View
0
Download
0
Category
Preview:
Citation preview
PREDICTORS OF LOW NUMBER KNOWLEDGE
Early Developmental Trajectories of Number Knowledge and Math Achievement From 4
to 10 Years: Low-Persistent Profile and Early-Life Predictors
Gabrielle Garon-Carrier1, Michel Boivin1,2, Jean-Pascal Lemelin3, Yulia Kovas4,5, Sophie Parent6,
Jean Séguin7,8, Frank Vitaro6, Richard E. Tremblay2,9,10, & Ginette Dionne1
1 School of Psychology, Université Laval, Canada2 Institute of Genetic, Neurobiological, and Social Foundations of Child Development, Tomsk
State University, Tomsk, Russian Federation3 Department of Psychoeducation, Université de Sherbrooke, Canada
4 Department of Psychology, University of London, Goldsmiths, England5 Laboratory for Cognitive Investigations and Behavioural Genetics, Tomsk State University,
Russian Federation6 Department of Psychoeducation, Université de Montréal, Montréal, Canada
7 Department of Psychiatry, Université de Montréal, Canada8 CHU Ste-Justine Research Center, Université de Montréal, Montréal, Canada
9 Department of Pediatrics and Psychology, Université de Montréal, Canada10 School of Public Health, Physiotherapy and Population Sciences, University College Dublin,
Ireland
Correspondence concerning this article should be sent to Michel Boivin, CRC in Child
Development, Professor, École de psychologie, Université Laval, Québec, Canada, G1K 7P4.
Email: Michel.Boivin@psy.ulaval.ca
1
PREDICTORS OF LOW NUMBER KNOWLEDGE
Research Highlights
Abstract
Little is known about the development of number knowledge (NK) and the antecedents of low-
persistent NK profiles in early childhood. We documented the developmental trajectories of NK
across the transition from preschool to elementary school, their predictive validity with respect to
Number knowledge development is not linear, and vary in onset level and rate of progression from preschool to school entry.
10% of preschool children show developmental lag in number knowledge, signaling a clear risk of math underachievement.
Poor early cognitive development, memory and visual-spatial skills uniquely predict children with low number knowledge.
2
PREDICTORS OF LOW NUMBER KNOWLEDGE
later math achievement, and the child and family early-life factors associated with low NK
profiles. Children’s NK was assessed four times at regular intervals between the ages 4 and 7
years in a large, representative population-based sample. Developmental trajectories of NK were
established for 1,597 children. These children were also assessed with respect to several features
of their family environment at 5, 17, and 29 months, as well as their cognitive skills at age 41
months. Analyses revealed a best-fitting 4-trajectory model, characterized by Low-Increasing
(10% of the children), Moderate-Increasing (39%), Moderate-Fast Increasing (32%) and High-
Increasing (19%) groups. Children of these trajectory groups differed significantly with respect
to math achievement at ages 8 and 10 years, with the Low-Increasing group persistently scoring
lower than the other groups throughout these years. Children of Low-Increasing NK group were
from household of lower income and father with low educational background, poorer early
cognitive development, and more importantly, reduced visual-spatial skills and memory span.
Children displaying reduced cognitive abilities and impoverished living conditions early in life
are at greater risk of low NK throughout late preschool and school entry, with ensuing
difficulties in math achievement. They deserve early preventive attention to help alleviate later
mathematic difficulties.
Keywords: Number knowledge, Mathematics achievement, Developmental trajectories, Early-
life predictors, Longitudinal study
Early Developmental Trajectories of Number Knowledge and Math Achievement From 4 to 10
Years: Low-Persistent Profile and Early-Life Predictors
Knowledge and skills in mathematics have been shown to predict later academic
achievement (Duncan et al., 2007; Jordan, Kaplan, Ramineni, & Locuniak, 2009; Magnuson,
3
PREDICTORS OF LOW NUMBER KNOWLEDGE
Duncan, Lee, & Metzger, 2016; National Research Council, 2009; Nguyen et al., 2016; Siegler et
al., 2012), and more generally, later educational attainment (Magnuson et al., 2016). Mathematic
abilities are paramount for college entry (Sadler & Tai, 2007) and degree completion in STEM
fields (science, technology, engineering and mathematics; Wolniak, 2016). Unfortunately, the
negative side of this predictive association is that individuals with poor mathematic abilities have
reduced educational and employment opportunities (e.g., low rates of full-time employment and
promotions, high rates of low-paying occupations; Lundetræ, Gabrielsen, & Mykletun, 2010;
Parsons & Bynner, 1997), and might even experience difficulties in common day-to-day
activities in adulthood.
The early signs of these difficulties are quite prevalent. For instance, between 6% and
10% of children suffer from learning disabilities in mathematics (Barbaresi, Katusic, Colligan,
Weaver, & Jacobsen, 2005; Shalev, Auerbach, Manor, & Gross-Tsur, 2000), and many more
struggle with mathematics without a formal diagnosis. Children showing persistent difficulties in
math may never catch up to their grade-level peers. There is also good evidence that these
difficulties emerge quite early, even before school entry (Mazzocco & Thompson, 2005; von
Aster & Shalev, 2007).
Early Development of Mathematic Skills
Children develop a broad array of early mathematic skills well before school entry
(Libertus, Feigenson, & Halberda, 2011). As early as 6 months, they possess an intuitive
impression of numbers – a number sense – that allows them to approximate magnitude difference
between small set of objects (e.g., 4 vs. 8; Feigenson, Dehaene, & Spelke, 2004; Hyde & Spelke,
2011; Lipton & Spelke, 2003; McCrink & Wynn, 2007; vanMarle, 2013; Xu, Spelke, &
Goddard, 2005). Number sense is assumed to provide the basic meaning of number and quantity
4
PREDICTORS OF LOW NUMBER KNOWLEDGE
in infancy (von Aster & Shalev, 2007). From these initial number sense skills, children learn
counting through linking numbers to objects. Children’s numerical development is thus
characterized by an increasing capacity to solve mathematic problems using tangible materials
(e.g., blocks). Around age 3 years, children typically learn number words, counting principles
(e.g., one-to-one counting, cardinality, ordinal number, numerical identification; Bermejo, 1996),
and to carry out simple operations (e.g., number combinations, see Geary, 2004, for a review).
This early knowledge about numbers is key to later mathematical development (Duncan et al.,
2007; Göbel, Watson, Lervåg, & Hulme, 2014; National Research Council, 2009; Nguyen et al.,
2016; Watts, Duncan, Siegler, & Davis-Kean, 2014).
Number knowledge (NK), that is, the conceptual and procedural understanding of whole
numbers (Okamoto & Case, 1996), has been posited to develop steadily and gradually
throughout early childhood (Piaget, 1977), and to lead to more sophisticated mathematic abilities
(Duncan et al., 2007; Göbel et al., 2014; National Research Council, 2009; Nguyen et al., 2016;
Watts et al., 2014). Specifically, it has been suggested that early NK partially originates from an
integration of number sense skills and the symbolic numerical system taught at home or in
school (Feigenson, Libertus, & Halberda, 2013; Libertus et al., 2011). Number sense skills, such
as subitizing and approximating would prepare children to associate quantities with Arabic digits
(i.e., numeral symbols, such as 0, 1, 2 or 3; von Aster & Shalev, 2007). This, in turns, leads to
hierarchically order numbers, a stepping-stone in children’s mastering of numbers and growing
math abilities (Friso-van den Bos et al., 2015; Siegler & Booth, 2004).
Thus, there is both theoretical and empirical support for the view that early NK and the
ensuing mathematic skills are developmentally interlocked (Duncan et al., 2007; Göbel et al.,
2014; National Research Council, 2009; Nguyen et al., 2016; Piaget, 1977; Watts et al., 2014).
5
PREDICTORS OF LOW NUMBER KNOWLEDGE
However, very little is known about individual differences in the development of these skills. If
the development from NK to mathematics generally follows age and grade levels, the rate at
which children transit through this period may differ significantly; some children quickly master
mathematic concepts and operations while others struggle.
Documenting early NK skills is especially crucial during the transition from late
preschool to school entry, as this period is characterized by substantial developmental changes.
This transition not only coincides with shifts in physical, cognitive, emotional, and behavioral
capacities (Blair & Raver, 2015; Sasser, Bierman, & Heinrichs, 2015; Welsh, Nix, Blair,
Bierman, & Nelson, 2010), but it also entails a modification of the learning context, as children
start to be systematically exposed to formal training in early mathematic skills. Accordingly, one
goal of the present study is to assess inter-individual variations in trends of NK development
across the period from late preschool to school entry.
Early Cognitive Correlates of Number Knowledge
An important related question to that of early individual differences in NK is the issue of
their associated cognitive factors and putative environmental determinants. Previous studies of
normative samples, as well as of low math achievers and children with mathematic learning
disability have consistently revealed that poor language, visual-spatial skills, and memory-span
are correlates, if not precursors of low math skills (Bull, Espy, & Wiebe, 2008; Dehaene, Piazza,
Pinel, & Cohen, 2003; Dehaene, Spelke, Stanescu, Pinel, & Tsivkin, 1999; Geary, 2004; Geary,
2011; Geary, Hoard, Byrd-Craven, & DeSoto, 2004; LeFevre et al., 2010; Mazzocco &
Thompson, 2005; Soto-Calvo, Simmons, Willis, & Adams, 2015). Language is involved when
manipulating information within working memory, as well as when counting forward and
backward. Difficulties in processing numbers have been associated with reading difficulties,
6
PREDICTORS OF LOW NUMBER KNOWLEDGE
language impairment, or both (Jordan, Hanich, & Kaplan, 2003). Clearly, the phonetic and
semantic systems are activated when counting, if only to connect the quantities with number
words (Vukovic & Lesaux, 2013), and to solve arithmetic problems (Jordan et al., 2003).
Deficits in these systems might result in difficulties in counting and arithmetic reasoning, as well
as in concurrent reading difficulties (Dehaene & Cohen, 1995, 1997). The visual-spatial system
is also solicited when representing conceptual knowledge; visual-spatial skills are indeed
involved in basic geometry problems, in magnitude comparisons (Dehaene et al., 1999), in
logical thinking and problem solving, as well as in estimating and mentally manipulating
numbers encoded as spatial forms (e.g., number line; see Zorzi, Priftis, & Umiltà, 2002).
Children experiencing difficulties in mathematics also fail to correctly encode and/or retrieve
basic arithmetic facts from memory. Their problem typically arises from persistent difficulties in
memorizing relevant and inhibiting irrelevant information during facts retrieval (Bull et al.,
2008; Geary, 2011; Geary et al., 2004).
Thus, several studies clearly support the association of language, visual-spatial skills, and
memory with different math components (Dehaene et al., 2003; Dehaene et al., 1999; Geary,
2011; LeFevre et al., 2010; Soto- Calvo et al., 2015). However, the prediction of NK from these
cognitive abilities very early in life has not been well documented. This question will be
investigated in the current study.
Early Family Correlates of Number Knowledge
Children’s early NK is also context-dependent. Environmental adversities, such as poor
living and educational conditions (e.g., low socioeconomic status) have been associated with low
math skills (Griffin, Case, & Siegler, 1994; Jordan & Levine, 2009; Jordan, Kaplan, Oláh, &
Locuniak, 2006; Klibanoff, Levine, Huttenlocher, Vasilyeva, & Hedges, 2006; Siegler, 2009).
