proportional reasoning instead
of numerical magnitude representation (Barth & Paladino, 2011) and that the effects of number com-
parison versus number line training are not generalized to the other task (Maertens et al., 2016). To
address this controversy, a three-factor model (the number line tasks and the number comparison
tasks formed different factors) was included in the CFA so that its model fit could be compared with
that of the two-factor model (both number line tasks and number comparison tasks belonged to a sin-
gle factor). The results suggested that the two-factor model fit the data better than the three-factor
model and, hence, provided support to the idea that both number comparison and number line tasks
captured the same underlying construct—numerical magnitude representation.
A related issue is that when the four rational number magnitude measures were included as sep-
arate variables in the regression models, only number comparison tasks, but not number line tasks,
significantly correlated with the mathematical outcomes. One potential reason is that number com-
parison may be a more sensitive measure of numerical magnitude because it involves the representa-
tion of two varying numerical magnitudes (e.g., the comparison between 3/4 and 5/6) instead of one
numerical magnitude (i.e., in the number line task, the only varying numerical magnitude was the tar-
get; the two end points were always fixed). The same logic should apply to all kinds of numbers. Yet, it
should be noted that the opposite pattern has been observed in other studies (e.g., Torbeyns et al.,
2015). Given the differences in task design (e.g., number range, presence of a fixed anchor) as well
as the age group of the participants, no clear conclusion can be made here. Future studies may further
look into this issue by including different versions of number line and number comparison tasks in a
single study.

Theoretical and educational implications

By demonstrating that both the representation of numerical magnitude and the understanding of
arithmetic operations uniquely contributed to children’s mathematical outcomes, the current study
has significant theoretical and practical implications. Whereas the field of research on children’s
mathematics learning has been expanding during recent decades, many of the studies focused on
the representation of numerical magnitude (Chen & Li, 2014; Schneider et al., 2017; Siegler, 2016).
The current findings suggest that, besides the magnitude construct, the operation construct also plays
a significant role in children’s mathematics learning. The operation construct should receive much
more attention than it actually does. A lot more can be done to investigate children’s understanding
of arithmetic operations. By a quick review of the literature, the author found that most of the liter-
ature on arithmetic principle understanding focused on some of the arithmetic principles (e.g., com-
mutativity, inversion; Canobi et al., 2003; Gilmore & Bryant, 2006; Robinson & Dubé, 2009) but not on
others (e.g., relation to operand, direction of effect). Furthermore, these studies used different single-
T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86 83

facet assessment in assessing arithmetic principle understanding. For instance, some studies ask the
participants to evaluate the use of principle-derived procedures (Canobi et al., 2002; Patel & Canobi,
2010), whereas other studies require the participants to apply these principles in problem solving
(Andersson, 2008; Dowker, 2014; Gilmore & Bryant, 2006; Robinson & Dubé, 2013). These assess-
ments reflect different facets of principle understanding, and the use of single-facet assessment might
not reflect the complete picture of participants’ understanding. For this reason, Prather and Alibali
(2009) advocated the use of multifaceted assessment in assessing people’s arithmetic principle under-
standing. By including a more comprehensive multifaceted assessment of arithmetic principle under-
standing, we can have a much better understanding about the important factors facilitating or
hindering children’s mathematics learning. The knowledge can then be translated into practice. Our
mathematics curriculum can be improved by incorporating the relevant elements about arithmetic
operation understanding. Previous studies suggested that exposure to both principle-consistent and
principle-inconsistent arithmetic errors facilitates the mastery of the relevant arithmetic principle
(Prather & Alibali, 2011) and that blocking the principle-inconsistent errors makes the training more
effective (Prather, 2012). Directing children’s attention to the relative magnitudes of the numbers in
the equations may also aid in the acquisition of arithmetic principles (Prather & Alibali, 2011). Incor-
porating these elements into our mathematics curriculum may be a cost-effective way of improving
children’s mathematics achievement. Yet, it should be noted that this kind of intervention needs to
be carefully designed so that children’s attention is directed to the key components. Otherwise, the
intervention may promote mindless computation, which can be detrimental to the understanding
of mathematical concepts (McNeil, 2008).

