Vous êtes sur la page 1sur 15

Educational Psychology, Vol. 19, No.

2, 1999 149

Overcoming Obstacles to Understanding and


Solving Word Problems in Mathematics

ALLAN B. I. BERNARDO, College of Education, De La Salle University, Manila,


Philippines

ABSTRACT The trouble students have with solving word problems often comes from the
difficulty in understanding the problem structure embedded in the problem text. This research
was conducted to study the interactive effects of learner and instructional variables on
understanding and solving word problems among Filipino-English bilingual grade school
students in the Philippines. The results showed better understanding and solution performance
(1) when problems were written in the students' first language, (2) when the problems were
re-worded to state more explicitly the relationship among the known and unknown quantities,
(3) for students in higher levels of schooling and (4) for students with higher levels of academic
achievement. The effects of re-wording were mediated by the effects of the language; and the
effects of language also varied between the high- and low-achievement students. In most cases
where the students' performance improved, the improvement seems to be due to better
comprehension of the text (or fewer basic comprehension errors); except with the improvement
due to re-wording, which seems to be due to fewer misinterpretation errors. The results of the
study provide further support for earlier findings, but also extend and modify the theoretical
ramifications of the earlier findings and point to some important implications for educational
practice.

Word problems are often considered to be the most challenging problems students have
to solve in mathematics education. In solving word problems, students have to as-
semble concepts and procedures and apply them towards the solution of one problem.
Failure to solve the problem can come from an error in just one of the range of concepts
and procedures applied. However, the most basic difficulty students have in solving
word problems lies in the ability to understand the mathematical problem structure that
is embedded in the problem text. Difficulties with comprehending the problem struc-
ture often lead to errors in the choice of the solution strategy. This study looks into a
number of factors that may constrain or facilitate the process of understanding and

0144-3410/99/020149-15 © 1999 Taylor & Francis Ltd


150 A. B. I. Bernardo

solving word problems in mathematics among Filipino-English bilingual grade school


students in the Philippines.
The task of understanding the word problem is most critical in successful word
problem solving. Various models of word problem solving (e.g. Kintsch & Greeno,
1985; Carpenter et al., 1988) assume that the first task for the problem solver is to
derive from the text an accurate mental representation of the problem. This mental
representation of the problem will be used as the basis for choosing the solution
processes that will operate. Several studies (e.g. Clarkson, 1991; Dark & Benbow,
1990; Stern, 1993) have documented how difficulties associated with comprehending
the problem text are associated with corresponding difficulties in problem solution. For
example, Cummins et al. (1988) found that students had more difficulty comprehend-
ing certain types of texts (i.e. those that have abstract and/or ambiguous language) and
committed more solution errors with these types of texts. They concluded that language
comprehension processes determine whether the student will be able to understand
correctly the different pieces of problem information and how these relate to each other.
Their understanding of these elements then determines the choice of solution proce-
dures (see also, Riley et al., 1983; Riley & Greeno, 1988).
Several studies (e.g. De Corte & Verschaffel, 1987; Davis-Dorsey et al., 1991) also
show that changes in the linguistic form of word problem texts affect a student's
efficacy in solving the word problems. For example, De Corte et al. (1985) re-worded
subtraction problems so that the relations among the known and unknown quantities
were made more explicit. Their subjects solved the re-worded problems much better
than those with standard wording. They proposed that these linguistic changes led to
variations in the ease with which the relationships among the quantitative information
are understood, which affected the solution procedures.
For bilingual students, another factor that is known to be related to problem
comprehension and solution is the student's proficiency in the language in which the
problem is stated (Adetula, 1990; Clarkson, 1992). In studies involving Filipino-En-
glish bilinguals, Bernardo (1996a, 1999) found that students were better at compre-
hending the problem text when it was written in the student's first and more proficient
language (whether Filipino or English). Moreover, students made more comprehension
errors that violate the problem structure, and consequently made more solution errors,
when the problem was written in the student's second and less proficient language.
The ability to understand word problems is also affected by internal variables or
factors related to the student's prerequisite knowledge and information-processing skills
related to the use of this knowledge (Kinstch & Greeno, 1985; Cummins et al., 1988).
In particular, students may vary in whether they have acquired more advanced knowl-
edge representations or schemas that allow for the efficient access and use of relevant
problem information. In this regard, others (e.g. Krutetskii, 1976; Briars, 1983; Mayer,
1985) have proposed that differences in ability and/or achievement actually correspond
to differences in the efficiency of access to and use of the relevant knowledge and
information-processing skills.
In the classroom, all these internal (student-related) and external (task or instruc-
tional) variables come into play simultaneously, mediating the effects of each other. For
example, De Corte et al. (1985) and Davis-Dorsey et al. (1991) found that the benefits
of re-wording the arithmetic word problems for easier comprehension were greater for
students in lower grades compared to those in higher grades. They argued that less
experienced problem solvers have more difficulty in comprehending the problems
because they lack the schematic problem-solving knowledge that more experienced
Word Problems in Mathematics 151