7
PREDICTORS OF LOW NUMBER KNOWLEDGE
For example, Levine and colleagues (2010) have documented substantial socioeconomic status
(SES) differences in mothers’ number talk to children aged 14 and 30 months. Compared to
mothers of high SES, mothers of low SES provided more input of simple verbal counting, but
less of advanced NK skills (e.g., numerical magnitude estimation), and this difference predicted
knowledge of number words cardinality at 46 months (Levine, Suriyakham, Rowe, Huttenlocher,
& Gunderson, 2010).
However, research linking SES and NK is limited with respect to the key factors
underlying this association. For instance, SES subsumes income, a family-level characteristic,
and educational levels, which may vary within family. These dimensions could have different
contributions to NK (Melhuish et al., 2008), and point to the specific role of mother and father in
the development of NK (McBride, Dyer, Liu, Brown, & Hong, 2009; Viljaranta, Lazarides,
Aunola, Räikkönen, & Nurmi, 2015). Furthermore, parents’ outcome expectancies (Bandura,
1997; Benasich & Brooks- Gunn, 1996; Bornstein, 2002; Kouimtzi & Stogiannidou, 2009; Parke
& Buriel, 1998), that is, the belief that their parenting behaviors are important for their child’s
development could account for providing a stimulating environment, such as exposing and
engaging young children in numerical activities (e.g., cooking, board games, moving along
number paths, Laski & Siegler, 2014; LeFevre et al., 2009; Levine et al., 2010; Ramani, Siegler,
& Hitti, 2012; Siegler & Ramani, 2009). This should be investigated more systematically.
The Present Study
Few studies have longitudinally documented individual variations in the development of
NK across the late preschool and school transition periods. Previous studies have provided
evidence for the relevance of mathematical skills at school entry for academic outcomes up to
age 9 and 10 (Aunola, Leskinen, Lerkkanen, & Nurmi, 2004; Duncan et al., 2007; Jordan et al.,
8
PREDICTORS OF LOW NUMBER KNOWLEDGE
2009; Jordan, Glutting, & Ramineni, 2010; Nguyen et al., 2016). However, few studies have
been conducted on NK as early as age 4 years. NK has only been studied from age 5-6 years
and/or only at single time-points (Aunola et al., 2004; Jordan et al., 2009; Jordan et al., 2010;
Nguyen et al., 2016), providing little information on its early development and validity in
predicting later math achievement and difficulties.
In the present study, we argue for the identification of subsets of developmental
trajectories based on repeated measures of evolving NK during the preschool and early school
years. To reflect the evolving and discrete nature of NK, we used the Number Knowledge Test
(Okamoto & Case, 1996), which provides a four-level assessment of several aspects of numerical
competence (see Method). Such an approach posits heterogeneity in the developmental
trajectories, and thus identifies homogeneous subgroups with respect to the evolution of NK over
the targeted period of time, including children that show developmental lags in NK. Indeed, as
some children may experience learning difficulties at various times in their NK development, it
is therefore important to distinguish children with persistent difficulties from those with
normative transient difficulties. Previous studies have often used arbitrary cut-offs to establish
groups or patterns of NK and mathematic skills (e.g., above 10% vs. over 90%; 2 SD above and
over the mean). For example, one study classified children with mathematic scores under the 15th
percentile as children with mathematic learning disability, and those scoring between the 15th and
30th percentiles as math low achievers (Geary, Bailey, & Hoard, 2009). Here, we used a
clustering procedure based on semi-parametric modeling (Nagin, 1999). This quantitative
approach offers several advantages over uniform longitudinal analysis, such as growth curve
analysis (Jordan et al., 2003; Jordan et al., 2006; Jordan, Kaplan, Locuniak, & Ramineni, 2007).
First, it considers individual differences in trajectories to determine the optimal number of
9
PREDICTORS OF LOW NUMBER KNOWLEDGE
groups needed to describe different patterns of change over time. Second, it does not make
strong assumptions about the population distribution of the putative developmental trajectories.
Third, the model defines the form of the trajectory for each potential cluster of trajectories,
which allows for the possibility that subgroups of children show distinct developmental trends in
NK, thus providing a more nuanced view of NK development, and one that could reveal non-
linear changes in development (see Leblanc, Boivin, Dionne, Tremblay, & Pérusse, 2008, for a
more extensive discussion of these points).
Furthermore, from a preventive perspective, not only is it essential to document the
various early developmental pathways in NK, but it is also crucial to (a) estimate from a person-
centered perspective (Laursen & Hoff, 2006) whether and how some of these developmental
profiles foresee later difficulties in mathematics, (b) identify early risk factors that predict low
and enduring profiles of NK over time, and (c) forecast a possible evolution into mathematic
difficulties. Identifying specific predictors of low NK may help preschool and childcare
practitioners to better understand the cognitive and family profile of children with low NK, with
the hope of guiding preventive efforts aimed at children at risk of failure in school.
The study also provided a unique opportunity to test for possible sex differences in
number knowledge and achievement in mathematics. Previous studies found differences favoring
boys in number knowledge. Jordan et al. (2006) found small but significant sex effects on
calculation with objects and on numerical estimation in kindergarten (mean age = 5.8 years);
boys had an edge over girls even when income level, age, and reading ability were controlled for
in the analyses, and more boys than girls were classified in the highest performing group.
However, several studies found no such differences in later math performance (Lachance &
Mazzocco, 2006; Levine, Jordan, & Huttenlocher, 1992). To date, only a few longitudinal
10
PREDICTORS OF LOW NUMBER KNOWLEDGE
studies have tested for the possible sex difference in both number knowledge and to its
prediction to math achievement.
Accordingly, the following research questions were addressed in the present study: (1)
Using a group-based modeling approach, are there discrete patterns of developmental trajectories
of NK? At least three putative trajectories of NK were expected: a small group of children with
high level of NK, a larger group, likely the majority of children, with an average/moderate level
of NK, and a small but significant group of children with low NK profile (Jordan et al., 2006).
All these groups were expected to gain in NK over time, but we did not have a specific
hypothesis regarding the rate of their relative gains. Based on Jordan et al. (2006), we also
expected more boys than girls to fall on a high-level trajectory of NK.
(2) Does early NK predict later math achievement? To answer this question, we
compared school-based mathematic achievement in 2nd and 4th grade (age 8 and 10) according to
NK trajectories and sex. This period is a time when mathematics becomes more complex and
differentiated. We expected that children falling on the lowest trajectory would show the lowest
math achievement scores at both ages, whereas those of the highest NK trajectory-group would
display higher scores, again, boys having higher scores than girls.
(3) To what extent can we predict a child’s trajectory of low NK from specific early
cognitive and family predictors? Cognitive factors, specifically receptive vocabulary, visual-
spatial skills, and memory-span were considered central in the model. Also included was an
assessment of early general cognitive development at 29 months to account for initial
developmental gains in cognitive skills, and thus more precisely estimate the unique
contributions of these specific skills to NK. Because family factors may also play a role in the
development of both the specific cognitive factors and NK, they were assessed early in
11
PREDICTORS OF LOW NUMBER KNOWLEDGE
development. Here we distinguished between family income and parent’s education, and
considered early parenting outcome expectancies (i.e., the parents’ beliefs that their parenting
behaviors are important for their developing child). We expected both lower cognitive skills and
poorer family background factors to independently predict membership in a trajectory of low
NK.
Method
Participants and Setting
Participants were from the Quebec Longitudinal Study of Child Development (QLSCD),
a longitudinal population-based study aimed at understanding the role of early childhood in later
school adjustment and academic achievement. Participating families were recruited through the
Quebec Master Birth Registry of the Ministry of Health and Social Services to be representative
of children born in 1997-98 in the province of Quebec, Canada. For practical reasons, data could
not be collected on children living on Cree or Inuit territories, in Indian reserves, and in northern
Quebec, and thus, were excluded from the study. A stratified three-stage sampling design based
on living area and birth rate was used. The territory covered by the survey was first divided into
primary sampling units (regions), which were divided into second-stage units composed of one
or two county regional municipalities, and then further divided in third-stage units according to
the number of births registered in 1996. All selected infants had been born after October 1, 1997
to ensure that they would enter school the same school year. Families were excluded if the
mother could not speak French or English, and if babies were born before 24 weeks or after 42
weeks of gestation. Those with a gestation period of less than 24 weeks could not be retained
because they had a higher risk of mortality between entry in the register and the conducting of
the survey. Similarly, births occurring after 42 weeks of gestation had to be excluded because the
12
PREDICTORS OF LOW NUMBER KNOWLEDGE
delay in selection would have meant waiting until they became available for the sampling frame.
These two a priori exclusions represented approximately 0.1% of registered births of the date
collection.
A sample of 2,940 families with newborns was initially identified with the goal of
obtaining reliable longitudinal statistics within the budget. Selected families that could be located
(N = 2,675) were approached by mail and by phone. Of those, 2,223 families were first visited
when the child was 5 months old (83%), and 2,120 of them (79% response rate) were followed
longitudinally and regularly assessed on various child and family characteristics (Jetté & Des
Groseillers, 2000a, 2000b).
The QLSCD children were of Canadian, European, British or French origin in 82.5% of
the cases; 12.1% were of American, African, Haitian or another ancestry, and 0.5% were of an
unidentified ethnic origin. At 5 months, 31.4% reported an income lower than 30,000 Canadian
Dollar, 17.9% of mothers and 17.6% of fathers had no high school diploma, and 17% were
reconstituted or one-parent families. French was the most frequent language spoken at home
(75.2%), followed by English (10.1%). The remaining families spoke other languages or a
combination of languages (Jetté & Des Groseillers, 2000a, 2000b). Other details of the cohort
study can be found at: http://www.iamillbe.stat.gouv.qc.ca/publications/baby_no1.pdf
The present study focused on data relevant to the study aims and collected in infancy and
the early school period. The study covers the preschool and early school periods, and uses data
collected when children were ages 5 months (N = 2,120; age: M = 5.0 months, SD = .48), 17
months (N = 2,045; age: M = 17.1 months, SD = .49), 29 months (N = 1,997; age: M = 29.1
months, SD = .47), 41 months (N = 1,950; age: M = 41.1 months, SD = .52), and then at ages 4
years (N = 1,944; age: M = 50.0 months, SD = 3.09), 5 years (N = 1,759; age: M = 61.9 months,
13
PREDICTORS OF LOW NUMBER KNOWLEDGE
SD = 3.09), and 6 years (N = 1,492; age: M = 73.8 months, SD = 3.00), and in the following
school years: grade 1 (N = 1,528; age: M = 7.15 years, SD = 3.06), grade 2 (N = 1,451; age: M =
8.10 years, SD = .26), and grade 4 (N = 1,334; age: M = 10.14 years, SD = .26).
At age 4 years, a minority (16.7%, N = 324) of the participants attended a nursery school,
a preschool, or educational activities offered by a municipality or a community or recreational
center on a regular basis (M = 8.59 hours/week; SD = 8.88). At age 5 years, 47.5% (N = 835)
attended daycare, and 37.5% (N = 659) attended junior or senior kindergarten at a full day or
half-day frequency.
Starting at age 4, all data collections were done in the winter and spring of the year. The
average attrition rate from ages 5 months to 10 years was 4.0% per year, although it varied
slightly across measures.