Limitations and future directions

One of the potential limitations of the current study is that the control measures were not mea-
sured concurrently with the independent and dependent variables. Yet, these constructs have been
shown to be highly stable (Deary, 2014; Lee & Bull, 2016; Pan et al., 2011), and the time lag between
the measures should not affect the findings in any significant ways. The large amount of outcome vari-
ance being explained by these constructs (>35%) provided support to the claim. The readers should
also note that different notations of numbers were used in testing different kinds of numerical knowl-
edge (rational numbers for magnitude and natural numbers for arithmetic operations) to increase the
sensitivity of the measures. In fact, if magnitude knowledge of whole numbers was assessed, instead
of that of rational numbers, the contribution of numerical magnitude measures may be further
reduced. This possibility remains to be tested by aligning the number notations in the two kinds of
measures in future studies. Furthermore, shapes, instead of numbers, were used to test children’s
commutativity understanding. The use of shapes ensured that the children did not solve the problems
through computation, which could be a serious confounding factor. It should be noted that the use of
shapes inevitably elicited children’s abstract or algebraic reasoning. In fact, the understanding of com-
mutativity itself is a form of algebraic reasoning (Blanton & Kaput, 2005). The question that the author
sought to address in this study was whether this particular aspect of algebraic reasoning is important
for mathematics learning at the stage where formal algebra has not been introduced, and the findings
suggest that this understanding is important for children’s arithmetic computation even though none
of the items in the arithmetic computation task required algebraic reasoning.
Most important, although a significant relation between arithmetic operation understanding and
mathematics achievement was observed, the readers should interpret this relation with caution.
Due to the correlational nature of the current study, the direction of influence between the two con-
structs cannot be inferred. Although it makes sense that better understanding of the arithmetic oper-
ations facilitates children in solving mathematical problems in various ways (e.g., facilitating
arithmetic fact retrieval, rejection of arithmetic errors, selection of the right operations), the reverse
is also possible. In fact, Dowker (2009), Dowker (2014) suggested that children’s arithmetic skills
served as a prerequisite of the understanding of arithmetic principles. This makes sense because chil-
dren with better arithmetic skills may free up more attentional resources during mathematical prob-
lem solving, making the discovery of arithmetic principles more likely (Watchorn et al., 2014). In fact,
the relation between these two constructs may be similar to that between procedural and conceptual
84 T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86

mathematics skills; that is, the development of these two constructs may be iterative (Rittle-Johnson,
Siegler, & Alibali, 2001). The current findings do not allow us to tease apart different possibilities.
Intervention studies are needed to address the directionality issue.

Conclusion

By demonstrating that children’s understanding of arithmetic operation was at least equally impor-
tant as their numerical magnitude understanding for the development of mathematical skills, the cur-
rent study highlights the need for further research on children’s understanding of arithmetic
operations. Through conducting comprehensive multifaceted assessment of arithmetic principle
understanding (Prather and Alibali, 2009), the exact nature of the relation between arithmetic opera-
tion understanding and mathematics achievement can be unfolded, which may in turn provide new
insights for the development of novel mathematics interventions.

Acknowledgments

The current study was supported by the Internal Research Grant (Grant ID: RG 99/2014-2015) from
the Education University of Hong Kong. The author would like to thank all the participants for their
continuous support to this project, as well as the student interns who have devoted substantial effect
to the data collection process.

