You are on page 1of 11

Scandinavian Journal of Psychology, 2005, 46, 157–167

Number-word sequence skill and arithmetic performance


Blackwell Publishing, Ltd.

BO S. JOHANSSON
Department of Teacher Education, University of Uppsala, Sweden

Johansson, B. S. (2005). Number-word sequence skill and arithmetic performance. Scandinavian Journal of Psychology, 46, 157–167.
In two studies, the role of the number-word sequence skill for arithmetic performance was investigated. In the first, children between 4 and 8
years of age were asked to count forward and backward on the number-word sequence and to solve arithmetic problems followed by post-
solution interviews about solution procedures. The results demonstrated that the number-word sequence skill predicted both number of prob-
lems solved and strategy to solve the problems. In Study 2 it was found that solving doubles (e.g., 2 + 2 = ?) problems served as a link between
the number-word sequence skill and the number of arithmetic problems solved. The findings suggest that counting on the number-word sequence
may be an early solution procedure and that, with increasing counting skill, the child may detect regularities in the number-word sequence
that can be used to form new and more accurate strategies for solving arithmetic problems.
Key words: Number-word sequence skill, arithmetic performance, solution procedure, counting.
Bo S. Johansson, Department of Teacher Education Box 3621 S- 750 02 Uppsala, Sweden . E-mail: Bo.Johansson@ilu.uu.se

& Huttenlocher, 1992) that young children find arithmetic


NUMBER-WORD SEQUENCE SKILL AND problems, with the addends represented by concrete objects,
ARITHMETIC PERFORMANCE easier than story problems as e.g., “Joe had 3 apples. He got
A recurrent finding has been that children’s skill in reciting 2 more. How many does he have now?” or number fact
the number-word sequence is a strong predictor of perform- problems, “How much is 3 and 2?”. Therefore, the number-
ance on other numerical tasks. For example, a substantial word sequence skill as such is not regarded as the root of the
correlation between reciting the counting string and the arithmetic ability (e.g., Gelman & Gallistel, 1978; Steffe, von
dividing of objects into equally numerous sets was found in Glasersfeld, Richards, & Cobb, 1983; Steffe, 2002). Rather,
Pollio and Whitacre (1970). Secada, Fuson, and Hall (1983) availability of countable items has been considered an indis-
observed that the use of the counting-on procedure for pensable prerequisite for the development of the number
solving addition problems was related to the skill of count- concept. In a very influential paper Steffe et al. (1983)
ing forward from an arbitrary point in the number-word argued that initially the number-word sequence is acquired
sequence. The same skill has also been found to strongly as a recitation skill independent of the ability to form count-
predict performance on additive composition of number able units. The step from reciting to counting requires the
problems (Martins-Mourao & Cowan, 1998). Ho and Fuson construction of countable unit items and that each unit item
(1998) reported a correlation between length of counting is one-to-one coordinated with a number-word (Steffe et al.,
forward sequences and understanding of teen quantities and 1983; Steffe & Cobb, 1988). The authors have described the
performance on simple addition problems. In a study of development of five countable unit items allowing increas-
school-beginners’ numerical abilities Wigforss (1946) found ingly advanced procedures to solve arithmetic problems.
that counting forward was the best predictor of perform- First, perceptual unit items are constructed, for example
ance on the other numerical tests. Wigforss’ conclusion perceiving marbles or checkers as a plurality of objects that
was “The little respect that many teachers seem to give to can be counted. At this level, a child can only count if per-
the importance of the child’s ability to recite the number- ceptual unit items are available. With no such units at hand,
words is evidently unfounded” (Wigforss, 1946, p. 19, my “there is literally nothing for the child to count” (Steffe
translation). et al., 1983, p. 47). Next, figural unit items are created, fol-
lowed by motor unit items, verbal unit items, and finally,
abstract unit items. Common for the first four unit items is
Countable unit items and the development of that each number-word needs some sort of perceptual sup-
the arithmetic skill port, whereas abstract unit items means that a given number
The above findings suggest that the number-word sequence word “stands for a specific collection of individual unitary
skill may play a central role in the development of the child’s items” (Steffe et al., 1983, p. 42), for example, the word
arithmetic performance. However, research (e.g., Wynn, 1998) “seven” stands for seven marbles. Thus, according to this
has shown that infants, months before saying their first word, theory – the counting types theory – the child solves
can solve easy numerical problems represented as sets of arithmetic problems by counting explicitly or implicitly con-
physical objects. It has also been found (e.g., Levine, Jordan, structed unit items.

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd. Published by Blackwell Publishing Ltd., 9600 Garsington
Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA. ISSN 0036-5564.
158 B. S. Johansson Scand J Psychol 46 (2005)

