You are on page 1of 13

Journal of Educational Psychology Copyright 2004 by the American Psychological Association

2004, Vol. 96, No. 4, 635– 647 0022-0663/04/$12.00 DOI: 10.1037/0022-0663.96.4.635

Enhancing Mathematical Problem Solving Among Third-Grade Students


With Schema-Based Instruction
Lynn S. Fuchs, Douglas Fuchs, Karin Prentice, Carol L. Hamlett, Robin Finelli, and Susan J. Courey
Vanderbilt University

The purposes of this study were to assess the effects of schema-based instruction (SBI) in promoting
mathematical problem solving and to investigate schema induction as a mechanism in the development
of mathematical problem solving. Twenty-four 3rd-grade teachers, with 366 students, were assigned
randomly to conditions that provided instruction on 4 types of word problems. The 3 16-week conditions
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

were contrast, SBI, and SBI with practice in sorting word problems into schemas. Students were pre- and
This document is copyrighted by the American Psychological Association or one of its allied publishers.

posttested on mathematical problem-solving tests and were posttested on schema development. Students
receiving SBI, with and without sorting practice, improved more than the contrast group on problem-
solving measures. Concurrently, the SBI groups’ schema development exceeded that of the contrast
group, and schema development explained a substantial portion of unique variance in students’ post-
treatment problem-solving performance. Results also suggested the need for additional research testing
the contribution of practice in sorting word problems.

Mathematical problem solving is a form of transfer, which quiring the relevant problem–solution methods); when this hap-
requires students to apply the problem–solution rules they have pens, the transfer required for mathematical problem solving oc-
mastered to novel problems (e.g., Mawer & Sweller, 1985; Sweller curs. Novel problems differ from source problems in terms of
& Cooper, 1985). According to schema-induction theory, a major superficial features, which alter problems without modifying the
challenge in producing mathematical problem-solving expertise is problem-type schema (e.g., step-up function) or the required so-
the development of schemas. As defined by Gick and Holyoak lution methods. The most basic superficial problem feature is the
(1980), a schema is a general description of two or more problems, cover story (see Appendix), which changes a source problem in
which students use to group problems into types that require only minor ways and thus requires relatively narrow schema. As
similar solution methods (Chi, Feltovich, & Glaser, 1981; Gick & novel problems become increasingly different from source prob-
Holyoak, 1980; R. E. Mayer, 1992; Quilici & Mayer, 1996). lems, however, broader schemas are required to recognize novel
Theory suggests that the broader the schema (i.e., the more general
problems as belonging to established schema. Superficial features
the description of the problem category), the greater the probability
that render problems more novel than cover stories (and involve
individuals will recognize connections between novel and familiar
greater transfer distance) include the problem’s format, key vo-
problems, so they will know when to apply the solution methods
they have mastered. cabulary, question posed, and scope (i.e., where the target problem
Consider, for example, the problems in the Appendix. Each is a small piece of a larger problem; see Appendix).
belongs to a step-up function schema because its structure requires Unfortunately, the development of broad schemas for mathe-
a step-up function for solution. The first problem shows a source matical problem types does not occur naturally for many individ-
problem—for which a set of solution methods has been taught. uals (e.g., Bransford & Schwartz, 1999; Cooper & Sweller, 1987;
Because it is not possible to teach every step-up function problem, R. E. Mayer, Quilici, & Moreno, 1999). Moreover, the link be-
however, learners must formulate a schema, or category, that tween the development of schemas and problem solving, espe-
incorporates problems requiring step-up function solutions. When cially for young and low-performing students, awaits experimental
learners encounter a novel (i.e., untaught) problem, they must demonstration. Establishing such a link is potentially important for
recognize its structure as fitting an established schema (and re- extending our understanding of problem solving and for guiding
effective instructional design in this area. Consequently, the major
purposes of this study were to assess the effects of schema-based
instruction (SBI) in promoting mathematical problem solving and
Lynn S. Fuchs, Douglas Fuchs, Karin Prentice, Carol L. Hamlett, Robin
Finelli, and Susan J. Courey, Department of Special Education, Peabody to investigate schema induction as a mechanism in the devel-
College, Vanderbilt University. opment of mathematical problem solving among third graders.
This research was supported in part by U.S. Department of Education, We randomly assigned participants to a contrast group or to SBI
Office of Special Education Programs Grant H324C030115 and National and assessed problem-solving skill and schema development as
Institute of Child Health and Human Development Core Grant a function of condition. A secondary purpose was to examine
HD15052 to Vanderbilt University. Statements do not reflect the posi-
whether guided practice in sorting problems into schemas might
tion or policy of these agencies, and no official endorsement by them
should be inferred.
provide added value in promoting schema development and
Correspondence concerning this article should be addressed to Lynn S. problem-solving skill. In this introduction, we review prior
Fuchs, Department of Special Education, Vanderbilt University, Box 328 related work and clarify the major and secondary purposes of
Peabody, Nashville, TN 37203. E-mail: lynn.fuchs@vanderbilt.edu the present study.
635
636 FUCHS ET AL.

Prior Work on Schema Induction Related to Mathematical arithmetic word problems and suggests the potential for schema
Problem Solving construction theory to guide instruction on math problem solving.
Of course, findings do not provide the basis for separating the
Most prior work investigating schema induction and mathemat- effects of SBI from the use of instructional diagrams and do not
ical problem solving involves short-term, independent study of document schema induction as the mechanism by which problem
problems for the purpose of inducing schemas. A smaller body of solving improves.
research involves longer duration, explicit SBI designed to en- To isolate the contribution of SBI, Fuchs, Fuchs, Prentice,
hance mathematical problem solving. Burch, Hamlett, Owen, et al. (2003) randomly assigned third-grade
classes to contrast and experimental conditions, this time in whole-
Short-Term Independent Study class, rather than tutoring, format. The contrast condition involved
instruction designed and implemented by classroom teachers; in
Intervention work has examined how examples can be designed the experimental conditions, researchers designed and conducted
to build schemas related to mathematical problem solving. Quilici lessons. One of the researcher-designed conditions taught
and Mayer (1996), for example, demonstrated that college students problem–solution methods; the other researcher-designed condi-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

who independently studied statistics problems grouped by problem tions taught problem–solution methods along with SBI. SBI pro-
type (t test vs. chi square vs. correlation) sorted subsequent prob- duced superior problem-solving outcomes compared with teacher-
lems more effectively into problem types compared with students designed instruction and compared with researcher-designed
who studied without examples. These researchers also showed that problem–solution method instruction. Effect sizes ranged between
schema induction was strengthened when examples emphasized 1.16 and 1.82 depending on transfer distance. As with Jitendra et
structure rather than superficial features. Interestingly, the pattern al. (1998), however, results failed to document schema induction
was stronger for lower ability students who, unless primed to do as a mechanism that enhanced problem solving.
otherwise, may tend to focus on superficial features more than
higher ability peers. Of course, Quilici and Mayer operationalized
Purposes of the Present Study
low ability with SAT scores below 575, a definition with limited
generalizability to the public school-age population. Prior work suggests the potential for instruction based on
In related work conducted with elementary school children, schema-induction theory to enhance mathematical problem solv-
Chen (1999) examined how the variability of examples used for ing but fails to substantiate schema induction as the process by
independent study facilitates not only schema induction but also which problem-solving skill develops. The primary purposes of the
problem–solution accuracy and its relation to schema develop- present study, therefore, were to assess the effects of SBI in
ment. Results illustrated how a broader range of problems for promoting mathematical problem solving while investigating
independent study was more likely than a narrow range of exam- whether instruction inspired by schema theory actually results in
ples to build schemas and produce problem-solving transfer. This schema development. The target population was 8 –9-year-olds.
time, however, effects were mediated by age: Older children Understanding how to build schemas and the relation between
(10 –12-year-olds) were more effective than 8 –9-year-olds at ex- schema building and math problem solving for this population is
tracting schemas from the example problems and at solving sub- important in light of studies demonstrating the difficulty 8 –9-year-
sequent problems. olds experience extracting schemas and succeeding with transfer
As Quilici and Mayer’s (1996) and Chen’s (1999) studies illus- (Chen, 1999; Gentner, 1988, 1989; Gholson, Eymard, Morgan, &
trated, important questions remain about how to build schemas and Kamhi, 1987). Moreover, because previous work (Cooper &
mathematical problem solving among low-performing and Sweller, 1987; Fuchs, Fuchs, Karns, Hamlett, & Katzaroff, 1999;
younger students. Both studies, as is the case for much of the D. P. Mayer, 1998; Woodward & Baxter, 1997) demonstrated that
research on analogical problem solving and schema induction, mathematical problem solving (and the transfer involved) is more
relied on single-session interventions without explicit instruction difficult to effect with low-achieving students, we separated out-
to build students’ schemas. As Quilici and Mayer concluded, comes for students at different points on the achievement contin-
further study is warranted to examine whether instruction and uum. In a related way, because most students with disabilities
structured practice in schema-inducing activities, rather than inde- receive instruction in general education (U.S. Department of Ed-
pendent study of examples, might strengthen effects. ucation, 1999) and given the chronic, severe difficulties these
students experience in achieving transfer (e.g., White, 1984), we
Longer Duration, Explicit SBI examined results separately for these learners.
In addition, the experimental instruction used in the present
In fact, some research has explored the potential of explicit SBI study was an innovative form of SBI because it sought to broaden
as a method for promoting mathematical problem solving. For schemas for problem types (e.g., distinguishing step-up function
example, in a small-group tutoring study conducted with low- problems from half problems) by developing children’s schemas
achieving students in Grades 2– 6, Jitendra et al. (1998) tested a for superficial problem features (e.g., distinguishing a problem’s
two-step intervention that combined SBI with the use of diagrams. source of novelty as a different format vs. a different question). In
When presented with arithmetic word problems, students first this article, we refer to these two types of schemas, respectively, as
categorized the problem as a change, group, or compare problem problem-type schemas and superficial-feature schemas. To de-
type; this first step was designed to develop schemas. Then, velop superficial-feature schemas, we explicitly taught students
students used a diagram representing the relevant problem type to about transfer and how transfer involves changes in superficial
assist in problem solution. This multicomponent intervention re- features, which make problems seem unfamiliar without modify-
sulted in an effect size of 0.45 on students’ accuracy in solving ing the problem type or the required solution methods. Our SBI
SCHEMA DEVELOPMENT AND MATH PROBLEM SOLVING 637

