Psychonomic Bulletin & Review (2022) 29:699–720

[Link]

THEORETICAL/REVIEW

Examining the relations between spatial skills and mathematical

performance: A meta-analysis
Kinnari Atit 1 & Jason Richard Power 2 & Terri Pigott 3 & Jihyun Lee 4 & Elyssa A. Geer 5,6 & David H. Uttal 7 &
Colleen M. Ganley 5 & Sheryl A. Sorby 8

Accepted: 10 September 2021 / Published online: 19 November 2021

# The Psychonomic Society, Inc. 2021

Abstract
Much recent research has focused on the relation between spatial skills and mathematical skills, which has resulted in widely
reported links between these two skill sets. However, the magnitude of this relation is unclear. Furthermore, it is of interest
whether this relation differs in size based on key demographic variables, such as gender and grade-level, and the extent to which
this relation can be accounted for by shared domain-general reasoning skills across the two domains. Here we present the results
of two meta-analytic studies synthesizing the findings from 45 articles to identify the magnitude of the relation, as well as
potential moderators and mediators. The first meta-analysis employed correlated and hierarchical effects meta-regression models
to examine the magnitude of the relation between spatial and mathematical skills, and to understand the effect of gender and
grade-level on the association. The second meta-analysis employed meta-analytic structural equation modeling to determine how
domain-general reasoning skills, specifically fluid reasoning and verbal skills, influence the relationship. Results revealed a
positive moderate association between spatial and mathematical skills (r = .36, robust standard error = 0.035, τ2 = 0.039).
However, no significant effect of gender or grade-level on the association was found. Additionally, we found that fluid reasoning
and verbal skills mediated the relationship between spatial skills and mathematical skills, but a unique relation between the spatial
and mathematical skills remained. Implications of these findings include advancing our understanding for how to leverage and
bolster students’ spatial skills as a mechanism for improving mathematical outcomes.

Keywords Meta-analysis . Spatial skills . Mathematical skills . Fluid reasoning . Verbal skills

Introduction
* Kinnari Atit
[Link]@[Link] In the past three decades, much research has focused on the
relation between spatial skills and mathematical skills (e.g.,
1
School of Education, University of California, Riverside, 1207 Casey et al., 1995; Verdine et al., 2017). This effort has re-
Sproul Hall, Riverside, CA 92521, USA sulted in widely reported links between these two sets of
2
School of Education, University of Limerick, Limerick, Ireland skills, as well as a considerable body of evidence highlighting
3
College of Education and Human Development and School of Public the benefits of well-developed spatial skills in a wide variety
Health, Georgia State University, Atlanta, GA, USA of science, technology, engineering, and mathematics
4
College of Education, The University of Texas at Austin, Austin, TX, (STEM) fields (Atit et al., 2015; Newcombe, 2010; Uttal
USA et al., 2013). But to understand and perhaps leverage these
5
Department of Psychology, Florida State University, relations within educational contexts, the magnitude of the
Tallahassee, FL, USA relation as well as the moderators and mediators of the relation
6
Human Development and Family Sciences, Purdue University, West need to be better understood. The meta-analyses presented
Lafayette, IN, USA here aim to estimate the average magnitude of the reported
7
School of Education and Social Policy, Northwestern University, relationship between spatial skills and mathematical achieve-
Evanston, IL, USA ment, to identify potential moderators and mediators of this
8
College of Engineering and Applied Science, University of relationship, and to inform the design and execution of future
Cincinnati, Cincinnati, OH, USA research in this area.
700 Psychon Bull Rev (2022) 29:699–720

The current meta-analysis is particularly pressing in light of In addition to the performance-based evidence of the rela-
widespread underachievement of US students in STEM sub- tion between spatial skills and mathematical achievement, pri-
jects (Aud et al., 2012), specifically in mathematics or research suggests that the positive relation between spatial
(Organization for Economic Development, 2016). A report and mathematical reasoning found in many studies may be
from the Program of International Student Assessment based on shared cognitive or neural processes (Gunderson
(PISA) suggests that American students’ weaknesses in math- et al., 2012; Mix & Cheng, 2012; see Hawes & Ansari,
ematics are particularly evident when “[r]easoning in a geo- 2020, for a review). Some researchers assert that number in-
metric context – requiring authentic reasoning in a planar or formation is mentally represented in spatial formats (Mix,
spatial geometric context”(Organization for Economic 2019; Mix & Cheng, 2012). For example, work by
Development, 2012, p.3). This indicates that problems engag- Gunderson et al. (2012) indicates that quantitative magnitudes
ing one’s spatial skills may be particularly difficult for are represented in the mind spatially as a mental number line.
American students. As a concerted research and educational Further evidence comes from brain-imaging studies showing
effort rises to address this need to improve students’ mathe- similar areas of brain activation when individuals process both
matical outcomes, it is worth considering the existing wealth spatial and numerical information (Hubbard et al., 2005).
of research examining the relation between these two skill Lastly, successful interventions often focus on helping stu-
sets. A brief overview of relevant research, grouped by com- dents translate mathematical symbols or problem statements
mon themes, is presented to situate the current meta-analysis into spatial representations, including number lines, diagrams,
within the broader academic and societal context. concrete models, or hand gestures (Novack et al., 2014;
Valenzeno et al., 2003).
In line with the success of spatial interventions in
Spatial skills and mathematics supporting students’ mathematical reasoning, some re-
searchers believe that the relation between spatial skills and
Spatial skills enable us to mentally manipulate, organize, rea- mathematical skills is perhaps due to differences in problem-
son about, and make sense of spatial relationships in real and solving strategies (Delgado & Prieto, 2004). More specifical-
imagined spaces (e.g., Newcombe & Shipley, 2015; Uttal ly, studies show that many mathematical problems can be
et al., 2013). These skills are commonly employed when com- solved by using spatial visualization and/or analytic strategies,
pleting everyday tasks such as assembling furniture or navi- and there may be two different styles: one is based on algo-
gating from one location to another. Understanding spatial rithm memorization and automatic application, the other on
skills has been a topic of interest in psychology for much of the visuospatial representation of the problems. This suggests
the last 100 years (e.g., Bethell-Fox & Shepard, 1988; Carroll, that perhaps those who have better spatial skills would employ
1993; Linn & Petersen, 1985; Shepard & Metzler, 1971). the second style of problem solving (i.e., visuospatial strate-
Historically, the interest in spatial skills has roots in the study gies), which is often considered to be more efficient (Delgado
of mechanical aptitude (Cox, 1928; Paterson et al., 1930) and & Prieto, 2004).
in defining the factors of intelligence (Carroll, 1993;
Thurstone & Thurstone, 1941). Examples of spatial skills Domain-general cognitive processes and
commonly examined in studies of mathematical understand- mathematical achievement
ing include mental rotation (e.g., Geer et al., 2019) and spatial
visualization (e.g., Burte et al., 2017). Parallel to the research on spatial skills and mathematical
Many studies examining the relations between spatial skills achievement, researchers interested in mathematics learning
and mathematical achievement indicate that the two are sig- and outcomes have also focused their efforts on examining
nificantly correlated for students at all educational levels (e.g., the role of domain-general cognitive processes in students’
Casey et al., 1995; Delgado & Prieto, 2004; Verdine et al., mathematical achievement and understanding (e.g., Hawes
2017). For example, mental rotation performance relates to et al., 2019; Raghubar et al., 2010; Taub et al., 2008).
mathematical reasoning skills in elementary students (Geer Domain-general cognitive processes that have been found to
et al., 2019) and secondary students (Delgado & Prieto, have a strong positive relationship with mathematical achieve-
2004), as well as mathematical aptitude in undergraduates ment include executive functions, such as working memory
(Casey et al., 1995). Furthermore, some studies report that (Raghubar et al., 2010) and fluid reasoning (Green et al.,
spatial skills are predictive of students’ future mathematics 2017; Taub et al., 2008). Working memory refers to a mental
learning even after accounting for indicators of general rea- workspace that is involved in controlling, regulating, and ac-
soning skills, such as verbal skills and executive functioning tively maintaining relevant information to accomplish com-
skills (Verdine et al., 2017; Zhang et al., 2014). Yet, the details plex cognitive tasks (Miyake & Shah, 1999). Studies have
of the relations between spatial skills, mathematical skills, and found that working memory is fundamental for mathematics
factors of general intelligence are not well understood. learning and performance in school-aged children (Berg,
Psychon Bull Rev (2022) 29:699–720 701

2008; McKenzie et al., 2003) and adults (Imbo et al., 2007; mathematical skills requires accounting for their shared
Seitz & Schumann-Hengsteler, 2002). For instance, Berg relations with fluid reasoning.
(2008) examined the contribution of working memory in third Furthermore, though many studies simultaneously exam-
to sixth grade students to performance on a mathematical test ine the contribution of executive functions and spatial skills to
of a range of skills, such as single- and multi-digit arithmetic, mathematical achievement in their analyses (e.g., Green et al.,
fractions, and algebra. Working memory contributed unique 2017; Hawes et al., 2019; Verdine et al., 2017), the findings
variance to mathematical performance independent of chrono- regarding the relations between the three cognitive skill sets
logical age, short-term memory, reading, and processing have not been consistent. For instance, contrary to the findings
speed. In children with and without significant mathematics of Green et al. (2017), Hawes et al. (2019) found that spatial
difficulties, Swanson and Beebe-Frankenberger(2004) found skills were the strongest predictor of mathematical achieve-
that working memory predicted solution accuracy on word ment in children ages 4–11 years after controlling for age,
problems independent of several academic and cognitive while executive functioning skills (which included measures
variables, such as fluid reasoning, reading skills, and of working memory) was not a significant unique predictor.
phonological processing, among others. Additionally, Imbo Moreover, the patterns of the relations between these three
et al. (2007) found that working memory plays a significant factors remained stable across age and grade (Hawes et al.,
role during multi-digit arithmetic problem solving in adults. 2019). Thus, taken together, the differences in findings be-
Fluid reasoning is the ability to solve novel problems flex- tween these studies indicate that the nuances of the relations
ibly and deliberately without using previous information. between spatial skills, executive functioning skills, and
More specifically, it is the ability to analyze novel problems, mathematical skills are not well understood.
identify patterns and relationships that underpin these prob- In a recent theory of intelligence called process overlap
lems, and apply logic (Schneider & McGrew, 2012). In a theory, Kovacs and Conway (2016) postulate that any task
synthesis of studies investigating the concurrent relationships requires a number of domain-specific and domain-general
between cognitive abilities and achievement measures, fluid cognitive processes. The domain-specific cognitive tests mea-
reasoning was one of the three cognitive abilities that was suring different cognitive skill sets (e.g., tests of spatial skills,
consistently related to mathematical performance in calcula- tests of mathematical skills) all tap into a common pool of
tion and problem solving at all age ranges throughout devel- domain-general executive functions, such as working memory
opment (the other two were verbal comprehension and pro- and fluid reasoning (Kovacs & Conway, 2016). In particular
cessing speed). Additionally, fluid reasoning has been found with regard to the relations between spatial skills and mathe-
to predict future mathematical achievement (Green et al., matical achievement, the role of domain-general executive
2017; McGrew & Wendling, 2010). Using a longitudinal co- functions has yet to be ascertained. More specifically, it re-
hort sequential design, Green et al. (2017) examined how fluid mains unclear whether executive functions, such as working
reasoning measured at three assessment occasions, spaced 1.5 memory and fluid reasoning, mediate the relations between
years apart, predicted math outcomes for a group of 69 partic- spatial skills and mathematical performance.
ipants between ages 6 and 21 years across all three assessment In addition to executive functions, another domain-general
occasions. Results revealed that fluid reasoning was the only cognitive factor positively related to mathematical achieve-
significant predictor of future mathematical achievement, ment is general intelligence, or g (Taub et al., 2008;
while age, spatial skills, and vocabulary were not significant Wrigley, 1958). General intelligence includes the ability to
predictors. think logically and systematically (Embretson, 1995) and is
Green et al.’s (2017) finding that fluid reasoning, but the best individual predictor of achievement across academic
not spatial skills, predicts future mathematical achieve- domains, including mathematics (e.g., Deary et al., 2007;
ment is not surprising given that spatial skills and fluid Jensen, 1998; Stevenson et al., 1976; Taub et al., 2008;
reasoning skills are reported to be very strongly related Walberg, 1984). For instance, in a 5-year prospective study
(e.g., Fry & Hale, 1996; Green et al., 2017). Moreover, of more than 70,000 students, Deary et al. (2007) found that
many tests of fluid reasoning skills require processing general intelligence assessed at 11 years of age explained
spatial information, such as Matrix Reasoning and nearly 60% of the variation on national mathematics tests at
Raven’s Progressive Matrices. Despite the strong corre- 16 years of age. A study by Geary (2011) aimed at identifying
lation between fluid reasoning and spatial skills, there is cognitive factors and quantitative competencies in first-grade
evidence that suggests that the two skills rely on sepa- students that predict mathematical achievement in fifth-grade
rable cognitive processes and brain regions (e.g., students found that general intelligence, along with processing
Halford et al., 1998; Holyoak, 2012; Klingberg, 2006; speed and working memory, predicted fifth-grade mathemat-
Krawczyk, 2012; Vendetti & Bunge, 2014). Taken to- ical achievement as well as growth in achievement. While
gether, these findings indicate that gaining an accurate these studies solidify the contribution of general intelligence
understanding of the relation between spatial skills and on mathematical achievement and performance, what is
702 Psychon Bull Rev (2022) 29:699–720