The number-word sequence hypothesis pulatives available as solution aids. According to the counting
types theory, this should force children with a low level of
The hypothesis tested here is that counting on the number- number skill to either construct countable items, e.g. make
word sequence is an early, and increasingly efficient, proce- use of the fingers to construct the addends and sum, or refrain
dure to solve arithmetic problems. Initially, the child may from solving the problems. The number-word sequence
solve arithmetic problems using countable units to represent hypothesis, on the other hand, predicts that with no concrete
the addends, but as the sequence skill develops, alternative objects at hand, the children may try to solve the problems
strategies based on counting on the number-word sequence by counting on the number-word sequence.
are assumed to emerge. It is further assumed that task char- A second difference is the role played by objects when
acteristics determine which solution procedure the child will solving the problems. The counting types theory states that
use. With objects present during testing, the child may use the solution procedure rests on the construction of count-
them as solution aids; but with no physical props available, able unit items. Initially perceptual unit items are con-
the child may try to solve the problem by counting on the structed; later the child develops the capacity to construct
number-word sequence (see also Carpenter & Moser, 1984). abstract unit items allowing internalized strategies. There-
For example, a child who can only count forward, starting fore, the counting types theory predicts that the use of
with the number word “one”, may try to solve the (three objects is related to level of number skill: object counting
apples and two apples) problem mentioned above by count- strategies should decrease in frequency as the number skill
ing “1, 2, 3; 4, 5” (the Sequence counting-all strategy). As increases. According to the number-word sequence hypo-
the child’s number-word sequence skill increases, the child thesis, objects have to be used to monitor the count if the
may find structures in the sequence that can be used to solve addends exceed the stable portion of the sequence. There-
the problems. A first step is assumed to be the ability to fore, problem size should constitute the main factor deter-
break up the sequence and focus on individual number- mining the use of objects. Problems with addend size within
words. Mastering this, the child should be able to count for- the stable portion of the sequence should be solvable by
ward and backward, starting and ending the count at arbi- counting on the number-word sequence, whereas when the
trary number words, for example, counting from 7 to 3. This addends exceed the stable portion of the sequence it is
acquisition should allow the child to solve the apples prob- necessary to find a representation supporting the count.
lem with the Sequence counting-on strategy, that is, starting
the count from the number-word “three”: “3; 4, 5” (Secada
et al., 1983). Finally, the child may detect regularities in the Arithmetic solution procedures
sequence, such as the decade or the doubles (2 + 2 = 4, etc.) The children’s procedures to solve arithmetic problems have
regularity, which allows the child to solve the apples prob- been the subject of extensive research (see, e.g., Gal’perin, 1970;
lem by jumping on the sequence: “2 and 2 makes 4, 1 more Svenson & Broquist, 1975; Svenson, Hedenborg, & Lingman,
– 5” (Carpenter & Moser, 1984). 1976; Carpenter & Moser, 1983; Steffe et al., 1983; Steffe
Level of number-word sequence skill should determine the & Cobb, 1988). These procedures are usually grouped into
ease with which the number-word sequence can be used as a three developmental levels (e.g. Fuson, 1992) related to the
solution tool. Monitoring strategies, such as representing the counting types theory. At the first level the child models
second addend by a finger pattern or by an image of objects the addends with the help of objects. For example, to solve the
is assumed to be necessary with addends exceeding the problem “Jim has four apples, then he gets three more. How
thoroughly learned portion of the sequence. Children at a many apples does he have altogether?” the child first counts
high level of number-word sequence skill, on the other hand, out four objects, then three additional objects and finally,
should be able to represent the second addend by segment- joins the two sets to count the sum (the Concrete counting-
ing the number sequence. For example, children with an all strategy). A child at level two is assumed to be able to
incompletely elaborated sequence, when asked to solve the solve the same problem by counting on representations of
“17 + 7” problem, may have to represent the number-word the objects, for example, verbal unit items (Fuson, 1992). The
7 by extending seven fingers, whereas children at a more child may use the Sequence counting-on procedure, starting
advanced level may decompose 7 into 3 and 4 and then jump with the number-word for the first addend (4; 5, 6, 7). Thus,
on the number row, first to 20 then to 24. children at the first level need objects to represent the two
Thus, a first difference between the counting types theory addends and the sum, whereas children at the second level
and the number-word sequence hypothesis is that the former can use the number-words to represent the quantities
stipulates that initial solution procedures require countable (Fuson, 1992). In the first level, the counting of the addends
items, whereas the number-word sequence hypothesis identi- is separated from the counting of the sum, whereas in the
fies an additional early solution procedure: counting on the second level the counting of the second addend is embedded
number-word sequence. It was decided to test these contrast- within the counting of the sum. Embedded counting requires
ing hypotheses by asking children in the age interval 4–8 the child to be able to pair the number-words with some
years to solve simple arithmetic problems with no mani- kind of unit representing the second addend to keep track of

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd.


Scand J Psychol 46 (2005) Sequence skill and arithmetic performance 159

counting (Steffe et al., 1983; Fuson, 1992). In the third level, objects or images to keep track of the number-words said for
the problems are solved by known facts and derived facts the second addend.
procedures; for example, by doubles decompositions or Finally, the Bidirectional chain level is reached, at which
immediate retrieval “7, I know that 3 + 4 makes 7”. In the the child can count both forward and backward with ease,
present study these three levels were taken as the frame of accuracy and speed. New number skills are (Fuson, 1992)
reference when categorizing the solution strategies used to the understanding of the inverse relation between addition
solve the arithmetic problems. and subtraction and the ability to break down small num-
bers into addend pairs. According to the number-word
sequence hypothesis, children having reached this level
Level of elaboration of the number-word sequence and should be able to solve the problems by segmenting the
arithmetic solution procedures sequence. For example, they may use various decomposition
Fuson, Richards and Briars (1982) made an in-depth investi- strategies, such as making use of tens complements (e.g. “16
gation of the child’s acquisition and elaboration of the + 7 = 16 and 4 is 20, 3 more makes 23”), doubles decompo-
number-word sequence. They found five levels in the elabo- sitions (“7 − 3 = 4 because 3 + 3 = 6, 1 more makes 4”) or
ration of the sequence and also described the number skills immediate retrieval (Carpenter & Moser, 1984).
at each level. In her 1992 paper, Fuson summarized these In sum, Fuson’s (1992) analysis of the relation between
results and also specified the arithmetic problem solution elaboration level and arithmetic solution procedures offers
procedures available at each elaboration level. an experimental situation to test the two contrasting hypo-
At the two first – the String and Unbreakable list levels – theses presented above. In the present report this was made
the child can only produce the sequence by beginning with by presenting the children with two number-word sequence
the number-word “one”; the child has not yet acquired the tests (counting forward and counting backward), followed
capacity to start counting from an arbitrary entry point. by a series of simple addition and subtraction tasks. Count-
With the Unbreakable list level emerges the capacity to ing forward and counting backward performance was used
count a given number of objects to decide the cardinality of to categorize the children with respect to level of elaboration
the set, which, according to the counting types theory allows of the number-word sequence. Then, the relation between
the child to solve addition problems by counting perceptual level of number-word sequence skill and arithmetic perform-
unit items. Not finding objects to represent the addends, the ance was analyzed.
problems should be unsolvable. According to the number-
word sequence hypothesis, children at this level should be
able to solve arithmetic problems using the number-word STUDY 1
sequence, given that the sizes of the addends do not exceed The children were first given two number-word sequence tests,
the stable portion of the sequence. As children at this level followed by two arithmetic tests. The number-word sequence
have to start counting by the number-word “one”, the preferred tests were first used to investigate whether a precondition
solution procedure should be counting-all. for the subsequent analysis was met, namely that the elabo-
At the Breakable chain level, the children can produce ration levels in Fuson et al. (1982) could be replicated. If
parts of the beginning sections of the sequence from arbi- that was the case, the contrasting hypotheses stated above
trary entry points. The skill of counting forward from three were to be tested.
to six or backward from five to one constitutes a first step
towards the counting-on strategy (Secada et al., 1983).
Fuson (1992) pointed out that this might be a transitional Method
stage; the children can use the counting-on method but
Subjects. The subject pool consisted of 61 girls and 65 boys, aged
need objects to represent the second addend. According
4 to 8 years (mean age = 6 years and 9 months) and living in
to the number-word sequence hypothesis, these children middle-class areas. Of these, 75 were in pre-school (in mixed-age
should be able to count on the number-word sequence on groups) and 51 attended first grade. Because testing was performed
problems with small addends, but should require objects in early spring, the 51 school children had already completed one
or images of objects to monitor the count in the case of term of arithmetic teaching, with exercises on cardinal and ordinal
numbers, N + 1 / N − 1 and N + 2 / N − 2 practices, numeral writing
large addends.
exercises, and the introduction of the plus, minus and equal signs.
For children at the Numerable chain level the number- In the pre-schools no formal arithmetic training was given. All
words have become countable units (Fuson et al., 1982), parents, altogether 131, in the selected pre-schools and schools, were
which means that the addends can be represented by contacted and all but five responded favorably to the testing of their
number-words. Problems like “9 + 6 = ?” can be solved by child.
counting up 6 words from 9, and “15 − 4 = ?” by counting
Tests. Two sequence skills tests were constructed: Forward count-
down 4 words from 15. Here the counting types theory and ing from a to b (e.g., from 8 to 12) and backward counting from b
the number-word sequence hypothesis arrive at the same to a (e.g., from 13 to 9). The counting forward problems were count-
predictions. With large addends, the child may still need ing from 3 to 7, 5 to 9, 8 to 12, 17 to 21, 28 to 32, 47 to 51, and 98

