Journal of Computer Assisted Learning (2003) 19, 46-55

The effect of metacognitive training on

mathematical word-problem solving
S.K. Teong
National Institute of Education, Singapore

Abstract This study demonstrates how explicit metacognitive training

influences the mathematical word-problem solving of forty 11–12-year-
old low achievers in a cognitive-apprenticeship, computer-based
environment. Results from the experimental and case study designs
revealed that experimental students outperformed control students on
ability to solve word-problems on their individual written measures;
experimental students developed the ability to ascertain when to make
metacognitive decisions, and elicit better regulated metacognitive
decisions than control students; knowing when and how to use
metacognitive strategies is an important determinant to successful word-
problem solving; and the cognitive-apprenticeship-computer-based
environment appears to amplify low achievers’ metacognitive and
cognitive behaviours during word-problem solving.

Keywords: Cognition; Collaboration; control group; Low achievers;

Metacognition; Primary; Two-phase design; Word-problem solving

Introduction
Evidence from Schoenfeld (1985) suggests that compared with an ‘expert’ problem
solver, ‘novices’ lack essential metacognitive monitoring, assessing and decision
making skills. According to Schoenfeld, these are essential elements that determine
one’s success or failure in problem solving, and there is a need for students to be
explicitly trained to monitor their cognitive processes during problem solving.
Cardelle-Elawar (1995) trained low achievers with a metacognitive intervention
strategy and observed its influence on their mathematics performance. The
researcher reported that low achievers trained in learning to monitor and control their
own cognitive processes for solving mathematics problems do better than untrained
students. However, research that focuses on metacognition in problem solving are
usually carried out in a non-computer environment. Over the past decade, attempts
have been made to employ metacognitive training within computer environments
(Mevarech & Kapa, 1996; Kramarski & Ritkof, 2002). For example, Kramarski &
Ritkof (2002) used a teaching/learning strategy called IMPROVE when students
were learning about graphs in EXCEL environment. IMPROVE is the acronym for
the teaching/learning stages that require students to focus on their metacognition.
The researchers reported positively that students who were exposed to metacognitive

Accepted 28 June 2002

Correspondence: Su Kwang Teong, National Institute of Education, Science Building, 1 Nanyang Walk,
Singapore 637616 Email: skteong@nie.edu.sg

46  2003 Blackwell Publishing Ltd

 Mathematical word-problem solving 47
treatment with IMPROVE tended to construct graphs better and were able to reflect
better on their learning compared to those who were not exposed to such treatment.
A cognitive-apprenticeship model of instruction, an approach to instruction, is
also identified by Schoenfeld (1992) and other researchers like Verschaffel et al.
(1999) that may promote students’ ability to regulate during problem solving. These
studies were again conducted in traditional classrooms where the computer was not
used. WordMath, a computer software designed on the instructional approach of
cognitive apprenticeship methodology, was used as a tool for students to solve word-
problems that are particular to the Singapore mathematics curriculum (Looi & Tan
(1998). The software was tested on thirty-six 11–12-year-olds from two Singapore
schools and the researchers concluded that ‘WordMath has a good potential of
supporting thoughtful practice in students by enabling them to engage actively and
reflectively in problem solving and by providing informative feedback to students
about their performance’ (Looi & Tan, 1998; p. 353). However, Looi & Tan (1998)
did not explicitly address how the students’ cognitive skills in the cognitive
apprenticeship mode of instruction aided them in their word-problem solving. The
present study addresses the extent metacognitive training can promote metacognitive
behaviours of low achievers solving word-problems in WordMath computer
environment. The specific research questions of this study were:
• Do experimental low achievers outperform control low achievers on ability to
solve word-problems on their individual written measure?
• Can experimental low achievers develop the ability to ascertain when to make
metacognitive decisions, and elicit better regulated metacognitive decisions than
control lower achievers? If can, do they require some time to internalise the
positive benefits of metacognitive training.
• Does the occurrence of metacognitive behaviours on its own ensure successful
word-problem solving?
• What is the role of WordMath in low achievers’ cognitive activity?
These questions were answered for problems involving whole numbers and fractions.

