Professional Documents
Culture Documents
Equivalence in Symbolic and Nonsymbolic Contexts: Benefits of Solving Problems With Manipulatives
Equivalence in Symbolic and Nonsymbolic Contexts: Benefits of Solving Problems With Manipulatives
Equivalence in Symbolic and Nonsymbolic Contexts: Benefits of Solving Problems With Manipulatives
problems suggests that children fail to map their understanding of equivalence onto problems presented
with the symbols of arithmetic. For Study 2 (N ⫽ 32), we implemented a within-subject design to assess
whether experience with nonsymbolic problems would facilitate performance on symbolic problems.
This hypothesis was confirmed. Exposure to nonsymbolic problems may have enabled children to map
their successful concepts and strategies to symbolic equivalence problems.
Equivalence, in its broadest sense, implies a relation in which Not only is misunderstanding the equal sign prevalent, it may be
one item can be replaced by another (Kieran, 1981). In formal difficult to correct (Alibali, 1999; McNeil & Alibali, 2000; Seo &
arithmetic, equivalence involves recognizing that two sides of an Ginsburg, 2003). Frequent and intense exposure to canonical prob-
equation must have the same quantity (Alibali, 1999). Children’s lems (e.g., a ⫹ b ⫽ c) in both textbooks and classroom instruction
understanding of equivalence in arithmetic is generally tested appear to enforce the operator notion that the equal sign means “the
using worksheets with problems such as a ⫹ b ⫹ c ⫽ __ ⫹ c, answer comes next” or “do something,” rather than the relational
whereby children have to make both sides of the equal sign view that both sides are equivalent (Alibali, 1999; Baroody &
quantitatively identical. Despite success on canonical arithmetic Ginsburg, 1983; Behr, Erlwanger, & Nichols, 1980; Carpenter,
problems (e.g., a ⫹ b ⫹ c ⫽ __) of similar magnitude, perfor- Franke, & Levi, 2003; Kieran, 1981; McNeil & Alibali, 2004,
mance on equivalence problems is generally low. For example, 2005b; Perry, 1991; Perry et al., 1988; Renwick, 1932; Sáenz-
Perry (1985, as cited in Perry, Church, & Goldin-Meadow, 1988) Ludlow & Walgamuth, 1998; Seo & Ginsburg, 2003). An operator
found that approximately 88% of fourth- and fifth-grade students view of the equal sign makes solving for an unknown difficult, can
failed problems of this form. Both error patterns and children’s make noncanonical equations (e.g., __ ⫽ 2 ⫹ 3) look incorrect, and
verbal explanations indicate a misunderstanding of the symbol for may impede comprehension of more advanced mathematics, such as
equivalence (the equal sign). This misunderstanding could hinder algebra (Carpenter et al., 2003; Kieran, 1981; Knuth et al., 2005;
success in higher mathematics, yet it often goes undetected (Ki- Renwick, 1932; Seo & Ginsburg, 2003). Children’s exposure to
eran, 1981; Knuth, Alibali, McNeil, Weinberg, & Stephens, 2005; canonical examples may also lead to the formation of addition
McNeil & Alibali, 2005a). schemas that include the belief that (a) addends are always to the
left of the equal sign and the solutions are on the right, (b)
equations always involve operating on all terms, and (c) the equal
sign means “the answer comes next” (McNeil & Alibali, 2002,
Jody Sherman and Jeffrey Bisanz, Department of Psychology, Univer- 2004). Although applying addition schemas to canonical problems
sity of Alberta, Edmonton, Alberta, Canada. Jody Sherman is now at the
may yield correct solutions, these schemas limit children’s ability
Graduate School of Education, University of California, Berkeley.
Support for this research was provided in part by a scholarship from the to solve equations that require relational reasoning, such as equiv-
Social Sciences and Humanities Research Council of Canada to Jody alence problems and algebra.
Sherman and a grant from the Natural Sciences and Engineering Research Several researchers have attempted to improve children’s suc-
Council of Canada to Jeffrey Bisanz. We are grateful to Elena Nicoladis, cess on equivalence problems and promote a relational, rather than
Connie Varnhagen, Nancy Galambos, Jacqueline Leighton, and Martha operator, view of the equal sign. For example, Rittle-Johnson
Alibali for their helpful comments and to Elizabeth Jacob and Ozlem (2006) assigned third- and fourth-grade students to one of four
Cankaya for their assistance in data collection and entry. Finally, we conditions on the basis of crossing (a) instruction of the add–
especially appreciate the helpful cooperation of students, teachers, and staff
subtract procedure for equivalence problems versus children hav-
at both the Black Gold and St. Thomas Aquinas Roman Catholic school
districts. ing to invent the procedure, and (b) self-explanation versus no
Correspondence concerning this article should be addressed to Jody explanation. Children in the self-explanation condition saw fic-
Sherman, 4511 Tolman Hall, Graduate School of Education, University of tional students’ responses (both correct and incorrect) to equiva-
California, Berkeley, CA 94720-1670. E-mail: jodysherman@berkeley.edu lence problems and were asked to explain orally both how the
88
EQUIVALENCE PROBLEMS WITHOUT SYMBOLS 89
students solved the problem and why they were right or wrong. performance cannot be assessed, or several types of problems are
Both self-explanation and instruction helped children learn and used but performance across all types is collapsed. For example,
maintain a correct procedure for solving equivalence problems problems such as a ⫹ b ⫹ c ⫽ d ⫹ __ and a ⫹ b ⫹ c ⫽ ___ ⫹
(e.g., a ⫹ b ⫹ c ⫽ __ ⫹ d), as accuracy on equivalence problems c have been included in several studies (Alibali, 1999; McNeil &
improved significantly, but self-explanation had the additional Alibali, 2000, 2005a; Perry, 1991; Perry et al., 1988; Rittle-
benefit of promoting transfer to novel equivalence problems (e.g., Johnson & Alibali, 1999), but performance across these different
a ⫹ b ⫹ __ ⫽ c ⫹ d). Further research is needed to determine problem types is often collapsed and presented as a composite
whether improvements are long lasting and generalize to other score. In a notable exception, McNeil and Alibali (2004) deliber-
algebra-type problems. Some researchers have found limited suc- ately examined performance on two different problem types and
cess on equivalence problems even after various interventions or found that Grade 4 children have fewer difficulties encoding
instruction (e.g., Alibali, 1999). problems in the form a ⫹ b ⫹ c ⫽ __ ⫹ c than in the form a ⫹
Although research to date has shown that children do poorly on b ⫹ c ⫽ a ⫹ __. The performance discrepancy was likely due to
equivalence problems, the fact that tests of equivalence typically the similarity between the structure of the latter problem type and
require children to solve a limited range of equivalence problems canonical arithmetic problems. Children would often reconstruct
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
intervention that teachers could implement in the classroom to help month), 22 of whom were male and 26 were female. Four addi-
students learn to solve mathematical equivalence problems. tional children were tested but not included because of experi-
menter error (three children) or inadequate comprehension of
Study 1 English (one child). All children were from schools in a small,
largely middle-class town in the Canadian province of Alberta.