problem solvers have acquired. Hence, they benefit from interventions that aid the
basic comprehension processes.
Although it has not been shown in the literature, it is also conceivable that a variable
like student achievement would interact with an external variable like the language of
the problem; that is, it may be that the language of the problem will affect performance
of low-achieving students but not of high-achieving students. In other words, the
different variables that have been shown to affect problem-solving performance might
be considerably altered in the presence of other variables.
The present study investigated the effects of two external variables and two internal
variables on arithmetic word problem understanding and solution among Filipino-En-
glish bilingual grade school students. The first external variable chosen was problem
re-wording as defined and operationalised by De Corte et at. (1985). It was expected
that, generally, students would benefit from re-wording the word problems in ways that
make the known and unknown quantities more explicit. The second external variable
was the language format of the problem; that is, whether the problem was written either
in the student's first and more proficient language (Filipino) or second language
(English). It was hypothesised that, generally, students will be better at understanding
and solving word problems in Filipino. The first internal variable was the student's
grade level. Students from Grades 2 to 4 participated in the study. It was hypothesised
that students at higher grades would have acquired more schematised knowledge for
understanding and solving word problems, and would do better than those in the lower
grades. Finally, the variable of student's academic achievement was also studied.
Students were grouped into high-ability and low-ability groups based on their prior
academic performance. This variable was expected to have effects similar to the
preceding internal variable. More than the main effects of each of these variables, the
study sought to find how the effects of these variables interact with and mediate each
other. The goal was to determine how the facilitative effects of one variable may be
constrained by the effects of another variable, or may be enhanced by yet another
variable. In doing so, there can be a more rational understanding of the range of factors
that need to be dealt with in trying to promote students' understanding of word
problems in mathematics.

Method
Subjects and Procedure
The subjects were 283 students from Balara Elementary School; 97 were in Grade 2
(mean age: 8 years 4 months), 100 in Grade 3 (mean age: 9 years 4 months), and 86
in Grade 4 (mean age: 10 years 6 months). In this school, the students were assigned
to different sections in each grade level using ability-based groupings as determined by
performance in the previous grade level. In the study, the first and the third sections
(based on ability-based groupings) in each grade level were chosen. The first section in
each grade level was designated as the high academic achievement group and the third
section as the low-achievement group. All the students have Filipino as their first
language and are taught mathematics using English as the primary language of instruc-
tion and using English language materials.
The students participated in the study during class hours within a time period
allotted by the teacher and administrator-in-charge. Students in one section completed
the task all at once; different sections completed the tasks at different times depending
152 A. B. 1. Bernardo

TABLE 1. Sample word problems

Type of problem Conventional Re-worded

Change (start set unknown) Jose won 3 balls. Jose had some balls.
Now he has 5 balls. He won 3 more balls.
How many balls did Jose Now he has 5 balls.
have in the beginning? How many balls did
Jose have in the
beginning?

Combine (subset unknown) Rico and Pat have 9 Rico and Pat have 9
candies altogether. candies altogether.
Rico has 3 candies. 3 of these candies
How many candies does belong to Rico. The
Pat have? rest belong to Pat.
How many candies does
Pat have?

Compare (difference unknown) Martin has 8 mangoes. There are 8 spoons, but
Rita has 3 mangoes. there are only 3 forks.
How many mangoes does How many spoons do
Martin have more than not have a matching
Rita? fork?

on the allotted schedule. Each student was given one randomly chosen test booklet and
was told to read the instructions and complete the task at their own pace. The test
booklets were collected after all students were done.

Materials
Initially, a set of six one-step subtraction word problems were prepared in English,
based on the problems used in the studies of De Corte et at. (1985). Two problems
were change problems: A known set is added to an unknown start set to arrive at a
known final set (this type of problem is labelled as change 5 in the classification scheme
developed by Riley et at., 1983). Two problems were combine problems: Two sets, one
known and one unknown, are combined to form a known final set (combine 2 in Riley
et al.). The last two problems in the set were compare problems: Two known sets are
given and the difference between the two sets is unknown (compare 1). The two
analogous problems for each of these three problem types contained superficially
dissimilar information and all the problems were worded in the conventional form
found in most textbooks.
A second set of problems was derived from the first, following the procedures used
by De Corte et at. (1985) and Hudson (1983). In the second set, the problems were
re-worded so that that the relationships among the known and unknown quantities are
made more transparent. Table I shows examples of each type of problem in each set.
The six problems in each set were then translated from English to Filipino by a
research assistant who was fluent in both languages, and then translated back and
revised by another research assistant who was also fluent in both languages. Hence,
there were four sets of six problems each: one conventional set in Filipino, one
conventional set in English, one re-worded set in Filipino, and one re-worded set in
English.
Word Problems in Mathematics 153

From these four sets, individual test booklets were made. Each booklet consisted· of
12 problems, three problems from each set. The three problems taken from each set
represented each of the three problem types. The selection of the problems from each
set was done using the following rules: For each specific problem in the booklet, the
other three analogous problems that appeared in the booklet were (a) a superficially
different analogous problem in the same language and different wording version, (b) a
superficially similar analogous problem in the different language and different wording
version, and (c) a superficially different analogous problem in the different language
and same wording version.
Each test booklet included one page for instructions and student information, and
four pages each containing three problems from each of the four language and wording
sets. The three problems from each set were typed in one page leaving enough space
after each problem for the solution. The sequence of the four pages was arranged using
a partial counterbalancing scheme, wherein across all subjects each of the four sets has
an equal probability of occupying either the first, second, third, or last positions.