Methodology
Participants
Forty 11 to 12-year-old low achievers were involved in this intensive study over a
period of eight weeks. They were chosen according to their year-end mathematics
examination results. All low achievers had scores between 50% and 70% and this
was in accordance with the categories of scores in the national examination grading
system∗.
CRIME strategy
CRIME is an acronym for the word-problem solving stages: Careful Reading; Recall

∗
Singapore primary education consists of a foundation stage of four years (7 to 10 years old) and an
orientation stage of two years (11 and 12 years old). At the end of orientation stage at Primary Six,
students are required to sit a national examination, the Primary School Leaving Examination (PSLE). The
scores in this examination determine the type of programme (Special, Express, Normal(A) and
Normal(T)) students follow when they enter secondary school. The students are given a grade for each
subject in their PSLE result. The grades and corresponding scores are defined as follows:
A* 91%* and above: A 75% to 90%: B 60% to 74%; C 50% to 59%: D 35% to 49%: E 20% to 34%;
U Below 20%.
The present study focused on students who had scored at least a grade C in their year-end mathematics
examination.

 2003 Blackwell Publishing Ltd, Journal of Computer Assisted Learning, 19, 46-55
48 S.K. Teong
Possible Strategies; Implement Possible Strategies; Monitor, and Evaluation which
was developed by the author. Its aim is to develop low achievers’ ability to monitor
and evaluate their actions during word-problem solving using WordMath (for details
see http://math.nie.edu.sg/skteong/WordMath.doc). At each stage of the word-
problem solving, there are questions to direct the students to regulate and monitor
their solution. In the present study, students solved word-problems in the WordMath
environment with the metacognitive strategy, CRIME. It was piloted with two
students and there were indications that CRIME offered an effective metacognitive
strategy to make students more aware of their cognitive processes during word-
problem solving. (see Appendix A for further description)

Design
The study adopts a two-phase design, combining an experimental design and a case
study design. The experimental design used the analysis of students’ mathematical
achievement test data, whereas the case study design used the analysis of
collaborative think-aloud protocol data.

Instruments
The experimental design was based on word-problem tests, where the students’
performance in completing word-problem items (see Appendix B) were used to test
the hypothesis generated in the experimental design. The word-problems were the
kinds of tasks students were expected to ‘master’ in school and on which their
mathematical performance was systematically monitored and evaluated. They were
also designed to reflect the problems used in the School Leaving Examination.
For the case study, students solved word-problems in WordMath and their think
aloud protocols were video-recorded and analysed to elicit students’ word-problem
solving activity. Criteria for choice of word-problems were similar to those for the
mathematics achievement test items.

Procedure
The experimental design. Students were assigned to two groups: one group had
explicit metacognitive training with CRIME before solving word-problems in
WordMath (Experimental or E); and one group solved word-problems with
WordMath without CRIME (Control or C). In the experimental design, all groups
took a pre-test consisting of 10 word-problems on numbers and fractions (see
Appendix B). Written post-test 1 was immediately administered after intervention.
Six weeks after post-test 1, a final written delayed post-test 1 was administered.
The case study design. Two pairs of students were selected from each group, based
on their ability to work together. The four pairs of students underwent two extra
training sessions where the students solved four word-problems for each session.
Their word-problem solving behaviours were video-recorded but not analysed. The
purpose of these extra sessions was to enable students to feel comfortable working
collaboratively in front of the video camera. After these sessions, the pairs of
students’ word-problem solving think-aloud protocols of four word-problems (post-
test 2) were video-recorded. Six weeks after post-test 2, due to logistics and time
constraints, only two pairs of experimental students were granted permission to
participate in the case study. Their word-problem solving think-aloud protocols of
four parallel word-problems (delayed post-test 2) were video-recorded and analysed.

 2003 Blackwell Publishing Ltd, Journal of Computer Assisted Learning, 19, 46-55
 Mathematical word-problem solving 49
Intervention
Both experimental and control groups were trained separately, but each group had
four 60-minute training sessions on the use of WordMath with/without metacognitive
training over a period of two weeks. All students worked collaboratively in pairs
throughout the study. During the first training session, experimental students were
taught how to use WordMath and encouraged to think aloud. During the second
training session, experimental students were given direct explanation, explicit
demonstration, and ‘scaffolded’ (Bruner, 1985) practice in using CRIME while
solving word-problems. Students had the CRIME acronym card at all times and they
were encouraged to refer to CRIME during word-problem solving. During the last
two training sessions, experimental students practised using CRIME to solve word-
problems in WordMath. In contrast, control students received no training, modelling
or instructions in using CRIME. They discussed concepts only in the context of
solving WordMath word-problems.

