See discussions, stats, and author profiles for this publication at: [Link]

net/publication/232472405

The Model Method: Singapore Children's Tool for

Representing and Solving Algebraic Word Problems

Article in Journal for Research in Mathematics Education · May 2009

DOI: 10.5951/jresematheduc.40.3.0282

CITATIONS READS

144 10,608

2 authors, including:

Kerry Lee
The Education University of Hong Kong
82 PUBLICATIONS 2,564 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

SKIP Singapore Kindergarten Impact Project View project

OER 17/11 KL: Executive Functioning and Academic Performance: From the Laboratory to the Classroom View project

All content following this page was uploaded by Kerry Lee on 28 April 2016.

The user has requested enhancement of the downloaded file.

Journal for Research in Mathematics Education
2009, Vol. 40, No. 3, 282–313

The Model Method:

Singapore Children’s Tool for
Representing and Solving
Algebraic Word Problems
Swee Fong Ng and Kerry Lee
Nanyang Technological University

Solving arithmetic and algebraic word problems is a key component of the Singapore
elementary mathematics curriculum. One heuristic taught, the model method, involves
drawing a diagram to represent key information in the problem. We describe the model
method and a three-phase theoretical framework supporting its use. We conducted 2
studies to examine teachers’ perceptions and children’s application of the model
method. The subjects were 14 primary teachers from 4 schools and 151 Primary 5 chil-
dren. The model method affords higher ability children without access to letter-
symbolic algebra a means to represent and solve algebraic word problems. Partly
correct solutions suggest that representation is not an all-or-nothing process in which
model drawing is either completely correct or completely incorrect. Instead, an incor-
rect solution could be the consequence of misrepresentation of a single piece of
information. Our findings offer avenues of support in word problem solving to chil-
dren of average ability.
Key words: Algebra; Elementary, K–8; Problem-solving; Representation, modeling;
Teaching practice

Solving arithmetic and algebraic word problems is a key component at every level
of the Singapore primary mathematics curriculum (Curriculum Planning Division,
1990; Curriculum Planning & Development Division (CPDD), 2000a). However,
the solution of word problems is one of the most problematic areas in the mathe-
matics curriculum. Why are children challenged by word problems? What sort of
instructional strategies can be provided to support children in such tasks?
Success in solving elementary word problems requires more than sound concep-
tual knowledge (Riley, Greeno, & Heller, 1983). Findings from the solution of
physics word problems show how skilled physics problem solvers solved problems
first by constructing elaborate representations of the problem, rather than solving
directly from the text-based problem description. These representations often
include diagrams that make certain relationships and constraints conspicuous

This study was partly funded by an EDRF grant, EP2/02KL, and a CRPP grant,
CRP22/03KL. The authors thank Ee-Lynn Ng, Zee-Ying Lim, and Yu-Xi Peh for
their assistance in data collection, and Leone Burton for her comments on earlier
drafts of this article.

Copyright © 2009 The National Council of Teachers of Mathematics, Inc. [Link]. All rights reserved.
This material may not be copied or distributed electroncally or in any other format without written permission from NCTM.
Swee Fong Ng and Kerry Lee 283

(Larkin, McDermott, Simon, & Simon, 1980). Other studies, conducted with both
adults and children, show that it is the ability to represent a given word problem
and the ability to identify the appropriate computation, rather than algorithmic skill,
that determines success in solving word problems (De Corte, Verschaffel, & De
Win, 1985; Lee, Ng, & Ng, in press; Lewis, 1989).
The representation of problems should provide opportunities to (a) reflect on the
representation, (b) make modifications, and (c) select a solution strategy (Briars &
Larkin, 1984; Riley et al., 1983; Willis & Fuson, 1988). Of particular importance
to our study are the challenges involved in children’s acquisition and usage of
various forms of algebraic representations.
Children encounter a variety of obstacles in using formal symbolic algebra to
represent word problems. These include (a) understanding the meaning of letters
used in symbolic algebra (Küchemann, 1981); (b) translating natural language
into equations (e.g., Carpenter, Corbitt, Kepner, Lindquist, & Reys, 1981; Clement,
1982; Herscovics, 1989; MacGregor, 1991; Stacey & MacGregor, 2000); (c) under-
standing the semantic structure of word problems, in particular the nature of the rela-
tionships between quantities and how they are linked (Bednarz & Janvier, 1996);
and (d) using text-based semantic cues in the construction of equations (Martin &
Bassok, 2005).
In view of the cognitive demands posed by algebraic representations, with what
other form of representation could young children work? A number of studies have
shown that the use of visual and concrete representations improves performance
in solving word problems. In a study conducted with adults, Lewis (1989) taught
96 college students in the United States to use a diagramming method for repre-
senting compare word problems. This method involved drawing a number line and
asking students first to place on it a known value, then an unknown value related
to the known value. Students’ representation skills improved after the teaching
experiment. Willis and Fuson (1988) taught two classes of second graders of
average and above-average ability to use schematic drawings to solve word prob-
lems. Again, the results were positive. Further evidence comes from research on
the Russian elementary curriculum by Davydov (Davydov, 1962; Davydov &
Steffe, 1991; Freudenthal, 1974). In this curriculum, children learn to use lines to
model part-whole relationships between quantities in their 1st year of primary
schooling. They also use letters to express relationships between quantities and solve
for missing wholes and parts using addition and subtraction. Children then learn to
use equations to represent and solve two-step word problems. By the 3rd year of
primary school, they are able to represent and solve word problems requiring
proportional reasoning. Davydov’s curriculum develops children’s skill in solving
complex word problems by drawing upon their ability to use visual models to
analyze and express quantitative relationships, and to manipulate these relationships
symbolically.
In Singapore, primary school children are taught a visual and concrete approach
to solve arithmetic word problems and algebraic—the term we use for start-
unknown—word problems. In 1983, the Singapore Ministry of Education (MOE)
284 Model Method and Algebraic Word Problems

officially introduced a heuristic involving diagram or model drawing known as the

model method into the primary mathematics curriculum. The model method can
be used as a tool for solving both arithmetic and algebraic word problems involving
whole numbers, fractions, ratios, and percents (Kho, 1987). It was believed that if
children were provided with the means to visualize a word problem—be it a simple
arithmetic word problem or an algebraic word problem—the structure underlying
the problem would be made overt. Once children understood the structure of the
problem, they were more likely to be able to solve it (Kho, 1987).
The model method focuses on the heart of the word problem solution—the
importance of representation. The model method also requires children to work with
three modes of representation: text, pictorial, and symbolic. It builds upon children’s
knowledge of part-whole relationships in numbers. In the lower grades, objects,
pictures, and symbols are used to model combinations of numbers. In the higher
grades, children are taught to use appropriately sized rectangles to represent the
information presented in word problems. In arithmetic word problems, the rectan-
gles represent specific numbers. By changing the role of the rectangles and using
them to represent unknown quantities, the model method also can be used to depict
algebraic word problems. Such representations effectively serve as “pictorial equa-
tions” (Cai et al. 2005, p. 8). Although the model method is very similar to
Davydov’s approach, there are two main differences. Davydov emphasized the use
of letters as variables and the construction of equations to solve word problems.
These are not the foci of the Singapore primary mathematics curriculum. In the
model method, the rectangles represent unknown values, but the solution for the
unknown is still grounded in children’s knowledge of arithmetic.
In this article, we first provide details of the model method and show how it is used
to solve arithmetic and algebraic word problems that require the construction of only
one model drawing. Second, we discuss the theoretical framework highlighting the
three phases of problem solving in which children engage when they use the model
method to solve word problems. Third, we provide evidence from two studies
showing how teachers and children use the model method to solve algebraic word
problems. Fourth, we use examples of children’s partly correct solutions to show that
representation is not an all-or-nothing process. Such solutions are used to illustrate
what may have prevented children from using the model method successfully. In the
concluding section we discuss the implications of the findings from these two
studies for mathematics education, in particular, how average-ability children can
be supported in the model representation process of word problem solving.

THE MODEL METHOD—A HEURISTIC FOR SOLVING

ARITHMETIC AS WELL AS ALGEBRAIC WORD PROBLEMS

In Singapore, a cartoon-like version of the model method is introduced in Primary

1 (entrance age of 6) or Primary 2. Children are taught to use pictures of bears and
other familiar objects to model information presented in arithmetic word problems.
Later, to increase the level of abstraction, rectangles are used to replace pictures.
Swee Fong Ng and Kerry Lee 285

Beginning in Primary 3, children are taught to apply the same heuristic to solve alge-
braic word problems involving whole numbers. This heuristic—the model method—
gives children early access to such problems by circumventing the need to construct
and solve linear algebraic equations. Algebraic word problems involving fractions
and ratios are introduced in Primary 5 and 6.
The concept of letters as variables and the construction of algebraic expressions
with one unknown are taught to Primary 6 children (CPDD, 2000a). The construc-
tion and solution of algebraic equations with one variable are taught in the 1st
year of high school: Secondary 1. However, even younger children (9- to 11-year-
olds) are taught to solve algebraic word problems that in algebra texts would be
represented by equations such as x ± a = c, x × a = c, (c > a, a ∈ N), x ÷ a = c,
(a, c ∈ N). (In this section, the letters a, b, c, and n are used to represent known
values, and x is used to represent unknown values.)
These models serve as visual analogues that capture all the information in a word
problem. Their structure consists of a series of rectangles in which the relationships
of the rectangles are specified and are presented globally. Children use the struc-
ture of the model to help them construct appropriate sets of step-by-step arithmetic
procedures to solve given problems.