Results
Quantitative Analysis
The subjects' solutions for each of the 12 word problems were scored for accuracy; a
correct answer was scored as 1 point and an incorrect answer was scored 0 points. The
percentage of correct answers for each set of problems is summarised in Table II. The
percentage were then analysed using a 2 X 2 X 2 X 3 Analysis of Variance for mixed
factorial designs (with language and wording of problem as within-group variables, and
academic achievement and grade level of student as between-group variables). [1]
The analysis showed that all the hypothesised main effects were verified. There was
a statistically significant main effect of grade level, F(2, 277) = 34.88, MSE = 0.27, P <
0.0001. Overall, those in Grade 4 had the highest percentage of correct answers
(M=68.51, SE=3.15), followed by those in Grade 3 (M=54.58, SE=2.86), and
those in Grade 2 performed the poorest (M = 35.22, SE = 3.05). There was also a
statistically significant main effect of academic achievement, F(I, 277) = 71.81,
MSE=0.27, p<O.OOOI, with the high academic achievement students (M=66.60,
SE = 2.28) outperforming their low-achievement counterparts (M = 39.39, SE = 2.57).
This pattern is simply a validation of the operational definition of the variable and
seems to be consistent at all grade levels. (Note that the low-achievement students
in Grade 4 performed at about the same level as the high-achievement students in
Grade 2.)
There was also a statistically significant interaction effect between grade level and
academic achievement levels, F(2, 277) = 3.40, MSE = 0.27, P < 0.035. The differ-
ences in performance between the high-and low-achievement students was very much
marked at the Grade 2 level (high achievement: M= 55.37, SE= 3.91; low achieve-
ment: M = 17.79, SE = 2.89), then stabilised at more constant levels in Grade 3 (high
achievement: M=66.29, SE=3.83; low achievement: M=45.39, SE=3.70) and
Grade 4 (high achievement: M=78.41, SE=3.34; low achievement: M=58.13,
SE= 4.96).
Moving to the task variables, there was also a statistically significant main effect of
problem wording, F(I, 277) = 95.27, MSE = 0.05, p< 0.0001. The results show that,
overall, the students did better with the re-worded problems (M = 58.95, SE = 2.01),
-
..,.
UI

~
~
~

ra.
TABLE II. Percentage of correct answers

Filipino problems English problems $:I

Conventional Re-worded Conventional Re-worded <:>

Standard Standard Standard Standard


Grade/achievement level Mean error Mean error Mean error Mean error

Grade 2
Low achievement 18.59 4.43 28.85 3.89 10.26 2.97 13.46 3.21
High achievement 45.19 5.10 68.15 4.96 48.89 5.25 59.26 4.96
Grade 3
Low achievement 41.67 4.27 61.90 4.20 33.33 4.41 44.64 4.97
High achievement 59.85 4.54 78.03 4.05 56.06 5.06 71.21 4.28

Grade 4
Low achievement 53.17 5.69 74.60 5.18 47.62 6.03 57.14 6.04
High achievement 75.00 3.93 84.09 3.51 71.21 4.92 83.33 3.34

*For each subject, the percentage of correct answers out of three items was computed for each type of problem. The mean
percentages indicated in the table represent the arithmetic average of the percentages across all subjects in each group.
Word Problems in Mathematics 155

compared to those worded in the conventional form (M = 45.41, SE = 2.06), which


replicates earlier findings (De Corte et al., 1985; Davis-Dorsey et aI., 1991).
The main effects of language of the problem was also statistically significant, F(l,
277) = 40.19, MSE = 0.04, P < 0.0001. Consistent with earlier research (Adetula,
1990; Bernardo, 1996a, 1999), students performed better when the problems were in
their first language, Filipino (M = 56.18, SE = 1.90), compared to when the problems
were in their second language, English (M = 48.17, SE = 2.11). This result was ob-
tained in spite of the fact that all the students were being instructed in mathematics
using English as the primary medium of instruction and that they were being taught
word problem solving using English language problems.
There were also statistically significant interaction effects between the wording and
the language of the problem, F(1, 277) = 10.84, MSE = 0.03, P < 0.0015. In particular,
it seems that the benefit of re-wording was greater when the problems were written in
Filipino (conventional: M= 47.70; SE= 2.15; re-worded: M = 64.66, SE= 2.06), com-
pared to when the problems were in English (conventional: M = 43.11, SE = 2.24;
re-worded: M = 53.24, SE = 2.28). Another way of describing this interaction is that
with the problems worded conventionally, it did not make much of a difference whether
the problems were in Filipino (M=47.70) or English (M=43.11). But for the
reworded problems, the students were doing much better with the Filipino problems
(M = 64.66), compared to the English ones (M = 53.24). Hence, it seems that, overall,
the subjects were performing their best with the re-worded problems in Filipino. This
observation is true for all grade levels, for both academic achievement groups, as shown
in Table II.
The effect oflanguage was also affected by the students' academic achievement level,
F(1, 277) = 12.65, MSE = 0.04, P < 0.0005. For students with a record of high aca-
demic achievement, the language of the problem seems to have made no difference
(Filipino: M = 68.30, SE = 2.28; English: M = 64.91, SE = 2.50). But for students with
a record of low academic achievement, presenting the problems in Filipino led to better
performance (Filipino: M=45.44, SE=2.68; English: M=33.33, SE=2.79). As
Table II shows, this pattern seems to be consistent at all grade levels.

