Content uploaded by Layla Unger
Author content
All content in this area was uploaded by Layla Unger on Jul 30, 2018
Content may be subject to copyright.
All Creatures Great and Small: Category-Relevant Statistical Regularities in
Children’s Books
Layla Unger (unger.114@osu.edu)
Ohio State University, Department of Psychology, 1835 Neil Avenue
Columbus, OH 43210 USA
Anna V. Fisher (Fisher49@andrew.cmu.edu)
Carnegie Mellon University, Department of Psychology, 5000 Forbes Avenue
Pittsburgh, PA 15214 USA
Abstract
Sensitivity to statistical co-occurrence regularities is present
from infancy. This sensitivity may contribute to learning in
many domains, including category learning. However, prior
research has not examined whether everyday input conveys
category-relevant statistical regularities. This study assessed
whether statistical regularities relevant to real-world categories
are present in a commonly experienced source input –
children’s picture books. We focused on animal categories
because this is a domain in which children receive much
exposure from an early age, while simultaneously holding
persistent misconceptions about category membership beyond
preschool years. Analysis of 80 books revealed that they: 1)
Were likely to contain regularities from which individual
species categories (e.g., “chicken”) might be learned, but 2)
Were unlikely to contain regularities from which broader
taxonomic categories (e.g., “bird”) might be learned. These
findings point to a paucity of taxonomically-relevant statistical
regularities that may contribute to persistent taxonomic
misconceptions.
Keywords: Cognitive Development; Semantic Knowledge;
Semantic Development; Category learning
Introduction
Over the course of development, we are faced with the
challenge of acquiring an extensive body of knowledge about
the world. For instance, we must learn to classify the entities
we perceive around us into meaningful categories, segment
continuous visual input into events, and acquire our ambient
language. Research in the field of cognitive development
over recent decades has highlighted how these feats of
knowledge acquisition may be facilitated in part by a
sensitivity to statistical regularities in the environment. In
some domains, this research has included investigations into
the nature of the statistical regularities that are available in
the environment for learners to pick up on. For example, in
the domain of language acquisition, studies have investigated
the degree to which spoken and written language input
contain statistical regularities to which learners are sensitive,
such as the reliable co-occurrence of speech sounds (Mattys
& Jusczyk, 2001; Pelucchi, Hay, & Saffran, 2009) In
contrast, in other domains where sensitivity to statistical
regularities may play a critical role in knowledge acquisition,
such as category learning, we know little about the statistical
regularities present in the environment.
The overall aim of the present study was to investigate the
degree to which statistical regularities that are relevant to
real-world category learning are actually present in the
environmental input that learners receive, starting early in
development. To conduct this study, we made a set of
evidence-based choices about which real-world categories,
statistical regularities, and sources of environmental input to
investigate.
First, several factors prompted us to focus on statistical
regularities that may foster knowledge of taxonomic
categories of animals, such as “bird”, “mammal”, and
“reptile”. We focused on taxonomic categories because they
are cognitively useful for processes such as inductive
reasoning and integrating new information into memory. The
utility in inductive reasoning comes from the fact that
membership in a taxonomic category tends to reliably predict
the properties that an entity possesses (e.g., if an animal is a
mammal, it will have a four-chambered heart), and therefore
facilitates inferences about such properties without the need
for direct observation (e.g., Heit, 2000). Moreover, recent
evidence suggests that taxonomic category knowledge may
facilitate the integration of newly-learned information about
members of the category into memory (Pinkham, Kaefer, &
Neuman, 2014).
We specifically focused on taxonomic categories the
animal domain in part because it is one in which
environmental input is common and familiar from an early
age. Moreover, the possibility that statistical regularities can
shape category knowledge in the animal domain is supported
by evidence that children show earlier and stronger
knowledge that animals from the same taxonomic category
are related to each other when those animals also co-occur,
versus when they do not (Unger, Fisher, Nugent, Ventura, &
MacLellan, 2016). However, evidence from other lines of
research suggest that the statistical regularities children
experience in the animal domain may often instead lead them
to form an inaccurate understanding of taxonomic category
composition. Specifically, findings from multiple learning
sciences studies suggest that children possess persistent
misconceptions about taxonomic category membership in the
domain of animals that is suggestive of the contributions of
learning from statistical regularities. In comparison to
children’s robust knowledge of the habitat groups within
which animals regularly co-occur (e.g., farm animals, water
animals, etc.) (Crowe & Prescott, 2003; Kattmann, 1998,
2001; Storm, 1980), their knowledge of taxonomic category
membership (e.g., mammal, bird, reptile, etc.) is often poor.