unclear is the magnitude of its contribution to the relation of understanding in the field about if and how the relations
between spatial skills and mathematical achievement. between spatial skills and mathematical skills vary between
In sum, existing research suggests that spatial skills and ages or grades. Identifying if and how the relation between
mathematical skills may not be directly related. Theories of spatial skills and mathematical skills varies across develop-
intelligence indicate that other domain-general cognitive fac- ment is critical because it could shed light on when
tors, such as working memory, fluid reasoning, and general implementing spatial interventions may be most effective in
intelligence, may underlie their relation. Thus, a more accurate improving students’ mathematical outcomes.
understanding of how spatial skills and mathematical skills are Just as the relations between spatial skills and mathematics
related requires examining if and how associated skillsets in- outcomes could vary by age or grade level, studies have sug-
fluence the relation. gested that gender could also be a moderating factor. Women
are under-represented at the highest levels of STEM occupa-
The influence of participants’ age or grade-level and tions (Ceci & Williams, 2011; Halpern et al., 2007), especially
gender on the relations between spatial and mathe- in math-intensive STEM domains (e.g., computer and
matical skills information sciences, engineering, and physical and
technical sciences; see Wang & Degol, 2017, for a review).
Just as the influence of domain-general cognitive processes on Moreover, boys outperform girls on tests of mathematical ap-
the relation between spatial and mathematical skills is unclear, titude, such as on the SAT mathematics subtest (SAT-M;
whether the relation between the skill sets varies across devel- Halpern et al., 2007), especially on the most complex prob-
opment is also unknown. Much prior research suggests that lems (Hyde et al., 1990). Underlying the gender differences in
the relation between spatial skills and mathematical skills mathematical outcomes are reported differences in men and
could vary depending on participants’ age or grade level women’s spatial skills (e.g., Casey et al., 1995; Nuttall et al.,
(Battista, 1990; Stannard et al., 2001; Wolfgang et al., 2005), as well as differences in the relations between spatial
2001). For instance, in a longitudinal study by Li and Geary skills and mathematical skills (e.g., Tartre & Fennema, 1995).
(2013), first- to fifth-grade gains in visuospatial memory pre- For instance, Casey et al. (1995) found that gender differences
dicted the end of fifth-grade mathematical achievement. on the mathematics section of the SAT-M were related to
However, visuospatial memory was not related to mathemat- gender differences on a spatial skills measure in high-
ics in first grade (Li & Geary, 2013). Similarly, in a study by achieving students. Boys/men performed better than girls/
Hanline et al. (2010), block construction scores, an indicator women on both mental rotation and the SAT-M (Casey
of students’ understanding of spatial relations and geometric et al., 1995).
knowledge, predicted preschoolers’ scores on the Test of Beyond these gender differences in levels of mathematical
Early Reading Ability at age 8 years and the growth rate on or spatial skills, there is also some evidence for gender differ-
the same test from age 5–8 years. However, there was no ences in the strength of the relations between spatial skills and
significant relation between block construction scores and mathematical skills (e.g., Tartre & Fennema, 1995). Tartre and
scores on the Test of Early Math Ability at age 8 or the growth Fennema (1995) examined the relationship between spatial
rate for the Test of Early Math Ability from ages 5–8. skills, verbal skills, and mathematical achievement in 60 stu-
Similarly, in a cross-sectional study, Mix et al. (2016) exam- dents as they progressed from sixth to 12th grades. Unlike
ined the relations between spatial and mathematical skills in Casey et al. (1995), they found no significant difference be-
854 students from kindergarten, and third and sixth grades and tween boys’ and girls’ spatial skills, verbal skills, or their
found that mental rotation skills were strongly related to math- mathematical achievement. However, they found that the re-
ematics performance in kindergarten and third-grade students, lations between the three factors varied by gender. Spatial
but was not a significant predictor of mathematics perfor- skills were a consistent significant predictor of mathematical
mance in sixth-grade students. Instead, working memory and achievement for girls across the years, but not for boys, where-
visual motor integration showed the strongest relations with as verbal skills were a significant predictor of mathematics
mathematical performance in sixth-grade students. across the years for boys, but not for girls (Tartre &
While some studies show that the relations between spatial Fennema, 1995). On the contrary, Ganley and Vasilyeva
and mathematical skills vary across development, there are (2011) examined the relations between spatial skills and math-
some studies that show that the relations remain stable ematical performance in middle school students and found
across ages. For example, as previously discussed, Hawes that despite similar levels of mathematical performance for
et al. (2019) found that the strong relations between spatial boys and girls, spatial skills predicted mathematical perfor-
skills and mathematical skills were evident in children from mance in boys, but not in girls. As the research findings on
ages 4 to 11 years. Specifically, the relation between spatial the role of gender in the association between spatial skills and
skills and mathematical skills remained consistent across the mathematical skills is not consistent, a systematic
ages. Taken together, these mixed findings underline the lack
Psychon Bull Rev (2022) 29:699–720 703

investigation synthesizing prior findings is necessary to eluci- between spatial skills and mathematical skills? In the second
date how gender impacts the relationship between these skill meta-analysis, we used the meta-analytic structural equation
sets. modeling (MASEM) approach to examine how accounting
for domain-general reasoning skills (working memory, fluid
Prior meta-analytic findings on the relations between reasoning, and verbal skills) impacts the relation between spa-
spatial skills and mathematical skills tial skills and mathematical skills (depicted in Fig. 1). Based
on Carrol’s (1993) model of cognitive abilities, Wai et al.
With the aim of better understanding the relations between (2009) posit that cognitive abilities center around three cogni-
spatial skills and mathematical reasoning skills, a meta- tive domains: quantitative/numerical, spatial/pictorial, and
analysis was recently conducted by Xie et al. (2019) summa- verbal/linguistic (or mathematical, spatial, and verbal do-
rizing the findings of 73 studies reporting correlations be- mains, respectively). The shared variance across these three
tween these two factors. The aims of their study included: content domains can be attributed to the higher order construct
(1) determining whether there is a significant association be- of general intelligence (g) (Wai et al., 2009). Thus, we includ-
tween spatial skills and mathematical skills, and (2) determin- ed measures of verbal skills in our analyses to account for the
ing whether the domains of spatial skills, mathematical skills, variance contributed by g in the relationship between spatial
age, and developmental disability status moderated this rela- skills and mathematical skills.
tionship. In this meta-analysis, Xie et al. (2019) report finding
a medium positive association between spatial skills and
mathematical skills (r = .27; 95% confidence interval (CI) Method
[0.24, 0.32]), which did not differ by age, developmental dis-
ability status, or type of spatial skill. However, they did find Literature search/information sources
that mathematical domain did moderate the relations between
mathematical and spatial skills. Specifically, logical reasoning We began with electronic searches of PsycINFO, ProQuest,
showed the strongest association with spatial skills in compar- and ERIC databases. We searched all available records pub-
ison with numerical skills or arithmetic skills (Xie et al., lished since 1997 to the date of the search: 1 May 2018. We
2019). Important to the research reported here, Xie et al. limited our search to this 20-year window for three reasons.
(2019) did not examine if domain-general reasoning skills, First, the time frame was large enough to provide a wide range
such as working memory or fluid reasoning skills, mediate of studies and would encompass the large increase in studies
the relation between spatial and mathematical skills, or wheth- on the topic of mathematics and spatial skills that has recently
er the magnitude of the relation between the two factors differ occurred. Second, the time frame is narrow enough to allow us
by gender. Thus, the role of domain-general reasoning skills to gather most of the relevant published and unpublished data
and participants’ gender on the relations between spatial skills for analysis. Third, educational structures (i.e., schooling and
and mathematical outcomes is still unknown. curriculum) have remained mostly unchanged during this time
frame (Payne, 2008).
The current study We used the following search terms: (Intervention OR
training OR practice OR class OR enhancement OR
As discussed above, the nuances of the relationship between education OR quasi-experimental OR experimental) AND
spatial skills and mathematical skills remains unclear. (“spatial skills” OR “spatial ability” OR “spatial cognition”
Specifically, it is unknown if the relation differs by gender OR “spatial thinking” OR visuospatial) AND (mathematics
and how much domain-general reasoning skills may account OR math OR “math skills” OR “math ability” OR “mathemat-
for the association. Furthermore, although Xie et al. (2019) ical reasoning”). Where possible, database search filters were
found that age was not a significant moderator in their meta- employed prior to screening. For example, when searching the
analysis, we decided to include the factor in our analyses to ERIC database, the following filters were used: Peer-
see if we could replicate their finding. Reviewed Journal Publications, and Within 21 Years of the
In this study, we examined two sets of questions that each Search Date. To obtain unpublished data, we conducted a
required different analytical tools. As such, we chose to con- search in ProQuest Dissertations and Abstracts. When
duct two separate meta-analyses that would each address one searching ProQuest Dissertations and Abstracts, we limited
set of questions. In the first meta-analysis, we employed cor- our search term to within the abstracts due to the expansive
related and hierarchical effects (CHE) meta-regression models nature of dissertations, which significantly reduces the effec-
(Pustejovsky & Tipton, 2021) to answer the following re- tiveness of search terms. In addition to an electronic search of
search questions: (1) What is the magnitude of the relation databases, we acquired unpublished data through requests in
between spatial skills and mathematical skills? and (2) What social media posts and emails to various relevant research
is the effect of gender and grade-level on the association groups (e.g., Cognitive Development Society, The Spatial
704 Psychon Bull Rev (2022) 29:699–720

Fluid
Reasoning

Working
Memory

Verbal
Skills

Mathematical
Spatial Skills Skills

Fig. 1 Hypothesized path model between cognitive constructs. Note. An using MASEM analysis. Fluid reasoning, working memory, and verbal
illustration of the hypothesized theoretical relations between spatial skills skills are hypothesized to be moderators between spatial skills and
and mathematical skills, and general reasoning skills to be examined mathematical skills in this model

Intelligence and Learning Center network listserv). satisfactory levels of inter-rater reliability were observed (k =
Furthermore, prominent researchers in the field were 0.77, p<.01) (Altman, 1990). After inter-rater reliability was
contacted directly via email for any current or previous un- calculated for the abstract coding, any disagreements in study
published work. selection were discussed and agreed upon prior to final desig-
Searching the databases using the Boolean phrases and nation. The remaining 75% of the studies were randomly
filters mentioned above resulted in the acquisition of 858 ar- assigned to each reviewer for the completion of the abstract-
ticles. All resulting articles were compiled within an Endnote coding process. 592 studies were removed from further anal-
database and duplicates were removed with 760 unique arti- ysis during the abstract-evaluation phase. After the study ab-
cles remaining. An additional nine articles were acquired stracts were evaluated, manuscripts coded as included were
through social media and email requests, resulting in a total retained for full manuscript review.
of 769 articles.