In Study 1, we presented nonsymbolic equivalence problems— Self-reported levels of education, economic vitality, and living
equivalence problems without conventional symbols—to Grade 2 standards by town residents were mostly at or above comparison
children. The nonsymbolic context, in this case, refers to the use of data (see Leduc Genuine Wealth Accounting Project, 2005, for
manipulatives to iconically represent the quantities of the addends details).
in the equation. If children can solve these problems, then their Problem sets were manipulated in terms of presentation format
poor performance is likely not due to a fundamental deficit in (Figure 1). Children were assigned unsystematically to either the
understanding the concept of equivalence in arithmetic. Rather, the nonsymbolic or symbolic condition (11 boys and 13 girls in each
well-documented poor performance by school children on equiv- condition) with the constraint that children within each of the six
alence problems is likely due to a limited understanding of the classrooms were distributed across conditions as evenly as possible.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Nonsymbolic
Symbolic (Study 1)
5 + 2 = 4 + ____
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.
Symbolic (Study 2)
up to two additional problems were administered, until the child commutativity (a ⫹ b ⫽ b ⫹ __), four part–whole (a ⫹ b ⫽
answered correctly, prior to the equivalence problem. c ⫹ __), and four combination (a ⫹ b ⫹ c ⫽ a ⫹ __) problems.
Children assigned to the nonsymbolic condition were presented Ten problems, two of each type, were presented in each session.
with the same problems but with different materials (Figure 1). A Within each set of 10, a, b, c, and correct solutions never appeared
6-cm-high piece of blue cardboard, folded to look like a tent, more than twice in a sequence; in half the problems, a ⫽ b ⫽ c and
served to separate the two sides of the equations. On the child’s left a ⬍ b. Two such problem sets were created and presented in
side of the tent was a piece of orange construction paper and on the forward and reverse orders, resulting in four problem orders.
right was purple construction paper, to make the two sides distinct. Problem order was counterbalanced as evenly as possible within
Wooden cylinder blocks were placed in opaque plastic bins ap- condition and gender. Problems were constructed so that magni-
proximately 3 cm high to represent each term of the arithmetic tude, as indexed by a ⫹ b, was balanced across the five types of
expressions. A screen was erected so the child could not see the problems. Addends ranged from 1 to 6. Instructions identical to
materials, and the experimenter arranged the bins and blocks to those used for practice problems were provided for the first ex-
represent 1 ⫹ 1 ⫽ __. The experimenter solved the first problem ample of each of the five problem types for both sessions. In this
by placing two blocks in the bin, and children were asked if the way, children heard the same instructions for each new type of
answer was correct. Children’s responses were recorded, and feed- problem. Feedback was not provided, performance was scored as
back was provided. Following confirmation that children under- correct or incorrect, and children’s justifications were recorded on
stood the task, the remaining problems were presented. Feedback all problems.
was provided, and the screen was replaced between trials. For the symbolic format, problems were presented using a
For the sample equivalence problem (1 ⫹ 1 ⫽ 1 ⫹ __), worksheet on which all problems were covered except for the one
children in the symbolic condition were instructed to “put a being solved. For the nonsymbolic format, problems were pre-
number on the line so that when you put together these on this sented as described for the practice arithmetic problems and the
side of the equal sign (experimenter pointed to numbers on left), screen was placed so that children could not watch the next
you’ll have the same number as when you put together these on problem being arranged.