For example, in a large-scale study of over 450 U.S. students
from elementary school, middle school, high school, and
college, Trowbridge and Mintzes (1988) found that
substantial numbers of students throughout the age-range
miscategorized many animals, such as identifying all animals
that live in water as fish, including whales and marine
invertebrates. Such misconceptions persisted even in college
students who majored in biology (e.g., 20% of biology majors
identified whales as fish). Such misconceptions appear cross-
cultural (Yen, Yao, & Chiu, 2004). Taken together, these
findings suggest that children form categories of animals in
part based on the regularities with which different animals
co-occur, and that such categories are often inconsistent with
taxonomic category membership. By the same token, these
findings suggest that the statistical regularities that children
experience in the animal domain are not informative about
the composition of taxonomic category membership.
Second, we chose as our statistical regularity of interest the
regularity with which a set of perceptual inputs co-occur
more reliably with each other than with others. Throughout
the remainder of this paper, we refer to these as “co-
occurrence regularities”. This choice was motivated in part
by prior research that has shown that a sensitivity to this
regularity is present from very early in development
(possibly birth, e.g., Bulf, Johnson, & Valenza, 2011).
Additional motivation for this choice came from cognitive
neuroscience, computational modeling and behavioral
evidence that co-occurrence may foster category knowledge
by increasing the similarity of mental representations of co-
occurring entities (Huebner & Willits, 2017; Schapiro,
Kustner, & Turk-Browne, 2012; Schapiro, Rogers, Cordova,
Turk-Browne, & Botvinick, 2013).
Finally, the source of environmental input in which we
investigated animal taxonomic category-relevant co-
occurrence regularities was children’s picture books. One
motivation for this choice was evidence that picture books
represent a common source of environmental input from very
early in development (e.g., Young, Davis, Schoen, & Parker,
1998) This choice was also motivated by evidence that
children’s picture books are a key source from which children
learn about animals (Ganea, Ma, & DeLoache, 2011). Lastly,
we focused on children’s books because they readily lend
themselves to quantifying co-occurrence regularities by
recording which items co-occur on the same page.
Present Study
For the reasons outlined above, we investigated the degree to
which category-relevant statistical regularities are present in
environmental input by measuring the degree to which
animals from the same taxonomic category are likely co-
occur on the same page in a large sample of children’s picture
books. Moreover, we analyzed the degree to which such co-
occurrences are likely to be informative about the
composition and breadth of taxonomic categories.
Specifically, evidence from prior category learning research
(Oakes, Coppage, & Dingel, 1997; Perry, Samuelson,
Malloy, & Schiffer, 2010) suggests that examples of many
different items from the same category are more informative
about the composition and breadth of that category than are
multiple examples of the same item. Therefore, we tested
whether co-occurrences between members of the same
taxonomic category are primarily between animals of the
same species (e.g., multiple cows depicted on the same page),
or are also between diverse members of the category (e.g.,
multiple different kinds of mammal depicted on the same
page). Finally, to investigate whether the nature of animal
category-relevant statistical regularities to which learners are
exposed changes with age, we measured the relationship
between the regularities present in children’s books and the
target age range for which each book was written.
Methods
Book Sample
The sample of books analyzed consisted of 80 children’s
books about or featuring animals. To afford an evaluation of
co-occurrences between depicted animals, only books in
which multiple pages depicted more than one animal were
included in the sample. To ensure that books were
representative of those to which children are typically
exposed, books were selected on the basis of librarian
recommendations, circulation statistics for children’s books
from the local public library, and amazon.com best sellers.
The selection of books was additionally guided by a
consideration of the diversity of books to which children may
be exposed. Specifically, books were selected to assemble a
sample that included both fiction and non-fiction books
written for a wide range of target age groups (early childhood
through middle school-age). However, due to the real-world
distribution of picture books about animals across genres and
target age groups, the sample was biased towards non-fiction
books written for pre- to early-school-age children. Both the
genre and target age group of each book was recorded for use
in analyses, as described in the Results and Discussion
section below.