Some Generic Examples of Models Commonly Taught to Primary Children

In this section, we present three different types of model drawings commonly

taught to children. These are the (a) part-whole, (b) comparison, and (c) multipli-
cation and division models. In the model method, letters are not used to represent
numerical values, known or unknown.

The Part-Whole Model

The part-whole model can be used to solve arithmetic problems similar to the
following Cruise Problem or algebraic word problems of the form depicted by the
following Orchid Problem.
Cruise Problem: On Saturday, 1050 people went on a cruise. On Sunday, 1608
people went on a cruise. How many people went on the cruise over the 2 days?
Orchid Problem: There were 2659 visitors at Orchid Gardens. 447 of them were
adults and the rest were children. How many children visited Orchid Gardens?
Such problems are commonly found in Primary 3 textbooks (for examples, see
Collars, Koay, Lee, & Tan, 2007, p. 43).
The part-whole model can be used to represent the arithmetic situation represented
by a + b = x or the algebraic case of x + a = b. In the arithmetic situation, a + b = x,
the model consists of two rectangles of different lengths representing the quanti-
ties a and b (see Figure 1, at left). The unknown total, x, is represented by a brace
that links the two rectangles together. The unknown is found by summing a and b.
286 Model Method and Algebraic Word Problems

Hence, for the Cruise Problem, a and b represent the number of people for Saturday
and Sunday and x, the total number of people who went on the cruise.
The model drawing for the algebraic word problem (see Figure 1, at right), has
the same structure as the model drawing for the arithmetic word problem. The alge-
braic model can be used to solve for x in x + a = b. Where a and b are given, x could
be found by subtracting a from b. In the algebraic model drawing, no letters are used
to represent unknowns. Instead, a rectangle with a question mark within it signi-
fies an unknown. Such a rectangle is known as one unit, and this unit rectangle
assumes the role of letters as variables. Rectangles also can be used to represent
specific numerical values, such as the numerical relationships between two vari-
ables, but such relationships are indicated using appropriately sized rectangles. In
Figure 1 at right, the rectangle with the letter a represents the given numerical rela-
tionship between b and the unknown. In the Orchid Problem, an example of a start-
unknown problem, the rectangle with the letter a could represent the number of
adults and the rectangle with the letter b the total number of visitors. The number
of children, represented by x, can be found by subtracting a from b.

? b

a b ? a

a+b=x x+a=b

Figure 1. Part-whole models: Arithmetic model (at left) and algebraic model (at right).

The Comparison Model

The comparison model can be used to solve the following problems. The first,
identified as the Enrollment Problem, is an arithmetic word problem; the second,
the Animal Problem, is an algebraic or start-unknown word problem.
Enrollment Problem: Dunearn Primary School has 280 pupils. Sunshine Primary
School has 89 pupils more than Dunearn Primary. Excellent Primary has 62 pupils
more than Dunearn Primary. How many pupils are there altogether?
Animal Problem: A cow weighs 150 kg more than a dog. A goat weighs 130 kg
less than the cow. Altogether the three animals weigh 410 kg. What is the mass
of the cow?1
The comparison model shows the relationship between two or more quantities
when they are compared. The varying lengths of the rectangles show that one quan-

1 Given the age of the students in the study, we did not distinguish between mass and weight.
Swee Fong Ng and Kerry Lee 287

tity is bigger than another, and the difference between the quantities is indicated
by the difference in lengths of the rectangles. The arithmetic model (see Figure 2,
at left) represents the situation in which a = b + d and the sum of a and b is x. With
all the information made overt, the unknown total can be found by summing a and
b. An algebraic model drawing with the same structure as the arithmetic model
drawing can be used to solve the algebraic problem of the form x + b = c, where
the rectangle with the question mark is the unit rectangle that assumes the role of
the letter x (see Figure 2, at right).

a ?
? c

b d b d

a+b=x x+b=c

Figure 2. Comparison models: Arithmetic model (at left) and algebraic model (at right).

The model at left in Figure 3 depicts a more complex model, which is used to
solve the Enrollment Problem, a type of problem commonly found in Primary 4 text-
books. This model drawing has three sets of rectangles. Each set of rectangles is
labeled, for example, DP for Dunearn Primary. The numerical value represented
by each rectangle is noted and what has to be calculated is indicated with a ques-
tion mark. The total enrollment of the three schools is found by summing all the
numbers in the rectangles. The algebraic model (see Figure 3, at right) with the same
structure can be used to solve the Animal Problem.

DP 280 dog 150

SP 280 89 ? cow ? 410

EP 280 62 goat

20

Figure 3. Comparison model used to represent the Enrollment Problem (at left) and the
Animal Problem (at right).
288 Model Method and Algebraic Word Problems

In the Animal Problem, the comparison model is used to show the relationships
among the masses of the three animals. Here the mass of the dog is used as the
generator (Bednarz & Janvier, 1996) of the model drawing. In the model method,
the rectangle representing the mass of the dog is the unknown unit. Representation
of the masses of the two remaining animals is based on that of the dog. The differ-
ences in mass between the animals are recorded by rectangles indicating the
known values. The total mass of the three animals is indicated on the right. The
model drawing becomes a pictorial equation representing 3 units + 170 = 410. The
value of one unit can be found by solving the equation 3 units + 170 = 410. The
mass of the cow is found by summing the value of one unit and the difference
between the mass of the cow and the dog. The pictorial equation represents an alge-
braic equation of the form x + x + 150 + x + 20 = 410, where x represents the mass
of the dog.
Although a model with a similar structure can be used to solve arithmetic as well
as algebraic word problems, the solution to the algebraic word problem is more chal-
lenging. Using the model to solve arithmetic problems is a direct process since it
involves summing the numerical inputs represented by the rectangles, as demon-
strated by the solution to the Enrollment Problem depicted in Figure 3, at left. In
algebraic word problems, the outputs are given and the intent is to solve for the input.
Solving for the unknown input requires a different way of thinking. In the Animal
Problem depicted in Figure 3, at right, masses of the animals cannot be calculated
using step-by-step procedures with a selected input. Instead, to solve for the
unknown input, all three unknown states or rectangles have to be considered
together. Also, any of the three unknown states can be used as an entry point to the
problem. The mass of the cow can be found directly if it is chosen as the generator,
and the cow rectangle is the unit rectangle. Alternatively, it can be found indirectly
if the mass of the dog or the mass of the goat is chosen as the generator. This alter-
native solution is discussed further in the context of children’s responses.

The Multiplication and Division Model

Models can also be used to represent problems involving multiplication and divi-
sion as well as fractions. The model at left in Figure 4 represents the arithmetic situ-
ation 3a = b and a + b = x. The following is an example of a Primary 3 word problem
that can be solved using this arithmetic model.
Bala took 24 pictures. David took 3 times as many pictures as Bala. How many
pictures did the two boys take in all? (Collars, Koay, Lee, & Tan, 2003, p. 84)
Figure 4, at right, shows the model representing the algebraic situation x + b = a
and b = nx, where n represents the multiplicative relationship between a and b.
In this instance, the total is given but one of the parts is an unknown. This alge-
braic model can be used to solve this word problem introduced at Primary 5. The
question mark in the bottom rectangle indicates the base used for comparison, in
this case, the sum of money held by Mary.
Swee Fong Ng and Kerry Lee 289

b b

a a a
? a
a ?

a+b=x

Figure 4. Multiplication and division models for arithmetic word problem (at left) and alge-
braic word problem (at right).

Mary and John have $48 altogether. John has three times as much money as Mary.
How much money has Mary? (Collars, Koay, Lee, Ong, & Tan, 2008, p. 33).
The multiplication and division model illustrates the efficacy of the model
method in solving problems involving fractions. With the model method, rather than
operating on fractions, children are provided the option of using concepts of, and
operating on, whole numbers to solve problems involving fractions. For example,
using the model method, children need not operate on fractions to solve the
following word problem involving fractions introduced at Primary 5.
If the volume of water in container A is 1/4 the volume of water in container B,
and the total volume of water in both containers is 250 litres, find the volume of
water in container A.
By working backward, children identify the volume of B as four times that of A.
Hence, one rectangle is used to represent the volume of A while B is represented
by four rectangles of the same size. The algebraic model in Figure 5 illustrates the
relationship between A and B and all the information provided in the question. The
model drawing shows that since there are five equal parts, then 5 units = 250 l.
Hence, 1 unit = 50 l and container A’s volume is 50 l. As with the other models,
the multiplication and division model can be used to solve algebraic problems in
which x + b = c and x = (1/n)b.

THE EFFICACY OF THE MODEL METHOD: THEORETICAL CONTEXT

We hypothesized that children who use the model method to solve word prob-
lems are guided through three phases of problem solution presented schematically
in Figure 6. This theoretical framework is constructed based on the processing model
for solving arithmetic word problems presented by Kintsch and Greeno (1985) and
on our observations of teachers and children who used the model method to solve
word problems. The theoretical model we present is for consistent-language
290 Model Method and Algebraic Word Problems

Figure 5. Example of an algebraic multiplication and division model.