this side of the equal sign (experimenter pointed to number and Children’s justifications were coded first according to criteria
blank line on right)”. Children in the nonsymbolic condition described in Appendix B and then reclassified as either equiva-
were instructed to “put blocks in the empty bin so that when you lence or nonequivalence for parsimony and consistency with pre-
put together these (experimenter pointed to blocks on left) on vious research (see Alibali, 1999; Perry et al., 1988). This dichot-
this side of the blue tent, you’ll have the same number as when omy is used throughout with one exception: Identical numerosity
you put together these (experimenter pointed to blocks and justifications are discussed separately in some comparisons to
empty bin) on this side of the blue tent.” The experimenter then highlight instances in which children made explicit reference to
prompted the children to justify their responses by asking, making both sides of the expression quantitatively equivalent. It is
“Why is your answer right?” Feedback was given. important to note that justifications do not necessarily denote
Arithmetic problems. Each child solved 20 problems, includ- success or failure on a problem. That is, children may answer a
ing four addition (a ⫹ b ⫽ __), four identity (a ⫹ b ⫽ a ⫹ __), four problem correctly but say that they have solved the problem by
92 SHERMAN AND BISANZ
making both sides look the same (identity justification) or that they lems in school. Problem type interacted with presentation format,
solved it by adding the terms together (arithmetic justification), F(4, 176) ⫽ 14.76, p ⬍ .001, 2p ⫽ .25, such that children in the
which would yield a correct solution on addition problems. A nonsymbolic group had higher accuracy than children in the sym-
stringent criterion for judgments of identical numerosity (Appen- bolic group for every problem type, Fs(1, 220) ⬎ 26.20, ps ⬍ .001,
dix B) was useful for identifying situations in which children except addition.
specifically stated that they made both sides have the same Children differed widely not only in accuracy but also in the
amount. Justifications on all problems for 10 children were coded frequency with which they gave equivalence justifications. A 2
by two researchers as equivalence, nonequivalence, and other with (condition) ⫻ 2 (gender) ⫻ 5 (problem type) ANOVA, with
an interrater reliability of 89.4%. Discrepancies were resolved repeated measures on the last variable, revealed main effects of
through discussion. condition, F(1, 44) ⫽ 39.50, p ⬍ .001, 2p ⫽ .47, and problem type,
F(4, 176) ⫽ 17.53, p ⬍ .001, 2p ⫽ .29, but no interactions. Data
Results are presented in Figure 3. Children in the nonsymbolic condition
gave equivalence justifications on nearly three quarters of all
Performance in Symbolic Versus Nonsymbolic Conditions
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
problems and three times more often than children in the symbolic
This document is copyrighted by the American Psychological Association or one of its allied publishers.
To determine whether performance relates to the presence of condition. Tests of simple effects revealed that the means for
symbols, we compared accuracy of Grade 2 children in the non- identity and commutativity problems were indistinguishable from
symbolic condition to that of Grade 2 children in the symbolic each other but higher than the means for part–whole and combi-
condition. A 2 (condition) ⫻ 2 (gender) ⫻ 5 (problem type) nation problems combined, F(1, 176) ⫽ 21.42, p ⬍ .001), and that
analysis of variance (ANOVA) with repeated measures on the last the latter two means also were similar to each other but higher than
variable was conducted. Tests of simple effects were used to explore the mean for addition problems, F(1, 176) ⫽ 74.16, p ⬍ .001).
main effects and interactions. Data are presented in Figure 2. Children Possibly the most interesting approach used by children in the
in the nonsymbolic condition solved twice as many problems symbolic condition was the carry-after-equal strategy, as indicated
correctly (M ⫽ 16.92, SD ⫽ 4.56, summed across all five condi- by their justifications. Children who used this approach stated that
tions) as children in the symbolic condition (M ⫽ 8.33, SD ⫽ their answer was correct because the number after the equal sign in
4.05), F(1, 44) ⫽ 48.05, p ⬍ .001, 2p ⫽ .52. This difference problems such as 1 ⫹ 3 ⫹ 2 ⫽ 1 ⫹ __ “had” to be the correct
cannot be attributed to a general discrepancy in arithmetic skill solution; therefore, all they had to do was copy the number, in this
because children in the two conditions did not differ in perfor- case 1, to the line. Their understanding that the equal sign means
mance on practice arithmetic problems, F(1, 44) ⬍ 1. that the solution comes next apparently hindered their ability to
Furthermore, performance differed according to problem type, think of the expression in terms of relational equivalence. Thus,
F(4, 176) ⫽ 29.09, p ⬍ .001, 2p ⫽ .40, with the highest scores for children in the nonsymbolic condition appeared to solve by delib-
addition problems (M ⫽ 3.67, SD ⫽ .72) and the lowest scores for erately making both sides equal, whereas children in the symbolic
part–whole problems (M ⫽ 1.73, SD ⫽ 1.72). Differences in condition responded with a variety of justifications including
accuracy were negligible between identity and commutativity carry, arithmetic, and carry-after-equal. Clearly, children in the
problems and also between part–whole and combination problems, two conditions differed not only in accuracy but also in how they
but the first pair differed from the second, F(1, 176) ⫽ 29.53, p ⬍ reportedly solved the problems.
.001. Performance on addition problems exceeded that of all other To identify individual children who used relational equivalence
problem types, Fs(1, 176) ⬎ 26.98, ps ⬍ .001. This high level of consistently, we used accuracy and justifications together for part–
success on the addition problems, compared with the other prob- whole problems. A joint criterion for this purpose was essential
lem types, is not surprising considering the amount of experience because interpretations based on accuracy or justifications alone
children normally have with symbolic, canonical addition prob- could be misleading. For example, accuracy may have been de-
3.5
3
Mean Performance
2.5
nonsymbolic
2
symbolic
1.5
0.5
0
identity commutativity part-whole combination addition
Problem Type
Figure 2. Mean performance (with standard error bars) across all five problem types by Grade 2 children in
the nonsymbolic and symbolic conditions (Study 1).