Book Coding Schemes
The books in the sample were coded by hypothesis-blind
research assistants in order to yield data from which rates of
co-occurrences of taxonomically related versus unrelated
animals could be measured. The coding procedure was
therefore designed to record the taxonomic category of each
animal that appeared on each page of each book. The
taxonomic categories identified were: Mammal, Bird, Fish,
Reptile, Amphibian, Insect, Arachnid, Crustacean, Mollusk,
and Other Invertebrate. Additionally, pairs of contiguous
pages that are visible at the same time were treated as a single
“page”, because the animals they depict would be
experienced as co-occurring. When large numbers of animals
(e.g., more than 50) were depicted on a page, coders were
instructed to estimate the total number by counting the
Figure 1. Example of the taxonomic categories of animals
that would be recorded on a page using the All Animals
scheme (top) and the Animal Groups scheme (bottom).
number within a square inch of the page, and multiplying this
number by the number of square inches that comprised the
area over which the animals were depicted.
This general approach was used to code the contents of
books in the sample according to two coding schemes. Two
coding schemes were developed in order to take into account
the possibility that co-occurrences between a diverse set of
animals from the same taxonomic category may be more
informative about the composition of the category than co-
occurrences between very similar animals. For example, a
learner may acquire a more complete and accurate sense of
what animals are mammals when presented with a variety of
mammals, rather than multiple exemplars of the same
mammal. This possibility is consistent with evidence that
children learn more inclusive categories when presented with
heterogeneous versus homogeneous exemplars (Oakes et al.,
1997; Perry et al., 2010).
The two coding schemes therefore consisted of: 1) An “All
Animals” scheme, in which the taxonomic category of each
individual animal on each page was recorded, and 2) An
“Animal Groups” scheme, in which multiple exemplars of the
same animal on a page were all grouped as a single instance
of their respective taxonomic category (Figure 1). For
example, if a page depicted three cats, it would be coded as
depicting three instances of the Mammal category according
to the All Animals scheme, but only one instance according
to the Animal Groups scheme. However, if a page depicted
three different mammals such as a cat, moose, and rabbit, it
would be coded as depicting three instances of the Mammal
category according to both schemes. The degree to which the
different coding schemes yield different descriptions of the
rate at which children experience co-occurrences between
1
For the purpose of illustration: Consider the three mammals as M1,
M2, and M3, and the two fish as F1 and F2. The same category co-
occurrences would be M1-M2, M1-M3, M2-M3, and F1-F2. The
animals from the same taxonomic category is evaluated in the
Results and Discussion section below.
Results
Data collected via the All Animals and Animal Groups
coding schemes were processed separately following the
same approach. First, for each page, the number of co-
occurrences between pairs of animals from the same and from
different taxonomic categories was calculated. For example,
if a page was coded as depicting three instances of the
Mammal category and two instance of the Fish category, the
number of same taxonomic category co-occurrences would
be recorded as four, and the number of different taxonomic
category co-occurrences would be recorded as six
1
. Next, the
total number of same and different taxonomic category co-
occurrences was calculated for each book. Finally, the
proportion of same taxonomic category co-occurrences out of
total number of co-occurrences was then calculated for each
book in order to assess the rate at which children are exposed
to taxonomic co-occurrence regularities.
Figure 2. Distributions of proportions of same taxonomic
category co-occurrences across books according to the
Individual and All Animals coding schemes.
different category co-occurrences would be M1-F1, M1-F2, M2-F1,
M2-F2, M3-F1 and M3-F2.
Figure 3. Scatterplot with best-fit lines showing increase
in rates of same taxonomic category co-occurrences with
age in data collected using the All animals and Animal
Groups coding schemes.
The distributions of same taxonomic category co-
occurrence proportions across books in the sample according
to the All Animals and Animal Groups coding schemes is
depicted in Figure 2. Examination of these distributions
reveals that the shape of the distribution appears influenced
by the coding scheme. According to the All Animals coding
scheme, books with high proportions of same taxonomic
category co-occurrences are relatively common, whereas
according to the Animal Groups coding scheme, they are
uncommon. These patterns indicate that the high rates of
same taxonomic category co-occurrences in books at the
upper end of the distribution for the All Animals coding
scheme data are likely to have been recorded due to the
depiction of large numbers of the same animal (e.g., schools
of fish, colonies of ants, etc.), rather than multiple different
exemplars of a taxonomic category. When such depictions
are recorded as one instance of a given category in the Animal
Groups coding scheme, rates of same taxonomic category co-
occurrence appear much lower.