Think aloud protocol framework

The framework to analyse low achievers’ think aloud protocols was designed to
differentiate explicitly between cognitive and metacognitive problem solving
behaviours observed within different episodes of problem solving in small group
settings (Artzt & Armour-Thomas, 1992). The development of this framework was
based on earlier work (Schoenfeld, 1985), but was amended. One of the amendments
was categorising the six distinctive types of problem solving behaviour as cognitive
or metacognitive.
. . . metacognitive behaviours could be exhibited by statements made about the
problem or about the problem solving process while cognitive behaviours could be
exhibited by verbal or nonverbal actions that indicated actual processing of
information. (Artzt & Armour-Thomas, 1992, p. 141)
This means that Artzt and Armour-Thomas’ framework had 11 episodes to partition
group think aloud protocols. They were: read (cognitive), understand
(metacognitive), analyse (metacognitive), explore (cognitive or metacognitive), plan
(metacognitive), implement (cognitive or metacognitive), verify (cognitive or
metacognitive), and watch and listen. The author modified their framework for
highlighting major strategic decisions made by pairs of students. The first
modification involved recategorising the descriptors for the categories
‘understanding the problem’ and ‘analysing the problem’ to ‘analysing the word-
problem’. This is because the author drew on Schoenfeld’s (1985) observation —
that was, in analysing a problem, ‘an attempt is made to fully understand the
problem’. The second modification involved eliminating the category ‘watch and
listen’ which categorised students who were not involved in group problem solving
discourse. In the present study, students were paired to elicit think aloud protocol
data during word-problem solving. It appeared that continual verbal interactions
were evident between students during word-problem solving. Each student was also
involved in either typing or controlling the mouse. Hence, the category ‘watch and
listen’ was not appropriate in pair protocols in the WordMath environment.

Analysis and results

Experimental design data
For each group, Table 1 shows the mean and standard deviation of the students’

 2003 Blackwell Publishing Ltd, Journal of Computer Assisted Learning, 19, 46-55
50 S.K. Teong
scores on the three tests. As the groups were different in the pre-test scores, this table
also shows the adjusted data for the post-test and delayed post-test scores where the
pre-test scores were used as the covariate. A repeated measures analysis of
covariance (ANCOVA) was carried out on the post-test and delayed post-test with
corresponding pre-test scores again used as a covariant. The repeated measures
analysis of covariance indicated significant differences between experimental and
control groups (F1,37 = 9.008 , p = 0.005).
Table 1. Means and s.d. of low achievers’ mathematical achievements (n = 20 in both groups)

Before Adjustment After Adjustment, evaluated

at covariate of pre-test = 2.30
Metacognitive Pre-test Post-test Delayed post- Post-test Delayed Post-
training scores scores test scores scores test scores
Experimental group Mean 2.00 3.40 4.45 3.63 4.68
s.d. 2.18 2.44 2.46 0.34 0.35.
Control group Mean 2.60 2.75 3.50 2.52 3.28
s.d. 1.54 1.62 1.61 0.34 0.35

Case study data

After the students’ think aloud transcripts were coded into episodes, timeline
representations were drawn to map the flow of the students’ cognitive processes in
word-problem solving against the episodes in the modified Artzt and Armour-
Thomas framework. It was observed that each pair of students had a unique
progression of word-problem solving activity for the four word-problems (Teong,
2000; p. 77). The analysis of the following Q2 word-problem will be used as an
illustration of the students’ word-problem solving.
Joe Ee, Mun Fai & Jing Hao shared 400 marbles amongst themselves. Joe Ee received
28 marbles. Jing Hao received seven times the total number of marbles Joe Ee and
Mun Fai received. How many more marbles did Jing Hao receive than Mun Fai?
Figures 1, 2, 3 and 4 demonstrate the overall structure of the solution analysis of low
achievers S1 & S2, S3 & S4, S5 & S6, and S7 & S8, respectively. The interrater
reliability coefficient for the four word-problems was 0.86 (Teong, 2000; p. 76).
R e a d (C ) B e h a v i o u r s T im e ta k e n b y p a ir
M e ta c o g n it iv e 4 9 0 s
A n a ly s e ( M ) C o g n itiv e 1 7 9 s
T o ta l tim e ta k e n = 6 6 9 s
E x p lo r e ( C ) % o f t im e t a k e n o n m e t a c o g n i t i v e b e h a v io u r s = 7 3 .3 %