EQUIVALENCE PROBLEMS WITHOUT SYMBOLS 93
Justifications
2.5
nonsymbolic
2
symbolic
1.5
1
0.5
0
identity commutativity part-whole combination addition
Problem Type
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Figure 3. Mean number of equivalence justifications (with standard error bars) for each problem type by Grade
This document is copyrighted by the American Psychological Association or one of its allied publishers.
pressed due to arithmetic demands or faulty counting, or inflated Relations Among Problem Types
due to chance guessing. Similarly, justifications may not have
consistently reflected children’s procedures because of constraints Among children in the nonsymbolic condition, performance
on expressive language. Using both types of evidence in conjunc- across the four equivalence problem types was correlated strongly
tion and only using instances in which there was a clear match (Table 1). Children in the symbolic condition did not fit this
between accuracy and equivalence justifications increased our pattern, however. Instead, performance on the two qualitative
confidence in identifying children who understood and used equiv- problem types (identity and commutativity) was related, as was
alence. Performance on part–whole problems was used because performance on the two quantitative problem types (part–whole
and combination), but correlations between quantitative and non-
children could not solve these problems correctly by using identity
quantitative types were small ( ps ⬎ .25). Conceivably, there could
cues or by carrying terms from one part of the equation to another.
be a developmental progression in understanding equivalence such
To meet the joint criteria for consistent understanding and use of
that children first appreciate similarity across two sides of an
relational equivalence, children had to (a) solve all four part–whole
expression in a more qualitative manner by detecting patterns and
problems correctly (an answer within one of the correct answer
later in a fully quantitative manner, whereby using arithmetic to
was considered correct) and (b) provide justifications consistent
ensure equivalence becomes a dominant strategy. If this hypothesis
with relational equivalence. Children who were correct but who
is correct, then we would expect some children to have poor
gave nonequivalence justifications, such as guessing, were not
performance on all problem types, some children to solve just the
included. This stringent criterion provided a conservative estimate
two qualitative problem types accurately, and some children to be
of the number of children who consistently solved the most diffi-
successful on all problem types. Other patterns, such as children
cult problems in a relational manner.
solving only the quantitative problems successfully, would not fit
Thirteen of 24 children in the nonsymbolic group solved all four
the progression.
part–whole problems correctly by making both sides numerically
We examined individual performance to determine whether
identical and gave equivalence justifications, whereas only 1 of 24
children in the symbolic condition fit one of the expected patterns
children in the symbolic group met this criterion 2(1, N ⫽ 48) ⫽
of performance. We used at least three correct of four problems on
11.12, p ⬍ .001. Thus, accuracy levels, justifications, and exam-
both qualitative problem types and at least three correct of four on
ination of performance on part–whole problems all suggest that
both of the quantitative problem types as criteria. Seventeen (71%)
children who solved the problems in the nonsymbolic format were
of the 24 children did poorly on both the qualitative and quantitative
far more likely to approach and solve the problems using relational
equivalence. Children in the symbolic group, in contrast, had low
levels of accuracy and appeared to try a variety of strategies when Table 1
solving the problems. Correlations for Performance Across the Five Problem Types by
When Grade 2 children in the symbolic condition also were Grade 2 Children in the Nonsymbolic (Above the Diagonal) and
asked to tell the experimenter what the equal sign meant during Symbolic (Below the Diagonal) Groups in Study 1
both sessions, 94% of their responses were consistent with an
operator interpretation. Only 1 child during one of the two sessions Problem types children
solved correctly 1 2 3 4 5
gave a response consistent with a relational view of the equal sign.
ⴱ ⴱ ⴱ
He also gave exclusively equivalence justifications during both 1. Identity — .843 .696 .865 .554ⴱ
sessions, yet his solutions on all problems except the addition 2. Commutativity .697ⴱ — .667ⴱ .677ⴱ .286
problems were consistent with an add-to-equal procedure. His 3. Part–whole .231 .186 — .757ⴱ .356
4. Combination .107 .234 .700ⴱ — .710ⴱ
equivalence justifications may have reflected his attempt to repeat 5. Addition ⫺.129 ⫺.126 ⫺.244 .162 —
the experimenter’s instructions rather than an effort to make both
ⴱ
sides of the expressions the same. p ⬍ .01.
94 SHERMAN AND BISANZ
problems, thus showing little evidence of understanding equivalence search. Children’s context-dependent performance fits with what
in any form. Six children (25%) solved at least three of four problems Lawler (1981) referred to as having distinct microworlds of un-
on both of the qualitative problem types yet did poorly on the derstanding, in which children reason very differently about the
quantitative problems. One child had perfect performance on the same task in one condition or context compared with another. In
quantitative problems and, as predicted, had perfect performance on his case study, Lawler noted that his young daughter’s accuracy on
the qualitative problems as well. None of the children fit other addition problems was significantly lower than when the same
patterns, such as doing well on the quantitative problems only. Taken problems were presented with coins. That is, the child’s experience
together, children’s performance appears to fit one of three patterns: with money made this “money world” of addition more available
they understand equivalence and can demonstrate understanding for solving arithmetic problems than experience with standard,
across a variety of problems; they have a less sophisticated under- paper-and-pencil addition problems. Whether children can actually
standing and can only solve problems that look similar on both sides benefit from experiencing problems in the nonsymbolic condition
of the equal sign; or they do not understand at all. and transfer their successful strategies to the symbolic context
remains to be tested. In Study 2, we attempted to address this
question using similar methodology but a within-subjects design.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Discussion
This document is copyrighted by the American Psychological Association or one of its allied publishers.
than other strategies in both sessions, whereas children who one of the practice problems but for none of the equivalence
began with symbolic problems would tend to use operator problems in the problem set. In the problem set, we determined the
strategies in the symbolic session and the relational strategy in order in which the choice cards were presented from left to right
the nonsymbolic session. In another modification, we chose using a Latin square for the first and second half (8 problems each)
problems that were most challenging for Grade 2 children of the problem set. Order of choice cards was never the same for
(part–whole and combination) rather than comparing across all two questions in a row. Half the problems were part–whole types
types. These two problem types are frequently used to assess and half were combination types, and no more than 2 problems of
children’s performance on equivalence problems. a single type were ever presented in a row.