Predictors of Same Taxonomic Category Co-
Occurrence
To quantitatively evaluate the influence of both coding
scheme and book characteristics, including target age group
and genre (fiction versus non-fiction), we followed a multi-
level mixed effects modeling approach in which we first
predicted the outcome variable of Same taxonomic co-
occurrence proportion using a null model that included one
random effect of Book, then tested whether the addition of
Coding Scheme (0=All Animals, 1=Animal Groups), Genre
(0=Fiction, 1=Non-Fiction), and Target age group as fixed
effects improved the model fit. For the purpose of this
analysis, the Target age group for a given book was taken as
the age at the midpoint of the range of target ages for that
book. For example, the Target age group a book written for
children aged 3-5 years was recorded as 4.
The results of this analysis revealed that, as inspection of
the distributions for the two coding schemes in Figure 2
suggests, the addition of Coding Scheme to the null model
improved model fit (2(3,4)=40.27, p<.001). The additional
inclusion of Target age group further improved model fit
(2(4,5)=5.19, p<.05). However, model fit was not improved
by either the inclusion of an interaction between Coding
Scheme and Target age (2(5,6)=2.67, p=.10), or the
inclusion of Genre (2(5,6)=0.05, p=.83). The parameters of
the null and two models that incrementally improved fit are
reported in Table 1. The estimates for the fixed effect of
Coding Scheme indicate that, in comparison to the All
Animals scheme, the Animal Groups scheme yielded lower
proportions of same taxonomic category co-occurrences
(which is consistent with distributions depicted in Figure 2).
Moreover, the estimates for the fixed effect of Target age
Table 1.
Estimates of effects for Proportion of same taxonomic category co-occurrences
Null Model
Coding Scheme Model
Coding Scheme+Target Age Model
Estimate
SE
t
Estimate
SE
t
Estimate
SE
t
Fixed Effects
(Intercept)
0.55
.03
21.54
0.67
0.03
21.99***
0.48
0.09
5.49***
Coding Scheme
-0.24
0.03
7.19***
-0.24
0.03
7.19***
Target Age
0.03
0.01
2.89*
Variance
SD
Variance
SD
Variance
SD
Random Effects
Book
0.02
0.13
0.03
0.17
0.03
0.17
Note. *p<.05, **p<.01, ***p<.001
indicate that proportions of same taxonomic category co-
occurrences increased with the age group for which a book
was written (see Figure 3 for graphical depiction of these
patterns).
Discussion
Knowledge of taxonomic categories categories is cognitively
useful for processes such as inductive inference and
integrating newly learned information into memory (Heit,
2000; Pinkham et al., 2014). However, even in a domain that
is commonly experienced and familiar from an early age –
i.e., the domain of animals, evidence from large-scale studies
with students at many levels of education suggests that many
students fail to acquire accurate knowledge about the
composition of taxonomic categories (Trowbridge &
Mintzes, 1985, 1988; Yen et al., 2004). The purpose of this
study was to investigate the degree to which statistical
regularities that are informative about the composition of
taxonomic categories in the animal domain are present in
source of input commonly experienced during development
– i.e., children’s books. Specifically, we investigated the
degree to which animals from the same taxonomic category
tended to co-occur on the same page of children’s books.
Our findings indicate that current children’s books about
animals are not likely to provide a source of co-occurrence
regularities that are informative about the composition of
taxonomic categories. Specifically, books in which the
number of same taxonomic category co-occurrences
outnumber different taxonomic category co-occurrences
could provide a source of input that fosters taxonomic
category learning. Our analyses revealed that co-occurrences
between members of the same taxonomic category primarily
consisted of co-occurrences between members of the same
species. Such same-species co-occurrences are likely to be
less informative about the composition of taxonomic
categories that co-occurrences between animals of different
species from the same taxonomic category (e.g., Oakes et al.,
1997; Perry et al., 2010). When co-occurrences between
members of the same species are eliminated from the tally of
same taxonomic category co-occurrences, the number of
books in which same taxonomic category co-occurrences
outnumber different taxonomic category co-occurrences is
small.