treatment taught children about four superficial-feature schemas: we randomly assigned 8 teachers to each condition: contrast, SBI, SBI plus
format, key vocabulary, question, and scope (see the Appendix for sorting practice.1 Teacher groups were comparable on age, degree, and
examples). years teaching (see Table 1).
One method for assessing the degree to which students possess Student participants were the 366 children in these classrooms for whom
we had obtained parental consent and who were present for each pre- and
schemas is sorting tasks (Chi et al., 1981; Cummins, 1992; Hins-
posttesting. On the basis of pretreatment scores on the immediate-transfer
ley, Hayes, & Simon, 1977; Quilici & Mayer, 1996), which ask
problem-solving measure (see below), we designated students as low
students to group problems with similar solution methods. In the performing (LO; less than 0.5 SDs below the mean), average performing
present study, we relied on a variation of the sorting task by asking (AV; within 0.5 SDs of the mean), or high performing (HI; greater than 0.5
students to identify the problem-type schema to which problems SDs above the mean). Students were distributed across LO, AV, and HI
belonged. We also asked students to identify whether novel prob- status, respectively, as follows: in contrast (n ⫽ 113), 27 (24%), 54 (48%),
lems involved transfer (i.e., representing novelty beyond a cover and 32 (28%); in SBI (n ⫽ 127), 25 (20%), 63 (49%), and 39 (31%); and
story change) and, if so, to identify the superficial-feature schema in SBI plus sorting (n ⫽ 126), 32 (25%), 54 (43%), and 40 (32%). Within
(i.e., different format or different key vocabulary or different each classroom, 12 to 19 students were represented in the database. Within
question or larger scope) that produced the problem’s novelty. We LO, AV, and HI groups, students were comparable across conditions on
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

gender, subsidized lunch, race, and English language learner status (see
This document is copyrighted by the American Psychological Association or one of its allied publishers.

randomly assigned participants to a contrast condition (teacher-


Table 1). The 366 children in the complete data set were demographically
designed and implemented instruction on four problem types) or to
comparable with the remaining 83 pupils who were absent on 1 or more
experimental conditions (researcher-designed and implemented in- testing days.
struction on the same problem types). The researcher-designed
instruction, inspired by schema-induction theory, attempted to Treatment
expand children’s problem-type schemas and their superficial-
feature schemas via explicit instruction on superficial problem Treatment shared by all conditions. All three conditions (contrast,
features. We assessed schema development for problem types and SBI, SBI plus sorting) shared three instructional features: the district
curriculum, a basal text, and a 3-week unit on general math problem-
for superficial features via this sorting task as well as their
solving strategies.2 The district curriculum required teachers to address the
problem-solving growth.
same content each week of the year. To teach this curriculum, teachers
At the same time, we wondered whether practice, mediated by relied primarily on the basal text, Math Advantage (Burton & Maletsky,
teacher feedback and explanations, in sorting problems into their 1999).
problem-type schemas and their superficial-feature schemas might We chose our four problem types from the curriculum to ensure that
help build stronger schemas and superior problem solving. Sorting contrast group students had instruction relevant to the study. For the
practice is designed to help students analyze and articulate simi- students, we labeled the four problem types (or schemas) as follows:
larities and differences among problems and to categorize those shopping list (e.g., “Joe needs supplies for the science project. He needs 2
problems into problem-type schemas that require the same solution batteries, 3 wires, and 1 board. Batteries cost $4 each, wires cost $2 each,
methods—without the cognitive load demanded with actual prob- and boards cost $6 each. How much money does he need to buy sup-
plies?”), half (e.g., “Marcy will buy 14 baseball cards. She’ll give her
lem solution. The possibility that such sorting practice might
brother half the cards. How many cards will Mary have?”), buying bags
enhance schema development and problem solution has been
(e.g., “Jose needs 32 party hats for his party. Party hats come in bags of 4.
raised in previous work (Schwartz & Bransford, 1998), but has not How many bags of party hats does Jose need?”), and pictograph (e.g.,
been tested with primary-grade children of varying achievement “Mary keeps track of the number of chores she does on this chart [picto-
status or for students with disabilities, for whom schema develop- graph is shown with label: each picture stands for 3 chores]. She also took
ment is more challenging (Chen, 1999; Gentner, 1988, 1989; her grandmother to the market 3 times last week. How many chores has
Gholson et al., 1987). So, in the present study, we included another Mary done?”).
experimental condition to which participants were randomly as- All students also received a 3-week unit (two lessons per week), de-
signed: SBI plus guided sorting practice. signed and delivered by the researchers, that addressed general math
In sum, the purposes of this study were to assess the effects of problem-solving strategies: making sure answers make sense; lining up
numbers from text to perform math operations; checking computation; and
SBI in promoting mathematical problem solving while investigat-
labeling work with words, monetary signs, and mathematical symbols. This
ing schema induction as a mechanism for the development of
unit, common across all three conditions, was not based on schema-
mathematical problem solving and to examine the added value of induction theory. The six lessons in this unit relied on worked examples
guided sorting practice on schema development and problem- with explicit instruction, dyadic practice, independent work with checking,
solving skill. Across these purposes, we addressed internal validity and homework. Research assistants conducted the lessons in whole-class
via random assignment of classrooms to treatments (using class- format, but teachers were present for every lesson and assisted the research
room as the unit of analysis). We addressed external validity by assistants in answering questions, monitoring participation, and checking
relying on whole-class instruction delivered in general education work. The six lessons each lasted 30 – 40 min, for a total of 180 –240 min
for 16 weeks and by separating effects for high-, average-, and of researcher-designed instruction across 6 school weeks.
low-achieving students as well as those with disabilities.
1
We chose the term contrast (rather than control) group because the
Method contrast group involved instruction relevant to the four problem types
addressed in the experimental conditions. This included teacher-designed
Participants and implemented instruction on the four problem types as well as a 3-week
researcher-designed and implemented unit on general problem-solving
In six schools in a southeastern urban district in the United States, 24 strategies.
2
third-grade teachers (all female) volunteered to participate. Stratifying so For a manual with teaching scripts and materials, contact Lynn S.
that each condition was represented approximately equally in each school, Fuchs.
638 FUCHS ET AL.