Figure 6. Possible phases of problem solution by children who used the model method to
solve word problems.
Note: In the TS route, children represent information embedded in the text in the structure of a model.
In the SP route (the solid line), to solve the problem, children represent the relationships embedded in
the model in a series of arithmetic equations. In the TP route, some children bypass the TSP route, moving
directly from T to P. In the TPS route, children represent the textual information in a set of arithmetic
expressions or equations, which they then use to draw the model. This alternative route is represented
by the path in dotted lines. Double-headed arrows are used to signify that children may alternate
between representations to check the accuracy of the representations.

compare word problems that require the construction of a single model drawing that
typify word problems for Primary 5 children in Singapore. Consistent-language
word problems are those in which the comparative relationships such as more than
are in agreement with the arithmetic operation for solution.
Kintsch and Greeno’s (1985) processing model for solving arithmetic word
problems works with two modes of representation, text-based information and
abstract-problem representation. The text-based representation comprises sets of
information abstracted from the word problem, which are then organized into a
meaningful macrostructure highlighting pertinent concepts and relationships
between sets of information. The abstract-problem representation contains relevant
information derived from the text but presented in an abstract form that with appli-
cation of suitable calculation strategies will result in the solution of the problem.
Swee Fong Ng and Kerry Lee 291

Although the Kintsch and Greeno model discusses the processes and the knowl-
edge involved at each phase of representation, the exact form of the abstract-
problem representation is not specified. With the model method, however, the
abstract-problem representation is delineated into two specific phases, the structural
phase and the procedural-symbolic phase, with the nature of each phase clearly spec-
ified. The model drawing becomes a macrostructure capturing the inputs, the rela-
tionships between the inputs, and the output of the problem. This information is then
translated into a series of arithmetic equations. The model drawing is one layer of
the multilayered representation necessary for the solution of word problems. Thus,
this model extends the work by Kintsch and Greeno.

Phase 1: Text Phase (T)

Children read the information presented in text form.

Phase 2: Structural Phase (S)

In this phase, children represent the text information in the structure of the
model. Children can alternate between text and the model to check that the model
accurately depicts the textual information.

Phase 3: Procedural-Symbolic Phase (P)

Once they have constructed a model, children then use the model to plan and
develop a sequence of logical arithmetic equations, which allows for the solution
of the problem. Again, alternating between the two representations, structural and
procedural (SP), allows for the accuracy of the arithmetic equations to be checked
against the model.
In the text phase, many iterations of the reading are conducted. The first reading
provides the context of the story. In subsequent readings of the text, information is
further processed to ascertain what the givens are and what is to be found. Each bit
of information is termed a chunk. Each chunk may refer to a specific numerical quan-
tity or specific information related to an unknown quantity. In the Enrollment
Problem, the information “Dunearn Primary School has 280 students” is a chunk.
The chunk “Sunshine Primary School has 89 more students than Dunearn Primary”
stipulates the comparative relationship between the two schools. Each chunk is
represented by at least one rectangle. If the comparative relationship between two
chunks is one of more than, then two rectangles may be used to represent that infor-
mation. For example, two rectangles can be used to represent the enrollment of
Sunshine Primary or the mass of the cow in the Animal Problem. With the model
method, there is continual coordination between the text phase and the structural
phase. After representing one chunk of information, the child returns to the text for
the next chunk of information to be represented. In the structural phase, the focus
is on how to represent each chunk; hence, alternations between the text phase and
292 Model Method and Algebraic Word Problems

the structural phase will continue until all the information has been represented. Each
rectangle or set of rectangles becomes a pictorial chunk representing text infor-
mation. Translation between the structural phase and the procedural-symbolic
phase takes place once the model drawing is completed. The arithmetic equations
are the foci of the final phase, and the information processing takes place in several
steps. Each alternation between the structural phase and the procedural phase may
result in an arithmetic equation that captures the information of a pictorial chunk
or two chunks.
Rectangles of arithmetic word problems represent specific numerical inputs. If
the model method is used to solve algebraic word problems, the rectangles can either
represent unknown inputs, such as the mass of the animals in the Animal Problem,
or specific numerical values, such as the difference in mass between two particular
animals. Whether the model method is used to solve arithmetic or algebraic word
problems, children go through these three phases of problem solution, albeit not
necessarily in a fixed order. Where appropriate, children may first translate the text
information into a set of arithmetic expressions (TP) and then use these to construct
a model (TPS). The objects of the procedural phase could either be a series of arith-
metic equations or a set of algebraic equations, the latter being dependent on chil-
dren’s cognizance with equations. Regardless of the routes taken, the products at
the second phase provide evidence as to whether children have understood the word
problem. Similarly, the arithmetic expression presented in the final phase provides
evidence of children’s interpretation of the model.
Solution of word problems is contingent upon, inter alia, sound conceptual
knowledge (Vergnaud, 1982), knowledge of part–part–whole relationships of
numbers (Carpenter & Moser, 1982), an integrated and well-organized knowledge
base, and metacognitive processes (Verschaffel, Greer, & De Corte, 2000). Sound
conceptual knowledge is vital because though different word problems may have
the same mathematical structure, their underlying conceptual structures may be very
different. The discussion of generic examples of models shows that although
certain arithmetic and algebraic word problems may share the same model, models
for algebraic word problems have very different underlying conceptual structures
from those underpinning arithmetic word problems.
The model method utilizes knowledge of the part–part–whole relationship of
numbers and that a rectangle is used to represent each part, a relatively longer
rectangle for a bigger number, a shorter rectangle for a smaller number, or a
rectangle of arbitrary length for an unknown unit. An integrated and well-organized
knowledge base of how to represent information such as a known unit, an unknown
unit, and comparative relationships such as more than, less than, and as many as
are vital for the construction of models.
Successful utilization of the model method is also contingent upon metacogni-
tive processes. For example, it may be necessary first to represent the text infor-
mation with an appropriate set of arithmetic expressions that can be used to draw
the model (T → P → S). Also, as any unknown could be used to solve an algebraic
word problem, a decision has to be made to select the appropriate generator. In both
Swee Fong Ng and Kerry Lee 293

of these cases, metacognition is necessary (a) to decide whether to use a TSP route
or a TPS route and (b) in generator selection.
The affordance of a visual and concrete receptacle for given numbers or unknowns
means that primary children without prior knowledge of letter-symbolic algebra can
use the model method to solve algebraic word problems. The large numbers
involved make it unlikely that problem-solving procedures such as counting all or
counting on can be used. Rather the method capitalizes on children’s facility with
arithmetic operations to solve for the value of the unknown unit represented by
rectangles. By a process of undoing, the value of the unknown unit is found. Hence,
children circumvent the cognitive demands that are inherent in solving algebraic
equations, which normally require the construction of a system of equivalent equa-
tions. Children can manipulate and operate on the rectangles without having to
engage with many of the abstractions and difficulties associated with symbolic repre-
sentations reported in the research literature (Behr, Erlwanger, & Nichols, 1980;
Booth, 1988; Kieran, 1981; Küchemann, 1981; Schoenfeld & Arcavi, 1993; Stacey
& MacGregor, 1999; Steinberg, Sleeman, & Ktorza, 1991; Usiskin, 1988).

THE PRESENT STUDIES

Singapore has a centralized education system with the Ministry of Education

spearheading curriculum development and implementation. Hence, all schools
follow a common mathematics curriculum and a common syllabus provided by the
Ministry. Prior to 2006, all schools used a common textbook series. Although
schools can now select textbooks from a number of commercially produced series,
all these series are approved by the Ministry. All textbook series introduce and
develop the model method as a problem-solving tool.
There is a dearth of research on the delivery of the model method and on chil-
dren’s use of the model method. Two studies were conducted: Study 1 examined
teachers’ perception of and facility with the model method, and Study 2 explored
how children used the model method. Although it would have been preferable to
observe teachers in their teaching of the model method, funding and timing
constraints made this impractical. Instead, Primary 5 teachers were asked to demon-
strate and explain how they used the model method to solve a set of algebraic word
problems that were used in the second study. Because data for the second study were
collected as part of a larger study on the relationship between working memory and
problem-solving proficiency (Lee, Ng, Ng, & Lim, 2004), constraints on the amount
of time we had with each child precluded detailed interviews with the children.
Instead, teachers were asked to help us understand children’s responses.

Study 1: Teachers

We interviewed four heads of department (HODs) and 14 Primary 5 mathematics

teachers, 4 of whom were from the five schools participating in Study 2. The
objectives of interviews with the HODs were to find out how and when the model
294 Model Method and Algebraic Word Problems

method was introduced to children. Interviews with the teachers were conducted
to help us understand the key ideas on which they believed children had to focus
when using the model method.

Method

Participants
Contact with schools was made via teachers who participated in professional
development courses conducted by the first author. All HODs and one Primary 5
mathematics teacher from each of the five participating schools were invited to take
part in the interview, and all but those from one school agreed to participate. Ten
primary teachers, two from each of five other neighborhood schools, also were inter-
viewed. All the HODs had held their positions for more than 10 years and all 14
teachers had been teaching Primary 5 mathematics for at least 5 years. Primary 5
mathematics teachers were chosen because it was at this grade that the model method
was used to solve a wider variety of word problems.

