You are on page 1of 9

Children's Conformity to the Cardinality Rule as

a Funetion of Set Size and Counting Accuraey

Karen C. Fuson and Geraldine C. Pergament


Northwestern University

Barbara G. Lyons
University of Illinois at Chicago

James W. Hall
Northwestern University

FUSON, KAREN C ; PERGAMENT, CERALDINE C ; LYONS, BARBARA C ; AND HALL, JAMES W. Chil-
dren's Conformity to the Cardinality Rule as a Function of Set Size and Counting Accuracy. CHILD
DEVELOPMENT, 1985, 56, 1429-1436. Young children must leam to count objects accurately, and they
must leam that the last word said in counting a set of ohjects tells how many objects there are in that
set (the Ccirdinality rule). In 3 experiments with preschoolers no difference across sets ranging from 2
to 19 ohjects was found in giving the last counted word as the answer to the question "How many
ohjects are there?" The relationship hetween children's response of the last counted word to a how-
many question and counting accurately varied with set size. For very small sets (2, 3,4), significantly
more children counted accurately and did not give last-word responses than vice versa. For sets from
4 to 7 ohjects, approximately equal numhers of children demonstrated only 1 of these ahilities and
not the other. For sets fTom 9 to 19, significantly more children gave last-word responses and did not
count accurately than vice versa. Children did not exhibit difficulty in coordinating last-word re-
sponding and counting accurately.

The three experiments reported here mean that the child knows the cardinality
were designed to examine how young chil- rule. An altemative possibility, pointed out
dren's conformity to the cardinality mle is af- earlier by Fuson and Hall (1983), is that the
fected by the size of the sets counted and by child has leamed a simpler mle: the last
counting accuracy. The literature presently counted word is the appropriate response to a
contains conflicting results about both size of how-many question. This simpler mle is a
sets and counting accuracy (see below). necessary but not sufficient part of fhe cardin-
ality rule, which also requires knowing that
The cardinality rule (Schaeffer, Eggles- fhe answer to the how-many question refers
ton, & Scott, 1974), or cardinality principle to fhe set as a whole and to die numerosity of
(Gelman & Gallistel, 1978), refers to the the set. Typical testing conditions for the car-
understanding that the final number-word dinality mle have not addressed the child's
spoken in counting a set of objects denotes knowledge of these cardinal attributes, and
the cardinality (numerosity) of tiiat set. Chil- thus those conditions are not sufficient to dis-
dren's conformity to the cardinality rule has tinguish between knowledge of the simpler
been tested by having children count a set of how-many rule and knowledge of the cardin-
objects and then answer the question "How ality rule. Because of that ambiguity, fhe more
many objects are there?" A response with the neutral and operational term "last-word re-
last word used in counting has been taken to sponse" will be used in this paper for a child's