Full manuscript review and data extraction

Abstract screening
One hundred and seventy-seven full manuscripts were
All studies found through our initial search procedure were retained for full manuscript review and data extraction, which
compiled into a numbered study catalogue. As noted by was again conducted by the first and second authors. Each
Berman and Parker (2002), “Failure to blind review could lead manuscript was reviewed and the following information was
to biases similar to those in a record review when subjects are extracted: (1) measure(s) of spatial skills, mathematical skills,
selected by investigators who are not blinded to the outcomes working memory, fluid reasoning, and verbal skills adminis-
of interest” (p.4). In this catalogue, each study was assigned a tered; (2) bivariate correlations between measures of spatial
unique identifier and only the study title and abstract were skills and measures of mathematical skills; (3) bivariate cor-
displayed for evaluation in line with best practice (Berman relations between measures of spatial skills and measures of
& Parker, 2002). The abstracts for all acquired studies were working memory, fluid reasoning, and verbal skills; (4) bivar-
then evaluated by the first and second authors to establish iate correlations between measures of mathematical skills and
whether they included at least one measure of spatial skills measures of working memory, fluid reasoning, and verbal
and at least one measure of mathematical skills. Studies that skills; (5) the reported sample size for each correlation; (6)
did not mention a mathematical measure or a spatial measure the age or grade-level of the participants; and (6) the propor-
in the study abstract were excluded from further analysis. tion of males relative to female participants. If during full
Additionally, studies where manuscripts were not written in manuscript review, a study was found to be missing informa-
English, or an English translation was not able to be acquired, tion about the spatial skills measure(s) administered, the math-
were also excluded from further analysis. ematical skills measure(s) administered, or the bivariate cor-
To ensure reliability of the abstract coding and selection relation(s) between them, the manuscript authors were
process, 25% of the abstracts were randomly selected and contacted for the information. If no response was provided,
screened by both reviewers for inclusion of the appropriate the study was excluded from further data analysis. Measures
measures. Inter-rater reliability using Cohen’s Kappa was then of working memory, fluid reasoning, or verbal skills were not
calculated for this subset of studies, which indicated that required to retain the manuscript for further data analysis. A
Psychon Bull Rev (2022) 29:699–720 705

Table 1 Articles included in the meta-analyses

% Male Sample Spatial measure description Math measure description

sizea

Ackerman and 58.45% 142 Paper folding Math approximation

Wolman (2007)
Spatial orientation Math knowledge
Spatial analogies Number series
Verbal test of spatial ability Arithmetic
Bonny and 52.50% 40 Area discrimination task Multiple choice geometry task
Lourenco
(2015)
Carr et al. (2008) nr 227 Vandenberg test of mental rotation Criterion referenced competency test
Casey et al. (2011) 51.61% 124 Spatial visualization task and a mental rotation test Formula Based Measurement
Spatial Conceptual Measurement
Numerical Skills (Items from NAEP and TIMSS
4th grade)
Casey et al. (2015) 0.00% 127 Block Design of the WISC-IV Addition task
2D Mental Transformation Task Subtraction task
3D Mental Rotation Task
Casey et al. (1997) 45.74% 94 Mental rotation test Geometry test
SAT math
Cirino et al. 53.00% 150-162 Vandenberg and Kuse Spatial Test Estimation task
(2016)
Procedural computation
Fraction competency
Proportional reasoning
State test
Fact-based retrieval
Clifford (2008) 49.00% 112 WISC Block Design WJ III
Picture Completion
WISC Visual Spatial
Form Patterns
Position & Direction
Visual Spatial Factor
Cui et al. (2017) 40.36 223 Mental rotation test Exact computation
Approximate Computation
Math achievement test
Edens and Potter 50.00% 214 Case's Framework General math test
(2007)
Frick (2019) 57.14 119 Perspective taking task T1 Basic calculation
Mental rotation T2 number line task
Diagrammatic representations task T2 Proportional reasoning
Cross-sectioning task T3 mathematics (HRT composite score)
Spatial scaling task T3 Geometry
Mental transformation task
T2 mental rotation
T3 mental rotation (figure rotation)
Geer and Ganley 65.67 67 Spatial Orientation Task Math fluency
(2021)
Money RoadMap Task
Gilles and 57.40% 122 Spatial sub-test from the Ability Factor Battery Numerical subtest from the Ability Factor Battery
Bailleux (2001)
706 Psychon Bull Rev (2022) 29:699–720

Table 1 (continued)

% Male Sample Spatial measure description Math measure description

sizea

MAST: Mathematical achievement score on

scholastic test
General math test
Gilligan et al. nr 12,099 Pattern Construction subscale of the BAS-II Progress in Maths (NFER PiM)
(2017)
Green et al. (2017) 48 69 Spatial span & block design Number series
Applied problems
Math fluency
Haciomeroglu 54.60% 150 Cube comparison Calculus
(2015)
Card Rotations
Form Board
Paper Folding
Haciomeroglu 60.29% 348 KIT Reference Tests for Cognitive Factors AP calculus test
(2016)
Hegarty 100.00% 33 Space Subtest from PMA Space General math test
and Kozhevni-
kov (1999)
Block Design Subtest from WISC-R(Wechsler, 1976)
Johnson (2017) 53.00% 304 Thurstone PMA CRCT math competency test
Kirby and Boulter 54.50% 44 Hidden Patterns, Card Rotations, Surface Development Geometry test
(1999)
Kyttälä et al. 43.50% 46 Mental rotation task Early Numeracy Test
(2003)
Kytällä and Lehto 54.69% 128 VSWM task (mental rotation) General math test
(2008)
Lowrie et al. nr 66 Spatial Reasoning Instrument SRI MathT test, which we developed using released
(2017) items from Australia’s National Assessment
Program (NAPLAN)
Mix et al. (2016) 47.81% 274 Mental rotations, Visual Spatial Working Memory, General math test
Visual Motor Integration, Block Design and Map
Reading Perspective Taking
Möhring et al. 50.00% 50 Spatial localization task Proportional reasoning task
(2015)
Morosanova et al. 49.69% 318 Mental Rotation Task Problem Verification Task
(2016)
Understanding Numbers test
Olver (2013) 47.60% 21 Mental Transformation Test Number Knowledge test
Number line estimation
Oostermeijer et al. 50.00% 128 Picture rotation task Math word problem solving
(2014)
Paunonen and 30.11% 176 Spatial subscale of the Multidimensional Aptitude Arithmetic subscale of the Multidimensional
Hong (2010) Battery Aptitude Battery
Peng et al. (2017) 58.3 94 Blocks reconstruction subtest Approximate addition task
exact addition task
rapid digit naming
number identification
numerical reasoning
Pyke (2003) nr 174 The Gestalt completion test Algebra and geometry test
Reuhkala (2001) 58.06% 62 A mental rotation task based off of Shepard & Metzler total score of a national math test
Rodán et al. 51.11% 45 Spatial Ability Subtest, EFAI-3 Numerical Ability Subtest, EFAI-3
(2016)
Simms et al. 51.95% 77 Arrows subtest from NEPSY WIAT-II Mathematical Reasoning and Numerical
(2016) Operations subscales
Psychon Bull Rev (2022) 29:699–720 707

Table 1 (continued)

% Male Sample Spatial measure description Math measure description

sizea

Mr. X mental rotation

Skagerlund and 39.80% 133 Mental rotation; paper folding Multidigit calculation; arithmetic equations;
Träff (2016) arithmetic fact retrieval;
Tolar et al. (2009) 49.74% 195 Differential Aptitude Tests (DAT) space relations subtest AAIMS algebra content test
van Garderen 40.90% 22 Block Design subtest from the WISC-III WJ-III ACH Applied Problems sub-test
(2006)
MGMP-SVT MPI - mathematical word-problem solving per-
formance
Block Design WISC-III Mathematical Processing Instrument (MPI)
Spatial Visualization Test MGMP-SVT
Ventura et al. 39.94% 323 Spatial orientation test SAT math test
(2013)
Mental rotation test EMAS age 3 y
Verdine et al. 53.92% 81-102 CMTT age 5 y EMAS age 4 y
(2017)
TOSA age 3 y High Count age 3 y
TOSA age 4 y HIgh Count age 4 y
WPPSI Block design subtest age 4 y WIAT Math Problem solving test age 4 y
WJ III space relations subtest age 4 y Number Sense Brief
WJ III space relations subtest age 5 y WIAT age 5 y
Verdine et al. 53.92% 102 Test of spatial assembly (TOSA) ( Dimensional) Early Mathematics Assessment System: Number
(2014) and Operations subtest
Test of spatial assembly (TOSA) (Match)
Voyer (1998) 43.09% 123 The Vandenberg and Kuse Mental Rotations Test General math test
The Water Level Task General math test
Wei et al. (2012) 50.00% 80 3D rotation Advanced Math Concepts
Wong (2017) 0-100% 83 - 99 Levine's Mental transformation Task Diagnostic Tests for Metacognitions and
Mathematics
Zhang and Lin 38.67 70-75 Spatial perception Arithmetic - non symbolic
(2018)
38.67 70 Spatial visualisation Arithmetic - word problems
Arithmetic – written
Zhang and Lin 46.23 71-106 Spatial Perception subtest from the Test of Visual- Arithmetic word problems
(2017) Perceptual Skills Revised
Visual-spatial subtest from the Test of Visual- Perceptual
Skills Revised
Spatial Visualization subtest of Spatial Relations
Mental Rotation

Note. If a commonly used acronym or name was used in reference to a measure this was included as used in the original paper. Where possible non-
widely used tests were described using the language of the original paper. For the sake of conciseness full measurement titles and sources are not listed
here but can be found within the relevant papers. More complete notations regarding measure composition and additional extracted variables are included
in the overall dataset available upon request. Where relevant information was not provided it has been recorded as nr: not reported
708 Psychon Bull Rev (2022) 29:699–720

list of the 45 included manuscripts as well as the study char- grade-level than age, so grade-level was used as a moderator
acteristics and measure information extracted from each one of effect size in the analysis.
(except for the bivariate correlations) is provided in Table 1. A
summary of the steps carried out in the abstract screening and Percent of male participants To examine whether gender
full manuscript review components of this study, based on the moderates the relationship between spatial skills and mathe-
PRISMA (Preferred Reporting Items for Systematic Reviews matical skills, the percent of male participants was extracted
and Meta-Analyses) flowchart template (Moher et al., 2015), from each study. Few studies reported the correlation between
are outlined in Fig. 2. spatial skills and mathematical skills by gender. Thus, we use
percent of male participants as an indirect measure of whether
the magnitude of the correlation between spatial and mathe-
Study characteristics and measures of interest matical skills differs as a function of the percent of male
participants.
Participant grade Both the age of the participants and their
grade-level reported within each study was extracted. If stud- Spatial skills Tasks that were considered as measures of spatial
ies collected data on participants at multiple time points, age skills assessed participants’ skills in visualizing and/or men-
information at each administration of the relevant measures tally manipulating objects or figures or navigating spaces
was extracted. As many studies examining mathematical un- (Atit, Uttal, & Stieff, 2020b; Newcombe & Shipley, 2015).
derstanding occur in a classroom setting (e.g., Verdine et al., Common examples of measures of spatial skills include the
2017), if participants’ grade-level was reported without age- Mental Rotations Test (Peters et al., 1995; Vandenberg &
related data, then age was approximated based on typical age Kuse, 1978), the Wechsler Preschool and Primary Scale of
per grade defined by the National Center for Education Intelligence (WPPSI-IV) Block Design subtest (Wechsler,
Statistics (2018). More studies reported on participants’ 2012), and the Paper Folding test (Ekstrom et al., 1976).