Instrument
An instrument comprising 10 questions was designed for Studies 1 and 2. Except
for Question 1, which is an arithmetic word problem, the remaining questions are
algebraic. One arithmetic word problem was used to ensure that as many children
as possible were able to participate in the test. Curricular materials such as textbooks
and worksheets used by different schools guided construction of the items. Various
versions of the test instrument were designed and piloted, and instruments that
resulted in extreme scores were rejected. The items listed in Table 1 were chosen
because the piloted version showed that children gave varied responses to the same
items. A group of primary mathematics teachers attending a professional develop-
ment course, but who were not involved in the Study with Teachers, also was asked
to comment on the suitability of the final set of items. The final set of items was
chosen because they required the application of a variety of mathematical concepts
taught to children by the end of Primary 4. Solution of the problems required profi-
ciency with concepts of fraction and number theory up to and including factors and
multiples of numbers. No questions tested the concepts of ratio and percent because
the study was conducted before such concepts were taught. The first 5 problems can
be solved with the construction of a single model drawing. The last 5 problems are
more challenging than the first 5, in part because more than one model drawing would
be needed if the model method were used to solve the last 5 problems.

Interview With the HODs and the Teachers

Although the mathematics questions were e-mailed to the respective teachers
ahead of the interview, the teachers chose to solve the questions during the interview
itself. In this article, we report teachers’ responses to two questions: (a) What are the
Swee Fong Ng and Kerry Lee 295

Table 1
Word Problems Presented to Children and Success Rates as a Function of Group
Success rates (percents)
Entire
group EM1 EM2
The problems (n = 151) (n = 31) (n = 120)
11. Enrollment Problem:a Dunearn Primary School has 63 87 57
280 pupils. Sunshine Primary School has 89 more
pupils than Dunearn Primary. Excellent Primary
has 62 more pupils than Dunearn Primary. How
many pupils are there altogether?
12. Furniture Problem: At a sale, Mrs. Tan spent $530 44 84 33
on a table, a chair, and an iron. The chair cost $60
more than the iron. The table cost $80 more than
the chair. How much did the chair cost?
13. Animal Problem: A cow weighs 150 kg more than 37 74 28
a dog. A goat weighs 130 kg less than the cow.
Altogether the three animals weigh 410 kg. What
is the mass of the cow?
14. Water Problem: A tank of water with 171 liters of 21 58 11
water is divided into three containers, A, B, and C.
Container B has three times as much water as con-
tainer A. Container C has 1/4 as much water as con-
tainer B. How much water is there in container B?
15. Book Problem: A school bought some mathematics 21 55 12
books and four times as many science books. The
cost of a mathematics book was $12 while a science
book cost $8. Altogether the school spent $528.
How many science books did the school buy?
16. Cat Problem: Mae Ling bought a new box of cat 11 42 3
biscuits. In the first week, she gave the cat half the
biscuits and three more. In the second week, she
gave the cat half of the remaining biscuits and 3
more. In the third week, she gave the cat half of the
remaining biscuits and 3 more. There was only 1
piece of biscuit left. How many pieces of biscuits
were there in the new box?
17. Age Problem: Mr. Raman is 45 years older than 16 42 9
his son, Muthu. In 6 years time, Muthu will be 1/4
his father’s age. How old is Muthu now?
18. Stamp Problem: Vincent bought a total of 62 pieces 0 0 0
of $3 and $5 stamps. Altogether Vincent spent $254.
How many $3 stamps were there?
19. Money Problem: Each week, Yah Hui gets $6 more 10 16 8
pocket money than Philip. Each week, Yah Hui and
Philip each spend $19 on books. After some weeks,
Yah Hui saved $98 while Philip saved $56. How
much pocket money does Philip get each week?
10. Class Problem: During a class game, Peter threw a 5 16 2
ball to 4 times as many boys as girls. Mei Lin threw
the ball to 5 times as many boys as girls. If Peter
and Mei Lin threw the ball to every student in the
class, how many boys were there in the class?
a Problems are named for ease of reference.
296 Model Method and Algebraic Word Problems

important points to look for when drawing the model for a given problem? and (b)
How are children taught to solve for the value of the unknown unit if they are neither
taught to construct nor taught to transform equations? The HODs and the mathe-
matics teachers were interviewed on separate occasions. Each interview session was
audiotaped and required approximately 1 hour to complete.

Findings

Both the HODs and the teachers reported that the model method was taught to
their students. All but one school introduced the model method at the Primary 2 level.
The remaining school did so at Primary 1. Other problem-solving heuristics were
introduced to children as they progressed through the primary years. The HODs
explained that an even less abstract precursor activity to the model method, such
as pictures, was used to introduce the part-whole model to Primary 1 children. For
example, pictures of five brown teddy bears and three white teddy bears could be
used to represent the part-whole concept of addition of 5 + 3 at Primary 1. With
older children, the pictures were replaced by rectangles within the models.
What are the important points to look for when using the model method to solve
a given problem? The teachers explained that care had to be exercised when
drawing the rectangles so that the entire model drawing was meaningful. For
example, when the model method is used to solve arithmetic word problems, the
lengths of the rectangles should be in proportion to each other. This was possible
for arithmetic word problems because the rectangles represent specific numbers.
In the Enrollment Problem, for example, the length of the rectangle representing
the difference in enrollment between Dunearn Primary and Sunshine Primary is 89,
so the length of this rectangle should be about one third of the rectangle representing
the enrollment of Dunearn Primary. Also, all the known information should be
recorded onto the model and question marks used to identify unknown values that
are to be evaluated. When it was necessary to partition a rectangle into smaller
rectangles, dotted lines were used. The use of different types of lines helped chil-
dren keep track of the problem structure, whether the multiplication or division oper-
ation was involved.
How are children taught to solve for the value of the unknown unit if they are
neither taught to construct nor taught to transform equations? Because primary
children were neither taught to construct nor taught to transform equations, they
were taught the unitary method as a strategy to evaluate the value of the unknown
unit. This strategy, which takes place in the procedural-symbolic phase of the model
method, involves the construction of a series of arithmetic equations that repre-
sent the unknown units. This set of arithmetic equations is solved by undoing
related operations. In the Animal Problem, for example, if the mass of the dog was
the chosen generator, then the rectangle representing the mass of the dog is the base
for comparison. Children were taught to find the sum represented by the three
rectangles by undoing the differences from the total mass (410 – 170). This suggests
Swee Fong Ng and Kerry Lee 297

that the model drawing was a form of pictorial equation in which if a certain
known amount was removed from the left side of the brace, then the same amount
needed to be removed from the right of the brace. The brace functions as if it were
the pivot of a balance or the equal sign of an algebraic equation. If three unknown
units were equivalent to 240, then the value of one unknown unit was found by
dividing the difference by three (240 ÷ 3). In this way, children were required to
undo the multiplication operation. The mass of the cow was found by summing the
value of one unknown unit with its difference with the dog (80 + 150 = 230). At
no time were children required to indicate what each individual expression repre-
sented, but they were expected to know the meaning of each expression (e.g., that
the expression 410 – 170 represented the total value of three units).

Discussion
Because we did not have access to classroom teaching, we cannot comment on
the nature of the delivery of lessons across the years. This is a limitation of this study.
Nevertheless, the four schools in this study had a structure to ensure the teaching
of heuristics. Although HODs and teachers were from different schools, they
agreed that the visual and concrete nature of the model method made it a useful
problem-solving tool. Because they used the unitary method strategy to solve for
the unknown unit, they avoided the need to construct equivalent equations. The
objects of the procedural phase were a series of logical arithmetic equations, not a
set of equivalent algebraic equations. Such a strategy circumvented the need to teach
transformation of equations, which the teachers explained was beyond the primary
school syllabus.

Study 2: Children
In this study, we examined how 151 children from five primary schools chose
to use the model method to solve five word problems of increasing difficulty.

Method
Participants
Of the 151 Primary 5 children recruited from five neighborhood schools—non-
selective schools situated in housing estates—in Singapore, 77 (51%) were boys,
and 74 were girls (49%). The children’s average age was 10.7 years (SD = 0.65).
All children participated with parental consent. Most children in this study were
drawn from middle-class to lower middle-class areas. In Singapore, Primary 5
classes are streamed, or ability-tracked, according to performance in languages and
mathematics (Ministry of Education, 2005). In this study, 21% of the children were
from the top stream (henceforth known as EM1 children), and the remainder was
from the middle stream (EM2 children). In all subjects, EM1 and EM2 children
shared the same syllabus and the same Primary School Leaving Examination. In
298 Model Method and Algebraic Word Problems

addition, EM1 children completed a more demanding mother tongue examination.

The lowest stream, EM3, was not recruited because these children followed a
different mathematics syllabus. Many children in this study were nonnative speakers
of English but had at least 7 years of education, with English as the medium of
instruction. All examinations were written in English.

Instrument
Table 1 lists the 10 problems presented to the children in Study 2. Because teachers
and children shared the same instrument, the discussion on instrument design
appears in the description of Study 1.

Administration of the Test

Using the pilot studies as a guide, the children were given 1 hour to complete
the 10-item mathematics test. They sat for the test in a classroom assigned by the
teacher. Although children may have been familiar with other problem-solving
heuristics, they were asked to use only the model method to solve these problems
because the use of this method was the focus of this study. Some children asked
the researcher what actions they should take if they were unable to use the model
method. These children were advised to use any method they could think of to solve
the word problems.