We wish to thank Mrs. Elaine Spanier, the director, and the teachers, staff, and children of the
B'nai Torah Nursery School, Highland Park, IL, and the teachers and children of the Baker Demon-
stration School, Evanston, IL. Special thanks go to Lloyd Kohler and Cinny Neal—wonderful
experimenters with children. The flrst experiment in this paper is based on a Ph.D. dissertation
submitted hy the second author to the School of Education, Northwestern University. The second
experiment and an earlier draft of the paper were completed while the flrst author was a Visiting
Scholar at the Wisconsin Center for Education Research, University of Wisconsin. Parts of this paper
were presented at the Piaget Society Annual Meeting, Philadelphia, June 1983. Requests for re-
prints should he sent to Karen C. Fuson at the School of Education, Northwestern University,
Evanston, IL 60201.
[Child Development, 1985, 56, 1429-1436. © 1985 by the Society for Research in Child Development, Inc.
All rights reserved. 0009-3920/85/5606-0007$01.00]
1430 Child Development
answering a how-many question with the last size 3 through 26, though last-word responses
counted word. were slightiy higher for sets of 3 than for 6
and were slightiy lower for sets of 19 and 26
The evidence is presently mixed with re- than for 6, 9, and 12.
spect to the effect of the size of the set
counted on the last-word response. Schaeffer Several different positions also exist in
et al. (1974) reported a considerably higher fhe literature with respect to the relationship
last-word response rate for the small sets of 2, between counting accurately and last-word
3, and 4 than for sets of 5, 6, and 7; that is, responding. Two studies suggest that count-
they reported a decrease of the last-word re- ing accurately precedes giving last-word re-
sponse with an increase in set size. However, sponses. Schaeffer et al. (1974) proposed that
Schaeffer et al. included as correct last-word children integrate their use of pattem recog-
responses the many cases for the small sets in nition of small numbers (subitizing) and the
which a child did not count the set but simply counting procedure to form the cardinality
recognized the small set as a pattem and an- rule and then extend the use of the cardinality
swered the how-many question based on this rule to numbers larger than four only when
pattem. Ginsburg and Russell (1981) reported their counting of such numbers is accurate.
the opposite result, that the last-word re- Consistent with this prediction, Schaeffer et
sponse did not seem to be related to the size al. found many children who could count ac-
of the set counted over sets of 3, 5, 8, and 11. curately but did not give the last-word re-
However, some of their children also used sponse, while they did not find children who
pattern recognition, guessing, and other strat- showed the reverse pattem (did not count ac-
egies rather than counting before being given curately but did give the last-word response).
the how-many question, so it is not clear
whether these children could give the last- Gelman and Gallistel (1978) identified
word response. Thus, neither of these studies three how-to-count principles: the cardinality
provides entirely clear data on this issue. principle and two principles that underlie ac-
Celman and Gallistel (1978) observed a curate counting—the stable-order principle
very sharp decline in the extent to which chil- and the one-one principle. The cardinality
dren gave the last-word response as sets in- principle was credited if a child gave last-
creased from 2 through 19. However, their word responses (or exhibited behaviors from
procedure had two attributes that may have which last-word responses could be inferred),
resulted in underestimating last-word re- the stable-order principle was credited if a
sponses. First, the children were not explic- child produced counting words in the same
itiy given a how-many question on each of the stable order from trial to trial, and the one-one
trials on which last-word responses were as- principle was credited if a child gave each
sessed. Instead, on some trials the children counted object one and only one counting
were just told to "Count fhem" or to "Co word. Gelman and Gallistel reported that
ahead." Therefore, on these trials the child's children produce counting words in a stable
volunteering of the last-word response was as- order and count one-one before they give a
sessed rather than the child's ability to give last-word response and that, as set size in-
the last-word response when asked. Second, creases, children stop giving a last-word re-
when a how-many question was used, it was sponse before they stop saying words in a sta-
given only before counting began. Thus it ble order and stop counting one-one (also see