Measures
Number knowledge. NK was measured through an adapted version of Okamoto and
Case (1996) Number Knowledge Test (NKT; Côté et al., 2013; Duncan et al., 2007; Romano,
Babchishin, Pagani, & Kohen, 2010). The NKT was developed with the goal of documenting
children’s understanding of whole numbers and basic operations, and as a tool for teachers to
identify children with mathematic difficulties (Gersten, Clarke, & Jordan, 2007). The NKT has
four levels of complexity (from 0 to 3). Each level of the test reflects a current developmental
stage of children’s NK conceptions (Gersten et al., 2007; Gersten, Jordan, & Flojo, 2005). The
baseline level (level 0, around ages 3-4) requires children to count small sets of tangible objects
(e.g., The experimenter shows five unordered chips to the child, and then asks him to count the
chips; The experimenter shows image with two piles of chips, and then asks the child which pile
has more chips). The first level (around age 6) measures uni-dimensional mental representations
14
PREDICTORS OF LOW NUMBER KNOWLEDGE
of numbers, i.e., numerical comparison, with items that were designed to probe for the “mental
number line” structure (e.g., Which number is closer to 5: 6 or 2?). The second level (about age
8) reflects bi-dimensional representations, i.e., understanding numerical “difference” (e.g., How
many numbers are there in between 7 and 9?), and the base-ten system with double-digit
numbers (e.g., What number comes 5 numbers after 49?). The third level (age 10) reflects
integrated bi-dimensional mental representations of numbers, i.e. constructing and comparing
two sums or differences rather than just one (e.g., which difference is smaller: the difference
between 48 and 36 or the difference between 84 and 73?). At this level, children also have to
manage triple-digit numbers and/or to solve more complex problems involving double-digit
numbers.
In the present study, the adapted NKT (Côté et al., 2013; Duncan et al., 2007; Romano et
al., 2010) was administered by a trained research assistant at ages 4 (N = 1,768), 5 (N = 989), 6
(N = 1,189) and 7 (N = 1,461) years. Five items of the baseline level (level 0) were slightly
modified to facilitate the administration of the test, but aimed for the same knowledge and skills
as the original items (see Table S1 in Supplementary materials). The items of the next levels
were identical to the original test. To reduce the length of the assessments, alleviate fatigue, and
preserve children’s motivation, the administration of the NKT slightly departed from Okamoto
and Case (1996). According to Okamoto and Case (1996), only children who correctly answered
at least 60% of the items at one level should move to the next. Here, the procedure was
systematized differently so that the same levels would be administered to all children at a given
age, thus insuring within-year standardization. Specifically, at ages 4, 5, and 6 years, only the
levels 0 and 1 were systematically administered, and the test was stopped if the child made three
consecutive errors at level 1. Thus, up to age 6, the procedure started at the baseline level and
15
PREDICTORS OF LOW NUMBER KNOWLEDGE
moved to level 1 until the child failed three consecutives items. At 7 years, the baseline level was
omitted and levels 1 and 2 were administered. The test was stopped if the child made three
consecutives errors at level 2. The score consisted of the total number of correct items across
levels. Points were automatically assigned for the baseline level items if testing began at level 1
(for children aged 7 years).
With the exception of the low reliability at age 5 (α = .55), the internal consistency of the
adapted-NKT in our sample was found to be adequate (α = .68 at age 4, .92 at age 6, and .79 at
age 7 years); and the test-retest stability was high across all time points: r = .74 between ages 4
and 5; r = .92 between ages 5 and 6; and r = .82 between ages 6 and 7. We also tested the
measurement-invariance of the adapted-NKT across time using Mplus 7.11 (Muthén & Muthén,
1998–2012). The test was invariant over time, with non-significant difference in the model fit
when comparing the non-constrain model to model with constrained to equality factor loadings
for matching items of NK, ∆χ2 (1) = .008, p = .78 (results available from the authors). Scores
were normally distributed at ages 6 and 7 years, but a small trend toward bimodal distribution
was found at both ages 4 and 5 years, which again justifies the use of a modeling approach that
does not assume normal distribution (Nagin, 1999).
Achievement in mathematics in elementary school. Achievement in mathematics in
elementary was assessed through the Mathematics subtest of the Canadian Achievement Test
(CAT), an evolving Canadian-based subtest reflecting grade appropriate math achievement in
elementary school (Canadian Test Center, 1992; Carter, Dubois, & Ramsay, 2010). Using a
multiple-choice response and a pen-and-paper formats, the CAT was administered when children
were in grade 2 (age 8) and in grade 4 (age 10). The grade 2 version of the CAT assessed
children’s capacity to perform addition (7 items), subtraction (8 items), and multiplication (6
16
PREDICTORS OF LOW NUMBER KNOWLEDGE
items) operations. The grade 4 version assessed the ability to compute slightly more complex
(compared to grade 2) addition (5 items), subtraction (5 items), multiplication (6 items)
operations, and also required to compute division (4 items) operations. Children had to select the
right answer out of four choices within a given time (between 45 and 60 seconds depending on
the item difficulty). Children had to move to the next item when the time was up. The test ended
when three successive errors were made within a type of operation (e.g. three successive errors
in subtraction operations). We computed a total raw score for each child by summing the number
of correct items across all types of operation. The internal consistency of this measure in our
sample at ages 8 and 10 was α = .76, and α = .81, respectively, and its grade 2-4 stability was r =
.49 (Garon-Carrier, Boivin, Ouellet, Tremblay, & Dionne, in preparation).
Family predictors. Data on two categories of family predictors—socio-demographic
factors and parenting outcome expectancies—were collected.
Socio-demographic factors. We collected socio-demographic data through questions
from the National Longitudinal Survey of Children and Youth (see Willms & Shields, 1996)
when the child was 5-month-old. Household income was assessed on a 9-point scale ranging
from 1 ($10,000 or less/year) to 9 ($80,000 or more/year). Mother and father school attainment
was measured as their highest diploma, as revealed through a 4-point self-report scale ranging
from 1 (no high school diploma) to 4 (undergraduate degree). These rank-ordered variables were
normally distributed, and were thus treated as continuous variables in the following analyses (see
Pasta, 2009; Powers & Xie, 2008).
Parenting outcome expectancies. Mother’s and father’s parenting outcome expectancies
were measured through a subscale (5 items) of the Parental Cognitions and Conduct Toward the
Infant Scale (PACOTIS) (Boivin et al., 2005), a self-report questionnaire designed to assess
17
PREDICTORS OF LOW NUMBER KNOWLEDGE
parenting perceptions and behavior tendencies toward a recently born infant. For most children
(i.e., 82% of them), both mothers and fathers filled out the questionnaire when their child was
aged 5 months, 17 months and 29 months. They had to indicate on a continuous 11-point scale
(i.e., from 0 to 10) to what extent each statement regarding the perceived impact of their
behavior on their developing child accurately described their actions or thoughts (e.g.,
Regardless of what I do, my baby will develop on his/her own; My behavior has little effect on
the development of emotions (happiness, fear, anger) in my baby, etc.). The item scores were
reversed to reflect perceived parenting impact/expectancies. These parenting expectancy items
presumably reflect the quality of parents’ involvement vis-à-vis their child (see Boivin et al.,
2005 for a detailed description of the construction and validity of the scale). The measure
revealed acceptable internal consistency in our sample from 5 to 29 months (α = .71–.78). For
both mother and father, we averaged the three time-specific scores into a global parenting
outcome expectancy score.
Child Cognitive Abilities. Data on four child cognitive abilities—early cognitive
development, memory span, receptive vocabulary, visual-spatial skills—were collected.
Early cognitive development. When the child was aged 29 months, the person “most
knowledgeable about the child” (the mother 99.7% of the time) reported the child’s early
cognitive development through a 3-item scale (naming four colors, counting three objects and
pronouncing partial sentence of three words or more). The three items were moderately
correlated (.26 and .31) and summed to compute a total score of early cognitive development.
This measure was also correlated to the other measures of child cognitive abilities at 41 months,
i.e., with memory-span (r = .22), receptive vocabulary (r = .33), and visual-spatial skills (r
= .20).
18
PREDICTORS OF LOW NUMBER KNOWLEDGE
Memory-span. Children were also assessed with respect to their memory-span at age 41
months through the Visually Cued Recall task (VCR; Zelazo, Jacques, Burack, & Frye, 2002), a
reliable (α = .95 in our sample) measure of the child’s incremental capacity to encode visual
items and to recall the spatial locations of the items after a short delay. In each trial, a cardboard
with pictures of 12–18 objects was shown to the child by a research assistant who pointed to
items and asked the child to remember them. The research assistant then flipped the cardboard
for a short delay. When flipped back, the child was cued to identify the items pointed to
previously. The number of items to remember increased after each trial, up to 12 different levels
of difficulty. The test ended when the child made two errors on two subsequent levels. The score
consisted of the highest level reached by children.
Receptive vocabulary. Children’s receptive vocabulary at age 41 months was evaluated
with the Peabody Picture Vocabulary Test-Third Edition (PPVT-III; Dunn & Dunn, 1997; Dunn,
Theriault-Whalen, & Dunn, 1993), a standardized language test that assessed phonological
recognition and semantic understanding of words upon hearing them. This test has high internal
consistency (α = .93; Dunn et al., 1993), and is valid for use with French and English speakers
(Dunn et al., 1993; Flipsen, 1998). The test consisted of 170 cards each depicting four different
objects, actions, or emotions. Children had to identify the correct corresponding image to the
word said by the experimenter. One point was allowed for each correct answer. The test stopped
when children made six errors within a sequence of eight cards.
Visual-spatial skills. The visual-spatial skills were measured at 41 months through the
Block Design subtest of the Wechsler Preschool and Primary Scale of Intelligence–Revised
(WPPSI-R; Wechsler, 1989). This subtest was composed of 14 models depicted on pictures and
that children had to re-create using blocks. Bonus points could be gained for some models as a
19
PREDICTORS OF LOW NUMBER KNOWLEDGE
function of time, and the test ended after three consecutive failures. Raw scores varied from 0 to
42. The Block Design mainly required visual processing ability, including perceptions of spatial
relations and mental manipulations of visual patterns. This subscale was highly correlated with
the Full Scale WPPSI-R score (r = .62) in the norming sample. It had excellent internal
consistency and test–retest reliability (see Sattler, 2001). Cronbach alpha for this subtest was .89
in our sample.
Procedure
Interviewers from an independent research firm were hired for the data collections. They
were trained, filmed, provided feedback, and tested for administration reliability by our research
coordinator. Specifically, they were trained by the research coordinator to administer the
measures, and were filmed when practicing with children that were not part of the study. The
following week, they would watched the video with, and provided feedback by the research
coordinator. Feedback to the interviewers would also be provided regularly after administrations
on real participants. Measures were administered in French (most families) or in English. All
French speaking children were assessed using French norms and validation procedures. Table 1
summaries the data collection procedures: the variables (predictors and outcomes), child’s age-
at-assessment, informant, and instruments.
When children were 5 months old, a face-to-face interview with the person most
knowledgeable about the child provided data on the household income, and the mother’s and the
father’s diploma. The parenting expectancies’ subscale was filled out separately by both parents
at home when their child was 5, 17 and 29 months, each time after a home visit.