Findings

Because the children were not interviewed, interpretations are based on analysis
of children’s written work. Compared to the first 5 problems, children’s rate of
response to problems 6 through 10 is low (M = 0.63). Because of the low rate of
response to problems 6 through 10, only responses to the first 5 questions (see Table
1) are discussed in detail. One reason for this low performance is that more than
one model drawing is required to capture the problem state. The use of such
models, which depict the before and after states, is more common after the intro-
duction of ratio. When the children participated in this study, they had yet to be
taught this concept.
Children’s written work was coded at different levels. Responses were first
coded question by question. Each question was coded for accuracy. The correct
responses were then checked for variation in model drawings. Next, the errors were
coded into four categories: (a) partly correct models with essential information
missing or misrepresented, (b) changing generators midway, (c) correct model
drawing and correct arithmetic equations but failure to keep the goal in mind, and
(d) lack of the conceptual knowledge necessary for the solution of the five prob-
lems. Finally, we compared and contrasted the difference in responses between the
EM1 and EM2 groups. Responses using other methods were coded as “Did not use
the model method” and, for the purpose of this study, were coded as wrong. Even
when model drawings were wrong, children’s computations were checked to ascer-
Swee Fong Ng and Kerry Lee 299

tain whether the children were proficient with arithmetic computations. In this
section, we first provide descriptive statistics for the data. Second, we provide exam-
ples of how some children successfully used the model method to solve the five
problems. Finally, we use examples of children’s errors to illustrate how children
failed to use the model method successfully.
Table 1 provides children’s overall success rate in using the model method to solve
each of the 10 problems, and success rates by academic stream. Children were more
successful on the Enrollment Problem than on the other problems. This is not
surprising, given that the former is an arithmetic word problem and the rest are alge-
braic. Children found the Furniture Problem to be the easiest of all the algebraic
word problems. If symbolic algebra is used, then linear equations with one vari-
able are needed to solve Questions 2, 3, and 4, whereas the last question would
require solution of two simultaneous linear equations. That the EM1 children
performed better than the EM2 children is also not surprising because they were
streamed according to their mathematics and language performance. The mean score
of EM1 children was 3.58 compared to a mean of 1.43 for the EM2 children.
Analysis of the correct solutions by these two groups of children showed no differ-
ences in the models they constructed and the undoing method they used to solve
for the unknown unit.
Table 2 shows the performance of children with the model method. About 23%
of the children did not answer any of the first 5 questions correctly. Overall, 9% of
the children obtained the full score. About 15% of the children used the model
method to solve the Enrollment Problem, the only arithmetic problem, and at least
one algebraic word problem.

Table 2
Children’s Performance on the First Five Problems
Number correct 0 1 2 3 4 5
Percent of children 23 27 17 15 9 9

Children from one school were taught the model method in first grade, and chil-
dren from the other four schools were taught it in second grade. Performance in this
school did not differ significantly from the average performance of the other four,
t(58.106) = –0.798, p = .428. Difference in the amount of experience with the model
method did not seem to have an impact on children’s success rates.

Successful Use of the Model Method

Arithmetic word problem. The Enrollment Problem is an arithmetic word problem
involving the comparative term more than. Rectangles for Sunshine Primary and
Excellent Primary were drawn based on the rectangle for Dunearn. Unlike the
rectangles in the other four questions, all the rectangles in the Enrollment Problem
300 Model Method and Algebraic Word Problems

carried the given values. About 63% of the children used the model method to
answer this question correctly. Figure 7 shows the two types of correct responses
offered by children. Although the model drawings were similar, the arithmetic solu-
tions were not. Of the 63%, 26% presented the solution shown on the left, which
is the more efficient of the two. Here, children first used multiplication to find the
total of the values represented by the three identical rectangles and then summed
this total with the differences in enrollment between pairs of schools. Solutions of
the type in the right panel consisted of a series of independent arithmetic expres-
sions that matched, step-by-step, the structure of the model. A total of 73% of the
correct responses were of this type.

Figure 7. Examples of correct solutions to the Enrollment Problem: Efficient use of the arith-
metic model (at left) and step-by-step use of the arithmetic model (at right).

The teachers were asked to comment on these two types of solutions. They
confirmed that they would have presented similar model drawings and that both sets
of solutions were acceptable. They explained that many children preferred the step-
by-step solution shown on the right since it meant that they found the enrollment of
each school first; its logic appeared to be more amenable to children. The solution
on the left was more complicated because it no longer referred to the individual enroll-
ment of the schools. Instead, it required children to move away from the context of
the story and to focus on the structure of the arithmetic relationships—three groups
of 280 and then the individual numbers giving the differences in enrollment.
Algebraic word problems involving whole numbers. Children offered two types
of correct models to the Animal Problem. About 29% of the children provided the
correct response as shown on the left in Figure 8, where the mass of the dog was
taken as the generator. Nine percent of the children gave the correct solution in which
the mass of the cow was used as the generator, as in right panel of Figure 8. A
Swee Fong Ng and Kerry Lee 301

rectangle was drawn to represent the mass of the cow. Because the mass of both
the dog and the goat were less than the mass of the cow, rectangles shorter than that
for the cow were drawn to represent their respective masses, and the difference in
mass was indicated on the respective rectangles. By using the mass of the cow as
the generator, a homogeneous or common operation—addition of the differences
in mass to the total mass—was used to solve for the mass of the cow. This solution
is more efficient and elegant than the former, in which the mass of the dog was the
generator, because the mass of the cow can be found directly without having first
to refer to the mass of the dog.

Figure 8. Solutions to the Animal Problem: Use of the mass of the dog as generator (at left)
and use of the mass of the cow as generator (at right).

It is noteworthy that while 9% of the children chose the mass of the cow as a gener-
ator, which gave the solution of the mass of the cow directly, none of these chil-
dren used the cost of the chair as a generator for the furniture problem. If they had
done so, they would have found the cost of the chair directly. This choice, however,
would necessitate the use of nonhomogeneous operations—addition and subtrac-
tion of the difference in cost between the table and the chair and the difference in
cost of the iron and the chair, respectively, to and from the total cost. These chil-
dren who selected the mass of the cow as the generator were able to recognize which
generator to use and for which type of questions.
Children’s models showed that they represented all the unknown units and the
known values as specified rectangles to the left of the brace. The total known output
was kept to the right of the brace. Children could then solve for the unknown unit
in one of two ways. The procedures on the right in Figure 8, offered by one child,
showed that by equalizing the lengths of the rectangles representing the mass of the
302 Model Method and Algebraic Word Problems

goat and the dog to that of the cow there were three rectangles of equivalent length.
This equalizing process was represented numerically by incrementing the total mass
of the three animals by their differences (410 + 150 + 130). This allowed the direct
computation of the mass of the cow. Figure 8 shows how children worked part by
part from the model until they found the value of the unknown unit.
The most common form of solution was of the type shown on the left in Figure
8. Here because no algebraic equation was constructed to represent the informa-
tion captured by the models, the unknown to be found was not made explicit. These
children, however, knew they had to calculate the value of the unit, which was
knowledge they held mentally. The children then proceeded to solve for the
unknown value by undoing, step by step, the operations suggested by the relation-
ships of the rectangles to the left of the brace. A rectangle with a specific rectangle
appended would suggest subtraction of that specified value from the given total
while three identical rectangles would suggest division of the total on the right. The
work of these children suggests that they exhibited good metacognition skills
because they knew what to do next after each operation and knew when they had
the answer. When asked to comment on these solutions, the teachers explained that
the solution on the left was the easier of the two and hence they preferred to teach
this to solve similar problems. The teachers explained that children who gave the
solution in the right may have received extra tutoring from either their parents or
private tutors.
Algebraic word problems involving rational numbers. In the Water Problem,
concepts of fractions are involved. With such problems, children are taught to use
the inverse of operations, thus converting the relationship involving unit fractions
into a relationship involving whole numbers. Which unknown is selected as the
generator depends on which relationship is under consideration. We have selected
three children’s solutions to demonstrate how the model method can be used to solve
the Water Problem.
A total of 57% of the children who answered this question correctly drew models
with the exact number of rectangles in which the number of rectangles for each
container was ascertained by evaluating the relationship between the volume of
water in A, B, and C. For example, the solution of Student 126 (see top, left of Figure
9) suggested that the child’s thinking alternated between the volume of B and A,
then B and C. Twelve is a common multiple of 3 and 4. Hence, if B has 12 rectan-
gles, then A has 4 rectangles and C, 3. It is possible that the numerical information
could have been refined as the model was being drawn—alternating between the
model drawing and the numerical relationships. The volume of container B could
be found once the value of one unit was ascertained.
A comparison of Student 126’s solution with that of Student 147 (see bottom,
left of Figure 9) suggests that the latter might have processed the information as
follows: (a) If A is the base, then there should be three rectangles for B; (b) If C is
one quarter of B, then using its inverse, the volume of B is four times C; (c) How
many parts must C be so that B is 4 times C? (d) Look for the lowest common
Swee Fong Ng and Kerry Lee 303