Fig. 2 The process of article selection for inclusion in analysis. Note. After carrying out this selection process, data from 45 articles were included for
analysis
Psychon Bull Rev (2022) 29:699–720 709

Mathematical skills Tasks that were considered measures of skills and mathematical skills), we converted all correlation
mathematical skills assessed participants’ skills in reasoning coefficients to the Fisher’s z scale (Borenstein et al., 2005).
about “numbers and their operations, interrelations, combina- After the analysis, all the results in Fisher’s z scale were then
tions, generalizations, and abstractions and of space configu- converted back to correlation coefficients for the ease of inter-
rations and their structure, measurement, transformations, and pretation. For our second meta-analysis, a MASEM approach
generalizations” (Merriam Webster, n.d.). Examples of mea- was used to examine the contribution of domain-general rea-
sures assessing mathematical skills include assessments of soning skills to the relations between spatial skills and math-
arithmetic (e.g., Ackerman & Wolman, 2007), geometry tests ematical skills, and the original bivariate correlation coeffi-
(e.g., Bonny & Lourenco, 2015), and the mathematics portion cients were utilized.
of the Criterion Referenced Competency Test (e.g., Carr et al.,
2008). Analysis

Working memory capacity Tasks that were considered mea- Study 1: What is the magnitude of the relation between spa-
sures of working memory capacity assessed participants’ abil- tial skills and mathematical skills? To answer the above re-
ity to both retain and process information simultaneously search question, we first fit an intercept-only random-effects
(Conway et al., 2003). Examples of working memory mea- model and estimated the pooled correlation coefficient be-
sures include Backwards Digit Span (GrÉGoire & Van Der tween spatial and mathematical skills. Next, we examined
Linden, 1997), Automated Symmetry Span (Oswald et al., whether the grade and gender (i.e., proportion of males in
2015), and the Corsi Block Task (Corsi, 1972; Miyake et al., the primary study sample) explained further variability in ef-
2001). fect sizes across different studies using a meta-regression
model. Because the unit of analysis is the study, the grade
Fluid reasoning Fluid reasoning tasks measured participants’ and gender are the study and sample characteristics, not an
ability to flexibly and deliberately solve novel problems with- individual participant characteristic.
out using previous information. More specifically, they mea- Given that there is more than one kind of spatial skill and,
sured participants’ ability to analyze novel problems, identify thus, more than one kind of spatial test (Linn & Petersen,
patterns and relationships that underpin these problems, and 1985; Newcombe & Shipley, 2015), and given that there is
apply logic (Schneider & McGrew, 2012). Common measures more than one kind of mathematical domain, many of the
of fluid reasoning include the Matrix Reasoning from the studies included in our meta-analyses administered more than
Perceptual Reasoning Index in the Wechsler Intelligence one measure of spatial skills or mathematical skills. Therefore,
Scale for Children, Fourth Edition (WISC-IV; Flanagan & multiple effect sizes between different types of spatial tests
Kaufman, 2004), the Culture Fair test (Cattell, 1971), and and measures of mathematical skills were reported based on
the Diagramming Relationships Test (Ekstrom et al., 1976). the same sample. To account for the dependent structure of
effect sizes within studies, robust variance estimation (RVE)
Verbal skills Measures of verbal skills assess participants’ vo- with the small-sample correction technique (Hedges et al.,
cabulary, verbal comprehension skills, and/or verbal reason- 2010; Tipton, 2015) was used. Specifically, we incorporated
ing capacity, and are strong predictors of achievement in an extended working model for RVE that models the correlat-
school and the ability to learn in non-school settings ed and hierarchical structure of effect size estimates, called the
(Gottfredson & Deary, 2004). Examples of verbal measures Correlated and Hierarchical Effects (CHE) working model
include the WAIS-R Information (Wechsler, 1981), the (Pustejovsky & Tipton, 2021). Many studies included in the
Woodcock-Johnson III Picture Vocabulary (Woodcock synthesis report multiple correlations from the same partici-
et al., 2001), and the verbal portion of the SAT (SAT-V; pants (e.g., the correlation between algebra performance and
College Board, 2020). mental rotation skills and algebra performance and
perspective-taking skills) leading to dependencies among cor-
Effect size measures relations nested within individuals and studies. The CHE
working model represents our data structure and is described
Correlation coefficients representing the bivariate correlation by Pustejovsky and Tipton (2021) as the most broadly appli-
between spatial skills and mathematical skills, spatial skills cable model in social science systematic reviews. In applying
and a measure of domain-general reasoning skills (i.e., fluid CHE to the data, we assumed a constant correlation of .8 for
reasoning, working memory capacity, and verbal skills), and the effect sizes within studies.
mathematical skills and a measure of domain-general reason- The random-effects meta-regression models were fit using
ing skills were used as measures of effect size for this study. the CHE working model to estimate the mean correlation be-
Prior to conducting any analyses, for our first meta-analysis tween spatial skills and mathematical skills, and to examine
(examining the magnitude of the relation between spatial whether the correlation between spatial skills and
710 Psychon Bull Rev (2022) 29:699–720

mathematical skills was moderated by grade or gender. The R the same approaches as we used in our meta-analysis (i.e.,
packages metafor (Viechtbauer, 2010) and clubSandwich RVE with CHE working model using Fisher’s z scale effect
(Pustejovsky, 2020) were used for the analyses, and the vari- sizes was used to reduce any artifactual correlation between
ance components were estimated using restricted maximum the effect size and its variance estimates (i.e., measure of pre-
likelihood (REML). cision).1 The results suggested that funnel plot asymmetry was
not present (p = .642), indicating no evidence of publication
Study 2: How do domain-general reasoning skills influence bias.
the relation between spatial skills and mathematical skills?
To examine whether domain-general reasoning skills account
for some of the relation between spatial skills and mathemat- Results
ical skills, we hypothesized a path model with verbal skills,
and fluid reasoning as mediators of the relationship from spa- Sampling characteristics
tial skills to mathematical skills as seen in Fig. 3. We had
originally planned to use working memory in the path model, Of the 45 studies examined for the two meta-analyses, 18 of
but we decided to remove it due to the insufficient numbers of the studies were conducted with participants outside of the
studies that reported correlations with working memory. For USA. A summary of the different countries from which par-
example, the nine correlation coefficients including working ticipants were recruited is provided in Table 3. Regarding the
memory and verbal skills are derived from only three studies distribution of studies across educational levels, 18 of the
(see Table 2). This adversely affects the degrees of freedom of studies were conducted with participants in preschool/
the pooled correlation coefficients. The degrees of freedom of primary grades (i.e., preschool to fifth grades), 18 of the stud-
the pooled correlation were less than 4, which indicates unre- ies were conducted with participants in secondary grades
liable estimates when using RVE (Tipton, 2015). Thus, we (sixth–12th grades), and one study was conducted with par-
removed working memory from the path model. ticipants in both preschool/primary and secondary grades. The
We used MASEM with a two-stage structural equation remaining eight studies were conducted in participants at the
modeling (TSSEM) process to estimate the path model post-secondary level.
(Cheung & Chan, 2005). In the first stage of MASEM, the
correlation matrices are pooled across studies. In the second Descriptive
stage, the pooled correlation matrix from the first stage is used
to fit the structural equation model. In the first stage, we im- In total, the data include 568 correlation coefficients across ten
plemented RVE with the small-sample correction using the types of correlations from 45 studies. Table 2 shows the num-
CHE working model (Pustejovsky & Tipton, 2021) to pool ber of reported correlation coefficient effect sizes and the
the correlations in order to incorporate dependent effect-size number of studies by the type of correlation. Again, because
estimates within study. Our approach differs from a standard the unit of analysis is the study, not the individual participants,
TSSEM application, which assumes one correlation matrix is the number of studies is our focused sample size in meta-
reported per study, a limitation of current MASEM applica- analysis. Note that for Study 1, we used the subset of the
tions (Wilson et al., 2016). Thus, we implemented RVE so dataset only using the correlations between spatial and math-
that the variance-covariance matrix of the pooled correlation ematical skills.
matrix takes into account the dependent structure of correla-
tions within studies. As in the meta-regression analysis, the R
packages metafor (Viechtbauer, 2010) and clubSandwich
Study 1
(Pustejovsky, 2020) were used. In the second stage, we fit
In study 1, we estimated the pooled correlations between spa-
the path model using the 4 × 4 pooled correlation and
tial and mathematical skills and examined whether grade and
variance-covariance matrices. We used the metaSEM R pack-
gender moderated the relationship between spatial and math-
age (Cheung, 2015) to estimate the path model provided in
ematical skills. To estimate the pooled correlation between
Fig. 3.
spatial and mathematical skills, we fit an intercept-only me-
ta-regression under the CHE working model. The pooled ef-
Publication bias and selective reporting
fect size in the correlation coefficient metric was .358 (95% CI
[.295, .419]) and it is statistically significantly different from
To examine possible publication bias in the data, we imple-
zero (β^ = .375, SE = .035, t (43.277) = 10.635, p < 001). The
mented a modified version of Egger’s test of funnel plot asym-
metry (Egger et al., 1997), or Egger’s sandwich (Rodgers & 1
We adopted this idea from Pustejovsky and Rodgers (2019), who focused on
Pustejovsky, 2020), that can be used when dependent effect the standardized mean difference. In the current study, we applied this idea to
sizes exist in the meta-analysis data. Egger’s sandwich uses correlation coefficients and Fisher’s z transformation.
Psychon Bull Rev (2022) 29:699–720 711

Table 2 The number of correlations, the number of studies, and the total average number of participants for each pairing

Pairing Number of correlations Number of studies Total averaged n

Math and Fluid Reasoning 48 14 1,711

Math and Verbal 81 18 14,286
Math and Working Memory 67 11 1,716.86
Spatial and Fluid Reasoning 32 12 1364
Spatial and Math 181 45 17833.93
Spatial and Verbal 96 19 14326
Spatial and Working Memory 25 10 1401
Verbal and Fluid Reasoning 20 6 596
Working Memory and Fluid Reasoning 9 6 976
Working Memory and Verbal 9 3 377

Note. The “Total Averaged n” is the sum of the averaged sample size across the studies reporting the relevant correlation. Averaged sample size is the
mean of the sample size for each relevant correlation within each study
Spatial denotes spatial skills, Math denotes mathematical skills, and Verbal denotes verbal skills

estimated between-study heterogeneity (τ2) was .039 and the studies in this variable. As discussed by Hedges et al. (2010)
estimated within-study heterogeneity (ω2) was .017. as well as Tipton (2015), results from RVE with small degrees
Next, we conducted a meta-regression using grade and of freedom should be interpreted cautiously. In sum, results
proportion of males as moderator variables. There were three indicated a positive moderate association between spatial
missing values in grade from two studies and five missing skills and mathematical achievement. However, neither grade
values in proportion of males from four studies among 181 nor gender significantly moderated the relationship.
effect size estimates. The effect sizes with missing moderators
were deleted to fit the meta-regression model. Thus, 173 cor- Study 2
relation coefficients from 41 studies were used to examine the
effects of grade and proportion of males. The results of the In study 2, a path model was examined to understand the
meta-regression model are presented in Table 4. Neither mod- relationships between spatial and mathematical skills as well
erator was significantly related to effect size. The degrees of as fluid reasoning and verbal skills as mediators. In this set of
freedom for the test of proportion of males in the sample is less analyses, the 4 × 4 correlation matrix of these four variables is
than 4, indicating little information in the data about the rela- the unit of analysis.
tionship between the proportion of males in the sample and the For the first stage of TSSEM, we pooled correlation matri-
correlation between spatial skills and mathematical skills. The ces across studies and estimated the variance-covariance ma-
proportion of males in the studies has a mean of 0.496 with a trix using the CHE model given the presence of dependent
standard deviation of 0.13, indicating little variation among correlations within studies (Pustejovsky & Tipton, 2021).