If children chose the same type of solution option (e.g., add-all)
Method on at least 13 of the 16 equivalence problems (approximately
80%), they were classified as consistently solving in that manner.
Participants and Design The binomial probability of randomly selecting the same type of
solution option on 13 of 16 problems—with an alpha of .05 and the
Participants consisted of 32 second-grade students, including 16
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
much higher accuracy on symbolic problems than children who left bar). Of the 16 children in this group, 5 consistently used a
solved symbolic problems first, F(1, 56) ⫽ 16.96, p ⬍ .01. Thus relational strategy, 3 used an operator strategy, 3 tried to solve by
solving problems in the nonsymbolic condition facilitated perfor- making patterns or matching terms on both sides of the equations,
mance on subsequent symbolic problems even though the sessions and 5 used a variety of approaches or solved part–whole versus
were approximately a week apart and experimenters did not deliber- combination problems differently (i.e., part–whole problems
ately or explicitly draw children’s attention to the similarity between solved with an add-to-equal strategy, combination problems solved
the two “tasks” across sessions. Accuracy did not differ between relationally). These same children presented a very different pat-
part–whole and combination problems, and there were no other sig- tern of strategy use when solving the nonsymbolic problems a
nificant main effects or interactions. Scores on the six arithmetic week later (Figure 5, right bar), with 12 of the children using a
problems were quite high (M ⫽ 94.8%, SD ⫽ 18.7% in the symbolic relational strategy consistently.
condition; M ⫽ 95.8%, SD ⫽ 10.3% in the nonsymbolic condition) Children in the nonsymbolic/symbolic group could be character-
and are not discussed further. ized as using one of three strategies to solve the nonsymbolic prob-
The probability of guessing correctly was considered to be 20% lems (Figure 6, right bar), with the majority solving the problems
because five solution options were available on every trial (add-all, relationally. When solving symbolic problems a week later, children
add-to-equal, relational, small-number, and other). For children in the could be classified as using one of five strategies (Figure 6, left bar),
symbolic/nonsymbolic group, performance on symbolic problems (M ⫽ but the majority of children (11 of 16) still solved the problems
36.5%, SD ⫽ 42.3%) was particularly low. A 95% confidence interval relationally. Relational justifications almost never occurred in the
indicated that the true mean was between 15.9% and 57.9%, a range that absence of relational solution choices (⬍2% of problems).
includes the observed accuracy level that could be expected with random
guessing. The other three means (children in symbolic/nonsymbolic Verbal Interpretations of the Equal Sign
group on nonsymbolic problems, and children in nonsymbolic/symbolic
group on both symbolic and nonsymbolic problems) were higher than None of the children in either group gave a relational verbal
would be expected by chance (Figure 4), with confidence intervals interpretation of the equal sign in the symbolic condition. Contrary
ranging from a low of 43.3% to a high of 97.1%. to our prediction that children’s verbal interpretations of the equal
sign might differ, children who solved nonsymbolic problems first
proceeded to describe the equal sign nonrelationally in the arith-
Strategy Use
metic context a week later. In the symbolic/nonsymbolic group, 15
Having solution options that corresponded to specific ways of thinking of the children gave operator interpretations, and 1 gave an ambiguous
about and solving equivalence problems allowed us to examine whether interpretation. In the nonsymbolic/symbolic group, 13 children gave
100
90
80
Percent Correct
70
60
Symbolic
50
Nonsymbolic
40
30
20
10
0
symbolic/nonsymbolic nonsymbolic/symbolic
Session Order
16
14
Variety
Number of Children
12
Amiguous
10
Part-whole vs. Combination
8 Pattern/Matching
6 Add-to-equal
Add-all
4
Relational
2
0
Symbolic Nonsymbolic
Session
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Figure 5. Number of the 16 second-grade children in the symbolic/nonsymbolic group who used specific
This document is copyrighted by the American Psychological Association or one of its allied publishers.
operator interpretations, 2 gave ambiguous interpretations, and 1 child equal sign) can help children map their successful approaches to
said that she did not know what the symbol meant. This particular symbolic problems. Having children learn to reason about equiv-
child, not surprisingly, did not solve any of the equivalence problems alence problems in a nonsymbolic format would be a relatively
correctly in either condition, despite solving 92% of the arithmetic easy, inexpensive intervention that could have a positive impact on
problems correctly. Thus, children’s performance on equivalence children’s ability to solve similar problems presented symboli-
problems can be high even though they verbally describe the equal cally. Examining potential boundary conditions surrounding the
sign in operator terms in the context of an arithmetic equation. effectiveness of this intervention would be an important step
Perhaps a procedural improvement (i.e., using arithmetic to make both before suggesting the widespread use of manipulatives for teach-
sides of an expression equivalent) precedes an ability to verbally ing children to think about equations relationally, however. There
define the symbol, or the procedure of solving relationally is more may have been other unique factors, in addition to the nonsymbolic
directly modified by exposure to nonsymbolic problems than verbal presentation format, that helped children map their relational strategy
interpretations of the equal sign. to symbolic problems, such as the presence of multiple-choice op-
tions. Directly examining the potential benefits of using multiple-
Discussion choice options rather than having children respond from an unre-
In this study, we confirmed what was only suspicion based on stricted range should be examined in future research.
previous research (Study 1): Experience with nonsymbolic equiv- The within-subject design was instrumental for examining order
alence problems can actually lead to improvements on subsequent effects and detecting the benefit of having children solve nonsym-
symbolic problems. This result is particularly surprising consider- bolic problems prior to symbolic problems. Children in the non-
ing that sessions were approximately 1 week apart, and the exper- symbolic/symbolic group may have remembered the approach
imenters made no direct references linking the two sessions to- they used to solve the nonsymbolic problems and transferred the
gether. The potential for application to classroom instruction procedures to the same symbolic problems presented a week later.
seems considerable. Specifically, experience with the complex Perhaps children had no idea how to solve the symbolic equiva-
equivalence problems without the potentially misleading cues lence problems (an explanation that is supported by the literature),
from conventional symbols (due to a misunderstanding of the so they used whatever they could construct from their strategy
16
14
Number of Children
12 Variety
10 Amiguous
Pattern/Matching
8
Add-to-equal
6 Add-all
4 Relational
2
0
Symboli c Nonsymbolic
Session
Figure 6. Number of the 16 second-grade children in the nonsymbolic/symbolic group who used specific
strategies on both symbolic and nonsymbolic sessions (Study 2).