References

Andersson, U. (2008). Mathematical competencies in children with different types of learning difficulties. Journal of Educational
Psychology, 100, 48–66.
Barth, H. C., & Paladino, A. M. (2011). The development of numerical estimation: Evidence against a representational shift.
Developmental Science, 14, 125–135.
Blanton, M. L., & Kaput, J. J. (2005). Characterizing a classroom practice that promotes algebraic reasoning. Journal of Research in
Mathematics Education, 36, 412–446.
Booth, J. L., & Siegler, R. S. (2006). Developmental and individual differences in pure numerical estimation. Developmental
Psychology, 42, 189–201.
Canobi, K. H. (2005). Children’s profiles of addition and subtraction understanding. Journal of Experimental Child Psychology, 92,
220–246.
Canobi, K. H., Reeve, R. A., & Pattison, P. E. (1998). The role of conceptual understanding in children’s addition problem solving.
Developmental Psychology, 34, 882–891.
Canobi, K. H., Reeve, R. A., & Pattison, P. E. (2002). Young children’s understanding of addition concepts. Educational Psychology,
22, 513–532.
Canobi, K. H., Reeve, R. A., & Pattison, P. E. (2003). Patterns of knowledge in children’s addition. Developmental Psychology, 39,
521–534.
Carraher, T., Carraher, D., & Schliemann, A. (1985). Mathematics in the streets and in schools. British Journal of Developmental
Psychology, 3(1), 21–29.
Chen, Q., & Li, J. (2014). Association between individual differences in non-symbolic number acuity and math performance: A
meta-analysis. Acta Psychologica, 148, 163–172.
Colom, R., Rebollo, I., Palacios, A., Juan-Espinosa, M., & Kyllonen, P. C. (2004). Working memory is (almost) perfectly predicted by
g. Intelligence, 32, 277–296.
Deary, I. J. (2014). The stability of intelligence from childhood to old age. Current Directions in Psychological Science, 23, 239–245.
Dehaene, S. (2001). Précis of ‘‘The number sense”. Mind & Language, 16, 16–36.
Dehaene, S. (2005). Evolution of human cortical circuits for reading and arithmetic: The ‘‘neuronal recycling” hypothesis. In S.
Dehaene, J. R. Duhamel, M. Hauser, & G. Rizzolatti (Eds.), From monkey brain to human brain: A Fyssen Foundation symposium
(pp. 133–157). Cambridge, MA: MIT Press.
DeWolf, M., Bassok, M., & Holyoak, K. J. (2015). From rational numbers to algebra: Separable contributions of decimal
magnitude and relational understanding of fractions. Journal of Experimental Child Psychology, 133, 72–84.
DeWolf, M., Grounds, M. A., Bassok, M., & Holyoak, K. J. (2014). Magnitude comparison with different types of rational numbers.
Journal of Experimental Psychology: Human Perception and Performance, 40, 71–82.
Dixon, J. A., Deets, J. K., & Bangert, A. (2001). The representations of the arithmetic operations include functional relationships.
Memory & Cognition, 29, 462–477.
Dowker, A. (2009). Use of derived fact strategies by children with mathematical difficulties. Cognitive Development, 24, 401–410.
Dowker, A. (2014). Young children’s use of derived fact strategies for addition and subtraction. Frontiers in Human Neuroscience,
7. http://dx.doi.org/10.3389/fnhum.2013.00924.
Ekstrom, R. B., French, J. W., & Harman, H. H. (1976). Manual for Kit of Factor-Referenced Cognitive Tests. Princeton, NJ:
Educational Testing Service.
Farrington-Flint, L., Canobi, K. H., Wood, C., & Faulkner, D. (2010). Children’s patterns of reasoning about reading and addition
concepts. British Journal of Developmental Psychology, 28(Pt 2), 427–448.
T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86 85