The person most knowledgeable about the child also provided data on the child’s early
cognitive development at 29 months, and a trained research assistant administered the VCR, the
20
PREDICTORS OF LOW NUMBER KNOWLEDGE
PPVT, and the Block Design test following a standard procedure in a face-to-face interview
when children born in 1997-98 were 41 months.
The adapted-NKT was orally administered one-on-one by a trained research assistant at
school or at home. The baseline and first levels of the adapted-NKT were administered at ages 4,
5 (preschool period), and 6 (kindergarten); the second level was added to the previous ones at
age 7 (grade 1). The CAT was individually assessed at school or at home by a trained research
assistant at ages 8 (grade 2) and 10 (grade 4).
Data Analysis
Treatment of Missing Data
Missing NK, cognitive, and family scores ranged between less than 1% and 8.5%. Based
on the Little’s MCAR test, the missing data were not completely at random (χ2 = 1020.86, df =
516, p < .001). A series of t tests obtained with the MVA module in SPSS 20.0 for Windows
(SPSS Inc, Chicago, IL) showed that children with missing data tended to have lower NK, and
were from a significantly lower socioeconomic background. To control for this potential bias, we
used full information maximum likelihood (FIML) to treat missing data. All the statistics
reported were estimated using FIML.
Missing data on the math achievement measures ranged between 13.5% and 21.2%. We
used multiple imputations to produce estimates for missing data through PROC MI procedure in
SAS (SAS Institute, Inc, Cary, NC).
Developmental Trajectories of NK
Developmental trajectories of NK from ages 4 years to 7 years were estimated using
semi-parametric mixture models in the PROC TRAJ procedure in SAS (SAS Institute, Inc, Cary,
NC; Jones & Nagin, 2007). This procedure identifies clusters of individuals following similar
21
PREDICTORS OF LOW NUMBER KNOWLEDGE
progressions of an outcome over time by fitting a group-based model. A quadratic relationship is
used to model the link between age and behavior (here NK). Model estimation results in three
key outputs: (a) the shape of each group’s trajectory, (b) the estimated proportion of the
population belonging to each trajectory group, and (c) for each individual in the estimation
sample, an estimate of the probability that he/she belongs to each of the trajectory groups
identified (based on posterior probability of group membership).
In our model, subjects were included when data were available for at least two time-
points out of four to minimize attrition due to missing data. Thus, a total of 1,597 children were
included in the analysis. Solutions yielding two to five trajectory-groups of various shapes
(intercept -0-, linear -1-, quadratic -2-, cubic -3-) were estimated, and the best fitting solution
was derived based on the values of the Bayesian Information Criterion (BIC), which reflects the
parsimony of the model, the Akaike information criterion (AIC), and the theoretical likelihood
(L). Once the best-fit model established, each child was assigned to a specific NK trajectory-
group based on the highest probability of belonging to a trajectory.
We also performed a three-way repeated-measure ANOVA 4 (time) X 4 (group) X 2
(sex) on the NK scores from ages 4 to 7 to evaluate within and between groups differences
across time. The results were adjusted with Bonferronni correction for multiple comparisons.
Prediction of Achievement in Mathematics from NK Trajectories
Using the PROC MIXED SAS procedure, we performed a 2 (time) X 4 (group) X 2 (sex)
repeated-measures ANOVA on math scores at ages 8 and 10 to test whether boys and girls of
various trajectories maintained their relative position from early NK to math achievement in
grades 2 and 4. The results were adjusted for multiple comparisons.
Prediction of Low NK Trajectory
22
PREDICTORS OF LOW NUMBER KNOWLEDGE
The predictive associations between early-life family and child-cognitive factors and the
low trajectory of NK were first tested with SPSS 20.0 for Windows (SPSS Inc, Chicago, IL) by
running chi-square tests (for categorical variables) or ANOVAs (for continuous variables) in
which the low trajectory was compared to the other trajectories of NK (see Figure 1).
We next conducted a binary logistic regression analysis to test the unique prediction of
early-life family factors and child-cognitive risk factors to membership in the low trajectory of
NK. The models were tested with Mplus 7.11 (Muthén & Muthén, 2012). The predictors were
grouped in three blocks following a developmental-hierarchical time order. Early-life family
factors were considered in the first two blocks, first, household income, and both mother and
father diploma, then mother and father outcome expectancies. Child cognitive factors, including
the index of early cognitive development, the Block Design visual-spatial score, the VCR
memory-span score, and the PPVT receptive vocabulary score, were introduced in the last block.
Each block of predictors was entered sequentially. Non-significant variables were removed from
the regressive model, before adding a new block of variables. To derive the best fitting model,
the regressive models were compared to a baseline model that did not include predictors.
Goodness-of-fit indices were quantified using the Akaike’s information criterion (AIC), the
variance explained (R2), and the -2 Log Likelihood ratio (∆-2LL). The lowest AIC index, the
highest R2, and a significant deterioration of the model fit suggested by the ∆-2LL value
indicated a better fit of the regression model compared to the baseline model.
Results
Developmental Trajectories of NK
As presented in Figure 1, the solution yielding a four-group model (one linear, three
quadratic trajectories) was found to best fit the NK data (BIC = -12886.15; AIC = -12845.83; L =
23
PREDICTORS OF LOW NUMBER KNOWLEDGE
-12830.83). Trajectory models of two (BIC = -13119.94; AIC = -13098.44; L = -13090.44), three
(BIC = -12922.03; AIC = -12889.77; L = -12877.77) and five groups (BIC = -12951.53; AIC = -
12897.77; L = -12877.77) were also examined, but not retained on the basis of the BIC, AIC and
Likelihood indices.
The four trajectory groups consisted of (a) a Low-Increasing group1 (10% of children, [n
= 153; boys = 71, girls = 82]), (b) a Moderate-Increasing group (39%, [n = 630; boys = 334, girls
= 296]), (c) a Moderate-Fast Increasing group (32%, [n = 506; boys = 241, girls = 265]), and (d)
a High-Increasing group (19%, [n = 308; boys = 132, girls = 176]). All trajectory groups
displayed significant increasing trends over time (mostly quadratic, with exception of the Low-
Increasing trajectory which showed linear trend). A chi-square analysis showed significant sex
differences in NK trajectory group membership, χ2 (3) = 9.46, p < .05, with a higher proportion
of boys in the Moderate-Increasing trajectory, and a higher proportion of girls in the High-
Increasing trajectory.
Results from the repeated-measure ANOVA on NK scores showed significant group by
time interaction, F(8.48, 1848.43) = 73.93, p < .001, η2 = .25, and sex by time interaction,
F(2.83, 1848.43) = 3.19, p < .05, η2 = .005, but no significant group by sex interaction.
Significant differences in NK across all trajectories were found (p < .001), except between the
Moderate-Fast Increasing trajectory (M = 15.45, SD = 1.55) and the High-Increasing trajectory
(M = 15.26, SD = 2.25) at age 6. The Low-Increasing trajectory improved significantly from ages
5 to 6, and ages 6 to 7 (p < .05), but not from ages 4 to 5 (p > .05). The Moderate-Increasing
trajectory improved between all ages (p < .05); the Moderate-Fast Increasing trajectory improved
on NK from ages 4 to 5, and ages 5 to 6 (p < .05); and the High-Increasing trajectory improved
1 By convention, two terms are used to qualify the trajectory; the first reflects the initial level (e.g., low, moderate, high) of the NK score, whereas the second indicates the developmental course (i.e., slope) of the phenotype (e.g., increase, decrease, no-change or persistent).
24
PREDICTORS OF LOW NUMBER KNOWLEDGE
on NK from ages 4 to 5, and from 5 to 6 (p < .05), but not from ages 6 to 7 (p > .05).
Developmental patterns of NK also differed across sex at age 7 (p < .01). Boys significantly
progressed in their NK from ages 6 to 7 (p < .01), whereas girls did not (p > .05).
Children’s NK was interpreted in reference to the developmental level conversion chart
of the Number Knowledge Test (Okamoto & Case, 1996). This conversion chart provides a
rough index of the developmental levels normally reached as a function of age. Children
following the Low-Increasing trajectory, although progressing over time, systematically trailed
in terms of expected acquisition of NK. These children were characterized by a persistent lower
performance in NK from the end of preschool to grade 1. Over these three years (age 4 to 7
years), these children maintained the equivalent of a two-year delay in expected NK, thus
signaling a substantial risk of mathematics underachievement in later grades.
Prediction of Later Math Achievement from NK Trajectories
Results from the repeated-measures ANOVA on math scores at ages 8 and 10 showed
significant group by time interaction, confirming the expected difference in later mathematic
achievement according to NK trajectories (p < .01), but with the exception of the Moderate-Fast
Increasing and the High-Increasing trajectories at age 10 (p >.05). The significant sex by time
interactions, but no significant sex by group interaction, further painted an evolving picture of
sex differences in math: boys performed better in math than girls at age 8 (p < .05), but not at age
10 (p < .05). Most importantly, boys and girls of the Low-Increasing group both systematically
trailed in mastering mathematical skills in comparison to the other trajectory groups, which also
brings additional support to the developmental trajectories of NK.
Family and Child Cognitive Predictors of Low NK Trajectory
25
PREDICTORS OF LOW NUMBER KNOWLEDGE
Descriptive statistics of the cognitive and family variables for each trajectory-group are
provided in Table 2. Table 3 shows the early family and child-cognitive profiles of the low
versus other trajectories of NK. Compared to children of other trajectories, children of the low
NK trajectory were from household of lower income, and were more likely to have parents
without high school diploma (30% of mothers, 38% of fathers); both mothers and fathers of
children in the Low-Increasing trajectory also perceived they had lower impact on their child
development than other parents. Children of the low NK trajectory also had lower memory-span,
visual-spatial skills, receptive vocabulary, as well as hampered cognitive development at 29
months.
Table 4 shows the results of the binary logistic regression models, which tested the
unique longitudinal contributions of these factors to membership in the Low-Increasing
trajectory of NK, again using the three other trajectories as the basis for comparison. The final,
best-fitting regression model accounted for 26% of the variance of membership in the Low-
Increasing trajectory. A combination of family and child cognitive factors accounted for this
prediction. Father diploma (but not mother diploma) and household income uniquely predicted
membership in the Low-Increasing trajectory, and their unique contribution was maintained even
after taking into account the other factors in the model, including cognitive variables. Mother
parenting expectancies (but not father’s) added to the prediction of low NK trajectory, but this
contribution was no longer significant in the final model, when child cognitive factors were
taken into account. Three of the four cognitive variables, early cognitive development, visual-
spatial skills, and memory span uniquely predicted membership in Low-Increasing trajectory of
NK. Thus, children from family with low income and father with poor education backgrounds,
who, at 29 months, trailed in reaching early learning milestones, and who had lower visual-
26
PREDICTORS OF LOW NUMBER KNOWLEDGE
spatial and memory skills at 41 months were more likely to fall on the Low-Increasing trajectory
of NK. According to the Odds ratio statistics, low early cognitive development best predicted
low profiles of NK (OR = .55), followed by the father diploma (OR = .70), the memory span
(OR= .80), the visual-spatial skills (OR = .86), and household income (OR = .87).
Discussion
This study was among the first to identify discrete patterns of developmental trajectories
of NK from late preschool to school entry, to document their related mathematic achievement in
elementary school, as well as their early family and cognitive predictors. Using a large
representative sample of children, four distinct developmental trajectories of NK were revealed
from ages 4 years to 7 years. Early development of NK was not linear, but rather varied in onset
level and rate of progression during the transition from late preschool to school entry.