98 SHERMAN AND BISANZ
arsenal. For children in the nonsymbolic/symbolic group, their arse- lems are conventionally presented can excel when numerically
nal would consist of the strategies used for problems they had seen a identical problems are presented in the absence of arithmetic
week before that seemed similar in some important and salient as- symbols. That is, Grade 2 children are quite capable of using
pects, such as having quantities on both sides that needed to be reasoning relationally about complex, algebra-like problems when
considered when solving. Rittle-Johnson (2006) also found that after the problems are presented in a nonsymbolic context. This result
a brief intervention, some children showed a significant improvement has important psychological and academic implications.
on equivalence problems, yet did not show similar improvements on Psychologically, children’s differential performance across the
tests of conceptual understanding of the equal sign. She too hypoth- two contexts suggests that children may encode the equivalence
esized that children’s procedural approaches, and their self- problems in distinct ways, depending directly on the presence or
explanations of those procedures, may have served as mechanisms of absence of conventional symbols. McNeil and Alibali (2002, 2004,
change even though children did not explicitly think about the con- 2005b) have described the formation of addition schemas, based
ceptual rationale underlying the problems. on operational patterns common within early arithmetic, including
Although children in the nonsymbolic/symbolic group showed the need to operate on all numbers and the belief that the solution
marked improvement on symbolic problems in their second session, must come immediately after the equal sign. Symbolic equivalence
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.
children who began with symbolic problems had extremely low problems look very much like the canonical problems that children
accuracy. The low performance of children in this group is consistent solve regularly, causing children’s representations of the opera-
with previous research findings and helps to support the argument that tional patterns to become activated. These representations control
elementary school children are in need of specific instruction, inter- attention, determine what is encoded, and influence how informa-
vention, or experience to help them not only solve equivalence prob- tion is interpreted (McNeil & Alibali, 2005b). Unfortunately, the
lems successfully but, more important, understand the equal sign in result of this process in the context of familiar symbols is that
all, but especially in arithmetic, contexts. Use of the multiple-choice children use an operator approach for solving equivalence prob-
format was helpful in making judgments about the relative success, or lems. In contrast, nonsymbolic equivalence problems lack conven-
in this case failure, on the equivalence problems. We recommend the tional symbols, and therefore children’s addition schemas may not
use of such procedures for assessing children’s performance on equiv- have been activated for problems in this context. Thus children
alence problems for this very reason. could override, or avoid altogether, their addition schemas, en-
Contrary to our initial hypothesis, children who solved nonsym- abling relational approaches to be used.
bolic problems first did not give more relational verbal interpretations Academically, children’s accuracy on nonsymbolic equivalence
of the equal sign than children who solved symbolic problems first. In problems suggests that children can reason about arithmetic, even
fact, no children from either group described the equal sign relation- complex problems such as equivalence problems, in a relational
ally. This outcome is surprising considering that children in the manner, long before formal algebra is introduced in the classroom.
nonsymbolic/symbolic group performed significantly higher on sym- Manipulatives may be an important tool for not only assessing chil-
bolic problems than children from the symbolic/nonsymbolic group. dren’s relational reasoning but perhaps for instructing students about
Thus, children’s performance on equivalence problems can be high equivalence and relational thought (Taylor-Cox, 2003). The blocks
even though they use operator terms to describe the equal sign in the and bins used in this nonsymbolic context offered the benefit of not
context of an arithmetic equation. If the ability to give a verbal being associated with misconceptions, as are conventional symbols
definition constitutes conceptual understanding, then perhaps the fact such as the equal sign, perhaps allowing children to focus on the
that children were able to use the equal sign in a relational manner yet relationships between quantities and sides of the equations, rather than
not be able to explicitly define the symbol relationally is another on applying addition schemas. Despite the success by children who
example of children changing procedurally before they change con- solved nonsymbolic problems, however, it is important to note that
ceptually (Baroody, 2003; Rittle-Johnson, 2006; Rittle-Johnson & manipulatives are not always recommended for teaching new con-
Siegler, 1998). However, with the present data, we cannot speak to the cepts. In fact, some have argued that manipulatives themselves are
order or type of understanding. Rather, children’s disconnect between simply new symbols systems and, as such, are open to being misun-
solving equivalence problems and verbally defining the equal sign derstood, misapplied, and nontransparent (Uttal, Liu, & DeLoache,
may be another example of Lawler’s (1981) microworlds of under- 2006; Uttal, Scudder, & DeLoache, 1997).