seems possible that a child might forget the Gelman, 1982, and Gelman & Baillargeon,
how-many question during the course of 1983).
counting. Such forgetting might be especially A recent study (Celman & Meek, 1983)
likely with larger sets where more time and indicated that on large sets children are cor-
effort would be required for counting than rect in judging errors made by others in giv-
with small sets; such differential forgetting ing last-word responses even though fheir
that a how-many question had been asked own performance on those sets is incorrect in
could have contributed to the reported de- some way (either in counting accuracy or in
crease in last-word responses with an increase giving last-word responses). However, in that
in set size. study, errors in giving last-word responses
Wilkinson (1984) used a procedure in were not separated from errors in counting.
which children were explicitiy asked the Therefore it is not clear whether the giving of
how-many question on each trial and were last-word responses declined with increasing
asked the question after the counting was set size nor is it clear whether last-word re-
completed. He reported little effect of set size sponses were given regardless of counting ac-
on children's last-word responses over sets of curacy.
Fuson et al. 1431
Two studies failed to find a dependency For each trial, the board containing the
of last-word responses on accurate counting. row of objects was placed on the table in front
Cinsburg and Russell (1981) reported that of fhe child. Ghildren were told to count the
preschool children's last-word responses on objects, and after the counting was completed
sets of 3, 5, 8, and 11 did not seem to depend were asked, "How many X's are there?"
on counting accurately. However, this result where X was the type of object counted (ani-
might have been affected by the overgener- mals, dolls, soldiers, or blocks). If the child
ous criterion for last-word responses dis- counted again rather than responding to the
cussed above with respect to set-size effects. how-many question with a single number
Wilkinson (1984) did observe a relationship word, the how-many question was repeated at
between counting accuracy and last-word re- the end of this second count in order to en-
sponses, and the nature of that relationship sure fhat a failure to volunteer a single-word
depended on set size. For sets of 3, there response at the end of this second count was
were many more trials on which counting was not just due to forgetting that a how-many
accurate and last-word responses were not question had been asked. The experimenter
given than vice versa, whereas for sets of 6 recorded fhe number words said by each
and more there were many more trials on child, marked on a prepared drawing any ob-
which last-word responses were given and jects that did not receive exactiy one word
counting was inaccurate than vice versa. Wil- when the child was counting, and recorded
kinson proposed that counting skills may the child's response to the how-many ques-
emerge before last-word responses emerge, tion.
but that the latter may reach functional matu-
rity sooner. Furthermore, he suggested that Each child received six small sets, two of
children may have trouble coordinating their each size; 7 (or, occasionally, 10) days later
last-word responding with fheir execution of each child received six of the larger arrays,
accurate counting. Thus, even after both have two of each size. Order of set size was not
been acquired, only one may be displayed on counterbalanced in fhis first experiment be-
any given trial, especially on trials involving cause we did not want to discourage children
counting larger sets. This possibility was ex- by giving the larger arrays first. (This was also
amined in the present experiments. the procedure used by Celman and Gallis-
tel—their sets were given in increasing order
of size.)
In Experiment 1, three questions were
addressed: (1) Do last-word responses de- Results and Discussion
crease in frequency from sets just at and be- A child was scored as giving a last-word
yond the pattem recognition size (4, 5, 6) to response on a given set if the answer to the
sets considerably larger (9, 12, 14)? (2) How is how-many question was fhe last word the
last-word responding related to counting ac- child had said in counting that set. Note that
curately? and (3) To what extent do children this counting did not have to be accurate. A
have difficulty coordinating accurate counting child was classified as "an accurate counter"
and last-word responding? The same three is- if the child counted accurately on all of the
sues were addressed in the second experi- trials and did not give a last-word response on
ment for very small pattern-recognizable sets all of the trials, and was classified as "a last-