E x p lo r e ( M )

P la n (M )

Im p le m e n t ( C )

Im p le m e n t (M )

V e r if y ( C )

V e r if y ( M )

0 4 8 1 0 1 2 1 4

Fig. 1. A timeline representation of S1 and S2 (E) solving Q2 Word-problem

R e a d (C )

A n a ly s e (M )

E x p lo re (C )

E x p lo re (M )

P la n (M )

I m p le m e n t ( C )

Im p le m e n t ( M )
B e h a v io u r s T i m e t a k e n b y p a ir
V e rify (C ) M e t a c o g n it i v e 9 7 8 s
C o g n it i v e 4 2 8 s
V e rify (M ) T o ta l tim e ta k e n = 1 4 0 6 s
% o f tim e ta k e n o n m e ta c o g n itiv e b e h a v io u rs = 6 9 .7 %

8 1 6 2 0 2 8

Fig. 2. A timeline representation of S3 and S4 (E) solving Q2 Word-problem

 2003 Blackwell Publishing Ltd, Journal of Computer Assisted Learning, 19, 46-55
 Mathematical word-problem solving 51
R e a d (C )

A n a ly s e (M )

E x p lo re (C )

E x p lo re (M )

P la n (M )

Im p le m e n t ( C )

I m p le m e n t ( M ) B e h a v io u r s T i m e t a k e n b y p a ir
M e ta c o g n itiv e 3 0 4 s
V e r if y (C ) C o g n it iv e 3 3 5 s
T o ta l tim e t a k e n = 6 3 9 s
V e rif y (M ) % o f tim e ta k e n o n m e ta c o g n itiv e b e h a v io u r s = 4 7 .6 %

0 4 6 8 1 0
T im e (m in u te s )

Fig. 3. A timeline representation of S5 and S6 (C) solving Q2 Word-problem

R e a d (C )

A n a ly s e (M )

E x p lo re (C )

E x p lo re (M )

P la n ( M )

Im p le m e n t ( C )

I m p le m e n t ( M ) B e h a v io u rs T im e ta k e n b y p a ir
M e t a c o g n itiv e 1 33 s
V e rify (C ) C o g n itiv e 2 94
T o ta l tim e t a k e n = 4 2 7 s
V e rify (M ) % o f tim e ta k e n o n m e ta c o g n itiv e b e h a v io u r s = 3 1 .2 %

2
T im e ( m i n u t e s )

Fig. 4. A timeline representation of S7 and S8 (C) solving Q2 Word-problem

Table 2 illustrates the time in seconds and percentage devoted to cognitive and
metacognitive behaviours by low achievers while solving the Q2 word-problem
during post-test 2 and a parallel Q2 word-problem during delayed post-test 2.
Table 2. Time (secs. and %) devoted to cognitive and metacognitive behaviours by pairs
Behaviour Post-test 2 Delayed post-test 2
category S1 & S2* (E) S3 & S4 (E) S5 & S6 (C) S7 & S8* (C) S1 & S2* (E) S3 & S4* (E)
Metacognitive 490 978 304 133 612 740
(73.3) (69.7) (47.6) (31.2) (95.9) (93.1)
Cognitive 179 428 335 294 26 55
(26.8) (30.3) (52.4) (68.8) (4.1) (6.9)
Total 669 1406 639 427 638 795
(100.0) (100.0) (100.0) (100.0) (100.0) (100.0)
* correct solution
The extracts below illustrate portions of S1 & S2’s (E), and S3 & S4’s (E) discourse
while solving the Q2 word-problem with WordMath.
S1 The question asked how many more marbles did Jing Hao receive than Mun Fai.
S2 So, we have to find Mun Fai.
S1 Let me see (pauses for 3 secs).
This is the unknown (pointing to the diagram)/unknown because of Mun Fai.
So let say this is one small unit/
S2 Okay
S3 Draw part-whole model
S4 You know how to draw?
S3 {shaking his head}
S4 I don’t know how to draw.
You just take 400 minus 28 first/400 minus 28. Then you have the left over