Fazio, L. K., Bailey, D. H., Thompson, C. A., & Siegler, R. S. (2014). Relations of different types of numerical magnitude
representations to each other and to mathematics achievement. Journal of Experimental Child Psychology, 123, 53–72.
Feigenson, L., Dehaene, S., & Spelke, E. (2004). Core systems of number. Trends in Cognitive Sciences, 8, 307–314.
Fuchs, L. S., Schumacher, R. F., Sterba, S. K., Long, J., Namkung, J., Malone, A., et al (2014). Does working memory moderate the
effects of fraction intervention? An aptitude–treatment interaction. Journal of Educational Psychology, 106, 499–514.
Geary, D. C. (1996). International differences in mathematical achievement: Their nature, causes, and consequences. Current
Directions in Psychological Science, 5, 133–137.
Geary, D. C. (2010). Mathematical disabilities: Reflections on cognitive, neuropsychological, and genetic components. Learning
and Individual Differences, 20, 130–133.
Geary, D. C. (2013). Early foundations for mathematics learning and their relations to learning disabilities. Current Directions in
Psychological Science, 22, 23–27.
Gilmore, C. K., & Bryant, P. (2006). Individual differences in children’s understanding of inversion and arithmetical skill. British
Journal of Educational Psychology, 76(Pt 2), 309–331.
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.
Halberda, J., Mazzocco, M. M. M., & Feigenson, L. (2008). Individual differences in non-verbal number acuity correlate with
maths achievement. Nature, 455, 665–668.
Hamdan, N., & Gunderson, E. A. (2017). The number line is a critical spatial–numerical representation: Evidence from a fraction
intervention. Developmental Psychology, 53, 587–596.
Ho, C. S.-H., Chan, D. W.-O., Chung, K. K., Tsang, S.-M., Lee, S.-H., & Cheng, R. W.-Y. (2007). The Hong Kong Test of Specific Learning
Difficulties in Reading and Writing for Primary School Students-Second Edition [HKT-P(II)]. Hong Kong: Hong Kong Specific
Learning Difficulties Research Team.
Hong Kong Education Bureau (2015). The Learning and Achievement Measurement Kit 3.0. Hong Kong: Author.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new
alternatives. Structural Equation Modeling, 6, 1–55.
Hyde, D. C., Khanum, S., & Spelke, E. S. (2014). Brief non-symbolic, approximate number practice enhances subsequent exact
symbolic arithmetic in children. Cognition, 131, 92–107.
Kucian, K., Grond, U., Rotzer, S., Henzi, B., Schönmann, C., Plangger, F., et al (2011). Mental number line training in children with
developmental dyscalculia. NeuroImage, 57, 782–795.
Lee, K., & Bull, R. (2016). Developmental changes in working memory, updating, and math achievement. Journal of Educational
Psychology, 108, 869–882.
Leung, F. K. S. (2005). Some characteristics of East Asian mathematics classrooms based on data from the TIMSS 1999 video
study. Educational Studies in Mathematics, 60, 199–215.
Lewis, A. B., & Mayer, R. E. (1987). Students’ miscomprehension of relational statements in arithmetic word problems. Journal of
Educational Psychology, 79, 363–371.
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.
Libertus, M. E., Odic, D., Feigenson, L., & Halberda, J. (2016). The precision of mapping between number words and the
approximate number system predicts children’s formal math abilities. Journal of Experimental Child Psychology, 150,
207–226.
Lyons, I. M., Price, G. R., Vaessen, A., Blomert, L., & Ansari, D. (2014). Numerical predictors of arithmetic success in Grades 1–6.
Developmental Science, 5, 714–726.
Maertens, B., De Smedt, B., Sasanguie, D., Elen, J., & Reynvoet, B. (2016). Enhancing arithmetic in pre-schoolers with comparison
or number line estimation training: Does it matter? Learning and Instruction, 46, 1–11.
McNeil, N. M. (2008). Limitations to teaching children 2 + 2 = 4: Typical arithmetic problems can hinder learning of
mathematical equivalence. Child Development, 79, 1524–1537.
Miller, K. F., Kelly, M., & Zhou, X. (2005). Learning mathematics in China and the United States: Cross-cultural insights into the
nature and course of mathematical development. In J. I. D. Campbell (Ed.), Handbook of mathematical cognition
(pp. 163–177). New York: Psychology Press.
Miller, K. F., Smith, C. C. M., Zhu, J., & Zhang, H. (1995). Preschool origins of cross-national differences in mathematical
competence: The role of number-naming systems. Psychological Science, 6, 56–60.
Mullis, I. V. S., Martin, M. O., Foy, P., & Hooper, M. (2016). TIMSS 2015 international results in mathematics. <http://timssandpirls.
bc.edu/timss2015/international-results/>.
Muth, K. D. (1984). Solving arithmetic word problems: Role of reading and computational skills. Journal of Educational
Psychology, 76, 205–210.
Ni, Y., & Zhou, Y.-D. (2005). Teaching and learning fraction and rational numbers: The origins and implications of whole number
bias. Educational Psychologist, 40, 27–52.
Pan, J., McBride-Chang, C., Shu, H., Liu, H., Zhang, Y., & Li, H. (2011). What is in the naming? A 5-year longitudinal study of early
rapid naming and phonological sensitivity in relation to subsequent reading skills in both native Chinese and English as a
second language. Journal of Educational Psychology, 103, 897–908.
Park, J., & Brannon, E. M. (2013). Training the approximate number system improves math proficiency. Psychological Science, 24,
2013–2019.
Parsons, S., & Bynner, J. (2005). Does numeracy matter more? London: National Research and Development Centre for Adult
Literacy and Numeracy.
Patel, P., & Canobi, K. H. (2010). The role of number words in preschoolers’ addition concepts and problem-solving procedures.
Educational Psychology, 30, 107–124.
Piazza, M., Facoetti, A., Trussardi, A. N., Berteletti, I., Conte, S., Lucangeli, D., et al (2010). Developmental trajectory of number
acuity reveals a severe impairment in developmental dyscalculia. Cognition, 116, 33–41.
Prather, R. W. (2012). Implicit learning of arithmetic regularities is facilitated by proximal contrast. PLoS ONE, 7(10), e48868.
86 T.T.-Y. Wong / Journal of Experimental Child Psychology 164 (2017) 68–86