Error Analysis
To explore further the relationships among the different variables, an error analysis was
conducted. Answers to each problem were coded in terms of the following six cate-
gories: Correct answer (CA); Operation error (OE)-the solution added the two given
numbers in the problem; all the problems were subtraction problems, hence, adding the
numbers was the incorrect arithmetic operation; First given number error (FGNE)-
the answer to the problem was the first given number in the problem; Second given
number error (SGNE)-the answer to the problem was the second given number in the
problem; Miscellaneous error (ME)-errors for which there were no ready explana-
tions; and no answer (NA).
The error analysis involved looking at differences in the error patterns related to the
main effects and interaction effects found in the quantitative analysis. There seems to
be differences in the likelihood of producing only two types of errors: the OE and the
FGNE; there seems to be no marked difference in the distribution of the other types of
errors. Hence, all results in this error analysis will refer to the OE and FGNE.
The OE is most likely a result of a gross inability to parse the meaning of the problem
text. The most probable scenario that would lead the subjects to select to add the given
156 A. B. 1. Bernardo

numbers in the problem, rather than to subtract them, is when the subjects totally
missed on the expressed relationships between the known and unknown quantities.
Hence, it is assumed that the OE is an indication of a gross failure in comprehending
the problem text.
On the other hand, the FGNE could be a result of some minor misinterpretation on
the part of the text. Consider the conventionally worded change problem in Table I. It
is possible that in parsing the question, "How many balls did Jose have in the
beginning?", the subjects incorrectly interpreted "beginning" to mean the beginning of
the problem text. Similarly, in the conventionally worded compare problem, the subject
might have interpreted the question, "How many mangoes does Martin have more than
Rita?" as "How many mangoes does Martin have?" In both cases of misinterpretation,
the first given number error would be the 'appropriate' answer. Such misinterpretations
are not wholly unreasonable if the subject is not too familiar with the schemas for
wording problem texts and for interpreting different types of word problems. For these
subjects, the text was probably ambiguous or lacked explicitness. Hence, in this error
analysis, the FGNE is assumed to indicate the lack of proficiency with parsing the
ambiguities in word problem texts.
Of course, these two types of incorrect understanding associated with OEs and
FGNEs are not mutually exclusive. For purposes of this analysis, however, differentiat-
ing the errors in this way can reveal some insights about the specific difficulties that
subjects may have had and the factors that may have brought these about. The
assertions in this qualitative analysis would most certainly be strengthened with more
explicit tests in future studies.
The task variables were considered first. It seems that re-wording led to improved
problem-solving performance because the FGNEs decreased (see right part of Table
III). There were no marked differences in the distribution of errors for the conventional
and re-worded problems, except for the FGNE. This result is consistent with the
notion that re-wording makes the text more explicit, leading to fewer FGNE errors.
On the other hand, using Filipino as the language of the problem seems to have led
to better problem-solving performance because the OEs decreased (see right part of
Table III). There were no marked differences in the distribution of errors for the
Filipino and English problems, except for the OE. It seems that the difficulties students
have with English problem texts involve gross difficulties in parsing the problem texts.
This is not surprising considering most of the students are still in the process of
developing their proficiency in English.
Interestingly, the interaction effect found between re-wording and language seems to
reflect the trends in the preceding observations. There were more FGNEs for the
conventional problems and more OEs with the English problems (see bottom right part
of Table IV). However, the proportion of OEs was lower for the re-worded Filipino
problems compared to the conventional Filipino problems (13.43 vs 17.90), but the
proportion of OEs was about the same for conventional and re-worded English
problems (24.62 vs 23.91). When the students read English problem texts, it seems that
they have much difficulty with the basic parsing of the text, so much so that increasing
the transparency of the text as regards the known and unknown quantities makes no
difference at all. But when the students read Filipino problem texts, they have an easier
time parsing these; hence, their comprehension becomes more sensitive to changes in
wording. The students can take advantage of the more explicit wording in the text to
avoid the gross miscomprehension errors of which the OEs are symptoms.
The error analysis related to the student grouping variables reveal interesting rela-
TABLE III. Distribution (in percent) of answers for the different categories