Fig. 3 Path model examining the relations between spatial skills and mathematical skills. Note. This figure shows the paths between spatial skills,
mathematical skills, and general reasoning skills (i.e., fluid reasoning and verbal skills) examined using a modified TSSEM approach for MASEM.
712 Psychon Bull Rev (2022) 29:699–720

Table 3 Summary of the study locations for the studies included in the Table 5 The pooled correlation matrix in the lower triangle and its
meta-analyses standard error estimates in the upper triangle

Country No. of studies Variables Spatial Verbal Fluid Reasoning Math

China 6 Spatial .036 .031 .031

Switzerland 1 Verbal .286* .038 .045
Finland 3 Fluid Reasoning .418* .277* .032
Australia 1 Math .346* .313* .334*
Russia 1
Spatial denotes spatial skills, Math denotes mathematical skills, and
Canada 2
Verbal denotes verbal skills. * p < .05
Netherlands 1
Spain 1 coefficient from different sample sizes. The hypothesized path
United Kingdom 1 model included a direct path from spatial skills to mathemat-
Sweden 1 ical skills and indirect paths through verbal skills and fluid
United States 27 reasoning (see Fig. 3). We implemented likelihood-based con-
fidence intervals (Neale & Miller, 1997) to test indirect and
Note. Study location is the country from which participants were recruited
in the study direct effects. The approximate goodness-of-fit indices for the
model were mixed with root mean square error of approxima-
tion (RMSEA; Brown & Cudeck, 1993) of .043 (95% CI
Table 5 provides the pooled correlation matrix and Table 6 [.031, .056]), and standardized root mean residual (SRMR;
provides its corresponding standard error estimates. The Hu & Bentler, 1999) of .110. Cheung (2015) suggested
pooled correlation coefficients among four variables ranged RMSEA and SRMR to evaluate the second stage of
between .277 (verbal skill and fluid reasoning) and .418 (spa- TSSEM. RMSEA less than or close to 0.06 and SRMR less
tial skill and fluid reasoning). All six correlation coefficients than or close to 0.08 support good fit (Hu & Bentler, 1999).
were statistically significant at α = .05 and generally moder- The model fit indices showed somewhat mixed conclusions
ate in magnitude. about the hypothesized path model. Given the exploratory
We also examined whether the pooled correlation coeffi- nature of this analysis, we treat the interpretation of the path
cients differed due to the moderators of grade and gender (i.e., coefficients including direct and indirect effects among the
proportion of males in a sample). Neither of the moderators skills as suggestive. Table 7 shows the results of the path
were significantly related to the magnitude of the correlation model including direct and indirect effects. All direct and in-
coefficients. Thus, we used a single pooled correlation matrix direct paths were statistically significant (p < .05). The stron-
to fit a path model in the second stage. gest direct path was spatial skills to fluid reasoning. The direct
In the second stage of TSSEM, we fit a path model using path from spatial skills to mathematical skills remained signif-
the pooled correlation matrix and variance-covariance matrix icant, even after accounting for the indirect paths that included
in the first stage. The total sample size used in this stage was fluid reasoning or verbal skills. Fluid reasoning mediated
17,824.80 obtained by averaging sample size within studies more of the effect from spatial skills to mathematical skills
and summing across studies. The total averaged sample size than did verbal skills, and the relation that remained between
within study per pooled correlation coefficient is presented in spatial skills and mathematical skills was larger than both of
Table 2. The second stage ensures that the parameter estimates these indirect relations. Figure 3 shows the path model with
take into account the precision of each pooled correlation the path coefficients.

Table 4 Results of a meta-regression examining the effect of age and

gender on the relations between mathematics and spatial skills Discussion
Coefficient r SE t df p
This study investigated the relations between spatial skills and
Intercept 0.27 0.06 4.32 3.12 0.02 mathematical skills. Furthermore, it examined the effect of
Grade -0.0002 0.01 -0.03 22.72 0.97 gender and age on this association, and the role of domain-
Proportion Male 0.20 0.07 2.76 1.65 0.14 general reasoning skills in the relations between these two
constructs. Results from synthesizing the reported effect sizes
Note. The reported estimated coefficients (r) are Inverse-Fisher’s z trans-
formed to correlation metric and test results are in Fisher’s z metric. The
between spatial skills and mathematical skills from 45 studies
degrees of freedom (df) and p values are based on a Satterthwaite approx- revealed a positive moderate association between the two skill
imation for small-sample correction sets (r = .36, robust standard error = 0.035, τ2 = 0.039).
Psychon Bull Rev (2022) 29:699–720 713

Table 6 Correlation estimates between spatial skills, mathematical et al., 2016, 2017; Verdine et al., 2017). More specifically,
skills, and verbal skills
our study verifies findings of previous studies conducted
Correlation r SE t df p using experimental, factor analytic, and meta-analytic
methods showing that spatial skills and mathematical skills
Spatial – Math 0.35 0.03 11.21 43.21 <0.001 have a direct positive association (e.g., Lombardi et al.,
Spatial – Verbal 0.29 0.04 7.98 30.71 <0.001 2019; Mix et al., 2016, 2017; Xie et al., 2019).
Spatial – Fluid Reasoning 0.42 0.03 13.30 17.32 <0.001 The magnitude of the relation between spatial skills and
Math – Verbal 0.31 0.05 6.95 31.26 <0.001 mathematical achievement found in our study (r = .36) is
Math – Fluid Reasoning 0.33 0.03 10.49 21.28 <0.001 meaningful as it is consistent with findings of existing meta-
Verbal – Fluid Reasoning 0.28 0.04 7.31 7.19 <0.001 analyses examining the relations between various cognitive
skills and mathematical achievement. For instance, Jacob
Spatial denotes spatial skills, Math denotes mathematical skills, and
Verbal denotes verbal skills and Parkinson (2015) used meta-analytic techniques to iden-
The degrees of freedom (df) and p values are based on a Satterthwaite tify the relation between executive functioning skills and
approximation for small-sample correction mathematical achievement in children aged 2–18 years.
Their results showed that executive functioning skills and
mathematical achievement are positively and moderately cor-
However, no significant effect of gender or grade on the as- related (r = 0.31). Similarly, Peng et al. (2019) used meta-
sociation was found. analytic techniques to determine the relation between fluid
By implementing the meta-analytic SEM (MASEM) ap- intelligence and mathematical achievement in individuals
proach, the current study was able to model directional paths aged 3–80 years. Their analyses revealed a positive moderate
among the variables with direct and indirect effects simulta- association between fluid intelligence and mathematical
neously. In addition, we utilized RVE with the CHE working achievement (r = 0.41). Lastly, our findings replicate the find-
model approach (Pustejovsky & Tipton, 2021) to handle de- ings of the meta-analysis conducted by Xie et al. (2019) who
pendent effect sizes within studies when estimating the pooled also found a moderate positive relation between spatial skills
correlation matrix and variance-covariance matrix. Results in- and mathematical achievement (r = 0.27). These findings sug-
dicated that fluid reasoning and verbal skills mediated the gest that spatial skills are similarly related to students’ math-
relationships between spatial skills and mathematical skills, ematical understanding as other cognitive processes, such as
and these indirect effects were small but statistically signifi- executive functioning skills and fluid intelligence. Future re-
cant. In addition to the indirect effects, the direct relation from search should examine how to leverage these pertinent cogni-
spatial skills to mathematical skills was also statistically sig- tive skill sets to bolster students’ mathematics understanding
nificant and larger than the indirect effects. and performance.
Consistent with prior research, this meta-analysis confirms However, unlike many studies closely examining the rela-
that spatial skills are significantly related to mathematical tionship between spatial skills and mathematical achievement
skills (Atit, Power, et al., 2020a; Casey et al., 1995; Mix (e.g., Casey et al., 1997; Li & Geary, 2013; Stannard et al.,

Table 7 The results of the analyzed path model

Path Estimate 95% confidence interval

CI.L CI.U

Direct effect
Spatial → Math .18 .13 .24
Indirect effect
Spatial → Fluid Reasoning → Math .08 .05 .11
Spatial → Fluid Reasoning .41 .33 .48
Fluid Reasoning → Math .19 .13 .25
Spatial → Verbal → Math .02 .01 .05
Spatial → Verbal .20 .13 .28
Verbal → Math .12 .06 .19

Spatial denotes spatial skills, Math denotes mathematical skills, and Verbal denotes verbal skills
CI.L is the lower bound of the 95% confidence interval, CI.U is the upper bound of the 95% confidence interval
714 Psychon Bull Rev (2022) 29:699–720

2001), we found that the association between these two skill inhibition (Miyake et al., 2000), on the relations be-
sets was not influenced by participants’ grade or gender. It is tween spatial skills and mathematical performance.
well established that there is more than one kind of spatial skill One limitation of this meta-analysis is that while we tried to
(e.g., Linn & Petersen, 1985; Newcombe & Shipley, 2015) include all relevant studies in our analysis, some eligible stud-
and more than one kind of mathematical domain (American ies may have been missed due to our search strategy. We used
Mathematical Society, 2020). Many of the studies investigat- generalized terms such as “spatial skills” or “spatial ability” or
ing the effects of gender and age/grade on the association “mathematics” or “math skills.” However, as many studies
between the two constructs focus only on one kind of spatial focus on a specific kind of spatial skill or mathematical skill,
skill (e.g., Casey et al., 2017; Li & Geary, 2013) and/or one they may have only used such specific terms, such as mental
kind of mathematical domain (e.g., Lombardi et al., 2019). In rotation or algebra, in their articles. Therefore, future studies
our study, we did not examine whether the relations differed should consider using a broad variety of search terms to re-
for different kinds of spatial skills or different kinds of math- duce the possibility of missing eligible articles. Secondly, the
ematical concepts. The meta-analysis conducted by Xie et al. small number of effect sizes limited our ability to examine the
(2019), which did consider domains of spatial skills and math- relations between different kinds of spatial skills and different
ematical skills in their analyses, did not examine the effect of kinds of mathematical domains. Thus, more effect sizes be-
participants’ age/grade or gender for each subarea. Thus, the tween sub-constructs are needed to examine these relation-
nuances of how gender and age/grade influence the relations ships in future studies.
between spatial skills and mathematical skills have yet to be Another limitation of this research is that the studies includ-
understood. Future meta-analyses should aim to disentangle ed in these meta-analyses were largely from the USA and
how different kinds of spatial skills are related to different other Western nations. Only a small number of studies, spe-
kinds of mathematical skills, and identify the role of age/ cifically seven studies, were conducted in Eastern countries
grade and gender in these various associations. (i.e., China and Russia). The lack of international diversity of
In line with findings on the importance of executive func- the participants in the selected studies potentially biases our
tioning skills in mathematical achievement (e.g., Green et al., results as existing data indicates that mathematical achieve-
2017; McGrew & Wendling, 2010), our study found that fluid ment varies greatly between students from eastern and western
reasoning skills were an essential component of mediating the nations (Mullis et al., 2012). Thus, more primary studies ex-
relation between spatial and mathematical skills. In a synthesis amining the relations between spatial skills and mathematical
of studies investigating the concurrent relationships between skills need to be conducted in samples across a broader range
various cognitive abilities and achievement measures, of countries. Furthermore, the search and inclusion criteria for
McGrew and Wendling (2010) found that fluid reasoning literature to be examined in future meta-analyses needs to be
was one of the three broad cognitive abilities (fluid reasoning, revised, perhaps by expanding the literature search to studies
verbal comprehension, and processing speed) that was consis- published in languages other than English, so that a broader
tently related to mathematical performance at all age ranges. range of international studies are represented in the analyses.
Furthermore, in a longitudinal study, Green et al. (2017) found The findings from these meta-analyses have multiple im-
that fluid reasoning, spatial skills, and verbal skills accounted plications for mathematics education research and practice.
for 90% of the variance in future math achievement in indi- First, this research highlights that integrating practices that
viduals ranging from 5 to 15 years of age, with fluid reasoning develop and support students’ spatial skills could benefit stu-
skills being the only significant predictor. In our study, the dents’ mathematical understanding at all educational levels.
indirect relation from spatial skills to fluid reasoning to math- Many of the efforts aimed at improving students’ mathemat-
ematical skills was larger in magnitude compared to the indi- ical achievement by way of bolstering their spatial skills have
rect relation from spatial skills to verbal skills to mathematical been carried out with students at the preschool/primary and
skills. However, the direct path from spatial skills to mathe- secondary levels (e.g., Cheng & Mix, 2014; Hawes et al.,
matical skills was larger than the two indirect paths. These 2015; Lowrie et al., 2017; Lowrie et al., 2019; Schmitt et al.,
results are consistent with Kovacs and Conway’s (2016) 2018). Only a handful of efforts have been made to bolster
process overlap theory, which poses that domain- students’ mathematical outcomes by improving their spatial
general executive processes, such as fluid reasoning, skills at the postsecondary level (e.g., Sorby, 2007; Sorby
underlie performance on domain-specific cognitive et al., 2013). However, our research shows that the relations
tasks, such as spatial and mathematical measures. between spatial skills and mathematical skills do not vary by
There were not enough studies that measured working participants’ grade level. Therefore, future research should
memory capacity to include the construct in our estimat- examine how to support mathematics instructors at the post-
ed model. Thus, an area of focus for future research is secondary level in developing and supporting their students’
to ascertain the role of additional executive functioning spatial skills.
processes, such as working memory capacity and
Psychon Bull Rev (2022) 29:699–720 715