(2, 3, 4) and sets just larger than this (5, 6, 7) word responder" if the child gave a last-word
and, in the third experiment, for larger num- response on all of the trials and did not count
bers (16, 18, 19). accurately on all of the trials. For all of
the experiments, fhe analyses using these
classifications were also done using the more
Experiment 1 lenient criterion of correct performance on
Method two-thirds rather than on all of the trials. In all
Suhjects.—Forty-eight middle-class chil- but two cases the results of these analyses
dren ranging in age from 3-2 to 4-7 (mean 4-0) were consistent with those done with the
attending preschool in a suburb of Chicago stricter criterion and, therefore, are not re-
served as subjects. ported here. The results with both criteria are
reported for the two exceptional cases.
Materials and procedure.—The smaller To facilitate comparisons vi^ith the earlier
arrays contained 4, 5, or 6 objects; the larger Gelman and Gallistel results, in all experi-
contained 9, 12, or 14. Objects used were ments all analyses were redone with counting
small toys (animals, dolls, and soldiers) and 1- being classified as accurate if it satisfied the
inch wooden blocks. Objects were arranged one-one and the stable-order principles (i.e.,
in a single row and glued to a board. the use of a consistent but incorrect sequence
1432 Child Development
of number words was allowed). This re- (2), x^(l,20) = 12.8, p < .01. Thus, for larger
classification did not change any significant set sizes, giving last-word responses seems
results. Thus, in fhe following report and dis- developmentally to precede rather tiian fol-
cussion sections, relationships between accu- low the ability to count tiiose sets accurately.
rate counting and last-word responses also There was littie evidence that children
refiect relationships between use of the stable were experiencing difficulty coordinating
order and one-one principles and the use of counting accurately and giving last-word re-
the cardinality principle. sponses. If there is competition between
Turning to fhe data, children gave a last- these two aspects of counting, one would ex-
word response on 70% of the smaller and on pect fhat children would at times display each
67% of fhe larger arrays. Thus, there was no of them without the other, especially for
significant decrease in last-word responses larger sets where the competition might be
wifh increasing set size. As in earlier studies, intensified. In fact, for the smaller sets only
counting was much more accurate for the one out of the 48 children sometimes showed
smaller than for the larger sets: 75% vs. 44% each aspect of counting without the other
sets counted accurately. (counted accurately and did not give a last-
word response on one trial and on a different
Children were fairly consistent in giving trial gave a last-word response and did not
last-word responses across sets (see Table 1). count accurately). For the larger sets, only
Across all 12 trials of the six different sets, four of the 48 children showed each aspect of
85% of the children either consistentiy gave counting without the other.
or consistentiy did not give a last-word re-
sponse on 11 of the 12 trials. Experiment 2
On the smaller set sizes (4, 5, 6), children To test the conclusion of Schaeffer et al.
counted accurately and gave last-word re- (1974) that children first give the last-word re-
sponses to about the same extent (75% and sponse on very small pattern-recognizable
70%). However, an examination of the perfor- sets, in Experiment 2 last-word response per-
mance of individual children revealed fhat formance on sets of 2, 3, and 4 objects was
many children consistentiy demonstrated compared to that on 5, 6, and 7 objects. For
only one of these skills and not the other. each set size, last-word responding was also
Seven children were accurate counters and compared to accuracy of counting. Younger
13 children were last-word responders. children (2-year-olds) were also included in
McNemar's test indicated that the number order to see whether there were any differ-
of last-word responders did not differ ences in the relationships due to age.
significantiy from the number of accurate
counters. Thus, for sets of 4, 5, and 6 neither The process that Schaeffer et al. called
of fhese aspects of counting seems to precede pattem recognition is sometimes called sub-
the other developmentally, and each oft^n ap- itizing: the immediate apprehension of the
pears without the other. For the larger sets numerosity of small sets, generally sets of four
(9, 12, 14), McNemar's test indicated that or fewer objects. It is not clear just what is
significantiy more children (18) were last- involved in this process, and different re-
w^ord responders than were accurate counters searchers have made somewhat different pro-