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd.


160 B. S. Johansson Scand J Psychol 46 (2005)

to 102, with the first problem serving as a practice task. Seven • Immediate recall. The child relied on facts recalled from memory,
counting backward problems were constructed: counting from 5 to with no signs of counting. To be defined as Immediate recall, the
1, 8 to 4, 13 to 9, 18 to 13, 23 to 18, 32 to 27, and 203 to 198, with response had to be given within about 3 sec.
the first problem a practice task. • Not identifiable. Some children could not describe their solution
The arithmetic tasks were simple addition and subtraction prob- procedure, others said that they had guessed the answer or did not
lems of type Change 1 and Change 2 (see Riley, Greeno, & Heller, complete the solution and some descriptions were too vague to be
1983). Four lists with six addition and six subtraction problems categorized. All these responses were scored as Not identifiable.
each were constructed with the objectives to obtain a great variation • Not given. The problem was considered beyond the child’s current
in problem difficulty and to stimulate different solution procedures. performance level and was therefore not given.
Commutative pairs were not used to avoid problem recognition.
The problems, except the easiest ones, required counting beyond a On a few occasions the children modeled the second addend before
decade. The lists are given in the Appendix. Each child was testing solving the problem, named CE strategy (Fuson, 1982; Baroody,
with one list made up by six addition and six subtraction problems. 1987). For example, when given the problem “5 + 3 = ?”, the child
The problems were given in order of difficulty. Testing started first counted out three fingers and then counted from 1 to 8.
with the two number-word sequence tests, counterbalanced for Similarly for the problem “26 + 15 = ?”; a few children extended all
order, followed by first the addition then the subtraction test. fingers on one hand three times (5 + 5 + 5) and then used Sequence
counting-on. The number of observed CE strategies was only seven;
Procedure. The children were tested individually in a separate therefore no such category was defined. Whether the children solved
room, the sessions were video recorded using a Panasonic WP-G1 the problems by counting from first or from larger addend was not
camera with a built-in timer to measure response times. The analyzed, because the problem list was not constructed to test this
sequence instructions were, “Can you count forward from a to b/ effect.
backward from b to a?” If necessary, the meaning of the terms was The codes for the subtraction solution procedures were (illustrated
explained, followed by a single practice trial. The instructions for with 7 − 2 = ?):
the addition problems were, “Imagine that you have 6 pears/crowns
(etc.) and then get 7 more. How many do you have altogether?” • Take away. The child counted out the minuend on the fingers
Instructions for the subtraction problems were, “Imagine that you (counting out 7 fingers), took away the given subtrahend entities
have 13 pears/crowns (etc.) and then give away 7. How many do you from the minuend (folding down 2 of the fingers), and left the
have left?” After the solution of each problem, the child was asked, remaining extended fingers to be counted.
“How did you figure out the answer to this problem?” If the child’s • Adding up to minuend. The child counted out the given subtrahend
answer was unclear, the experimenter would pose follow-up ques- (counted out 2 fingers); and then added on entities until the minuend
tions. Testing was terminated when the child had failed on two was reached (5 more fingers were counted out, separately form the
consecutive problems, or if the experimenters judged that the child’s 2 first). Finally the child counted the added-on entities.
highest performance had been attained. The experimenters did their • Sequence counting-down. The child started with the minuend (7)
best to create a supportive and relaxed atmosphere, which encour- and then counted down the number of words given by the subtra-
aged the children to try to solve many problems, even those beyond hend (6, 5).
their normal capacity. • Sequence counting-up. The child began with the subtrahend (2)
and counted up to the minuend, counting the number of counts.
Sequence skill. Correct responding was defined as beginning with
the first given number-word, counting in the conventional order, The categories of Decomposition, Immediate recall, Not identifiable
and ending with the second number-word. Response time for all and problems Not given were defined as for addition.
correct responses was measured and standard deviations were Using the above definitions, a research assistant and the author
computed. Correct responses with a response time greater than scored the interviews. The inter-rater reliability was 91%. Following
two standard deviations were coded as incorrect, to obtain that Siegler (1987), coding was based on overt behavior when that behavior
extremely slow, but correct, solutions were not included. unambiguously indicated the strategy and on immediate, retrospect-
ive verbal reports when behavior was absent or ambiguous.
Arithmetic solution codes. The distinctions between concrete For the addition problems solved by Sequence counting-all and
counting, verbal counting, and recalled or derived facts strategies Sequence counting-on, and the subtraction problems solved by Sequence
(e.g., Carpenter & Moser, 1984; Fuson, 1992) were used to define counting-up and Sequence counting-down, was noted whether the
the solution procedures on the addition and subtraction problems. fingers or other objects were used or reported as solutions aids.
The addition strategies were:

• Concrete counting-all. Fingers or other countable objects were Results


counted out one by one to represent one of the addends followed
by a similar process for the other addend. Then, all the objects Sequence skill levels. The analysis focused on whether the
or fingers were counted to determine the sum (see Baroody, 1987). Fuson et al. (1982) elaboration levels could be found. The
Also short-cut Concrete counting-all strategies (Baroody, 1987)
were included.
counting backward data were taken as point of departure,
• Sequence counting-all. The child started with “one” and counted because advances in the ability to produce the number-
up to the cardinal value of one of the addends. Then, the child words in a backward order constitute a key factor in the
continued to count a number of steps equal to the other addend. definition of these levels. The present counting backward
• Sequence counting-on. The child started with the cardinal value of results indicated four distinct levels. The first consisted of
one of the addends and counted on from there while the other
addend was enumerated.
the 38 children with 0 correct. At the second were placed the
• Decomposition. The child decomposed the given numbers into 45 children solving 1 or 2 problems. The third comprised
easier number combinations, using doubles or tens complements. the 24 children being able to solve 3 or 4 problems and the

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd.