Second, while our findings underline the importance of References

developing students’ spatial skills to support their mathemat-
ics learning and achievement, efforts should also focus on Ackerman, P. L., & Wolman, S. D. (2007). Determinants and validity of
self-estimates of abilities and self-concept measures. Journal of
developing students’ domain-general cognitive processes,
Experimental Psychology. Applied, 13(2), 57–78. [Link]
such as their verbal skills and fluid reasoning skills. Our re- 10.1037/1076-898X.13.2.57
search shows that domain-general cognitive processes ac- Altman, D. G. (1990). Practical Statistics for Medical Research. CRC
count for some of the variance between spatial skills and Press.
mathematical skills. This suggests that bolstering spatial skills American Mathematical Society (2020). Homepage. Retrieved
November 24, 2020, from [Link]
in conjunction with fluid reasoning and verbal skills may pro-
Atit, K., Gagnier, K., & Shipley, T. F. (2015). Student gestures aid pen-
vide a greater boost to students’ mathematical achievement etrative thinking. Journal of Geoscience Education, 63(1), 66–72.
than developing spatial skills alone. Thus, future research [Link]
should aim to identify pedagogical practices focused on de- Atit, K., Power, J. R., Veurink, N., Uttal, D. H., Sorby, S., Panther, G.,
veloping students’ domain-general reasoning skills as well as Msall, C., Fiorella, L., & Carr, M. (2020a). Examining the role of
spatial skills and mathematics motivation on middle school mathe-
their spatial skills, which can be integrated into preschool to matical achievement. International Journal of STEM Education,
postsecondary mathematics curricula. 7(1), 38. [Link]
Atit, K., Uttal, D. H., & Stieff, M. (2020b). Situating space: Using a
discipline-focused lens to examine spatial thinking skills.
Conclusions Cognitive Research: Principles and Implications, 5(1), 19. https://
[Link]/10.1186/s41235-020-00210-z
In conclusion, the current meta-analysis affirms that spatial Aud, S., Hussar, W., Johnson, F., Kena, G., Roth, E., Manning, E., Wang,
X., & Zhang, J. (2012). The condition of education 2012. NCES
skills and mathematical skills are positively related and, more- 2012-045. National Center for Education Statistics. [Link]
over, there is a direct relationship between the two constructs [Link]/fulltext/[Link]
that did not vary based on grade or gender. Furthermore, we Battista, M. T. (1990). Spatial visualization and gender differences in
found that other cognitive processes, specifically fluid reason- high school geometry. Journal for Research in Mathematics
Education, 21(1), 47–60. [Link]
ing skills and verbal skills, mediated the relations between
Berg, D. H. (2008). Working memory and arithmetic calculation in chil-
spatial skills and mathematical skills, and yet a direct relation dren: The contributory roles of processing speed, short-term mem-
between the two constructs independent of other general cog- ory, and reading. Journal of Experimental Child Psychology, 99(4),
nitive processes remained. These findings indicate that efforts 288–308. [Link]
to improve mathematical skills should include bolstering spa- Berman, N. G., & Parker, R. A. (2002). Meta-analysis: Neither quick nor
easy. BMC Medical Research Methodology, 2, 10. [Link]
tial skills but may also consider support for other pertinent 10.1186/1471-2288-2-10
cognitive skill sets. Bethell-Fox, C. E., & Shepard, R. N. (1988). Mental rotation: Effects of
stimulus complexity and familiarity. Journal of Experimental
Psychology. Human Perception and Performance, 14(1), 12–23.
Acknowledgements We would like to acknowledge Dr. Martha Carr [Link]
who contributed to the initial stages of this research but sadly passed away Bonny, J. W., & Lourenco, S. F. (2015). Individual differences in chil-
during the summer of 2017. dren’s approximations of area correlate with competence in basic
geometry. Learning and Individual Differences, 44, 16–24. https://
Availability of data and materials The dataset generated during and/or [Link]/10.1016/[Link].2015.11.001
analyzed during the current study are available at the Open Science Borenstein, M., Hedges, L., Higgins, J., & Rothstein, H. (2005).
Framework, [Link]/h3w7j. Comprehensive Meta-analysis (Version 2). Biostat.
Brown, M. W., & Cudeck, R. (1993). Alternative ways of assessing
Code availability The coding scheme and script used in conducting the model fit. Sage Focus Editions, 154, 136-136.
analysis is also available from the corresponding author upon reasonable Burte, H., Gardony, A. L., Hutton, A., & Taylor, H. A. (2017). Think3d!:
request. The dataset and analysis code associated with this project are Improving mathematics learning through embodied spatial training.
available at the Open Science Framework, [Link]/h3w7j. Cognitive Research: Principles and Implications, 2(1), 13. https://
[Link]/10.1186/s41235-017-0052-9
Funding This research was partially supported by a grant from the US Carr, M., Steiner, H. H., Kyser, B., & Biddlecomb, B. (2008). A com-
Department of Education (award number R305A150365). parison of predictors of early emerging gender differences in math-
ematics competency. Learning and Individual Differences, 18(1),
61–75. [Link]
Declarations
Carroll, J. B. (1993). Human Cognitive Abilities: A Survey of Factor-
Analytic Studies. Cambridge University Press. [Link]
Conflicts of interest/Competing interests The authors have no financial [Link]/details?id=book-jp9dt4_0_cIC
or non-financial interests to disclose.
Casey, M. B., Nuttall, R., Pezaris, E., & Benbow, C. P. (1995). The
influence of spatial ability on gender differences in mathematics
Ethics approval NA college entrance test scores across diverse samples. Developmental
Psychology, 31(4), 697–705. [Link]
Consent to participate NA 4.697
716 Psychon Bull Rev (2022) 29:699–720

Casey, M. B., Nuttall, R. L., & Pezaris, E. (1997). Mediators of gender Education, 48(3), 282–298. [Link]
differences in mathematics college entrance test scores: a compari- 2007.11650106
son of spatial skills with internalized beliefs and anxieties. Egger, M., Davey Smith, G., Schneider, M., & Minder, C. (1997). Bias in
Developmental Psychology, 33(4), 669–680. [Link] meta-analysis detected by a simple, graphical test. BMJ, 315(7109),
1037//0012-1649.33.4.669 629–634. [Link]
Casey, B. M., Dearing, E., Vasilyeva, M., Ganley, C. M., & Tine, M. Ekstrom, R. B., French, J. W., Harman, H., & Derman, D. (1976). Kit of
(2011). Spatial and numerical predictors of measurement perfor- Factor-referenced cognitive tests (revised edition). Educational
mance: The moderating effects of community income and gender. Testing Service, Princeton, NJ.
Journal of Educational Psychology, 103(2), 296–311. [Link] Embretson, S. E. (1995). The role of working memory capacity and
org/10.1037/a0022516 general control processes in intelligence. Intelligence, 20(2), 169–
Casey, B. M., Pezaris, E., Fineman, B., Pollock, A., Demers, L., & 189. [Link]
Dearing, E. (2015). A longitudinal analysis of early spatial skills Flanagan, D. P. & Kaufman, A. S. (2004). Essentials of WISC-IV
compared to arithmetic and verbal skills as predictors of fifth- Assessment. John Wiley & Sons.
grade girls’ math reasoning. Learning and Individual Differences, Frick, A. (2019). Spatial transformation abilities and their relation to later
40, 90–100. [Link] mathematics performance. Psychological Research, 83(7), 1465–
Casey, B. M., Lombardi, C. M., Pollock, A., Fineman, B., & Pezaris, E. 1484. [Link]
(2017). Girls’ spatial skills and arithmetic strategies in first grade as Fry, A. F. & Hale, S. (1996). Processing speed, working memory, and
predictors of fifth-grade analytical math reasoning. Journal of fluid intelligence: Evidence for a developmental cascade.
Cognition and Development: Official Journal of the Cognitive Psychological Science, 7(4), 237-241.
Development Society, 18(5), 530–555. [Link] Ganley, C. M, & Vasilyeva, M. (2011). Sex differences in the relation
15248372.2017.1363044 between math performance, spatial skills, and attitudes. Journal of
Cattell, R. B. (1971). Abilities: Their Structure, Growth, and Action. Applied Developmental Psychology, 32(4), 235–242. [Link]
Houghton Mifflin. org/10.1016/[Link].2011.04.001
Ceci, S. J., & Williams, W. M. (2011). Understanding current causes of Geary, D. C. (2011). Cognitive predictors of achievement growth in
women’s underrepresentation in science. Proceedings of the mathematics: A five-year longitudinal study. Developmental
National Academy of Sciences of the United States of America, Psychology, 47(6), 1539-1552. [Link]
108(8), 3157–3162. [Link] Geer, E. A., & Ganley, C. M. (2021). Sex differences in social and spatial
Cheng, Y. L., & Mix, K. S. (2014). Spatial training improves children’s perspective taking: A replication and extension of Tarampi et al.,
mathematics ability. Journal of Cognition and Development, 15(1), (2016). Unpublished manuscript.
2-11. Geer, E. A., Quinn, J. M., & Ganley, C. M. (2019). Relations between
Cheung, M.W.L. (2015). {metaSEM}: An R package for meta-analysis spatial skills and math performance in elementary school children: A
using structural equation modeling. Frontiers in Psychology, 5, 1-7. longitudinal investigation. Developmental Psychology, 55(3), 637–
[Link] 652. [Link]
Cheung, M. W.-L., & Chan, W. (2005). Meta-analytic structural equation Gilles, P.-Y., & Bailleux, C. (2001). Personality traits and abilities as
modeling: A two-stage approach. Psychological Methods, 10(1), predictors of academic achievement. European Journal of
40–64. [Link] Psychology of Education, 16(1), 3–15. [Link]
Cirino, P. T., Tolar, T. D., Fuchs, L. S., & Huston-Warren, E. (2016). BF03172991
Cognitive and numerosity predictors of mathematical skills in mid- Gilligan, K. A., Flouri, E., & Farran, E. K. (2017). The contribution of
dle school. Journal of Experimental Child Psychology, 145, 95– spatial ability to mathematical achievement in middle childhood.
119. [Link] Journal of Experimental Child Psychology, 163, 107–125. https://
Clifford, E. (2008). Visual-spatial processing and mathematics achieve- [Link]/10.1016/[Link].2017.04.016
ment: the predictive ability of the visual-spatial measures of the Gottfredson, L. S., & Deary, I. J. (2004). Intelligence predicts health and
Stanford-Binet intelligence scales, and the Wechsler intelligence longevity, but why? Current Directions in Psychological Science,
scale for children. University of South Dakota. [Link] 13(1), 1–4. [Link]
[Link]/openview/7909403d7b16e45eca563dbc3826e3eb/1? Green, C. T., Bunge, S. A., Briones Chiongbian, V., Barrow, M., &
pq-origsite=gscholar&cbl=18750&diss=y Ferrer, E. (2017). Fluid reasoning predicts future mathematical per-
College Board (2020). SAT Math [Link]://collegereadiness. formance among children and adolescents. Journal of Experimental
[Link]/sat/inside-the-test/math Child Psychology, 157, 125–143. [Link]
Conway, A. R. A., Kane, M. J., & Engle, R. W. (2003). Working memory 2016.12.005
capacity and its relation to general intelligence. Trends in Cognitive GrÉGoire, J., & Van Der Linden, M. (1997). Effect of age on forward and
Sciences, 7(12), 547–552. [Link] backward digit spans. Neuropsychology, Development, and
005 Cognition. Section B, Aging, Neuropsychology and Cognition,
Corsi, P. M. (1972). Human Memory and the Medial Temporal Region of 4(2), 140–149. [Link]
the Brain. McGill University. Gunderson, E. A., Ramirez, G., Beilock, S. L., & Levine, S. C. (2012).
Cox, J. W. (1928). Mechanical aptitude. Methuen. The relation between spatial skill and early number knowledge: the
Cui, J., Zhang, Y., Cheng, D., Li, D., & Zhou, X. (2017). Visual form role of the linear number line. Developmental Psychology, 48(5),
perception can be a cognitive correlate of lower-level math catego- 1229–1241. [Link]
ries for teenagers. Frontiers in Psychology, 8, 1336. [Link] Haciomeroglu, E. S. (2015). The role of cognitive ability and preferred
10.3389/fpsyg.2017.01336 mode of processing in students’ calculus performance. Eurasia
Deary, I. J., Strand, S., Smith, P., & Fernandes, C. (2007). Intelligence Journal of Mathematics, Science & Technology Education, 11(5),
and educational achievement. Intelligence, 35(1), 13-21. 1165–1179. [Link]
Delgado, A. R., & Prieto, G. (2004). Cognitive mediators and sex-related The%20Role%20of%[Link]
differences in mathematics. Intelligence, 32(1), 25–32. [Link] Haciomeroglu, E. S. (2016). Object-spatial visualization and verbal cog-
org/10.1016/S0160-2896(03)00061-8 nitive styles, and their relation to cognitive abilities and mathemat-
Edens, K., & Potter, E. (2007). The relationship of drawing and mathe- ical performance. Educational Sciences: Theory and Practice,
matical problem solving: Draw for math tasks. Studies in Art 16(3), 987–1003. [Link]
Psychon Bull Rev (2022) 29:699–720 717