Academic Language
Grade level achievement Problem wording ofprob!em

Answer categories Grade 2 Grade 3 Grade 4 High Low Conventional Re-worded Filipino English

CA 35.22 54.58 68.51 66.60 39.39 45.41 58.95 56.18 48.17


OE 27.66 21.58 9.40 8.83 29.83 20.91 19.02 15.67 24.26
FGNE 24.74 17.92 17.64 19.74 20.56 26.21 14.13 22.26 18.08
SGNE 4.64 1.66 1.26 1.69 3.33 1.94 3.18 2.30 2.82
ME 6.36 3.25 1.94 2.19 5.44 3.65 4.18 2.71
~
5.12
NA 1.37 1.00 1.26 0.94 1.44 1.88 0.53 0.88 1.53
a.
Note: CA = correct answer; OE = operation error; FGNE = first given number error; SGNE = second given number error; ~
~
ME = miscellaneous error; and NA = no answer. Percentages do not always add up to 100% because of rounding. ~
~

~
s..
~
I:l
:::
.....
...,

-
U\
-.J
158 A. B. 1. Bernardo

TABLE IV. Distribution (in percent) of answers for the different categories across interaction groupings

Academic Achievement X Grade Level

High academic achievement Low academic achievement

Answer categories Grade 2 Grade 3 Grade 4 Grade 2 Grade 3 Grade 4

CA 55.37 66.29 78.41 17.79 45.39 58.13


OE 13.52 6.63 6.25 39.90 33.33 12.70
FGNE 24.07 23.30 11.74 25.32 13.69 23.81
SGNE 3.33 1.33 0.38 5.78 1.93 2.18
ME 3.33 2.08 1.14 8.97 4.17 2.78
NA 0.37 0.38 2.08 2.24 1.49 0.40

Achievement X Language Wording X Language

High achievement Low achievement Conventional Re-worded

Filipino English Filipino English Filipino English Filipino English

CA 68.30 64.91 45.44 33.33 47.70 43.11 64.66 53.24


OE 6.64 11.03 23.67 36.00 17.90 23.91 13.43 24.62
FGNE 20.30 19.17 24.00 17.11 28.86 23.56 15.67 12.60
SGNE 1.13 2.26 3.33 3.33 1.88 2.00 2.71 3.65
ME 2.76 1.63 2.67 8.22 2.59 4.71 2.83 5.54
NA 0.88 1.00 0.89 2.00 1.06 2.71 0.71 0.35

Note: CA = correct answer; OE = operation error; FGNE = first given number error; SGNE = second
given number error; ME = miscellaneous error; and NA = no answer. Percentages do not always add up
to 100% because of rounding.

tionships between academic achievement, grade level and understanding word prob-
lems. The main difference between the high- and low-achievement students seems to
lie in the tendency to use an incorrect operation (see middle part of Table III). If we
assume that such errors indicate failures in comprehension, then it would seem that
what distinguishes the high- and low-achievement students is their basic skill or ability
to parse word problem texts. More interestingly, if one considers the interaction effect
between academic achievement and language (see bottom left part of Table IV), it
seems that the interaction lies in the distribution of these errors, as well. Again, there
are clearly more gross comprehension failures with the English problem texts and
among the low-achievement students. However, whereas there is a marked difference
in the proportion of OEs between Filipino and English problems among low-achieve-
ment students (23.67 vs 36.00), the corresponding difference among high-achievement
students is definitely not comparable in size (6.64 vs 11.03). This result suggests that
the major incidences of gross comprehension failures occur among low-achievement
students reading English texts.
The interaction between achievement and grade levels seems to reveal characteristics
of how the underlying comprehension skills develop among high- and low-achieving
students. The distribution of OEs (top part of Table III) shows that among high-
achieving students, there is a clear drop in such errors at Grade 3, and the errors level
off at Grade 4. But among low-achieving students, this error drops only at Grade 4.
Moreover, at Grade 4, the low-achieving students are still making as many gross
comprehension errors as the high-achieving students in Grade 2. This seems to suggest
Word Problems in Mathematics 159

that the basic comprehension skills of the low-achieving students are not only low, they
are also developing much slower. One could also assert that their comprehension skills
are at comparably lower levels because they develop these skills slowly. This observa-
tion might not be surprising to many, but the point that needs to be underscored is that
these problems with developing comprehension skills have strong effects on the devel-
opment of mathematical skills of the students.
Overall, the error analysis suggests that differences in accuracy levels may be
associated with the subjects' basic ability to comprehend the problem texts. The effects
of grade level, achievement level and language of the text seem to be associated by the
distribution of OEs that are assumed to indicate gross comprehension errors. The only
effect that seems to be associated with another comprehension error is that of re-word-
ing. The benefit of re-wording seems to be related to minimising the misinterpretation
errors due to the lack of explicitness in the problem text-this, not coincidentally, is the
very problem that re-wording is intended to remedy. However, the benefit of re-word-
ing seems to depend on the basic comprehension abilities. That is, if there is too much
difficulty in the basic comprehension process in the first place, the students will not be
able to take advantage of the clarity afforded by re-wording.