Scand J Psychol 46 (2005) Sequence skill and arithmetic performance 161

addition, they counted faster than the level 1 and 2 children,


indicating a separate skill level.
The children categorized as level 4 solved about 5 on both
the counting forward and backward problems. The present
children differed from those at level 3, besides scoring higher
on the counting backward problems, in rate of counting. If
rate and correctness of counting is taken to indicate ease and
flexibility in producing the number-words, these children
had reached the Bidirectional chain level.
In sum, performance at levels 1, 2, and 4 closely resembles
performance at the String and Unbreakable list levels, the
Breakable list level and the Bidirectional chain level, respect-
ively. The response profile for the level 3 children could not
be directly related to the Numerable chain level because the
Fig. 1. Performance on the counting forward problems as a func- measures used to define that level in Fuson et al. (1982)
tion of number of correct on the backward counting problems. differ from those used here. However, considering that per-
formance is located between levels 2 and 4, it seems reason-
able to assume that level 3 approximates the Numerable
fourth the remaining 19 children. To examine whether the chain level. In conclusion, it was possible to categorize the
characteristics for these four levels agree with the Fuson children according to the four main elaboration levels described
et al. (1982) elaboration levels, the counting forward perform- by Fuson et al. (1982). This provides a basis for the crucial
ance for each level was analyzed (see Figure 1). step in the analysis, that is, whether the arithmetic solution
An analysis of variance was run on the correct response procedures at each level could be predicted from the counting
data with level as a between-subjects variable. A significant, types theory or from the number-word sequence hypothesis.
F(3, 122) = 39,21, p < 0.0001, main effect was obtained, the
Tukey HSD procedure showed significant increases ( p < Level of number-word sequence skill and arithmetic performance.
0.05) from level 1 to 2 and from 2 to 3. To analyze the relation between level of number-word
Mean forward counting time for the four levels were: 3.6, sequence skill and arithmetic performance two measures
3.7, 3.3 and 2.6 sec. An analysis of variance with level as a were analyzed: frequency of report of each strategy and
between-subjects variable demonstrated a significant main strategy accuracy. The results were summed over the six
effect, F(3, 105) = 5.81, p < 0.0001. Tukey’s HSD indicated problem sizes. In the analysis of report frequency, all reports
that response time decreased ( p < 0.05) from level 2 to 3 and of a given strategy were scored as 1 and all other responses
from 3 to 4. as 0, and mean number of reports of that strategy were
The above forward and backward counting results mean computed for each child. These means were then subjected
that typical for the level 1 children were that they solved to a one-way analysis of variance with level of number-word
none of the counting backward tasks and a mean of about sequence skill as a between-subjects variable (F-statistics, see
one and a half of the counting forward problems. This Table 1). In the analyses of accuracy, only the reports of a
agrees with performance at the String and Unbreakable list given strategy were analyzed, using the same model of ana-
levels (Fuson et al., 1982), in that children at these levels had lysis of variance as above, with correct solutions scored as
great difficulties starting the count with other number-words 1 and incorrect as 0. In Table 1 are given mean frequency of
than “one”. That is, asked to count from “3 to 7” or from report of each strategy and also mean accuracy (as proportions)
“5 to 3” the present children should, in both cases, start for each strategy.
counting with the number-word “one”. An error analysis First, the relation between number-word sequence skill
showed that the present children produced 51% of all such and overall accuracy was analyzed (see bottom lines for
“running start” errors observed. That correct responding addition and subtraction, respectively, in Table 1). For both
did not exceed 50% on any of the counting forward problem addition and subtraction, overall accuracy increased mono-
(see Figure 1) further underlines that the ability to count tonically with increasing level of number-word sequence skill,
from an arbitrary number-word was very unstable. as indicated by the significant correlations ( p < 0.001) for
The children at level 2 could solve 1 or 2 of the counting both addition (0.64) and subtraction (0.66). Partialling
backward and about 3 of the counting forward problems, out age reduced the coefficients somewhat, but they were
with a high score of correct responding on the two first still significant (0.49 and 0.53, respectively). Thus, level of
problems (see Figure 1). Evidently, the beginning sections of number-word sequence skill is a significant factor for accur-
the sequence were elaborated to the Breakable chain level. acy on simple addition and subtraction problems. However,
The level 3 children could solve 3 or 4 of the counting this result may be explained in two ways: either there is a direct
backward and about 5 of the counting forward problems. In relation between number-word sequence skill and accuracy,

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd.


162 B. S. Johansson Scand J Psychol 46 (2005)

Table 1. Mean report frequency (max = 6.0) and accuracy (proportion correct) for each response category. Addition and subtraction results are
presented separately