Halford, G. S., Wilson, W. H., Phillips, S. (1998). Processing capacity Kirby, J. R., & Boulter, D. R. (1999). Spatial ability and transformational
defined by relational complexity: Implications for comparative, de- geometry. European Journal of Psychology of Education, 14(2),
velopmental, and cognitive psychology. Behavioral and Brain 283. [Link]
Sciences, 21(6), 803-831. Klingberg, T. (2006). Development of a superior frontal-intraparietal net-
Halpern, D. F., Benbow, C. P., Geary, D. C., Gur, R. C., Hyde, J. S., & work for visuo-spatial working memory. Neuropsychologia, 44(11),
Gernsbacher, M. A. (2007). The science of sex differences in sci- 2171-2177.
ence and mathematics. Psychological Science in the Public Interest: Kovacs, K., & Conway, A. R. A. (2016). Process overlap theory: A
A Journal of the American Psychological Society, 8(1), 1–51. unified account of the general factor of intelligence. Psychological
[Link] Inquiry, 27(3), 151–177. [Link]
Hanline, M. F., Milton, S., & Phelps, P. C. (2010). The relationship 1153946
between preschool block play and reading and maths abilities in Krawczyk, D. C. (2012). The cognition and neuroscience of relational
early elementary school: A longitudinal study of children with and reasoning. Brain Research, 1428, 13-23.
without disabilities. Early Child Development and Care, 180(8), Kytällä, M., & Lehto, J. E. (2008). Some factors underlying mathematical
1005–1017. [Link] performance: The role of visuospatial working memory and non-
Harter, C. A. & Ku, H. (2008) The effects of spatial contiguity within verbal intelligence. European Journal of Psychology of Education,
computer-based instruction of group personalized two-step mathe- 23(1), 77. [Link]
matics word problems. Computers in Human Behavior,24(4), 1668- Kyttälä, M., Aunio, P., Lehto, J. E., Van Luit, J., & Hautamäki, J. (2003).
1685. [Link] Visuospatial working memory and early numeracy. Educational
Hawes, Z. & Ansari, D. (2020). What explains the relationship between and Child Psychology, 20(3), 65–76.
spatial and mathematical skills? A review of evidence from brain Li, Y., & Geary, D. C. (2013). Developmental gains in visuospatial mem-
and behavior. Psychonomic Bulletin & Review, 27, 465-482. https:// ory predict gains in mathematical achievement. In PLoS ONE (Vol.
[Link]/10.3758/s13423-019-01694-7 8, Issue 7, p. e70160). [Link]
Hawes, Z., Moss, J., Caswell, B., & Poliszczuk, D. (2015). Effects of 0070160
mental rotation training on children’s spatial and mathematics per- Linn, M. C., & Petersen, A. C. (1985). Emergence and characterization of
formance: A randomized controlled study. Trends in Neuroscience sex differences in spatial ability: a meta-analysis. Child
and Education, 4(3). 60-68. Development, 56(6), 1479–1498. [Link]
Hawes, Z., Moss, J., Caswell, B., Seo, J., Ansari, D. (2019). Relations pubmed/4075870
between numerical, spatial, and executive function skills and math- Lombardi, C. M., Casey, B. M., Pezaris, E., Shadmehr, M., & Jong, M.
ematical achievement: A latent-variable approach. Cognitive (2019). Longitudinal analysis of associations between 3-D mental
Psychology, 109, 68-90. rotation and mathematics reasoning skills during middle school:
Hedges, L. V., Tipton, E., & Johnson, M. C. (2010). Robust variance Across and within genders. Journal of Cognition and
estimation in meta-regression with dependent effect size estimates. Development: Official Journal of the Cognitive Development
Research Synthesis Methods, 1(1), 39–65. [Link] Society, 20(4), 487–509. [Link]
jrsm.5 1614592
Hegarty, M., & Kozhevnikov, M. (1999). Types of visual–spatial repre- Lowrie, T., Logan, T., & Ramful, A. (2017). Visuospatial training im-
sentations and mathematical problem solving. Journal of proves elementary students’ mathematics performance. British
Educational Psychology, 91(4), 684–689. [Link] Journal of Educational Technology: Journal of the Council for
0022-0663.91.4.684 Educational Technology. [Link]
Holyoak, K. J. (2012). Analogy and relational reasoning. In K. J. Holyoak Lowrie, T., Logan, T., & Hegarty, M. (2019). The influence of spatial
& R. G. Morrison (Eds.), The Oxford Handbook of Thinking and visualization training on students’ spatial reasoning and mathemat-
Reasoning (pp. 234–259). Oxford University Press. ics performance. Journal of Cognition and [Link]://doi.
Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in org/10.1080/15248372.2019.1653298
covariance structure analysis: Conventional criteria versus new al- McGrew, K. S., & Wendling, B. J. (2010). Cattell–Horn–
ternatives. Structural Equation Modeling: A Multidisciplinary Carrollcognitive-achievement relations: What we have learned from
Journal, 6(1), 1-55. the past 20 years of research. Psychology in the Schools, 47(7), 651–
Hubbard, E. M., Piazza, M., Pinel, P., & Dehaene, S. (2005). Interactions 675. [Link]
between number and space in parietal cortex. In Nature Reviews McKenzie, B., Bull, R., & Gray, C. (2003). The effects of phonological
Neuroscience (Vol. 6, Issue 6, pp. 435–448). [Link] and visual-spatial interference on children’s arithmetical perfor-
1038/nrn1684 mance. Educational and Child Psychology, 20(3), 93–108.
Hyde, J. S., Fennema, E., & Lamon, S. J. (1990). Gender differences in Merriam Webster (n.d.) Definition of MATHEMATICS. Retrieved
mathematics performance: a meta-analysis. Psychological Bulletin, November 18, 2020, from [Link]
107(2), 139–155. [Link] dictionary/mathematics
Imbo, I., Vandierendonck, A., & De Rammelaere, S. (2007). The role of Mix, K. S. (2019). Why are spatial skill and mathematics related? In
working memory in the carry operation of mental arithmetic: num- Child Development Perspectives (Vol. 13, Issue 2, pp. 121–126).
ber and value of the carry. Quarterly Journal of Experimental [Link]
Psychology, 60(5), 708–731. [Link] Mix, K. S., & Cheng, Y.-L. (2012). The relation between space and math:
17470210600762447 developmental and educational implications. Advances in Child
Jacob, R. & Parkinson, J. (2015). The potential for school-based inter- Development and Behavior, 42, 197–243. [Link]
ventions that target executive function to improve academic b978-0-12-394388-0.00006-x
achievement: A review. Review of Educational Research, 85(4), Mix, K. S., Levine, S. C., Cheng, Y.-L., Young, C., Hambrick, D. Z.,
512-552. Ping, R., & Konstantopoulos, S. (2016). Separate but correlated:
Jensen, A. R. (1998). The Science of Mental Ability. Praeger Publishers. The latent structure of space and mathematics across development.
Johnson, C. M. (2017). Mathematics performance of sixth-grade students Journal of Experimental Psychology. General, 145(9), 1206–1227.
in single-gender and mixed classrooms in a large urban school [Link]
system (Publication No. 10604090) [Doctoral dissertation, Grand Mix, K. S., Levine, S. C., Cheng, Y.-L., Young, C. J., Hambrick, D. Z., &
Canyon University]. ProQuest Dissertation Publishing. Konstantopoulos, S. (2017). The latent structure of spatial skills and
718 Psychon Bull Rev (2022) 29:699–720

mathematics: A replication of the two-factor model. Journal of analysis in sixth-grade students. Frontiers in Psychology, 5, 782.
Cognition and Development: Official Journal of the Cognitive [Link]
Development Society, 18(4), 465–492. [Link] Organization for Economic Cooperation and Development (2012).
15248372.2017.1346658 Programme for International Student Assessment (PISA) Results
Miyake, A., & Shah, P. (1999). Models of working memory: from PISA [Link]://[Link]/unitedstates/PISA-2012-
Mechanisms of active maintenance and executive control. [Link]
Cambridge University Press. Organization for Economic Development (2016). Country Note: Key
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, Findings from PISA 2015 for the United [Link]/pisa/
A., & Wager, T. D. (2000). The unity and diversity of executive [Link]
functions and their contributions to complex “frontal lobe” tasks: A Oswald, F. L., McAbee, S. T., Redick, T. S., & Hambrick, D. Z. (2015).
latent variable analysis. Cognitive Psychology, 41(1), 49–100. The development of a short domain-general measure of working
[Link] memory capacity. Behavior Research Methods, 47(4), 1343–1355.
Miyake, A., Friedman, N. P., Rettinger, D. A., Shah, P., & Hegarty, M. [Link]
(2001). How are visuospatial working memory, executive function- Paterson, D. G., Elliot, R., Anderson, L. D., Toops, H. A., & Heidbreder,
ing, and spatial abilities related? A latent-variable analysis. Journal E. (1930). Minnesota mechanical ability tests: The report of a re-
of Experimental Psychology. General, 130(4), 621–640. [Link] search investigation subsidized by the committee on human migra-
org/10.1037//0096-3445.130.4.621 tions of the national research council and conducted in the depart-
Moher, D., Shamseer, L., Clarke, M., Ghersi, D., Liberati, A., Petticrew, ment of psychology of the University of Minnesota. University of
M., Shekelle, P., Stewart, L. A., & PRISMA-P Group. (2015). Minnesota Press.
Preferred reporting items for systematic review and meta-analysis Paunonen, S. V., & Hong, R. Y. (2010). Self-efficacy and the prediction
protocols (PRISMA-P) 2015 statement. Systematic Reviews, 4, 1. of domain-specific cognitive abilities. Journal of Personality, 78(1),
[Link] 339–360. [Link]
Möhring, W., Newcombe, N. S., & Frick, A. (2015). The relation be- Payne, C. M. (2008). So Much Reform, So Little Change: The
tween spatial thinking and proportional reasoning in preschoolers. Persistence of Failure in Urban Schools. Harvard Education Press.
Journal of Experimental Child Psychology, 132, 213–220. https:// Peng, P., Yang, X., & Meng, X. (2017). The relation between approxi-
[Link]/10.1016/[Link].2015.01.005 mate number system and early arithmetic: The mediation role of
Morosanova, V. I., Fomina, T. G., Kovas, Y., & Bogdanova, O. Y. numerical knowledge. Journal of Experimental Child Psychology,
(2016). Cognitive and regulatory characteristics and mathematical 157, 111–124. [Link]
performance in high school students. Personality and Individual Peng, P., Wang, T., Wang, C., Lin, X. (2019). A meta-analysis on the
Differences, 90, 177–186. [Link] relation between fluid intelligence and reading/mathematics: Effects
034 of tasks, age, and social economics status. Psychological Bulletin,
145(2), 189-236.
Mullis, I. V. S., Martin, M. O., Foy, P., Arora, A. (2012). TIMSS 2011
Peters, M., Laeng, B., Latham, K., Jackson, M., Zaiyouna, R., &
International Results in Mathematics. International Association for
Richardson, C. (1995). A redrawn Vandenberg and Kuse mental
the Evaluation of Educational Achievement. [Link]
rotations test: different versions and factors that affect performance.
ed544554
Brain and Cognition, 28(1), 39–58. [Link]
National Center for Education Statistics (2018). Annual reports program
1995.1032
2018. [Link]
Pustejovsky, J. E. (2020). clubSandwich: Cluster-Robust(Sandwich)
referrer=figures
Variance Estimators with Small-Sample Corrections. R package
Neale, M. C., & Miller, M. B. (1997). The use of likelihood-based con- version 0.5.1. [Link]
fidence intervals in genetic models. Behavior genetics, 27(2), 113– Pustejovsky, J. E., & Rodgers, M. A. (2019). Testing for funnel plot
120. asymmetry of standardized mean differences. Research Synthesis
Newcombe, N. S. (2010). Picture this: Increasing math and science learn- Methods, 10(1), 57–71. [Link]
ing by improving spatial thinking. American Educator, 34(2), 29. Pustejovsky, J. E., & Tipton, E. (2021). Meta-analysis with robust vari-
[Link] ance estimation: Expanding the range of working models.
Newcombe, N. S., & Shipley, T. F. (2015). Thinking about spatial think- Manuscript under review. [Link]. [Link]
ing: New typology, new assessments. Studying Visual and Spatial Pyke, C. L. (2003). The use of symbols, words, and diagrams as indica-
Reasoning for Design Creativity, 179–192. [Link] tors of mathematical cognition: A causal model. Journal for
978-94-017-9297-4_10 Research in Mathematics Education, 34(5), 406–432. [Link]
Novack, M. A., Congdon, E. L., Hemani-Lopez, N., & Goldin-Meadow, org/10.2307/30034794
S. (2014). From action to abstraction: using the hands to learn math. Raghubar, K. P., Barnes, M. A., & Hecht, S. A. (2010). Working memory
Psychological Science, 25(4), 903–910. [Link] and mathematics: A review of developmental, individual difference,
0956797613518351 and cognitive approaches. Learning and Individual Differences,
Nuttall, R. L., Casey, M. B., & Pezaris, E. (2005). Spatial ability as a 20(2), 110–122. [Link]
mediator of gender differences on mathematics tests: A biological- Reuhkala, M. (2001). Mathematical skills in ninth graders: Relationship
environmental framework. In A. M. Gallagher (Ed.), Gender differ- with visuo-spatial abilities and working memory. Educational
ences in mathematics: An integrative psychological approach, (Vol. Psychology Review, 21(4), 387–399. [Link]
351, pp. 121–142). Cambridge University Press, xvi. https:// 01443410120090786
[Link]/fulltext/[Link] Rodán, A., Contreras, M. J., Elosúa, M. R., & Gimeno, P. (2016).
Olver, A. (2013). Investigating early spatial and numerical skills in ju- Experimental but not sex differences of a mental rotation training
nior kindergarten children learning in an inquiry-and play-based program on adolescents. Frontiers in Psychology, 7, 1050. https://
environment [Master's Thesis, University of Toronto]. Tspace. [Link]/10.3389/fpsyg.2016.01050
[Link] Rodgers, M. A., & Pustejovsky, J. E. (2020). Evaluating meta-analytic
Oostermeijer, M., Boonen, A. J. H., & Jolles, J. (2014). The relation methods to detect selective reporting in the presence of dependent
between children’s constructive play activities, spatial ability, and effect sizes. Psychological Methods. [Link]
mathematical word problem-solving performance: a mediation met0000300
Psychon Bull Rev (2022) 29:699–720 719