Table 1
Demographics by Initial Achievement Status and Condition

Contrast SBI SBI plus sorting

Variable n % M SD n % M SD n % M SD ␹2 (df) F(2, 21)

Teacher age (years) 3.40 (6)


21–29 2 25 2 25 1 13
30–39 1 13 2 25 1 13
40–49 2 25 0 0 1 13
50–59 3 38 4 50 5 63
Teacher degree 1.07 (2)
Bachelor’s 4 50 2 25 3 38
Master’s 4 50 6 75 5 63
Years teaching 12.50 8.72 12.88 7.20 15.63 6.00 0.43
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

No. of male studentsa


This document is copyrighted by the American Psychological Association or one of its allied publishers.

LO 13 48 11 44 12 38 0.70 (2)
AV 24 44 31 49 25 46 0.27 (2)
HI 16 50 20 51 19 48 0.12 (2)
Student subsidized lunch
LO 22 81 20 80 27 84 0.20 (2)
AV 20 37 17 27 18 33 1.40 (2)
HI 4 13 3 8 4 10 0.46 (2)
No. of Black studentsa
LO 20 74 17 68 20 63 1.00 (2)
AV 19 35 25 40 23 43 0.63 (2)
HI 4 13 7 18 4 10 1.11 (2)
English language learner
LO 3 11 3 12 1 3 1.85 (2)
AV 4 7 7 11 5 9 0.47 (2)
HI 3 9 2 5 2 5 0.48 (2)

Note. SBI ⫽ schema-based instruction; LO ⫽ low-performing students; AV ⫽ average-performing students; HI ⫽ high-performing students.
a
Others were White.

SBI. All students in SBI and in SBI plus sorting also received four example, and as they referred to the poster, explained how each step of the
additional 3-week units. Each 3-week unit comprised six lessons, with two solution method had been applied in the worked example. Students re-
lessons per week (i.e., 24 lessons across the four 3-week units). SBI and sponded frequently to questions. After reviewing the concepts and present-
SBI plus sorting students also received two review lessons the week after ing several worked examples in this way, research assistants presented
winter break, for a total of 26 SBI lessons (plus the six general math partially worked examples while students applied steps of the solution.
problem-solving strategies lessons); by contrast, the contrast condition Students then completed 1– 4 problems in dyads, where stronger students
only received the six general math problem-solving strategies lessons. helped weaker students solve problems and check work with answer keys.
Researchers designed the lessons, and research assistants implemented the Lessons 2– 4 were structured similarly, with a greater proportion of time
lessons in whole-class format. Classroom teachers were present for every spent on partially worked examples and dyadic practice. Also, at the end of
lesson and assisted the research assistants in answering questions, moni- Lessons 2– 4, students completed one problem independently, checked
toring participation, and checking work. work against an answer key, and were assigned a problem for homework.
SBI lessons addressed problem–solution methods and schema- Classroom teachers provided students feedback on homework.
enhancing methods. Each 3-week unit was dedicated to one of the four Lessons 5– 6 of each unit continued to focus on the unit’s problem type
problem types: shopping list problems, half problems, buying bags prob- but taught students about how four superficial problem features make
lems, and pictograph problems. In each 3-week unit, Lessons 1– 4 taught problems seem unfamiliar even though the problem type stays the same
children the problem–solution methods required for the problem type; and even though the same solution methods are required. (Research assis-
Lessons 5– 6 were designed to broaden the problem-type schema by tants used the term problem types rather than problem-type schemas and
developing four superficial-feature schemas that make a problem seem used the term transfer features rather than superficial problem features.)
novel without changing the problem type or the required solution method. Each problem used for instruction in Lessons 5– 6 varied the cover story as
In each unit, the first problem–solution lesson (Lesson 1) and the first well as one superficial problem feature. (Research assistants referred to
schema-enhancing lesson (Lesson 5) lasted about 40 min; others lasted these problems as transfer problems.)
25–30 min. So, the number of instructional minutes delivered by research In Lessons 5– 6, the research assistant first taught that transfer means to
assistants for SBI units approximated 800 (including the two cumulative move: Just as we transfer (move) to a different school, we can transfer
review lessons following winter break) across 14 school weeks (or 50 min (move) skills we learn to new situations. They also presented examples of
per week). how children transfer skills; for example, children learn to drink from a
All 26 lessons relied on worked examples with explicit instruction, toddler cup, then move this skill to a real cup, then move this skill to a
dyadic practice, independent work with checking, and homework. glass, then move this skill to a soda bottle. Other examples were presented
Problem–solution methods were taught in Lessons 1– 4 of each unit, using from everyday life, and students volunteered examples. Math examples
problems that varied only cover stories. A poster listing the steps of the were included (e.g., we learn to add two-digit horizontal problems, then
solution method was displayed in the classroom. In Lesson 1 of each move this skill to add two-digit vertical problems, then move this skill to
3-week unit, research assistants taught students about the underlying con- add three-digit problems, then move this skill to the supermarket check-out
cepts and structure inherent in the problem type, presented a worked line).
SCHEMA DEVELOPMENT AND MATH PROBLEM SOLVING 639

After discussing the meaning of transfer, research assistants taught conditions, respectively, were 303.13 (SD ⫽ 41.26), 289.25 (SD ⫽ 44.65),
students about the four superficial features that change a problem without and 301.25 (SD ⫽ 38.49), F(2, 21) ⫽ 0.61, ns.
altering its problem type or solution. Each unit covered the same four Treatment fidelity. Prior to the first delivery of each lesson in each
superficial problem features but the superficial problem features were condition, the research assistants agreed on the essential points of infor-
taught in relation to the unit’s problem type. Students were taught that a mation in the script and made a checklist of these points (i.e., each point of
familiar problem type, for which a solution is known, can (a) be formatted essential information equaled 1 point on the fidelity scale). Every session
so that the problem looks novel, (b) use an unfamiliar key word, (c) pose was audiotaped. At the end of the study, 2 research assistants indepen-
a different or additional question, or (d) be placed within a larger problem- dently listened to tapes while completing the checklist to identify the
solving context. A poster, “Transfer: Four Ways Problems Change,” was percentage of essential points addressed (i.e., 1 point ⫽ one piece of
displayed in the classroom. In Lesson 5, research assistants explained the essential information). We sampled tapes so that, within treatments, re-
poster by illustrating each superficial feature with a worked example. They search assistants and lesson types were sampled equitably. For each con-
gradually moved to partially worked examples. Then, students worked in trast class, 1–2 tapes were sampled (all Unit 1); in each of the other two
pairs to apply the solution method to problems that varied superficial treatments, 5–7 tapes were sampled per class (distributed equally across
features. Finally, homework was assigned, and classroom teachers de- Units 1–5). The percentage of essential points addressed averaged 94.3
briefed on the homework the following day. In Lesson 6, research assis- (SD ⫽ 9.56) for contrast, 96.4 (SD ⫽ 7.43) for SBI, and 94.6 (SD ⫽ 6.91)
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

for SBI plus sorting.


This document is copyrighted by the American Psychological Association or one of its allied publishers.