Discussion
The present study compared the effects of two learning conditions in a cognitive-
apprenticeship, computer-based environment on word-problem solving. Specifically,
it aimed to answer the following questions:

 2003 Blackwell Publishing Ltd, Journal of Computer Assisted Learning, 19, 46-55
52 S.K. Teong
• Do experimental students outperform control students on ability to solve word-
problems on their individual written test?
• Can experimental low achievers develop the ability to ascertain when to make
metacognitive decisions, and elicit better regulated metacognitive decisions than
lower achievers under the control condition? If so, do they require some time to
internalise the positive benefits of metacognitive training.
• Does the occurrence of metacognitive behaviours on its own ensure successful
word-problem solving?
• What is the role of WordMath in low achievers’ cognitive activity?
Important findings emerged from this study. First, the experimental students
outperformed control students on ability to solve word-problems on their individual
test measures. Second, the experimental low achievers developed the ability to
ascertain when to make metacognitive decisions and elicit better regulated
metacognitive decisions than control lower achievers, but they appear to require
some time for internalisation to occur before the positive benefits of metacognitive
training could prevail. Third, possessing metacognitive strategies is not sufficient. It
is the ability to know when and how to use metacognitive behaviours, when they are
needed, that is an important determinant to successful word-problem solving. Fourth,
WordMath appears to enhance experimental low achievers’ metacognitive
behaviours during word-problem solving, and also appears to enhance low
achievers’ metacognitive and cognitive behaviours during word-problem solving.
These findings are examined below in more detail.

Do experimental students outperform control students on ability to solve word-

problems on their individual written measure?
Empirical evidence from the experimental design supports the potential of meta-
cognitive training on the mathematical word-problem solving performance of low
achievers in a cognitive-apprenticeship, computer-based environment (see Table 1).
The findings concur with results of previous studies that focused on the effects of
metacognitive training on mathematical problem solving in a non-computer
environment (Schoenfeld, 1985; Cardelle-Elawar, 1995; Verschaffel et al., 1999).

Can experimental low achievers develop the ability to ascertain when to make
metacognitive decisions, and elicit better regulated metacognitive decisions than
control lower achievers?
There is no evidence from the case study design, however, that metacognitive
training inevitably leads to effectiveness in mathematical word-problem solving
performance. Visual inspection suggests that some treatment lower achievers needed
some time for internalisation to occur before positive metacognitive influence could
prevail in their word-problem solving from post-test to delayed post-test. For
example, there was evidence that S3 & S4 (E) were more successful, and were
devoting more time to regulating and monitoring their word-problem solving process
even after a prolonged period of six weeks with metacognitive instruction (see
Table 2). However, it was not known, in the present study, if this delayed
improvement on the pair of experimental low achievers would also prevail in their
word-problem solving over an extended period of time (e.g. after several months).
This issue merits further attention.

 2003 Blackwell Publishing Ltd, Journal of Computer Assisted Learning, 19, 46-55
 Mathematical word-problem solving 53
Does the occurrence of metacognitive behaviours on its own ensure successful
word-problem solving?
The analysis of transcripts of collaborative word-problem solving of four pairs of
students by using modified Artzt and Armour-Thomas’ framework also revealed that
the ability to know when and how to use metacognitive behaviours when they are
needed are important determinants to successful word-problem solving. Mayer
(1998) posits that one of the factors that discriminate successful problem solvers
from unsuccessful ones is their ability to know not only what to do, but also when to
do it with regard to using their cognitive skills. In this study, S3 & S4 (E), who
devoted 69.7%, and S5 & S6 (C), who devoted 47.6%, to metacognitive activities
were unsuccessful in solving the Q2 word-problem (see Table 2 and Figs. 2 & 3).
Though they monitored their activities, evidenced by the occasion ‘It cannot be’,
they failed to obtain the correct solution. In contrast, the protocol for S1 & S2 (E)
could be summarised as a well-regulated progression of activity (which led to their
success in their word-problem solving (see Fig. 1). They were also in control of their
cognitive actions throughout their word-problem solving.
The above examples suggest that the occurrence of metacognitive behaviours on
its own does little to ensure successful word-problem solving. This concurs with a
study (Stillman & Galbraith, 1998) which reveals that although successful groups in
displayed a high number of key points where metacognitive decisions would
influence cognitive actions, this alone was not a guarantee for success. Their advice
is that the focus of metacognitive training is to provide facilities that enable students
to develop a rich store of metacognitive strategies over an extended period of use.