Schmitt, S. A., Korucu, I., Napoli, A. R., Bryant, L. M., & Purpura, D. J. Science, 22(5), 367-373. [Link]
(2018). Using block play to enhance preschool children’s mathemat- 0963721413484756
ics and executive functioning: A randomized controlled trial. Early Valenzeno, L., Alibali, M. W., & Klatzky, R. (2003). Teachers’ gestures
Childhood Research Quarterly, 44, 181–191. facilitate students’ learning: A lesson in symmetry. Contemporary
Schneider, W. J., & McGrew, K. S. (2012). The Cattell-Horn-Carroll Educational Psychology, 28(2), 187–204. [Link]
model of intelligence. In D. P. Flanagan & P. L. Harrison (Eds.), [Link]/science/article/pii/S0361476X02000073
Contemporary Intellectual Assessment: Theories, Tests, and Issues van Garderen, D. (2006). Spatial visualization, visual imagery, and math-
(pp. 99–144). The Guilford Press. ematical problem solving of students with varying abilities. Journal
Seitz, K., & Schumann-Hengsteler, R. (2002). Phonological loop and of Learning Disabilities, 39(6), 496–506. [Link]
central executive processes in mental addition and multiplication. 00222194060390060201
Psychologische Beitrage, 44(2), 275–303. Vandenberg, S. G., & Kuse, A. R. (1978). Mental rotations, a group test
Shepard, R. N., & Metzler, J. (1971). Mental rotation of three- of three-dimensional spatial visualization. Perceptual and Motor
dimensional objects. Science, 171(3972), 701–703. [Link] Skills, 47(2), 599–604. [Link]
10.1126/science.171.3972.701 Vendetti, M. S., & Bunge, S. A. (2014). Evolutionary and developmental
Simms, V., Clayton, S., Cragg, L., Gilmore, C., & Johnson, S. (2016). changes in the lateral frontoparietal network: a little goes a long way
Explaining the relationship between number line estimation and for higher-level cognition. Neuron, 84(5), 906–917. [Link]
mathematical achievement: The role of visuomotor integration and 10.1016/[Link].2014.09.035
visuospatial skills. Journal of Experimental Child Psychology, 145, Ventura, M., Shute, V., Wright, T., & Zhao, W. (2013). An investigation
22–33. [Link] of the validity of the virtual spatial navigation assessment. Frontiers
Skagerlund, K., & Träff, U. (2016). Processing of space, time, and num- in Psychology, 4, 852. [Link]
ber contributes to mathematical abilities above and beyond domain- Verdine, B. N., Golinkoff, R. M., Hirsh-Pasek, K., Newcombe, N. S.,
general cognitive abilities. Journal of Experimental Child Filipowicz, A. T., & Chang, A. (2014). Deconstructing building
Psychology, 143, 85–101. [Link] blocks: Preschoolers’ spatial assembly performance relates to early
016 mathematical skills. Child Development, 85(3), 1062–1076. https://
Sorby, S. A. (2007). Developing 3D spatial skills for engineering stu- [Link]/10.1111/cdev.12165
dents. Australasian Journal of Engineering Education, 13(1), 1– Verdine, B. N., Golinkoff, R. M., Hirsh-Pasek, K., & Newcombe, N.
11. [Link] (2017). Links between spatial and mathematical skills across the
Sorby, S., Casey, B., Veurink, N., & Dulaney, A. (2013). The role of preschool years. Monographs of Society for Research in Child
spatial training in improving spatial and calculus performance in Development, 82(1), 7–30. [Link]
engineering students. Learning and Individual Differences, 26, Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor
20–29. [Link] package. Journal of Statistical Software, 36. [Link]
Stannard, L., Wolfgang, C. H., Jones, I., & Phelps, P. (2001). A longitu- 18637/jss.v036.i03[Link]
dinal study of the predictive relations among construction play and Voyer, D. (1998). Mathematics, gender, spatial performance, and cerebral
mathematical achievement. Early Child Development and Care, organization: A suppression effect in talented students. Roeper
167(1), 115–125. [Link] Review, 20(4), 251–258. [Link]
Stevenson, H. W., Parker, T., Wilkinson, A., Hegion, A., & Fish, E. 02783199809553902
(1976). Longitudinal study of individual differences in cognitive Wai, J., Lubinski, D., & Benbow, C. P. (2009). Spatial ability for STEM
development and scholastic achievement. Journal of Educational domains: Aligning over 50 years of cumulative psychological
Psychology, 68(4), 377–400. [Link] knowledge solidifies its importance. Journal of Educational
68.4.377 Psychology, 101(4), 817–835. [Link]
Swanson, H. L., & Beebe-Frankenberger, M. (2004). The relationship Walberg, H. J. (1984). Improving the productivity of America’s schools.
between working memory and mathematical problem solving in Educational Leadership, 41, 19–27.
children at risk and not at risk for serious math difficulties. Wang, M.-T., & Degol, J. L. (2017). Gender gap in science, technology,
Journal of Educational Psychology, 96(3), 471–491. [Link] engineering, and mathematics (STEM): Current knowledge, impli-
org/10.1037/0022-0663.96.3.471 cations for practice, policy, and future directions. Educational
Tartre, L. A., & Fennema, E. (1995). Mathematics achievement and gen- Psychology Review, 29(1), 119–140. [Link]
der: A longitudinal study of selected cognitive and affective vari- s10648-015-9355-x
ables [grades 6-12]. Educational Studies in Mathematics, 28(3), Wechsler, D. (1976). Wechsler intelligence scale for children-revised.
199–217. [Link] San Antonio, TX: Psychological Corporation.
Taub, G. E., Floyd, R. G., Keith, T. Z., & McGrew, K. S. (2008). Effects Wechsler, D. (1981). WAIS-R manual: Wechsler adult intelligence scale-
of general and broad cognitive abilities on mathematical achieve- revised. Psychological Corporation.
ment. School Psychology Quarterly, 23, 187-198. Wechsler, D. (2012). Wechsler preschool and primary scale of
Thurstone, L. L., & Thurstone, T. G. (1941). Factorial studies of intelli- intelligence—fourth edition. The Psychological Corporation San
gence. Psychometric Monographs, 2, 94. [Link] Antonio.
fulltext/[Link] Wei, W., Yuan, H., Chen, C., & Zhou, X. (2012). Cognitive correlates of
Tipton, E. (2015). Small sample adjustments for robust variance estima- performance in advanced mathematics. The British Journal of
tion with meta-regression. Psychological Methods, 20(3), 375–393. Educational Psychology, 82(1), 157–181. [Link]
[Link] 2044-8279.2011.02049.x
Tolar, T. D., Lederberg, A. R., & Fletcher, J. M. (2009). A structural Wilson, S. J., Polanin, J. R., & Lipsey, M. W. (2016). Fitting meta-
model of algebra achievement: computational fluency and spatial analytic structural equation models with complex datasets.
visualization as mediators of the effect of working memory on alge- Research Synthesis Methods, 7(2), 121–139. [Link]
bra achievement. Educational Psychology Review, 29(2), 239–266. 1002/jrsm.1199
[Link] Wolfgang, C. H., Stannard, L. L., & Jones, I. (2001). Block play perfor-
Uttal, D. H., Miller, D. I., & Newcombe, N. S. (2013). Exploring and mance among preschoolers as a predictor of later school achieve-
enhancing spatial thinking. Current Directions in Psychological ment in mathematics. Journal of Research in Childhood Education:
720 Psychon Bull Rev (2022) 29:699–720

JRCE / Association for Childhood Education International, 15(2), Zhang, X., & Lin, D. (2017). Does growth rate in spatial ability matter in
173–180. [Link] predicting early arithmetic competence? Learning and Instruction,
Wong, W. I. (2017). The space-math link in preschool boys and girls: 49, 232–241. [Link]
Importance of mental transformation, targeting accuracy, and spatial Zhang, X., & Lin, D. (2018). Cognitive precursors of word reading versus
anxiety. The British Journal of Developmental Psychology, 35(2), arithmetic competencies in young Chinese children. Early
249–266. [Link] Childhood Research Quarterly, 42, 55–65. [Link]
Woodcock, R. W., McGrew, K. S., Mather, N., & Schrank, F. A. (2001). [Link].2017.08.006
Woodcock Johnson III Tests of Cognitive Abilities. Riverside Zhang, X., Koponen, T., Räsänen, P., Aunola, K., Lerkkanen, M.-K., &
Publishing. Nurmi, J.-E. (2014). Linguistic and spatial skills predict early arith-
Wrigley, J. (1958). Factorial nature of ability in elementary mathematics. metic development via counting sequence knowledge. Child
British Journal of Educational Psychology, 28, 61–78. Development, 85(3), 1091–1107. [Link]
Xie, F., Zhang, L., Chen, X., & Xin, Z. (2019). Is spatial ability related to 12173
mathematical ability: a meta-analysis. Educational Psychology
Review. [Link] Publisher’s note Springer Nature remains neutral with regard to jurisdic-
tional claims in published maps and institutional affiliations.