TABLE 1
NUMBER OF CHILDREN CIVING LAST-WORD RESPONSES ACROSS TRIALS (Experiment 1)

NUMBER OF TRIALS OVERALL OVERALL


LAST-WORD COUNTING
SIZE OF SET 6 5 4 3 2 1 0 RESPONDING (%) AGGURACY(%)

4 5 6 31 9, 0 1 0 9. 12 70 75
9, 12, 14 26 4 3 1 1 0 13 67 44

NUMBER OF TRIALS

12 11 10 9 8 7 6 5 4 3 2 1 0

4, 5, 6, 9, 12, 14 .. .. 24 5 1 2 0 1 1 1 1 0 0 1 11 69 60
Fuson et al. 1433
posals about this (Beckwith & Restie, 1966; Results and Discussion
Chi & Klahr, 1975; Klahr & Wallace, 1976; Because some authors consider only sets
Mandler & Shebo, 1982; Schaeffer et al., of 2 and 3 to be subitizable by very young
1974). The term will be used here in its oper- children, performance on sets of 2 and 3 was
ational sense, with no position implied con- also compared to performance on sets of 5
ceming just what is involved in fhis process. and 7 (sampling from the nonsubitizable
sets). Last-word responding did not differ
Method significantiy across the subitizable and non-
The subjects were 45 middle- to upper- subitizable sets for either age group, regard-
middle-class children attending a preschool less of whefher subitizable sets were consid-
in Madison, Wisconsin. There were 19 2-year- ered to be sets of 2, 3, and 4 or just sets of 2
olds ranging from 2-5 to 2-11 (mean age 2-8) and 3 (see Table 2). Thus no support was
and 26 3- and 4-year-olds ranging from 3-0 to found for the Schaeffer et al. suggestion that
4-3 (mean age 3-6). children first give last-word responses for sub-
The stimuli were small objects (animals, itizable sets and fhen later give such re-
girl dolls, boy dolls, trucks, pennies, blocks, sponses for nonsubitizable sets.
and pigs). Six trials were given on linearly As in Experiment 1, the children were
arranged objects of set sizes 2 through 7. Sets fairly consistent in their giving of last-word
of size 2, 3, and 4 (subitizable sets) were alter- responses. The distribution in Table 2 is
nated wifh sets of 5, 6, and 7 (nonsubitizable bimodal, with 87% of the sample consistentiy
sets). Half of the sample received a subitiz- giving or consistently not giving last-word re-
able set first and half received a nonsubitiz- sponses on five of the six trials.
able setfirst.The questioning procedure and
recording procedures were identical to those The results comparing accurate counting
used in the first experiment. The experi- and last-word responding by age and set size
menter was naive with respect to the issues in are presented in Table 2. A 2 (age: 2- vs. 3-
this experiment. and 4-year-olds) X 2 (skill: last-word respond-

TABLE 2
NUMBER OF CHILDREN CIVING LAST-WORD RESPONSES AGROSS TRIALS (Experiment 2)

NUMBER OF TRIALS OVERALL OVERALL


LAST-WORD COUNTING
AGE 3 2 1 0 RESPONDING (%) AGGURAGY(%)

2-year-olds:
Set sizes 2 3 4 1 3 2 13 19 49
1 2 9. 14 16 21
3- and 4-year-olds:
Set sizes 2 3 4 16 0 3 7 65 92
Set sizes 5 6 7 . . 13 2 5 6 63 69

NUMBER
OF TRIALS

2 1 0

2-year-olds:
Set sizes 2 3 2 4 13 21 53
Set sizes 5 7 . . . . 3 2 14 21 16
3- and 4-year-olds:
Set sizes 2 3 15 4 7 65 92
Set sizes 5 7 . . . . 14 5 7 63 67