What is the role of WordMath in low achievers’ cognitive activity?

WordMath appears to encourage low achievers to focus on the meaning of the
representation constructed in the computer environment (Noss & Hoyles, 1996; p.
228). For example, in the analysis of S1 & S2’s (E) word-problem solving, they
devoted 41.1% to analysis (see Fig. 1), which was mainly focused on ensuring that
all the vital information was represented on the diagram with WordMath tools, and
they were able to derive meanings from the representation they had drawn. This is
illustrated by the above exchange after the pair had drawn the diagram that
represented the Q2 word-problem with WordMath tools.
This example suggests that using WordMath enables students to reorganise their
cognitive activity and shift their focus to a (higher) ‘meta-level’ (Dörfler, 1993). It is
possible that drawing the diagram can be done without the tool considered. However,
these drawing actions have to be carried out in full and might entail considerable
problems (e.g. erasing a diagram using a rubber if the relationship between the
blocks are not accurate). The tool, WordMath, appears to free the students from that
workload by condensing the unfolded action into a single click of the button. On the
other hand, there were occasions when WordMath tools had neither helped students
reorganise their cognitive activity nor enabled them shift their focus to a higher
‘meta-level’. For example, S3 & S4 (E) failed to solve the Q2 word-problem and
they appeared to have difficulties in representing the word-problem situation using
WordMath tools. In most of S3 & S4′s word-problem solving during post-test 2, the
students tried to avoid drawing diagrams using WordMath tools. The above
exchange illustrates their reluctance to draw diagrams using WordMath tools while
solving the Q2 word-problem in post-test 2.

 2003 Blackwell Publishing Ltd, Journal of Computer Assisted Learning, 19, 46-55
54 S.K. Teong
During delayed post-test 2, S3 & S4 used WordMath tools to draw diagrams in
order to represent the word-problem. Their representations were appropriate and
they were successful in 50% of their word-problem solving. Based on the above
observations, it is appropriate to suggest that WordMath appears to amplify
experimental low achievers’ metacognitive behaviours during word-problem solving,
and also appears to amplify low achievers’ metacognitive and cognitive behaviours
during word-problem solving. These assertions merit further research.

Conclusion
The findings of this study provide evidence that the role of low achievers’
metacognition, influenced by metacognitive training in a cognitive-apprenticeship-
computer-based environment, can play in contributing to low achievers’ word-
problem solving performance. This represents a contribution to research on low
achievers’ metacognition influenced by metacognitive training in a computer
environment as well as providing a way forward in understanding how this influence
can help low achievers’ in word-problem solving.

References
Artzt, A. & Armour-Thomas, E. (1992) Development of a cognitive-metacognitive framework
for protocol analysis of mathematical problem solving in small groups. Cognition and
Instruction, 9, 137–175.
Bruner, J. (1985) Vygotsky: a historical and conceptual perspective. Culture, Communication
and Cognition: Vygotskian Perspectives (ed. J. Wertsch) pp. 21-34. Cambridge
University Press, Cambridge.
Cardelle-Elawar, M. (1995) Effects of teaching metacognitive skills to students with low
mathematics ability. Teaching and Teacher Education, 8, 109–121.
Dörfler, W. (1993) Computer Use and Views of the Mind. Learning from Computers. In
Mathematics Education and Technology (eds. C. Keitel & K. Ruthven) pp. 159–186.
Springer-Verlag, Berlin.
Kramarski, B. & Ritkof, R. (2002) The effects of metacognitive and email interactions on
learning graphing. Journal of Computer Assisted Learning, 18, 1, 33–43.
Looi, C.K. & Tan, B.T. (1998) A Cognitive-Apprenticeship-Based Environment for Learning
Word Problem Solving. Journal of Computers in Mathematics and Science Teaching, 17,
4, 339–354.
Mayer, R. (1998) Cognitive, Metacognitive, and Motivational Aspects of Problem Solving.
Instructional Science, 26, 1–2, 49–63.
Mevarech, Z. & Kapa, E. (1996) The effects of a problem-solving based Logo environment
on children’s information processing components. British Journal of Educational
Psychology, 66, 181–195.
Noss, R. & Hoyles, C. (1996) Windows on Mathematical Meanings. Learning Cultures and
Computers. Kluwer Academic Publishers, Dordrecht.
Schoenfeld, A. (1985) Mathematical Problem Solving. Academic Press, San Diego.
Schoenfeld, A. (1992) Learning to think mathematically: problem solving, metacognition and
sense making in mathematics. In Handbook of Research on Mathematics Thinking and
Learning (ed. D. Grouws) pp. 334–370. MacMillan, New York.
Stillman, G. & Galbraith, P. (1998) Applying Mathematics with Real World Connections:
Metacognitive Characteristics of Secondary Students. Educational Studies in
Mathematics, 36, 157–195.