Expertise level F statistics

Response category 1 2 3 4 F-ratio d.f. Sig. level

Addition
Immediate retrieval
Response frequency 0.18 0.90 1.32 1.58 32.13 3,122 0.0001
Accuracy 0.86 1.0 0.98 1.0 – – –
Decomposition
Response frequency 0.05 0.64 1.21 1.63 10.88 3,122 0.0001
Accuracy 1.00 0.98 0.79 0.79 1.51 3,38 n.s.
Sequence counting-on
Response frequency 0.58 1.45 1.54 2.16 6.25 3,122 0.001
Accuracy 0.71 0.57 0.94 0.74 3.84 3,65 0.05
Sequence counting-all
Response frequency 1.18 0.69 0.41 0.16 2.96 3,122 0.05
Accuracy 0.40 0.61 0.58 0.50 0.63 3,32 n.s.
Concrete counting-all
Response frequency 0.11 0.09 0.00 0.00 – – –
Accuracy 0.00 0.33 – – – – –
Not identified
Response frequency 1.34 1.11 1.13 0.42 3.83 3,122 0.05
Accuracy 0.11 0.13 0.12 0.14 0.03 3,83 n.s.
Not given
Response frequency 2.55 1.11 0.38 0.05 27.12 3,122 0.0001
Accuracy – – – – – – –
Overall accuracy 0.23 0.50 0.72 0.78 32.00 3,122 0.0001
Subtraction
Immediate retrieval
Response frequency 0.16 0.67 1.29 1.37 18.38 3,122 0.0001
Accuracy 1.0 1.0 1.0 1.0 – – –
Decomposition
Response frequency 0.08 0.29 0.71 0.74 4.94 3,122 0.01
Accuracy 1.00 1.00 0.58 0.94 3.06 3,26 0.05
Sequence counting-down
Response frequency 0.34 0.93 1.62 2.53 13.16 3,122 0.0001
Accuracy 0.46 0.58 0.50 0.63 0.60 3,81 n.s.
Take away
Response frequency 0.58 0.60 0.13 0.00 3.66 3,122 0.05
Accuracy 0.42 0.72 0.67 – 1.75 2,27 n.s.
Adding up to minuend
Response frequency 0.08 0.07 0.00 0.00 – – –
Accuracy 0.25 0.67 – – – – –
Sequence counting-up
Response frequency 0.02 0.04 0.04 0.04 – – –
Accuracy 1.0 0 1.0 1.0 – – –
Not identified
Response frequency 1.76 1.31 1.29 0.95 3.11 3,122 0.05
Accuracy 0.13 0.15 0.28 0.38 2.57 3,95 n.s.
Not given
Response frequency 2.97 2.09 0.92 0.37 31.10 3,122 0.0001
Accuracy – – – – – – –
Overall accuracy 0.17 0.37 0.52 0.68 31.96 3,122 0.0001

or the result is due to a selection of progressively more effi- cantly more frequent with increasing number-word sequence
cient solution procedures with increasing sequence skill. The skill: Immediate retrieval, Decomposition and Sequence
next series of analyses was run to test these two alternatives. counting-on, whereas Sequence counting-all, Concrete counting-
The data in Table 1 demonstrate a strong relation between all and Not identified were reported in decreasing frequencies.
level of number-word sequence skill and frequency of strat- In addition, the response category Not given decreased strongly
egy report. For addition, the following strategies got signifi- with increasing sequence skill.

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd.


Scand J Psychol 46 (2005) Sequence skill and arithmetic performance 163

The response frequency data for subtraction present a similar sum exceeding 10. Recalling that the forward and backward
picture. Strategies that increased in frequency were: Immedi- counting results demonstrated that the level 2 children could
ate retrieval, Decomposition and Sequence counting-down, produce only the beginning sections of the sequence from
whereas Take away and Not identified decreased. The strat- arbitrary entry points, it seems that the level 2 children tried
egies Adding up-to-minuend and Sequence counting-up to use this strategy on problems exceeding the stable portion
appeared in too few numbers to permit statistical analysis. of their sequences. In effect, level of accuracy on the two first
These results allow two important conclusions. First, the (below 10) arithmetic problems was about the same for both
strategy reports were strongly related to level of number-word level 2 (88%) and level 3 (100%) children, but differed
sequence skill. Strategies increasing in frequency indicated sharply on the four most difficult problems (43% and 92%,
that the child had learned the number facts in the problem respectively). Further support was obtained by an error
(Immediate recall), could start counting from an arbitrary analysis, which showed that the level 2 children produced
number-word in the sequence (Sequence counting-on and 82% of the erroneous responses under this strategy, but only
Sequence counting-down) or had acquired the ability to jump 39% of the Sequence counting-on strategy reports. Thus, the
on the numbers in the number-word sequence (Decomposi- dip in accuracy for the level 2 children was due to an exten-
tion). Strategies decreasing in frequency indicate a low level sion of the Sequence counting-on strategy to unstable
of mastery of the number-word sequence: The child had to portions of the number-word sequence.
start counting from the number-word “one” with or without A similar explanation holds for the level 3 children’s
relying on concrete counting (Sequence counting-all, Concrete Decomposition reports under subtraction in that their accur-
counting-all or Take-away), or the child produced an answer acy on the three most difficult problems was exceptionally
to the problem without being able to describe the solution low (0.50). In addition the level 3 subjects produced all erro-
procedure (Not identified). neous responses given under this strategy. In sum, the above
A second conclusion is that the strategies Concrete counting- results support the conclusion that once a strategy is found;
all and Take away were very infrequent, probably because no it can be used with about a constant level of accuracy,
physical props were available during testing. The absence independently of level of number-word sequence skill, given
of manipulatives did not prevent the children in the lowest that the strategy is not applied on unstable portions of the
number-word sequence skill level from trying to solve the number-word sequence. In other words, the data presented
problems, using all sorts of strategies. According to the counting- indicate that the detection of new and more efficient solution
types theory, children at this skill level should only be able strategies is the main factor responsible for the overall
to solve the problems by counting on physical objects. How- accuracy increase (see Table 1); the skill of applying a given
ever, the present results demonstrate that these children had strategy seems to play a secondary role. The detection of new
access to both object counting and number-word sequence strategies, finally, seems to be heavily related to an increased
strategies. number-word sequence skill, a finding difficult to explain in
The analysis of the accuracy data for the individual terms of the counting types theory.
strategies revealed very few significant results. No tests
were made of accuracy for Immediate recall because of ceiling Finger counting with the Sequence counting-all, Sequence
effects. For addition, accuracy varied with expertise level counting-on and Sequence counting-down strategies. As stated
for the strategy Sequence counting-on (a significant increase in the introduction, a significant feature distinguishing the
from level 2 to level 3 according to the Tukey HSD test); for counting types theory from the number-word sequence
subtraction, a similar finding was obtained for Decomposi- hypothesis is the role of objects. According to the former,
tion (a significant decrease from level 2 to 3 according to the the child has to construct countable unit items in order to
Tukey HSD test). This allows the preliminary conclusion solve arithmetic problems. In particular, counting on objects
that once a solution procedure has been found it can be used should be the only possible solution procedure for children,
at about a constant level of accuracy irrespective of level of who only have elaborated the number-word sequence to the
number-word sequence skill. To test the generality of this String or Unbreakable list levels. The relative use of object
conclusion, the two exceptions above were reanalyzed counting strategies, however, should decrease with the
including problem size in the analysis. elaboration of the number-word sequence (Fuson, 1992).
For Sequence counting-on there was a strong increase in The number-word sequence hypothesis, in contrast, states
response frequency from level 1 to level 2 (significant at the that children at a low elaboration level should be able to
0.05 level in a Tukey HSD procedure). Analyzing the prob- solve arithmetic problems by counting on the number-word
lem sizes on which Sequence counting-on was reported by sequence, as long as the addends do not exceed the stable
the level 1 children showed that about 50% of their reports portions of the sequence. Counting on objects should be
were given on the four most difficult problems, that is, prob- advantageous mainly as a means to monitor the count in the
lems with sum or minuend exceeding 10. In comparison, case of large addends. Therefore, the number-word sequence
about two-thirds of the Sequence counting-on reports by the hypothesis predicts that object counting should increase
level 2 children were given on the problems with minuend or with increasing addend size. To test this prediction problem

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd.