standing. In the present study, children’s solving, or procedural, world The fact that Grade 2 children can solve equivalence problems with
might be disconnected from the defining world, such that children manipulatives may not seem surprising to some, given that even
were able to construct strategies for solving equivalence problems yet preschool children can solve simple arithmetic problems presented
did not transfer their knowledge of equivalence to the task of defining with objects or events (Mix, 1999a, 1999b). However, given the
the equal sign. increased attention to children’s poor performance on algebra in
higher mathematics (Knuth et al., 2005), and the recent movement
General Discussion towards introducing algebra earlier in the curriculum (Carpenter et al.,
2003; Carraher & Schliemann, 2007; Taylor-Cox, 2003), document-
Various equivalence problems and procedures were used to ing children’s specific abilities and limitations becomes critical. These
determine the extent to which Grade 2 children can reason rela- studies are the first to conclusively show that young children are
tionally about arithmetic problems in both symbolic (i.e., conven- capable, with very little instruction or feedback, of solving complex
tional) and nonsymbolic (i.e., manipulatives) contexts. In prior equivalence problems when presented in a nonsymbolic context.
studies, researchers have found fairly consistent, and dismal, per- Building upon such competencies may serve as an important educa-
formance on symbolic equivalence problems by elementary school tional tool for helping to remediate children’s failure on equivalence
children. We found that children who fail when equivalence prob- problems and misunderstanding of the equal sign.
EQUIVALENCE PROBLEMS WITHOUT SYMBOLS 99
In Study 2, we found that children who solved nonsymbolic equiv- Hyde, J. S., Fennema, E., & Lamon, S. J. (1990). Gender differences in
alence problems in the first session went on to have impression mathematics performance: A meta-analysis. Psychological Bulletin,
accuracy on the same problems presented symbolically a week later. 107, 139 –155.
This result supports the idea that early success, specifically success on Hyde, J. S., Fennema, E., Ryan, M., Frost, L., & Hopp, C. (1990). Gender
nonsymbolic equivalence problems, can be useful for improving comparisons of mathematics attitudes and affect: A meta-analysis. Psy-
performance on subsequent symbolic problems. Having children chology of Women Quarterly, 14, 299 –324.
Kieran, C. (1981). Concepts associated with the equality symbol. Educa-
solve nonsymbolic problems first may have allowed them to map
tional Studies in Mathematics, 12, 317–326.
their experiences onto the symbolic equivalence problems presented a
Knuth, E. J., Alibali, M. W., McNeil, N. M., Weinberg, A., & Stephens,
week later. Future research is needed to assess how well children can A. C. (2005). Middle school students’ understanding of core algebraic
maintain and transfer their success on symbolic equivalence problems concepts: Equality and variable. Zentralblatt für Didaktik der Mathema-
after the nonsymbolic intervention. tik [International Reviews on Mathematical Education], 37, 1–9.
Although the main result of Study 2 appears to be the impressive Lawler, R. W. (1981). The progressive construction of mind. Cognitive
success of children in the nonsymbolic/symbolic group on sym- Science, 5, 1–30.
bolic equivalence problems, children from the opposite order
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
group also showed notable behavior. Specifically, children who genuine well-being report. Retrieved April 14, 2008, from http://
began with symbolic problems had low accuracy and used operator www.leduc.ca/Leduc/1024/News_and_Events/initiatives/gwa.asp
strategies on their first session, yet approached the same problems McNeil, N. M., & Alibali, M. W. (2000). Learning mathematics from
very differently when the problems were presented with manipu- procedural instruction: Externally imposed goals influence what is
latives a week later. The effect of context is particularly striking learned. Journal of Educational Psychology, 92, 734 –744.
among children in this group. It is not the case that solving McNeil, N. M., & Alibali, M. W. (2002). A well-established schema can
interfere with learning: The case of children’s typical addition schema.
symbolic problems first affects performance on subsequent non-
In C. D. Schunn & W. Gray (Eds.), Proceedings of the 24th annual
symbolic problems. Rather, children switch from viewing the
conference of the Cognitive Science Society (pp. 661– 666). Mahwah,
problems in an operator fashion to viewing the problems relation-
NJ: Erlbaum.
ally. The influence of conventional symbols is particularly evident McNeil, N. M., & Alibali, M. W. (2004). You’ll see what you mean:
among children in this group. Students encode equations based on their knowledge of arithmetic.
These studies mark the first step in identifying (a) children’s Cognitive Science, 28, 451– 466.
significant success on equivalence problems in the absence of McNeil, N. M., & Alibali, M. W. (2005a). Knowledge change as a function
symbols, (b) their ability to benefit from prior experience with of mathematics experience: All contexts are not created equal. Journal
manipulatives on subsequent symbolic problems, and (c) how of Cognition and Development, 6, 285–306.
differently children approach equivalence problems depending McNeil, N. M., & Alibali, M. W. (2005b). Why won’t you change your
on the context, as evident in both accuracy and oral justifica- mind? Knowledge of operational patterns hinders learning and perfor-
tions. Children have competencies hidden by their misconcep- mance on equations. Child Development, 76, 883– 899.
tions of mathematical symbols, particularly the equal sign, and Mix, K. S. (1999a). Preschoolers’ recognition of numerical equivalence: Se-
both discovering and building upon these competencies are quential sets. Journal of Experimental Child Psychology, 74, 309 –332.
important for remediating children’s difficulty with equivalence Mix, K. S. (1999b). Similarity and numerical equivalence: Appearances
count. Cognitive Development, 14, 269 –297.
and for preparing the way to higher algebraic and symbolic
Perry, M. (1991). Learning and transfer: Instructional conditions and
thought.
conceptual change. Cognitive Development, 6, 449 – 468.
Perry, M., Church, R. B., & Goldin-Meadow, S. (1988). Transitional
References knowledge in the acquisition of concepts. Cognitive Development, 3,
359 – 400.