tants reviewed the four superficial features using similar procedures, except
students spent more time working in dyads and then completed a transfer
problem independently and scored work against a key. Throughout Lessons Measures
5– 6, teachers also reminded students to search novel problems, in and out
of school, for superficial features to recognize familiar problem types so Problem solving. We used three measures of problem solving: imme-
they could apply the solutions they knew. diate transfer, near transfer, and far transfer. For problems on the
SBI sorting practice. In the SBI plus sorting condition, research assis- immediate-transfer measure, the only source of novelty was the cover
tants also incorporated guided schema-based sorting practice. Sorting prac- story; otherwise, the problems resembled those used for instruction in
tice began in Unit 3 when instruction had addressed more than one Lessons 1– 4 of each unit. On the near-transfer measure, each problem
incorporated a novel cover story as well as a novel superficial feature,
problem-type schema for categorization and had covered all four
resembling problems in Lessons 5– 6 of each unit, but in an unfamiliar way.
superficial-feature schemas for categorization. The sorting activity, con-
The far-transfer measure introduced multiple sources of novelty and in no
ducted during each lesson in Units 3–5, was called “What Kind of Prob-
way resembled any problem used during instruction.
lem?” The research assistant read and showed a problem while students
Each measure had two alternate forms; the problems in both forms
followed along on paper. For each problem, the research assistant asked
required the same operations and presented text with the same number and
students to answer three multiple-choice questions independently on paper:
length of words. The immediate-transfer alternate forms incorporated the
same numbers. The near-transfer alternate forms incorporated the same
1. What kind of problem is this? (Response options were buying
numbers. The far-transfer alternate forms used similar numbers. In half of
bags, shopping list, half, or pictographs.)
the classes in each condition, we used Form A at pretest and Form B at
2. Is this a transfer problem? (Response options were yes or no.) posttest; in the other half, forms were reversed.
Immediate transfer. The immediate-transfer measure comprised 10
3. What kind of transfer problem is it? (Response options were this word problems in the style of problems used for problem–solution method
is not a transfer problem, different look, different words, differ- instruction but with novel cover stories. The problems represented all of the
ent question, or small piece of a bigger problem.) problem-solving units. Across the 10 problems, the maximum score was
43. The scoring rubric awarded points for correct computation and correct
So, the first question corresponded to the problem-type schemas; the last labeling of each step of each problem. For example, for the following
two questions to the superficial-feature schemas. Response options to problem, “You want to buy some crayons. Crayons come in bags with six
Question 1 included the problem-type schemas that had been addressed crayons in each bag. How many bags of crayons do you need to buy to get
instructionally (i.e., in Unit 3, buying bags and shopping list problem-type 17 crayons?”, the rubric awarded 2 points for finding three bags (or 1 point
schemas were included; in Unit 4, buying bags, shopping list, and half for finding two bags) plus 1 point for labeling the answer with a word.
problem-type schemas; in Unit 5, buying bags, shopping list, half, and Cronbach’s alpha was .95; concurrent validity with the TerraNova (CTB/
McGraw-Hill, 1997) was .58. Interscorer agreement, computed on 20% of
pictograph problem-type schemas). Problems were sampled systematically
protocols by two independent blind scorers was .983 at pretreatment and
to represent the different problem-type schemas and the different
.977 at posttreatment.
superficial-feature schemas.
Near transfer. The near-transfer measure comprised seven problems: a
After students answered the questions, research assistants debriefed by
shopping list problem with a novel format, one shopping list problem with
discussing answers in a large-group format and reviewing strategies for
a novel question, a buying bags problem with a different key word, a
recognizing different problem types and superficial features. The review
buying bags problem with a novel question, a half problem with unfamiliar
emphasized the structural features that defined each problem type and the
vocabulary, a pictograph problem with a novel question, and one picto-
nature of the superficial problem features. Teachers also reminded students
graph problem with a novel question. Across items, the maximum score
that when presented with math problems, they should search for a familiar
was 47. The scoring rubric awarded points for correct computation and
problem type before initiating work. During each of the 18 relevant lessons correct labeling of each step of each problem. For example, for the
(i.e., Units 3–5, with six lessons per unit), the sorting activity took approx- following problem,
imately 10 min. It was added to SBI.
Delivery. Each of 4 research assistants had responsibility for classes in Kelly had $25 to spend on earrings. The sign at the store looked like
every condition. All sessions were scripted to ensure consistency of infor- this: “Earrings! One pair costs $12.” When Kelly got up to the
mation; however, to permit natural teaching styles, scripts were studied, not counter, the clerk told her that earrings were half the price on the sign.
read. At the end of the study, classroom teachers reported the number of Kelly bought one pair. How much money did Kelly have left?,
minutes per week they spent on math, including this project (which
involved approximately 50 min per week for SBI and 70 min per week for the rubric awarded 1 point for calculating the cost of the earrings, 1 point
SBI plus sorting). Means for the contrast, SBI, and SBI plus sorting for labeling the cost with a word, 1 point for labeling the cost with a money
640 FUCHS ET AL.

sign, 1 point for calculating the amount left, 1 point for labeling the amount Data Collection
left with a word, and 1 point for labeling the amount left with a money sign.
Cronbach’s alpha was .94; concurrent validity with the TerraNova was .55. Trained research assistants collected data in whole-class arrangement.
Interscorer agreement, computed on 20% of protocols by two independent On the problem-solving measures, students were pre- and posttested; on
blind scorers was .958 at pretreatment and .962 at posttreatment. schema-development, students were posttested. Pretesting occurred during
Far transfer. The far-transfer measure introduced multiple sources of the 3 weeks before treatment; posttesting occurred during the 3 weeks
novelty by (a) combining all four problem types, (b) simultaneously following treatment. For the immediate- and near-transfer problem-solving
varying all four superficial features, (c) incorporating irrelevant text and measures and for the schema-development measure, research assistants
numbers, and (d) assessing application of six additional essential skills read aloud each item and provided students time to complete work before
from the district curriculum. Also, to decrease association with experimen- progressing to the next item. For the far-transfer problem-solving measure,
tal treatment, the measure was formatted to look like a commercial test research assistants read the entire assessment and reread portions to indi-
(printed with a formal cover, on green paper, with photographs and viduals, on request, as they worked. To avoid familiar research assistants
prompting awareness that experimental conditions might be applicable,
graphics interspersed throughout the test booklet). The measure included a
they did not collect data in classrooms where they had delivered lessons.
multiparagraph narrative with charts and graphs, and it posed four inter-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

related questions. Two assessments were constructed as alternate forms:


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Although the context of the problem situations differed, the structure of the Data Analysis
problem situation and the questions were identical, and the problem solu-
tions and reading demands were equivalent. Performance was scored Because teachers were assigned randomly to conditions, we used teacher
according to a rubric with four dimensions: conceptual underpinnings, as the unit of analysis and treated students’ initial status (LO, AV, HI) as
computational applications, problem-solving strategies, and communica- a within-teacher factor. This allowed us to compare effects for students of
tive value. The original rubric (Kansas State Board of Education, 1991) varying status in the same classrooms and to test interactions between
scored responses on a 6-point scale. To enhance reliability, we awarded initial status and condition. For each initial status group of children, each
points on a finer basis (e.g., the problem-solving strategies score included teacher had a mean pre- and posttreatment score on each problem-solving
points for finding relevant information, accumulating to a total, showing all measure and a mean set of posttreatment schema-development scores.
computation, working the answer in distinct multiple parts, labeling at least For each problem-solving and schema-development score, we conducted
half of the multiple parts, and labeling work with monetary and operation a two-factor mixed model analysis of variance (ANOVA). Condition was
signs). The maximum score across dimensions and questions was 70. the between-teachers variable; initial student status was the within-teacher
Cronbach’s alpha was .94; concurrent validity with the TerraNova was .67. variable. For problem-solving measures, these analyses were applied to (a)
Interscorer agreement, computed on 20% of protocols by two independent pretreatment scores to examine treatment group comparability and (b)
blind scorers was .903 at pretreatment and .928 at posttreatment. Given improvement scores to investigate treatment efficacy. (Results were anal-
work (Fuchs et al., 2000) demonstrating the deleterious effects of student ogous using change scores and analysis of covariance. Although these
unfamiliarity with performance assessments, research assistants delivered a procedures are equally acceptable for analyzing two-wave data, with each
45-min “test-wiseness” lesson immediately before pre- and posttesting in presenting a different set of problems, we opted for change scores because
their interpretation is more straightforward.)
all treatment groups.
To evaluate pairwise comparisons for significant effects, we used the
Schema development. The schema-development measure presented
Fisher least significant difference (LSD) post hoc procedure (Seaman,
students with 16 novel problems. Eight problems were structured like
Levin, & Serlin, 1991). To estimate the practical significance of effects, we
problems on the immediate-transfer problem-solving measure (represent-
computed effect sizes (ESs): on the problem-solving scores, by subtracting
ing the four problem-type schemas with novel cover stories). Eight prob-
the difference between improvement means and then dividing by the
lems were structured like problems on the near-transfer problem-solving pooled standard deviation of the improvement divided by the square root of
measure (each representing one of the four problem-type schemas with a 2(1 – rxy), where x is the pretest and y is the posttest (Glass, McGaw, &
novel cover story and with a novel variation in one of the four superficial- Smith, 1981); on the schema-development scores, by subtracting the dif-
feature schemas). ference between means and dividing by the pooled standard deviation
A research assistant read and showed a problem while students followed (Hedges & Olkin, 1985).
along on paper copies. For each problem, students answered three multiple- For students with disabilities, we ran a one-way ANOVA on each
choice questions: problem-solving and schema-development score, using condition as the
factor. We used student as the unit of analysis because of the small sample
1. What kind of problem is this? (Response options were buying size and uneven distribution of students across classrooms. We evaluated
bags, shopping list, half, or pictographs.) pairwise comparisons and computed ESs as for the larger sample.
To explore the contribution of schema development to problem-solving
2. Is this a transfer problem? (Response options were yes or no.) development, we conducted a series of planned regression analyses, using
a combined (summed) score across problem-type and superficial-feature
3. If this is a transfer problem, what kind of transfer problem is it? schema development. For each posttreatment problem-solving measure
(Response options were this is not a transfer problem, different (immediate, near, and far transfer), we conducted a pair of regressions, first
look, different words, different question, or small piece of a entering the pretreatment problem-solving score and seeing how much
bigger problem.) additional variance was explained by schema development; next entering
schema development and seeing how much additional variance was ex-
Two scores were derived: one for problem-type schema development plained by the pretreatment problem-solving score.
and another for superficial-feature schema development. Scores were cor-
rected for guessing with a standard correction formula {number correct –
[number incorrect/(number of response choices – 1)]}. Cronbach’s alpha Results
was .88 for problem-type schema development and .75 for superficial-
feature schema development. Interscorer agreement, computed on 20% of Table 2 shows means and standard deviations for the schema-
protocols by two independent blind scorers was .996 (agreements divided development and problem-solving measures, Table 3 shows results
by items). of the correlation analyses, Table 4 shows results of the planned
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 2
Performance by Initial Achievement Status and Condition