164 B. S. Johansson Scand J Psychol 46 (2005)

Table 2. Mean finger use (proportions) under the three strategies In sum, solution procedures with no signs of finger count-
Sequence counting-all, Sequence counting-on and Sequence count- ing were observed already for the level 1A children, contrary
down as a function of number-word sequence skill and problem size to what can be predicted from the counting types theory.
Furthermore, for both Sequence counting-on and Sequence
Strategy
Number-word sequence counting-down, finger use increased with increasing addend
skill and problem size SCA SCO SCD size, as predicted by the number-word sequence hypothesis.

Proportion of finger use


Level 1A Discussion
Size 1 0.56 0.17 0.50
Size 2 0.38 0.50 – The discussion will center on the arithmetic solution pro-
Level 1B cedures found at each number-word sequence skill level and
Size 1 0.56 0.20 0.40 on the transitions from less to more advanced solution pro-
Size 2 0.70 0.60 0.67 cedures. For level 1, Concrete counting-all was surprisingly
Level 2
infrequent, considering that the counting types theory pre-
Size 1 0.73 0.31 0.33
Size 2 0.82 0.37 0.71 dicts that this solution procedure should be the dominating
Level 3 and 4 alternative for children at a low level of number skill. The most
Size 1 0.83 0.09 0.36 frequent addition and subtraction strategies were Sequence
Size 2 1.00 0.52 0.63 counting-all and Take away, respectively, both strategies with
a low level of accuracy. Finally, the present children often used
Sequence counting-all with no signs of finger counting. That
size was included in the analysis. To obtain a sufficient Sequence counting-all was common on the addition problems
number of observations only two problem sizes were distin- is in agreement with the number-word sequence hypothesis.
guished: the two easiest problems with sum or minuend Their number-word sequence skill was in an initial stage;
below 10 (Size 1) versus the remaining four problems (Size they could solve 1 or 2 of the counting forward problems,
2), for the same reason performance was summed over levels none of the counting backward problems and often started
3 and 4. The analysis included only the strategies Sequence to count from the number-word “one”. Thus, their main
counting-all, Sequence counting-on and Sequence counting- method to solve the addition problems seems a straight-
down, because these strategies were the only ones, observed forward application of their number-word sequence skill. The
in sufficient numbers, under which the child may or may not use of Take away on subtraction problems, on the other
use finger counting. hand, agrees with the counting types theory in that minu-
To analyze initial finger use in detail the level 1 subjects end, subtrahend and result were represented by objects; in
were partitioned into subgroup 1A (n = 20), who did not the present case the fingers. In sum, the results indicate that
solve the practice problem “Count from 3 to 1” and sub- the children, already at this level of number-word sequence
group 1B (n = 18) who did. The data are given in Table 2. skill, have access to both object counting and number-word
The table presents mean proportion of finger use under each sequence strategies.
of the three strategies. Two-way analyses of variance were The main difference between the expertise level 1 and 2
run on each of the three strategy data series with problem children was a statistically reliable ( p < 0.05, Tukey HSD test)
size as a within-subjects variable and sequence skill level as increase in the use of Sequence counting-on and Sequence
a between-subjects variable. counting-down. Both the counting types theory and the number-
The analysis of the Sequence counting-all data showed that word sequence hypothesis predicts that development, because
finger use did not vary significantly with problem size, F(1, 54) it is not until level 2 that the children can count backward
= 1.05, p > 0.05, but increased with increasing sequence from a given to a given number-word.
skill, F(3, 32) = 15.47, p < 0.05, contrary to the counting A major advance from the two first to the third expertise
types theory. Levine et al. (1992) have also reported that level, as to sequence skill, was the ability to count backward
some children, at low expertise levels, can solve addition and from a number-word greater than ten, which indicates that
subtraction problems with no signs of concrete counting. the number-words in the initial sections of the sequence
For Sequence counting-on mean proportion of finger use are well differentiated. This agrees with the finding that
did not vary with number-word sequence skill, F(3, 65) = solving the problems by starting the count from the number-
0.19, p > 0.05, but was more frequent on the four largest word “one” (Sequence counting-all, Concrete counting-all
than on the two smallest problems, F(1, 96) = 6.95, p < 0.05. and Take away) almost disappeared. Instead Decom-
The same results were obtained for Sequence counting- position and Immediate retrieval strategies, in addition to
down, although with different F-values: F(3, 53) = 0.22, Sequence counting-on and Sequence counting-down,
p > 0.05 for sequence skill, and F(1, 74) = 6.40, p < 0.05 for started to appear in great numbers. These strategies certainly
problem size. In the latter analysis the level 1A data were deleted require a high skill of entering the sequence at arbitrary
because of lack of observations in the Size 2 condition. number-words.

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd.