NUMBER OF TRIALS

6 5 4 3 2 1 0

2-year-olds 0 2 1 1 1 1 13 18 31
3- and 4-
12 2 2 0 1 5 4 64 81
1434 Child Development
ing vs. accurate counting) X 2 (set size) analy- difficulty coordinating accurate counting and
sis of variance with repeated measures for the last-word responses. Only one child some-
last two factors indicated that the main effect times demonstrated each skill without the
of age apparent in the Table (the older chil- other. Thus, the results from Experiment 1 on
dren did considerably and uniformly better in the relationship between counting accuracy
all cells) was significant, F(l,43) = 29.35, p < and last-word responding for sets just beyond
.001. The interaction between set size and the subitizing range were replicated here.
skill was significant, F(l,43) = 16.11, p <
.001. More subitizable sets were counted ac- Experiment 3
curately fhan were given last-word responses, In Experiment 1, last-word responses did
while the nonsubitizable sets were counted not decrease with set size and did not depend
accurately and given last-word responses to on counting accuracy for sets up to 14. In Ex-
fhe same extent. periment 3, somewhat larger set sizes were
Children who counted fhe subitizable examined and the order of set size was coun-
sets accurately could have answered the how- terbalanced. The smaller sets were given first
many question by subitizing the objects (the in Experiment 1, and there seemed to be
objects were not covered) rather than by giv- some chance that this might have induced
ing their last counted word. Such subitized more last-word responses on the larger sets.
responses would be quite difficult to distin- Method
guish from last-word responses, since the sub- The subjects were 24 middle- to upper-
itized answer would be the same as their last middle-class children between the ages of 3-6
counted word. However, only two children and 4-5 who attended a preschool in a differ-
showed response pattems that possibly re- ent suburb from fhat used in the first experi-
fiected such subitizing, fhat is, they each gave ment. The experimenter was also different
a last-word response on one subitizable set from the early experimenters and was un-
and no such responses on a nonsubitizable aware of the issues in this experiment.
set. Thus there was very littie evidence of
children subitizing after they had counted a The counting stimuli were irregularly
set. spaced red or blue dots arrayed in a row.
Cardboard strips contained 7, 8, 9, 16, 18, or
The analysis of perfonnance within indi- 19 dots. Smaller and larger strips were alter-
vidual children indicated no age difference in nated. Half of the children received a smaller
the proportion of children demonstrating only number (7, 8, 9) first. Two fixed orders of pre-
one aspect of counting and not the otiier. sentation were used. The questioning proce-
Therefore the data were pooled across age for dure, recording, and scoring were like those
further analyses. For the subitizable sets (2, 3, used in the first experiment.
4), there was littie evidence that children
were having trouble coordinating last-word Results and Discussion
responding and accurate counting. Only two There was no set-size difference in the
children sometimes exhibited each of these giving of last-word responses: children gave
skills without the other. McNemar's test at the last-word responses on 92% of the smaller and
more lenient two-fhirds criterion indicated on 92% of the larger sets. Counting accuracy
that significantiy more children (16) were did differ by set size; children counted accu-
accurate counters than were last-word re- rately on 78% of the sets of 7, 8, and 9 objects
sponders (1), x^(l,17) = 13.24, p < .01. At tiie and on 19% of tiie sets of 16, 18, and 19 ob-
stricter criterion of use of a skill on all three jects.
trials, the respective numbers of single-skill
users were in the same direction (8 and 2), but Children's last-word responding was per-
McNemar's test was not significant. Thus, ap- fectiy consistent: 22 children gave such re-
parentiy children typically count subitizable sponses on all six sets and two children never
sets fairly, though not necessarily perfectiy, gave them. Thus there was no evidence of
accurately before they give a last-word re- difficulfy in coordinating accurate counting
sponse for these sets. and last-word responding.
The above relationship did not hold for At both set sizes the relationship be-
the nonsubitizable sets (5, 6, 7). For these the tween last-word responses and counting accu-
nonsignificant McNemar test indicated that racy was as it had been for the sets of 9, 12,
the number of accurate counters (six children) and 14 in Experiment 1. For both the smaller
did not differ significantiy from the number of and larger sets with the strict criterion,
last-word responders (six children). There McNemar's test indicated that significantiy
was littie evidence of children experiencing more children were last-word responders
Fuson et al. 1435
thian were accurate counters, 8 versus 1 and rule. Children did not give more last-word re-
19 versus 0, x^(l,9) = 5.44, p < .05, and x^(l, sponses on subitizable sets than on non-
19) = 19.00, p < .01, respectively. For tiie subitizable sets regardless of whetiier 2, 3,
smaller sets at the more lenient two-thirds cri- and 4 or just 2 and 3 were considered to be
terion, the number of single-skill users did subitizable sets. Individual response pattems