Figure 9. Solutions to the Water Problem: Student 126 (at top, left), Student 105 (at top,
right), and Student 147 (at bottom, left) and global relationships among 3, 4, and 12 (at bottom,
right).

multiple of 3 and 4. If it is 12, then B has 12 rectangles and A has 4; (e) The dotted
lines are then worked into the original rectangles; and (f) Find the volume for B.
Although the solutions of Students 126 and 147 appear similar, the process used
by Student 147 suggests that perhaps he or she was thinking more globally than
Student 126, who was focusing more on the details. Perhaps Student 147 saw the
structure in the bottom, right in Figure 9, which gave rise to the concise solution.
A total of 23% of the children with the correct answer gave this particular solution.
Although the model drawings of Students 105 and 126 were accompanied by the
corresponding arithmetic expressions, that of Student 147 showed only two arith-
metic equations, one showing how the value of one subunit was found and the final
equation, 171 ÷ 19 = 9, showing how the volume of container B was found.
Comparing these three solutions suggests that Student 147 carried out many
processes mentally, which meant that many meaningful and conceptual thought
processes are hidden.
Of all the different solutions, Student 105’s solution (see top, right in Figure 9)
was the clearest, and 20% of those with correct answers used this approach. Student
105’s structural phase comprised two representations—symbolic and model. He
used the symbolic representation to help him draw the model. His solution showed
that container B was taken as a whole and hence was selected as the generator. If
B is one whole, then A is one third of B (working backward) and C is one quarter
of B (using the given information). This mode of thinking allowed the construc-
tion of this mathematical sentence 1 + 1/4 + 1/3 = 12/12 + 3/12 + 4/12 = 19/12. Once
304 Model Method and Algebraic Word Problems

the symbolic representation was constructed, drawing the model was not difficult.
With the aid of the model he found the value of one unit (171 ÷ 19) and then the
volume of container B (9 × 12).
The teachers explained that without the model drawing, children would not be
able to solve for the volume of B because they had not been taught how to solve
the equation 19 units/12 = 171. They had yet to learn how to transform equations
and to divide whole numbers by a rational number. The model method provided
children with a way to work with whole numbers rather than fractions and this
simplified the solution process because it now involved dividing 171 by 19 and then
multiplying the result by 12. This meant that children were able to circumvent the
need to evaluate the expression 171 × 12/19, a skill taught only in secondary school
mathematics (CPDD, 2000b).
Algebraic word problems involving proportional reasoning. The Book Problem
involves two unknowns and is the most challenging of the questions under consid-
eration because two levels of relationships are presented in it. First, there is a direct
proportional relationship involving the number of mathematics books and science
books: For every 1 mathematics book, there are 4 science books. Hence, the model
shows the relationship between the number of mathematics books and the number
of science books. Second, the costs of the books, and not the number of books, are
represented by each rectangle. Therefore, rectangles of proportionate lengths should
be used to represent the cost of the mathematics and science books. With the
exception of the Book Problem, all the rectangles for models drawn for the Furniture,
Animal, and Water Problems are of the same size because each rectangle represents
a unit that is the unknown that needs to be determined.
In a standard algebra text, the Book Problem can be solved by construction and
solution of two simultaneous linear equations. With the model method, however,
children can dispense with this necessity. The solution presented at left in Figure
10 shows that children used the individual rectangles of the same size to reflect the
costs of the mathematics and science books. Each set of 1 mathematics book and
4 science books has a total cost of $44. The problem is solved by asking how many
such sets of $44 there are in $528. There are 12 such sets, meaning that there are
12 mathematics books and 48 science books. Perhaps these children had a mental
picture (see the diagram at right in Figure 10) of the question “How many $44 are
there in the total of $528?”
Although many of the children’s answers were correct, their models were not.
All children drew erroneous models in which the rectangles representing the math-
ematics books were the same size as those used to represent science books. This
suggests that children were using the model method as an algorithm to solve such
problems. Teachers confirmed that for such problems, it was not necessary to
show the relative lengths of the rectangles because the objective of the model
drawing was to help children make the information overt. For such problems, the
meaning of the rectangles need not be known so long as the method consistently
gives the correct answer.
Swee Fong Ng and Kerry Lee 305

Figure 10. A common solution to the Book Problem (at left) and grouping method held
mentally (at right).

How did children fail to use the model method successfully? Not all children
acquiesced to our request to use the model method to solve the word problems. Some
attempted to use other heuristics, but they were not always successful, even when
using the method of their choice. Although some children who participated in the
pilot studies used the model method to solve the Stamp Problem, none of the chil-
dren in the Study 2 did. Instead, guess and check and systematic listing were the
two heuristics of choice. The rate of success was low because they either failed to
satisfy the condition of a total of 62 stamps or the total amount spent.
The visual nature of model drawing enabled us to probe which aspects of the trans-
lation and representation processes were problematic for children. We compared
and contrasted model drawings of correct solutions with those that resulted in
wrong answers. What differences exist between these model drawings? The model
drawings that resulted in correct solutions accurately represented the information
presented in that word problem. Teachers explained that care is needed to construct
accurate models. To what level of care were they alluding?
Correct model drawings were elaborate in that they indicated clearly and precisely
every piece of information presented in the text that was necessary for the solution
of the word problem: the selected generator, the difference rectangles, and what was
to be evaluated. In contrast, partly correct model drawing representations, which
formed about 30% of the errors, showed how a lack of attention to a single detail
or a misrepresentation of a single piece of information could result in an incorrect
solution. Furthermore, construction of correct model drawings was underpinned by
sound conceptual knowledge.
The discussion in this section is divided into four categories according to the type
of error made: (a) partly correct models with essential information missing or
306 Model Method and Algebraic Word Problems

misrepresented, (b) changing generators midway, (c) correct model drawing and
correct arithmetic expressions but failure to keep the goal in mind, and (d) lack of
conceptual knowledge necessary for the solution of the five problems.
Partly correct models with essential information missing or misrepresented.
Errors of this nature were common among algebraic word problems. Correct model
drawings identified and represented information accurately, but partly correct
model drawings often misrepresented a detail or omitted it entirely. Partly correct
model drawings appeared in about 7% of the Furniture Problem solutions and 11%
of the Animal Problem solutions. The difference between a correct model drawing
and one with missing or misrepresented information is very subtle. At a superfi-
cial level, the partly correct model drawings appear very similar to a correct repre-
sentation, but upon closer inspection, missing details or misrepresented informa-
tion can be noted. In the case of correct solutions for the Furniture Problem, since
the cost of each item is based on the cost of the iron, each representation clearly
shows the rectangle for the cost of the iron and the difference rectangle. Partly correct
model drawings with missing information did not reflect this level of detail. The
item that was furthest from the chosen generator carried fewer details. Although
the representation of the chair was accurate, the cost of the table did not show how
it was related to the generator. Because of the lack of detail at each subsequent level,
the resulting arithmetic equation was incorrect.
In the Furniture Problem, the homogeneous relationship more than relates the
remaining two items to the cost of the iron. In the Animal Problem, however, the
nonhomogeneous relationships more than and less than relate the mass of the cow
with the mass of the dog, and the mass of the goat with that of the cow, respectively.
This difference between the two problems could account for the lower accuracy rate
on the Animal Problem than on the Furniture Problem. Accurate model drawings
clearly showed the generator and the difference rectangle. This was not the case
for partly correct model drawings. Although the overall representation for 13% of
the Animal Problem solutions was correct, the arithmetic equations were wrong.
The model drawing at top, left of Figure 11 shows a single rectangle was used
to represent the mass of each animal. A dotted rectangle was used to represent the
difference in the masses of the related animals. This model drawing could have
resulted in a correct solution if the child had equalized the length of the rectangles
for the mass of the dog and the mass of the goat to that of the cow, effectively treating
the mass of the cow as the generator. The first arithmetic equation, 150 – 130 = 20,
suggests that the child was trying to find the difference in mass between the goat
and the dog, indicating an attempt to use the mass of the dog as the generator.
Perhaps this child forgot which was the selected generator, and thus the subsequent
set of arithmetic equations was incorrect (see Figure 11, top left). Another 11% of
the solutions clearly identified the mass of the dog as the generator for the mass of
the cow, but because there was no clear distinction between the representations for
less than and more than, the goat is then 130 kg heavier than the dog. Hence, the
set of arithmetic equations was incorrect (see Figure 11, top right).
Swee Fong Ng and Kerry Lee 307

What is the mass of the cow? What is the mass of the cow?

Figure 11. Examples of how representation of a problem is not an all-or-nothing process.

Note. For the Animal Problem, a correct model that would have resulted in a correct solution if the inten-
tion were to equalize the mass of the animals to that of the cow (at top, left). For the Animal Problem,
a correct model but incorrect set of arithmetic equations (at top, right). For the Furniture Problem, an
incorrect base used for comparison but correct numerical answer based on the model (at bottom, left).
For the Book Problem, a correct overall representation but incorrect understanding of the role of the
rectangle (at bottom, right).