Scand J Psychol 46 (2005) Sequence skill and arithmetic performance 165

Typical for the expertise level 4 children was that they the four first decades was taken to indicate a minimum grasp of
exhibited the ability to select the optimal strategy for each the decade structure. This test was considered an equivalent
though less time-consuming version of the counting forward test
problem. For example, they used Immediate recall on prob-
in Experiment 1. The last number-word reported in correct order
lems with small addends, decomposition by doubles when served as the score given to the child.
the difference between the two addends (or minuend and • Counting backward. The counting backward problems were count-
subtrahend) was small, and tens decomposition or Sequence ing from 3 to 1, 7 to 5, 13 to 8, and 22 to 16. Pilot tests indicated
counting-on/Sequence counting-down in the case of problems that the present children would probably not solve the two most
difficult problems in the former experiment. The same criterion
with large addends.
for correct solutions as in Experiment 1 was used: flawless count-
Some of the children combined the Sequence counting- ing from first to last given number-word. The sum score included
on/Sequence counting-down strategies with the Decomposi- performance on the 3 − 1 problem.
tion strategy. One example is the previously mentioned chil- • Doubles test. The experimenter introduced the word “doubles” by
dren who extended the fingers on one hand three times (5 + the problems “1 + 1 = ?” and “2 + 2 = ?” solving them together
with the child. Next, the child was given: “3 + 3”, “4 + 4”, “5 +
5 + 5) and then solved the 26 + 15 problem by counting-on.
5”, “6 + 6”, “7 + 7”, “8 + 8”, “9 + 9”, “12 + 12”, “13 + 13”, and
Another example is a boy who formed dice patterns of the “17 + 17”, pausing after each problem for the child to respond.
smaller addend. When solving the 15 + 26 problem he imag- Correct responding was defined as giving the correct answer
ined 15 as two dice of 6 and one die of 3, using Sequence without using fingers or other objects as solution aids.
counting-on to solve the problem. These transitional • Arithmetic tests. The same addition and subtraction word
problems were used as in Experiment 1. The problems were
solution procedures indicate how the Decomposition
presented verbally and no concrete objects or other solution aids
strategy may develop out of the Sequence counting-on/ were allowed.
counting-down strategies.
Distinguishing the expertise level 1 and 2 children from Procedure. The children were tested individually in a separate room
those at the two highest levels was that the former often with the child’s response recorded in a booklet. No post-solution
solved the problems by counting one-by-one on the number- strategy interviews were conducted. Testing was terminated when the
child had failed on two consecutive problems, or if the experimenter
word sequence, for one or both addends, whereas the latter
judged that the child’s highest performance had been attained. The
could segment the sequence, as manifested by the decompo- experimenter exerted considerable effort to establish an atmosphere
sition strategies. In particular, decompositions with doubles without strain and anxiety. The order of the tests was randomized.
increased from the former to the latter expertise levels; the
former group of children gave a total of 26 such strategy
reports, compared to 56 by the latter group. Therefore, the Results
discovery of the doubles regularity may be an important LISREL (Jöreskog & Sörbom, 1993) was used to test the
step in the development of the decomposition strategies. If hypothesis that the skill of solving doubles problems serves
this is the case, the ability to solve doubles problems should as a link between the number-word sequence skill and the
contribute to the ability of solving arithmetic problems. ability to solve arithmetic problems. In the LISREL language
Experiment 2 was run as a preliminary test of that hypothesis. Sequence skill, Doubles and Arithmetic ability are latent
variables that cannot be observed directly. In this study,
the first was operationalized by the forward and backward
STUDY 2 counting problems, the second by the doubles problems,
The purposes with Experiment 2 were to replicate the Study and the third by the addition and subtraction problems.
1 results and to test the hypothesis that the skill of solving In all three cases performance was measured as number
doubles problems plays an important role in the develop- of correct responses. This allowed the formulation of the
ment of the arithmetic ability. model in Figure 2 and the testing of whether this model
was consistent with the empirical data (Tuijnman &
Keeves, 1997).
Method LISREL provides a series of measures of the fit between
the theoretical model and the empirical data. The ones
Subjects. The subjects were 129 children: 93 were 5- and 6-year-old
recommended by Jöreskog and Sörbom (1993) indicate an
pre-school children and 36 were 7-year-old school children. Because
testing was done in late autumn, the school children had only had a almost perfect fit between model and data: ( χ2 d.f. = 3) =
few months of arithmetic teaching. In the pre-schools no formal 0.77, p = 0.86, root mean square error of approximation
arithmetic training took place. Parental permission was asked for all (RMSEA) = 0.0, and p-value for test of close fit (RMSEA)
children in the selected pre-schools and schools; of 133 parents, = 0.90.
129 permitted the testing of their child.
This result replicates Study 1, providing strong support
for the hypothesis that the doubles regularity is a structural
Tests
feature that children can detect and use as a solution aid,
• Counting forward. The child was asked to count forward until the given that the number-word sequence skill has reached a
experimenter said, “Stop” (at number-word 45). Counting beyond minimum level.

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd.


166 B. S. Johansson Scand J Psychol 46 (2005)

Fig. 2. LISREL-analysis of the results in Study 2.

CONCLUDING DISCUSSION the discovery of more advanced strategies (e.g., Siegler,


1991), that is, the child detects a new structural feature in
In the present study the answers to two problems were the number-word sequence that can easily be used to solve
sought: Do the initial arithmetic solution procedures agree arithmetic problems in new and more efficient ways. In
with the counting types theory or with the number-word conclusion, this line of reasoning is an argument for the
sequence hypothesis, and do the children use objects to con- importance of number-word sequence exercises in introductory
struct the addends or do they mainly use objects to monitor mathematics education.
the count? The results demonstrated that already children at The weakest spot in the present paper is that the represen-
a low level of numeral skill could use number-word sequence tational mechanism when counting on the number-word
procedures to solve arithmetic problems, a finding contrary sequence has not been specified. In comparison, the count-
to the counting types theory. Furthermore, counting on ing types theory has a clear answer to the representational
objects (in the present case, the fingers) mainly fulfilled the question: The numbers are represented as objects or images
role of monitoring the count. The Concrete counting-all and of objects. There is, of course, an obvious answer to this
the Take away strategies, which can be interpreted to mean criticism, namely representation by means of the numerals.
that the child makes use of objects to construct the addends, The empirical test of this speculation, however, will be the
occurred in too low frequencies to support the counting subject of a separate study. In addition, it may be argued
types theory (Steffe et al., 1983). that no control of potentially important contributing factors,
Thus, the results support a dual representation theory for example, phonological skills have been made. However,
(e.g., Uttal et al., 1999). The child may represent the numbers the results presented here can easily be followed up by such
either by means of objects or by means of the number-word parametric investigations.
sequence. The findings that representations by means of
objects were very infrequent and that the accuracy of object- This research was supported by a grant from Humanistisk-
counting strategies was low support the arguments against samhällsvetenskapliga forskningsrådet (Dnr F0579/2000).
early mathematics education leaning heavily on the manipu-
lation of concrete materials. For example, evaluations (e.g., REFERENCES
Resnick & Omanson, 1987) of the effects of having young Baroody, A. J. (1987). The development of counting strategies for
schoolchildren to solve mathematical problems with the help single-digit addition. Journal for Research in Mathematics Edu-
of Dienes (1960) blocks give no evidence of transfer of cation, 18, 141–157.
approaches learned when solving problems with manipulat- Carpenter, T. P., & Moser, J. M. (1983). The acquisition of addition
ives to the solution of similar problems involving written and subtraction concepts. In R. Lesh & M. Landau (Eds.),
Acquisition of mathematics: Concepts and processes (pp. 7–44).
symbols. New York: Academic Press.
The present results indicate a parallel phenomenon, in Carpenter, T. P., & Moser, J. M. (1984). The acquisition of addition
that object counting does not seem to evolve into more and subtraction concepts in grades one through three. Journal
advanced strategies, whereas such transitions were identified for Research in Mathematics Education, 15, 179–202.
in the case of the number-word sequence strategies. For Dienes, Z. P. (1960). Building up mathematics. London: Hutchinson.
Fuson, K. (1982). An analysis of the counting-on solution pro-
example, the Sequence counting-on/counting-down strat- cedure in addition. In T. P. Carpenter, J. M. Moser, & T. A.
egies require that the child can start counting, forward or Romberg (Eds.), Addition and subtraction: A cognitive perspective
backward, from an arbitrary point in the sequence, and the (pp. 67–81). Hillsdale, NJ: Lawrence Erlbaum Associates.
development from the Sequence counting-on/counting- Fuson, K. C. (1992). Research on whole number addition and
down strategies to Decomposition strategies seems to be subtraction. In D. A. Grouws (Ed.), Handbook of research on
mathematics teaching and learning (pp. 243–275). New York:
related to the ability to segment the sequence, with the Macmillan.
detection of the doubles regularity playing a significant Fuson, K. C., Richards, J., & Briars, D. J. (1982). The acquisition
role. Thus, the findings point to a possible mechanism in and elaboration of the number word sequence. In C. J. Brainerd