not differ significandy but the direction of the also revealed littie evidence of subitizing to
difference was the same. answer the how-many question.
These experiments indicate that few chil-
General Diseussion dren evidence any difficulty in coordinating
The three experiments here showed no accurate counting and giving a last-word re-
difference in the extent to which last-word re- sponse. Rather, for smallish sets, a substantial
sponses are given across sets ranging from 2 number of children exhibit one but not the
to 19 objects. Over set sizes and over repeated other of these. Thus neither of these aspects
trials of fhe same set size, individual children of counting requires the other. Why some
were fairly consistent in their last-word re- children begin to count accurately before they
sponding. Thus, last-word responses do seem leam that the last counted word answers fhe
to be based on a mle or a principle that, once how-many question while others leam this
leamed, is applied over a wide range of set last-word response before counting accurately
sizes. is an interesting question for future research.
The relationship between giving last- Taken together, these results suggest that
word responses and counting accurately initially the last-word response may not
varies with set size. For fhe very small sets of necessarily be linked to notions of cardinality.
2, 3, and 4, children ordinarily count accu- First, after counting, few children seem to
rately before they give last-word responses. subitize very small sets to answer the how-
For sets of 4 through 7, many children count many question. Second, that giving last-word
accurately before they give last-word re- responses does not depend on accurate count-
sponses, but a roughly similar proportion give ing reveals a lack of understanding that the
last-word responses before they count accu- specific numerosity of a set depends on accu-
rately. For sets larger than this, many children rate counting (and/or reveeJs a conviction that
give last-word responses without counting ac- one's counting is always correct). Third, even
curately and almost no children do the re- though it would seem to be a bit more
verse. This reversal with set size in the rela- difficult to conceptualize a numerosity of a
tionship between last-word responding and large set than of a much smaller set, no set-
accurate counting is consistent with the size difference was found for giving last-word
findings of Wilkinson (1984), except that in responses. Thus, the last-word response may
that study no set sizes were found with initially be for many children only a rule
roughly equivalent demonstrations of these about the answer to a how-many question (a
skills. This may be because half of the count- how-many rule) rather than a cardinality rule.
ing tasks in that study were more difficult
than those used here. Referenees
The use of a criterion that assessed count- Beckwith, M., & Restie, F. (1966). The process of
ing as acctirate if the counting satisfied the enumeration. Psychological Review, 73, 437-
stable-order principle and the one-one princi- 444.
ple resulted for each experiment in the same Chi, M. T. C , & Klahr, D. (1975). Span and rate of
pattem of results as the criterion of fully accu- apprehension in children and adults. Joumal
rate counting. Thus, these experiments indi- of Experimental Child Psychology, 19, 434—
cate that the relationships among Celman and 439.
Gallistel's three how-to-count principles vary Fnson, K. C , & Hall, J. W. (1983). The acquisition
by set size in the manner discussed above. In of early number word meanings: A conceptual
particular, for sets larger than 8, many chil- analysis and review. In H. Cinsburg (Ed.), The
dren give last-word responses but do not ex- development of children's mathematical think-
hibit both the stable order and one-one prin- ing (pp. 49-107). New York: Academic Press.
ciples, whereas the reverse is rarely observed. Celman, R. (1982). Basic numerical ahilities. In R.
Stemherg (Ed.), Advances in the psychology of
The present evidence did not provide human intelligence (pp. 181-205). Hillsdale,
support for the Schaeffer et al. (1974) pro- NI: Erlhaum.
posed hierarchic integration of pattem recog- Celman, R., & Raillargeon, R. (1983). A review of
nition of small numbers (subitizing) with the some Piagetian concepts. In I. H. Flavell & E.
counting procedure to form the cardinality Markman (Eds.), P. H. Mussen (Series Ed.),
1436 Child Development
Handbook of child psychology: Vol. 3. Cogni- Klahr, D., & Wallace, I. C. (1976). CognUive de-
tive development (pp. 167-230). New York: Wiley. velopment: An information-processing view.
Celman, R., & Callistel, C. R. (1978). The child's Hillsdale, NI: Erlhaum.
understanding of number. Camhridge, MA: Mandler, C , & Sheho, B. J. (1982). Suhitizing:
Harvard University Press. An analysis of its component processes. Jour-
Celman, R., & Meek, E. (1983). Preschoolers' nal of Experimental Psychology: General, 3,
counting: Principles hefore skill. Cognition, 1-22.
13, 343-359. Schaefifer, B., Eggleston, V. H., & Scott, J. L. (1974).
Cinshurg, H. P., & Russell, R. L. (1981). Social class Number development in young children. Cog-
and racial influences on early mathematical nitive Psychology, 6, 357-379.
thinking. Monographs of the Society for Re- Wilkinson, A. C. (1984). Children's partial knowl-
search in Child Development, 46(2, Serial No. edge of the cognitive skills of counting. Cogni-
193). tive Psychology, 16, 28-64.

You might also like