LO AV HI Across

SBI plus SBI plus SBI plus SBI plus


Contrast SBI sorting Contrast SBI sorting Contrast SBI sorting Contrast SBI sorting

Variable M SD M SD M SD M SD M SD M SD M SD M SD M SD M SD M SD M SD

Schema development
Problem-type
schemas 7.79 1.29 11.92 1.23 11.64 1.99 8.61 1.31 11.04 2.56 12.97 1.31 9.62 2.67 14.23 1.21 13.61 1.55 8.67 1.55 12.39 0.88 12.72 1.17
Superficial-feature
schema 3.11 0.22 3.65 0.64 4.41 0.50 2.81 0.48 3.80 0.82 4.56 0.71 2.91 0.73 4.72 0.50 5.05 0.91 2.94 0.37 4.06 0.40 4.67 0.31
Problem solving
Immediate transfer
Pre 2.50 1.40 2.82 0.79 3.49 1.10 7.36 2.04 7.01 1.52 5.99 0.60 11.28 2.30 14.05 4.28 12.96 3.05 7.05 0.92 7.96 1.57 7.48 1.37
Post 11.24 4.23 31.10 3.29 31.90 4.57 18.10 5.58 34.24 7.65 36.52 3.54 23.95 7.52 40.73 1.41 39.45 3.23 17.76 4.79 35.36 3.03 35.96 2.91
Improve 8.74 3.39 28.27 3.01 28.41 4.06 10.74 5.04 27.22 7.29 30.53 3.34 12.96 3.05 26.68 3.42 26.48 3.56 10.71 4.04 27.39 3.25 28.47 1.99
Near transfer
Pre 2.10 0.73 1.75 0.73 1.97 1.17 5.68 2.60 4.97 1.50 4.83 1.48 7.10 2.17 7.91 2.72 7.10 1.53 4.96 1.51 4.88 1.05 4.63 1.12
Post 5.64 1.55 21.29 2.33 19.65 3.59 9.77 3.06 25.12 6.80 25.37 5.09 13.60 5.08 33.26 2.83 29.04 6.88 9.67 2.24 26.56 2.39 24.69 4.37
Improve 3.54 1.39 19.53 2.42 17.68 3.20 4.09 2.94 20.15 6.14 20.54 4.53 6.50 4.34 25.35 4.58 21.94 6.39 4.71 2.12 21.68 3.21 20.06 3.82
Far transfer
Pre 7.93 2.64 8.78 4.40 7.52 3.85 9.89 5.92 10.58 7.62 9.11 3.99 15.79 8.33 19.24 7.64 17.31 6.30 11.20 4.85 12.87 4.48 11.31 4.05
Post 12.74 3.29 21.06 4.65 18.98 3.68 18.33 7.80 27.48 10.25 26.96 8.23 32.01 13.98 45.72 2.27 42.06 8.29 21.03 6.32 31.42 3.73 29.33 5.33
SCHEMA DEVELOPMENT AND MATH PROBLEM SOLVING

Improve 4.81 3.01 12.28 5.90 11.45 3.60 8.44 5.30 16.90 7.69 17.85 8.02 16.22 8.75 26.48 6.77 24.75 6.77 9.82 2.17 18.55 6.24 18.02 5.03

Note. SBI ⫽ schema-based instruction; LO ⫽ low-performing students; AV ⫽ average-performing students; HI ⫽ high-performing students; Pre ⫽ pretest; Post ⫽ posttest; Improve ⫽ posttest score
minus pretest score.
641
642 FUCHS ET AL.

Table 3
Correlations Among Pretreatment Problem Solving, Schema Development, and Posttreatment
Problem Solving

Variable M SD 1 2 3 4 5 6 7

1. Schema development 13.90 3.98 — .37 .69 .25 .71 .37 .55
2. Pretest immediate 7.02 5.66 — .41 .67 .48 .56 .62
3. Posttest immediate 29.19 12.62 — .33 .85 .39 .61
4. Pretest near 4.47 3.98 — .39 .49 .52
5. Posttest near 19.84 11.73 — .45 .67
6. Pretest far 11.19 9.96 — .61
7. Posttest far 25.25 16.66 —

Note. n ⫽ 366. All ps ⬍ .01. Immediate, near, and far refer to immediate-transfer, near-transfer, and
far-transfer problem solving, respectively.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

regression analyses, and Table 5 shows descriptive data on demo- – 0.35). Again, this main effect was not mediated by students’
graphics and performance variables for students with disabilities. initial status, F(4, 42) ⫽ 0.80, ns.5
Similarly, for far-transfer problem solving, the condition main
Pretreatment Differences Among Treatment Groups effect was significant, F(2, 21) ⫽ 6.65, p ⬍ .01. Fisher’s LSD post
hoc procedure again showed that contrast group improvement was
Groups were comparable prior to the study, as manifested by the less than that of the SBI group (ES ⫽ 1.55) and less than that of
lack of significant main effects for condition or interactions be- the SBI plus sorting group (ES ⫽ 2.29), but the improvement of
tween condition and students’ initial status on pretreatment the SBI and the SBI plus sorting groups was comparable (ES ⫽
immediate-, near-, or far-transfer problem-solving measures. The 0.65). Again, this main effect was not mediated by students’ initial
F(2, 21) values for the condition main effect on the three respec- status, F(4, 42) ⫽ 1.90, ns.6
tive measures were 0.97, 0.15, and 0.90. The F(4, 42) values for
the interaction effect on the three respective measures were 1.79,
0.62, and 1.60. Differential Schema Development as a Function of
As expected, at pretreatment, there were significant main effects Treatment
for students’ initial achievement status: For the three respective
measures, F(2, 42) values were 126.24, 75.48, and 18.62, all ps ⬍ Schema development (i.e., students’ ability to categorize novel
.001. LSD (using a critical p value of .016) follow-up tests showed problems into the four problem types) differed as a function of
that for immediate transfer, HI (M ⫽ 12.76, SD ⫽ 3.37) ⬎ AV treatment. The pattern of effects differed for the two schema-
(M ⫽ 6.77, SD ⫽ 1.65) ⬎ LO (M ⫽ 2.94, SD ⫽ 1.15); for near development scores. Problem-type schema development tapped
transfer, HI (M ⫽ 7.37, SD ⫽ 2.13) ⬎ AV (M ⫽ 5.16, SD ⫽ students’ capacity to categorize problems into the four problem
1.88) ⬎ LO (M ⫽ 1.94, SD ⫽ 3.64); for far transfer, HI (M ⫽ types (buying bags, shopping list, half, pictographs). For problem-
17.45, SD ⫽ 7.83) ⬎ AV (M ⫽ 9.86, SD ⫽ 5.80) ⫽ LO (M ⫽ type schema development, the significant main effect for condi-
8.08, SD ⫽ 3.58). These effects pertain across conditions; thus, tion, F(2, 21) ⫽ 26.69, p ⬍ .001, was not mediated by students’
they do not threaten the study’s validity.3 initial achievement status, F(4, 42) ⫽ 1.75, ns. Follow-ups to the
condition main effect indicated that both treatment groups ex-
ceeded the contrast group but were comparable with each other.
Differential Problem-Solving Learning as a Function of
For SBI versus contrast, the ES was 3.06; for SBI plus sorting
Treatment versus contrast, 2.99; for SBI plus sorting versus SBI, 0.32.7
Improvement on problem solving varied as a function of con- Superficial-feature schema development tapped students’ ability
dition, favoring both SBI groups over the contrast group. On to categorize superficial features of problems (i.e., as different
immediate-transfer problem solving, Fisher’s LSD post hoc pro-
cedure for the significant condition main effect, F(2, 21) ⫽ 77.05, 3
p ⬍ .001, showed that the contrast group’s improvement was less On the pretreatment problem-solving scores, the main effects for initial
achievement status are relevant. On the remaining scores, however, these
than that of the SBI group (ES ⫽ 3.17) and less than that of the SBI
main effects are not relevant to the study questions (because they pertain
plus sorting group (ES ⫽ 3.47), but the improvement of the SBI across conditions). Consequently, we report them as footnotes.
and the SBI plus sorting groups was comparable (ES ⫽ 0.45). This 4
On immediate transfer, the main effect for initial status was not
main effect was not mediated by students’ initial achievement significant, F(2, 42) ⫽ 0.47.
status, F(4, 42) ⫽ 1.95, ns.4 5
On near transfer, the main effect for initial status was significant, F(2,
For near-transfer problem solving, results were similar, again 42) ⫽ 9.23, p ⬍ .001; LO comparable with AV but less than HI.
favoring both SBI groups over the contrast group. Fisher’s LSD 6
On far transfer, the main effect for initial status was significant, F(2,
post hoc procedure for the significant condition main effect, F(2, 42) ⫽ 30.31; LO comparable with AV but less than HI.
21) ⫽ 71.48, p ⬍ .001, showed that contrast group improvement 7
On problem-type schema development, the main effect for initial status
was less than that of the SBI group (ES ⫽ 3.65) and less than that was significant, F(2, 42) ⫽ 11.47, p ⬍ .001; LO and AV student scores
of the SBI plus sorting group (ES ⫽ 3.10), but the improvement of were lower than those of HI students, but LO and AV student scores were
the SBI and SBI plus sorting groups was comparable (ES ⫽ comparable.
SCHEMA DEVELOPMENT AND MATH PROBLEM SOLVING 643