Discussion
This research was conducted to investigate the effects of different learner and instruc-
tional variables on understanding and solving arithmetic word problems among Fil-
ipino-English bilingual grade school students. The results showed better understanding
and solution performance when problems were written in the students first language
(L1) and when the problems were re-worded to state more explicitly the relationship
among the known and unknown quantities. Performance was also better for students in
higher levels of schooling and with higher levels of academic achievement. The effects
of re-wording were mediated by the effects of the language format. The effects of
language also varied between the high- and low-achievement students. In most cases
where the students' problem-solving performance improved, the improvement seems to
have been brought about by better comprehension of the text (or fewer basic compre-
hension difficulties). The exception is with the improved performance due to re-word-
ing, which seems to be a result of fewer misinterpretation errors due to ambiguities in
the text. The results were derived from a sample of bilingual students from one school,
hence, caution should be taken in generalising some of the findings, particularly those
that relate to the grouping variables. But the generalisability of the findings regarding
the relationship among specific variables is strengthened by the fact that the results of
the study replicate, and provide further support for, earlier findings. More important,
the results also extend and modify the theoretical ramifications of the earlier findings
and point to some important implications for educational practice.
The effect of re-wording found in previous studies (e.g. De Corte et al., 1985;
Davis-Dorsey et ai., 1991) was replicated. This finding is significant in at least two
ways. Firstly, it suggests very clear ways by which mathematical problem-solving
performance of students can be aided by making minor revisions in the instructional
materials. Secondly, the result suggests that poor performance or even failure on the
part of the students in doing word problems may not necessarily indicate a lack of
mathematical (computational or conceptual) skill, but may indicate difficulties in
understanding the text in which the mathematical information is embedded.
In this connection, the effects of the language of the text on comprehension and
160 A. B. 1. Bernardo

problem solving (replicating Adetula, 1990; Bernardo, 1996a, 1999) are also
significant. In the Philippines, students are instructed in mathematics (and also in other
subjects) using the English language and English language materials. English is the
second or third language of the overwhelming majority of the student population.
Difficulties that students have in understanding and solving word problems in math-
ematics may be brought about by their difficulties with the English language, among
other things. Hence, the language of instruction may be creating noise in the learning
process.
The lower levels of understanding and problem-solving accuracy in English persist
until Grade 4, after at least 4 continuous years of instruction using English. The
improvement in performance from Grade 2 to 4 seems to be mainly due to the
increasing proficiency in the language (i.e. students make fewer gross comprehension
errors). In their earlier studies, De Corte et at. (1985) asserted that such improvements'
across grade levels indicate the emergence of more advanced problem-solving schemas
acquired through more problem-solving experience. It seems that among the Filipino
students sampled, the basic proficiency with the language of the problem and instruc-
tion is an additional factor that constrains the process of understanding and solving the
problems. Even if students are taught in English, and even as they may acquire better
proficiency in English, they are always better in comprehending problems in Filipino.
This particular finding relates to an earlier finding by Bernardo (1999) which showed
that Grade 4 Filipino students whose Ll is English were better at understanding and
solving word problems compared to their counterparts whose Ll is Filipino. Such an
advantage for those whose Ll is English was not found among Grade 1 and 2 students.
The students with different Ll s come from the same schools and classes, and studied
word problem solving under the same teachers. Yet those whose Ll was English later
emerge as better than the other group. A possible interpretation of that finding is that
those whose Ll happens to be English end up learning the word problem-solving skills
better because the language of instruction is English, whereas those whose Ll is
Filipino or any other non-English language continuously have difficulties learning the
word problem-solving skills because of their lack of proficiency in English. Differences
in the students' proficiency in English seem to be a reasonable factor to implicate, as
most other pertinent instructional factors are apparently equal.
In addition the effects of wording and language, a new finding in the study was the
interaction between the two variables. Students seem to have difficulty with problems
worded in the conventional form, whatever the language of the problem. Even when the
subjects were proficient in the language, the information in most word problem texts is
too sparse to make sense. However, the benefit ofre-wording was stronger with Filipino
problems. With efficient comprehension skills in the language, students are apparently
more sensitive to the effects of altering the problem texts. This result also points to an
important consideration for those thinking about working on improving instructional
materials. The clear lesson is that changes in the instructional variables will have effects
on performance if the changes complement the information-processing capabilities of
the learner. In the particular domain of the study, re-wording of word problems will not
benefit most students if the problems are still in English. One wonders whether other
instructional reforms will also have limited or no benefits for Filipino students as long
as English is used as the language of instruction for students whose first language is not
English.
Interestingly, the language of the problem does not seem to be an important factor
for the high-achieving students. These students are much less likely to commit gross
Ward Prablems in Mathematics 161