Specifically, heterogeneity in children’s NK development was organized around a small Low-
Increasing performing group (10%), a Moderate-Increasing group (39%), a Moderate-Fast
Increasing group (32%) and a High-Increasing performing group (19%). Thus, as expected, two
groups were identified at the extreme (i.e., the Low-Increasing and the High-Increasing groups),
each accounting for a significant minority of children. Two other groups representing the
majority of children fell between these two groups and showed different change in NK over time.
As early as age 4, between-group differences in children’s level of NK were significant.
These differences were maintained throughout the elementary school years, suggesting long-term
prediction from early NK to later achievement. Of specific interest, children in the Low-
Increasing trajectory fell well behind other children – about a two-years behind –, and the gap
between these children and those in the other trajectories did not narrow during the course of the
school years. In grades 2 and 4, and compared to children in the other trajectories, children of the
27
PREDICTORS OF LOW NUMBER KNOWLEDGE
Low-Increasing trajectory were trailing in mathematics, reflecting a stable pattern of low early
NK leading to low mathematic achievement during elementary school. This low and persistent
trajectory depicted the most atypical course of NK in this sample, and its prevalence (10%) was
consistent with those previously reported for mathematics learning disability (Barbaresi et al.,
2005; Shalev et al., 2000). In other words, these children were already at risk of mathematic
difficulties in preschool, and these difficulties were generally maintained in elementary school,
suggesting that deficits in NK likely hampers later mathematic and academic achievement.
Compared to children of the other trajectories of NK, children in the Low-Increasing
trajectory-group were characterized by lower early cognitive development. Already at 29
months, they had not acquired basic verbal and counting competencies (e.g., naming colors,
counting objects and making small sentences), and at 41 months visual-spatial skills and memory
span, to the same level as the other children. These hampered cognitive skills were predictive
even after controlling for household income, parents education and outcome expectancies.
These predictive associations were only partly consistent with previous findings showing
that child cognitive abilities such as visual-spatial skills and memory-span underlie the
emergence of early numeracy (Bull et al., 2008; Geary, 2011; Soto-Calvo et al., 2015). As
suggested by the triple code model (Dehaene & Cohen, 1995), the symbolic representation of
numbers develops sequentially. Numerical quantities are handled at first in a non-symbolic way
by the approximate magnitude mechanism (analog code), used for estimating quantities. Then,
increasing exposure to language and informal educational activities lead children to learn number
words (verbal code) and to identify number symbols (visual code), e.g., Arabic digits,
corresponding to the quantities (Dehaene et al., 1999). Accordingly, we found hindered visual-
28
PREDICTORS OF LOW NUMBER KNOWLEDGE
spatial skills (which may be involved in the analog and visual codes) to predict low NK.
However, early language did not emerge as a significant predictor of NK.
Indeed, our results did not support poor receptive vocabulary skills (verbal code) at 41
months as a significant cognitive risk factor for low-NK profiles. One possible reason for this
failed prediction is that disruption in receptive vocabulary might characterize children with
severe mathematic learning disability, but not children at-risk of early NK difficulties. As for
NK, receptive vocabulary relies on both phonological and semantic components of language.
Deficiency was found in the language-based phonological loop for children diagnosed with a
severe mathematic learning disability, i.e., dyscalculia (Geary, 2011; Geary, Hoard, Byrd-
Craven, Nugent, & Numtee, 2007). However, contrary to children with dyscalculia, those
experiencing mathematic difficulties, but not reaching the clinical criteria for dyscalculia, were
found with a superior phonemic system to keep information in memory (Geary, 2011; Geary et
al., 2007).
It is also possible that specific language components may be differentially associated with
specific mathematic skills, such as NK. For example, a recent longitudinal study following
children from ages 6 to 9 years showed that vocabulary and listening comprehension predict to
geometry, but not arithmetic or algebra (Vukovic & Lesaux, 2013). LeFevre et al. (2010) have
also shown that linguistic skills such as elision and receptive vocabulary (PPVT) of preschool
and kindergarten children uniquely predicted number naming, but not nonlinguistic arithmetic
during the same year. Although strongly associated with numeration, linguistic skills were
weakly associated with magnitude comparison two years later. Accordingly, receptive
vocabulary may be associated to some components of NK such as counting skills, but not with
magnitude comparisons, or procedural understanding of whole numbers. The content of language
29
PREDICTORS OF LOW NUMBER KNOWLEDGE
input also impact on NK development. Children who hear great amounts of spatial language
outperform on nonverbal spatial tasks associated to NK development, such as the Block Design
subtest of the WPPSI (Pruden, Levine, & Huttenlocher, 2011). This shows that the role of
language in NK development vary depending upon the language inputs and the NK outcomes.
One final reason for this lack of significance may be that poor receptive vocabulary may
have been accounted for by the early cognitive development index at 29 months, which also had
a strong language component. To evaluate this possibility, we simply removed this measure from
the regression, and found that receptive vocabulary at 41 months became significant (although
with an odd ratio barely different from one; results not shown). Thus, early cognitive
development at 29 months, which not only included language but also predicted receptive
language at 41 months, appeared as a key general factor in predicting low NK.
Interestingly, even when controlling for early-life family factors, father diploma, but not
mother diploma, still uniquely predicted low NK. Children of fathers with low educational
background were more at risk of low NK, suggesting that fathers with high education might
prevent the development of low numeracy and later mathematic difficulties. It does not mean
that mother diploma is not important for early NK, but it rather shows the increment of higher
paternal education, once the covariance with maternal education is taken into account. One
possible explanation for this result is that fathers with higher degrees might provide more support
for learning. Indeed, compared to parents from lower income and schooling, parents with higher
educational background were found to engage more frequently with their children in a broader
range of explicit mathematical-related activities (LeFevre et al., 2009; Levine et al., 2010;
Siegler, 2009; Siegler & Ramani, 2009). Accordingly, it is possible that fathers with low
educational background were less engaged with their children in mathematic-related activities,
30
PREDICTORS OF LOW NUMBER KNOWLEDGE
resulting in slow and/or age inappropriate development of specific cognitive abilities involved in
NK and mathematics. Beyond parental involvement, fathers with low education and their low
NK child could share common vulnerability to poor cognitive abilities. The role of fathers in the
development of NK should be further explored in future research to address this question.
Fathers’ expectancies regarding their children development was, however, not
significantly associated to NK, whereas low mothers’ outcome expectancies were initially
predictive to low NK trajectory. However, it did not uniquely predict to low NK development
once child cognitive variables were entered in the model. Parenting style and perceptions had
previously been associated with children’s school-trajectory (LeFevre et al., 2009; Kouimtzi &
Stogiannidou, 2009). Here, mothers’ expectancies about her developing child may have been
associated to NK through one or many child-cognitive factors and thus, lost its significance when
entering the child cognitive abilities into the analyses. Parental beliefs and behaviors have been
associated to knowledge such as counting objects and learning vocabulary (Glascoe & Leew,
2010; Kouimtzi & Stogiannidou, 2009; LeFevre et al., 2009). Therefore, it is possible that
mother’s expectancies promote the development of NK through children’s cognitive abilities and
knowledge implied in mathematics. It may also be that our measure of parenting outcome
expectancies was too general, not specifically tapping into children’s learning of NK, but rather
providing a global perspective.
Implications for Research and Practices
These findings have implications for early identification of children at risk of persistent
mathematic difficulties and school underachievement, as well as for preventive intervention.
First, it gives a clear portrait of children from preschool age that, without a formal diagnosis,
showed persistently lower math skills compared to their typically achieving peers. The specific
31
PREDICTORS OF LOW NUMBER KNOWLEDGE
cognitive skills – memory span, and visual-spatial skills predicted NK over and above the family
characteristics, such as household income and father education, and the early cognitive
development and knowledge (of colors, counting and speaking) at 29 months. Thus, children
already performing poorly on these characteristics at early age should deserve special attention to
alleviate later NK and math difficulties. The attention given to these children “at risk” should
also persist in early grade school. One potential approach to promote achievement among
children with initial risk of underachievement is to capitalize on teacher’s instructions. For
example, higher order instruction, such as teaching strategies that highlight deduction and critical
thinking in conjunction with skill acquisition, tends to improve achievement more than a rote
basic skills approach, especially among students academically at risk (Clements, Agodini, &
Harris, 2013; Connor, Morrison, Fishman, Schatschneider, & Underwood, 2007; Hamre &
Pianta, 2005; Xue & Meisels, 2004).
Second, this study also revealed continuity from early NK to mathematical skills during
elementary school, and points to early NK as a potential target for early prevention. This result
reinforces the predictive value of NK and highlights the transition from late preschool to school
entry as a crucial period to assess and promote early numeracy. Given this, children’s NK skills
should be assessed before school starts in order to provide additional support as soon as
difficulties emerge in this area.
Third, the present study identified the specific cognitive and family predictors of low NK
skills from preschool to school entry. However, to improve our understanding of the mechanisms
underlying early mathematical development, future studies should further investigate the
association between the approximate magnitude system and the level of NK, over and above
those cognitive and family predictors. Previous research supported the approximate magnitude
32
PREDICTORS OF LOW NUMBER KNOWLEDGE
system at preschool age to uniquely predict later performance in mathematics (Fuchs & McNeil,
2013; Libertus et al., 2011; Mazzocco, Feigneson, & Halberda, 2011). However, the extend to
which children with low ability to approximate magnitude difference also perform poorly in their
early NK is still unknown.
Limitations
The present study should be interpreted in the context of its limitations. First, the
administration of the adapted-NKT departed from Okamoto and Case (1996), which may have
resulted in a ceiling effect at specific ages. At ages 4, 5, and 7 this procedure did not result in a
significant divergence from Okamoto and Case (1996), as most children did not reach the 60%
mark allowing them to move to the next level (see Method). Post-hoc examination of the score
distributions did not signal a meaningful ceiling effect, as only 0.3%, 0.6%, and 1.5% of children
reached the scale, respectively. However, at age 6 years, a majority of children successfully
reached 60% of success at the top level (i.e., Level 1), and 5.6% of children scored the scale
maximum. This suggests a ceiling effect, that is, a compression of scores at the top of the scale.
This ceiling effect at age 6, may have partly accounted for the curvilinear and asymptotic form of
the two highest trajectories (i.e., the High-Increasing and the Moderate-Fast Increasing
trajectories; see Figure 1), as well as of their decreasing gap at ages 6 and 7. Although this may
have affected the shape of these higher-performing trajectory groups, it was unlikely to play an
important role in children’s trajectory membership, especially for the Low-Increasing group, as it
was based on four longitudinal data points. What may have played against the reliability of all
trajectory groups was the lower internal consistency of the adapted-NKT at age 5. But again, the
impact of this limited reliability on the proper NK characterization of children is likely to have
been attenuated by its longitudinal aggregation to three other, more reliable data points.
33
PREDICTORS OF LOW NUMBER KNOWLEDGE
Interestingly, despite the low reliability at age 5, individual differences in the adapted-NKT were
highly stable across the four ages (r = .74 - .92).