Table 4
Regressions Among Pretreatment Problem Solving, Schema Development, and Posttreatment
Problem Solving

Predicted and predictor R R2 Adj. R2 SEE ⌬R2 ⌬F a

Posttest immediate
Schema development (first) .693 .481 .479 9.115 .481 336.92
Pretest immediate (second) .714 .510 .507 8.869 .029 21.45
Pretest immediate (first) .415 .172 .170 11.509 .172 75.64
Schema development (second) .714 .510 .507 8.869 .338 249.93
Posttest near
Schema development (first) .706 .498 .497 8.299 .498 362.23
Pretest near (second) .738 .545 .543 7.911 .047 37.65
Pretest near (first) .388 .151 .149 10.794 .151 64.84
Schema development (second) .738 .545 .543 7.911 .394 315.53
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Posttest far
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Schema development (first) .550 .302 .300 13.940 .302 157.25


Pretest far (second) .697 .485 .482 11.991 .183 128.64
Pretest far (first) .609 .371 .369 13.239 .371 213.86
Schema development (second) .697 .485 .482 11.991 .114 80.49

Note. n ⫽ 366. Immediate, near, and far refer to immediate-transfer, near-transfer, and far-transfer problem
solving, respectively. First and second refer to predictor order. Adj. ⫽ adjusted; SEE ⫽ standard error of
estimate.
a
dfs ⫽ 1, 364 and 1, 363 for the first and second predictors, respectively; p ⬍ .01.

format, different key vocabulary, different or additional question, What About Students With Disabilities?
or different scope) that render them novel. For superficial-feature
schema development, the significant condition main effect, F(2, For students with disabilities, treatment groups were comparable
21) ⫽ 46.66, p ⬍ .001, was not mediated by initial treatment on demographics (see Table 5) and on pretreatment problem-
status, F(4, 42) ⫽ 2.01, ns. Follow-ups to the condition effect solving scores, F(2, 31) ⫽ 0.39, for immediate transfer, 1.03 for
indicated that the SBI plus sorting group exceeded the SBI group, near transfer, and 1.29 for far transfer (see Table 5).
which in turn exceeded the contrast group. For SBI versus contrast, On problem-solving improvement, the effect for condition was
the ES was 2.88; for SBI plus sorting versus contrast, 3.99; for SBI significant on immediate-transfer problem solving, F(2, 31) ⫽
plus sorting versus SBI, 1.24.8 12.25, p ⬍ .001 (contrast ⬍ SBI ⫽ SBI plus sorting); on near-
transfer problem solving, F(2, 31) ⫽ 8.57, p ⬍ .001 (contrast ⬍
The Contribution of Schema Development to SBI ⫽ SBI plus sorting); and on far-transfer problem solving, F(2,
31) ⫽ 4.84 (contrast ⬍ SBI plus sorting, with SBI equal to both
Problem-Solving Learning
other groups).
For each posttreatment problem-solving measure (immediate, For problem-type schema development, the effect for condition
near, far transfer), we conducted a pair of regressions, first entering was not significant, F(2, 31) ⫽ 1.39. However, for superficial-
the pretreatment problem-solving score to see how much addi- feature schema development, the effect for condition was signifi-
tional variance was explained by schema development; next en- cant, F(2, 31) ⫽ 4.02, p ⬍ .05, with Fisher’s LSD post hoc
tering schema development to see how much additional variance procedure showing that the SBI plus sorting group outperformed
was explained by the pretreatment problem-solving score. Results, the contrast group.
along with the zero-order correlations, are shown in Tables 3 For problem-type schema development and superficial-feature
and 4. schema development and for immediate-, near-, and far-transfer
As shown, in predicting posttreatment immediate-transfer problem-solving measures, respectively, the ESs for SBI versus
problem-solving, pretreatment problem solving and schema devel- contrast were 0.27, 0.20, 1.55, 1.78, and 0.87; for SBI plus sorting
opment jointly accounted for 14.0% of variance (i.e., total adjusted versus contrast, 1.01, 1.13, 1.79, 1.96, and 1.04; and for SBI versus
R2 minus [unique ⌬R2 for schema development plus unique ⌬R2 SBI plus sorting, 0.78, 0.95, 0.55, 0.38, and 0.39.
for pretreatment immediate-transfer problem-solving]), but
schema development accounted for substantially more unique vari-
ance than did initial problem-solving competence (33.8% vs. Discussion
2.9%). For near-transfer problem solving, results were similar:
One major purpose of this study was to assess the effects of SBI
Pretreatment score and schema development jointly accounted for
in promoting mathematical problem solving among young students
10.2% of variance, but schema development again accounted for
of varying initial achievement status, including those with disabil-
more unique variance (39.4% vs. 4.7%). For far-transfer problem
solving, pretreatment score and schema development jointly ac-
counted for 18.5% of variance, with schema development account- 8
On superficial-feature schema development, the main effect for initial
ing for 11.4% unique variance and with the pretreatment score status was significant, F(2, 42) ⫽ 4.73, p ⬍ .05, but follow-ups failed to
accounting for 18.3% unique variance. detect differences among the student groups.
644 FUCHS ET AL.

Table 5
Demographics and Performance for Students With Disabilities by Condition

SBI plus sorting


Contrast (n ⫽ 12) SBI (n ⫽ 12) (n ⫽ 10)

Variable n M SD n M SD n M SD

No. of male students 10 10 7


Subsidized lunch 5 7 6
Race
African American 8 8 5
European American 4 3 4
Hispanic 0 1 1
English language learner 0 1 0
Disability
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Learning disability 6 7 7
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Mental retardation 0 1 0
Behavior disorder 0 0 1
Speech disorder 4 4 2
Attention-deficit/hyperactivity disorder 2 0 0
Schema development
Problem-type schemas 8.45 4.00 9.39 3.72 11.07 3.51
Superficial-feature schemas 2.80 1.00 3.08 0.94 4.10 1.39
Problem solving
Immediate transfer
Pretest 2.92 3.00 2.17 3.33 3.40 3.72
Posttest 4.58 4.52 9.67 5.14 13.20 6.29
Improve 1.67 3.75 7.50 3.37 9.80 4.96
Near transfer
Pretest 3.21 3.19 1.68 1.94 2.55 2.48
Posttest 2.94 2.94 6.06 3.23 8.30 5.93
Improve ⫺0.27 2.45 4.37 3.52 5.75 4.80
Far transfer
Pretest 4.71 4.90 3.17 4.78 5.55 6.92
Posttest 8.42 8.08 13.00 9.27 20.10 15.45
Improve 3.71 5.63 9.83 8.41 14.55 10.35