comprehension errors of the text that lead to incorrect problem solution with both
Filipino and English problems. The high-achieving students apparently have no
difficulties with parsing English texts. There are many possible ways of interpreting this
result. It is possible that those who start out in school with better English proficiency
are better able to learn and benefit from instruction in the English language. It may be
that good learners are able to acquire English language skills better and this achieve-
ment allows the same students to benefit more from mathematics instruction in
English. It is also possible that children who are better learners, learn the material in all
their subjects equally effectively, hence the high-achieving students have no difficulty
either in English or mathematics.
All these interpretations seem plausible and the subtle differences might be difficult
to distinguish empirically. But a point that seems difficult to ignore is the apparent
reality that general academic achievement and mathematical learning seem to be
strongly related with proficiency in the English language. It seems that for student
whose Ll is not English, lack of proficiency in English constrains learning and
academic achievement when the medium of instruction is English. If one considers the
fact that basic language proficiencies are usually set during the first 5 or 6 years of life,
often before formal schooling begins, it seems unfair that the present Philippine
educational system has incorporated a feature that systematically disadvantages most of
its students.
That abstract mathematical knowledge is often remote and difficult to understand for
most students is an important factor to consider in mathematics education (see e.g.
Bernardo, 1997). Unfortunately, many instructional practices serve as additional obsta-
cles to students' understanding of mathematical concepts and procedures; practices
that relate to the linguistic aspect of mathematics problems. The results of the study
show that such difficulties in understanding and solving word problems in mathematics
can be overcome if features of the mathematical tasks are suited to the information-pro-
cessing capabilities of the students. In particular, students perform better when the text
of the problem is written in their more proficient language and is worded in ways that
allow more easy comprehension of the problem structure.
Although the study focused on the performance of Filipino-English bilingual stu-
dents, the results and their implications are relevant to other bilingual students in
educational systems that are considering the implications of using one language over
another in formal education. Such is the case in a growing number of countries in Asia,
Africa, Europe and North America. Even countries that were previously dominantly
monolingual, like the United States, are now confronting the language of instruction
issue as a result of growing ethnic diversity in the communities and schools (see e.g.
Hakuta, 1986). In many countries, educators are now pondering questions like: Which
language should be used for teaching? For which subjects? Does the language used
affect achievement and learning? Does the language used in teaching affect the stu-
dent's self-efficacy and self-esteem? Of course, the specific character of the bilingual
situation in different countries varies. However, educators and psychologists from one
bilingual country can still draw insights from the functional relations established
between the language of instruction and student learning and achievement in another
bilingual country. Hence, although the findings of these study are specifically focused
on mathematics achievement of bilingual Filipino students, the conclusions made
regarding the effects of using one language over another and on the effects of changing
the linguistic structure of problems could also be relevant to the learning experiences of
other bilingual students in other subjects.
162 A. B. l. Bernardo

Indeed, if we attempt to abstract more general principles from the specific findings
of the study, the emerging picture of the effective learning situation is one where
features of the instructional environment are anchored on characteristics of the learner.
Such a leamer-centred approach is integral to many contemporary perspectives in
educational theory (Bernardo, 1996b; Black & Atkin, 1996; Bruer, 1993; Entwistle,
1995; Lambert & McCombs, 1998; Marshall, 1996; McGilly, 1994) and in mathemat-
ics education, in particular (Cobb, 1994; De Corte, 1995). The challenge that faces
educators now lies in designing and implementing programmes that draw from our
emerging understanding of the information-processing capabilities of our students and
of the complex processes involved in their development.

Acknowledgements
This research was supported by the Don Eduardo Cojuangco Sr. Professorial Chair in
Liberal Arts awarded to the author by De La Salle University, Manila. The author
thanks Nino Jose Mateo, Alma San Buenaventura and Milagros Tan for their assistance
in preparing the research materials and in administering the test; the administrators,
teachers and students of the Balara Elementary School, Quezon City, for their support
and enthusiasm during the data-gathering phase; and the two reviewers for their useful
and detailed comments.

Correspondence: Allan B. 1. Bernardo, College of Education, De La Salle University,


2401 Taft Avenue, Manila 1004, Philippines. Email: c1aabb@mai1.dlsu.edu.ph

NOTE
[1] Following the procedures for preparing the test booklets, roughly half of the subjects used one set
of items and the rest used another set. To determine whether using different sets had an effect on
performance, the scores were first analysed using 2 X 2 X 2 X 3 X 2 AN OVA with the problem set
as an additional between-group factor. The ANOVA showed that the problem set had no
systematic effect on performance, F(1, 277) = 0.52, MSE = 0.33, P = 0.47. In fact, the means for
the two problem sets were nearly identical, suggesting that the two problem sets were fairly
equivalent.