It is also possible that the NK measure did not reflect what is learn in math at school, and
that children in the low trajectory of NK would have looked different if assessed through other
measures. However, with respect to the assessment of performance in mathematics (NK and
standardized test of mathematics achievement), our findings suggest that the information
provided by multiple informants was roughly equivalent.
Second, some family predictors did not specifically tap into the mathematic domain and
were only indirectly linked to NK (e.g., parenting outcome expectancies). Our parenting outcome
expectancies also had less than optimal internal consistency. Similarly, our child cognitive
predictors (e.g., visual-spatial measure) were not numerical in nature and only one item in the
early cognitive development measure was mathematically focused. The early cognitive
development measure was confined to three items and based on the assessment of only one rater,
the person “most knowledgeable about the child”, limiting its quantitative and conceptual
validity. Finally, our conclusions may only apply to children from the province of Quebec,
Canada. It should also be noted that the present results may be specific to this developmental
window, and predictors of low-NK trajectories could differ at different ages (Haworth, Kovas,
Petrill, & Plomin, 2007).
To conclude, this study convincingly showed that children with low cognitive abilities
and impoverished living conditions early in life are at greater risk of low NK from late preschool
and school entry, and of persistent lower math achievement in elementary school.
Acknowledgements
This research was supported by grants from the Québec Ministry of Health, the Fonds
34
PREDICTORS OF LOW NUMBER KNOWLEDGE
Québécois de la Recherche sur la Société et la Culture (FQRSC), the Social Science and
Humanities Research Council (SSHRC), the Canadian Institutes for Health Research (CIHR),
and grant 11.G34.31.003 from the Russian Federation. We are grateful to the parents of the
children participants to the Québec Longitudinal Study of Child Development (QLSCD).
References
Aunola, K., Leskinen, E., Lerkkanen, M.-K., & Nurmi, J.-E. (2004). Developmental dynamics of
math performance from preschool to grade 2. Journal of Educational Psychology, 96,
699–713. doi:10.1037/0022-0663.96.4.699
Bandura, A. (1997). Self-efficacy. The exercise of control. W. H. Freeman and Co, New York.
Barbaresi, W. J., Katusic, S. K., Colligan, R. C., Weaver, A. L., & Jacobsen, S. J. (2005). Math
learning disorder: Incidence in a population-based birth cohort, 1976–82, Rochester,
Minn. Ambulatory Pediatrics , 5, 281–289. doi:10.1367/A04-209R.1
Benasich, A. A., & Brooks-Gunn, J. (1996). Maternal attitudes and knowledge of child-rearing:
Associations with family and child outcomes. Child Development, 67, 1186–1205.
doi:10.1111/j.1467-8624.1996.tb01790.x
Bermejo, V. (1996). Cardinality development and counting. Developmental Psychology, 32,
263–268. doi:10.1037/0012-1649.32.2.263
Boivin, M., Pérusse, D., Dionne, G., Saysset, V., Zoccolillo, M., Tarabulsy, G. M., …Tremblay,
R. (2005). The genetic-environmental etiology of parents’ perceptions and self-assessed
behaviours toward their 5-month-old infants in a large twin and singleton sample.
Journal of Child Psychology and Psychiatry, 46, 12–630. doi:10.1111/j.1469-
7610.2004.00375.x
Bornstein, M. C. (2002). Handbook of parenting (vols 1–4). Mahwah, NJ: Lawrence Erlbaum.
35
PREDICTORS OF LOW NUMBER KNOWLEDGE
Blair, C., & Raver, C. C. (2015). School readiness and self-regulation: A developmental
psychobiological approach. Annual Review of Psychology, 66, 711–731.
doi:10.1146/annurev-psych-010814-015221
Bull, R., Espy, K. A., & Wiebe, S. A. (2008). Short-term memory, working memory, and
executive functioning in preschoolers: Longitudinal predictors of mathematical
achievement at age 7 years. Developmental Neuropsychology, 33, 205–228.
doi:10.1080/87565640801982312
Canadian Test Center (1992). Canadian Achievement Test, Second Edition. Retrieved from
http://www.canadiantestcentre.com/
Carter, M. A., Dubois, L., & Ramsay, T. (2010). Examining the relationship between obesity and
math performance among Canadian school children: A prospective analysis.
International Journal of Pediatric Obesity, 5, 412–419. doi:10.3109/174771609034969
Clements, D. H., Agodini, R., & Harris, B. (2013). Instructional practices and student math
achievement: Correlations from a study of math curricula. (NCEE Evaluation Brief ED-
04-CO-0112/0003). Washington, DC: National Center for Education Evaluation and
Regional Assistance, Institute of Education Sciences, U.S. Department of Education.
Connor, C. M., Morrison, F. J., Fishman, B. J., Schatschneider, C., & Underwood, P. (2007). The
early years: Algorithm-guided individualized reading instruction. Science, 315, 464–465.
doi:10.1126/science.1134513
Côté, S. M., Mongeau, C., Japel, C., Xu, Q., Seguin, J. R., & Tremblay, R. E. (2013). Child care
quality and cognitive development: Trajectories leading to better preacademic skills.
Child Development, 84, 752–766. doi:10.1111/cdev.12007
Dehaene, S., & Cohen, L. (1995). Towards an anatomical and functional model of number
36
PREDICTORS OF LOW NUMBER KNOWLEDGE
processing. Mathematical Cognition, 1, 83–120.
Dehaene, S., & Cohen, L. (1997). Cerebral pathways for calculation: Double dissociation
between rote verbal and quantitative knowledge of arithmetic. Cortex, 33, 219–250.
doi:10.1016/S0010-9452(08)70002-9
Dehaene, S., Piazza, M., Pinel, P., & Cohen, L. (2003). Three parietal circuits for number
processing. Cognitive Neuropsychology, 20, 487–506. doi:10.1080/02643290244000239
Dehaene, S., Spelke, E., Pinel, P., Stanescu, R., & Tsivkin, S. (1999). Sources of mathematical
thinking: Behavioral and brain-imaging evidence. Science, 284, 970–974.
doi:10.1126/science.284.5416.970
Duncan, G. J., Dowsett, C. J., Claessens, A., Magnuson, K., Huston, A. C., Klebanov, P., …
Japel, C. (2007). School readiness and later achievement. Developmental Psychology, 43,
1428–1446. doi:10.1037/0012-1649.43.6.1428
Dunn, L. M. & Dunn, L. M. (1997). Peabody Picture Vocabulary Test (3rd ed.). Circle Pines,
MN, US: American Guidance Service.
Dunn, L. M., Theriault-Whalen, C. M., & Dunn, L. M. (1993). Échelle de Vocabulaire en
Images Peabody. Adaptation française du Peabody Picture Vocabulary Test-Revised.
Manuel pour les formes A et B. Toronto, Ontario, CANADA: PSYCAN.
Feigenson, L., Dehaene, S., & Spelke, E. (2004). Core systems of number. TRENDS in Cognitive
Sciences, 8, 307–314. doi:10.1016/j.tics.2004.05.002
Feigenson, L., Libertus, M. E., & Halberda, J. (2013). Links between the intuitive sense of
number and formal mathematics ability. Child Development Perspectives, 7, 74–79.
doi:10.1111/cdep.12019
Flipsen, P. (1998). Assessing receptive vocabulary in small-town Canadian kindergarten
37
PREDICTORS OF LOW NUMBER KNOWLEDGE
children: Findings for the PPVT-R. Journal of Speech-Language Pathology and
Audiology, 22, 88–93.
Friso-van den Bos, I., Kroesbergen, E. H., Van Luit, J. E., Xenidou-Dervou, I., Jonkman, L.
M., Van der Schoot, M., Van Lieshout, E. C. (2015). Longitudinal development of
number line estimation and mathematics performance in primary school children. Journal
of Experimental Child Psychology, 134, 12–29. doi:10.1016/j.jecp.2015.02.002
Fuchs, M. V. & McNeil, N. M. (2013). ANS acuity and mathematics ability in preschoolers from
low-income homes: contributions of inhibitory control. Developmental Science, 16, 136–
148. doi:10.1111/desc.12013.
Garon-Carrier, G., Boivin, M., Ouellet, E., Tremblay, R. E., & Dionne, G. (in preparation).
Assessing children’s computational skills: Validation of an adapted version of the
Canadian Achievement Test - Second Edition for 10 year olds.
Geary, D. C. (2004). Mathematics and learning disabilities. Journal of Learning Disabilities, 37,
4–15. doi:10.1177/00222194040370010201
Geary, D. C. (2011). Consequences, characteristics, and causes of mathematical learning
disabilities and persistent low achievement in mathematics. Journal of Developmental &
Behavioral Pediatrics, 32, 250–263. doi:10.1097/DBP.0b013e318209edef.
Geary, D. C., Bailey, D. H., & Hoard, M. K. (2009). Predicting mathematical achievement and
mathematical learning disability with a simple screening tool: The number sets test.
Journal of Psychoeducational Assessment. 27, 265–279. doi:10.1177/0734282908330592
Geary, D. C., Hoard, M. K., Byrd-Craven, J., & DeSoto, M. C. (2004). Strategy choices in
38
PREDICTORS OF LOW NUMBER KNOWLEDGE
simple and complex addition: Contributions of working memory and counting knowledge
for children with mathematical disability. Journal of Experimental Child Psychology, 88,
121–151. doi:10.1016/j.jecp.2004.03.002
Geary, D. C., Hoard, M. K., Byrd-Craven, J., Nugent, L., & Numtee, C. (2007). Cognitive
mechanisms underlying achievement deficits in children with mathematical learning
disability. Child Development, 78, 1343–1359. doi:10.1111/j.1467-8624.2007.01069.x
Gersten, R., Jordan, N. C., & Flojo, J. R. (2005). Early identification and interventions for
students with mathematics difficulties. Journal of Learning Disabilities, 38, 293–304.
doi:10.1177/00222194050380040301
Gersten, R., Clarke, B. S., & Jordan, N. C. (2007). Screening for mathematics difficulties in K-3
students. NH: RMC. Research Corporation, Center on Instruction. Retrieved from
http://www.centeroninstruction.org/files/COI%20Math%20Screening1.pdf
Glascoe, F. P., & Leew, S. (2010). Parenting Behaviors, Perceptions and Psychosocial Risk:
Impact on Child Development. Pediatrics, 125, 313–319. doi:10.1542/peds.2008-3129
Göbel, S. M., Watson, S. E. Lervåg, A., & Hulme, C. (2014). Children’s arithmetic development:
It is number knowledge, not the approximate number sense, that counts. Psychological
Science, 25, 789–798. doi:10.1177/0956797613516471
Griffin, S., Case, R., & Siegler, R. (1994). Rightstart: Providing the central conceptual
prerequisites for the first formal learning of arithmetic to students at risk for school
failure. Classroom lessons: Integrating cognitive theory and classroom practice.
Cambridge, MA, US: The MIT Press.