Changing generators midway. This category of error was common to algebraic

and arithmetic word problems. In the Enrollment Problem, instead of using the
enrollment of Dunearn to construct the representation for Excellent Primary, 25%
of the solutions based the enrollment of Excellent Primary on that of Sunshine
Primary. In the Furniture Problem, 26% of the solutions made the error of comparing
the cost of the table with the cost of the chair rather than with the cost of the iron
(see Figure 11, bottom left). Although the SP translation was correct, the answer
was incorrect because of an erroneous TS translation.
Correct model drawing, correct arithmetic expressions but failure to keep the goal
in mind. Such errors were common to the arithmetic and algebraic word problems
but comprised less than 3% of the errors. Children who made such errors constructed
a correct model drawing and its related arithmetic equations but failed to answer
the question as stated. With the Enrollment Problem, children found the enrollment
for each school but did not continue to find the total enrollment of all three schools.
For the Furniture Problem, instead of proceeding to find the cost of the chair, these
children stopped after finding the cost of the iron, which was used as the generator
308 Model Method and Algebraic Word Problems

for the problem. Children who made such errors may not have made the extra effort
to check whether they had answered the question. More important, model draw-
ings of correct solutions were more likely to indicate the unknown to be found. For
the Enrollment Problem, more correct solutions than incorrect solutions carried a
vertical brace and the related question mark to indicate that the total was required.
For the Furniture and Animal Problems, a question mark in the selected rectangle
identified the unknown to be found. Such details were missing from model draw-
ings with incomplete solutions.
Partly correct model drawings may be the result of poor metacognitive practices.
If children check their drawings against the text-based information, they may
realize that the model drawing is incomplete or partly correct, or they may realize
that they have chosen the wrong generator or failed to answer the question. Correct
solutions may be evidence of good practices in problem solving, in which the infor-
mation captured in the model is checked against the text-based information.
Lack of conceptual knowledge necessary for the solution of the five problems.
For the Water Problem, comparison of correct solutions against those that were
incorrect highlights how correct solutions offer evidence of sound conceptual
understanding of part-whole relationships and fractions. They also illustrate an inte-
grated and well-organized knowledge base of number facts including factors and
multiples of numbers as well as skill in operating with fractions.
The Water Problem shows how about 34% of the children were challenged by
the concept of fraction. These children constructed the correct model drawings for
the volume of containers A and B, but the representation of the relationship between
B and C was faulty. This could be because these children were more comfortable
working with whole numbers, and the relationship between container A and B was
stated in terms of whole numbers, hence the representation was relatively simple.
In contrast, the relationship between container B and C was expressed in terms of
a fraction. A concept of fraction is necessary to transform the relationships to ones
in which operations with whole numbers sufficed. First, the children had to be able
to translate the relationship “Container C has 1/4 as much water as container B” to
its equivalent “Container B has four times as much water as C.” Second, they had
to determine the least common multiple of 4 and 3. If they were able to do this, the
volume of C could be represented. Erroneous model representations of the volume
for container C showed that children were unable to transform the original infor-
mation to its equivalent. They represented the volume of container C additively by
attaching a rectangle representing the value 1/4 to three rectangles representing the
volume of container B. They may have misinterpreted the volume of container C
as 1/4 more than the volume of B. If these children had a sound knowledge of frac-
tions, multiplicative reasoning, and knowledge of factors and multiples of 3 and 4,
they might have been able to solve the Water Problem.
Another example of the importance of conceptual knowledge is that rectangles
play different roles. In the first four problems, the rectangles could be used to repre-
sent either specific or unknown quantities. In the Book Problem, the children
needed to rethink the role of the rectangles because they represent the cost of the
Swee Fong Ng and Kerry Lee 309

books, even though the unknown is not located within any of the rectangles.
Instead, the unknown is the question “How many $44 are there in $528?” and this
has to be abstracted from the representation. Approximately 21% of the children
answered the Book Problem correctly, and another 32% of the children constructed
a correct overall representation of the problem—one rectangle for the mathematics
books and four rectangles for the science books—but the SP translation was wrong.
The SP translation showed how these children continued to treat these rectangles
as unknown units and found an answer by dividing the total cost by the total
number of rectangles (see Figure 11, bottom right). Perhaps if these children had
a clearer understanding of the different roles of the rectangles, they might have
solved the problem correctly.
Children who used the model method also successfully demonstrated a sound
conceptual knowledge of comparative phrases such as more than, less than, and as
many as, since their model drawings showed that there should be a base for compar-
ison. In contrast, about 7% of the children who did not have the conceptual under-
standing of such relationships drew models that were devoid of any base for
comparison. In the Enrollment Problem, the rectangles represented the differences
in enrollment between schools; the children may have treated the difference as the
enrollment for each school. Because no base was used in the comparison, the rela-
tionships more than and less than in the Furniture and Animal Problems were irrel-
evant, and these algebraic word problems were converted into arithmetic word prob-
lems. This practice of ignoring the base for comparison may be the result of a local
practice in which it is quite common to hear children say “I have more” without
referring to a base, and more is treated as a noun. Therefore, they may have focused
only on “Sunshine Primary School has 89 more pupils” without attending to
Dunearn Primary, the base for comparison. These children, who simply used
numbers as they were presented in the problems rather than as comparisons, need
more support solving arithmetic problems. Work with algebraic word problems
should be deferred until they are confident with arithmetic word problems.

Discussion

Study 2 shows that the model method can be used to solve some, but not all, alge-
braic word problems. Algebraic word problems involving whole numbers and those
that can be solved with the construction and solution of a system of linear equa-
tions in one unknown are more amenable to the model method than those that
require the construction and solution of a system of linear equations in two
unknowns. The children’s solutions to the Book Problem show how they have
adapted the model method to solve algebraic word problems involving two
unknowns.
This study also showed that although most EM1 children were able to use the
model method to solve arithmetic and algebraic word problems, many EM2 chil-
dren could not. We are not sure why this is the case. At the end of Primary 4, the
children completed a streaming examination. Based on their proficiency in their
310 Model Method and Algebraic Word Problems

mother tongue language (Chinese, Malay, or Tamil), English, and mathematics,

these children were streamed into either the EM1 or EM2 class. Could it be that
the streaming examination effectively identified those who were more mathemat-
ically inclined from those who were less so? Or could it be that the EM2 children
faced language difficulties? It could be that the EM2 children had more difficul-
ties than their EM1 peers in understanding the word problems. This would suggest
that they had difficulties with the first phase of the three-phase problem-solving
process. That about a quarter of the children were unsuccessful with any of the five
problems suggests that they do need support with arithmetic word problems and
that algebraic word problems should be deferred.
Study 2 showed the challenges that children face when using model representa-
tions to solve word problems. Those who were successful with the model method
drew detailed models and constructed accurate arithmetic equations to represent the
information represented by the model drawing. Such practices are consistent with
the behaviors of skilled physics problem solvers (Larkin et al., 1980). In addition,
the children’s success may be reflective of good monitoring practices in which they
countercheck the representation of the model against the text before proceeding to
construct the corresponding set of arithmetic equations. Alternatively, the moni-
toring activity could be the penultimate part of the solution process in which they
checked the arithmetic equations against the model method and the text before
working out the final answer. Because children were not interviewed, we can only
speculate on the sequence of events. Further research is needed to confirm these
speculations.
More important, this study shows that using the model method to solve word prob-
lems is not an all-or-nothing process. Consistent with the research literature (De
Corte et al., 1985), children’s errors were due more to erroneous representation of
the problem than to computational errors. One third of the partially correct solu-
tions show that children were neither completely incorrect nor completely correct
in their problem-solving activity. Although the overall representation may be
correct, a single error in the representation can result in an incorrect solution. The
error could be a result of misinterpretation of information correctly captured in the
model, misrepresentation of a piece of information, or changing the generator
midway through the solution.
Successful representation of word problems requires an integrated and well-orga-
nized knowledge base of number facts, conceptual understanding of part-whole rela-
tionships and fraction, multiplicative reasoning, and knowledge of the four opera-
tions (e.g., subtraction is the inverse of addition; division is the inverse of
multiplication). This study shows that conceptual understanding of fraction chal-
lenged half the children. Albeit fewer in numbers (about 7%), there were children
who drew models that illustrated their lack of understanding that comparisons in-
volving the four operations must be made against a base. Without this conceptual
understanding, children may continue to draw models that are incorrect.
Swee Fong Ng and Kerry Lee 311

IMPLICATIONS FOR TEACHING

The visual nature of the model representation means that it is possible for teachers
and children to point to specific errors as they appear in the incorrect model draw-
ings. How can teachers help children who drew partially correct models understand
the source of their errors? Teachers could offer children a set of correct solutions
and ask children to compare and contrast these against those that are partially correct.
Children could be asked to discuss from where those errors came and how they could
improve upon their model drawings.
Teachers could also use the variations among correct solutions to a particular
problem as a stimulus to provoke discussion among children to seek alternate
ways to solve the same problem. For example, in the Animal Problem teachers could
ask children to compare and contrast solutions in which the mass of the dog was a
generator and the mass of the cow was a generator. Children who used the mass of
the dog as a generator could be encouraged to see that the mass of the cow might
be used as a generator instead, because these different model solutions are within
their zone of proximal development (Vygotsky, 1978). Similarly, for the Enrollment
Problem, children who found the total enrollment by summing the individual
enrollments of the three schools could be encouraged to shift to the alternative solu-
tion. Offering children the various Water Problem solutions for discussion may help
those who are challenged by the procedural aspects of model drawing to articulate
their thoughts. Furthermore, children with errors in the final phase of problem
solving could be encouraged to check their arithmetic representations against the
model drawing.