REFERENCES
ADETUI.A, L.O. (1990) Language factor: does it effect children's performance on word problems?,
Educational Studies in Mathematics, 21, pp. 351-365.
BERNARDO, A.B.I. (1996a) Language and understanding in mathematical problem solving, Philippine
Journal of Educational Measurement, 7, pp. 29-54.
BERNARDO, A.B.!. (1996b) Using contemporary psychological perspectives to intensify the involvement
of psychologists in Philippine schools, Philippine Journal of Psychology, 29, pp. 39-64.
BERNARDO, A.B.I. (1997) The psychology of mathematics learning and performance: implications for
educational practice, Layag, 2, pp. 57-81.
BERNARDO, A.B.!. (1999) The Effects of Language Format on Understanding Wont Problems and on
Developing Problem-Solving Skills in Mathematics. Manuscript submitted for publication.
BLACK, P. & ATKIN, J.M. (1996) (Eds) Changing the Subject: innovations in science, mathematics, and
technology education (London, Routledge).
BRIARS, D.l (1983) An information-processing analysis of mathematical ability, in: RF. DILLON & RR
SCHMECK (Eds) Individual Differences in Cognition, Vol. 1, pp. 181-204 (New York, Academic Press).
BRUER, IT. (1993) Schools for Thought (Cambridge, MA, MIT Press).
Word Problems in Mathematics 163

CARPENTER, T.P., MOSER, J.M. & BEBOUT, H.C. (1988) Representation of addition and subtraction
word problems, Journal for Research in Mathematics Education, 19, pp. 345-357.
ClARKSON, P.C. (1991) Language comprehension errors-a further investigation, Mathematics Edu-
cation Research Journal, 3, pp. 24-33.
ClARKSON, P.C. (1992) Language and mathematics: a comparison of bilingual and monolingual
students in mathematics, Educational Studies in Mathematics, 23, pp. 417--430.
COBB, P. (1994) Where is the mind? Constructivist and sociocultural perspectives on mathematical
development, Educational Researcher, 23, pp. 13-20.
CUMMINS, D.D., KINTSCH, W., REUSSER, K. & WEIMER, R (1988) The role of understanding in solving
word problems, Cognitive Psychology, 20, pp. 405--438.
DARK, V.I. & BENBOW, C.P. (1990) Enhanced problem translation and short-term memory: compo-
nents of mathematical talent, Journal of Educational Psychology, 82, pp. 420--429.
DAVIS-DORSEY, J., Ross, S.M. & MORRlSON, G.R (1991) The role of rewording and context personal-
ization in the solving of mathematical word problems, Journal of Educational Psychology, 83, pp.
61-68.
DE CORTE, E. (1995) Fostering cognitive growth: a perspective from research on mathematics learning
and instruction, Educational Psychologist, 31, pp. 37--46.
DE CORTE, E. & VERSCHAFFEL, L. (1987) The effect of semantic structure on first-graders' strategies for
. solving addition and subtraction word problems, Journal for Research in Mathematics Education, 18,
pp. 363-381.
DE CORTE, E., VERSCHAFFEL, L. & DE WIN, L. (1985) Influence of the semantic structure of word
problems on second graders' eye movements, Journal of Educational Psychology, 82, pp. 359-365.
ENTWISTLE, N. (1995) (Ed.) Influences of instructional settings on learning and cognitive develop-
ment: findings from European research programs [Special issue], Educational Psychologist, 30(1), pp.
1-54.
HAKUTA, K. (1986) Mirror of Language: the debate on bilingualism (New York, Basic Books).
HUDSON, T. (1983) Correspondences and numerical differences between disjoint sets, Child Develop-
ment, 54, pp. 84-90.
KINTSCH, W. & GREENO, J.G. (1985) Understanding and solving word arithmetic problems, Psycholog-
ical Review, 92, pp. 109-129.
KRUTETSKII, V.A. (1976) The Psychology of Mathematical Abilities in Schoolchildren (Chicago, The
University of Chicago Press).
LAMBERT, N.M. & MCCOMBS, B.L. (1998) (Eds) How Students Learn: reforming schools through learner-
centered education (Washington, DC, American Psychological Association).
MARSHAll, H.H. (1996) (Ed.) Recent and emerging theoretical frameworks for research on classroom
learning: contributions and limitations [Special issue], Educational Psychologist, 31 (3/4), pp. 147-340.
MAYER, RE. (1985) Mathematical ability, in: RJ. STERNBERG (Ed.) Human Abilities: an information-
processing perspective (New York, Freeman).
MCGILLY, K. (1994) (Ed.) Classroom Lessons: integrating cognitive theory and classroom practice (Cam-
bridge, MA, MIT Press).
RILEY, M. & GREENO, J.G. (1988) Developmental analysis of understanding language about quantities
and of solving problems, Cognition and Instruction, 5, pp. 49-101.
RILEY, M.S., GREENO, J.G. & HELLER, J.I. (1983) Development of children's problem-solving ability in
arithmetic, in: H.P. GINSBURG (Ed.) The Development Of Mathematical Thinking, pp. 153-196 (New
York, Academic Press).
STERN, E. (1993) What makes certain arithmetic word problems involving the comparison of sets so
difficult for children?, Journal of Educational Psychology, 85, pp. 7-23.

Vous aimerez peut-être aussi