Hamre, B. K., & Pianta, R. C. (2005). Can instructional and emotional support in the first grade
39
PREDICTORS OF LOW NUMBER KNOWLEDGE
classroom make a difference for children at risk of school failure? Child Development,
76, 949–967. doi:10.1111/j.1467-8624.2005.00889.x
Haworth, C. M. A., Kovas, Y., Petrill, S. A., & Plomin, R. (2007). Developmental origins of low
mathematics performance and normal variation in twins from 7 to 9 years. Twin Research
and Human Genetics, 10, 106–117. doi:10.1375/twin.10.1.106
Hyde, D. C., & Spelke, E. S. (2011). Neural signatures of number processing in human infants:
Evidence for two core systems underlying numerical cognition. Developmental Science,
14, 360–371. doi:10.1111/j.1467-7687.2010.00987.x
Jetté, M., & Des Groseillers, L. (2000a). Survey description and methodology. In Longitudinal
Study of Child Development in Quebec (ELDEQ 1998-2002) (Vol. 1, No. 1). Quebec
City, Quebec, Canada: Institut de la statistique du Québec. Retrieved from
http://www.jesuisjeserai.stat.gouv.qc.ca/bebe/pdf/baby_no1-1.pdf
Jetté, M., & Des Groseillers, L. (2000b). Family, child care, and neighbourhood characteristics.
In Longitudinal Study of Child Development in Quebec (ELDEQ 1998-2002) (Vol. 1,
No. 2). Quebec City, Quebec, Canada: Institut de la statistique du Québec. Retrieved
from http://www.iamillbe.stat.gouv.qc.ca/bebe/pdf/baby_no2-1.pdf
Jones, B. L., & Nagin, D. S. (2007). Advances in Group-Based Trajectory Modeling and an SAS
Procedure for Estimating Them. Sociological Methods Research, 35, 542–571.
doi:10.1177/0049124106292364
Jordan, N. C., Glutting, J., & Ramineni, C. (2010). The importance of number sense to
mathematics achievement in first and third grades. Learning and Individual Differences,
20, 82–88. doi:10.1016/j.lindif.2009.07.004
Jordan, N. C., Hanich, L. B., & Kaplan, D. (2003). Arithmetic fact mastery in young children: A
40
PREDICTORS OF LOW NUMBER KNOWLEDGE
longitudinal investigation. Journal of Experimental Child Psychology, 85, 103–119.
doi:10.1016/S0022-0965(03)00032-8
Jordan, N. C., Kaplan, D., Locuniak, M., & Ramineni, C. (2007). Predicting first-grade math
achievement from developmental number sense trajectories. Learning Disabilities
Research & Practice, 22, 36–46. doi:10.1111/j.1540-5826.2007.00229.x
Jordan, N. C., Kaplan, D., Oláh, L. N., & Locuniak, M. N. (2006). Number sense growth in
kindergarten: A longitudinal investigation of children at risk for mathematics difficulties.
Child Development, 77, 153–175. doi:10.1111/j.1467-8624.2006.00862.x
Jordan, N. C., Kaplan, D., Ramineni, C., & Locuniak, M. N. (2009). Early math matters:
Kindergarten number competence and later mathematics outcomes. Developmental
Psychology, 45, 850–867. doi:10.1037/a0014939
Jordan, N. C., & Levine, S. C. (2009). Socioeconomic variation, number competence, and
mathematics learning difficulties in young children. Developmental Disabilities Research
Reviews, 15, 60–68. doi:10.1002/ddrr.46
Klibanoff, R. S., Levine, S. C., Huttenlocher, J., Vasilveva, M., & Hedges, L. V. (2006).
Preschool children's mathematical knowledge: The effect of teacher "math talk".
Developmental Psychology, 42, 59–69. doi:10.1037/0012-1649.42.1.59
Kouimtzi, E-M., & Stogiannidou, A. (2009). Mothers of low achievers: Their perception as to
their role in their children's school success and the willingness to seek psychoeducational
support. Psychology: The Journal of the Hellenic Psychological Society, 16, 280–301.
Lachance, J. A., & Mazzocco, M. M. (2006). A longitudinal analysis of sex differences in math
and spatial skills in primary school age children. Learning and Individual Differences,
16, 195–216. doi:10.1016/j.lindif.2005.12.001
41
PREDICTORS OF LOW NUMBER KNOWLEDGE
Laski, E. V., & Siegler, R. S. (2014). Learning From Number Board Games: You Learn What
You Encode. Developmental Psychology, 50, 853–864. doi:10.1037/a0034321
Laursen, B. P., & Hoff, E. (2006). Person-centered and variable-centered approaches to
longitudinal data. Merrill-Palmer Quarterly, 52, 377–389. doi:10.1353/mpq.2006.0029
Leblanc, N., Boivin, M., Dionne, G., Tremblay, R., & Pérusse, D. (2008). The development of
hyperactive/impulsive behaviors during the preschool years: The predictive validity of
parental assessments. Journal of Abnormal Child Psychology, 36, 977–987.
doi:10.1007/s10802-008-9227-7
LeFevre, J.-A., Fast, L., Skwarchuk, S.-L., Smith-Chant, B. L., Bisanz, J., Kamawar, D., …
Penner-Wilger, M. (2010). Pathways to mathematics: Longitudinal predictors of
performance. Child Development, 81, 1753–1767. doi:10.1111/j.1467-8624.2010.01508.x
LeFevre, J-A., Skwarchuk, S-L., Smith-Chant, B. L., Fast, L., Kamawar, D., & Bisanz, J. (2009).
Home numeracy experiences and children’s math performance in the early school years.
Canadian Journal of Behavioural Science, 41, 55–66. doi:10.1037/a0014532
Levine, S. C., Jordan, N. C., & Huttenlocher, J. (1992). Development of calculation abilities in
young children. Journal of Experimental Child Psychology, 53, 72–103.
doi:10.1016/S0022-0965(05)80005-0
Levine, S. C., Suriyakham, L. W., Rowe, M. L., Huttenlocher, J., & Gunderson, E. A. (2010).
What counts in the development of young children’s number knowledge? Developmental
Psychology, 46, 1309–1319. doi:10.1037/a0019671
Libertus, M. E., Feigenson, L., & Halberda, J. (2011). Preschool acuity of the approximate
number system correlates with school math ability. Developmental Science, 14, 1292–
1300. doi:10.1111/j.1467-7687.2011.01080.x
42
PREDICTORS OF LOW NUMBER KNOWLEDGE
Lipton, J. S, & Spelke, E. S. (2003). Origins of number sense: Large number discrimination in
human infants. Psychological Science, 14, 396–401. doi:10.1111/1467-9280.01453
Lundetræ, K., Gabrielsen, E., & Mykletun, R. (2010). Do basic skills predict youth
unemployment (16‐ to 24‐year‐olds) also when controlled for accomplished upper‐
secondary school? A cross‐country comparison. Journal of Education and Work, 23,
233–254. doi:10.1080/13639081003745439
Magnuson, K., Duncan, G. J., Lee, K. T. H., & Metzger, M. W. (2016). Early school adjustment
and educational attainment. American Educational Research Journal, 53, 1198–1228.
doi:10.3102/0002831216634658
Mazzocco, M. M., Feigenson, L., & Halberda, J. (2011). Preschoolers' precision of the
approximate number system predicts later school mathematics performance. PLoS One,
6(9):e23749. doi:10.1371/journal.pone.0023749
Mazzocco, M. M., & Thompson, R. E. (2005). Kindergarten predictors of math learning
disability. Learning Disabilities Research & Practice, 20, 142–155. doi:10.1111/j.1540-
5826.2005.00129.x
McBride, B. A., Dyer, J. W., Liu, Y., Brown, G. L., & Hong, S. (2009). The differential impact
of early father and later student achievement. Journal of Educational Psychology, 101,
498–508. doi:10.1037/a0014238
McCrink, K., & Wynn, K. (2007). Ratio abstraction by 6-month-old infants. Psychological
Science, 18, 740–746. doi:10.1111/j.1467-9280.2007.01969.x
Melhuish, E. C., Sylva, K., Sammons, P., Siraj-Blatchford, I., Taggart, B., Phan, M. B., & Malin,
(2008). Preschool influences on mathematics achievement. Science, 321, 1161–1162.
doi:10.1126/science.1158808
43
PREDICTORS OF LOW NUMBER KNOWLEDGE
Muthén, L. K., & Muthén, B. O. (1998-2012). Mplus User’s Guide. Seventh Edition. Los
Angeles, CA: Muthén & Muthén
Nagin, D. S. (1999). Analyzing developmental trajectories: A semiparametric, group-based
approach. Psychological Methods, 4, 139–157. doi:10.1037/1082-989X.4.2.139
National Research Council. (2009). Mathematics in early childhood: learning paths toward
excellence and equity. Washington, DC: National Academy Press.
Nguyen, T., Watts, T. W., Duncan, G. J., Clements, D. H., Sarama, J. S., Wolfe, C., & Spitler,
M. E. (2016). Which preschool mathematics competencies are most predictive of fifth
grade achievement? Early Childhood Research Quartely, 36, 550–560.
doi:10.1016/j.ecresq.2016.02.003
Okamoto, Y., & Case, R. (1996). II. Exploring the microstructure of children’s central
conceptual structures in the domain of number. Monographs of the Society for Research
in Child Development, 61, 27–58. doi:10.1111/j.1540-5834.1996.tb00536.x
Parke, R. D., & Buriel, R. (1998). Socialization in the family: Ethnic and
ecological perspectives.
In W. Damon (Ed.), Handbook of child psychology (5th ed, vol. 3 pp.
135–210). New York: John Wiley & Sons.
Parsons, S., & Bynner, J. (1997). Numeracy and employment. Education + Training, 39, 43–
51. doi:10.1108/00400919710164125
Pasta, D. J. (2009). Learning when to be discrete: Continuous vs. categorical predictors. SAS
Global Forum 2009. Statistics and data analysis, 248, 1–10.
Piaget, J. (1977). The Development of Thought: Equilibration of Cognitive Structures. New-
York: The Viking Press
44
PREDICTORS OF LOW NUMBER KNOWLEDGE
Powers, D. A., & Xie, Y. (2008). Statistical methods for categorical data analysis, 2nd Edition.
Howard House, England: Emerald.
Pruden, S. M., Levine, S. C., & Huttenlocher, J. (2011). Children’s spatial thinking: Does talk
about the spatial world matter? Developmental Science, 14, 1417–1430.
doi:10.1111/j.1467-7687.2011.01088.x
Ramani, G. B., Siegler, R. S., & Hitti, A. (2012). Taking it to the classroom: Number board
games as a small group learning activity. Journal of Educational Psychology, 104, 661–
672. doi:10.1037/a0028995
Romano, E., Babchishin, L., Pagani, L. S., & Kohen, D. (2010). School readiness and later
achievement: Replication and extension using a nationwide canadian survey.
Developmental Psychology, 46, 995–1007. doi:10.1037/a0018880
Sadler, P. M., & Tai, R. H. (2007). The two high-school pillars supporting college science.
Science, 317, 457−458. doi:10.1126/science.1144214
Sasser, T. R., Bierman, K. L., & Heinrichs, B. (2015). Executive functioning and school
adjustment: The mediational role of pre-kindergarten learning-related behaviors. Early
Childhood Research Quarterly, 30(Part A), 70–79. doi:10.1016/j.ecresq.2014.09.001
Sattler, J. M. (2001). Assessment of children: Cognitive applications. San Diego, CA: Sattler.
Shalev, R. S., Auerbach, J., Manor, O., & Gross-Tsur, V. (2000). Developmental dyscalculia:
prevalence and prognosis. European Child & Adolescent Psychiatry, 9(Suppl 2): S58.
doi:10.1007/s007870070009
Siegler, R. (2009). Improving the numerical understanding of children from low-income
families. Child Development, 3, 118−124. doi:10.1111/j.1750-8606.2009.00090.x
Siegler, R. S., & Booth, J. L. (2004). Development of numerical estimation in young children.