Prather, R. W., & Alibali, M. W. (2009). The development of arithmetic principle knowledge: How do we know what learners
know? Developmental Review, 29, 221–248.
Prather, R. W., & Alibali, M. W. (2011). Children’s acquisition of arithmetic principles: The role of experience. Journal of Cognition
and Development, 12, 332–354.
Ramani, G. B., Siegler, R. S., & Hitti, A. (2012). Taking it to the classroom: Number board games as a small group learning activity.
Journal of Educational Psychology, 104, 661–672.
Raven, J. C. (1956). Standard progressive matrices Sets A, B, C, D, & E. Oxford, UK: Psychologists Press.
Rickard, T. C., Healy, A. F., & Bourne, L. E. Jr., (1994). On the cognitive structure of basic arithmetic skills: Operation, order, and
symbol transfer effects. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 1139–1153.
Ritchie, S. J., & Bates, T. C. (2013). Enduring links from childhood mathematics and reading achievement to adult socioeconomic
status. Psychological Science, 24, 1301–1308.
Rittle-Johnson, B., Siegler, R. S., & Alibali, M. W. (2001). Developing conceptual understanding and procedural skill in
mathematics: An iterative process. Journal of Educational Psychology, 93, 346–362.
Robinson, K. M., & Dubé, A. K. (2009). Children’s understanding of the inverse relation between multiplication and division.
Cognitive Development, 24, 310–321.
Robinson, K. M., & Dubé, A. K. (2013). Children’s additive concepts: Promoting understanding and the role of inhibition. Learning
and Individual Differences, 23, 101–107.
Rosseel, Y. (2012). Lavaan: An R package for structural equation. Journal of Statistical Software, 48(2), 1–36.
Schneider, M., Beeres, K., Coban, L., Merz, S., Schmidt, S. S., Stricker, J., & De Smedt, B. (2017). Associations of non-symbolic and
symbolic numerical magnitude processing with mathematical competence: A meta-analysis. Developmental Science.
Advance online publication. doi: http://dx.doi.org/10.1111/desc.12372.
Siegler, R. S. (2016). Magnitude knowledge: The common core of numerical development. Developmental Science, 19, 341–361.
Siegler, R. S., & Lortie-Forgues, H. (2015). Conceptual knowledge of fraction arithmetic. Journal of Educational Psychology, 107,
909–918.
Siegler, R. S., & Mu, Y. (2008). Chinese children excel on novel mathematics problems even before elementary school.
Psychological Science, 19, 759–763.
Siegler, R. S., & Ramani, G. B. (2009). Playing linear number board games—but not circular ones—improves low-income
preschoolers’ numerical understanding. Journal of Educational Psychology, 101, 545–560.
Siegler, R. S., Thompson, C. A., & Schneider, M. (2011). An integrated theory of whole number and fractions development.
Cognitive Psychology, 62, 273–296.
Snow, R. E. (1980). Aptitude processes. In R. E. Snow, P.-A. Federico, & W. E. Montague (Eds.). Aptitude, learning, and instruction,
Vol. 1: Cognitive process analyses of aptitude (1, pp. 27–63). Hillsdale, NJ: Lawrence Erbaum.
Stigler, J. W., Gallimore, R., & Hiebert, J. (2000). Using video surveys to compare classrooms and teaching across cultures:
Examples and lessons from the TIMSS video studies. Educational Psychologist, 35, 87–100.
Thompson, C. A., & Opfer, J. E. (2010). How 15 hundred is like 15 cherries: Effect of progressive alignment on representational
changes in numerical cognition. Child Development, 81, 1768–1786.
Torbeyns, J., Schneider, M., Xin, Z., & Siegler, R. S. (2015). Bridging the gap: Fraction understanding is central to mathematics
achievement in students from three different continents. Learning and Instruction, 37, 5–13.
Watchorn, R. P. D., Bisanz, J., Fast, L., LeFevre, J.-A., Skwarchuk, S.-L., & Smith-Chant, B. L. (2014). Development of mathematical
knowledge in young children: Attentional skill and the use of inversion. Journal of Cognition and Development, 15, 161–180.
Wilkins, J. L., Baroody, A. J., & Tiilikainen, S. (2001). Kindergartners’ understanding of additive commutativity within the context
of word problems. Journal of Experimental Child Psychology, 79, 23–36.
Wong, T. T.-Y., Ho, C. S.-H., & Tang, J. (2014). Identification of children with mathematics learning disabilities (MLDs) using
latent class growth analysis. Research in Developmental Disabilities, 35, 2906–2920.
Wong, T. T.-Y., Ho, C. S.-H., & Tang, J. (2016). The relation between ANS and symbolic arithmetic skills: The mediating role of
number–numerosity mappings. Contemporary Educational Psychology, 46, 208–217.
Xu, F., & Spelke, E. S. (2000). Large number discrimination in 6-month-old infants. Cognition, 74, B1–B11.

You might also like