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd.


Scand J Psychol 46 (2005) Sequence skill and arithmetic performance 167

(Ed.), Progress in cognitive development research : Vol. 1 . Svenson, O., & Broquist, S. (1975). Strategies for solving simple
Children’s logical and mathematical cognition (pp. 33–92). New addition problems: A comparison of normal and subnormal
York: Springer-Verlag. children. Scandinavian Journal of Psychology, 16, 143–151.
Gal’perin, P. Ia. (1970). An experimental study in the formation of Svenson, O., Hedenborg, M.-L., & Lingman, L. (1976). On children’s
mental actions. In E. Stones (Ed.), Readings in educational heuristics for solving simple additions. Scandinavian Journal of
psychology (pp. 142–154). London: Methuen & Co Ltd. Psychology, 20, 161–173.
Gelman, R., & Gallistel, C. R. (1978). The child’s understanding of Tuijnman, A. C., & Keeves, J. P. (1997). Path analysis and linear
number. Cambridge, MA: Harvard University Press. structural relations analysis. In J. P. Keeves (Ed.), Educational
Ho, C. S.-H., & Fuson, K. C. (1998). Children’s knowledge of teen research, methodology, and measurement: An international hand-
quantities as tens and ones: Comparisons of Chinese, British, and book (pp. 621– 633). New York: Pergamon.
American Kindergartners. Journal of Educational Psychology, Uttal, D. H., Liu, L. L., & DeLoache, J. S. (1999) Taking a hard
90, 536–544. look at concreteness: Do concrete objects help young children
Jöreskog, K. G., & Sörbom, D. (1993). LISREL 8: Structural equ- learn symbolic relations? In L. Balter & C. S. Tamis-LeMonda
ation modeling with the SIMPLIS command language. Hillsdale, (Eds.), Child psychology: A handbook of contemporary issues
NJ: Lawrence Erlbaum Associates. (pp. 177–192). Ann Arbor: Taylor and Francis.
Levine, S. C., Jordan, N. C., & Huttenlocher, J. (1992). Develop- Wigforss, F. (1946). Barnens färdighet i räkning vid skolgångens början
ment of calculation abilities in young children. Journal of Experi- [Children’s counting skills at school entrance]. Häfte 191, Pedago-
mental Psychology, 53, 72–103. giska skrifter utgivna av Sveriges allmänna folkskollärarförenings
Martins-Mourao, A., & Cowan, R. (1998). The emergence of litteratursällskap. Göteborg: Elanders.
additive composition of number. Educational Psychology, 18, Wynn, K. (1998). Numerical conpetence in infants. In C. Donlan (Ed.),
377–390. The development of mathematical skills ( pp. 3 –25). East Sussex:
Pollio, H. R., & Whitacre, J. D. (1970). Some observations on the Psychology Press.
use of natural numbers by preschool children. Perceptual and
Motor Skills, 30, 167–174. Received 17 March 2003, accepted 3 October 2003
Resnick, J. B., & Omanson, S. F. (1987). Learning to understand
arithmetic. In R. Glaser (Ed.), Advances in instructional psycho- APPENDIX A
logy (pp. 41–96). Hillsdale, NJ: Erlbaum.
Riley, M. S., Greeno, J. G., & Heller, J. I. (1983). Development of Table A1. Addition and subtraction problems
children’s problem solving ability in arithmetic. In H. Ginsburg
(Ed.), The development of mathematical thinking (pp. 153–196). Lists
New York: Academic Press.
Secada, W. G., Fuson, K. C., & Hall, J. W. (1983). The transition Problems 1 2 3 4
from counting-all to counting-on in addition. Journal for
Research in Mathematics Education, 14, 47–57. Addition
Siegler, R. S. (1987). The perils of averaging data over strategies: 1. 1 + 4 4 + 1 2 + 3 3 + 2
An example from children’s addition. Journal of Experimental 2. 5 + 4 4 + 5 3 + 4 4 + 3
Psychology: General, 116, 250 –264. 3. 7 + 6 6 + 7 5 + 6 6 + 5
Siegler, R. S. (1991). Strategy choice and strategy discovery. Learning 4. 8 + 7 7 + 8 7 + 8 8 + 7
and instruction, 1, 89–102. 5. 6 + 18 18 + 6 6 + 17 7 + 16
Steffe, L. P. (2002). PSSM from a constructivist perspective. Paper 6. 15 + 26 26 + 15 13 + 28 14 + 27
presented at the Research Symposium Linking Research and the Subtraction
New Early Childhood Mathematics standards, held at the 2000 1. 5 − 2 5 − 4 5 − 1 5 − 3
NCTM Research Presession, Chicago IL. 2. 8 − 3 7 − 4 7 − 3 8 − 4
Steffe, L. P., & Cobb, P. (1988). Construction of arithmetical meanings 3. 12 − 5 12 − 7 13 − 6 13 − 6
and strategies. New York: Springer-Verlag. 4. 17 − 8 17 − 8 15 − 8 15 − 8
Steffe, L. P., von Glasersfeld, E., Richards, J., & Cobb, P. (1983). 5. 22 − 7 23 − 6 23 − 7 23 − 7
Children’s counting types: Philosophy, theory, and application. 6. 43 − 14 41 − 15 42 − 13 42 − 13
New York: Praeger Scientific.

© 2005 The Scandinavian Psychological Associations/Blackwell Publishing Ltd.

You might also like