CONCLUSIONS

This study contributes to a corpus of research in which the teaching of repre-

sentation skills supports higher ability children in their work with word prob-
lems—arithmetic as well as algebraic. More important, this study offers avenues
for average ability children to be supported in their word-problem-solving activity.
The partially correct model drawings show that this kind of representation is not
an all-or-nothing process. Average ability children’s solution of word problems
involving whole numbers could be improved if they learn to exercise more care in
the construction of related models. Furthermore, by deepening their conceptual
knowledge of fraction and its related skills, they may extend their competency to
word problems that involve fractions. Their solution of word problems involving
two unknowns also could be enhanced if they expanded their appreciation of the
different roles that the rectangles can have in model drawings.
This study also shows that although the children may treat the model method as
an algorithm, it is not an algorithm learned by rote that is intended to replace other
rote algorithms for solving word problems. Instead, it is a problem-solving heuristic
that requires children to reflect on how they could accurately represent the infor-
mation presented in word problems, first in terms of a drawing and then as a series
312 Model Method and Algebraic Word Problems

of arithmetic equations. This art of representation first has to be taught, but it is then
the children’s responsibility how they choose to use this heuristic effectively.

REFERENCES

Bednarz, N., & Janvier, B. (1996). Emergence and development of algebra as a problem-solving tool:
Continuities and discontinuities with arithmetic. In N. Bednarz, C. Kieran, & L. Lee. (Eds.),
Approaches to algebra: Perspectives for research and teaching (pp. 115–136). Dordrecht, The
Netherlands: Kluwer.
Behr, M., Erlwanger, S., & Nichols, E. (1980). How children view the equals sign. Mathematics
Teaching, 92, 13–15.
Booth, L. (1988). Children’s difficulties in beginning algebra. In A. F. Coxford & A. P. Shulte (Eds.),
The ideas of algebra, K–12: 1988 yearbook (pp. 20–32). Reston, VA: National Council of Teachers
of Mathematics.
Briars, D. J., & Larkin, J. H. (1984). An integrated model of skill in solving elementary word problems.
Cognition and Instruction, 1, 245–296.
Cai, J., Lew, H. C., Morris, A., Moyer, J. C., Ng, S. F., & Schmittau, J. (2005). The development of
students’ algebraic thinking in earlier grades: A cross-cultural comparative perspective. Zentralblatt
für Didaktik der Mathematik, 37, 5–15.
Carpenter, T. P., Corbitt, M. K., Kepner, H. S., Jr., Lindquist, M. M., & Reys, R. E. (1981). Results from
the Second Mathematics Assessment of the National Assessment of Educational Progress. Reston,
VA: National Council of Teachers of Mathematics.
Carpenter, T. P., & Moser, J. M. (1982). The development of addition and subtraction problem-solving
skills. In T. P. Carpenter, J. M. Moser, & T. A. Romberg (Eds.), Addition and subtraction: A cogni-
tive perspective (pp. 9–24). Hillsdale, NJ: Erlbaum.
Clement, J. (1982). Algebra word problem solutions: Thought processes underlying a common miscon-
ception. Journal for Research in Mathematics Education, 13, 16–30.
Collars, C., Koay, P. L., Lee, N. H., Ong, B. L., & Tan, T. S. (2008). Shaping maths—Coursebook 5A.
Singapore: Marshall Cavendish.
Collars, C., Koay, P. L., Lee, N. H., & Tan, T. S. (2003). Shaping maths—Coursebook 3A. Singapore:
Federal Publications.
Collars, C., Koay, P. L., Lee, N. H., & Tan, T. S. (2007). Shaping maths—Coursebook 3A. Singapore:
Marshall Cavendish.
Curriculum Planning and Development Division (2000a). 2001 Mathematics syllabus: Primary.
Singapore: Ministry of Education.
Curriculum Planning and Development Division (2000b). 2001 Mathematics syllabus: Lower secondary.
Singapore: Ministry of Education.
Curriculum Planning Division (1990). Mathematics syllabus: Primary. Singapore: Ministry of Education.
Davydov, V. V. (1962). An experiment in introducing elements of algebra in elementary school.
Sovetskaia Pedagogika, 8, 27–37.
Davydov, V. V., & Steffe, L. P. (1991). Soviet studies in mathematical education, Vol. 6: Psychological
abilities of primary school children in learning mathematics. Reston, VA: National Council of
Teachers of Mathematics.
De Corte, E., Verschaffel, L., & De Win, L. (1985). Influence of rewording verbal problems on chil-
dren’s problem representations and solutions. Journal of Educational Psychology, 77, 460–470.
Freudenthal, H. (1974). Soviet research on teaching algebra at the lower grades of the elementary school.
Educational Studies in Mathematics, 5, 391–412.
Herscovics, N. (1989). Cognitive obstacles encountered in the learning of algebra. In S. Wagner & C.
Kieran (Eds.), Research issues in the learning and teaching of algebra: Vol. 4, Research agenda for
mathematics education (pp. 60–86). Reston, VA: National Council of Teachers of Mathematics.
Kho, T. H. (1987). Mathematical models for solving arithmetic problems. In Proceedings of the Fourth
Southeast Asian Conference on Mathematical Education (ICMI-SEAMS). Mathematical Education
in the 1990’s (Vol. 4, pp. 345–351). Singapore: Institute of Education.
 Swee Fong Ng and Kerry Lee 313

Kieran, C. (1981). Concepts associated with the equality symbol. Educational Studies in Mathematics,
12, 317–326.
Kintsch, W., & Greeno, J. G. (1985). Understanding and solving word arithmetic problems. Psychological
Review, 92, 109–129.
Küchemann, D. (1981). Algebra. In K. M. Hart (Ed.), Children’s understanding of mathematics: 11–16
(pp. 102–119). London: John Murray.
Larkin, J., McDermott, J., Simon, D. P., & Simon, H. A. (1980). Expert and novice performance in solving
physics problems. Science, 208(4450), 1335–1342.
Lee, K., Ng, E. L., & Ng, S. F. (in press). The contributions of working memory and executive func-
tioning to problem representation and solution generation in algebraic word problems. Journal of
Educational Psychology.
Lee, K., Ng, S. F., Ng, E. L., & Lim, Z. Y. (2004). Working memory and literacy as predictors of perfor-
mance on algebraic word problems. Journal of Experimental Child Psychology, 89, 140–158.
Lewis, A. B. (1989). Training students to represent arithmetic word problems. Journal of Educational
Psychology, 81, 521–531.
MacGregor, M. E. (1991). Making sense of algebra: Cognitive processes influencing comprehension.
Geelong, Australia: Deakin University.
Martin, S. A., & Bassok, M. (2005). Effects of semantic cues on mathematical modeling: Evidence from
word-problem solving and equation construction tasks. Memory & Cognition, 33, 471–478.
Ministry of Education. (2005). Singapore education milestones 2004–2005. Retrieved March 10, 2009,
from [Link]
Riley, M. S., Greeno, J. G., & Heller, J. I. (1983). Development of children’s problem-solving ability.
In H. P. Ginsberg (Ed.), The development of mathematical thinking (pp. 153–196). New York:
Academic Press.
Schoenfeld, A., & Arcavi, A. (1988). On the meaning of variable. Mathematics Teacher, 81, 420–427.
Stacey, K., & MacGregor, M. (1999). Ideas about symbolism that students bring to algebra. In B. Moses
(Ed.), Algebraic thinking (pp. 308–312). Reston, VA: National Council of Teachers of Mathematics.
Stacey, K., & MacGregor, M. (2000). Learning the algebraic method of solving problems. Journal of
Mathematical Behaviour, 18, 149–167.
Steinberg, R. M., Sleeman, D. H., & Ktorza, D. (1991). Algebra students’ knowledge of equivalence
of equations. Journal for Research in Mathematics Education, 22, 112–121.
Usiskin, Z. (1988). Conceptions of school algebra and uses of variables. In A. F. Coxford & A. P. Shulte
(Eds.), The ideas of algebra, K–12: 1988 yearbook (pp. 8–19). Reston, VA: National Council of
Teachers of Mathematics.
Vergnaud, G. (1982). A classification of cognitive tasks and operations of thought involved in addition
and subtraction problems. In T. P. Carpenter, J. M. Moser, & T. A. Romberg (Eds.), Addition and
subtraction: A cognitive perspective (pp. 39–59). Hillsdale, NJ: Erlbaum.
Verschaffel, L., Greer, B., & De Corte, E. (2000). Making sense of word problems. Lisse, The
Netherlands: Swets & Zeitlinger.
Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Cambridge,
MA: Harvard University Press.
Willis, G. B., & Fuson, K. C. (1988). Teaching children to use schematic drawings to solve addition
and subtraction word problems. Journal of Educational Psychology, 80, 192–201.

Authors
Swee Fong Ng, Mathematics and Mathematics Education Academic Group, National Institute of
Education, Nanyang Technological University, 1 Nanyang Walk, Singapore 637 616; [Link]
@[Link]
Kerry Lee, Psychological Studies Academic Group, National Institute of Education, Nanyang
Technological University, 1 Nanyang Walk, Singapore 637 616; [Link]@[Link]

View publication stats