These results suggest that children’s books in the
representative sample analyzed in this study do not provide a
source of co-occurrence regularities that are informative
about taxonomic category membership. Moreover, this
deficit is particularly pronounced in books written for young
children. The lack of exposure to informative co-occurrence
regularities in early childhood may play a critical role in the
emergence of persistent inaccurate conceptions of taxonomic
category composition, given that recent research suggests
early childhood is the period during which taxonomic
misconceptions in the domain of animals are formed (Allen,
2015). These findings are consistent with the possibility that,
even in a domain in which children receive a great deal of
exposure from an early age, the input children receive from
their environment may not provide them with co-occurrence
regularities that can readily scaffold taxonomic category
knowledge.
Open Questions & Future Directions
To date, the sources of environmental input of interest in
research on taxonomic category acquisition have primarily
consisted of visual similarity (Kloos & Sloutsky, 2008;
Quinn, Eimas, & Rosenkrantz, 1993; Younger & Cohen,
1983) and shared labels (Fulkerson & Waxman, 2007;
Robinson & Sloutsky, 2007). In contrast, the contributions of
statistical regularities such as co-occurrence remain the
subject of less direct study. Instead, multiple lines of evidence
have strongly suggested a role for statistical regularities in
shaping the way that children learn to group entities. For
example, in the domain of animal knowledge, multiple
studies have investigated the organization of children’s
knowledge about animals into groups by analyzing the order
in which children spontaneously list animals when asked to
say all the animals that come to mind. Findings from these
studies have revealed that, across childhood, animals that
appear in close sequential order in lists are those that co-occur
in the same habitat (e.g., animals that co-occur in farms,
water, zoos, etc.). Similarly, a series of studies conducted by
Kattmann (Kattmann, 1998, 2001) revealed that, when
prompted to organize animals into groups, children at
multiple levels of education (ranging from ~9 to 13 years of
age) made grouping decisions based on which animals
typically co-occur. Finally, recent studies by Unger et al.
(2016) have revealed that children perceive members of the
same taxonomic category as related both earlier and to a
greater extent when they commonly co-occur in the
environment, versus when they do not.
Taken together, these findings support the possibility that
a sensitivity to statistical co-occurrence regularities
contributes to the way in which children’s knowledge about
the world becomes organized into groups or categories.
However, a more direct investigation of this possible role for
co-occurrence sensitivity ideally consist of research in which
children are exposed to co-occurrence regularities, and the
effects of this exposure on category knowledge are measured.
For example, a follow-on to the present research might
involve comparing the effects on children’s knowledge of
taxonomic animal categories of reading existing books with
high versus low same taxonomic category co-occurrences, or
a books developed in-lab to convey high versus low same
taxonomic category co-occurrences while keeping other
characteristics constant.
Conclusions
Sensitivities to statistical regularities in the environment may
substantially contribute to our development of knowledge
about the world around us. A key part of understanding the
degree to which such contributions do indeed transpire is to
assess the degree to which environmental input actually
contains statistical regularities relevant to a given facet of
knowledge about the world, such as the division of entities
into meaningful categories. Our present investigation yielded
evidence that, even in a source of environmental input that is
commonly experienced in childhood - i.e., children’s books,
and in a domain that is familiar from an early age – i.e.,
animals, statistical regularities that are informative about the
composition of meaningful categories are relatively rare.
Acknowledgements
This work was supported by a Graduate Training Grant
awarded to Carnegie Mellon University by the Department of
Education, Institute of Education Sciences (R305B040063)
and by the James S. McDonnell Foundation 21st Century
Science Initiative in Understanding Human Cognition –
Scholar Award (220020401) to the second author. We also
thank Anna Vande Velde, Eden Hu, Rebeka Almasi, Clara
Lee, Camille Warner, Smriti Chauhan, and Sara Jahanian for
their vital contributions to this research.
References
Allen, M. (2015). Preschool children’s taxonomic knowledge
of animal species. Journal of Research in Science
Teaching, 52, 107-134.
Bulf, H., Johnson, S. P., & Valenza, E. (2011). Visual
statistical learning in the newborn infant. Cognition, 121,
127-132.
Crowe, S. J., & Prescott, T. J. (2003). Continuity and change
in the development of category structure: Insights from the
semantic fluency task. International Journal of Behavioral
Development, 27, 467-479.