Note. SBI ⫽ schema-based instruction; LO ⫽ low-performing students; AV ⫽ average-performing students;


HI ⫽ high-performing students; Improve ⫽ posttest score minus pretest score.

ities. Our approach to SBI was innovative in that it attempted to Documenting the strength of the SBI treatment for third-grade
broaden problem-type schemas by inducing a supplementary form students across varying points on the achievement continuum on
of schemas—for superficial problem features. The hope was that this far-transfer measure is notable for two reasons. First, previous
as students learned to categorize problems according to the super- work illustrates the difficulty of effecting mathematical problem
ficial features that render problems novel, students would concur- solving for primary-grade children (e.g., Chen, 1999), especially
rently improve at identifying problem-type schemas and the solu- those with initial low achievement (e.g., Cooper & Sweller, 1987;
tions required to solve novel problems. Woodward & Baxter, 1997), probably because acquired knowl-
Results demonstrated that this approach to SBI improved math- edge for young children is likely to be closely tied to specific
ematical problem solving among third-grade students of varying situations (e.g., Brown & Campione, 1984; Rozin, 1976). Second,
initial achievement status. On immediate-, near-, and far-transfer in contrast to earlier related work also conducted with third graders
problem-solving measures, students in the SBI treatment reliably (Fuchs, Fuchs, Prentice, Burch, Hamlett, Owen, et al., 2003;
outgrew those in the contrast group. Across students with differing Fuchs, Fuchs, Prentice, Burch, Hamlett, Owen, & Schroeter,
levels of initial problem-solving competence, ESs were large: 3.17 2003), the present study controlled for the potential effects of
on immediate transfer, 3.65 on near transfer, and 1.55 on far teaching general problem-solving strategies (making sure answers
transfer. The effect sizes exceeding 3 SDs for immediate and near make sense; lining up numbers from text to perform math opera-
transfer are in part due to the alignment between the treatment and tions; checking computation; and labeling work with words, mon-
those two measures, although it is important to note that the etary signs, and mathematical symbols) by incorporating that con-
problem types were borrowed from the standard curriculum, which tent into all three conditions. Therefore, findings isolate the effects
the contrast teachers taught. These very large ESs demonstrate that of SBI from the conceptually unrelated, more general problem-
SBI with explicit teacher-guided instruction and with sufficient solving strategies.
duration can impact students’ mathematical problem solving in Of course, present findings also extend the literature on schema
dramatic fashion. Of course, the strong effects on far transfer of induction and mathematical problem solving in another important
1.55 SDs are particularly noteworthy because this measure in no way. A second major purpose of the present study was to assess
way resembled any task used for instruction; moreover, by incor- whether SBI actually induces schema development concurrent
porating multiple sources of novelty, it represented a contextually with mathematical problem solving in these relatively young chil-
realistic form of mathematical problem solving. dren. Results showed that students in both SBI groups did, in fact,
SCHEMA DEVELOPMENT AND MATH PROBLEM SOLVING 645

develop stronger schemas, relative to the contrast group, and the and thereby enhance children’s capacity to recognize and solve
resulting ESs were large: 2.88 –3.06 for the SBI group and 2.99 – novel problems for which solutions had been taught.
3.99 for the SBI plus sorting group. Interestingly, both SBI groups, with and without the sorting
Moreover, results of the planned regression analyses docu- practice, developed reliably and dramatically stronger schemas for
mented the contribution of schema induction to problem-solving categorizing the four problem types, compared with the contrast
competence at treatment’s end. In predicting posttreatment group (where the curriculum incorporated the same problem
immediate-transfer problem-solving performance, schema devel- types). These ESs were large, hovering around 3 SDs. At the same
opment accounted for substantially more unique variance than did time, the sorting activity provided no added value in inducing
initial problem-solving competence (33.8% vs. 2.9%). Results problem-type schemas— despite the fact that the sorting practice
were similar for the near-transfer problem-solving outcome, and task directly mirrored the schema-development measure. And, in
schema development again accounted for similarly more unique corresponding fashion, we saw no reliable advantage to SBI plus
variance (39.4% vs. 4.7%). At the same time, the proportion of sorting over SBI on the immediate- and near-transfer problem-
unique variance accounted for by schema development on the solving measures, where again ESs approximated 3 SDs over the
far-transfer problem-solving measure, although smaller, was nota- contrast group for both SBI conditions, underscoring the need for
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

ble (11.4%), with pretreatment score explaining a similar propor- additional research with larger samples.
tion of unique variance (18.3%). The weaker contribution of We did, nevertheless, see some evidence to suggest potential for
schema development on far transfer is understandable given that the sorting activity. On the superficial-feature schema-
this outcome task introduced multiple sources of novelty, includ- development measure, students in the SBI plus sorting condition
ing simultaneous inclusion of all four problem types, simultaneous identified the kinds of superficial problem features that render
manipulation of all four superficial problem features, inclusion of problems novel with greater success compared with the SBI (with-
irrelevant text and numbers, incorporation of six additional essen- out sorting practice) group, which in turn exceeded the contrast
tial skills from the district curriculum, and formatting that resem- group. These ESs were large (3.99 for SBI plus sorting over
bled a commercial test. contrast, 2.88 for SBI over contrast, and 1.24 for SBI plus sorting
Although effects on the far-transfer task were more muted, over SBI). So, we looked to the far-transfer problem-solving
findings clearly document, across all three measures, the contri- measure for additional evidence that the SBI plus sorting group’s
bution of schema induction to problem-solving competence at superior superficial-feature schema development may have en-
treatment’s end. These results extend the work of Chen (1999), hanced problem-solving performance. However, both SBI groups
who showed that exposure to examples enhanced schema devel- outperformed the contrast group on the far-transfer problem-
opment and problem solving in a single session for 10 –12-year- solving measure, without a statistically significant difference be-
olds but not for 8 –9-year-olds. Whereas Chen’s 8 –9-year-olds tween the two experimental conditions. Of course, the effect size
failed to build schemas or improve problem solving in response to of 0.65 SDs on the far-transfer problem-solving measure did favor
a brief, inductive exposure to examples that varied structural the SBI plus sorting condition.
features, we used more explicit instruction of longer duration to And what about students with disabilities, a group of children
effect schema enhancement concurrent with mathematical problem who receive most of their instruction in regular classrooms (U.S.
solving with this younger population, including children with Department of Education, 1999) and for whom transfer is ex-
initially low achievement and those with disabilities. Of course, tremely difficult to effect (e.g., White, 1984)? No advantage ac-
the somewhat weaker demonstration of the contribution of schema crued as a function of experimental treatment for problem-type
development to performance on the far-transfer measure suggests schema development, where effects were not statistically signifi-
that additional work is required to promote even broader schemas cant (although the ES comparing the SBI plus sorting and the
in the service of mathematical problem solving. The link estab- contrast groups was an encouraging 1.01). By contrast, superficial-
lished in this study between SBI and the actual development of feature schema development was reliably stronger for the SBI plus
schemas concurrent with mathematical problem-solving growth sorting treatment over the contrast group, with an ES of 1.13.
for young children of varying achievement status provides the Interestingly, although not statistically significant, the ES compar-
basis for extending the use of SBI with a larger repertoire of ing the two SBI groups (0.95) suggests that for this lowest per-
problem-type schemas, via a greater variety of superficial-feature forming group of students, the sorting practice may have added
schemas and through other related schema-building activities. value, at least for inducing superficial-feature schemas. In accord
Toward the goal of exploring related schema-building activities, with this possibility, it is interesting to note that among students
the third, and secondary, purpose of this study was to assess the with disabilities, only those in the SBI plus sorting condition
added value of providing third graders with practice in sorting reliably exceeded the contrast group on far-transfer problem solv-
problems into their problem types and into the superficial problem ing, with an ES of 1.04. Consequently, results suggest that for
features. We were interested in whether this practice, which was students with very low achievement profiles, such as students with
mediated by teacher explanations, might result in superior schema disabilities, the sorting practice does carry the added value neces-
development and problem solving. This possibility has been raised sary to induce schemas and promote problem-solving competence.
in previous work (Schwartz & Bransford, 1998) but has not been This, along with the full sample ES of 0.65 favoring the SBI plus
tested with children. So, in the present study, we included a sorting group over SBI on the far-transfer problem-solving mea-
condition that supplemented SBI with sorting practice that in- sure, supports continued study of practice in sorting problems into
cluded debriefing on the structural variables by which problem problem-type schemas and into superficial-feature schemas. That
types and superficial problem features might be categorized. The work should be conducted with larger samples. Future research
hope was that this guided sorting practice would solidify students’ should also explore the performance and cognitive variables asso-
capacity to group problems requiring the same solution methods ciated with responsiveness to SBI treatment.
646 FUCHS ET AL.