45
PREDICTORS OF LOW NUMBER KNOWLEDGE
Child Development, 75, 428−44. doi:10.1111/j.1467-8624.2004.00684.x
Siegler, R. S., Duncan, G. J., Davis-Kean, P., Duckworth, K., Claessens, A., Engel, M., …Chen,
M. (2012). Early predictors of high school mathematics achievement. Psychological
Science, 23, 691−697. doi:10.1177/0956797612440101
Siegler, R. S., & Ramani, G. B. (2009). Playing linear number board games—but not circular
ones—improves low-income preschoolers’ numerical understanding. Journal of
Educational Psychology, 101, 545–560. doi:10.1037/a0014239
Soto-Calvo, E., Simmons, F. R., Willis, C., & Adams, A-M. (2015). Identifying the cognitive
predictors of early counting and calculation skills: Evidence from a longitudinal study.
Journal of Experimental Child Psychology, 140, 16–37. doi:10.1016/j.jecp.2015.06.011
SPSS Inc. (2011). SPSS Base 20.0 for Windows User's Guide. SPSS Inc., Chicago, IL.
vanMarle, K. (2013). Infants use different mechanisms to make small and large number ordinal
judgments. Journal of Experimental Child Psychology, 114, 102–110. doi:
10.1016/j.jecp.2012.04.007
Viljaranta, J., Lazarides, R., Aunola, K., Räikkönen, E., & Nurmi, J.-E. (2015). The different
role of mothers' and fathers' beliefs in the development of adolescents' mathematics and
literacy task values. International Journal of Gender, Science and Technology, 7, 297–
317.
von Aster, M. G., & Shalev, R. S. (2007). Number development and developmental dyscalculia.
Developmental Medicine & Child Neurology, 49, 868–873. doi:10.1111/j.1469-
8749.2007.00868.x
Vukovic, R. K., & Lesaux, N. K. (2013). The language of mathematics: Investigating the ways
46
PREDICTORS OF LOW NUMBER KNOWLEDGE
language counts for children’s mathematical development. Journal of Experimental Child
Psychology, 115, 227–224. doi:10.1016/j.jecp.2013.02.002
Watts, T. W., Duncan, G. J., Siegler, R. S., & Davis-Kean, P. E. (2014). What's past is prologue:
Relations between early mathematics knowledge and high school. Educational
Researcher, 43, 352–360. doi:10.3102/0013189X14553660
Wechsler, D. (1989). Manual for the Wechsler Preschool and Primary Scale of Intelligence–
Revised. San Antonio, TX, US: Psychological Corporation.
Welsh, J. A., Nix, R. L., Blair, C., Bierman, K. L., & Nelson, K. E. (2010). The development of
cognitive skills and gains in academic school readiness for children from low-income
families. Journal of Educational Psychology, 102, 43–53. doi:10.1037/a0016738
Willms, J. D., & Shields, M. (1996). A measure of socioeconomic status for the National
Longitudinal Study of Children. Draft prepared as a reference for researchers conducting
analysis of the first wave of data from the Canadian National Longitudinal Study of
Children and Youth (NLSCY).
Wolniak, G. C. (2016). Examining STEM bachelor’s degree completion for students with
differing propensities at college entry. Journal of College Student Retention: Research,
Theory & Practice, 18, 287–309. doi:10.1177/1521025115622782
Xu, F., Spelke, E. S., & Goddard, S. (2005). Number sense in human infants. Developmental
Science, 8, 88–101. doi:10.1111/j.1467-7687.2005.00395.x
Xue, Y., & Meisels, S. J. (2004). Early literacy instruction and learning in kindergarten:
Evidence from the Early Childhood Longitudinal Study–Kindergarten Class of 1998–
1999. American Educational Research Journal, 41, 191–229.
Zelazo, P. D., Jacques, S., Burack, J. A., & Frye, D. (2002). The relation between theory of mind
47
PREDICTORS OF LOW NUMBER KNOWLEDGE
and rule use: Evidence from persons with autism-spectrum disorders. Infant & Child
Development, 11, 171–195. doi:10.1002/icd.304
Zorzi, M., Priftis, K., & Umiltà, C. (2002). Neglect disrupts the mental number line. Nature, 417,
138–139. doi:10.1038/417138a
48
PREDICTORS OF LOW NUMBER KNOWLEDGE
4 5 6 70
2
4
6
8
10
12
14
16
18High/increasingModerate/fast increasingModerate/increasingLow/increasing median scores
Age
Num
ber K
now
ledg
e Sc
ores
Figure 1. Developmental trajectories of number knowledge from 4 to 7 years of age (N = 1597):
Low-Increasing (n =153; 10%), Moderate-Increasing (n =630; 39%), Moderate-Fast Increasing
(n =506; 32%), High-Increasing (n =308; 19%), and the median number knowledge score. Data
courtesy of the Quebec Institute of Statistics.
49
Running Head: PREDICTORS OF LOW NUMBER KNOWLEDGE
Table 1Longitudinal measures by multi-informants and child’s age-at-assessment
Note. PMK = Person most knowledgeable about the child
Variables Informant Child age at data collection
Method of Assessment
Family income PMK 5 months Questionnaire completed during aface-to-face interview
Mother diploma PMK 5 months Questionnaire completed during aface-to-face interview
Father diploma PMK 5 months Questionnaire completed during aface-to-face interview
Outcome expectancies (mother)
Mother 5, 17, 29 monthsAverage score across time
Self-administered questionnaire
Outcome expectancies (father)
Father 5, 17, 29 monthsAverage score across time
Self-administered questionnaire
Early cognitive development
PMK 29 months Questionnaire completed during aface-to-face interview
Memory-span Child 41 months Visually Cued Recall TaskReceptive vocabulary Child 41 months Peabody Picture Vocabulary TestVisual-spatial skills Child 41 months Block Design subtest of the WPPSI-RNumber knowledge Child 4, 5, 6, 7 years Number Knowledge TestMath achievement Child 8, 10 years Mathematics subtest of the Canadian
Achievement Test
50
PREDICTORS OF LOW NUMBER KNOWLEDGE
Table 2Descriptive statistics of family and children variables for each trajectory-group
Note. Low-I refers to the Low-Increasing trajectory of NK, Moderate-I refers to the Moderate-Increasing trajectory of NK, Moderate-Fast I refers to the Moderate-Fast Increasing trajectory of NK, High-I refers to the High-Increasing trajectory of NK.Percentage is reported for categorical variables (%); Mean and Standard deviation are reported for continuous variables† [means (SD)].Data are courtesy of the Quebec Institute of Statistics
Family and children variables Low-I(10%, n = 153)
Moderate-I(39%, n = 630)
Moderate-Fast I(32%, n = 506)
High-I(19%, n = 308)
Socio-demographic (5 months)Household income, 30 000$ or less per year 44.44 34.00 19.40 13.40Maternal education, no high school diploma 30.06 22.70 10.50 7.50Paternal education, no high school diploma 37.87 23.00 16.80 8.60Parenting expectancies (5 to 29 months)Outcome expectancies (mother)† 7.81 (1.79) 8.28 (1.59) 8.59 (1.38) 8.69 (1.25)Outcome expectancies (father)† 7.90 (1.87) 8.34 (1.51) 8.56 (1.33) 8.80 (1.17)Child cognitive abilities (41 months)Visual-spatial skills† 4.03 (3.16) 5.44 (3.21) 6.88 (3.75) 8.54 (4.23)Receptive vocabulary† 21.20 (10.58) 26.69 (12.67) 32.90 (14.56) 38.22 (15.27)Memory-span† 1.98 (1.37) 2.65 (1.80) 3.72 (2.36) 4.25 (2.38)Early cognitive development† 1.58 (0.99) 2.07 (0.92) 2.40 (0.80) 2.61 (0.67)
48
PREDICTORS OF LOW NUMBER KNOWLEDGE
Table 3 Family and children characteristics associated with trajectories of low number knowledge from 4 to 7 years of age
Family and children variablesMissing Low-I Other trajectories
P(n = 1597) (10%, n = 153) (90%, n = 1444)
Socio-demographic (5 months) Household income, 30 000$ or less per year 1.25 44.44 24.56 <.001 Maternal education, no high school diploma 0.06 30.06 15.18 <.001 Paternal education, no high school diploma 7.70 37.87 17.66 <.001Parenting expectancies (5 to 29 months) Outcome expectancies (mother)† 0.25 7.81 (1.79) 8.47 (1.46) <.001 Outcome expectancies (father)† 7.14 7.90 (1.87) 8.52 (1.39) <.001Child cognitive abilities (41 months) Visual-spatial skills† 5.57 4.03 (3.16) 6.60 (3.82) <.001 Receptive vocabulary† 6.51 21.20 (10.58) 31.38 (14.66) <.001 Memory-span† 8.45 1.98 (1.37) 3.37 (2.24) <.001 Early cognitive development† 0.56 1.58 (0.99) 2.30 (0.86) <.001Note. Significant at p < .05; Low-I refers to the Low-Increasing trajectory of NK. Other trajectories refers to the Moderate-Increasing, Moderate-Fast increasing, and the High-Increasing trajectories of NK. Missing indicates the percent of missing data (%). Chi-square tests between the Low trajectory and the other trajectories were used for categorical variables (%).ANOVAs between the Low trajectory and the other trajectories were used for continuous variables† [mean, (SD)].Data are courtesy of the Quebec Institute of Statistics
49
PREDICTORS OF LOW NUMBER KNOWLEDGE
Table 4Associations between significant covariates (from Table 2) and the low trajectory of number knowledge (n=153)
Variables β p OR -2LL Δ -2LL p AIC R2
Block 1: Socio-demographic (5 months) -29681.14 71.67 0.00 59406.28 0.09Household income -0.20 0.00 0.85Mother diploma -0.08 0.16 0.87Father diploma -0.22 0.00 0.68Block 2: Parenting expectancies (5-29 months) -29675.01 84.33 0.00 59396.02 0.11Household income -0.19 0.00 0.85Father diploma -0.23 0.00 0.66Outcome expectancies (mother)† -0.11 0.00 0.87Outcome expectancies (father)† -0.08 0.08 0.90Block 3: Child cognitive abilities (41 months) -29621.23 180.81 0.00 59294.47 0.24Household income -0.13 0.01 0.89Father Diploma -0.17 0.00 0.72Outcome expectancies (mother)† -0.07 0.11 0.91Visual-spatial skills† -0.28 0.00 0.86Receptive vocabulary† -0.12 0.06 0.98Memory-span† -0.20 0.01 0.83Early cognitive development† -0.23 0.00 0.58Final model -29624.90 173.61 0.00 59297.79 0.26Household income -0.15 0.00 0.87Father Diploma -0.18 0.00 0.70Visual-spatial skills† -0.29 0.00 0.86Memory-span† -0.23 0.00 0.80Early cognitive development† -0.26 0.00 0.55
Note. Significant at p < .05; β: Standardized parameter estimates, OR: odds ratio, and †: continuous variables The -2 Log Likelihood ratio (∆-2LL), the Akaike’s information criterion (AIC), and the variance explained (R2) indicates the adequacy of the model. Data are courtesy of the Quebec Institute of Statistics
50
Recommended