Fulkerson, A. L., & Waxman, S. R. (2007). Words (but not
tones) facilitate object categorization: Evidence from 6-and
12-month-olds. Cognition, 105, 218-228.
Ganea, P. A., Ma, L., & DeLoache, J. S. (2011). Young
children’s learning and transfer of biological information
from picture books to real animals. Child Development, 82,
1421-1433.
Heit, E. (2000). Properties of inductive reasoning.
Psychonomic Bulletin & Review, 7, 569-592.
Huebner, P., & Willits, J. (2017). Structured semantic
knowledge can emerge automatically from predicting word
sequences in child-directed speech.
Kattmann, U. (1998). Do students have an implicit theory of
animal kinship. In B. Anderson (Ed.), Research in Didaktik
of Biology (pp. 61-83). Göteborg, , Sweden.
Kattmann, U. (2001). Aquatics, Flyers, Creepers and
Terrestrials—students' conceptions of animal
classification. Journal of Biological Education, 35, 141-
147.
Kloos, H., & Sloutsky, V. M. (2008). What's behind different
kinds of kinds: Effects of statistical density on learning and
representation of categories. Journal of Experimental
Psychology: General, 137, 52.
Mattys, S. L., & Jusczyk, P. W. (2001). Phonotactic cues for
segmentation of fluent speech by infants. Cognition, 78,
91-121.
Oakes, L. M., Coppage, D. J., & Dingel, A. (1997). By land
or by sea: the role of perceptual similarity in infants'
categorization of animals. Developmental Psychology, 33,
396.
Pelucchi, B., Hay, J. F., & Saffran, J. R. (2009). Statistical
Learning in a Natural Language by 8‐Month‐Old Infants.
Child development, 80, 674-685.
Perry, L. K., Samuelson, L. K., Malloy, L. M., & Schiffer, R.
N. (2010). Learn locally, think globally exemplar
variability supports higher-order generalization and word
learning. Psychological Science, 21, 1894-1902.
Pinkham, A. M., Kaefer, T., & Neuman, S. B. (2014).
Taxonomies Support Preschoolers’ Knowledge
Acquisition from Storybooks. Child Development
Research, 2014.
Quinn, P. C., Eimas, P. D., & Rosenkrantz, S. L. (1993).
Evidence for representations of perceptually similar natural
categories by 3-month-old and 4-month-old infants.
Perception, 22, 463-475.
Robinson, C. W., & Sloutsky, V. M. (2007). Linguistic labels
and categorization in infancy: Do labels facilitate or
hinder? Infancy, 11, 233-253.
Schapiro, A. C., Kustner, L. V., & Turk-Browne, N. B.
(2012). Shaping of object representations in the human
medial temporal lobe based on temporal regularities.
Current Biology, 22, 1622-1627.
Schapiro, A. C., Rogers, T. T., Cordova, N. I., Turk-Browne,
N. B., & Botvinick, M. M. (2013). Neural representations
of events arise from temporal community structure. Nature
Neuroscience, 16, 486-492.
Storm, C. (1980). The semantic structure of animal terms: A
developmental study. International Journal of Behavioral
Development, 3, 381-407.
Trowbridge, J. E., & Mintzes, J. J. (1985). Students'
alternative conceptions of animals and animal
classification. School Science and Mathematics, 85, 304-
316.
Trowbridge, J. E., & Mintzes, J. J. (1988). Alternative
conceptions in animal classification: A cross‐age study.
Journal of Research in Science Teaching, 25, 547-571.
Unger, L., Fisher, A. V., Nugent, R., Ventura, S. L., &
MacLellan, C. J. (2016). Developmental Changes in
Semantic Knowledge Organization. Journal of
Experimental Child Psychology, 146, 202-222.
Yen, C.-F., Yao, T.-W., & Chiu, Y.-C. (2004). Alternative
conceptions in animal classification focusing on
amphibians and reptiles: A cross-age study. International
Journal of Science and Mathematics Education, 2, 159-
174.
Young, K. T., Davis, K., Schoen, C., & Parker, S. (1998).
Listening to parents: a national survey of parents with
young children. Archives of Pediatrics & Adolescent
Medicine, 152, 255-262.
Younger, B. A., & Cohen, L. B. (1983). Infant perception of
correlations among attributes. Child Development, 54, 858-
867.