In the meantime, present findings provide some guidance to Gholson, B., Eymard, L. A., Morgan, D., & Kamhi, A. G. (1987). Problem
third-grade teachers as they struggle with the challenging of pro- solving, recall, and isomorphic transfer among third-grade and sixth-
moting mathematical problem solving. The instructional principles grade children. Journal of Experimental Child Psychology, 43, 227–243.
incorporated in this study’s intervention can be applied to the Gick, M. L., & Holyoak, K. J. (1980). Analogical problem solving.
Cognitive Psychologist, 12, 306 –355.
range of problem types typically found in third-grade curricula,
Glass, G. V., McGaw, B., & Smith, M. L. (1981). Meta-analysis in social
and results show how SBI represents one approach for dramati-
research. Beverly Hills, CA: Sage.
cally enhancing mathematical problem-solving outcomes for third- Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis.
grade students across the achievement continuum. Orlando, FL: Academic Press.
Hinsley, D. A., Hayes, J. R., & Simon, H. A. (1977). From words to
References equations: Meaning and representation in algebra word problems. In
Bransford, J. D., & Schwartz, D. L. (1999). Rethinking transfer: A simple M. A. Just & P. A. Carpenter (Eds.), Cognitive processes in compre-
proposal with multiple implications. In A. Iran-Nejad & P. D. Pearson hension (pp. 89 –108). Hillsdale, NJ: Erlbaum.
(Eds.), Review of research in education (pp. 61–100). Washington, DC: Jitendra, A. K., Griffin, C. C., McGoey, K., Gardill, M. C., Bhat, P., &
Riley, T. (1998). Effects of mathematical word problem solving by
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

American Educational Research Association.


students at risk or with mild disabilities. Journal of Educational Re-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Brown, A. L., & Campione, J. C. (1984). Three faces of transfer: Impli-


cations for early competence, individual differences, and instruction. In search, 91, 345–355.
M. Lamb, A. Brown, & B. Rogoff (Eds.), Advances in developmental Kansas State Board of Education. (1991). Kansas Quality Performance
psychology (Vol. 3, pp. 143–192). Hillsdale, NJ: Erlbaum. Accreditation. Topeka, KS: Author.
Burton, G. M., & Maletsky, E. M. (1999). Math advantage. Orlando, FL: Mawer, R., & Sweller, J. (1985). What do students learn while solving
Harcourt, Brace, Jovanovich. mathematics problems? Journal of Educational Psychology, 77, 272–
Chen, Z. (1999). Schema induction in children’s analogical problem solv- 284.
ing. Journal of Educational Psychology, 91, 703–715. Mayer, D. P. (1998). Do new teaching standards undermine performance
Chi, M. T. H., Feltovich, P. J., & Glaser, R. (1981). Categorization and on old tests? Educational Evaluation and Policy Analysis, 15, 1–16.
representation of physics problems by experts and novices. Cognitive Mayer, R. E. (1992). Thinking, problem solving, cognition (2nd ed.). New
Science, 5, 121–152. York: Freeman.
Cooper, G., & Sweller, J. (1987). Effects of schema acquisition and rule Mayer, R. E., Quilici, J. L., & Moreno, R. (1999). What is learned in an
automation on mathematical problem-solving transfer. Journal of Edu- after-school computer club? Journal of Educational Computing Re-
cational Psychology, 79, 347–362. search, 20, 223–235.
CTB/McGraw-Hill. (1997). TerraNova technical manual. Monterey, CA: Quilici, J. L., & Mayer, R. E. (1996). Role of examples in how students
Author. learn to categorize statistics word problems. Journal of Educational
Cummins, D. D. (1992). Role of analogical reasoning in the induction of Psychology, 88, 144 –161.
problem categories. Journal of Experimental Psychology: Learning, Rozin, P. (1976). The evolution of intelligence and access to the cognitive
Memory, and Cognition, 18, 1103–1124. unconscious. Progress in Psychobiology and Physical Psychology, 6,
Fuchs, L. S., Fuchs, D., Karns, K., Hamlett, C. L., Dutka, S., & Katzaroff, 245–280.
M. (2000). The importance of providing background information on the Schwartz, D. L., & Bransford, J. D. (1998). A time for telling. Cognition
structure and scoring of performance assessments. Applied Measurement and Instruction, 16, 475–522.
in Education, 13, 1–34. Seaman, M. A., Levin, J. R., & Serlin, R. C. (1991). New developments in
Fuchs, L. S., Fuchs, D., Karns, K., Hamlett, C. L., & Katzaroff, M. (1999). pairwise multiple comparisons: Some powerful and practical problems.
Mathematics performance assessment in the classroom: Effects on Psychological Bulletin, 110, 577–586.
teacher planning and student learning. American Educational Research Sweller, J., & Cooper, G. A. (1985). The use of worked examples as a
Journal, 36, 609 – 646. substitute for problem solving in learning algebra. Cognition and In-
Fuchs, L. S., Fuchs, D., Prentice, K., Burch, M., Hamlett, C. L., Owen, R., struction, 2, 59 – 89.
Hosp, M., & Janeck, D. (2003). Explicitly teaching for transfer: Effects U.S. Department of Education. (1999). To assure the free, appropriate
on third-grade students’ mathematical problem solving. Journal of Ed- public education of all children with disabilities: Twenty-first report to
ucational Psychology, 95, 293–304. Congress on the implementation of the Individuals With Disabilities
Fuchs, L. S., Fuchs, D., Prentice, K., Burch, M., Hamlett, C. L., Owen, R., Education Act. Washington, DC: Government Printing Office.
& Schroeter, K. (2003). Enhancing third-grade students’ mathematical White, O. R. (1984). Descriptive analysis of extant research literature
problem solving with self-regulated learning strategies. Journal of Ed- concerning skill generalization and the handicapped. In M. Boer (Ed.),
ucational Psychology, 95, 306 –315. Investigating the problem of skill generalization: Literature review (pp.
Gentner, D. (1988). Metaphor as structure mapping: The relational shift. 1–19). Seattle: University of Washington, Washington Research
Child Development, 59, 47–59. Organization.
Gentner, D. (1989). The mechanisms of analogical learning. In S. Vosni- Woodward, J., & Baxter, J. (1997). The effects of an innovative approach
adou & A. Ortony (Eds.), Similarity and analogical reasoning (pp. to mathematics on academically low-achieving students in inclusive
199 –241). Cambridge, England: Cambridge University Press. settings. Exceptional Children, 63, 373–388.
SCHEMA DEVELOPMENT AND MATH PROBLEM SOLVING 647

Appendix

Sample Problems for Step-Up Function Problem Schema

Source (Taught) Problem Different Vocabulary


Greg needs 16 balloons for his birthday party. Balloons come in bags of Francis is buying eggs for her dinner party. She needs 26 eggs for the
10. How many bags does Greg need? dishes she’s making. Eggs are sold in dozens. How many dozens does
Francis need?
Novel Problem Illustrating a Different Cover Story
Different Question
Harriet serves ice cream pops to her club for a treat. She invites 12
friends. Ice cream pops are sold in bags of 3. How many bags does Harriet Jose has $25 to spend on the ice hockey pucks he’s buying for the team.
need? He needs 7 pucks. Pucks are sold in bags of 3, and each bag costs $8. After
buying the pucks, how much money will Jose have?
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Novel Problems Illustrating Four More Demanding Superficial


Scope of the Problem
Problem Features
Frieda is on her way to the store that sells office supplies. She needs to
Different Format buy 15 pencils, 2 pencil sharpeners, and 4 notebooks. Pencils come in bags
of 7, and each bag costs $3. Each pencil sharpener costs $10, and each
The sign at the supermarket reads:
notebook costs $3. How much money will she spend?
!!!Come and get it! Four frozen pizzas in a bag for one low price!!!
You see the sign and decide to serve pizza for your next dinner party.
You figure you’ll need 10 pizzas. How many bags do you need? (Select Received June 13, 2003
one answer.) Revision received October 21, 2003
2 3 4 5 Accepted April 29, 2004 䡲

You might also like