Alibali, M. W. (1999). How children change their minds: Strategy change
Renwick, E. M. (1932). Children’s misconceptions concerning the symbols
can be gradual or abrupt. Developmental Psychology, 35, 127–145.
for mathematical equality. British Journal of Educational Psychology, 2,
Baroody, A. J. (2003). The developmental of adaptive expertise and
173–183.
flexibility: The integration of conceptual and procedural knowledge. In
Rittle-Johnson, B. (2006). Promoting transfer: Effects of self-explanation
A. J. Baroody and A. Dowker (Eds.), The developmental of arithmetic
and direct instruction. Child Development, 77, 1–15.
concepts and skills: Constructing adaptive expertise (pp. 1–33). Mah-
Rittle-Johnson, B., & Alibali, M. W. (1999). Conceptual and procedural
wah, NJ: Erlbaum.
Baroody, A. J., & Ginsburg, H. P. (1983). The effects of instruction on knowledge of mathematics: Does one lead to the other? Journal of
children’s understanding of the “equals” sign. Elementary School Jour- Educational Psychology, 91, 175–189.
nal, 84, 199 –212. Rittle-Johnson, B., & Siegler, R. S. (1998). The relation between concep-
Behr, M., Erlwanger, S., & Nichols, E. (1980). How children view the tual and procedural knowledge in learning mathematics: A review. In C.
equals sign. Mathematics Teaching, 92, 13–15. Donlan (Ed.), The development of mathematical skills: Studies in devel-
Carpenter, T. P., Franke, M. L., & Levi, L. (2003). Thinking mathemati- opmental psychology (pp. 75–110). Hove, UK: Psychology Press.
cally: Integrating arithmetic and algebra in elementary school. Ports- Sáenz-Ludlow, A., & Walgamuth, C. (1998). Third graders’ interpretations
mouth, NH: Heineman. of equality and the equal symbol. Educational Studies in Mathematics,
Carraher, D. W., & Schliemann, A. D. (2007). Early algebra and algebraic 35, 153–187.
reasoning. In F. Lester (Ed.), Second handbook of research on mathematics Seo, K., & Ginsburg, H. P. (2003). “You’ve got to carefully read the math
teaching and learning: A project of the National Council of Teachers of sentence . . . .”: Classroom context and children’s interpretations of the
Mathematics. (Vol. 2, pp. 669 –705). Charlotte, NC: Information Age. equals sign. In A. J. Baroody & A. Dowker (Eds.), The development of
Felson, R. B., & Trudeau, L. (1991). Gender differences in mathematics arithmetic concepts and skills: Constructing adaptive expertise. Mah-
performance. Social Psychology Quarterly, 54, 113–126. wah, NJ: Erlbaum.
100 SHERMAN AND BISANZ
Sherman, J. (2004). Constraints on understanding arithmetic equivalence: Child psychology: A handbook of contemporary issues (2nd ed., pp.
Effects of misunderstanding the equal sign. Unpublished manuscript. 167–184). New York: Psychology Press.
Taylor-Cox, J. (2003). Algebra in the early years? Young Children, 58, Uttal, D. H., Scudder, K. V., & DeLoache, J. S. (1997). Manipulatives
14 –21. as symbols: A new perspective on the use of concrete objects to teach
Uttal, D. H., Liu, L. L., & DeLoache, J. S. (2006). Concreteness and mathematics. Journal of Applied Developmental Psychology, 18,
symbolic development. In L. Balter and C. S. Tamis-LeMonda (Eds.), 37–54.
Appendix A
Relational view Child’s response includes the notion of both “The numbers on this side are the same as on this side.”
sides of an equation being the same in
value, or equal.
Operator view Child’s response indicates that the equal “It means you put it together and get the answer.” “It
sign signals what you’re supposed to do, means what the answer is.” “It means where the
where the answer should go, the end of answer is going to go.”
the expression, or what the answer is.
Ambiguous Child’s response was not interpretable. “It’s the number that’s going to come up.” “It’s like
saying, ‘What’s the question’?”
Appendix B
Equivalence
Identical numerosity Child refers to the same number on both “sides,” counts to ensure that both sides have the
same number, or comments on both sides being “even.” In a few cases, the “sides” that
the child refers to do not correspond to the blue tent or the equal sign. For example, in
the block condition, some children match their blocks with all the blocks already
present (e.g., you have 10 blocks, and so do I).
Identity Child refers to a pattern (e.g., 3 ⫹ 4 and 3 ⫹ 4), matching both sides to “look” the same,
or copying the sides. Occurs mainly with identity and commutativity problems.
Grouping Used for combination problems only. Child refers to adding only the b and c terms, and
may or may not explicitly state that the first term and the term after the equal sign (d
term) are the same. Example (2 ⫹ 5 ⫹ 3 ⫽ 2⫹__): “Because there’s 2 here [a term],
and 2 here [d term], so I don’t count these, then the rest (b ⫹ c) is 8, which I put.”
Nonequivalence
Add-all The child mentions adding all terms or overtly counts all terms.
Add-to-equal Child mentions adding or counts the terms on the left side of the blue tent or equal sign.
This justification does not apply to addition problems, where there are no terms on the
right side.
Carry Child mentions copying one of the terms in the expression. He or she may state that the
other numbers don’t count, or that they should be excluded for some reason.
Carry-after-equal Occurs mainly in the symbolic condition. Child states that he or she is taking the term
after the equal sign as being the answer, and writing it on the line. This justification is
typically accompanied by an justification that the term immediately following the equal
sign must be the answer; therefore, all they have to do is copy that term on to the line.
Other
Arithmetic Used for addition problems only. Child states that he or she added the terms together
without commenting on equality.
“I don’t know” Child states that he or she does not know how he or she solved the problem.
Guessing Child indicates that he or she guessed in order to solve.
Ambiguous Child’s justification can not be interpreted.