 2003 Blackwell Publishing Ltd, Journal of Computer Assisted Learning, 19, 46-55
 Mathematical word-problem solving 55
Teong, S.K. (2000) The Effect of Metacognitive Training on the Mathematical Word Problem
Solving of Singapore 11–12 year olds in a Computer Environment. PhD Thesis.
University of Leeds, UK.
Verschaffel, L., De Corte, E., Van Lasure, S.V.G., Bogaerts, H. & Ratinckx, E. (1999)
Learning to solve mathematical application problems: a design experiment with fifth
graders. Mathematical Thinking and Learning. An International Journal, 1, 3, 195–229.

Appendix A: CRIME

C AREFUL READING Have I read and understood what I am supposed to find?

R ECALL POSSIBLE Draw (solve by drawing models/diagrams)
STRATEGIES Small (simplify problem using small numbers)
Parts (solve part(s) of the problem first)
Before (use before-after concept) [list information given before and
after the action; compare the information to find the unknown]
Backwards (solve by working backwards) [used when final result is
given, e.g. budgets are made by working backwards from how much
money is needed.]
I MPLEMENT • Have I labelled the models/diagrams to show the relationship
STRATEGY/ between the knowns and unknowns?
STRATEGIES • Can I re-arrange the blocks to see if there is a clearer relationship?
• Can I write Mathematics statement(s) to work out my answer?
M ONITOR • Am I on the right track?
• Am I getting closer to the goal?
• Am I still using the strategy I have chosen?
• Do I need to reread the problem and use another strategy?
E VALUATION • Does the answer make sense?
• Can I check the answer by using a test?
• Can I solve the problem in a different way?

Appendix B: Word problem test items

1. A farmer bought a total of 85 ducks and geese for $445. When 15 ducks died, there was an equal
number of ducks and geese left. If each goose cost $3 more than each duck, find the cost of each
goose.
2. Mr Lim had a bag of sweets for his class. After giving each pupil 8 sweets, he had 3 sweets left. If he
had given only 5 sweets to each pupil, he would have 108 sweets left. How many pupils were there
in his class?
3. There are altogether 15 marbles in the three boxes A, B and C. If we move 3 marbles from A to B,
move 2 marbles from B to C, and then move 1 marble from C to A, there will be the same number of
marbles in the boxes. How many marbles are there in A, B and C at first?
4. A class of 43 pupils donated a total of $306 to the Community Chest. Each boy donated $2 more
than each girl. If there were 5 more boys than girls, how much did each girl donate?
5. There is a group of children. A boy in the group says, ‘Not counting me, the number of boys is half
the number of girls.’ A girl in the group says, ‘Not counting me, there is the same number of boys
and girls.’ How many children are there in the group?
6. There are some red, yellow and blue beads in a box. 1/3 of the beads is red and 1/4 of the beads is
yellow. If 35 blue beads are removed from the box, there is half as many blue beads as yellow beads
now. How many beads are there altogether in the box at first?
7. Alice read 1/4 of a book on Sunday. She read 6 more pages on Monday than on Sunday. If she still
had 36 pages to read, how many pages did she read on Sunday?
8. Jim had a sum of money. He spent 3/5 of it on a pair of shoes and 4 pairs of socks. Each pair of
socks cost 1/5 as much as the pair of shoes. He had $36 left. What was the cost of the pair of shoes?
9. David and Betty each had an equal amount of money at first. After David had spent $18 and Betty
had spent $42, Betty had 2/3 as much money as David. How much money did each of them have at
first?
10. A tank was 1/5 full. When 700 ml. of water was poured into the tank, it became 2/3 full. Find the
capacity of the tank.

 2003 Blackwell Publishing Ltd, Journal of Computer Assisted Learning, 19, 46-55